THE NEURON AND THE MIND

This book, a companion to William R. Uttal’s earlier work on macroneural

theories of mind–brain relationships, reviews another set of theories—those
based on microneuronal measurements. Microneural theories maintain the
integrity of individual neurons either in isolation or as participants in the great
neuronal networks that make up the physical brain. Despite an almost universal
acceptance by cognitive neuroscientists that the intangible mind must, in some
way, be encoded by network states, Uttal shows that the problem of how the
transformation occurs is not yet supported by empirical research findings at
the micro as well as at the macro levels of analysis. Theories of the neuronal
network survive more as metaphors than as robust explanations. This book also
places special emphasis on the technological developments that stimulate these
metaphors. A major conclusion drawn in this book is that it is not at all certain
that the mind–brain problem is solvable in the sense that many other grand
scientific problems are.

William R. Uttal is Professor Emeritus (Engineering) at Arizona State University

and Professor Emeritus (Psychology) at the University of Michigan. He was one
of the pioneering researchers in computational modeling and is the author of
numerous books and over 140 scholarly articles.
 Books by William R. Uttal

• Real Time Computers: Techniques and Applications in the Psychological Sciences

• Generative Computer Assisted Instruction (with Miriam Rogers, Ramelle
Hieronymus, and Timothy Pasich)
• Sensory Coding: Selected Readings (Editor)
• The Psychobiology of Sensory Coding
• Cellular Neurophysiology and Integration: An Interpretive Introduction
• An Autocorrelation Theory of Form Detection
• The Psychobiology of Mind
• A Taxonomy of Visual Processes
• Visual Form Detection in 3-Dimensional Space
• Foundations of Psychobiology (with Daniel N. Robinson)
• The Detection of Nonplanar Surfaces in Visual Space
• The Perception of Dotted Forms
• On Seeing Forms
• The Swimmer: An Integrated Computational Model of a Perceptual-Motor System
(with Gary Bradshaw, Sriram Dayanand, Robb Lovell, Thomas Shepherd,
Ramakrishna Kakarala, Kurt Skifsted, and Greg Tupper)
• Toward a New Behaviorism: The Case against Perceptual Reductionism
• Computational Modeling of Vision: The Role of Combination (with Ramakrishna
Kakarala, Sriram Dayanand, Thomas Shepherd, Jaggi Kalki, Charles Lunskis
Jr., and Ning Liu)
• The War between Mentalism and Behaviorism: On the Accessibility of Mental
Processes
• The New Phrenology: On the Localization of Cognitive Processes in the Brain
• A Behaviorist Looks at Form Recognition
• Psychomythics: Sources of Artifacts and Misrepresentations in Scientific Cognitive
Neuroscience
• Dualism: The Original Sin of Cognitivism
• Neural Theories of Mind: Why the Mind–Brain Problem May Never Be Solved
• Human Factors in the Courtroom: Mythology versus Science
• The Immeasurable Mind: The Real Science of Psychology
• Time, Space, and Number in Physics and Psychology
• Distributed Neural Systems: Beyond the New Phrenology
• Neuroscience in the Courtroom: What Every Lawyer Should Know about the Mind
and the Brain
• Mind and Brain: A Critical Appraisal of Cognitive Neuroscience
• Reliability in Cognitive Neuroscience: A Meta-Meta-Analysis
• Macroneural Theories in Cognitive Neuroscience
• The Neuron and the Mind: Microneuronal Theory and Practice in Cognitive Neuroscience
THE NEURON
AND THE MIND
Microneuronal Theory and Practice
in Cognitive Neuroscience

William R. Uttal
First published 2017
by Routledge
711 Third Avenue, New York, NY 10017
and by Routledge
2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
Routledge is an imprint of the Taylor & Francis Group, an informa business
© 2017 Taylor & Francis
The right of William R. Uttal to be identified as author of this work
has been asserted by him in accordance with sections 77 and 78 of the
Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced
or utilised in any form or by any electronic, mechanical, or other means,
now known or hereafter invented, including photocopying and recording,
or in any information storage or retrieval system, without permission in
writing from the publishers.
Trademark notice: Product or corporate names may be trademarks or
registered trademarks, and are used only for identification and explanation
without intent to infringe.
Library of Congress Cataloging-in-Publication Data
Names: Uttal, William R.
Title: The neuron and the mind : microneuronal theory and
practice in cognitive neuroscience / by William R. Uttal.
Description: New York, NY : Routledge, 2016. | Includes bibliographical
references and index.
Identifiers: LCCN 2015050181 | ISBN 9781138640191 (hb : alk. paper) |
ISBN 9781138640207 (pb : alk. paper) | ISBN 9781315617480
(e : alk. paper)
Subjects: LCSH: Cognitive neuroscience. | Brain—Physiology. |
Neurobiology.
Classification: LCC QP360.5 .U88 2016 | DDC 612.8/233—dc23
LC record available at http://lccn.loc.gov/2015050181
ISBN: 978-1-138-64019-1 (hbk)
ISBN: 978-1-138-64020-7 (pbk)
ISBN: 978-1-315-61748-0 (ebk)

Typeset in Bembo
by Apex CoVantage, LLC
. . . the relationship between brain and cognition is still only poorly
understood. Great progress notwithstanding, neuroscience still cannot
answer the “big questions” about mind and intelligence.
(Olaf Sporns, 2011, p. 179)

If psychological states are constructed, emergent phenomena, then they

will not reveal their more primitive elements, any more than a loaf of
bread reveals all the ingredients that constitute it.
(Lisa Feldman Barrett, 2011, p. 124)

To be brutally honest, scientists do not yet have even the remotest idea
of how visual experiences—or indeed any other kind of experiences—
arise from physical events in the brain.
(Stephen E. Palmer, 1999, p. 618)

The problem of consciousness is completely intractable. We will never

understand consciousness in the deeply satisfying way we’ve come to
expect from our sciences.
(Dietrich and Hardcastle, 2005, cited in Rakover, 2011, p. 1)

A review of the neuroimaging literature suggests that selective association

between mental processes and brain structures is currently impossible to
find.
(Poldrack, 2010, p. 754)

. . . the initial hope of associating each brain area with a particular func-
tion (Posner, Petersen, Fox, and Raichle, 1998) has not been realized.
(Fox and Friston, 2012, p. 408)
This page intentionally left blank
FOR MITCHAN
This page intentionally left blank
CONTENTS

Preface xi
Acknowledgments xv

1 Introduction 1

2 Single Neuron Practice and Theory 37

3 Microneuronal Network Theories—Technique

and Methods 67

4 Intermediate Level Neuronal Networks—The Role

of Models 113

5 Large-Scale Computer Simulations (Theories)

of Cognition 141

6 Emerging Conclusions 177

Bibliography 183
Index 199
This page intentionally left blank
PREFACE

The brain is arguably the most complex entity that has ever existed. It is made
up of a vast number of cellular components (neurons) whose abundance compares
with cosmological numbers. Recent estimates for the number of cells in the
normal human brain (based on measurements by Ananthanarayanan, Esser, Simon,
and Modha, 2009; Azevedo et al., 2009) are that there are about 86 to 100 bil-
lion neurons in the brain (along with possibly an equal number of glial cells).
Astronomers, for their part, estimate that there are between 10 and 100 billion
stars in a typical galaxy like ours. Thus, the numbers in a galaxy and a brain
are of roughly the same order of magnitude.
Such a comparison is misleading, however. Although the components of brains
and galaxies are both heavily interconnected, the natures of the respective inter-
active forces among their respective components are very different. Stars interact
by means of a single, uniform, and well-described force—gravity—that declines
uniformly with distance. Quite to the contrary, the brain’s neurons are both
locally and remotely interconnected with one another by what other estimates
suggest may be thousands or even tens of thousands of synapses with idiosyncratic
effects. Each of these synapses (or combinations of them) may produce distinc-
tively different interactive effects on a neuron quite unlike the uniform influence
of gravity between stars. Interconnections between the brain’s neurons, therefore,
are not simple or uniform nor are they likely to be as simple as the forces
operating among a galaxy’s stars. Thus, the combinatorial complexity of the
brain’s neuronal network is probably far greater than that of a galaxy!
Given this level of complexity as well as the microscopic nature of neurons
and synapses, formidable obstacles exist to learning about the mechanisms by
means of which tangible brain activity becomes or produces or is intangible thought.
Nevertheless, in the past few decades there have been some extraordinary
xii Preface

technical developments that allow us to begin to examine the brain in ways that
only a few years ago would have been considered impossible. Many of these
developments are just becoming known to both the general public and the
scientific community.
Twenty-five years ago, I set out the goals of my future research program in
an article entitled “On Some Two-Way Barriers between Models and Mecha-
nisms” (Uttal, 1990). In retrospect, I would have titled this article somewhat
differently. I now read much of what I was saying there as a critical appraisal of
neuroreductionism—in particular, how various aspects of neurophysiological
science produce or are related to cognition. This has been the central theme of
my work and is extensively discussed in my recent and not so recent books
(Uttal, 2001, 2005, 2009, 2011, 2013).
I argue here that none of these neuroreductionist theories proposed to
explain cognition at the neuronal level has yet matured to the point that we
can experimentally examine most of their postulates and implications. My
primary goal in this book is to evaluate the current state of theory in cogni-
tive neuroscience specifically at the level of the cellular (i.e., the neuronal)
components of the nervous system. Unfortunately, there are some profound
barriers to progress in this field. What has arisen in the place of solid data-
driven biological-based theory, I believe, are largely metaphors or analogs driven
by available instrumentation—an engineering field where astonishing progress
has been made. An emerging generalization is that our theories are, too a
much greater degree than we appreciate, creatures of whatever technology is
available at any stage of history.
Therefore, it is a secondary goal of this book to survey the new technologies
used by neuroscientists, to explain how they work, and to provide at least an
initial conceptual foundation for the basic principles they highlight. I also want
to consider what parts of the mind–brain conundrum they might help to solve
and which may be completely intractable. Novel applications of these new
techniques and the promise of even more exciting developments in the near
future offer promise of major breakthroughs in our understanding of at least the
anatomical and physiological aspects of brain, if not the “Holy Grail” of modern
brain research: the answer to the question—how does brain activity account for
mental activity?
Make no mistake, by the very act of applying some of these instruments, we
come up against some of the most perplexing problems in modern science.
Some of these problems are practical ones of dealing with the vast numbers of
neurons that are involved. However, others are matters of deep principle that
should be faced before any commitment can be responsibly made to such a
grand enterprise as the proposed billion-dollar investment in the “Brain Activity
Map” (Alivisatos et al., 2013), an approach whose goals are already coming into
conflict with mathematical and physical laws that the brain must obey (Under-
wood, 2014).
 Preface xiii

In this context of questioning which goals of cognitive neuroscience are

achievable using the best available technical developments and which are likely
to elude our best efforts, there are several caveats that I should make explicit so
that the comments presented here are not misinterpreted:

1. Nothing I say here about the difficulties arising in applying powerful new
tools should be interpreted to mean that I am suggesting that we can deter-
mine at this point whether the mind–brain problem is either solvable or
intractable. It is essentially a combined empirical and theoretical question that
may not be answered fully for many years. Of one thing we can be sure—we
are nowhere near even a tentative solution.
2. Should it turn out that the mind–brain problem is formally intractable, such a
finding would necessarily have no special significance outside of the domain
of the natural sciences. That we cannot understand something for practical
reasons cannot be used to argue that there is anything other than materialist
laws at work in the brain as well as in the cosmos. Nor does the absence of a
complete explanation of a complex system such as the mind–brain mean that
we are not able to manipulate that system in productive, practical ways.
3. Furthermore, no one can deny the possibility that some future discovery or
some scientific genius will lead to an overarching theory of mind–brain rela-
tions and overcome what are apparently insurmountable current barriers to
understanding. This book is only a contemporary evaluation of theory and
technology underlying a particular kind of neuroreductionism, the kind we
practice at the beginning of the 21st century. It makes no claims to the future
beyond reasonable extrapolations from existing knowledge and identification
of potential or real obstacles to progress.
4. The fact that there may be formidable technical and conceptual obstacles
lying ahead should not be interpreted as a call for cessation of research at
any level. We are learning much about single neurons and their properties in
complex networks, and, at the very least, some network models or computer
programs have led to useful simulations of human behavior of great value.
The engineering of new tools that simulate human behavior will move on
and whatever limits characterize current technology are likely to eventually, if
not immediately, correct themselves.
5. Nevertheless, there are some fundamental barriers to understanding
how the brain produces the mind that would be well to appreciate. To
attempt to overcome some otherwise well-established physical law, such
as the second law of thermodynamics, or the combinatorial explosion in
computation in an effort to build a perpetual motion machine would be
an enormous waste. We should not ignore the possibility that we might
encounter similar physical or computational limits in cognitive neurosci-
ence and beware of equally wasteful efforts to go beyond what we already
know to be impossible.
xiv Preface

In summary, there are two major questions that I seek to answer in this book.
The first is—are overarching microneuronal theories of cognition possible? The
second is—how does the available instrumentation constrain and dictate what
kind of theories we develop? Should I make some slight progress in answering
these two questions, I would consider my efforts worthwhile.
ACKNOWLEDGMENTS

As ever, I am extremely happy to acknowledge the support that I have received

from my home unit at Arizona State University—the School of Computing,
Informatics, and Decision Systems Engineering. I am grateful for the friendship
of my colleague and friend Professor John Reich during the trying time this
book was produced. I am also deeply grateful to my daughter, Taneil Uttal, who
assisted me in the editing of this, my final book. However, most of all, it is my
wife, Mitchan, to whom, in gratitude for her good sense, I dedicate this book
once again.
This page intentionally left blank
1
INTRODUCTION1

1.1 Explanation
Of all of the scientific mysteries confronting our inquisitive species, none is more
profound or challenging than understanding how the tangible brain can give
rise to intangible thought. This has traditionally been known as the mind-body
problem but, as we have become more sophisticated, referring to it as the mind–
brain problem seems more appropriate. Despite the enormous amount of research
being conducted that deals with the brain and the mind, it is almost universally
agreed that we actually have made little progress toward resolving the overarching
nature of this relationship. In place of creating anything approximating the grand
theory of cosmological physics, we have learned an enormous amount about
each domain separately but, unfortunately, piecemeal. Fortunately, knowledge
about the anatomy and physiology of the brain is succumbing to powerful new
research techniques, and psychologists have learned much about the ebb and
flow of mental activity by observing behavioral responses.
What we have not done is to have made any substantial progress toward build-
ing a conceptual bridge between the psychological and neurophysiological
domains. The reasons that we have not are multiple and complex; as a result
investigators from many fields have debated whether the problem is in principle
“solvable” or “tractable.” On the one side are the optimists. To them, to “explain”
how the brain makes the mind is just another scientific task, admittedly quite
complicated, but still achievable within the rubric of conventional scientific inquiry.
However difficult the problems may be, the obstacles to understanding them are
just practical matters awaiting the development of a new method or technology.
Most cognitive neuroscientists implicitly accept the ultimate achievability of the
task and assume that however slow it may be in coming, eventually we will
2 Introduction

understand or explain how the brain makes the mind. To think otherwise would
raise serious questions about the raison d’etre of cognitive neuroscience itself.
On the other side of the debate are the pessimists (or, if you prefer, the real-
ists); their countervailing opinion argues that the problem is not just difficult in
practice but is impossible in principle. That is, the complexity of the problem
is such that there is no strategy or tool that we could adopt or invent that would
ever lead to an analysis that would “explain” how the process of converting
brain activity to mental activity takes place. Philosophers, combinatorial mathema-
ticians, some behaviorally oriented psychologists, and even a few of those who
would happily accept the title Cognitive Neuroscientist are now beginning to
accept, implicitly if not explicitly, at least the possibility of impossibility.
The resolution of this dispute, of course, will ultimately come from empirical
studies. Someday we may solve the problem and that will be that—debate over.
Nothing proves the existence of a dimly seen phantom like Bigfoot better than
robust evidence of its existence although absence of such evidence does not
prove it does not exist. In this case, the “existence proof ” for an overarching
neuroreductionist theory would be the demonstration of our ability to construct
high-level behavior or intelligence or sentience or cognition (or whatever it is
that we may wish to call mind) from the properties of low-level neurophysiologi-
cal components. Of course, even then the uncertainty of what amounts to an
acceptable explanation would remain controversial.
This debate sets the stage for this book. In it, I strive to review and evaluate
the current state of a particular kind of mind–brain theory—the kind that is
based on the microscopic, cellular components (i.e., neurons) of the nervous
system. These theories are designated as microneuronal as distinguished from those
that are based on larger chunks of the brain—those designated as macroneural. I
have already dealt with macroneural theories in a previous book (Uttal, 2016),
to which this book is a companion, and will only deal with them casually in
passing in this chapter.
To start, it would be useful to define exactly what we mean by a theory. For
a number of reasons, however, categorizing existing theories is much more arbi-
trary than it may at first seem. Theories come in many kinds, each of which
may emphasize different foundation axioms and postulates. A rough taxonomy
of the major kinds of cognitive and cognitive neuroscience theories would include
the following:

• Behavioral or Descriptive Theories: Theories of this kind are not meant to be

reductive. That is, there is no effort to identify the cognitive or neurophysi-
ological component parts that make up an observable behavior pattern. A pure
behavioral theory, quite to the contrary, is only descriptive. It may be based
on a mathematical formulation that allows us to predict the trajectory of a
behavior or to represent the transformation between stimuli and responses, but
there is no attempt to delve into lower-level cognitive or neurophysiological
 Introduction 3

processes. Any en passant allusions to neurophysiological or inferred cognitive

components are at best rough metaphors, in the absence of additional data.
The mathematics itself, like behavior in a descriptive theory, is neutral with
regard to specific underlying mechanisms—cognitive or neural. Thus, no mat-
ter how accurate are the predictions made by a mathematical model, unless
there is additional data, it is not possible to produce a unique solution to the
mind–brain problem from the formulation. There are always many alternative
models that are sufficient to describe the behavior. Another way to say this is
that behavioral and mathematical descriptions are underdetermined.
• Cognitive Reductive Theories: Theories of this kind are intended to be reduc-
tive but not to neurophysiological mechanisms. The underlying components
are inferences implied by the behavioral data. Many theoreticians of this ilk
hold the assumption that behavior is transparent (i.e., can be inferred from)
the underlying cognitive process and that by careful experimental designs we
should be able to parse out these inferred cognitive components. These cog-
nitive components, modules, or faculties components have been designated as
“hypothetical constructs” (MacCorquodale and Meehl, 1948), emphasizing
the intrinsic difficulty of exactly defining them—a major handicap of such
“top-down” approaches in psychological research. Again, these hypothetical
constructs may allude in passing or by analogy to neural structures and pro-
cesses, but it is impossible in principle for this approach to rigorously designate
what these inferred components might be.
• Neuroreductive Theories: Because overt behavior and covert cognition are
not able to lead us deductively to the underlying brain components that
are necessary for mind, and because there is widespread agreement that all
cognitive processes are in some ultimate sense brain processes, the search for
the neural mechanisms of mental activity has motivated a vast amount of
research. With the development of modern neurophysiological and neuro-
anatomical techniques, the search for the brain mechanisms of mind has
greatly accelerated. There are two main threads of this kind of research, which
has been alternatively known as physiological psychology, psychobiology, and
most recently, cognitive neuroscience. These threads are respectively known
as macro-neuroreductionism and micro-neuroreductionism.
• Macro-neuroreductionism is based on data in which the cumulative,
pooled, or summed actions of many individual neurons are emphasized.
The details of what individual neurons are doing are lost since their neu-
roelectrical and neurochemical responses are additively lumped together
in the eyes of the investigator. This kind of data provides a powerful
impulse to develop theories that assume a kind of gross localization of
the component neural processes on or in the brain. It also stimulates ideas
about specialized cognitive roles for these macro-regions of the brain,
that is, phrenological and neophrenological associations of places and
cognitive phenomena.
4 Introduction

• Micro-neuroreductionism is based on data obtained from experiments

in which the activity and measurement of individual neurons is main-
tained either individually or in their participatory role as components of
a neuronal network. There are three levels of micro-neuroreductionism
within this part of this mini-taxonomy: (1) The action of single neurons
has been associated with cognitive functions; (2) the action of relatively
small, but computationally tractable, networks of a few neurons has been
used as the basis of an intermediate level of theorizing; and (3) with the
advent of supercomputers that can simulate the individual activity of
billions of neurons has come a new form of micro-neuroreductionism
approach in which the number of neurons involved in a theory can
approximate the number of neurons in the brain. The ultimate form
of a mind–brain theory, most cognitive neuroscientists probably agree,
would ideally be found in the properties of such a network; however, it is
still problematic whether we will be able to manipulate them in the way
needed to “solve” the mind–brain problem.

A distinction should also be made between top-down and bottom-up theories

in the context of this mini-taxonomy of theoretical types. The goal of top-down
theories is to measure events at higher levels and then to determine by various
methods the nature of the underlying neural and cognitive components that
might embody the higher-level processes. The goal is, thus, to determine the
underlying processes by drawing inferences from experiments on higher-level
processes. This is the essence of top-down, analytical neuroreductionism. The
underlying mechanisms are the unknown targets of this strategic approach.
Bottom-up theories have a different strategy; they attempt to synthesize
higher-level systems from preexisting knowledge of the underlying components.
This is essentially a synthetic approach to theory building. A classical example
of the bottom-up approach to theory development was Watson and Crick’s
(1953) reconstruction of the genetic code for DNA from what was then known
about the bonds between molecular structures. Their theory was embodied in
a mechanical structure that depended on laws of chemical bonding in a way
that opened up a new world in biological research.
Although both top-down and bottom-up methods seek to achieve the same
goal—explain higher-level processes by lower-level ones—there is a fundamental
difference in their basic approach. The analytical top-down method dotes on
parsing a molar process into its components, whereas the synthetic bottom-up
method aims to solve the problem by concatenating known lower-level processes.
Both are susceptible to errors, of course. Top-down analytical methods are prone
to mistake the sufficient for the necessary. Bottom-up synthetic methods are
often more robust but suffer from incomplete knowledge of how the components
may be interconnected.
 Introduction 5

All theories represent hopes for the future; however, each type of theory is
constrained in what it can ultimately accomplish. Not only are there differences
in the level of anatomical complexity at which a theory must operate but also
there are major differences in what relationships should be measured in an
experiment. Thus, the exact theoretical relation between a cognitive process and
a neural response is not always clear even in the most highly correlated data.
Theories also differ in their methodological origins. A “theory” may be built
on any one of the following relations:

• Statistical correlation between cognitive and neural responses

• The location of a neural mechanism of a cognitive process
• A mathematical description
• A physical model or simulation
• Reproduction by reconstruction, a rough verbal statement of some perceived
metaphorical relationship
• The psychoneural equivalent of a cognitive process

The psychoneural equivalent is the actual neural mechanism whose activi-

ties are supposed to be indistinguishable from the cognitive process itself.
In other words, the psychoneural equivalent is the neuronal embodiment of
the cognitive process. Although defined in the different languages of cogni-
tion and neurophysiology, the different words may denote exactly the same
thing. Determining the nature of the psychoneural equivalent is an extremely
challenging task facing many impediments and obstacles; nevertheless, it is
the holy grail of cognitive neuroscience. The most powerful interpretation—
literal equivalence—in all of their properties—is in the manner expressed by
“identity” theorists such as Place (1956) and Feigl (1958). In identity theory,
cognitive and neural activities have the strongest possible relationship—the
one is the other.

The range of theoretical issues with which I have dealt in my previous work
on neuroreductionism can be further clarified by the following list of assump-
tions, properties, postulates, restrictions, and if you wish, biases that have guided
my analysis:

1. Ontologically Physicalist: The prime postulate of modern cognitive neurosci-

ence is the physical or material origin of mental processes. That is, whatever
mind is, it is a manifestation or product of laws of the singular reality that
accounts for all other events, things, and processes in our real world. In other
words, cognitive neuroscience is fundamentally monist; it admits of no other
6 Introduction

separable and distinguishable reality of the kind proposed by some philoso-

phers. To accept any other (some kind of dualism) of a distinguishable reality
difference between the mental and the physical would totally invalidate the
science; it would require that we accept the possibility of forces that are not
controlled by our experiments influencing those experiments. Such experi-
ments, therefore, would invalidate the entire corpus of research in cognitive
neuroscience.
2. Epistemologically Indeterminate: Despite the basic materialist ontological
orientation of modern cognitive neuroscience, others believe that there are
practical or epistemological constraints on what an empirical science can
accomplish in the study of mind–brain relations. Some of these constraints
have to do with the complexity of the system; some have to do with the inac-
cessibility of mental activity; and others arise from interpretive and logical
errors. As a result of these practical problems, many proposed theories are
little more than rough metaphors or prototheories rather than full-blown
explanations. Thus, we may be in a situation in which despite being ontologi-
cal monists, we are simultaneously doomed to be epistemological dualists and
must study mind and brain separately.
3. Reductive: A reductive theory is based on the assumption that the properties
of the lower level can, if appropriately manipulated, produce the properties of
the higher level by concatenation. Nothing supernatural or additional hap-
pens between the two levels, just a rational flow of the processes and causes
leading from one level to the other. The word “emergence” (by which is
meant that new properties that have their origins in the lower level without
causal connections or forces) is often used to denote the process but its use
usually turns out to be a way of finessing the unknown without actually
explaining anything. The radical reductionist, to the contrary, assumes that
everything that is expressed at the higher level is at least implicit at the lower
level. The task of cognitive neuroscience is to make explicit those properties
that are implicit.
4. Bottom-Up Possibility: Bottom-up theories differ greatly from top-down
ones. It is possible, in principle, knowing the properties and the rules of inter-
action (even if not practicable for reasons to be discussed later) to construct
a bottom-up theory in which one reasons from the lower-level neurons to
higher-level properties and thus to develop a “necessary,” full, complete, and
uniquely accurate theoretical explanation. An example of a bottom-up the-
ory is how the human visual sensitivity to light of different wavelengths is
accounted for by the differential absorption of photons of different wave-
lengths by the retinal receptor photochemicals. Another example of a bottom-
up theory is the suggestion that Mach Bands (edge enhancements in visual
perception) are accounted for by mutual lateral inhibitory interaction among
retinal neurons. Another example of a successful scientific synthesis from a
 Introduction 7

totally different domain would explain how the microscopic structures of

atoms account for the macroscopic behavior of chemicals.
5. Top-Down Impossibility: However, no matter how empirically robust, full,
complete, and predictively accurate theoretical explanations they may be,
top-down, neuroreductionist theories are not in principle robust. There are
too many alternative, plausible, and possible “sufficient” explanations that
can be generated from any given set of data. There is no way in which the
underdetermined data of the macrocosm can be transformed into a unique
“necessary” neuron-level explanatory theory. Top-down theories are useful
for testing the plausibility of a prototheory but cannot distinguish between
the unique “necessary” explanation and the array of “sufficient” ones that all
fit the data equally well.
6. Neuronally Reductive: For the purposes of this book, the highest level of
analysis is that of cognitively related behavior and the lowest level with
which I deal is that of the specialized cells of the nervous system—the
neurons. This instantiates a very particular postulate of the kind of neuro-
reductionist analysis dealt with here—namely, it is assumed, both explicitly
and implicitly, throughout this discussion that any future neuroreductionist
explanation of cognition is going to be framed in terms of individual neu-
rons and their interactions at a microscopic level. It is further assumed that
the identity and measures of the activity of individual neurons is necessarily
preserved in any theory. This approach is distinguished from models that
dote on macroneural “chunks” of the brain—a topic on which I have writ-
ten extensively (e.g., Uttal, 2016)—and that have increasingly been rejected
as a strategy that will lead us to the solution of the mind–brain problem.
Other theoretical approaches, including the descriptive mathematical mod-
els commonly used in psychological theories, abound but are neither reduc-
tive nor neurophysiological no matter how well they correlate or predict
the trajectories of the data.
7. A major principle, permeating everything I present in this book, is that
available technology not only dictates what cognitive neuroscientists do in
the laboratory but also guides our theoretical explanations. It is important,
therefore, to distinguish between obvious and trivial metaphors and what are
potential theoretical explanations. My strategy for doing this is to deal with
the technology more thoroughly than is usual in a book of this kind.
8. Finally, to the extent possible, I remain an agnostic about the solvability of
the mind–brain problem at the microneuronal level with which this book
deals. As I discuss the empirical research on which my arguments are based, I
doubt that I shall be converted from what is essentially a pessimistic position
to an optimistic one, but at least some of the issues discussed will lead to a
better understanding of what is involved in the search for the solution to the
mind–brain problem.
8 Introduction

The foregoing, then, are some of the assumptions influencing my analysis of

microneuronal theories of cognition. To add substance to these initial perspec-
tives, we need to review and evaluate the spectrum of proposed microneuronal
theories and draw out their strengths and weaknesses. However, before inspecting
individual theories and particular empirical results, there are some other general
issues to be considered.

1.2 The Influence of Technology on Theory

It is an unambiguous cliché that science has always been driven by whatever
measuring instruments are available to probe and measure natural phenomena.
I argue here that the measuring instruments that can be marshalled by any
generation of scientists play an essential role in determining not only what can
be measured but also what theoretical inferences can and will be drawn. The
history of science is replete with instances in which newly available instrumenta-
tion opened the door to new knowledge and deep understanding. It is also
unfortunately true that, sometimes, whatever technology is fortuitously available
can misdirect and even restrict scientific activities to a narrow path of action
even when it is clear for other reasons that the path may be a dead end. It is
all too easy to concentrate on what is available or easy to do rather than what
is more fundamentally correct to do. Yet, as I have noted previously, we do what
we can do when we cannot do what we should do. There is an obvious truism built
into this aphorism, of course; one can only do what is doable and to ask one
to do that which is not doable is a recipe for magical and dualistic thinking.
The point is, nevertheless, that whatever tools are available at any point in sci-
ence can exert a powerful influence not only on what interpretations are made
of evidence but also on the nature of the evidence itself. That influence may
mislead as well as illuminate as evidenced throughout the history of science.
A brief glimpse at the course of scientific history quickly makes clear that
an array of influential instruments has guided the historical progress of the
biological sciences and lately the cognitive neurosciences. The impact of Anton
van Leeuwenhoek’s (1632–1723) invention of the high-magnification micro-
scope and its successors on biology in general and science in particular cannot
be overstated. For neuroscience, the microscope and clever staining techniques
(e.g., those developed by Golgi, 1906) clarified the cellular anatomy of the
nervous system, opened the door to the study of individual neurons and, thus,
provided the basis for theories emphasizing the role of neurons cum neurons
in coding cognitive processes. Before the microscope, the brain was a homo-
geneous and undifferentiated “bowl of porridge”; afterwards it was a mesh of
interwoven components.
Perhaps more than anything else, the discovery, measurement, and manipulation
of electricity became the impetus for an enormous variety of scientific develop-
ments. In the present context, this revolutionary advancement in mind–brain
 Introduction 9

thinking was based on two ideas. The first was that the nervous system’s covert
activity (as evidenced by the resulting overt behavioral responses) could be elicited
by electrical stimuli—an idea famously known since the work of Luigi Galvani
(1737–1798), in which a frog’s leg was activated by an electrical shock generated
by a primitive battery, the Voltaic pile. In addition to suggesting that there was
an electrical basis of nervous action, the efficacy of electrical stimulation meant
that very well-controlled stimuli could be used to generate well-defined responses
in the nervous system.
The second, but far more difficult task, was recording the minute signs of
nervous activity by measuring the electrical signals emitted by organic tissues
and cells. This meant that the entire panoply of electrical measuring instruments
developed over the centuries could be used to study the nervous system at several
different levels of magnification. Although we now know that this electrical
activity can be even more fundamentally explained in the language of ionic
chemistry, specifically of the concentrations and flow of sodium, potassium, and
chloride ions across the neuron’s cell membrane, it was the doorway opened by
electrical recording by the invention of some ingenious devices as the galva-
nometer and oscilloscope that made it possible to understand the operation of
individual neurons. It is still not possible to unravel the complex activity of
networks of many neurons, but progress looms on the horizon as new techno-
logical advances such as parallel-processing computers flood from engineering
laboratories.
The point is that, to an immeasurable but very large degree, both experimental
and theoretical progress in cognitive neuroscience, like all other sciences, has
been driven by available instrumentation. That is, the instrumentation available
at any moment in history largely determines our theoretical extrapolations and
the inferences we draw from our observations. As a result, it is all too easy for
us to be misled by tool availability to overvalue a particular line of theoretical
thought to the exclusion of other more plausible, but empirically unavailable,
alternatives. These biases also extend down into the empirical results themselves
as instrument-based theory also feeds back to guide and direct our experimental
designs and to select to which findings we will pay attention. Thus, for example,
it is argued that many theories may be heavily (if not more so) influenced by
the properties of whatever tool is available than some more germane but less
obvious attributes of the phenomenon under study.
Consider the following example: The most popular of the current tools—the
EEG and fMRI—used by cognitive neuroscientists mainly observe at the macro-
neural level. They record the activity of literally billions of neurons in the
form of cumulative activity localized at particular places (on or in the brain) or
occurring at particular times. This led to a number of prototheories of brain
organization that were couched almost exclusively in the terminology of mac-
roneural brain locations (using fMRI techniques) or temporal sequences (using
EEG techniques) rather than the properties of microscopic networks. This has
10 Introduction

opportunistically led to theories of macroscopic mind–brain relations. In point

of empirical fact, however, these cumulative, gross, macroneural measures may
have little utility as reductive explanations of cognitive activity. It is becoming
increasingly apparent that more valid explanations of mind–brain relations are
likely to be framed in the concepts and terminology of the microneuronal ele-
ments and their interactions.
Of course, this microneuronal line of thinking approach itself does not exoner-
ate any such theories from also misleading us; but, at least, there is also considerable
logical justification for thinking that the microneuronal, neuron-by-neuron–based
approaches are correct in principle and that the available macroneural approaches
may be misleading us in practice. In short, given that we do not know the answers
to most of the important mind–brain questions, I argue that we should be con-
stantly aware that our instruments (and whatever kind of data they currently
provide) can erroneously dictate our theoretical orientations.
Indeed, to a substantial degree, this is exactly what is happening (and has
happened throughout the history of science). Current theoretical cognitive
neuroscience thinking is heavily biased by the ease of availability of macroneural
measuring devices. Brain-imaging studies based on fMRI technology impel us
toward concepts of macroneural brain localization, and EEG studies drive us
equally powerfully to theories of temporal “synchronization.” Although these
macroneural theoretical options are not always exclusive of microneuronal con-
cepts, the amount of our intellectual energy committed to them is not justified
by robust empirical evidence or reasonable logic—according to an increasing
number of cognitive neuroscientists.
Thus, the caveat that the kind of instrumentation available exerts a powerful
impulse to think in a certain way must never be underestimated. The ability
to observe the response of an individual neuron has led to some fanciful
“single cell” theories of equivalence between thoughts and individual neurons.
On the other hand, according to most cognitive neuroscientists when pushed
on the matter, the most likely level at which the nervous system engenders
mind is that of the neuronal network, the microneuronal level at which there
is virtually no empirical evidence to substantiate such an explanation. It is
paradoxical that the least substantiated empirically is the most likely logically
and widely accepted.

1.3 Barriers to Cognitive Neuroscience Theory Development2

Given the enormity of the challenge posed by the mind–brain problem, it is
understandable that there are many obstacles and difficulties encountered
when one attempts to develop an explanatory theory of the mind, whether
it be cognitive, neural, or neuronal. The following discussion recapitulates
some of these barriers and challenges that bedevil the task of the
theoretician.
 Introduction 11

Levels of Analysis
A major barrier to theory building concerns the level at which a problem is
being studied. By level of analysis, I refer to both the scale at which brain activ-
ity becomes mental activity and the scale of our analyses and measurements. In
the present context, a microneuronal level of analysis entails investigating the
responses of cellular and subcellular units such as neurons and synapses and their
respective interactions. A macroneural level of analysis deals with signals that are
inseparable pools or accumulations of these lower-level responses such as fMRI
images or EEG recordings.
Theory development at a microscopic level is severely inhibited for several
very practical reasons. Primary among these, we are overwhelmed by so much
data that empirical-based theory is, for practical reasons, computationally unavail-
able to us. We have no way of both manipulating all of the necessary experi-
mental details and measuring the vast amount of activity in the neuronal network
whose collective but idiosyncratic states must in some sense correspond to cogni-
tive processes. Macroneural approaches to theory are also deeply flawed. They
trade the ability to measure for the convenience of pooled or mixed data, thus
losing the very information—the detailed microstructure of the brain—that they
would need to develop a microneuronal theory.
Thus, at both the microneuronal and macroneural levels, the possibility of
developing an authentic, valid, overarching theory that truly speaks to the question
of how the brain encodes mental activity is currently remote and is likely to be
for the foreseeable future. For these reasons, I suggest that both levels of analysis
are not promising foundations on which to build a theory of mind–brain relations.
However, unlike the microneuronal level, macroneural experiments cannot only
be conceived but actually implemented. This simple fact introduces a strong bias
into the entire conversation. What follows is a more detailed discussion of some
of the barriers that confront the cognitive neuroscience theoretician.

Complexity
It is a truism that the brain is an extraordinarily complex organ. Indeed, given
the idiosyncratic nature of its interconnections, an argument can be made that
it is the most complex entity in the universe. Regardless of which level of analysis
one is considering, the combinatorial complexity of the brain’s neuronal network
stresses any possible explanation of its function to its limits for a relatively simple
practical reason—there are too many neurons and too many interconnections
involved in cognitive processes to be processed by any conceivable information
processing machine. It surprises many how very few interacting neurons it takes
to pose an intractable combinatoric problem. The complexity issue has been
known for years. As long ago as the 1960s, Leon Harmon was demonstrating a
simple three-neuron simulator that produced unpredictable results.
12 Introduction

Complexity is a result of a number of factors. Obviously the sheer numerousness

of the components involved in any realistic neuronal network that might be capable
of instantiating a cognitive process is one factor. But, as we have already seen, even
a modest number of a few interconnected components can produce combinatorial
intractability. Furthermore, nonlinear neural mechanisms pose computational chal-
lenges that have not yet been solved by modern mathematics and remain among
the most profound problems confronted by cognitive neuroscience.
Beyond sheer numerousness and nonlinearity, many early theories of neural
nets did not scale up well. A simple neural network exhibiting, for example,
learning with less than a hundred neurons might suddenly saturate with all
simulated neurons responding simultaneously. This might occur when even a
few more neurons were added or a slightly different network configuration
introduced. The eventual collapse of the small neuronal network program in the
1980s (as discussed by Fodor and Pylyshyn, 1988), when attempts were made to
scale up the number of emulated neurons, was probably an unavoidable outcome
of this intrinsic instability. Although far more simulated neurons are involved in
today’s computational neuroscience models, these prototheories are also eventually
subject to the curse of numerousness and combinatorial complexity.
Sheer numerousness, however, pales in significance when one also considers
the idiosyncratic nature of the anatomy of the neurons and their interconnec-
tions that make up the brain. There are many different types of brain neurons
and they are interconnected in a plethora of irregular ways by an abundance of
synaptic variations. There are no shortcut methods to overcome this impediment
to explanation that do not result in a massive loss of the very information that
is presumably the psychoneural equivalent of cognitive processes. It has been
argued that any proposed attempt to study the brain at this minute level is
beyond both our technology and our mathematics (Lichtman and Denk, 2011).

Poor Definition of Cognitive Constructs

For science to achieve a high level of theoretical precision there is a profound
need that the dimensions and properties of its variables be precisely defined. Yet,
psychology is replete with poorly defined terms such as attention, thinking, emo-
tion, and consciousness at a high level of abstraction and words such as learning,
recognition, detection, retrieval, and conditioning defined more operationally.
Many cognitive terms have multiple meanings that vary with the particular goals
of the research. For example, Vimal (2009) tabulated 40 different meanings of
the word “consciousness.” As another example, I listed (Uttal, 2011) 41 different
meanings of the word “learning.” A further problem, however, is that none of
these fluctuating cognitive terms need necessarily map directly onto neural mecha-
nisms. That is, what is a fundamental elemental process or faculty to a psychologist
may mean something very different to the brain. In short, our behavioral taxono-
mies and those of the brain may not be speaking the same language.
 Introduction 13

A novel way of emphasizing how this lack of isomorphism between cognitive

and neural vocabularies affects our thinking has been provided by Oosterwijk
et al. (2012). They concluded that the quest to associate specific cognitive con-
structs with particular brain regions has generally been unfulfilled. They, therefore,
joined those who support the ideas embodied in an alternative hypothesis—namely
that psychological constructs are encoded by a distributed pattern of responses
executed by general purpose neural mechanisms. One implication of such a
conclusion is that however much we may seek to modularize our psychological
taxonomies (a tradition that goes back to Aristotle’s faculty psychology) there is
no reason to assume that the mental components are parsed in the same way
by the nervous system. In Oosterwijk et al.’s words, “the brain does not respect
faculty psychology categories” (p. 2110).
There are two ways to interpret this statement that are relevant to this discus-
sion. First, psychological processes are not sufficiently well defined to be linked
with the detailed responses of the underlying neural machinery. Second, as just
noted, the brain is not necessarily modularized in the same manner as are cogni-
tive processes. Either or both of these implications suggest that there is a fun-
damental mismatch between our psychological and neurophysiological languages
and concepts. To search for localized psychological constructs in the anatomical
brain, therefore, may be a search for a chimera.
Thus, if one seeks to find reliable equivalences or correlations, much less
causal relations, between cognitive phenomena and neural responses, it demands
an increased degree of precision in the definition of the key stimulus variables
in an experiment. Unfortunately, different investigators often use different ter-
minologies to specify what is actually the same cognitive process. One researcher’s
search for data backing up a theory of decision making, for example, may be
operationally indistinguishable from another’s effort to study attention. This is
the inverse of the additional problem in which the same words may be used to
define what may operationally be very different cognitive processes.
As a result, we have to take it as a given that cognitive neuroscience is as
profoundly influenced by the vagueness of psychological language as it is by the
availability of its tools. Clarifying psychological definitions and constructs should
also be a high-priority task, especially for this science. Nevertheless, this important
task is largely ignored by all but a few psychologists and cognitive neuroscientists.
Currently, psychological terms are largely defined by the operations involved in
carrying out an experiment. This can lead to serious confusions of definition
both creating hypothetical constructs and ignoring real constructs.

Inaccessibility
Why should this imprecision of definition and measurement of cognitive processes
exist? A major conceptual reason is that hypothetical cognitive constructs are not
only inherently difficult to define but also impossible to directly measure. From
14 Introduction

any point of view, mental experiences are private, intrapersonal, and inaccessible
psychological states that can only be experienced by the individual and are not
directly sharable with others. Neither introspection nor experimental assays can
provide us with direct evidence of the values, properties, and dimensions of mental
phenomena. Our reasoning concerning “other minds” is, therefore, limited to
analogies based on the Cartesian assumption that “Cogito ergo sum” and the
analogous conviction that our fellows must be just like us. This means that we
must use behavioral surrogates such as “percent correct” and “reaction time” as
indirect measures of cognitive response, thus isolating us from the phenomena of
primary interest to psychologists—our inner private cognitive responses.
Another reason that cognitive variables may be inaccessible is that they prob-
ably reflect the outcome of extremely complex multi-variate neural responses
that cannot be characterized by single (or a few) measures typical of modern
cognitive neuroscientific research. The net effect of inadequate definitions and
inaccessible, underdetermined responses is to place conceptual barriers between
the various levels of representation of cognitive activity and, thus, further limit
the possibilities of developing a theory of mind–brain relationships.

The Search for Objectivity

Despite the difficulty in defining or measuring inaccessible cognitive entities such
as thoughts, feelings, experiences, and perceptions, there has been a persistent historical
effort to seek out objective “hard science” correlates—in particular, brain responses—of
cognitive processes. The fundamental ontological premise of all such efforts has been
the physicalist assumption that psychological activity is the outcome of brain activity
although in some yet indefinable way. Few cognitive neuroscientists would disagree
with this point. However, there is an epistemological corollary of this assumption
that is a matter of great contention. Namely, can any measurement that is made of
brain processes be theoretically transparent to cognitive activity? That is, are objective
brain measurements capable of correlating sufficiently well with cognitive activity
to serve as an “explanation” of that activity? Or, in other words, are the parameters
of cognition inaccessible to science?
The main weakness of the a priori expectation that any objective brain
response should be able to inform psychology about cognitive processes is that
the multiple levels of brain coding mean that some objective measurements of
brain activity, however empirically sound, are simply irrelevant to the mind–brain
problem. Studies of the chemistry of individual neurons and synapses, microscopic
investigations of structure, and single-cell recordings tell us little about the cog-
nitively meaningful activity of what is increasingly likely to be widely distributed
portions of the whole brain. It is not the molecular chemistry of neurons nor
of synapses that might open the door to understanding how the brain works as
a cognitive engine; it is more likely to be the aggregate information processing
by a host of individual neurons whose collective (but not cumulative) states
 Introduction 15

account for mind. This is a formidable barrier to building neuroreductionist

theories of cognition.
Nevertheless, the search for objective measures of subjectively defined psy-
chological states continues. In doing so it perpetuates the myth that observable,
objective, neural brain responses representing cognitive processes are decodable.
Even more seriously, the search for objectivity can bias our experiments such
that the probability of false-positive results is inordinately enhanced.

False Analogies between Sensory, Motor, and Cognitive Processes

Another conceptual barrier encountered in the effort to formulate theories relat-
ing neural and cognitive processes is a false analogy being drawn between the
indisputable empirical and theoretical successes in explaining sensory and motor
codes, on the one hand, and the much more complicated matter of understanding
the neurophysiological representations of cognitive processes, on the other. Periph-
eral sensory and motor signal transmission is characterized by a mainly unidirec-
tional flow of information, relatively simple dimensional correlations between
stimuli and neural responses, as well as a dimensional isomorphism between stimuli
and perceived experiences. Perhaps most important is the fact that sensory and
motor signals are anchored to the full gamut of physical measures and forces.
Furthermore, the microanatomy of the peripheral sensory and motor pathways
is much simpler than of those more intricate and interwoven central mechanisms
that are presumably involved in cognitive processes. Sensory signals have a pre-
dominant afferent directionality, whereas motor signals have a predominant efferent
directionality. Furthermore, sensory mechanisms often have a repetitive anatomy
that is almost crystalline in nature, especially in invertebrates. As a result of the
relatively simple anatomy and transmission roles of the sensory and motor path-
ways, it has been much easier to develop theories, indeed, nearly complete expla-
nations, of transmission codes. The heyday of this kind of work was epitomized
by the work of Hartline and Ratliff (1957) and Hubel and Wiesel (1965) and an
army of sensory and motor neurophysiologists up to the present.
Thus, it is possible to ask and answer a question such as—what are the neu-
rophysiological codes (i.e., the pattern of neural responses) used by the peripheral
nervous system to transmit information about the magnitude or quality of an
acoustic stimulus to more central regions of the brain? The physical dimensions
of both the stimulus and the neural response are measurable in units for which
the physical sciences have provided well-established values such as lumens and
decibels on the one hand, and frequency and pattern of firing on the other.
Furthermore, psychophysicists have developed powerful techniques for quantify-
ing responses that require but the simplest of “Class A” discriminative responses
(Brindley, 1960). These simplifying conditions do not hold for cognitive processes
such as decision making or emotion, situations in which we are not really sure
what the proximal stimulus really is.
16 Introduction

Unfortunately, general disappointment was the main result when the same
protocols and concepts were used and as false analogies were drawn between
sensory and motor information transmission processes on the one hand, and cogni-
tive activities on the other. These optimistic (and probably false) analogies persist
to this day. In short, determining the neurophysiological basis of cognition is a
much more challenging process than is the analogous task for sensory and motor
processes. Progress in one does not necessarily portend progress in the other.

The Neurologizing of Psychological Language

A property of many psychological theories is the introduction of neurophysiologi-
cal terminology into their discussions without adequate empirical linkage. Behav-
ioral hypothetical constructs are supplemented by neural concepts and findings
based on little more than functional analogies and borrowed language. Skinner
(1950) referred to this use of unfounded neurophysiological concepts as a Con-
ceptual Nervous Systems (CNS) approach, noting that:

Many theorists point out that they are not talking about the nervous system
as an actual structure undergoing physiological or bio-chemical changes
but only as a system with a certain dynamic output.
(p. 194)

What it appears that many investigators (e.g., Anderson, 2010; Johnson, 2011)
are doing is to develop plausible, but highly speculative, “hypothetical constructs”
that could, in principle, describe the behavior but which do not enjoy the sup-
port of empirical linkages between the psychological and neural mechanisms.
The hypothetical mechanisms are then defined in neurophysiological terminology.
At best, theories of this type are heuristics that are not likely to rise from pro-
totheoretical hypotheses to robust neural theories. Perhaps we should be reevalu-
ating the important insight about the Conceptual Nervous System offered by
Skinner and distinguish between robust neurophysiological entities and “nick-
names” for hypothetical constructs.

1.4 The Macroneural and the Microneuronal

It is vital to make a very important distinction at this point. The ultimate goal
of the neuronal network theorist per se is either to develop a map of salient
networks; or, if that is not possible, then it is to understand the general principles
of networks such as those occurring in the brain. This is essentially a microscopic
approach. The current goal of the cognitive neuroscientist, on the other hand,
is to associate a particular mass of neurons (i.e., a location) with some cognitive
activity by comparing the outcomes of experiments that involve both anatomical
and functional methodologies. This is essentially a macroscopic approach. These
 Introduction 17

two goals, it should be emphasized, are not the same and, for the moment, no
bridge exists between the anatomical and the cognitive.
Each of these two approaches is challenged by technical and conceptual dif-
ficulties. Difficulties with the microneuronal approach (in which the integrity
of the individual neuronal response is preserved) are counterbalanced by difficul-
ties with the macroneural approach (in which the responses represent pooled or
cumulated information that represents cognitive processes). Technologically, these
two perspectives (the microneuronal and the macroneural) are very different in
terms of their basic assumptions and the ease with which their respective dif-
ficulties can be overcome. Macroneural measures are cumbersome, noise ridden,
and often quite arbitrary in their evaluation. They may require elaborate statistical
analysis to extract localized regions of activity that might correlate with cognitive
processes. This is the basic presumption forced on macroneural theorists by the
instruments used to measure them. Although often expensive and time consum-
ing, macroneural experiments are relatively easy to design to test their assumptions
using such devices as the EEG and the fMRI. The main problem with these
macroneural neural measures, however, is that they are cumulative responses
pooling or summating the responses of individual neurons and their intercon-
nections. Thus, much of the detailed information of microneuronal organization
is lost. I have discussed the weaknesses of the macroneural approach much more
completely elsewhere (Uttal, 2001, 2011, 2013, 2016).
Should the competing hypothesis—that the essential aspects of cognitive
encoding are to be found at the microneuronal network level—be correct, then
macroneural representation would be a deeply flawed approach to understanding
mind–brain relationships. Rather than measuring macroscopic blood-oxygen levels
with fMRIs as an indirect indicator of brain activity or summations of electrical
activity from scalp electrodes with EEGs, we should be studying the microscopic
action and interaction of neurons—and in particular the networks into which
they are organized.
Empirical studies of microneuronal events are, on the other hand, much more
difficult to implement experimentally. Most inhibiting to the microneuronal
approach is that we have no specific knowledge of how many neurons are needed
for a network to be “cognitively significant.” It is likely that the number is very
large—possibly as large as the number of neurons in the entire brain—but this
is pure conjecture.

The Technological Basis of Macroneural Theories

In the spirit of the basic premise of this book that our technology strongly
influences both our experimental results and our theories, I now briefly review
the specific technology that has stimulated the current excitement about the
macroneural approach using fMRI devices. The rush in recent years to develop
macroneural theories and findings in cognitive neuroscience was presaged by
18 Introduction

the discovery of Nuclear Magnetic Resonance (NMR) by Rabi, Zacharias, Mill-

man, and Kusch (1937, 1939). NMR was predicted by the emergence of quantum
mechanics as a dominating theoretical approach during the 1930s. Although
Rabi’s group worked with gases enclosed in a Crookes tube, other investigators
were shortly after able to reproduce the NMR phenomenon in solids and
liquids.
The term NMR is somewhat self-defining. It is a phenomenon of the nuclear
and subnuclear behavior of matter. The phenomenon depends on the magnetic
properties of particles such as protons, which act as tiny magnets whose poles
are in most situations randomly aligned. However, when a very strong magnetic
field is applied to a substance (which may include organic tissues as well as gases
and liquids), the magnetic orientation of all of the susceptible particles align in
the same direction. If the strong magnetic field is released after a measured
amount of time, the orderly proton magnetic alignments revert or relax back to
their random state. If a small magnetic pulse stimulates the material during this
“relaxation,” the protons will emit radio waves at resonant frequencies that are
dependent on the nature of the material of which the protons are a part and
the nature of the small magnetic stimulating field.
These emitted radio signals can be picked up and after complex mathematical
manipulations converted into three-dimensional spatial images of the material
under study. The key aspect of this process is that different organic tissues will
produce different patterns of radio frequency emissions. This was the major
contribution of Damadian (1971), who first applied NMR techniques to imaging
the human body. It was his inspiration that the NMR techniques developed in
the laboratory for various gases and solids might be able to distinguish between
normal and cancerous tissue in the clinic. Although his procedure was extremely
primitive compared with modern equipment, clearly Damadian was the first to
develop the imaging application of NMR. Subsequently, the brain-imaging
strategy based on NMR physics revolutionized modern neuroscience. Nowadays,
imaging methods based on NMR physics have largely replaced a wide variety
of traditional methods involving surgery, trauma, stimulation, and electrophysi-
ological recording.
The next step of consequence for cognitive neuroscience’s role was the devel-
opment of functional Magnetic Resonance Imaging (fMRI) by Ogawa, Lee, Kay,
and Tank (1990). They showed how slight differences in the oxygen level of
the blood could differentially determine the activity level of portions of the
brain at different times and locations. The Blood Oxygen Level Dependent
(BOLD) measures could thus be used to track functional brain activity as well
as anatomical structure.
The earliest application of the fMRI technique to cognitive neuroscience is
generally considered to have been published by Belliveau et al. (1991). (See the
article by Kwong, 2013, for an interesting historical discussion of this ground-
breaking experiment during which attempts were made to map out regions of
 Introduction 19

the visual cortex that were selectively responding to visual stimuli.3) From this
point in the history of this powerful tool, the influence of the fMRI technology
on cognitive neuroscience cannot be overstated. Theoretically correct or not,
fMRI technology changed the face of the science and created a zeitgeist in which
the macroscopic measurements of brain activity were to dominate both experi-
mental protocols and theoretical positions for years. Once again, the historical
role of brain-imaging systems illustrates the strong impact that whatever technol-
ogy is available to science can have on conceptual basis as well as its empirical
findings. It was only decades later that some of the conceptual and technical
deficiencies of the macroneural approach began to emerge.

Macroneural Theories—A Brief Review

Because of the current availability of particular kinds of instruments for measur-
ing brain responses, the macroneural approach is by far the most popular for
studying the functional (including the cognitive) role of the brain. The archetype
of these instruments is epitomized by the fMRI system. It has many advantages,
not the least of which is its noninvasive and benign nature. Currently, the number
of articles published each year that deal specifically with comparisons of cogni-
tive processes and fMRI brain images is difficult to determine because not all
research using this method is related to what is currently referred to as the
mind–brain problem (some have been directed quite successfully to related
problems in medicine, neuroanatomy, or neurophysiology). Nevertheless, it is
obvious that experiments comparing fMRI responses and cognitive processes
have gone from virtually nothing in the early 1990s to tens of thousands in the
2000s (e.g., Bandettini, 2007; Aue, Lavelle, and Cacioppo, 2009).4
Because they are brain responses putatively correlated with cognitive processes,
there is a compelling face validity presented to researchers by these new machines.
The promise is that we are on the verge of a breakthrough in the profound and
hitherto intractable mind–brain problem—how does the activity of tangible brain
tissue produce intangible cognitive processes? The hope persists that we will
eventually be able to meld what turns out to be an enormous database into a
coherent, valid, and comprehensive theory of mind–brain relations. This face
validity is, however, countervailed by a pervasive logical error that incorrectly
assumes that because the mind is a brain process, any brain activity correlated
with cognitive activity must be relevant to their interrelationship. At the least,
it has been hoped that eventually correlations will be found between cognitive
and neural responses that will serve as heuristics for plausible future theories of
mind–brain relationships.
The conceptual postulate that dominates much of current macroneural cognitive
neuroscience research is what we might refer to as the “traditional localization”
concept. Essentially, this type of research is empirical but not directly theoretical;
it is but a preliminary step in the acquisition of data that might help to build
20 Introduction

more comprehensive theories in the future. “Successful” experiments are those

that are able to consistently associate particular localized peaks of regional activa-
tions with particular cognitive stimuli or task conditions. To the degree that these
associations are robust and reliable, investigators try to add to a functional map of
brain locations as their prototype theory. However, the basic idea that there are
significant, localized activation peaks that selectively represent cognitive processes
is now under attack. New experiments suggest that the “peaks” may be artifacts
of sample size and statistical analysis procedures that prejudice the existence of the
peaks. In fact, if these studies are correct, peaks of brain activity identified by
brain-imaging equipment are artifacts in which our statistical techniques have
created a false orderliness out of what are actually stochastic processes.
There, thus, prevails a highly questionable assumption guiding the research
protocols in this macroneural mode of attack on the mind–brain problem, that
is, that the inferred modular components of a cognitive process will map in some
repeatable and neurophysiologically coherent way onto localized regions of the
brain. As widely accepted as this hypothesis is, there have been persistent logical
reasons and are now an increasing number of empirical reasons to question it.
The psychological processes and phenomena for which localized representations
are assumed are not necessarily dimensionally isomorphic with the brain’s natural
spatial layout. The cognitive processes are, it must also be remembered, themselves
typically the instantiation of our experimental designs and not necessarily of any
simple property of functional behavior. Whether they are divisible into the
“hypothetical constructs” or intervening “modules or faculties” that correspond
to specific anatomical regions or structures of the brain is uncertain.
The implications and inferences of hypotheses and theories must then be
empirically tested to determine if they continue to hold more generally. However
easy to put this essential step of the process into words, it is not that simple.
Indeed, it is conversely true that when you are dealing with a system whose
stimuli are at least multi-factorial and for which the triggering stimuli are obscure,
whose responses are multi-dimensional and redundant, and for which there may
be no direct relation of stimulus to cognitive response, the probability of finding
any kind of a response that satisfies the a priori theoretical judgments of inves-
tigators becomes greatly enhanced.
The primary questions that may be answerable with brain-imaging techniques
are variations on the theme of spatial localization; that is, where are the parts of
the brain that are concomitantly activated with particular cognitive processes and
how are they interconnected? This question can be approached from a number
of different points of view, each with subtly different connotations. Although
these differing connotations will become evident as I review the various kinds
of research carried out that attack the problem of mind–brain relations at this
macroneural level of analysis, it is useful at this point to tabulate some of the
issues that have been considered by investigators using brain-imaging devices in
their search for the foundations of a mind–brain theory at the macroneural level:
 Introduction 21

1. As noted, the archetype of the question asked by cognitive neuroscientists who

use brain-imaging equipment and techniques is—where are the parts of the
brain that become active when a particular cognitive process is under way?
2. Another level of inquiry is how, both in general and specific terms, are the
salient brain areas interconnected at the macroneural level for cognitive opera-
tions? This was initially conceptualized as an anatomical question dealing with
the nature of the interconnections (mediated by white, i.e., myelinated bands
of axons) between regions. However, there is a functional analog that has gen-
erated much current interest—how are the parts of the brain functionally inter-
connected when particular cognitive processes are being carried out? Theories
of this kind require much more elaborate analytical algorithms than simple
tabulation of activated regions. However, they are really just another version of
answers to the basic “where” question, one generalized to systems of interact-
ing nodes rather than unique function-specific locations. The ultimate goal
of this approach is to determine the connectivity among brain regions during
cognition and then describe the properties of the resulting networks.
3. Classical physiological psychology has traditionally been aimed at associating
anatomical brain mechanisms with particular cognitive processes by extirpa-
tive, stimulating, or recording techniques. However, in large part those kinds
of research are beset by ill-defined and invasive surgical procedures and pre-
existing assumptions about regional functions. In large part, such techniques
have been replaced by imaging techniques. Traditional work done using
lesioning as the main tool (e.g., Kennard, 1955) has been much reduced, to be
replaced mainly by fMRI techniques.
4. A classical problem that seemingly has been empirically resolved in the favor
of distributed responses is the debate between those who argued that the brain
correlates of cognitive processes were localized, function-specific regions and
those who thought that the responses were distributed over broad multifunc-
tional regions of the brain. The accumulating scientific evidence seems to
increasingly support the latter conclusion. However, the details of this debate
remain controversial, attracting the attention of a number of investigators.
5. Some cognitive scientists have proposed that current macroneural research
with brain images will be able to resolve some purely psychological controver-
sies. The question thus arising is—can neurophysiology inform psychology?
This is a form of hypothesis testing that depends on testable neurophysiologi-
cal postulates being included within what are otherwise purely psychological
theories. The debate over the applicability of brain-imaging data rages on.
(See the discussion between Coltheart, 2006, an opponent of the idea that
any psychological theory controversy has yet been resolved by brain-imaging
techniques, and Henson, 2006, who believes that they have.)
6. A major long-term and highly controversial issue in brain-imaging cognitive
neuroscience is—can these techniques be used to read the mind, that is, to
tell what a person was thinking about or perceiving by examining the fMRI
22 Introduction

data? Although there has been some progress in selecting alternate sensory
responses from among a limited set of brain images, much of this modest
achievement seems to be attributable to topologically preserved peripheral
sensory encoding. For example, visual stimuli are represented by retinotopic
maps in the primary sensory area that preserve the topology of stimuli and,
thus, may maintain accessible and useful information about the spatial pattern
of a stimulus. This is not possible with the symbolic brain representations
driven by higher-order cognitive processes that have no isomorphic relations
between brain activity and those cognitive processes.
7. Many cognitive neuroscientists, heavily influenced by earlier psychological tax-
onomies, are trying to use brain-imaging devices to determine something about
the localized brain mechanisms by which cognitive processes are carried out.
For example, how do we learn? What brain changes occur in learning? A major
unsolved problem is where and what is the engram? Others seek to understand
the emergence of consciousness, attention, and other vaguely defined high-level
cognitive processes by determining which brain regions are activated when
these cognitive processes are manipulated. This is the main theme of much of
brain-imaging research these days. However, there are compelling reasons to
believe that this simplistic concept of seeking correspondences between cogni-
tive processes and specific brain locations may be ill chosen.
8. Many technical issues occupy the time and energies of cognitive scientists,
not all of which are aimed at the great question of how the mind emerges
from brain processes. There is a continued effort to develop techniques to
extract the best possible and largest amount of data from noisy brain images.
Still other investigators are concerned with the technical matter of how we
can pool or combine methods with low statistical power to produce higher-
power experiments in order to yield more significant data—the meta-analysis
approach. There remains an unresolved question when we use this method—
can we plausibly combine experiments that are often varied in method and
conceptualization? Depending on the actual signal-to-noise relations and
whether a signal actually exists or is a manifestation of a stochastic system,
this may also be an ill-chosen expenditure of resources.
9. Most generally, cognitive neuroscience is currently aimed at establishing the
macroneural neural basis of cognition. In short, the main goal of this science
is to provide some insights into the great question—how does the brain make
the mind? To this overarching question, there is little in the way of either an
answer or a satisfactory theory yet available. It is highly problematic whether
we have even begun to answer this question given the likelihood that macro-
neural techniques such as the fMRI may actually obscure the critical micro-
neuronal information that would answer this question.

Obviously, whereas some of these questions and goals represent issues that are
of existential importance, others are of merely epistemic significance. Equally
 Introduction 23

obviously, most of the cognitive neuroscience questions posed here are currently
answerable. For those that can be answered, it is my hope that the subsequent
discussions in this book will at least help to clarify the issues arising.
Like all other empirical approaches directed at a solution of the mind–brain
problem, there are barriers and difficulties that impede progress toward a solution.
Some are general to all cognitive neuroscience strategies while others are specific
to the macroneural approach. One that falls into the latter class is the challenge
faced by the pooled nature of the responses, that is, by the fact that any macro-
scopic measure is a composite of the individual responses of many individual
neuronal responses. Currently, there is a prevailing opinion that it is the ensemble
coding or states of the individual responses that in some ultimate sense must be
understood if we are ever to make progress in unraveling the brain’s role in
cognition. Unfortunately for this approach, the pooled signals (the physical sum
of many neuronal responses) actually have lost the critical information.
In such a situation, in which an exhaustive microneuronal analysis is beyond
our technical capabilities, science tends to turn to available macroneural cumula-
tive measures (e.g., the fMRI or the EEG) and effectively let nature do the
accumulative analysis for approximate solutions to complex neuronal network
problems. The expectation is that the critical microneuronal information will
be at least partially preserved at the macroneural level. How the cumulative
process works neurophysiologically is not completely understood, but it may
involve summations of either local (e.g., Logothetis et al., 2001) or spike action
potentials (e.g., Mukamel et al., 2005).
Currently, it seems likely that the salient microneuronal information is not,
in fact, preserved in macroneural level measurements and, thus, the details of
neuronal responses and interactions cannot be retrieved from cumulative measures
such as fMRIs. The voltages and ionic currents that are the responses of individual
neurons must follow the laws of physics as they are added and subtracted from
each other. Once added or accumulated into pooled responses, basic thermody-
namic principles argue that the original initial responses of a given neuron cannot
be retrieved any more than an egg can be unscrambled. This constraint applies
directly to the pooling of neuronal or hemodynamic responses directly or indi-
rectly into signals like the fMRI or the EEG. Thus, there is an innate ambiguity
about the origins of any macroneural signal based on accumulations of micro-
neuronal activity—a huge variety of different configurations of microneuronal
neuronal activities may produce exactly the same macroneural response. This
means that scores based on differences between macroneural responses—for,
example, the subtractive method—produce data that cannot discriminate between
different microneuronal brain states. This issue is such an important matter that
it has been continuously discussed for almost two decades (Van Orden and Papp,
1997) and is still in need of defense (Roskies, 2010).
The implication of the fact that the same macroneural responses can be
produced by a variety of different microneuronal responses is that macroneural
24 Introduction

fMRI images are neutral with regard to their microneuronal origins. This is the
fundamental weakness of the macroneural approach; the critical microneuronal
information essential to understanding the transition from brain state to cognitive
activity may have been irretrievably lost as a result of the pooling process.
The one possible exception to this neutrality of fMRI images is that informa-
tion may be preserved concerning the respective locations of function-specific
nodes, regions, activation sites, or locales on and in the brain—if they actually
exist. Regardless of the truth or falsity of this fundamental postulate, finding the
brain loci associated with cognitive processes has been the core of the larger
portion of the research carried out with current fMRI equipment. Should it turn
out that the macroneural representations of specific cognitive activities are actually
valid and cognitive processes are localizable to particular nodes or activation sites
on the brain, the location of the respective cognitive responses might be considered
to be preserved information. However, even this most basic postulate of brain
organization—localization—may not be empirically supportable.
Another closely related barrier to macroneural theory building is generically
referred to as underdetermination. Underdetermination implies that there is
insufficient information in the results available from an experiment or group of
experiments to answer the salient questions for which the research was carried
out. Some of this information was lost as a result of pooling as previously
discussed—microneuronal information is not preserved in macroneural signals.
However, it is more likely that most of it is due to the fundamental conceptual
difficulty that the information necessary to answer a cognitive neuroscience
question was never present in available macroneural measurements. Both behavior
and formal mathematical models, for example, are underdetermined in that they
do not contain enough information to uniquely determine the specific underly-
ing mechanisms that produce the behavior (Moore, 1956). Indeed, neither behavior
nor mathematics can distinguish among what may be an innumerable number
of alternative possible and plausible underlying mechanisms.
Furthermore, additional experiments cannot always be carried out that will
resolve this deficiency. Hilgetag, O’Neill, and Young (1996), for example, have
argued on purely mathematical grounds that complex networks such as those
proposed by Van Essen, Anderson, and Felleman (1992) for the primate visual
system cannot be placed in a hierarchy of activation order because the outcomes
are fundamentally underdetermined. Carrying out additional experiments leads,
paradoxically according to Hilgetag and his co-workers, to an increase in the
number of plausible mechanisms at a rate that exceeds the number of new
experiments that can be designed to determine the response hierarchy. In short,
simply collecting more data may not help to resolve issues of underlying mecha-
nism if the data are underdetermined; indeed, it may exacerbate the problem.
Another way to describe underdetermination is closely linked to potential
neural mechanisms. I refer here to the “many to one” constraint. If, for
 Introduction 25

argument’s sake, we accept the possibility of explaining behavioral observa-

tions in terms of macroneural mechanisms, then there is no reason to assume
that a particular mechanism is solely capable of encoding any particular kind
of behavior. It is far more likely that there are many different mechanisms
that can equally well provide a satisfactory representation of the neural
mechanisms associated with a particular behavior, that is, many neural mecha-
nisms can produce the same behavior. Thus, a behavior is underdetermined
with regard to the identification of a particular explanatory neural mechanism.
It is also true that the inverse problem posed by the activation of a particular
neural mechanism cannot be used as an indicator of a particular cognitive
process. This is the problem of “reverse inference” highlighted by Poldrack
(2006).

New and Relevant Data Arguing Against Macroneural Theories

An obvious difficulty that has to be acknowledged by cognitive neuroscientists
is the undeniable (yet all too often denied) fact that the empirical foundation
for robust macroneural theory building is still in an underdeveloped state. Com-
prehensive and accurate theories extrapolating from specific experiments to
general laws are still relatively underdeveloped compared with the physical sci-
ences. It is not yet certain whether this is a fundamental intractability or just
the birth pangs of a new science like cognitive neuroscience.
If the ideal goal of a theory is to integrate data and abstract general principles,
then there is an a priori requirement that the empirical foundation be robust,
repeatable, and consistent; that whatever general principles are being expressed
should actually be represented by dependable data. In the past several years, some
investigators have reported experiments or analyses that challenge some of the
foundation assumptions of macroneural theory building. In their most extreme
versions, these assumptions include:

• That the brain is organized into regional subdivisions that encode or represent
cognitive functions
• That these regions will exhibit a distinctive local neurophysiological response
when a cognitive stimulus is activated
• That activation of one of these regions by, say, an electrical response will influ-
ence a relevant behavioral, cognitive, or neural response
• That reliable experiments can be reported in accord with robust experimental
and statistical paradigms that permit replication of those experiments

I now review some of the most compelling arguments against these assumptions.
Collectively these experiments suggest that serious questions are raised about
macroneural theories at their most basic conceptual levels.
26 Introduction

Thyreau et al. (2012) and Gonzalez Castillo et al. (2012)

Robust evidence supporting widely distributed brain states (as opposed to local-
ized activations) of cognitive activity has now begun to appear in the literature
(e.g., Gonzalez Castillo et al., 2012; Thyreau et al., 2012). Both of these groups
of investigators presented compelling data that the apparent separation of brain
responses into localized regions or functional nodes may be an artifact of inad-
equate sample sizes, arbitrary p values, statistical biases, and a prevailing presup-
position of sparse localization. Gonzalez Castillo and his colleagues, for example,
suggested that when adequate sample sizes (up to 500) were used to average
brain images, localized activation areas in an individual subject tended to disap-
pear, to be replaced by what is nearly a uniform overall activation of the entire
brain. They did not exclude some relatively large regional differences but they
concluded that:

. . . under optimal noise conditions, fMRI activations extend well beyond

areas of primary relationship to the task; and blood-oxygen level-dependent
signal changes correlated with task timing appear in over 95% of the brain
for a simple visual stimulation plus attention control task.
(p. 5487)

Their results indicated that the more data that were included in the analysis,
the greater was the extent of the distributed activity across the brain and less
evidence existed for localized function. To support this result, Thyreau et al.
(2012) reported a similar experimental result in which they also used a very
large subject sample (n = 1,326) and found generally the same result—widespread
distribution of responses across the entire averaged brain for even the most
elementary cognitive process. They pointed out that when very large samples
were used, even small responses began to achieve significance; this expansion
increased until virtually the whole brain was responding significantly. The
important implication of their work, like that of Gonzalez Castillo and his col-
leagues, was that the apparent segregation of the human brain into cognitively
specialized regions associated with particular cognitive processes or brain activa-
tion nodes may possibly have been an artifact of inadequate sample size.
Oosterwijk et al. (2012) also supported the idea that the brain mechanisms
of cognition must be more widely distributed in the brain than is currently held
by localization theorists. From their point of view, the assertion that the neural
mechanisms of cognitive processes are localized in the form of macroneural
function-specific regions scattered across the brain was fundamentally incorrect.
Specifically, they noted:

Our results also emphasize the importance of examining distributed pat-

terns of brain activation to understand mental states with different content
 Introduction 27

rather than focusing on single regions . . . isolated regions may serve

different psychological functions depending on what they are connected
to during a given instance.
(p. 2125)

The point of these studies is that the traditional search for localized macroneural
regions mapping onto behaviorally defined psychological constructs is likely to
be fruitless. The brain is not a system of isolated function-specific nodes; it is
more likely to be a broadly, even universally, distributed system of neural mecha-
nisms that are neither function-specific nor localizable to any particular region.
Such a whole-brain metaphor is vastly different than the current one guiding
much of macroneural thinking about how the brain is organized. In sum, the
general import of this work is that the search for narrowly localized brain regions
associable with cognitive processes may be sterile. The conflict between these
emerging data and the search for localizable neural mechanisms may require a
major change in whatever kind of theoretical model is being pursued.

Carp (2012a, 2012b)

How cognitive neuroscience data are reported is another general problem area
that has recently come to the notice of the scientific community. The problems
generated by the practical details of presentation and statistical analysis of fMRI
measurements may be as potentially distorting to our theories as our choice of
the level or kind of analysis at which we should work. We have recently been
reminded of two serious problems concerning reporting of brain image-based
experiments by Carp (2012a, 2012b). In the first of these two papers, he pointed
out that the replicability of research in this field depends on adequate informa-
tion being available concerning the design of an experiment so that the experi-
ments can be reproduced. Yet, despite this universal necessity for replicability,
when Carp analyzed 241 recent reports he found that “many did not report
critical methodological details” (p. 289) in sufficient detail to permit replication.
Not only was there a wide variety of procedures used for similar experiments,
but also even the same data could be analyzed and reported in quite different
ways, giving rise to diametrically opposed interpretations. Problems in reporting
were identified by Carp concerning the design, the data acquisition, and pre-
processing methods, as well as in modeling and display of the final results.
The problem of inadequate reporting, thus preventing adequate replication, is
serious; however, another problem of perhaps even greater import to fMRI-based
cognitive neuroscience research was also examined by Carp (2012b) in a follow-
up article aimed specifically at the variability of the analytical methods that were
used. He concluded that variations in method could lead to a cryptic increase in
false-positive results. This hazard is exacerbated by the multiple modes in which
data may be analyzed in an fMRI-based cognitive neuroscience experiment.
28 Introduction

According to Carp, there are many alternative parameters of available analysis

procedures with which to evaluate brain-imaging data. Depending on which
subset of the parameters was chosen for a particular experiment, alternative
conclusions could be drawn from a brain-imaging experiment. Carp suggested
that ten is a plausible number of the key analytical parameters (including, for
example, different kinds of normalization, filtering, and autocorrelation) as well
as two or four alternative means of dealing with each of these analytical param-
eters. On this basis, Carp computed that there were 6,912 unique analytical
“pipelines” that could be followed to a conclusion about the presence or absence
of a significant effect. As Carp pointed out, however, not all of these pipelines
lead to the same conclusion; indeed, there is a possibility of many competing
Type I and Type II errors emerging from what were essentially the same mea-
surements, depending on the investigators’ more or less arbitrary choice of the
particular pipeline that was used. In Carp’s words:

. . . results also varied considerably from one pipeline to another. Estimates

of activation strength were highly variable across analytic pipelines; in regions
of peak overall activation, significance estimates varied by 8 Z units.
(Carp, 2012b, p. 10)

Although Carp does not assert that all experimental results vary across all
pipelines, what he does argue is that “others varied widely from one pipeline
to another” (p. 12).The fact that some pipelines, even the preponderance of
them, were to some degree in agreement (a point made strongly in Carp, 2012a)
does not mitigate the fact that there was an increased tendency for false positives
to occur as a function of the variability in analytical methodology.
The conclusion to which we are drawn by this cogent critique and the concept
of alternative pipelines is that variation in analytical method might be driving
research conclusions (such as whether they are significant or not) to an unknown
degree. Thus, great uncertainty is injected into even basic decisions about the
significance or non-significance of the raw results. Should our scientific results
vary with method and not converge for different methods, they would not inspire
much confidence in their validity as measures of the underlying neural reality.
Similar points have been made by Simmons, Nelson, and Simonsohn (2011)
and Bakker and Wicherts (2011) for purely psychological studies. Both groups
also noted that there is such enormous flexibility in the way we collect and analyze
data that spurious false positives are encouraged. They emphasized several additional
factors as contributing to poor decisions that lead to error: (1) ambiguity about
the basis for the p < .05 criterion, (2) the arbitrariness of when to terminate data
collection, and (3) simple calculation errors. Bakker and Wicherts, for example,
stated that as many as 33% of the 281 reports they reviewed either misreported
their findings or made an error in calculation.
 Introduction 29

Horvath, Forte, and Carter (2015a, 2015b)

The general idea behind macroneuronal theories of cognition is that by appro-
priately activating some aspect of cognition, we can localize regions of the
brain that are in some way responsible for representing that cognitive process.
There is, however, a related idea that is based on the inverse of this idea—
namely that we can stimulate certain regions of the brain to manipulate, influ-
ence, or evoke specific cognitive processes. In the former case, we are searching
for the brain locations associated with predefined cognitive states. In the latter
case, we are assuming that certain brain areas are related to certain cognitive
states and then attempting to manipulate those cognitions by stimulating the
brain locales with electrical stimuli. Both approaches are based on the same
fundamental premise—namely that cognitive processes are mediated by specific
macroscopic brain locations.
Work using transcranial Direct Current Stimulation (tDCS) has blossomed in
the past decade with suggestions that electrical stimulation of localized regions
of the brain can have profound effects on cognitive functions and even psychi-
atric problems. Dozens if not hundreds of reports now purport to document
the psychological effects of electrical stimuli directly applied to the skull. Like
many other enthusiastic applications of new technologies, it now appears that
what seem to be a plethora of positive effects may actually be little more than
random noise. A team led by Jared Horvath has reviewed and statistically ana-
lyzed large samples of tDCS experiments that implemented this localization-by-
stimulation protocol. In two articles (Horvath, Forte, and Carter, 2015a, 2015b),
on the basis of an extensive statistical analysis of the germane literature, Horvath
and his colleagues concluded that there are no reliable effects of this kind of
electrical stimuli on neurophysiological and behavioral measures chosen from a
diverse sample of different types of experiments. Positive effects seemed to be
balanced by negative reports over a wide range of cognitive experiments. The
single exception to this general result was that motor twitches varied systemati-
cally in amplitude with tDCS stimulus amplitude. This is hardly a surprising
result given that the stimuli were being applied over motor areas of the brain.
Horvath, Carter, and Forte (2014) went on to suggest that this poor reliability
could be accounted for by five types of inadequate controls:

1. Intersubject variability
2. Intrasubject variability
3. Absence of sham stimulation and blind analysis techniques
4. Motor and cognitive interference
5. Controls for physical nature of the direct current stimuli

When examined in the light of the equally poor controls and variable results
obtained with the fMRI approach, this critique speaks strongly to the idea that
30 Introduction

macroneural theories based on localized brain regions associated with cognitive

processes, so widespread in current cognitive neuroscience, are inadequate.

Barrett (2015) and Barrett and Satpute (2013)

Another group that has been recently publishing research that is in contradiction
to the idea that there are specific locations in the brain that encode or represent
specific cognitive processes is led by Lisa Feldman Barrett of Northeastern Uni-
versity. She summarized this group’s conclusions about the localization of emotions
in an opinion piece in the New York Times (Barrett, 2015) in the following way:

. . . [we] collectively analyzed brain-imaging studies published from 1990

to 2011 that examined fear, sadness, anger, disgust, and happiness. We divided
the brain into tiny cubes, like 3-D pixels and computed the probability
that studies of each emotion found an increase in activation in each cube.
. . . Overall we found no brain region was dedicated to any single emotion.
We also found that every alleged “emotion” region of the brain increased
its activity during nonemotional thoughts and perceptions as well.
(p. 10)

As well as:

Although it has long been assumed that emotional, social, and cognitive
phenomena are realized in the operations of separate brain regions or brain
networks, we demonstrate that it is possible to understand the body of
neuroimaging evidence using a framework that relies on domain general,
distributed structure function mapping.
(Barrett and Satpute, 2013, p. 361)

Button et al. (2013) and Uttal (2013)

Recently, a general appreciation that not all was well with imaging as the basic
tool of cognitive neuroscience research—fMRI—has emerged. Two studies (But-
ton et al., 2013; Uttal, 2013) pointed to statistical problems with the reliability
of brain-imaging studies created by small sample size. The problem was that
because of the expense of running subjects, many experiments were deficient in
statistical power—a correlate of inadequate sample size. This permitted many
statistically significant studies to be interpreted as having demonstrated positive
results when, in fact, they had not.
Button et al. pointed out that low power led to “overestimates of effect size
and low reproducibility of results” (p. 365) in general and of brain-imaging
studies in particular. This meant that “the chance of discovering effects that are
genuinely true is low” (p. 366). Beyond statistical formalities, this means that
 Introduction 31

data from experiments will be highly variable even though individual experi-
ments might be significant. Uttal showed that this lack of reliability was per-
meating the entire cognitive science application of fMRI machines by comparing
a variety of experiments. The general effect was to produce variability and
inconsistency despite what must have been acceptable levels of significance.
By 2005, the problem had become widely recognized and resulted in such
hyperbolic statements as “Neuroscience research gets an “F” for reliability” in
a Scientific American blog, “Brain science is drowning in uncertainty” in New
Scientist, and “Brain-imaging is often wrong” in Vox, a “general news site.” More
seriously, submissions of this kind of research to various journals were down,
especially if they dealt with comparisons of cognitive and imaging data. Efforts
to overcome the technical and statistical problems resulted in grants to fund a
“Center for Reproducible Neuroscience” at Stanford University. Although there
may be some corrections to be made in the way data are collected and analyzed,
it is not certain that further investment with what otherwise may be a flawed
approach to the study of brain imaging carried out at the wrong level of analysis
as a pathway to understanding the mind–brain problem is justified.

1.5 The Rise of Distribution—the Fall of Phrenological

Localization
That we would be able to identify localized brain regions encoding specific
cognitive functions was the hope of the early work using fMRI systems. This
led to what some of us (e.g., Uttal, 2001) described as a neophrenological ori-
entation to our theories in which a narrowly localized area of the brain was
associated with poorly defined cognitive processes. However, increasingly, leading
investigators such as Barrett (2015), Fox and Friston (2012), Button et al. (2013),
and Poldrack (2010) have pointed out that the goal of assigning localized brain
regions to cognitive processes has not been and possibly cannot be achieved.
Fox and Friston look back over two decades in which a flood of papers sought
to relate macroscopic brain locations and cognitive processes and concluded:

Over the past twenty years, neuroimaging has been the predominant tech-
nique in behavioral and cognitive neuroscience. The volume of papers and
number of fields it pervades are unrivaled. Despite this, it is curiously
difficult to summarize its achievements in general terms. The simplest
attempts falls [sic] back on two principles that shaped brain mapping at
its inception; namely functional segregation and integration: Neuroimaging
has established functional segregation (the segregated or modular deploy-
ment of functional specialization within brain regions) as a fundament of
brain organization. . . . However the initial hope of associating each brain
area with a particular function (Posner et al., 1998) has not been realized.
While it is true that notions like the “motion sensitive center” and
32 Introduction

“fusiform face area” are part of common imaging parlance, the functionally
informed labeling of all but the smallest portion of cortex remains elusive.
Indeed people now prefer to talk about processing hierarchies, intrinsic
networks and default modes that have no clear association with discrete
cognitive processing components.
(p. 408, italics in original)

Fox and Friston argue that the field has changed or is in the process of
changing and we should no longer attempt to make simple associations between
brain locales and cognitive processes. Although we are likely to continue to use
brain-imaging equipment, they suggest that future work will increasingly be
concerned with the anatomical organization of brain exemplified by the search
for the connectome (e.g., Sporns, Tononi, and Kotter, 2005), the effort to deter-
mine how the brain responds when it is cognitively inactive (e.g., Fox et al.,
2005), or, more generally, “a more physiologically and anatomically informed
approach” (p. 408). Fox and Friston are not alone in drawing this conclusion.
Poldrack (2010), another prominent investigator, concluded that “A review of
the neuroimaging literature suggests that selective association between mental
processes and brain structures is currently impossible to find” (p. 754).
In other words, the field of cognitive neuroscience is gradually changing from
one that sought neural codes for cognitive processes to one more concerned
with the properties of neural networks per se. It seems to be that neuroimaging
is diminishing its activities related to cognition in favor of research that is more
concerned with anatomy and physiology of networks. The reasons for this are
clear:

• The solution to the mind–brain problem is more likely to be solved at the

level of microneuronal networks than of pooled macroneural responses.
• Therefore, the technological methods and tools we currently have available
to study the problem (most notably the brain-imaging technology) are inad-
equate and inappropriate for solving the mind–brain problem.
• Critical information is lost when neuronal responses are pooled or averaged.
• Cognitive processes are notoriously underdefined.
• Cognitive processes are underdetermined with regard to underlying neural
mechanisms.
• Statistical methods cannot handle the “low power” experimental designs
resulting from inadequate subject sample size.
• The problem of relating cognitive processes and neuronal components or
networks is far more difficult than is generally appreciated.

What a terrible waste of resources resulted from the fact that early critics of
brain imaging as a means of studying the neural basis of cognition were not
attended to (Uttal, 2001; Van Orden and Papp, 1997).
 Introduction 33

1.6 Microneuronal Theories—A Brief Introduction

The main handicap of the macroneural approach to theory building is the irre-
trievable loss of critical information about the states and responses of neurons
and synapses because of response summation. Likewise, the microneuronal
approach is similarly obstructed but by what many consider to be the antithetical
problem. Rather than having lost the necessary data, the difficulty facing micro-
neuronal approaches to theory building is a glut of data. And, what a glut it is!
The sheer number of neurons that must be involved in the representation of
even simple cognitive processes is almost certainly in the billions. Furthermore,
the number of relevant neurons pales in light of the number of salient synapses
between neurons for which we must account. Even more detrimental to our
hopes of building a microneuronal theory is that the nature of the interactions
between neurons is not constrained to a single forcing function. In its stead,
each neuron interacts with its neighbors by means of an idiosyncratic set of
connections encoding a myriad of forcing functions. No universal rule of inter-
action comparable to the force of gravity or any of the other attractive and
repulsive forces of physical nature operates in the same way between all of the
components of the neuronal network to simplify our computations. This means
that progress in developing microneuronal theories must be based on a brute-
force analysis in which the individual responses of each and every neuron must
be considered.
As we see later, this situation is only partially ameliorated by the advent of
supercomputers, in which the number of logical units is becoming comparable
to the number of brain neurons. If the microneuronal approach is correct, no
matter how large the computer, it will still be necessary to take into account
information about the states of the individual neurons. This means dealing with
the neurons individually and combinatorial issues immediately become salient.
It would still be necessary (if we succumbed to the poor judgment of actually
trying to execute the kind of Gedanken experiment proposed by Martin, Grim-
wood, and Morris, 2000) to not only measure the state of each neuron but also
to load the system with individual initial conditions for each neuron. Clearly,
it is not just a matter of measuring N (where N, the number of brain neurons,
is a large number itself) but also of evaluating the combinations of N that may
result in data explosions such as N! (where N! is N factorial, a much larger
number for even small Ns).
I hope it is not too pessimistic to suggest at this early point in this book that
there is little hope even in the distant future for a full-scale theory at the micro-
neuronal level. One must also keep in mind that Moore’s (1965) doubling law
of computer packing density (a surrogate measure of raw computer power) grows
far slower than any of the functions that designate the possible combinatorial
growth of complexity. Therefore, increases in computer power cannot be expected
to overcome the innate intractability of the microneuronal approach to the
mind–brain problem.
34 Introduction

1.7 Interim Summary

The generic question asked in this chapter is—is it possible to produce an over-
arching theory of the process by means of which mental processes are encoded
or represented by neural mechanisms? That they are, albeit in some so-far inex-
plicable manner, is the foundation postulate of modern cognitive neuroscience.
Without such a foundation postulate, none of the extensive empirical research
now being carried on would have any psychological meaning. Nevertheless,
although we have had millennia of concern and a recent outburst of technology
with this question, there are some indications that a positive answer to this generic
question may be less forthcoming than suggested by a superficial review of the
extraordinary new technological developments that make it possible for us to
even consider asking this question. The literature of cognitive neuroscience grows
each year as greater and greater investments in time and energy are made by the
scientific community, particularly with regard to the macroneural approach. There
seems to be an increasing feeling supported by new analytical and empirical
findings that the macroneural approach, dependent as it is on the brain-imaging
technology, is not going to provide the foundation for an overarching neurore-
ductionist theory of cognitive processes. Not only are the data unreliable, but
also the analysis techniques are arbitrary and inconsistent.
The following list of general conclusions may help to set the stage for under-
standing the arguments leading to this disappointing negative conclusion:

1. There is increasing reason to support the argument that the mind–brain prob-
lem may be intractable at both the macroneural and microneuronal levels.
Many of these reasons are based on conceptual or logical foundations, but
many others are empirical—simply considering the number of neuronal ele-
ments involved in any mental process raises practical issues of computability
that have no current solutions.
2. Theories come in several different kinds. Not all are neuroreductionist; some
are purely descriptive. However, there is considerable discussion about what
constitutes a neuroreductive theory or whether such an ideal understanding
at any level is actually possible. Scientists still argue over the descriptive versus
the reductive nature of theories.
3. Macroneural theories suffer from the loss of what may be critical data because
of the pooling or summation of microneuronal signals.
4. Microneuronal theories, conversely, suffer from the very large amount of data
that must be processed to answer the mind–brain question. Empirical testing
of the theories may be impossible for the most practical reasons.
5. Obviously, science is driven and constrained by the available measuring
instruments. It is not so obvious, however, that it is not only our empirical
results but also our theories that are so governed. Technology can both enable
and mislead explanations.
 Introduction 35

6. A main goal of macroneural theories is the localization of cognitive processes

in specific regions of the brain. This may be a dead end.
7. A main goal of microneuronal theories is the study of neuronal networks of
realistic cognitive size. This, too, may be a dead end.
8. Other factors that might mislead theoretical endeavors include:
• Incorrect attribution to the wrong level of analysis
• Poor definition of psychological constructs
• Inaccessibility of cognitive processes
• False analogies between transmission and integration processes

These summary points are presented here to clarify my prejudices as we set

off on a more detailed discussion of microneuronal theory in cognitive psychol-
ogy. Chapter 2, in accord with the general microneuronal goal of this book,
deals with theories involving single neurons. Chapter 3 is an introduction to
networks, the ways in which we study them, and the problems generated by
their intrinsic complexity. Chapter 4 deals with relatively simple theories in
which a few neurons are assumed to be interconnected. This intermediate level
of neuronal networks is then followed by what is probably the frontier of modern
neuroreductionist theory—the application of today’s supercomputers and the
problems that arise from their application to the potential solution of the mind–brain
problem—in Chapter 5. Chapter 6 summarizes the argument put forward in
this book.

Notes
1. In several places in this chapter I have abstracted material from my previous writing
to review and clarify a necessary point.
2. Some of the material in this section has been adapted and updated from an article by
Uttal (2015).
3. A readable technical discussion of the physics of fMRI recording can be found in Uttal
(2001).
4. This trend may not be continuing; there is some suggestion that the number of papers
that more or less naïvely attempt to localize particular cognitive functions in particular
regions of the brain is slowing. My general impression is that some of our most pres-
tigious journals (e.g., Science) are not publishing articles of the kind that contributed
to this statistic. Indeed, even some of the specialty journals (e.g., NeuroImage) long
associated with correlations of cognitive states and brain images seem to be more heavily
populated with technological issues rather than simplistic assignments of cognitive
processes to localized brain regions than previously has been the case.
This page intentionally left blank
2
SINGLE NEURON PRACTICE
AND THEORY

2.1 Single Neurons

Before the discovery that the nervous system was made up of cells that were
demarcated from one another and from their extracellular environment by a
gossamer membrane composed of lipid molecules, the nervous system was thought
to be a single syncytium with protoplasmic continuity among the neurons (an
idea attributed to Joseph von Gerlach, 1820–1896). A new idea—that neurons
were protoplasmically separated from one another by membranes—was first
promulgated in 1891 by Wilhelm Waldmeyer (1836–1921). The new concept
grew directly from the development of a new staining technique invented by
Camillo Golgi (1843–1926). The key to the Golgi stain’s importance was that
only a small proportion of the neuron’s bodies, axons, and dendrites were stained
by this technique. Furthermore, staining seemed to stop abruptly at what could
only be interpreted as membrane boundaries. The implications of the resulting
microscopic images were championed by Santiago Ramón y Cajal (1852–1934)
as reflecting a clear-cut separation of the intracellular protoplasmic material from
the extracellular fluids. Although Golgi never accepted this new idea (which
came to be called the Neuron Doctrine), both he and Cajal received the Nobel
Prize for their respective contributions in 1906.
The initial idea behind the Neuron Doctrine was a purely anatomical one—
namely that the cells of the brain are not protoplasmically interconnected. Instead,
as a result of a diversity of recording and microscopic staining techniques, it
became clear that what had been thought of as a reticulum of protoplasmically
interconnected neurons was actually a network of neurons that were only func-
tionally interconnected. Neurons, it became clear, maintained their individual
identities and the boundaries between adjacent neurons even as they participated
in the ensemble activities of the whole brain.
38 Single Neuron Practice and Theory

The enunciation of the Neuron Doctrine changed neurophysiological

thinking in profound ways. Not only did this lead to new experimental
approaches but also new modes of theorizing about mind–brain relations
eventually emerged. Impelled by the Neuron Doctrine, a fundamental question
immediately arose—how did these separate cells communicate with one another?
The answer was provided by Charles S. Sherrington (1852–1952) when he
named the main mechanism of interneuron communication—the synapse. Otto
Lowei (1873–1961) uncovered the chemical nature of synaptic connectivity,
and Henry H. Dale (1875–1968) specifically discovered the first chemical
transmitter substance (acetylcholine). At first the synapse could only be defined
functionally; it was so small that it could not be imaged until the development
of the electron microscope in the 1930s by Ernst Ruska, a contribution for
which he received the Nobel Prize in 1986. A full discussion of the history
and nature of synaptic activity can be found in many books (e.g., Bennett,
2001; Pickel and Segal, 2013).
Many of the properties of the neuron are determined by the electrochemistry
of the cell or bilipid plasma membrane. This membrane is the boundary between
the interior of the neuron and the outside environment. It is made up mainly
of two layers of a particular lipid molecule, the heads of which are attracted to
water (hydrophilic) and the tails of which are repelled by water (hydrophobic).
As a result of these attractive and repellant forces, membranes can spontaneously
organize themselves in an environment that contains the necessary lipid mole-
cules.1 It also is a powerful force to maintaining the integrity of the membrane
during mechanical insults—a factor that will become of central importance in
our later discussion.
The important influence these new discoveries had was to emphasize the role
of the individual neuron in both experimental protocols and theoretical thinking.
During the first half of the 20th century, a considerable amount of attention
was paid to determining the basic properties of neuroelectrical activity in terms
of the selective permeability of the cell membrane to ions. During this time,
neurophysiologists developed techniques for recording from single neurons. This
work culminated in the Hodgkin and Huxley (1952) model of neuronal activity,
a feat for which they also won the Nobel Prize in 1963.
Even more important to the development of theories of cognitive processes
was the discovery that patterned stimuli, more so than their raw energy, deter-
mined the responses of individual neurons. The corollary idea that the patterned
activity of individual neurons could encode or represent properties of the stimulus,
at least in the peripheral nervous system, gained wide popularity in the mid-20th
century. It was, however, a huge conceptual step to the assertion that single
neuron activity could explain much more subtle and complex symbolic cognitive
processes; in other words, that complex cognitive processes and events, not just
the basic parameters of magnitude, quality, or timing, could be encoded by the
activity of individual neurons.
 Single Neuron Practice and Theory 39

The idea that a single neuron could also encode more complex concepts such
as a particular face or object persists to this time. However, single neuron theories
remain contentious in the context of today’s network theories—theories that
emphasize the alternative view, namely that the pattern of activity in broadly
distributed, multi-neuron networks is the key to understanding cognitive infor-
mation processing by the brain.
This chapter has two parts. First, it tells the story of those single neuron
research technologies and their strengths and limitations in our search for under-
standing of the mind–brain problem. We start with an elaboration of the methods
and techniques that are used to determine the functions of single neurons. Of
one thing we can be sure—the availability of certain technological instruments
(e.g., high-impedance microelectrodes) had a defining influence on the set of
theories we consider in this chapter. The second part is an evaluation of the
theories that are based on single neurons.

2.2 The Technological Bases of Single Neuron Theories

As just noted, neurons are now appreciated to be the basic cellular elements of
the nervous system. Although they come in many shapes and sizes, neurons, like
all other cells of the body, have evolved to carry out a specific function—they
are specialized for information transmission and integration. That is, their role
is to convey neuronal signals representing information from one part of the body
to another and then to integrate this information so that coherent and adaptive
responses can be made by the organism. The ultimate role of the nervous system
is therefore to provide adaptive responses to environmental stimuli—a process
that requires the gathering of sensory information, its integration into meaningful
cognitive interpretations, and the selection of appropriate responses. Although
neurons, like any other cell of the body, may perform other supportive, metabolic,
or genetic functions, their main role is information processing.
Information, in the sense that I use it here, can be encoded in terms of patterns
of organization of many neurons as well as the states of individual neurons. Con-
veying information means the transmission of signals from one part of the nervous
system to another; integration refers to the mixing and mutual interaction of multiple
signals into coherent patterns that can produce adaptive behavior in the form of
both overt motor and covert mental activity. The currently most popular basic
postulate of cognitive neuroscience is that “cognition” (or any other word you may
choose to designate mental activity) is nothing more or less than the pattern of
neuronal activity created by these transmission and integrative processes.
In the past half-century we have learned an enormous amount about how
individual neurons function. We now know about the physics and electrochem-
istry of how they are stimulated, the kinds of chemical and electrical responses
they produce, and how they communicate with one another synaptically. What
we do not know very much about is how psychological processes arise from
40 Single Neuron Practice and Theory

this biochemical and electrical activity. Indeed, hopefully without falling over a
philosophical cliff, we might ask if psychological processes are necessary or just
a meaningless epiphenomenal byproduct of neuronal activity. This may seem
like an extraordinary question; however, considering that a fully functional brain
could in principle function without any consciousness (as a “zombie brain” as
discussed by Chalmers, 1996), the possibility that we are actually automata enjoy-
ing the unanticipated consequence of personal sentience as our nervous systems
go about their business cannot be ignored. The basic problem is, as I have noted,
that we do not yet know what the relationship is between neural activity and
cognitive activity either in terms of single or multiple neurons.
Many neurons specialized for information transmission over long distances have
evolved elongated processes (axons). For example, a single axon runs from the foot
to the spinal cord before synapsing, thus providing a communication line for an
irritation in the toe to be interpreted by the brain as a pain. The axon is an elon-
gation of the neuron’s cell body extending from the cell body or soma, as it oth-
erwise known, to a distant neural locus where it may communicate via a multitude
of synapses to the next neuron in an extensively interconnected network. Other
types of neurons specialized for local information processing may have very short
axons with many connections spread over its surface. In general, axons terminate
in close proximity to the receiving region or dendrite of the next neuron.
Synapses are now known to be of two types, either chemical or electrical.
Chemical synapses conduct by emitting specialized molecules known as “trans-
mitter substances” from the “presynaptic” axonal region of the neuron. These
molecules are received by specialized molecular receptor sites on the “postsyn-
aptic” dendrites. Electrical synapses are rarer, more tightly packed than chemical
synapses, and cannot amplify neuronal signals as do chemical synapses.
Because synapses are so small, their history is much more recent than are
those of other neuronal components. Observation of synapses requires the extreme
magnification provided by electron microscopes—a gain that must be as high
as 2 million compared with the maximum magnification of an optical microscope,
which is limited by the diffraction of light to about 2,000. In the past couple
of decades, ingenious fluorescence techniques (e.g., Hell and Wichmann, 1994)
suggest that new optical technologies may permit “superresolution” beyond what
is imposed by diffraction limits on optical microscopes.
To accelerate the conduction time, elongated axons are usually sheathed in a
multi-layered myelin coating composed of a series of specialized and overlapping
Schwann cells. This myelin sheath is discontinuous with narrow constrictions referred
to as “Nodes of Ranvier” spaced periodically along the length of the axon. The
electrical signals that produce the propagated neural response along an axon are
able to jump from node to node, producing a much higher conduction speed than
would be possible with an unmyelinated fiber. Cells primarily designed for localized
interaction have much shorter axons that are typically not myelinated—indeed,
some presynaptic axons are so short that they are hard to distinguish from the
 Single Neuron Practice and Theory 41

dendrites of the receiving or postsynaptic neuron. The cell body, or soma, is a

complex structure with many components and metabolically active microstructures,
including the intracellular mitochondria, the synaptic structures, and the capillaries
that convey blood to and from the neuron. It is also important to note that most
neurons are surrounded by supportive glial cells that have long been thought by
some (e.g., Galambos, 2007) to have some role in information processing beyond
the purely supportive role usually associated with them.
Penetrating the membrane are also some relatively large molecules mainly
concerned with the transport of metabolically significant chemicals from the
outside to the inside of the membrane and vice versa. This flow of materials
includes both simple ions, whose ebb and flow are dependent on the transmem-
brane potential, and synaptic transmitter substance of complex molecular con-
figurations. It is now thought that large penetrating molecules of this kind are
both the presynaptic source of both transmitter molecules and the postsynaptic
gates that permit the transmitter substance to enter a neuron. It is the intricate
pattern of interactions among neurons based on these robust molecular engines
that account for the neuronal activity that in some mysterious way produces all
of our cognitive processes.
The next question with which we must deal is how we go about observing
the activity of neurons, synapses, and membranes at the microneuronal level at
which they operate. Although much has been accomplished in understanding
their biochemistry and anatomy, it is at the interface between neurons and elec-
tronic equipment from which most of our information about their interneuronal
information processing functions has come. The method of stimulating and
recording from single neurons is a highly developed technology that once again
illustrates the close connection between our technologies and our theories.
Electrodes are the point of contact between our measuring instruments and
the activities of neurons and the brain. It is here the biology and electronics
interface and the micro- and macro-events of the nervous system become the stuff
of empirical science. Over the years, there has been a steady progression of more
and more sensitive instruments that permit recording from and sending stimuli to
the nervous system. Electrode design has become a highly developed technology
unto itself with different specialized electrodes serving different needs. Somewhat
arbitrarily I have distinguished between stimulating and recording electrodes
although in many cases the same kind of electrode may serve either function.

Recording Electrodes
The basic role of recording electrodes is to detect and measure the electrical
activities produced by the various components of the nervous system.2 Electrodes
come in a wide variety depending on the specific application. Different electrodes
are necessary for measuring the compound (i.e., pooled) neural responses than
for measuring the isolated responses of single neurons; electrodes that are optimal
42 Single Neuron Practice and Theory

for intraneuronal responses may be quite different than those that are useful for
extraneural ones. Other electrodes are specialized for recording the macroneural
responses such as those picked up by the electroencephalogram (Berger, 1929)
or whole nerve preparations (Dawson and Scott, 1949); these signals may require
only simple electrical contacts—a stainless steel button taped to the skull or
attached to the arm does a satisfactory job of communicating these macroneural
signals to our measuring equipment. In other contexts, there has been a persistent
effort over the years to make direct contact with brain tissue by means of surgi-
cal techniques that use implanted electrodes. (A comprehensive discussion of
electrode technology can be found in an informative article by Cogan, 2008.)
Our main interest in this section is concerned with the electrode types that
can be used to measure single neuron responses. (In Chapter 3, I will deal with
multiple arrays of electrodes that can be used to record from many neurons
simultaneously.)
Neurophysiological recording of the electrical activity of neurons is accom-
plished by measuring the minute voltages produced by shifting ionic distributions
across the cell membrane. To do this, some kind of an electrically sensitive
contact—an electrode—must be placed in or near the neural tissue. Subsequent
amplification of the obtained biological signals to a level that is sufficient to
drive recording equipment is accomplished by appropriate electronic devices.
The amplified signals can then be connected to a display such as a cathode ray
oscilloscope or entered into a computer for subsequent analysis.
The critical and most idiosyncratic component in such a system is the initial
contact point—the actual electrode interface that senses the tiny neuronal electro-
potentials and conducts the information they carry to signal processing equip-
ment. Various forms of electrodes capable of observing single neuron responses
have been used in the history of neuro-electrophysiology depending on the
particular neuroelectrical signal of interest. As we shall shortly see, electrode
technology has driven the course of modern neurophysiology and with it the
theories that have been engendered.
The first attempts to record electrical signals from single neurons were carried
out on surgically excised whole nerves consisting of a bundle of axonal fibers.
The electrodes used to record these signals were typically salt-saturated fiber wicks.
The experimental design was a direct one—whole nerves were surgically dissected
free of their surrounding tissue and laid across the wick; the detected voltage
responses were connected to an amplifier and recorded on whatever apparatus was
available for displaying and recording. Such techniques, however, produced signals
that were the accumulation of the responses of many neurons, a mélange that may
have hidden the activity of the individual neuron, as shown in Figure 3.2.
A step toward single fiber recordings occurred when investigators were able
to dissect single axonal fibers free from a compound nerve. Wick electrodes
were also at first used to detect the electrical activity generated by these single
dissected axons. However, the extracellular responses so recorded were very small
 Single Neuron Practice and Theory 43

(a few microvolts) compared with the actual transmembrane resting voltage dif-
ference that was ultimately measured across the membrane; these transmembrane
voltages could be as large as 70 or 80 millivolts or even larger in some cases.
The next important step in measuring neuronal activity occurred when Young
(1936) discovered that the squid Loligo vulgaris had evolved a neuron with a very
large axon—one that could be as large as one millimeter in diameter. This was
large enough to permit a reasonably sized glass tube electrode filled with a
conducting electrolyte solution to be inserted into the cut end of this giant axon
and, thus, into its internal environment. The glass tube, therefore, served as an
intracellular electrode. For the first time, using the accidental availability of the
anatomical anomaly in this model preparation, it became possible to observe the
transmembrane voltage in its full splendor, not just the much diminished version
obtained with extracellular electrodes. When a reference electrode (all voltages
are measured as potential differences between two points) was placed in the
extracellular fluid, the full transmembrane voltage could be measured as the
voltage between the inside and the outside of the neuron.
The enabling “trick” in this case was getting one relatively large electrode
into the interior of an unusually large neuron. Thus, the accident of an evolu-
tionary freak—an unusually large axon—made it possible to develop the ionic
theory of neuronal action. Using the giant neuron of the invertebrate squid L.
vulgaris as a model of vertebrate neuronal ionic mechanisms, Hodgkin and his
colleagues (Hodgkin and Katz, 1949; Hodgkin and Huxley, 1952; Hodgkin,
Huxley, and Katz, 1952) were able to work out many of the details of the ionic
flow across the membrane that accounted for its ability to both create local
potentials and to rapidly propagate a signal from one end of the axon to the
other—research that earned them the Nobel Prize in 1963. It also permitted
them to develop a comprehensive theory relating membrane potentials and ionic
concentrations that persists virtually unchanged up to the present time.

Stimulating Electrodes
Electrodes have to be more than just one-way devices. In addition to their
important uses as sensitive transducers of neural signals, they can, as noted previ-
ously, also be used to stimulate or activate neural tissue. For macroneural stimula-
tion, such as is necessary to evaluate percutaneously the properties of peripheral
nerves (e.g., Uttal, 1959), a simple, hemispherical, stainless steel button, buttressed
by some electrical gel, is all that is required.
Similarly, when stimulation of the brain with large implanted electrodes was
required, the material from which the electrodes were made was relatively
unimportant—almost any nontoxic and corrosion-resistant piece of wire would
be satisfactory. In some situations (e.g., Uttal and Krissoff, 1966), in which a
diffuse spatial stimulus was required, it was sufficient to simply insert two fingers
into separate test tubes filled with saline solution.
44 Single Neuron Practice and Theory

The Modern Microelectrode

The anatomical freak typified by the size of the squid’s giant axon, however, is not
available in vertebrate neuron preparations. Vertebrates, despite their larger size and
greater number, generally have much smaller neurons than do invertebrates. To study
their individual nature required an alternative to pushing a rather large glass pipette
into the cut end of a fortuitously supersized axon. It was at this point that the bilipid
molecular layer of the membrane played a critical role in continuing progress toward
a full understanding of how neurons operate, especially how the action potential—
the propagating “spike” of electrical activity accompanying the ionic flows—
communicated information from one part of the nervous system to another.
The attractive hydrophilic forces between the lipid molecules that made up
the cell membrane were sufficiently strong that it turned out to be possible to
punch an appropriately small (of the order of a micrometer) electrode right
through the cell membrane without immediately destroying it. The intermolecular
forces of the lipid molecules were strong enough to provide a robust seal around
the electrodes—a seal that could last for many hours.
Among the earliest to develop such a membrane-penetrating microelectrode
were Ling and Gerard (1949) in their pioneering study of the resting potential
of frog muscle cells. The technique had been under development for a number
of years in Gerard’s laboratory by Graham (Graham and Gerard, 1946), but was
finally brought to a mature reliable stage by Ling.3
The technique that Ling and Gerard developed to manufacture a microelec-
trode required that a glass pipette be heated at its middle until the glass began
to melt and then simply pulled from both ends. This procedure reduces both
the external diameter of the pipette and its internal diameter of the interior
down to approximately 1 μm in diameter—a size comparable to the diameter
of the neurons under study. Indeed it provided two electrodes as the “pulled”
glass pipette was carefully broken apart at the hot spot. The ability to produce
such a miniature electrode—a microelectrode—lies in the fact that the geometry
of a hollow glass pipette (like those of the millefloria of Venetian glass) is pre-
served even as it is vastly demagnified in diameter by the heating and pulling.
The important part of this procedure is to preserve the tiny (as small as a micron)
hole at the end of the glass tube so that the conducting fluid it contains can
make electrical contact with the intracellular fluid when it is punched through
the cell membrane of a neuron. The final role of glass in this manipulation is
simply to act as an insulator between the intracellular and extracellular fluid,
thus avoiding a short circuit across the membrane.
The miniaturized glass microelectrode is then filled with an electrolyte such
as Ringer’s solution—a mixture of any of several salts, potassium acetate, or potas-
sium chloride. Many other solutions can be used as the electrical interface as
long as they are conductive and do not disrupt the metabolism of the neuron
being studied by the unintentional injection of chemicals that may be toxic. The
 Single Neuron Practice and Theory 45

other, un-demagnified end of the glass tube (from which the electrode was pulled)
remained large enough for a conducting metal wire to be inserted to serve as a
reference electrode. The functional role of this intracellular microelectrode is to
convey the relatively large transmembrane potentials to a preamplifier; from there,
signals could be conveyed to display and computer analysis devices.
An early application of the new glass pipette–microelectrode technology to
neurons from the frog’s sciatic nerve was reported shortly after Ling and Gerard’s
publication. The first application of a microelectrode inserted into the human
brain was reported by Ward and Thomas (1955).
An alternative method of making intracellular micron-size microelectrodes is
to electrolytically polish a metal (usually tungsten, platinum, or iridium since steel
points are too brittle), wire down to a 1-μm point, and then coat the wire with
some kind of insulating varnish (Grundfest, Sengstaken, Oettinger, and Gurry, 1950;
Hubel, 1957). The tip is left uncoated so that electrical contact can be made with
the neuron, usually extracellularly. Both forms of electrode—glass micropipettes
and sharpened, then insulated, metal points—thus consist of a tiny conducting
core surrounded by insulation (glass or varnish) that can be used as a membrane-
penetrating, active electrode. The second point of electrical contact, the reference
electrode, can be placed virtually anywhere in the extracellular fluid environment
of the preparation and is decidedly low-tech—a clip or plug can be attached to
any available tissue in the specimen whose nervous system is being studied.
The glass microelectrode is particularly useful for intracellular recording. It is
capable of sensing a maximum amplitude transmembrane voltage from a single
neuron. Insulated electrodes made from many different kinds of metals seem to be
better adapted to extracellular recording, typically resulting in a mixture of neuronal
signals from several nearby neurons. How many neurons will be sensed is determined
by fortuitous accidents of the arrangement and propinquity of the neurons and the
electrode and is typically not under the control of the experimenter.
In sum, microelectrode design is still very much an intuitive art form, each
investigator often constructing their own design to meet the special needs of his
or her experiment. Even the original work reported by Ling and Gerard was
the result of a multi-year effort by several people before their design was stable
and reliable enough to be used in any quantitative manner. For reasons that are
still not entirely clear, minor details in the electrical design or in the material
from which a microelectrode is constructed may have significant effects on the
quality of the signals they detect.

Recording Devices
One of the earliest devices designed to store and display the electrical activity
of the nervous system was Willem Eintoven’s (1860–1927) invention—the
string galvanometer. Although first used to record the electrocardiogram,
eventually it was applied to the measurement of the conduction speeds of
46 Single Neuron Practice and Theory

nervous impulses in situations in which its predecessor—the capillary galva-

nometer (in which the height of a column of mercury was used as the indica-
tion of electrical activity magnitude)—was insufficiently sensitive. In either
case, displacement of the string or the meniscus of the mercury column by
neuronal response–generated electrostatic forces was recorded photographically.
These primitive displays of electrical response activity have been completely
replaced by the modern cathode ray oscilloscope, an evolutionary outcome of
the Crookes tube, first invented by William Crookes (1832–1919). In the
Crookes tube, a stream of electrons inside an evacuated glass tube is directed
onto a fluorescent surface. This directional control is affected by the applica-
tion of external magnetic fields. With such modern equipment, this basic idea
of changing the position of a spot on the face of an oscilloscope as a function
of the applied voltage makes it possible to plot the time course of the electrical
waveforms generated by neural activity.
It is something of a historical curiosity that the cathode ray oscilloscope
was not used in cellular neurophysiology until 1922 when it was first applied
by Joseph Erlanger (1875–1965) and Herbert Gasser (1888–1963) despite its
invention as early as 1897 by Karl Ferdinand Braun (1850–1918). This was
likely due to inadequate amplification of the tiny signals generated by neu-
rons, rather than any fundamental limitation on the part of the oscilloscope
itself.
An especially significant development for studying microneural events,
therefore, was the emergence of modern high-input impedance preamplifiers
to extend the range of the cathode ray oscilloscope to low voltage levels. The
small size of the electrical potentials recorded from the nervous system meant
that they had to be significantly amplified before they could be reliably mea-
sured. The invention of the triode vacuum tube amplifier by Lee De Forest
(1873–1961) was a milestone not only for amplifying music but also for
neurophysiology. The triode vacuum tube was constructed from an anode and
a cathode and a control electrode, all of which were encased in an evacuated
glass enclosure. The amplification of small signals was accomplished first by
the three-electrode vacuum tube and then by modern solid-state circuit designs.
By applying the preamplified, but still relatively small, signal generated by the
neuron to the control electrode situated between the cathode—a source of
electrons—and the anode—a sink for electrons—small signals could be ampli-
fied to a level that was adequate to drive the position of the electron beam.
The basic principle of this kind of amplification was accomplished because
the small signal placed on the control electrode could regulate a much larger
current flowing between the cathode and the anode. Modern amplifiers, of
course, are now constructed from solid-state components, but perform the
same amplification function albeit at ever-improved levels of amplification,
noise reduction, and frequency response.
 Single Neuron Practice and Theory 47

Amplifiers and Impedances

Unfortunately, in meeting the need for the ultraminiaturized size of an intracel-
lular microelectrode so that the neuron would remain reasonably intact when
impaled, another problem is created. Both the tips of glass and the metallic
microelectrodes are so small that they present very high electrical impedance
(resistive impendences of microelectrodes may be as high as 100 megohms); this
requires special electronic preamplifiers or head stages prior to the main amplifier
for signals to be measurable.
There are two problems with high-impedance electrodes. First, their resistive
impedance (Rm) is part of a voltage divider circuit with the input impedance of
the preamplifier Rp as shown in Figure 2.1. The voltage that is detected by the
preamplifier, therefore, is a function of the impedance of the electrode and that
of the amplifier in accord with Equation 2.1:

Vp = Vm × Rm/Rm + Rp Eqn. 2.1

where Vp is the voltage sensed by the preamplifier and Vm is the voltage produced
by the neuron. Thus, if Rm is much larger than Rp, then Vp (the portion of the
voltage) across the amplifier input will be very small—perhaps undetectably so

Input impedance
of preamplifier

Rp Cp
A

Rm
Input impedance
of microelectrode

Cm

B
Microelectrode

Neuron

FIGURE 2.1 The electrical impedances involved in a microelectrode circuit.

Reproduced from Uttal (1975).
48 Single Neuron Practice and Theory

by the main amplifier. The net effect of this will be to make the tiny voltage
from the neuron virtually undetectable by the amplifier unless the amplifier input
impedance is itself high. This is the reason that specialized high-input impedance
preamplifiers are so necessary in cellular neurophysiological research.
The second main problem with high-impedance microelectrodes is that they
operate in combination with the capacitance of the system to limit its temporal
response properties in accord with the RC rule:

TC = R × C

where TC is the time constant of the system, R is its resistance, and C is its
capacitance. The TC of such a circuit will determine the overall frequency
response of the recording system; the larger the TC, the less the high-frequency
components of the signal will pass through the system. The required high-
frequency response of a neuron-preamplifier combination must be measured at
least in kHz for neuronal responses to be detected and their waveform reproduced
although some modern experiments may require much higher limits.
Impedance mismatches can also introduce spurious noise of several different
kinds. The engineering of preamplifiers with sufficiently high-impedance prop-
erties to match microelectrodes is an art unto itself, but one that has matured
to such a high level of development that they are currently available off the shelf
from a number of different manufacturers.

Patch and Voltage Clamps

Other instrumentation tools that have become mainstays of single neuron physiologi-
cal reseach are the voltage clamp and the patch clamp. The problem arising is that
the several ion distributions that collectively sum to determine a neuron’s membrane
potential are each dependent on the membrane potential itself. Because this is a
dynamic processs with an intrinsic form of feedback, it is diffiult to determine such
factors as the specific relationship between the membrane potential and the flow of
the particular ions that collectively make up the actual neuronal response.
The voltage clamp overcomes this difficulty by generating electronic feedback
signals to stabilize the membrane potential at a predetermined constant level.
This essentially locks the membrane potential at any prespecified level. The
neurophysiological investigator is then able to derive the relationship between
stable membrane potentials and the ionic distributions. By varying the feedback
signal, the entire set of functions relating the several kinds of ion flow and
membrane potentials across the membrane can be determined.
The patch clamp is a derivative of the voltage clamp idea but is designed to
work only a small patch of the cell mebrane. This patch is often so small it is
possible to isolate a single one of the molecular channels through which ions
flow during neuronal membrane activity. Attachment to such a small region
 Single Neuron Practice and Theory 49

is accomplished by attaching the membrane to the tip of an electrode by means

of a vaccum. Both voltage and patch clamps have made it possible for modern
neuroscientists to understand how the flows of the various ions (mainly Na+,
K+, and Cl-) are dependent on the electrical properties of the membrane.
However, this detailed explanation of the ionic basis of neuronal activity does
not speak directly to the problem at hand—how cognition and mind are pro-
duced by this elaborated ionic dance. Putative explanations based on single
neurons are discussed in the next section, in which I review how some theories
of cognition attempt to link cognitive processes to individual neuronal responses.

2.3 The Single Neuron Theoretical Approach

to Solving the Mind–Brain Problem
In our search for the answer to the great mind–brain question, we seek to find
the neurophysiological properties and activities that “account for,” “encode,” or
“represent” our cognitive experiences. All of these words are ambiguous and ill-
defined, but the general idea is that there is in the brain some corresponding
neurophysiological activity that is not just correlated with mental activity but
“is” mental activity. In other words, mind and all of its accouterments are assumed
to be, in principle, identifiable in some way with the physiological and anatomical
properties and activities of neurons—either single neurons or networks of them
and possibly of other cell types in the brain. A better term to explain what is
meant by this relationship is the psychoneural equivalent—the neurophysiological
response that would be the necessary equivalent to cognitive experience. What
the psychoneural equivalent of a cognition is, of course, is only hinted at in cur-
rent cognitive neuroscience research but, under the influence of the methodologies
and technologies described in this book, speculative theories have proliferated.
As I begin this discussion, therefore, it is useful to be very specific about what
is meant by single neuron theory as well as its antithesis—neuronal network
theory. Single neuron theory (in its most extreme form) in cognitive neurosci-
ence refers to a particular set of possible explanations of the neural bases of
cognitive processes that are expressed in terms of the anatomy and physiology
of just one neuron (or, according to Barlow, 1972, just a few). According to the
most fundamental version of single neuron theory, the activation of a particular
neuron will elicit the experience of some object, person, or even a very complex
intangible concept such as “freedom.”
Once again, this illustrates the enormous influence that our technological tools
played in our theoretical development. It is all but certain that, without convincing
empirical support for the Neuron Doctrine and abundant single neuron findings,
the idea of individual neurons as the conveyers of our psychophysical experience
would have never occurred to anyone. Instead, more popular would have been the
holistic idea in which all or most of the brain is involved in mental activity.
(Holism, of course, is also driven by technology, namely the EEG and fMRI.)
50 Single Neuron Practice and Theory

There are several basic problems inherent in single neuron theoretical for-
mulation that suggest that it is not the ideal route for theorizing. First, we do
not know how many “concepts” there are in our mental life and so we cannot
tell if even the large number of neurons in the brain is adequate to represent
all of them, each with its own neuron whose activities serve as the psychoneural
equivalent of some mental experience. Counting the number of available neurons
and associating each with its cognitive equivalent, although possible in principle,
probably would not be a practical approach.
Second, unlike the first problem, which could in principle, if not in practice,
be validated by simply finding neurons whose response could be correlated with
a concept, there is no way that we could carry out the necessary experiments
to validate the association between specific cognitions and single neuronal activi-
ties. Although it is reasonable to assume that we could find a neuron that
responded to a particular stimulus, locating a particular neuron whose activation
provided empirical proof of a particular subjective awareness would be, to say
the least, a very challenging task.
Third, practical issues such as the manipulation of a particular neuron (isolat-
ing it, turning it off and on) would always conflate, confuse, and confound the
results of any experiment attempting to show single neuron representations of
cognitive processes.
An important related issue is—how do we account for the selective sensitivi-
ties and responsiveness of individual neurons to very complex “concepts”? In
general, the basic anatomical mechanism purported to explain these findings is
a progressive convergence from the properties of a stimulus onto fewer and fewer
neurons, with the result that more and more complex ideas or concepts can be
represented by fewer and fewer neurons. Ultimately, this convergence approach
would respond selectively to complex gnostic concepts and entities such as “yel-
low Volkswagens” or “grandmothers” or “altruism.” This extends the fundamental
idea of a hierarchy of increasingly complex responses in the visual system origi-
nally suggested by Hubel and Wiesel’s (1968) progression of simple, complex,
and hypercomplex cells. The concept of a convergent hierarchy suggested that
the lower, simpler, spatial-temporal properties defined at the stimulus level (such
as line orientation) might progressively converge with other properties onto
higher-level concepts, eventually as specific as a particular human face, at the
same time that neurons are becoming more and more selective and particular
to fewer and fewer tangible ideas.
The idea of a progressive convergence of constituent codes onto neurons
suggested, to hammer in this point, that the meaning represented by higher-level
neurons becomes more and more specific until complex cognitions such as
particular objects, people, or concepts are represented by the activity of a single
neuron. The important mechanism underlying further development of this idea
is, thus, a progressive convergence of the tuning of a neuron from the spatial-
temporal domain of simple stimuli to the gnostic, conceptual, or symbolic domain.
 Single Neuron Practice and Theory 51

At present, however, no one has been able to demonstrate this progressive neu-
rophysiological convergence once one proceeds beyond the peripheral sensory
systems.
In sum, the core of single neuron theory is that once it has been established
that a neuron has a particular sensitivity (such as responding to an individual
face) subsequent activation would be the necessary “cause” or “equivalent” of
the mental experience of that face. The basis of a pure single neuron theory,
therefore, is that the activation of a single neuron is the psychoneural equivalent
of entities of unlimited complexity.
Without a doubt, a statement that asserts that single neuronal activation is
tantamount to experience is replete with uncertainties, ambiguities, and unresolved
subissues. What are the other billion neurons doing while the critical one is
determining what we momentarily perceive or think? Is there some kind of
priority that one neuron has over all others? How did the neuron acquire the
particular sensitivity? Is the individual neuron itself conscious or its response
sufficient to determine a unique conscious experience? Is the role of psychoneural
equivalent being conflated with its equally plausible role as a “trigger”? For
example, if activation of a single neuron is simply acting as a trigger for a more
complex response actually embodied by a network of multiple neurons, then it
might well appear to an experimenter to be causal, but any such attribution
would be fallacious. In short, the single neuron theory, although superficially
simple, is actually extremely subtle and complicated by questions rarely asked
and almost never answered.
The antithesis to the single neuron theory is the neuronal network approach.
Rather than ascribing cognition to the activity of a single neuron, this alternative
proposes that the psychoneural equivalent is embodied in the interactions of a
network of many neurons, all of which are interacting with each other in com-
plex and dynamic manners. Controversies in this case concern how many neurons
are involved in a network and whether we can quantify network activity in a
mathematically tractable form. The raw conceptual simplicity of the single neuron
approach (neuronal response X is the equivalent experience Y) is an advantage
not shared with neuronal network theory. Even at the earliest stages of proto-
theory development describing how brain activity might make the transition
from neural function to psychological experience, preliminary anatomical evi-
dence made it clear that the process would be an overwhelming one that was
not likely to be solved in the foreseeable future if phrased in terms of the network
interactions.
Both single neuron and network theories propose an answer to the question—
what activities and properties of the brain produce, cause, or become cognition?
Both, however, raise many issues of meaning and interpretation that have to be
considered in detail. In subsequent chapters, I consider neural network theories
in detail; here I now discuss the practical and conceptual implications of single
neuron theories. However, before going on, there is one important clarifying
52 Single Neuron Practice and Theory

comment to be made—both the single cell and the neuronal network theories
have a common feature. That is, both maintain information about the status and
activities of individual neurons. Thus, both fall into the microneuronal category.
Although this may not seem to be the case when one observes the massive
neuronal networks that characterize the brain, this is their explanatory level. The
differences with the macroneural approach are profound.

Single Neuron Theories

The earliest single neuron theories antedated the neurophysiological findings
that were eventually to provide the putative evidence for what remains a
popular theoretical approach. William James, in his classical 1890 psychology
text, spoke of the “consciousness” of each and every “brain-cell.” To Charles
Sherrington (1940) is attributed the first use of the term “pontifical neuron,”
a phrase suggesting that an individual neuron might by itself represent a com-
plete cognitive experience in some way, dominating the current cognitive
experience.
More complete and up-to-date versions of single neuron theories were pre-
sented by Barlow (1961) and Konorski (1967) in the 1960s. In his version,
Konorski distinguished between transit (sensory and motor codes) and gnostic
(mental or cognitive representations) as distinguishable categories. This is an
important distinction that is all too often lost in evaluating today’s theories, since
great success in understanding peripheral transmission coding has been achieved,
whereas almost no compelling progress has been obtained in determining the
neuronal equivalents of complex gnostic or cognitive experiences. Barlow’s
approach was somewhat more eclectic, evolving from a single neuron approach
to one involving “a small multiplicity” of neurons—cells that he suggested
behaved as “cardinals” rather than “popes” (Barlow, 1972).
The single neuron approach to theory building was further stimulated in the
1950s and 1960s by the extraordinary neurophysiological discoveries of such
pioneers as Hartline and Ratliff (1957), Hubel and Wiesel (1957), and Lettvin,
Maturana, McCulloch, and Pitts (1959), among many others. All of this research
made the same point: response patterns could be observed in the responses of
single neurons that seemed to parallel much more complex behavioral responses.
Because many perceptual phenomena, such as visual illusions, had long been
described phenomenologically but without any neuroreductive theories, analogies
drawn between the associated peripheral transmission codes and the perceptual
responses had a certain face validity, as well as a seductive attractiveness. Based
on analogies between neuronal and psychophysical responses (and the success in
defining the peripheral transmission codes), available single neuron neurophysi-
ological findings quickly filled the vacuum as putative explanations of higher-level
cognitive processes. In the decades following the 1950s, based on these pioneering
studies of single neuron correlates of sensory information, the relatively simple
 Single Neuron Practice and Theory 53

idea that a single neuron could be responsible for representing complex perceptual
experiences became extremely popular, if not dominant, in mid-20th century
thinking.
Evidence provided support for the idea that cortical-level neurons become
selectively responsive to particular human beings (e.g., the well-known, but
hypothetical, “grandmother” neuron originally proposed by Lettvin, as cited
by Gross, 2002). Although some controversy exists about the robustness of
these findings (see the work of Tanaka, 1993, who showed specialization to
face-like geometrical forms rather than particular faces), there does appear to
be a relatively strong body of empirical evidence that suggests that these
observations should be taken very seriously—specifically, the not-to-be-
overlooked idea that there are neurons in the brain that respond only when
very specific and very complex stimulus conditions are met. These highly
specialized neurons, as I suggested earlier, are the end product of a process of
progressive convergence that starts with sensitivity to the physical dimensions
of a stimulus (e.g., its temporal-spatial properties) and progresses to neurons
whose responses are related to the highly particularized meaning of a stimulus
(e.g., a specific object, person, or idea).
A summary of a broad swath of current single neuron theories has been
provided in an important review of the field by Spillmann and Werner (1990).
In it, a distinguished group of investigators described their efforts to link the
neuronal properties of the sensory pathways with perceptual experience—that
is, to assume that these neurophysiological responses are the psychoneural equiva-
lents of cognitive experiences. However, even then, there were voices calling out
for a more thoughtful interpretation of the relation between these peripheral
neuronal codes and perceptual experience. For example, Westheimer (1990)
argued that:

Once the first discoveries were made . . . single cell researchers in the
mammalian visual system developed categorizations which do little for
psychophysicists.
(p. 7)

In retrospect, most if not all of the theories presented in Spillmann and

Werner’s book actually turned out to be associations of peripheral sensory
transmission codes that conveyed information but did not themselves encode
experience. The distinction being made, once again, is between transmission
and experiential representation codes. To put the distinction simply, we should
ask basic questions such as could the eye “perceive without a brain”? or could
the occipital region of the brain “perceive without an eye”? or could any other
isolated region of the brain “perceive” if activated in some indirect fashion?
Unfortunately, empirical answers to questions of this kind are currently impos-
sible to obtain.
54 Single Neuron Practice and Theory

The problem of how far up the sensory pathway one must go before we
find neural mechanisms that are the equivalent of cognitive experience (rather
than just communication codes) is rarely raised by cognitive neuroscientists.
An exception is the work of Crick and Koch (1995), who asked the question,
“Are we aware of neural activity in the visual cortex?” (p. 121). They sug-
gested that there is no awareness of neuronal activity in visual area V1 but
rather, that V1 is actually better considered as a part of the peripheral trans-
mission system. If their suggestion is correct, then the many conclusions of
cognitive encoding by single neurons based on correlations between psycho-
physical and neurophysiological responses in V1 may have been seriously
misinterpreted. The correlated neuronal responses would actually be reflecting
transmission codes, even as high as V1. Thus, the idea that neuronal responses
in V1 were the psychoneural equivalents of our perceptions would be based
on a false analogy—isomorphism of the dimensions of the stimulus and the
psychological responses. This issue is especially important in the present context
because any support of the cognitive efficacy of a single neuron’s responses in
V1 would become meaningless. Indeed, since the preponderance of the empiri-
cal work supporting single neuron theories has been carried out in the peripheral
systems, most of it does not actually deal with the neuronal coding of aware-
ness or experience per se.
Psychoneural equivalent codes at the level of cognitive activity are the answers
to a different question than the one concerning transmission codes. The equiva-
lence question asks—how do the neurophysiological responses become complex
perceptual phenomena? If Crick and Koch are correct, although there may be
correlations between neuronal activity and cognitive experiences, the relationship
is not a causal one. That is, the neuronal responses are not the psychoneural
equivalents of the experience but are conflations—both being driven by the
common stimulus, but without any direct relationship between the neuronal and
the psychophysical.
Without question, great progress has been made in the sensory domain and,
for all practical purposes, most codes for the key sensory dimensions (i.e., quality,
quantity, spatial extent, and temporal pattern) of most sensory stimuli in a variety
of animals have been broken (Uttal, 1973). If explicating the transmission codes
can be considered to be equivalent to theory development, then this is the most
compelling argument that successful neuroreductionist single neuron theories
have been forthcoming. However, this is not the same thing as demonstrating a
solid link between individual neurons and higher levels of experience. Far less
has been accomplished in understanding the relationships between complex
cognitive processes and their underlying neurophysiological equivalents. Further-
more, if Crick and Koch are correct and transmission mechanisms (as opposed
to cognitive ones) extend much farther up the ascending pathways into the visual
areas of the brain than has been appreciated, then the gap between neuronal and
cognitive responses remains as broad nowadays as it has been for decades.
 Single Neuron Practice and Theory 55

In the following discussion, I briefly review some of the theoretical associations

that have been made between single neurons and cognitive processes. I hope that
my readers will keep in mind the admonition that there are no direct and robust
links between neuronal responses and cognitive processes; rather, most of these
“theories” are little more than associations by analogy or metaphors.

1. Intraneuronal lateral inhibitions observed in the Limulus visual system

(Hartline and Graham, 1932) and in the frog by Barlow (1953) have been associ-
ated with visual phenomena such as the Mach Band (i.e., edge enhancement),
the dark spots at the intersection of the Hermann Grid illusion, and a spatially
broader perception of simultaneous contrast.
2. Lettvin et al. (1959) and Hubel and Wiesel (1965) observed that single
neurons were selectively sensitive to natural stimuli such as moving edge detec-
tors (flies?) and stimulus diminishment (looming predators?). The neurophysi-
ological sensitivities were then associated with practical ecological needs of the
entire organism such as “bug detection.”
3. Enroth-Cugell and Robson (1966) showed that individual neurons displayed
a differential sensitivity to spatial frequency—in particular the empirical result
called the perceptual “tuning” curve. This was extrapolated to the idea that the
Fourier frequency was associated with psychophysical detectability of spatial
frequency patterns.
4. One of the most interesting single neuron theories of the 1970s suggested
a near identity of neuronal and psychophysical responses. Jung and Spillmann
(1970) introduced the concept of a Perceptive Field (PF), the psychophysical
interaction between a spot stimulus and its surround. This perceptual phenom-
enon was deemed to be analogous to the Receptive Field (RF) of a single neuron
in a neurophysiological experiment. Jung and Spillmann suggested that compa-
rable perceptual results to the center-surround organization of a microscopic
visual neuron could be obtained at the molar level. Psychophysical responses
were measured to map out the surround of an isolated point stimulus. They
argued that the dimensions of the PF corresponded closely enough to the dimen-
sions of the RF of a single neuron to be the psychoneural equivalent of the PF.
There are, however, many differences between a single neuron’s response and
an analogous psychophysical phenomenon. One obvious functional difference is
the neurophysiologically measured output of the results from an individual neuron,
whereas the psychophysical experience must be the accumulated and overlapping
result of many neurons. A priori, therefore, there seems to be no obvious reason
why the two phenomena—the RF and the PF—should be of the same scale.
An obvious test of the PF hypothesis, therefore, is to ask—are the RF and
the PF at least the same size or, to the contrary, are any similarities between
them simply accidental responses occurring at different scales? Surprisingly,
according to Spillmann (2014) the psychophysical PF and the neuronal RF are
almost identical in size, a result obtained by comparing neurophysiological data
56 Single Neuron Practice and Theory

reported by DeMonasterio and Gouras (1975) and Spillmann’s psychophysical

data.
Neri and Levi (2006) provided further support for the similarity of RF and
PF sizes by reviewing seven pairs of experiments—one of each pair being an RF
neurophysiological experiment and the other being a psychophysical PF experi-
ment. They found several similarities between the two classes of experiments
beyond simple size. Although this finding seems to be counterintuitive, the
important point in the present context is that the RF of a single neuron is
presented by theorists of this persuasion as a direct cause of the psychophysical
phenomenon they call PF. However, even Neri and Levi point out that:

The similarity between receptive and perceptive fields highlighted here

should not be taken to imply that behavioral performance in these experi-
ments was completely determined by the response properties of individual
neurons in early cortex.
(p. 3472)

5. Single neuron theories of sensory processing have also been proposed in

the somatosensory system. The typical experimental paradigm in this case involves
stimulating the skin mechanically, probing a single neuron in a peripheral nerve
(e.g., the median nerve) with a microelectrode, and simultaneously asking subjects
to report when they sensed the stimulus (see Valbo and Hagbarth, 1968, for an
early example). Typical findings showed that the threshold of a subject was so
low that a single action potential in a single nerve fiber could be reliably detected
as a stimulus to the skin.
The inference drawn from this finding was that a single spike action potential
was capable of encoding a complex experience such as cutaneous touch or pain.
However, this inference is almost certainly misguided. The psychoneural equivalent
for the experience is not likely to be the single spike action potential recorded
from the peripheral nerve, but the greatly magnified burst of neural activity that
is triggered by that action potential at higher levels of the ascending (i.e., central)
nervous system. Thus, any attempt to use this as an argument for a single neuron
theory of perceptual experience is fragile. The fact that our sensory thresholds
are as low as they are is not evidence that single neurons are encoding experi-
ence. A more reasonable inference is that there is an enormous magnification
in the neural activity; an increase that ultimately rises to the level of perception
and ultimately to the relatively large amount of energy expressed in the verbal
motor response—“yes, I see [or feel or hear or taste or smell] it.” Indeed, mac-
roscopic evoked brain potentials, EEGs, and brain images of all kinds show that
vast numbers of neurons are generated in the regions of the brain associated
with barely threshold stimuli. That a minute stimulus activating a single neuron
might trigger a complex behavior is not evidence that the experience is encoded
by that neuron; in fact, it seems more supportive of the role of huge numbers
 Single Neuron Practice and Theory 57

of interacting neurons in a massive network being involved in virtually any

cognitive processes.
6. In the past few decades, a rush of reports has suggested that single neurons do
respond to a very specific and important kind of stimulus—individual faces. No
longer is the discussion dominated by the raw spatial or temporal properties of a
stimulus such as moving edges or corners. Instead, neurons are being reported in
various parts of the brain that respond selectively to faces. This may be in general
terms (responding to any face) or as specific as a response to a particular face. For
example, Perrett, Rolls, and Caan (1982) found temporal cortex neurons that
responded selectively to visually presented faces, as did Young and Yamane (1992)
and Leopold, Bondar, and Giese (2006). Fried, MacDonald, and Wilson (1997)
showed similar sensitivities in the hippocampus and amygdala. Kreiman, Koch, and
Fried (2000) also described temporal lobe neurons that responded selectively to
categories such as “faces, natural scenes and houses, famous people and animals.”
Similar “invariant” responses to faces (among other stimuli) have been described by
Quian Quiroga, Reddy, Kreiman, Koch, and Fried (2005) and Quian Quiroga (2012).
7. Although the idea of sensitivity to specific faces or objects dominates much
of the research into single neuron coding, other selectively sensitive neurons
encoding other cognitive processes have been recorded in various regions of the
brain. Jenison, Rangel, Oya, Kawasaki, and Howard (2011), for example, report
the observation of single amygdala neurons in awake humans that were linearly
related to the “value” assigned to the options in a decision-making task. This
was done by comparing psychophysical decision-making behavior with neuro-
physiological recordings from single neurons. Value, in this case, was used as a
metric of the desirability of the alternative choices in a purchase decision. The
authors claim that their findings were:

. . . consistent with the hypothesis that single neuron activity in the human
amygdala might play a role in either encoding or computing stimulus
values at the time of choice.
(p. 336)

8. Another popular single neuron theory of cognitive representation, but from

the motor rather than the sensory system, is based on the concept of the mirror
neuron. Acting on some early findings (di Pellogrino, Fadiga, Fogassi, Gallese, and
Rizzolatti, 1992; Gallese, Fadiga, Fogassi, and Rizzolatti, 1996) it was suggested that
single neurons in the F5 region of the prefrontal motor cortex respond not only
when a monkey performs a specific action (such as reaching for an object) but also
when it sees another monkey or a human perform that same action. The response
of the mirror neurons was said to “embody” the perception of the motor action.
Subsequent research defined some special conditions and limitations for this
selective response, but the important point in the present context is that a single
cortical neuron was deemed to represent the significance of a complex behavioral
58 Single Neuron Practice and Theory

response. Implicit in these observations is a theory of motor representation that

is comparable to the sensory and perceptual ones previously discussed—a single
neuron, when activated, could encode a complex motor experience. As with
other single cell approaches to explanation, the mind–brain problem would
essentially be solved should this interpretation be correct, since motor actions of
virtually any complexity would be produced by activation of the appropriate
neuron.
9. A number of ”single neuron theories” seem to conflate single neuron and
population activity. Sevush (2006), for example, has proposed a single neuron
theory that ascribes “consciousness” to individual neurons. These neurons do not
directly contribute to consciousness according to him but, acting in concert with
other neurons with which they are interconnected through their respective den-
dritic trees, their joint activation is tantamount to the concomitant production
of conscious psychological activity. These interconnections, according to Sevush,
solve the binding problem (how we aggregate the isolated responses of the nervous
system into a unified cognitive experience) by relegating it to the action of the
dendritic interconnections. A close reading of Sevush’s “single neuron theory,”
however, suggests that it is not very different from other network theories. He
notes that “the production of conscious behavior results from population activity
at the network level” (p. 705), which is also a principle of his theory.
Sevush also called attention to an even more microscopic problem—how does
a neuron learn to respond to the selective stimulus sensitivity? His answer is that
selective sensitivity is the outcome of “quantum entanglement” processes acting at
the level of a single elementary particle. This is an idea also supported by Argonov
(2012). Many of us would consider this idea unsupported by empirical evidence.
With this brief sampling and review of single neuron theories, we can now
turn to a critique of the single neuron approach to the neural coding of cogni-
tive processes.

2.4 A Critique of Single Neuron Theories

As the decades have passed by and many experiments have been carried out to
study the mind–brain relationship, there is no longer any doubt that there exist
neurons in the brain that are highly selective in the range and complexity of
stimuli to which they will respond. This selectivity seems to be specific enough
so that neurons can be found that respond to individual faces or objects or even
to complex ideas and concepts such as freedom or altruism. Evidence of an
extraordinarily precise selective sensitivity has been provided by a number of
investigators as evidenced in the preceding section. For a complex of reasons, a
theoretical approach has emerged that argues that the relationship between cog-
nition and neural activity is instantiated in the activity of neurons that are highly
selective to particular stimuli or whose responses are tantamount to experiencing
highly specific concepts—that is, that are the “psychoneural equivalents” of
 Single Neuron Practice and Theory 59

mental experiences. Some of the factors that have engendered this kind of single
neuron theoretical thinking include:

1. There is a well-developed technology with which the activity of individual

neurons has been studied. Much has been learned about the physiology and
anatomy of these neurons.
2. There is, as yet, little understanding of how cognition is encoded by the ner-
vous system. There is currently no overarching neuroreductionist theory of
cognition.
3. Thus, there is an explanatory vacuum attracting a variety of variously sup-
portable and plausible theoretical options. This vacuum has been partially
filled by cognitive neuroscience theories that dote upon extensive knowledge
of single neuron physiology.
4. Single neuron theories are conceptually simple and plausible from a cognitive
neuroscience point of view.
5. Because we want and need some explanation of how neuronal activity
becomes cognitive activity, there are strong extra-scientific reasons to use
whatever empirical evidence exists, no matter how sparse.
6. There is ample empirical evidence of a high degree of selectivity of the spe-
cific stimulus conditions that will activate a given neuron. This has to be
distinguished, however, from any Gedanken experiment in which it might be
shown that the activation of a particular neuron results in a particular cogni-
tive activity.
7. There has been substantial success in understanding the neuronal codes used
by the peripheral sensory and motor pathways to convey information to and
from the brain. Specifically, the codes used by individual neurons to transmit
the dimensions of stimuli are well known. Unfortunately, transmission codes
are not the psychoneural equivalents of cognition, and success in the former
field does not presage success in the latter.
8. Early work showed that there was a progressive convergence in the complex-
ity of information represented by single neurons as one ascended the sensory
transmission pathways. This convergence of geometrical properties was used
as a heuristic for convergence of abstract ideas and concepts.
9. Some of these analogies were misinterpreted as necessary and definitive
explanations of the neural representations of cognitive processes rather than
just sufficient alternatives.

How these neurons arrive at their high degree of stimulus selectivity is pre-
sumed, by analogy to Hubel and Wiesel’s geometrical properties in the periphery,
to be the result of progressive convergence of less selective stimulus properties
from one level of neural representation to more specialized and selective ones at
higher levels. Because it is not likely that specific sensitivities such as a “rabbit,”
“Albert Einstein,” a “yellow Volkswagen,” or any other entity never previously
60 Single Neuron Practice and Theory

encountered in the course of evolution, could be genetically programmed by

evolutionary processes, it is usually assumed that some degree of adaptive or
reinforced learning is also an important part of the process that determines the
selective stimulus sensitivity of the neuron.

Two Interpretations of “Gnostic” Neurons

Two ideas, in particular, are regularly conflated in dealing with single neuron theory.
We can characterize them as two postulates, one empirical and one speculative.

Empirical Postulate A: Neurons can be observed in the higher reaches of

the nervous system that respond selectively to complex and abstract stimuli.
An example would be the selective activation of a neuron by the presenta-
tion of a particular face or even a political idea.

Speculative Postulate B: The activation of a neuron tuned to selectively

respond to a complex stimulus (such as an individual face) produces the
psychological experience of that face.

There is no remaining doubt that individual neurons can be observed that

respond very selectively to complex stimuli—faces, houses, tools, ideas, ideals,
and so on. Thus, Empirical Postulate A is supported by an abundant literature
justifying its being considered as an empirical fact rather than a speculation or
hypothesis. Barring some subtle technical artifact mitigating the validity of these
observations, Empirical Postulate A seems incontrovertible—neurons with highly
specific sensitivities exist.
However, experimental support for the Speculative Postulate B—the assertion
that correlated psychophysical responses will occur when particular neurons are
activated—is a far more contentious matter. The experimental task of uniquely
linking the activity of a particular neuron to a particular cognitive response is
fraught with technical and empirical difficulties. Such an experiment would require
locating and controlling the same neuron previously demonstrated to respond
selectively to a particular concept, object, or face and then controlling it. This a
far more challenging task than locating neurons that respond selectively to complex
stimuli. It would then be necessary to demonstrate that the appropriate psycho-
logical response is causally associated with manipulation of the state of that particular
neuron. Experiments directly speaking in support of Speculative Postulate B have
never been carried out and probably never can be carried out. The relationship
between the neurophysiological and the psychophysical responses would be supported
only by functional analogy and not by some more robust kind of causal linkage.
Thus, we must ask—is the response of one of these high-level cognitive
neurons just a passive indication of what had been a series of information pro-
cesses that ultimately converged on that neuron? In this case, the correlation of
 Single Neuron Practice and Theory 61

the stimulus and the neuronal response would be cognitively irrelevant and
tantamount to committing Yule’s classical error of attributing causation to cor-
relation. The specific and selective neuronal response, in such a case, would
merely be an epiphenomenon of huge amounts of information that converged
on a particular neuron.
The idea of a stimulus-correlated neuronal response without cognitive sig-
nificance is a very different interpretation than one based on Speculative Postulate B—
the idea that activation of such a neuron is tantamount to experiencing its
meaning. In that context, a specific cognitive neuroscience (neuron to cognition)
connection would have been made as opposed to just a neuroscience connection
(stimulus to neuronal activity). To put these questions in other ways—is the
response of a single neuron, which is tuned to a complex concept, the necessary
code for the experience of that concept? That is, is it the psychoneural equiva-
lent? If the answers to such questions are affirmative, we would have come a
long way toward the solution of the mind–brain conundrum. Unfortunately,
since Speculative Postulate B is beset by practical considerations, effective testing
of single neuron theories of cognition seems unlikely in the near future.

Troubles in the Motor System—the Mirror Neuron

The mirror neuron concept is so fundamentally attractive as a psychological
theme that it has been applied to what is a fair sampling of almost everything
psychological. It reflects a widespread and persistent acceptance of what used to
be called a motor theory of cognitive embodiment. Most famously, it has been
used as a single cell explanation for “action understanding”—a somewhat
ambiguous term for the ability to predict future actions and thus to understand
“cognitive” activities. It is a theory providing a neural foundation for current
and future motor activities by identifying neurons whose responses are associated
with purpose and execution of motor activity.
However, the concept of a single “mirror” neuron as a necessary representation
of a complex cognitive process has come under strong criticism in recent days.
The difficulties are both empirical and conceptual. Efforts to find some correlate
of mirror neurons using fMRI techniques have so far produced negative results
(see the meta-review by Turella, Pierno, Tubaldi, and Castiello, 2009). The most
compelling, persuasive, and far-reaching critique of the concept of single mirror
neurons representing motor activity, however, has been offered by Hickok (2009).
He lists “Eight Problems for the Mirror Neuron Theory of Action Understanding
in Monkeys and Humans.” This authoritative analysis should be read by everyone
who is interested in single neuron theories but also to understand the logical
complexities of mind–brain theories in general. Hickok’s “problems” include:

• There is no empirical evidence in monkeys that mirror neurons sup-

port action understanding.
62 Single Neuron Practice and Theory

• Action understanding can be achieved via non-mirror neuron

mechanisms.
• M1 [primary motor cortex] contains mirror neurons. [This means that
the mirror neuron idea is “nothing more than the facilitation of the
motor system via learned associations.”] (p. 1237)
• The relation between macaque mirror neurons and the “mirror sys-
tem” in humans is either non-parallel or underdetermined.
• Action understanding in humans dissociates from neurophysiological
indices of the human “mirror system.”
• Action understanding and action production dissociate.
• Damage to the inferior frontal gyrus is not correlated with action
understanding deficits.
• Generalization of the mirror system to speech recognition fails on
empirical grounds.
(Hickok, 2009, pp. 1231–1238)

This is not the place to expand upon the details of each of these arguments, but
there are some general inferences that can be extracted from Hickok’s analysis
of the frailties of this particular single neuron theory of motor behavior. These
include the following (after Hickok, 2009):

• The primary criticism Hickok raises of this approach is that it is not sup-
ported by empirical evidence. A considerable amount of subsequent research
has failed to find the associations between perception of motor action and
neural responses.
• The hypothesis of “action understanding” neurons is not unique. The idea
of “mirror” neurons is not necessary and thus Ockham’s razor should be
invoked.
• Many of the so-called supportive experiments actively dissociate critical rela-
tionships between neural and behavioral parameters.
• When these neurophysiological findings failed to replicate, the crutch of
mirror “systems” as opposed to single neurons was invented to patch up the
failing theory.
• Despite these problems, the idea of neural representations of action under-
standing persists. (Action understanding is a new version of traditional motor
theories of cognition in which experiences are tied to actual but subliminal
motor responses.) This reflects the age-old problem faced by psychology of
poorly defined attributes or faculties.

All in all, single mirror neurons as the psychoneural equivalents of cognitive

processes appear to be another one of those popular fads that will persist despite
strong arguments against both the basic concept and the empirical findings.
 Single Neuron Practice and Theory 63

The Exclusivity Question

Another cogent counter-argument against single gnostic or concept coding neurons
arises when one asks—if the activation of a single “pontifical” neuron is capable
of creating an overall, unified cognitive response, what are all of the other neurons
of the brain doing when this pontifical neuron is dominating our thoughts? In
other words, how do we turn all of the other stimulus selective neurons off so that
their responses are not also striving to produce equally compelling mental experi-
ences? A more likely basic assumption, of course, is that vast numbers of neurons
arranged in a network are most likely to be collectively involved in defining our
thoughts. Obviously, there is a disconnect between the concept of the singular all-
powerful “pontifical,” “gnostic,” or “concept” neuron, on the one hand, and the
reality of the huge multitude of other neurons that are known to be active at any
moment in defining our mental states, on the other. I refer to this as the exclusivity
question. Unfortunately, the exclusivity question is also probably unanswerable and
may be a “bad” question emerging as an unintended consequence of the original
postulate of single cell coding. However, this is not the end of the story. There are
many other criticisms that challenge the conceptual as well as the empirical status
of single neuron theory. A few additional challenges include:

1. The conceptual leap from single cells to cognition is too long. It seems illogi-
cal that the vast and intricate neuronal networks should not play a key role in
representing complex ideas.
2. Neural network theories are actually simpler than single neuron theories
given the perplexities and unavoidable complications that follow single neu-
ron theories.
3. Despite the huge number of brain neurons, it seems unlikely that every object
of the external world would have its own neuronal representation.
4. Even if they did, the question of how selectivity for objects and concepts that
did not exist emerges during evolution remains salient.
5. Many cognitive processes do not have simple topological relationships to the
attributes of neural responses and, therefore, simple analogical isomorphism
is not likely to be a satisfactory criterion in determining the psychoneural
equivalent of a concept such as “democracy.” Furthermore, there are some
stimulus dimensions, such as stimulus chromaticity and responses such as per-
ceptual color experience, that cannot be linked together directly by any kind
of straightforward dimensional isomorphism. As a result we must be able to
search for symbolic relationships and interactions—information processing
mechanisms that are executed by the configuration of complex networks
rather than by single neuron states.
6. Perceptual responses are dependent on strong interactions between both the
dimensions of the stimuli and the meaning of the perceptual experiences;
interactions of this kind are not accounted for in single neuron theories.
64 Single Neuron Practice and Theory

7. The sensitivities reflected in the neuronal responses are actually not that good
a match to the concepts being encoded. As noted, what appears to be a face
coding neuron may actually be one responding to primitive shapes (e.g., the
t-shape implicit in the eye-nose-mouth configuration) that are inherent in
face-like stimuli.
8. Artifacts are common—artifacts that are produced by partial properties of the
stimulus.

Finally, like so many other examples of cognitive neuroscience research, replica-

tions of surprising results are rare. As experiments have been repeated, many of
the putative relations between single neurons and complex cognitive processes,
faces, or real objects have disappeared. In other words, the empirical basis for
single neuron theories is neither abundant nor reliable.

2.5 Interim Comment

As provocative as these challenging questions and problems for the single neuron
theory of cognition may be and despite the gradual evolution of the neurore-
ductionist zeitgeist away from pontifical and gnostic neurons toward neuronal
networks, there is a residuum of interest and support for single neuron theories
of cognition.
Single neuron theories, in general, are purported to represent an extreme
“sparse” version on the continuum between localized encoding in the nervous
system and distributed models of cognitive representation involving large numbers
of neurons acting collectively. “Sparse encoding” is another way of implying
that localized neural activity represents a concept with a relatively small number
of highly specialized neurons. Distributed encoding theories propose, on the
other hand, that very large numbers of unspecialized neurons working in concert
and probably involving large portions of the brain are the essential psychoneural
equivalents of cognition. If the latter, it would be extremely difficult to find a
particular neuron that represents a specific idea because the salient neural network
is composed of widely distributed neuronal activations, no one of which repre-
sents the entire concept.
Although the sparse single neuron theory is initially conceptually simpler and
promises to provide a superficially easy opening to the solution of the mind–
brain problem (one neuron–one concept), I have shown in this chapter that it
is also beset with a number of conceptual and technical difficulties. Not the
least of these is the problem posed by the rhetorical question—if a concept is
encoded by the activation of a single neuron or single class of neuron, what are
all of the other active neurons doing? Should the “one neuron–one cognition”
hypothesis be correct, what is the meaning of all of the other neuronal responses
simultaneously being executed when our thoughts are seemingly committed to
only one track?
 Single Neuron Practice and Theory 65

The “distributed” (i.e., neuronal network) theory (which is based on the idea
that neurons in many parts of the brain are responsible for our cognitive experi-
ences) does not have such a simple expression—from its perspective, the simple
one (neuron) to one (concept) idea is incorrect at its most basic level. Be not
misled, however; network theories confront us with the problem of how to
empirically analyze the actions and interactions of the multitude of neurons that
must be operating in concert to represent a particular concept. This poses a
combinatorial problem of network organization that is very likely to be com-
putationally intractable—but by involving all brain neurons in the process, it
finesses the “exclusivity” problem.
Single neuron theories also generate a false need for some kind of binding
process in which the responses of many neurons must be melded into the
single unified experience. However, when one reevaluates this idea, it seems
obvious that the distributed network hypothesis requires no such binding; all
neurons are incorporated and required to collectively produce the psychologi-
cal response. The state of the network—the configuration of all of the neurons
involved in the network—is itself the psychoneural equivalent. No longer do
we have one neuron equaling one concept but in its place we have the state
of an extended multi-neuron network that offers few opportunities for a
detailed analysis.
As a result, research attention has gradually shifted away from single neurons
encoding both simple and complex properties to networks and macroneural
brain locations as new devices (such as the microelectrode array and the fMRI,
respectively) have offered alternative measures of nervous system activity and its
relation to cognitive processes. Because network theories of various kinds rep-
resent the current consensus, in my opinion, most contemporary cognitive
neuroscientists look upon “grandmother” neurons and their ilk as antique curi-
osities. One reason they have not received more attention is that many of the
critical experiments are not executable in practice and in principle. Should one
be interested in considering single neuron theory further, reviews by Sevush
(2006) and Bowers (2009) present the opposing view. Our attention now shifts
to the antithesis of the single neuron theory—the technology and theory that
have developed to deal with multiple neurons interconnected in networks of
varying degrees of complexity.

Notes
1. This is an excellent example of how the basic chemical properties of a molecule can
lead to the production of organized cells and tissues. Any philosophy that assumes that
there is a neural foundation for mind must assume that such natural processes also
guide organic evolution.
2. Although the electrical activity of a neuron is a useful indicator of its activity, it is
important to note that these electrical signals are really epiphenomenal. The real
mechanisms of neuronal activity are chemical, specifically ionic and molecular, which
66 Single Neuron Practice and Theory

create electrical actions as the chemical activity takes place. It is far easier to measure
electrical activity than it is to measure chemical activity and, thus, much of the discus-
sion of neuronal action is presented in electrical terminology.
3. There is considerable controversy concerning the first use of glass, electrolyte-filled
microelectrodes. Galambos (1996) suggests in a history of neurosciences in his biography
that “glass micropipettes” may have been in use as early as 1910 (p. 187). Thus, their
use was certainly being explored elsewhere around the time Gerard and his group
published their methods.
3
MICRONEURONAL NETWORK
THEORIES—TECHNIQUE AND
METHODS

What is a microneuronal network theory of cognition? In its most extreme form,

it would be a complete statement of the structure and activity of each and all of
the relevant neurons involved in a cognitive process. It would show that particular
neuronal networks are both necessary and sufficient to produce a thought or expe-
rience. Activating the neuronal network would presumably produce an experience
and, then, if the neuronal network changed, that experience would change. In some
ideal experiment, we would be able to identify the state of each neuron and dem-
onstrate exactly the role it plays in creating the mental experience. The language
would be a low-level one phrased in the scale of neurons and synapses.
Obviously, this ideal conceptualization of a microneuronal network is not
achievable—any hope for such a detailed microneuronal “explanation” or “the-
ory” of anything mental dies quickly on the altar of simple numerousness. Not
even the most powerful conceivable computer would be able to process all of
the possible combinations and permutations of an appropriately sized and inter-
connected neuronal network. Thus, in the past, we have turned to different kinds
of prototheories such as the macroneural and single neuron theories just described
in Chapters 1 and 2. Metaphors, analogies, and simple recapitulation of the
observations take the place of complete, explanatory theory as we turn to what
in hindsight are only simplistic and trivial explanations. We construct “theories”
that are partial, explain nothing, or simply reiterate the basic principles of neural
function. It is an unavoidable conclusion, therefore, that a complete explanation
of how mental activity emerges from a tangle of interacting neurons is likely to
remain the greatest mystery in modern science.
However, there are some principles on which most cognitive neuroscientists
agree. The main one is that mind is probably the result of the operation of
matter (and, most probably, microneuronal network organization and activity).
68 Microneuronal Network Theories

This is the basic ontological premise of modern cognitive neuroscience. That

we can prove this assertion seems increasingly unlikely. However, this does not
mean that we cannot learn anything about the mind or the brain by continuing
to pursue some current paradigms. Psychologists have added enormously to our
understanding of behavior, and neurophysiologists and neuroanatomists have
explicated the nature of the components of the complex networks that make
up the brain in a way that produces deep insights into what may be going on
in the great networks of the brain.
Some of these insights come in the form of ever deeper metaphors for mind–
brain relationships expressed in the language of other sciences. I now propose
another metaphor—a much reduced version of what is probably going on in
the brain but one that clarifies some of its basic organizational properties. I call
this the “register” metaphor, a term borrowed from computer science.
First, what do I mean by a “register”? A register in a computer is a group
of binary elements that function together to select one of a large set of alterna-
tive outputs. For example, consider Figure 3.1. This register consists of 12 simple
elements, each of which can be either in the “on” or “off ” state. Computer
engineers call them bits. This means that there can be 212 or 4096 different
output states defined by the 12 input bits. In other words, a very small set of
input components can determine which of a greatly expanded range of possible
outcomes will occur. Yet each input component maintains its own integrity and
is used in many different configurations of the 12 components. By itself each
element tells us nothing about the state of the register, but as a member of the
group it is very powerful in defining a large range of outputs. Each bit in this
register is, thus, functionally connected to every other bit and participates in
defining the output of the register, yet is still independent.
Assume that that there are not 12 but millions of neurons in the brain’s
“superregister.” Assume further that each of these components of the register
can have 1,000 or more states defined by its synaptic inputs. Somewhere in this

4096
possible
output
states

1 2 3 4 5 6 7 8 9 10 11 12

12 bit register
IN

FIGURE 3.1 A 12-bit register can represent a much larger number (4096) of output
states.
 Microneuronal Network Theories 69

innumerable universe of possible brain states, there are emerging processes of

which we are aware only in the form of experience—of human sentience or
by observing the behavior of others. The fundamental postulate of the register
idea is that even if the number of neurons is small, the number of states they
can specify is vastly larger. Furthermore, given that the actual number of neurons
is itself large, a much larger number of configurations and states can be encoded
with the register than with any kind of single neuron. From one point of view,
the superregister model of the brain is a surprisingly complete metaphor for
what is happening in the brain. It incorporates many of the ideas that are widely
accepted these days for individual neurons and for interactions among those
neurons. It emphasizes a number of important aspects of mind–brain relations
such as the role of the states of activity of multiple numbers of individual neu-
rons. It does so without invoking any supernatural concepts or introducing any
incompatibilities with any physical law.
It also instantiates some other ideas that seem relevant. For example, the role
of any component may not be constant. Instead, what a “bit” means varies with
the task being confronted by the register. Likewise, what a bit means is dependent
on the role of all of the other bits in the register. Thus, there is a considerable
increase in the complexity and number of concepts—a multiplier that extends
the possible outcomes far beyond even what the 86 billion or so individual
neuronal states might encode. A further source of uncertainty is that individual
neurons may participate in several different networks at different times and in
different tasks. Finally, the register system is robust; if a bit fails the system may
still work—at least partially.
However, from an alternative point of view—practicability—register theory
is, by its most fundamental nature, incapable of complete fulfilment as a theory
of mind. The number of neurons involved in a realistic cognitive process raises
the impossibility of dealing with all of the details of the neurons and synaptic con-
nections and units of this vast network. As such, the most hopeful prognosis—
specifying the exact configurations and states of the components of the brain’s
superregister that embody a cognitive process—is beyond any hope of realization.
Thus, we are confronted with the unhappy conclusion that the main goal of
cognitive brain theory—“breaking the neuronal code”—is not possible unless
there is some kind of an intellectual and conceptual breakthrough not currently
imaginable.
This conclusion may be interpreted to mean that we cannot go beyond what
are essentially partial and incomplete metaphors of mind–brain transformations
that exist today. Despite the fact that superregister theories must be correct in
general principle, our analyses of them are currently constrained to simple toy
models (i.e., reduced versions of the brain’s actual neuronal network). In addi-
tion, they suffer from the same problem that all other theories of mind–brain
do. They finesse or eschew the most basic problem of all—how does intangible
mental activity arise from tangible neural activity? On this point, there is no
70 Microneuronal Network Theories

argument—we do not yet know how brain becomes mind. This lack of progress,
fortunately, has not obstructed thinking about neuronal networks. The starting
point for a consideration of a microneuronal theory of how the brain works is
Donald Hebb’s famous conjecture, the next topic in our discussion.

3.1 Hebb’s Conjecture

One of the most amazing, insightful, and creative examples of evidence-poor,
but widely accepted theoretical speculations about the way in which neurons
become mind is Donald Hebb’s (1949) formulation of what we would now
designate as a pioneering microneuronal network theory of cognitive processes.
The essence of his conjecture was that cognitive states are encoded, implemented,
or instantiated in the brain by the individual states of and interactions among
neurons as mediated by variations in the strength of synaptic junctions. The
synaptic variation is regulated by the use or disuse of those junctions in networks
that he referred to as cell assemblies and phase sequences. In Hebb’s words:

Any frequently repeated, particular stimulation will lead to the slow develop-
ment of a “cell-assembly,” a diffuse structure comprising cells in the cortex
and diencephalon (and also, perhaps in the basal ganglia of the cerebrum),
capable of acting briefly as a closed system, delivering facilitation to other
such systems and usually having a specific motor facilitation. A series of
such events constitutes a “phase sequence”—the thought process.
(p. xix)

This conjecture, based as it was on almost no empirical evidence other than

what was then known about the nature of individual neurons, was a milestone
in physiological psychology—the name by which cognitive neuroscience was
known at that time. In presenting his schema, Hebb explicitly rejected other
attempts to deal with psychological problems. In particular, he disavowed any
interest in a neuroreductionist behaviorism or classical stimulus-response schema.
What he offered, in their place, was the possibility of an explanation of cogni-
tion in terms of very specific neuronal concepts and measures. To understand
how profound his contribution was, it is desirable to list the basic properties of
his formulation:

1. All aspects of mind, cognition, consciousness, or whatever else you may wish
to call mental activity, are the results of interneuronal processes and mecha-
nisms. These processes and mechanisms are to be found at the microscopic
level of neurons, not larger anatomical units.
2. Specific patterns of neuronal activity are the “psychoneural equivalents” of
specific cognitive processes. Specification of the detailed state of participating
neurons produces a particular cognitive process.
 Microneuronal Network Theories 71

3. Hebb’s formulation is one of action and interaction of a complex system of

component parts—a network.
4. These component parts are, in Hebb’s formulation, neurons (cells), a constitu-
ency that defines the conceptual level of his model as a microneuronal one.
5. These mechanisms are not unique in that “. . . alternative pathways each
having the same function . . .” (p. 74) may exist. The system is therefore
redundant—multiple neurons, different cell assemblies, and phase sequences
can produce the same response.
6. The neuronal mechanisms of cognitive processes involved are distributed
throughout the brain, the only exceptions being sensory and motor regions
and the “. . . so-called speech area . . .” (p. 284).
7. The basic neuronal correlate of change in cognition is the establishment of
changes in the efficacy of the interneuronal connections as a result of use. In
Hebb’s words:

When an axon of Cell A is near enough to excite a Cell B and repeat-

edly or persistently takes part in firing it, some growth process or
metabolic change takes place in one or both cells such that A’s efficiency,
as one of the cells firing B, is increased.
(p. 62)

8. The “growth process” or “metabolic change” mechanism of changing inter-

connectivity between neurons is the increase in synaptic conductivity or effi-
cacy, specifically in the growth of “synaptic knobs.”
9. It was not known at the time of Hebb’s conjecture exactly what was the
functional equivalent of “change” in interneuronal efficiency. However, an
increase in synaptic efficacy is now often referred to as Long-Term Poten-
tiation (LTP)—the relatively persistent increase in synaptic conductivity that
occurs following repetitive stimuli. (We nowadays have a much more com-
plete understanding of how synaptic conductivity is modulated at the molec-
ular level.)
10. There was no allusion by Hebb to specific localized areas, but brain-wide
equipotentiality was not assumed.
11. Specific neurons do specific jobs in specific cell assemblies.
12. The system requires great numbers of neurons.
13. “There is no good evidence” for some of these ideas, nor do they provide “a
means of testing” them (p. 66).

There is no question that Hebb’s conjectures about cell assemblies were

stimulated by the outburst of neurophysiological research in the first half of the
20th century, especially by such notable neurophysiologists as Lorente de No
(1938) and Warren S. McCulloch (1944). It is somewhat more problematic
whether World War II developments in information theory and electronic circuits
72 Microneuronal Network Theories

had any direct effect on his conjecture. Notwithstanding the difficulty of

tracing the origins of ideas, there is no question that his integration of a
number of these ideas into a coherent conjecture was a monumental and
historic event for neuroreductionist psychology. Hebb’s conjecture, modestly
modified, still represents today what most cognitive neuroscientists consider to
be a working schema for the otherwise still unanswered mind–brain question.
It is surprisingly modern and almost all of these statements would find broad
agreement despite the fact that some of them (e.g., number 6) run against the
current zeitgeist.
Thus, it should come as something of a shock for people who are not at
the core of cognitive neuroscience thinking to realize how sparse is the empiri-
cal support for Hebb’s conjecture—that cognitive processes are the result of a
particular kind and arrangement of microneuronal activity as modulated by
variations in synaptic conductivity. Although plenty of evidence shows that
synaptic efficacy changes with use (LTP), only recently has some information
appeared (Nabavi et al., 2014) that links synaptic efficiency changes to behavior
such as learning. The history of the widely accepted, but scantily supported,
Hebb conjecture has been described in detail by Takeuchi, Duszkiewicz, and
Morris (2013). In that article they also make clear the paucity of direct support
for the conjecture. In general, therefore, the gap between behavior and Hebbian
neuronal networks remains obscure. However, this is what makes Hebb’s “con-
jecture” just that and not an empirically robust “theory.” As important as it
was in the evolution of thinking about the relation between neurons and cog-
nition, there is a real question about whether it is testable. We now examine
this important question.

Is Hebb’s Conjecture Testable?

Microneuronal interactions are the bases of the Hebb conjecture (Hebb, 1949)—a
conceptualization that assumes that cognitive processes emerge from the idio-
syncratic action of millions, if not billions, of individual neurons and the thousand-
fold greater number of synaptic connections; not in terms of their cumulative,
pooled, aggregated, or summarized values, but instead to the degree they maintain
their individual identity in the neuronal network. The main problem with the
Hebb hypothesis, despite its logical force, conceptual robustness, and near universal
acceptance, is that it has almost no empirical basis given the practical problems
preventing simultaneous and independent manipulation of a cognitively significant
number of neurons.
To establish the validity of the Hebb model would require an unobtainable
amount of individual control over an uncountable number of neurons. Consider
how a microneuronal experiment would have to be carried out. Ideally, the
brute force technique of manipulating each neuron would require that we
accomplish the following preliminary instrumentation steps:
 Microneuronal Network Theories 73

1. Identifying each relevant neuron and all of their interconnections that are
involved in a given cognitive process
2. Individually stimulating each of these identified neurons with a coherent
pattern of energies tailored to each neuron’s needs and roles in the relevant
network encoding the cognitive process
3. Recording individual responses from every neuron to determine its individual
and specific response in the relevant network

Obviously, carrying out these tasks would be a technical tour de force of the
first magnitude given that our ability to develop microelectrode arrays suitable
for stimulation or recording has so far been limited to a few hundred (e.g., Wark
et al., 2013). Furthermore, the amount of data obtained by even a few hundred
microelectrodes would tax available storage capabilities.
Should some extraordinary technical development allow us to carry out these
demanding experiments, we would then have to determine how well the criteria
for associating cognitive and neurophysiological measures proposed by Martin,
Grimwood, and Morris (2000) are satisfied. Meeting their four criteria would
authenticate that a microneuronal network response is the causal representation
of a cognitive process; in other words, a theory of how mind emerges from
brain.

1. Detectability: Stimulate the microneuronal network by driving it with a

cognitive stimulus. Measure the pattern of neuronal network activity that is
elicited by this “stimulus.”
2. Mimicry: Activate the previously recorded pattern of neuronal activity and
determine if it reproduces the cognitive process previously used as a stimulus.
3. Anterograde Alteration: If the pattern of network activity is blocked from
forming, the cognitive process should also be blocked.
4. Retrograde Alteration: Changes in the neuronal network state should pro-
duce corresponding changes in the cognitive process.
(Paraphrased from Martin, Grimwood, and Morris, 2000)

Obviously, for a host of current procedural and instrumentation reasons, the

technological tasks enumerated in the past several paragraphs cannot be carried out.
It has not proven possible to record simultaneously and independently from anywhere
near the number of identified neurons that are almost certainly involved in a cogni-
tive process. Nor are we able to selectively stimulate an identified group of individual
neurons to determine if we can mimic, block, or change a cognitive process.
To decode a cognitive state into a specific neuronal network, the implication
of Hebb’s first assumption, would require that Martin, Grimwood, and Morris’
second mimicry criterion be satisfied. That is, controlled replication of the spe-
cific synaptic conductivity pattern (and the resulting specific settings of the
individual neuronal state) should produce a specific cognitive state.
74 Microneuronal Network Theories

Because of these technical and procedural limitations, the brute force, micro-
neuronal strategy of stimulating and recording from a realistically large network of
individual neurons is probably not implementable and can only be conceived of as
a hypothetical Gedanken experiment. Even if by some technological magic we
were able to produce the necessary technical instruments, constraints on data accu-
mulation and analysis would overwhelm any conceivable computer solution to the
mind–brain for simple combinatoric reasons. Thus, there is a glaring inconsistency
between the microneuronal level at which neurons really operate and the macro-
neural level of data obtained at the level of current brain-imaging technology.
However, the second part of Hebb’s conjecture—that it is synaptic efficiency
that controls the state of the network (i.e., producing particular cell assemblies
and, thus, particular cognitive states)—has been tested, in general if not in indi-
vidual neuronal detail. Indeed, this is what has been accomplished by Nabavi
et al. (2014) in an extraordinarily important experiment during which this group
activated and deactivated a brain circuit purported to control a behavioral response
by means of genetic manipulations. Specifically they were able manipulate syn-
aptic activity in the form of LTPs to establish a conditioned fear response and
then reverse this “cognitive” state by producing a Long-Term Depression (LTD—a
reversal of the potentiation) with different optogenetic stimuli. This was accom-
plished by using optogenetic stimuli to alternatively activate and deactivate
synaptic activity, respectively. This procedure permitted Nabavi and his colleagues
to enhance (LTP) or reduce (LTD) activity in the amygdala—a region of the
brain associated with conditioned fear in the rat—by applying two different
trains of patterned light as opposing stimuli.1
The extraordinary outcome of this experiment was that by using this tech-
nique, the rat could be controlled to exhibit a conditioned response analogous
to fear under the LTP conditions, but behavioral passivity under the LTD condi-
tions. Furthermore, and most interesting, was that these researchers were able to
reversibly turn the fear behavior on and off by alternating between the LTP and
LTD synaptic conditions. The proximal independent variable in this experiment,
therefore, is the degree of conductivity of the collective responses of many syn-
apses in a region (the lateral amygdala) of the rat brain thought to be associated
with a well-known behavioral experiment: The dependent variable was a specific
behavior. This experiment provided the best evidence yet obtained to support
what only had been a hypothesis for many years—that the synaptic activity
regulated the behavior, thus supporting the Hebb conjecture.
Although we still do not know the exact synaptic or neuronal states at the
individual neuronal level for this model cognitive process, Nabavi and his col-
leagues have now shown that manipulation, albeit collective, of the degree of
synaptic potentiation produces correlated (and possibly causally related) changes
in an observable behavior. This is solid support for the parts of Hebb’s conjecture
that state that, in general, synaptic processes regulate neuronal states and thus
behavior. Given all of the challenges in this complex field of research, this is a
 Microneuronal Network Theories 75

remarkable achievement. This experiment meets the first, third, and fourth criteria
enunciated by Martin, Grimswood, and Morris. The one criterion that it does
not satisfy is mimicry, which requires information about specific patterns of
individual synaptic conductivity—information that still remains elusive.
What Nabavi et al. accomplished, to sum up their conclusion, is that synaptic
states can alter behavior by turning on and off a particular synaptic configura-
tion, thus confirming one of the prime assertions of Hebb’s conjecture. In their
words:

Thus, we have engineered inactivation and reactivation of a memory using

LTD and LTP, supporting a causal link between these synaptic processes
and memory.

Of course, this does not speak to the other part of Hebb’s conjecture—that is, how
the detailed pattern of the neuronal and synaptic states are decoded into specific
cognitive states. That would require meeting the mimicry criterion, a test that is
not applied in the Nabavi et al. experiment and may not be possible to satisfy.
Why should it be so difficult to provide an answer to the first part of Hebb’s
conjecture and to develop a microneuronal network theory of cognition? The
answer to this question is that some of the very properties that make the Hebb
conjecture plausible and logical also make it intractable to conventional labora-
tory experimentation. These properties include:

1. Microneuronal Level: The small size of the constituent components (neurons)

2. Numerousness: The large number of neurons involved in any cognitively
significant neuronal network
3. Mechanical Obstructions: Practical limits on the way in which recording and
stimulating electrodes can be brought to bear on neuronal tissues
4. Complexity: The rapid growth in the number of alternative arrangements of
only a few neurons, much less the number of involved neurons in the simplest
conceivable realistic network
5. Unpredictability: Responses are not deterministic but are stochastic; random-
ness would kill such an experimental approach even if the practical difficulties
just mentioned could be overcome
6. Intractability: The dearth of mathematical and computer methods for solving
even what seem to be relatively simple network problems, much less those
corresponding to a “thought”
7. Underdetermination: Multiple explanations exist for any observed macro
response
8. Redundancy: There may be no single answer to the question of psychoneural
equivalency
9. Hidden Causal Influences: Indirect interneuronal connections may simulate
direct connections
76 Microneuronal Network Theories

The only partial exceptions to these generalities are to be found in simple

invertebrate “model” preparations such as used in the work on Limulus (Hartline,
Wagner, and Ratliff, 1956), on Aplysia (Kandel and Tauc, 1964; Kandel, 1991)
and Drosophila (Vogelstein et al., 2014), or in the many examples of well-
structured peripheral sensory mechanisms such as the somatosensory cortex
(Schwarz et al., 2014) or the retina (Litke et al., 2004; Margolis, Gartland, Singer,
and Detwiler, 2014). It is in these locations or model preparations that the
neuronal networks are simple enough for us to have detailed knowledge of their
arrangement and how stimulus properties are specifically represented by neuronal
states. Of this topic, I have much more to say in Chapter 4.

3.2 Microelectrode Arrays

A long-term motivating goal of neurophysiologists and cognitive neuroscientists
is to observe the activities of neuronal networks. The conceptual basis for
attempting to meet this goal is the so far unverified (but also the most plausible)
conjecture that the cognitive information processing capabilities of the nervous
system are embodied in the action and interaction of networks of microscopic
neurons. One way to approach the goal of recording from many neurons is to
alter our investigative strategy, which has hitherto emphasized single microelec-
trodes, and strive to develop Multiple Electrode Arrays (MEAs) to observe the
activity of many neurons simultaneously. The hope is that by observing the
activity of many neurons, we may be able to understand how neuronal networks
operate and then to understand how the brain accomplishes its amazing func-
tions. Whether the mechanisms and analyses of the situation will ultimately
permit us to study simultaneously enough neurons to understand their relation
to cognition is an empirical question to be resolved in the future.
Although the future is unpredictable, we must acknowledge that the current
ideal of experimental manipulation of the vast numbers of neurons and their
synaptic connections involved in cognitive processes is obviously still a wish
rather than an accomplishment. Although we may learn an enormous amount
about interactions of a few neurons or the specifics of synaptic growth, it seems
unlikely given the enormous number of neurons, the redundancies, and the
irregular and idiosyncratic nature of the brain’s neuronal networks, that the
necessary criteria of the kind proposed by Martin, Grimwood, and Morris could
ever be satisfied.
This is not to deny that, at the present time, a considerable amount of progress
has been made in developing MEAs that can record from hundreds, if not
thousands of active neurons. This approach, however, poses its own challenges;
some are simply computational and arise because of the many ultramicroscopic
electrodes themselves that must be monitored at once. There is no assurance
that even an MEA with thousands of electrodes would be sufficient to solve the
problem of how neuronal networks account for cognition. Indeed, as we shortly
 Microneuronal Network Theories 77

see, just increasing numbers sometimes produces its own challenges. The need
for the ability to examine the function of many neurons simultaneously is obvi-
ous given the polyneuronal basis of current neuronal network theory.
However, there are several caveats that have to be made before we embark
on a discussion of the development of MEAs. First, no matter how many elec-
trodes may be combined into an MEA, the number of electrodes (and, thus the
number of neurons from which we may record) will probably always be smaller
than the enormous number of neurons probably involved in any cognitive
process.
Second, as we succeed in developing the technology to record from more
and more neurons, we will be facing an increasing counterforce—the increasing
combinatoric complexity and skyrocketing computational and storage require-
ments to analyze even modest-sized networks.
Third, despite some ingenious use of miniature manufacturing techniques
originally developed for computer chip fabrication, MEAs, in general, consist of
electrodes that are still too large to isolate the ultramicroscopic details of a cog-
nitively significant neuronal network. This is a technological issue that may be
overcome in the future, but, for the moment, represents a serious barrier to
progress.
Fourth, MEAs, which are almost universally extracellular, consist of electrodes
that connect to surrounding neurons in a way that is currently not predictable
or regular, but at least quasi-random. Furthermore, even if the connections were
deterministic rather than stochastic, it is by no means certain that the intricately
interconnected neuronal network of many neurons would map directly from
the MEA to the nervous system. Distant neurons may produce large responses
that are incongruent with the geometrical array implied by the MEA. Alterna-
tively, nearby ones may produce fortuitously small responses for what are usually
obscure reasons where one would have expected large responses. Thus, the goal
of inferring neural network organization from the observed responses may not
be achievable—we may not be able to solve the clustering or sorting problem
(to be introduced shortly) in which responses are associated with specific neurons
or to achieve our goal of determining the connections between particular neu-
rons. The problem of limited detailed knowledge of electrode-neuron connectivity
is exacerbated by the dynamic, transitory nature of neuronal networks.
In general, we do not have adequate control over which neurons will be
connected to which electrodes. Thus, associating the geometry of a particular
neuronal network with cognitive processes remains a challenging task. Therefore,
MEAs are likely to tantalize us by providing data that is conceptually relevant
(cognition is a product of a microneuronal network) but that will never be able
to fully satisfy criteria such as those proposed by Martin, Grimwood, and Morris
(2000) for robust empirical proof. Make no mistake, we are certain to learn
wonderful things about both the nervous systems and cognition, but, for the
moment at least, what seems currently promising may just be another dead end
78 Microneuronal Network Theories

in our quest to explain how brain matter becomes or produces intangible expe-
rience. For these reasons, it is unlikely that even the most advanced MEA will
ever be able to provide a foundation for a detailed explanatory theory at this
microlevel, however promising the idea may be. It may be necessary to concep-
tualize the problem in some completely different way that we cannot currently
anticipate to break the neuronal network–cognitive code.
Although it is currently impossible to simultaneously record from enough of the
enormous number of neurons that must be involved in even the simplest cognitively
relevant network, efforts have been made to construct assemblages of electrodes so
that the activity of a larger number of neurons can be recorded simultaneously than
is possible with a single microelectrode. Currently available MEAs consist of any-
where from a few to several thousand electrodes. The guiding presumption is that
by observing the sequence of activations or correlations among a modest number
of neurons we may achieve some insight into the basic properties of more realisti-
cally sized neuronal networks. The following section details some of the technological
progress that has been made in producing MEAs in the past decade or so.

Types of Microelectrode Arrays

Penetrating Microelectrode Arrays

Multiple electrode arrays come in two main types. The first is designed to
penetrate into tissue or chunks of tissue; the second is designed so that more or
less isolated neurons are laid on top of a printed array of electrodes. An example
of the first type simply binds a number of individual wire electrodes together
into a bundle (Figure 3.2). Because this type of MEA is designed to be thrust
into neural tissue, it must consist of mechanically robust shafts. These electrodes
are particularly appropriate for in vivo preparations in which the normal arrange-
ment and metabolism of the neural tissue must be preserved. Electrode arrays
of this type typically pick up a mix of signals (e.g., local potentials, spike action
potentials, and electronic noise), depending mainly on the random geometrical
relationships between the sampled neurons and the MEA. The individual elec-
trodes making up the MEAs are designed so that certain positions (the tip and,
in a few instances, other points along the shank) can be left uninsulated and
thus make electrical contact with parts of neurons.
The simplest and earliest kind of penetrating electrode was a brush-like device
in which individual wire electrodes were simply bundled together (Kruger and
Bach, 1981; Nicolelis and Chapin, 1994). Preparation of the bundled wire elec-
trodes is minimal. Other than stripping the insulation to expose an electrical
contact, and possibly plating the tips of the several electrodes with a metal such
as gold, silver, or carbon black for improved contact between the nervous tissue
and the microelectrode, little more need be done to prepare the individual wires
beyond what has already been described for individual microelectrodes.
 Microneuronal Network Theories 79

FIGURE 3.2 An early form of a multiple electrode array (MEA) formed by bundling a
group of wire electrodes (a) and the typical irregular and mixed responses recorded
with such a bundle (b).
Reproduced from Kruger and Bach (1981) with permission.

Figure 3.2A shows the arrangement of 30 bundled microelectrodes as constructed

by Kruger and Bach. Figure 3.2B shows a sample of the responses recorded by
them from 12 of their 18 electrodes. As was typical in all MEAs, not all of the
bundled electrodes picked up a neuronal signal. It is important to note the irregular
nature of the responses resulting from “noise” of various kinds obtained with these
early MEAs. A major problem, both with regard to these early records and with
later improved techniques, was that it was difficult to reliably determine which
neuronal responses went with which neuron. The idiosyncratic heterogeneity of
the responses picked up by simple “bundling” electrodes of this type is clearly
evident. This problem, then, engendered a problem special to MEAs—the lack of
a direct association of neural responses with their source neurons. We encounter
this problem repeatedly in the following discussion.
The wire “bundle” electrode, however, does have one very important
advantage—it is robust enough that it can be thrust into the brain of a living
animal for “in vivo” recording. Thus, it can record from nervous tissue under
relatively normal undamaged conditions. Its main disadvantage is that the indi-
vidual wire electrodes are irregularly mapped onto the brain tissue. A compromise
80 Microneuronal Network Theories

between a brush MEA with its unavoidable irregular electrode placement, on

the one hand, and more mechanically robust and regular MEAs, on the other,
was developed by a group at the University of Utah (Nordhausen, Maynard, and
Normann, 1996). This array has many of the advantages of the bundle electrode
but also overcomes many of its deficiencies.
The Utah electrode was constructed on a silicon substrate by a micromachin-
ing process similar to that used to produce computer circuits that etched away
unwanted material. These electrodes had some important advantages in that they
could pick up signals from a regular two-dimensional space and potentially also
in three-dimensional space, if active electrode points could be set up on the
shafts in addition to the one at its tip. Another advantage of the Utah array was
that these electrodes were robust enough to remain inserted in brain tissue for
periods that could be as long as 13 months (Rousche and Normann, 1998). An
example of this kind of electrode is depicted in Figure 3.3.
Effort is currently aimed at developing MEAs in which many of the multiple
electrodes in the array actually penetrate into the neurons, thus producing much
larger intracellular electrical signals than was possible with extracellular MEAs.
If this complex engineering task can be accomplished, then intracellular signals
orders of magnitude stronger than extracellular ones can be recorded; this may
open the door to MEA studies of synaptic potentials as well as the spike action
potentials (SAPs) now being routinely examined. All of this depends, of course,
on not being overwhelmed by the demands of the enormous amount of data
typically generated by experiments using MEAs.
Improving the capability of penetrating MEAs has been the goal of other
investigators. Kibler, Jamieson, and Durand (2012), for example, developed a

FIGURE 3.3 An MEA formed by a silicon micromachining technique consisting of

100 electrodes.
Reproduced from Nordhausen, Maynard, and Normann (1996) with permission.
 Microneuronal Network Theories 81

method for producing high-aspect (in which the length of the electrodes con-
siderably exceeds its width) MEAs so that the electrodes could be pushed deeply
into slabs of mouse hippocampus. This large penetrating depth was required
because the neurons of interest in their experiments were located deep in the
brain. Kibler and his colleagues used arrays consisting of as many as 64 electrodes.
Spira and Hai (2013) present a comprehensive discussion of these new develop-
ments in penetrating multielectrode technology and the problems they raise, as
well as their advantages for the simultaneous recording of multiple intracellular
responses.

Planar Microelectrode Arrays

The second type of MEA does not depend on penetrating more or less intact
neural tissue. Instead, it is designed to record from neurons that have been isolated
from their neighbors by dissection or chemical means. The idea when using this
alternative is to print a planar set of electrodes and allow them to make surface
contact with either dissociated or cultured neuronal preparations as shown in
Figure 3.4. This figure is of special interest because it shows not only the 36
printed microelectrodes (the tips of the dark lines) but also some neurons and
glia with which the electrodes are making contact.
Low-voltage, extracellular responses are the expected result of using this kind
of printed electrode since the effective diameter of the individual electrodes in
this kind of MEA is typically larger than the 1 μm diameter of glass or metal
microelectrodes; they are therefore not suitable for use in in vivo intracellular
recording but are primarily limited to in vitro extracellular recordings from
neuronal tissue that is simply laid on top of them or excised tissue on which
they can be placed. The responses recorded with such an electrode depend
upon the uncontrolled physical relationships between the neurons and the
electrodes.
Printed microelectrode arrays of this kind have been used to observe cellular
activity since the 1970s (Thomas, Springer, Loeb, Berwald-Netter, and Okun,
1972; Gross, Rieske, Kreutzberg, and Meyer, 1977; Pine, 1980; Gross, Williams,
and Lucas, 1982) and continue to be improved by current workers. The number
of electrodes that can be fabricated continues to rise, with 64 or 100 being
routine and 1,000 or more now being at hand. Indeed, Schwarz et al. (2014)
reported a system based on cubical arrays of electrodes that could, in principle,
record from as many as 1,792 electrodes, although random effects usually meant
that far fewer were active at any time. The work reported by this last group
was carried out on rhesus monkey primary somatosensory and motor cortices
in which topological relations with the external world were preserved. As a
result they were able to cluster neuronal SAP responses and to show that dif-
ferent areas of these peripheral regions of the brain correspond to different
sensory and behavioral motions.
FIGURE 3.4 An MEA formed by a photoetching technique consisting of 36 “printed”
electrodes. Neurons and other pieces of the mouse’s spinal neurons can be seen in
the background for scale.
Reproduced from Gross, Williams, and Lucas (1982) with permission.
 Microneuronal Network Theories 83

However, there remain profound difficulties in using such an array. The simple
act of mechanically connecting large numbers of multiplexed preamplifiers to such
an array is a challenging task unto itself. As we see, most modern research is still
limited to the analysis of relatively few neurons compared with the number of
neurons that must be involved in even the simplest kind of cognitive process.
The main problem with such an MEA, as mentioned earlier, is not
technological—indeed, advances in instrumentation, especially wireless transmis-
sion and microengineering procedures, are likely soon to overcome many or all
of these handicaps. Instead, it is the vast amount of information generated by
even a modest number of electrodes that may define the limits of their application.
What the future holds, of course, cannot be predicted and there has been
continuing progress in increasing the number of available electrodes, if not the
ability to process the large amount of data they could conceivably generate. In
particular, Eversmann et al. (2003) produced an MEA consisting of 16,384
electrodes arranged in a regular array of 128 × 128 individual electrical contacts.
Their MEAs were fabricated from complementary metal oxide semiconductor
(CMOS) technology. Figure 3.5 shows a small portion of Eversmann’s electrode
array and some disassociated neurons that have been deposited on it.
To the best of my knowledge, MEAs of this size have never been used in a
real experiment beyond limited demonstrations of the advanced technology

FIGURE 3.5 An MEA formed by a photoetching technique consisting of approximately

16,000 electrodes. A neuron can be seen lying on the MEA.
Reproduced from Eversmann et al. (2003) with permission.
84 Microneuronal Network Theories

developed by investigators such as Eversmann and colleagues. Nevertheless, MEAs

of this magnitude suggest the possibility of a whole new approach to research
on neuronal networks and contain the germ of applicability to cognitive neuro-
science when combined with the advent of powerful supercomputers.
Steidl, Neveu, Bertrand, and Buisson (2006) have succinctly summarized some
of the advantages that MEAs might have for neurobiological research:

• There is a need for micromanipulators to position electrodes in the tissue.

• Multiple electrodes can be used either for stimulation or recording.
• Printed MEAs are less sensitive to mechanical vibrations than are single glass
microelectrodes.
• Printed MEAs can be used several times if adequately cleaned after each
recording session.
• 3D-tip–shaped MEAs can pass through dead surface layers and record closer
to living neurons within an in vitro slice.
• Parallel recordings at multiple electrode sites in a single slice of tissue provide
the opportunity to observe region-specific effects, to increase individual data
points and to improve statistical analysis.
(Paraphrased from Steidl et al., 2006)

As I noted earlier, it seems that the largest body of MEA-related research

currently being carried out is aimed at engineering improved MEAs rather than
neurophysiological or cognitive research per se. This emphasis includes research
on topics that could generally be classified as housekeeping functions—for
example, efforts to optimize the functions, simplify the construction, or avoid
toxic effects of the materials used in MEAs. The following projects are examples
of some of this kind of housekeeping research. The important message in this
context is that although research on the design and construction of planar Multi-
ple Electrode Arrays (pMEAs) continues, relatively little of this activity has yet
been applied to the ultimate application (for cognitive neuroscientists) of these
devices—understanding how neurons function in networks to produce cognitive
activity. We can expect continued progress in improving the technology and in
understanding interneuronal interactions. Whether these remarkable technological
developments can be extended to solving the special problems of cognitive
neuroscience is yet to be established. There has been a continued effort to
develop improved pMEAs since the idea of using photolithographic techniques
was suggested in the 1970s. (For a detailed discussion of current photolithography
technology, see DeWeerth, Meacham, Giuly, Guo, and Hochman, 2008.)
Electrode arrays of those and earlier times were plagued by problems such
as signal-to-noise ratios and tissue damage, not to mention that failure in achiev-
ing the intended raison d’etre for which these electrode arrays were being
constructed: understanding how neural networks operate. The simple mechanical
task of fitting all of the devices, amplifiers, conducting leads, and connections
 Microneuronal Network Theories 85

into a small region perplexed early investigators who attempted to construct

MEAs. The eternal problem of impedance matching between the microelectrodes
and the preamplifiers also continues to be a formidable obstacle.
Some of the specific challenges faced by engineers involved optimizing the
spacing between electrodes, increasing the number of electrodes, increasing the
sensitivity of the electrode-neuron interface, decreasing their toxicity, modifying
electrode interface impedances, increasing the vitality of the neurons recorded
by the MEA, and other such housekeeping functions. The work of Heer et al.
(2004) exemplified the state of the art about a decade ago. Special emphasis was
directed at two problems—improving the fabrication techniques and integrating
the various electronic components.
Because of the complexity and expense of the standard photolithographic process
used to construct pMEAs, a substantial amount of work is now being carried out
to produce cheap and reusable MEAs by such workers as Charkhkar, Knaack, Gnade,
Keefer, and Pancrazio (2012). Their technique deposited gold electrodes on a plastic
(polyethylene napthalate) substrate. Their electrodes were insulated from one another
by another plastic (parylene). These electrodes were robust and could maintain
cultured neurons for up to 30 days and then could be sterilized and reused. As
many as 64 gold-plated electrodes made up their version of an MEA.
Other researchers, such as Seker and his colleagues (2010), also concentrated
their attention on the coating materials used at the active points on the electrodes
in a pMEA. Their goal was to overcome poor impedance-matching conditions
at the electrode-neuron interface. Gold was a fine contact material but efforts
to use it were challenged by difficulty in making gold adhere to their electrodes.
Seker et al. discovered that a form of porous gold was able to partially overcome
this difficulty and not only adhered to the electrode much better than ordinary
gold but also lowered the impedance of the electrode by a factor of 25. Similar
reductions in electrode impedance and adherence were reported by Keefer, Bot-
terman, Romero, Rossi, and Gross (2008) for thin layers of carbon nanotubes.
This material is increasingly used today to increase electrode efficiency both in
terms of lowered impedance and improved adherence to the surface of the
microlithographically produced electrodes.
Because of the microscopic size of the apparatuses used in MEA experiments
it is often difficult to provide a stable chemical environment in a perfusion
chamber. Researchers such as Blake et al. (2010) have developed a micro-perfusion
chamber that permits the necessary fluids to permeate both sides of a slab or
slice of neural tissue, thus prolonging its viability.
Among the most fundamental issues faced by those who wish to use this
technology is that there is no a priori one-to-one mapping from neurons to
electrodes. Because of the irregular nature of the neuronal networks and the
nature of the interneuronal environment, some electrodes would detect the
activity of multiple neurons and some would detect nothing; a few would for-
tuitously record the activity of a single, identifiable neuron, but this is probably
86 Microneuronal Network Theories

the exception rather than the rule. As we see later, one of the ultimate goals of
MEA research is to develop analytical techniques to separate mixed neuronal
responses and associate them with individual source neurons.

3.3 Applications of Microelectrode Arrays

“Housekeeping” research projects such as those just discussed are necessary to
develop the MEA technology, a technology that makes many promises as the
next best empirical approach to ultimately understanding how neural networks
function. However, at present the state of the art in MEA technology has not
yet developed to the point that we are able to parse the centrally mediated,
cognitively related neural activities. Despite the fact that technological develop-
ment is moving apace, it is unlikely that our ability to unravel such neural
networks will soon be available simply because the relevant neuronal networks
are without doubt going to be much larger and complex than we have hitherto
been able to handle analytically.
Nevertheless, the next question we must ask is, beyond housekeeping and
technology-based research discussed in the previous sections, what kind of neu-
roscience application-oriented research has been carried out so far with these
remarkable tools? Although, as I noted earlier, the preponderance of the research
being carried out these days is technology-directed, the field is not barren and a
considerable amount of what we may appropriately call “pioneering” projects has
been carried out in various fields in which the MEAs have been applied to specific
neurophysiological problems. Accepting the constraint that few of these applica-
tions are even remotely cognitively related, most current applications deal with the
simple geometry of peripheral sensory coding mechanisms or overall field effects
in which information concerning the details of network organization is lost.
An outstanding and rare example of a cognitive neuroscience application of
MEAs is the work on a sensory problem carried out by a collaborative group
from several California universities (Litke et al., 2004; Shlens et al., 2009) who
studied excised monkey retinae in vitro. The retinae were laid on top of a planar
MEA consisting of 512 electrodes. To carry their experiments, it was also neces-
sary for them to fabricate integrated circuits consisting of all the amplifiers,
analog-to-digital converters, and multiplexers necessary to record from such a
large number of electrodes. Using this system, the receptive fields were initially
described by Litke and his colleagues for 364 identifiable neurons. The neurons
were classified according to their on-off characteristics.
This research (Shlens et al., 2009) used the California integrated hardware
system to study patterns of synchronized firing in an effort to explore the orga-
nization of the retinal networks. In particular, they observed that the functional
response of retinal ganglion neurons depended largely on their immediate neigh-
bors, and the retinal ganglion cells seemed to be arranged in a relatively simple
anatomical pattern in which connections were limited to nearby neighbors.
 Microneuronal Network Theories 87

Shlens et al. wisely concluded that their observation of the relatively simple
pattern of organization in the retina would probably not be generalizable to
more central brain structures. Citing the work of Yoshimura, Dantzker, and
Callaway (2005) on the nonrandom network properties of the multilayer orga-
nization of cortical columns, they suggested that their ability to work out the
detailed network interactions in a “simple” structure such as the retina may not
be possible in more central structures.
MEAs have also been useful in determining the toxicity of various chemicals
on neuronal activity by examining the overall diminishment of neuronal activity
(Steidl et al., 2006). In their experiment, excised slabs of hippocampal tissue were
perfused with antagonists that specifically inhibited synaptic transmitter activity.
MEAs have advantages over single electrodes in such an application because they
can measure system-wide effects in multicellular preparations that may not be obvi-
ous in single cell responses. Furthermore, the specific interconnections of the neurons
from which they recorded were unimportant in light of their goal to determine a
system-wide effect. A comprehensive and up-to-date discussion of the use of MEAs
in neurotoxicity evaluations can be found in Johnstone et al. (2010).
The search for properties of central neuronal networks has been limited so
far because of their complexity. Currently, in most cases, only the overall or
average properties of a network can be determined or the data made available
only in some other cumulative form. Although it is possible with current MEA
instrumentation to observe the activity of a number of neurons simultaneously,
it is far more difficult to identify the sequential dependencies and, thus, inter-
connections of their activities, and, therefore, the specifics of their network
organization.
One very germane example of this kind of research that highlights both the
advantages and disadvantages of MEAs is the determination of the bursting activity
of cultured neuronal tissue. It is in this study that the capabilities of the MEA
technique are currently most clearly demonstrated. It is also in this context that
the great weakness of applying the MEA technique system is also made clear.
Bursting activity of neurons has been observed for many years (e.g., see
Murphy, Blatter, Wier, and Baraban, 1992). It is particularly noticeable in in
vitro clusters of cultured neurons. However, it can also be observed in intact
slabs of neural tissue. Bursting seems to be a ubiquitous property of isolated
neural tissue and may be due to an absence of inhibitory stimuli from other
parts of the brain or, quite to the contrary, as a result of exogenous stimulation
(e.g., see Wagenaar, Madhavan, Pine, and Potter, 2005).
The problem with any measurement of the activity of multiple neurons with
MEAs is that it is difficult to determine the sequential order (or hierarchy of
firing) in which the multiply interconnected neurons respond (a problem high-
lighted by Hilgetag, O’Neill, and Young, 1996). This is especially true in in vitro
preparations in which virtually all of the neurons in a specimen may ultimately
respond simultaneously and spontaneously. This near universal activity negates
88 Microneuronal Network Theories

one of the most promising hopes of those who use MEAs—the expectation that
they will allow us to examine the significant interactions of a network of neu-
rons. As a result, although there have been some associations of bursting activity
with behavioral learning, no obvious relationships have been observed that have
plausibly related bursting activity to controlled stimulation. In fact, in one of
the most notable (for reporting a negative result) articles on this problem using
the components of an experiment we have just been discussing (MEAs, cultured
neurons, and bursting activity) Wagenaar, Pine, and Potter (2006) reported no
difference in the development of bursting activity when stimulated and when
spontaneous. What we are left with in such situations is a dearth of information
describing the detailed properties of neuronal connections. Others (e.g., see
Belykh, de Lange, and Hasler, 2005) argue that synchrony of bursting neurons
depends only on the general topology of the network rather than the specific
interneuronal interactions. Specifically, they suggest that the synchrony of burst-
ing brain tissue depends solely on the number of inputs to a neuron and not
the specific details of the network. This interpretation suggests that “bursting”
may be an abnormal endogenous condition that occurs when some threshold
number of neurons is active. The obvious analogy, of course, is to seizure activity
in intact brains.
The inescapable conclusion one must draw from the extreme instability
exhibited by the bursting phenomenon is that not only is the brain complex,
but also it is not stable over time. Brain networks seem to reorganize themselves
for mysterious reasons in a way that may preclude any notion that we may be
able to decode a consistent neuronal network, much less a stable experimental
environment. Psychologists well understand the variability of our thoughts. Any
experiment, therefore, that seeks to correlate cognitive and neurophysiological
findings may be shooting at a moving target in a way that makes the goal of
understanding the network of neurons that makes up cognition impossible.
A useful and practical recent application of MEAs has been their use as
stimulators in sensory prosthetic devices that serve as supplements to deficient
vision and hearing. Among the most successful applications of this technology
has been in the auditory system in the form of chronic cochlear implants. Among
the first to experiment with this idea in a surgical context were Djourno and
Eyries (1957). In their procedure, electrodes of the kind shown in Figure 3.6
were inserted through the oval window directly into the snail shell–shaped
cochlear space following along its curving course. At various points, an electrode
shorn of its insulation permits electrical stimulation of the remnants of the
cochlear nerve and any surviving hair cells.
Cochlear implants have been remarkably successful, as anyone who has spoken
to a person with such a device will quickly appreciate. Evaluated behaviorally,
the success of such a system is really remarkable and conversation can take place
almost normally. Clinical evaluations by several laboratories generally show 85%
sentence comprehension (see the review by Zeng, Rebscher, Harrison, Sun, and
 Microneuronal Network Theories 89

FIGURE 3.6 A prosthetic MEA consisting of approximately 20 electrodes used for

electrical stimulation of the cochlea.
Reproduced from Zeng et al. (2008) with permission.

Feng, 2008), a notable improvement over patients who lost hearing due to hair
cell and early acoustic nerve damage.
The high level of success with cochlear implants is largely due to the one-
dimensional and orderly spatial manner in which the various acoustic frequencies
are linearly encoded along the cochlea—a discovery for which Georg von Bekesy
(1899–1972) received the Nobel Prize in 1961.
MEAs such as those shown in Figure 3.7 are also being pursued experimen-
tally as a prosthesis for vision (for an up-to-date review see Luo and da Cruz,
90 Microneuronal Network Theories

FIGURE 3.7 A prosthetic MEA consisting of 60 electrodes used for electrical stimula-
tion of the retina.
Reproduced from Luo and da Cruz (2014) with permission.

2014). However, the ambitious goal of applying an array of microelectrodes to

meaningfully stimulate many retinal neurons and reproduce functional vision of
a quality comparable to the results achieved in hearing remains a yet to be
perceived task. Perceptually, the few patients who have had this experimental
treatment report “seeing” nothing more than simple configurations of spots of
light. The main problem in producing artificial vision is the complex anatomy
and two- and three-dimensional information coding used by the neural network
of the peripheral visual system. The design of the visual prosthesis is hampered
by the very small size of the retinal neurons and thus the mismatch between
the resolving power of the best microelectrode and that of the neurons in the
retina (Zrenner, 2002). Figure 3.7 is a current intraretinal MEA as used by Luo
and da Cruz (2014).
 Microneuronal Network Theories 91

3.4 Analysis Techniques

The bare fact that we can record the responses of a larger number of neurons
with an MEA than with a single electrode does not necessarily mean that useful
information about the organization of the great neuronal networks of the brain
can be extracted from those recordings. Similarly, the ability to establish the
simple interactions between pairs of neurons does not guarantee that the specific
polyneuronal codes for cognitive processes, whatever they may be, can also be
determined. Currently, there is a qualitative mismatch between what our experi-
mental techniques can accomplish in mapping a relatively few components of
neuronal networks and the number of neurons in the networks that must account
for consciousness, behavior selection, learning and adaptation, and all of the other
cognitive activities that psychologists have probed and measured over the years.
How far we can go in pursuing this approach—analyzing (i.e., mapping) the
organization of brain networks—to solving the mind–brain problem is unknown
at present. The development of increasingly powerful computer techniques leaves
the door open to some extraordinary new developments in the future. However,
there are harbingers in the current literature that the task is going to be far more
difficult, if not intractable, than it may seem at the present.
This section deals with some of the progress that has been made in analyzing
neuronal networks using MEAs. The word “analysis” conveys different meanings
to different investigators and, for this reason, it seems appropriate to consider
what an analysis means in the present context. In its most pristine form, a
complete analysis of a neuronal network (or, for that matter, any network) would
be a map identifying not only the momentary state of each individual neuron
involved in the network but also the determination of how each neuron is con-
nected to any other. Network connectivity, in this case, implies that there is a
causal (i.e., functional) relationship between specific neurons or groups of neurons;
in other words, that a given neuron’s state is influenced by many other neurons.
This influence may be either positive, exciting the neuron’s activity, or negative,
inhibiting the neuronal activity. It may be observed functionally or anatomically,
but in the final analysis anatomical connectivity must trump any inferred func-
tional connectivity. Given that each brain neuron may have hundreds of synaptic
junctions converging on it, serious problems arise concerning how one would
represent, much less construct, such a functional map and how one might empiri-
cally confirm either its anatomy or its function.
The suggestion is that the goal of mapping the neuronal networks underlying
Hebb’s conjecture in particular and cognitively relevant neural networks generally
is probably going to be an overwhelmingly formidable task. We can know in
principle what an idealized version of such a network would be like—the meta-
phorical superregister introduced earlier in this chapter. Connectivity can be
depicted for small networks by a simple map characterized by a collection of
nodes (the neurons) and by lines (the connections) indicating the interneuronal
92 Microneuronal Network Theories

–5

–20
N1 N2

–5
+15 +10

N3 N4
+10

FIGURE 3.8 A simplified “toy” network consisting of four neurons, N(1-4). Arrows
indicate connections between neurons and the numerals indicate the strength of
these connections. Although relatively simple, this network is probably not amenable
to prediction and solution.

connections. The valence and magnitude of the connections must also be rep-
resented by appropriate values if the network is to be fully characterized. Neurons
not directly interacting would simply not be connected by lines, although this
stratagem might obscure indirect routing of signals that would produce correlated
activity among even very remote neurons in the network.
It is also generally assumed that such a map would have to indicate in some
way that the constituent neurons would be dynamic, that is, capable of being
in different states at different times (otherwise the network would not be func-
tional, just locked in a constant state). The momentary state and the sequence
of states are additional properties that have to be specified if a network theory
of cognition is to be complete. Figure 3.8 is an example of the neuronal network
for a highly simplified “toy network.”
In this figure, signed numerical values indicate the potency of the intercon-
nections between neurons (negative values being inhibitory and positive vales
being excitatory) and the arrows indicate the direction of interconnections. At
each moment in time, we assume that the ensemble of states of all of the neurons
(Nn) in this hypothetical “toy” network would determine the network’s function;
at other times and other states of the component neurons, different functions
would be encoded by different configurations of this same network.
Extrapolating this basic idea of a “toy” map to the brain would constitute a
neuroreductionist theory in which the momentary states and interconnections
of the network would presumably be associated with a particular mental process.
Despite some limited progress with simple invertebrate nervous systems (where
real networks contain a small number of neurons), extrapolation of the network
concept to realistic vertebrate neural networks has not proven to be possible
currently. The best we can do is speculate about general properties (e.g., neurons
change their state as a result of changing synaptic efficacy).
It is also important to remember that the behavior of even as simple a toy
network as the four-neuron example shown in Figure 3.8 may itself be unpre-
dictable. Indeed, it is well known in mathematical circles that even a three-neuron
 Microneuronal Network Theories 93

network may create a situation that is the formal equivalent of the three-body
problem in physics2—general solutions to which have long been known to be
intractable under the conditions of feedback and interaction that are characteristic
of all kinds of networks—biological and physical.
If even approximate, partial versions of cognitively relevant network maps are to
remain the immediate goal, then it is necessary to answer a number of initial ques-
tions concerning the data obtained with the MEA technique by means of appropriate
analytical procedures. The first question is—which neurons are members of the
network being examined? Of necessity, this task is complicated by a number of
issues. Trying to carry out such a task by repetitive probing with a single micro-
electrode would be tedious and ultimately unsuccessful because of the uncertainty
that successive penetrations would make contact with the same neuron. Such an
approach also ignores the loss of vitality of the neurons over extended periods of
time. The great potential advantage of the MEA technique with its ability to
simultaneously record from many neurons is that it promises to answer this first
question directly, if not completely. However, just how far one can go with even
1,000 electrodes in identifying a set of interacting neurons remains problematic.
Next, consider that the extracellular potentials picked up with an MEA are
not simply associated on a one-to-one basis with a particular electrode. Not
only will the activity of individual neurons be detected by more than one elec-
trode, but also one electrode will pick up the activity of several neurons. This
raises the second basic analytical question—which of the recorded responses go
with which of the identified neurons? This task is known variously as “response
sorting” or “clustering”—the assignment of responses that may have been picked
up by multiple electrodes to their source neurons. This may require the parsing
and then assignment of responses from among the noisy mixture of responses
recorded by each electrode. (See Figure 3.2B for a realistic example of responses
picked up from an array of electrodes.)
The third analytical question that has to be answered is what are the specific
functional interconnections and neuron states of the putative network that cor-
respond in what is still a mysterious way to some cognitive or mental activity?
This search for the neuronal network configuration or code is the “holy grail”
of cognitive neuroscience. The complete specification of the relationships between
mind and brain at this level would constitute fulfillment of the quest for a solu-
tion to the mind–brain problem. Unfortunately, for practical and conceptual
reasons we now consider, this “grail” may remain as remote and unobtainable
as the mythical one that Arthur’s knights failed to find.

Spike Action Potential (SAP) Clustering

Consider the basic task of finding which of several extracellular responses picked
up by an MEA can be segregated, sorted, or clustered in such a way that they
can be associated with individual neurons. In some idealized sense, it seems that
94 Microneuronal Network Theories

it should be possible to perform an initial segregation of neuronal response by

concentrating on a single property of the neuron’s response; for example, the
shape of the spike action potential (SAP), its amplitude, or the interval between
SAPs. In the latter case—SAP interval patterns—the suggestion is that we might
be able to use periodic intervals as a clue to which responses go with which
neurons. The basic postulate is that responses that are separated by equal intervals
are the product of the same neuron. Unfortunately, if interval periodicity is the
only clue, mathematicians (Cox and Smith, 1953) have determined that clustering
of neuronal responses is extremely difficult to do because information about
interval periodicity is lost when the intervals from a number of neurons are
pooled. Mixing neuronal interval responses is known as superposition and the
Cox and Smith superposition principle asserts that if intervals are pooled, the
resulting pattern of intervals in the pooled output meets all of the criteria of
random sequences! Thus, periodicity alone cannot be used to determine which
responses go with which neurons, simply because measures of this dimension of
the response (interval) are lost when SAP intervals are mixed. Cox and Smith
(1953) proved that even if the responses of a particular neuron are perfectly
periodic (i.e., their inter-SAP–intervals were all equal to some constant value),
if one pooled the interval information from a number of neurons:

. . . except in degenerative cases, the result is indistinguishable from that

for a random series, no matter what the form of the individual outputs.
In particular if the number of sources and the individual recurrence times
are large the output will be random over intervals containing many events.
(p. 6)

In a subsequent article that provided biological relevance to what otherwise

might have seemed to be but a bit of mathematical esoterica, Cox and Smith
(1954) showed that this restriction holds specifically for motor nerve ending
response data collected by Fatt and Katz (1952). Their conclusion was that “there
is no evidence that the [pooled] series is not completely random” (p. 99). What
this means is that it is impossible to use residual interval periodicity alone to
assign SAPs to individual neuronal sources—the intervals promptly converge to
a random sequence for relatively small numbers of intervals.
There remain other questions about whether other single properties (such as
amplitude or response shape) can be used to accomplish the SAP sorting task.
However, an implication of the Cox and Smith work is that it is probably not
possible to sort neuronal responses on the basis of any other single parameter
except in the most reduced “degenerate” cases. For example, consider another
simple SAP sorting procedure—one based on amplitude. Given the most reduced
case in which a collection of SAPs have a spread of very different amplitudes
and the signal-to-noise ratios are not themselves overwhelming, a simple measure
of the amplitudes of recorded extracellular SAPs could conceivably be used to
 Microneuronal Network Theories 95

segregate a few responses as belonging to the outputs of a small number of

individual neurons. Such a criterion might well work for a few neurons; however,
should the signal-to-noise ratio be such that the amplitude of the noise is close
to the amplitude of the signals, any such simple criterion is doomed to failure.
Neuronal variability would quickly overwhelm any ability on the part of the
analysis technique to assign detected SAPs to specific source neurons on the
basis of the single parameter of amplitude.
Because it is unlikely that simple criteria such as SAP interval pattern or
amplitude will work for more than a few neurons, other more advanced process-
ing techniques are necessary to solve the sorting problem. Several more sophis-
ticated analysis techniques have been suggested that involve measuring the shape
of the SAPs—a technique that also depends on relatively noise-free conditions.
In this technique, the shape of each of the recorded SAPs is digitized by an
analog-to-digital converter. The resulting tables of numbers representing the
shapes of the recorded SAPs are then compared to determine if they are similar
enough to be assigned to the same source neuron.
Shape recognition of SAPs and their assignment to source neurons can, in
principle, be carried out by a matching procedure in which the shape of a SAP
is compared with the shape of a prototypical “template” response shape. The
template may be a sparse one (a “window discriminator”) in which only a few
key elements from a template are used or a more complete one (a dense version—
a “vector” discriminator) consisting of all of the digital values. Characterizing
the SAP straightforward convolutional techniques may then be computed between
the digital samples representing the SAP and those of the template; correlational
values above some threshold are then considered to designate SAPs with com-
mon neuronal origin. This template-matching technique was in use as far back
as the 1960s by such workers as Gerstein and Clark (1964) and continues to
serve as the basis of much of the modern work in SAP sorting.
The SAP clustering task becomes increasingly difficult when multiple elec-
trodes are used to study multiple neuronal responses. However, several advantages
are obtained as the number of electrodes involved increases. For example, simul-
taneity of response can serve as a cue that the response recorded at one electrode
is the same as one recorded at another electrode. In this case, any idiosyncrasies
resulting from interneuronal distances become irrelevant. It is also possible, in
principle, to begin to study the spatial arrangement of neurons when using
multiple electrodes since response size is known to be strongly dependent on
the distance between an electrode and a neuron. Unfortunately, all such tech-
niques eventually fail at relatively modest numbers of neurons.
Other analytical procedures to assign or sort a mélange of SAPs to individual
neurons have also used familiar statistical methods (e.g., Principle Component
Analysis, Adamos, Kosmidis, and Neophilidis, 2008), new statistical techniques
(e.g., Continuous Basis Pursuit, Ekanadham, Tranchina, and Simoncelli, 2014),
or novel techniques such as wavelet transforms (e.g., Quiroga, Nadasdy, and
96 Microneuronal Network Theories

Ben-Shaul, 2004)—a method comparable to Fourier analysis but using different

basis functions. Others who have tackled the problem with new techniques
include Pillow, Shlens, Chichilnisky, and Simoncelli (2013). Their system, which
they have designated as “binary pursuit,” depends on a model of SAP shape that
they claim improves upon sorting produced by the less effective earlier methods.
A still timely review of the many published SAP clustering techniques has been
presented by Lewicki (1998). This article, along with more up-to-date discussions
of the topic (e.g., Brown, Kass, and Mitra, 2004; Wild, Prekopcsak, Sieger, Novak,
and Jech, 2012) provide useful tutorials on the efforts being made in the devel-
opment of techniques for sorting and clustering SAPs.
As formidable as these SAP sorting techniques appear to be at first glance, it
is clear that there are many difficulties in applying them. Some of these difficul-
ties are intrinsic to the problem (e.g., the loss of information that occurs as a
result of pooling interval data) whereas others are matters of practical details
(e.g., electrode movement induced variations in response properties occurring
over time). Matters of mechanical and electronic technique are always subject
to improvement and further developments are likely to overcome many of the
technical instrumentation problems in the near future. However, some problems
intrinsic to both the analysis methods and biological variability are probably
unavoidable and not likely to be solved soon. It is a good idea to distinguish
what the fundamental limits are and avoid wasting time and resources attempting
to make the intractable tractable.
Lewicki (1998) summarized some of the difficulties that inhibit solutions of
the SAP sorting problem. I paraphrase and update his list in the following
comments:

1. Lack of independence of the neuronal responses. If neurons are intercon-

nected in a way that their responses are not independent but highly correlated
with each other (such that neuronal response X always leads to neuronal
response Y), then the sorting problem becomes extremely difficult, if not
impossible.
2. Nonlinear mixing of the recorded responses may distort various attributes of
the SAP signal, including shape and the respective amplitudes of responses
occurring at different distances from the recording electrode.
3. Lewicki especially emphasized the fact that any technique for SAP sort-
ing that used the shape of the extracellular response is going to be severely
degraded if the shapes are not constant.
4. As we have seen, neurons in vitro tend to produce spontaneous bursting
activity—a response pattern that is likely to distort the shape of the SAP
by exhaustion of its metabolic and ionic resources, thus producing various
degrees of neuronal refractoriness.
5. Similarly, we can add that variability in virtually any other property that an
experimenter might use to sort SAPs onto neurons will tend to make the
 Microneuronal Network Theories 97

analysis quickly fail if the property is neither linear nor stable. Pillow et al.
(2013) have been especially critical in noting the deleterious effect of overlap-
ping SAPs on spike clustering techniques.
6. Simple mechanical and crowding problems of the ancillary signal recording
and transmission equipment are a continuing problem. Neurons are micro-
scopic entities; throughout the history of microelectrode use, both single and
when enhanced by MEAs, investigators have had to fight the battle of making
their instruments small enough so that they do not interfere with the target
neuron mechanically. Furthermore, just getting the necessary parts of the
recording system into the container holding the neuronal tissue remains a
continuing engineering challenge.
7. Similarly, any nonrigid electrode configuration that is free to move about can
lead to what Lewicki referred to as “electrode drift”—gradual changes in the
spatial relationships of neurons and electrodes. This will inevitably lead to
changes in the shape of the recorded responses and inhibit SAP sorting and
clustering. Stabilization of electrodes has been greatly helped by the applica-
tion of photolithographic techniques from computer fabrication engineering.
Indeed, in most in vitro experiments, printed electrodes are perfectly stabilized
in the containment well in which the (typically dissociated) neural tissue is
placed. Unfortunately, because of tissue flexibility, stabilizing the relative posi-
tion of the neurons and the electrodes either in vivo or in vitro experiments
is still an unsolved, but serious, problem. Thus, the spatial arrangement of the
interneuronal network is likely to fluctuate. Although some progress has been
made in “caging” single neurons near single electrodes (e.g., Erickson, Tooker,
Tai, and Pine, 2008), the problem remains a real one for those who wish to
study neuronal populations over prolonged periods of time.
8. Lewicki also pointed out that with any communication system, noise is
always recorded superimposed upon the signal information of main interest.
Depending on the signal-to-noise ratio, critical information may be fully lost
if this ratio is too small. Even worse, if the noise is not stable, changing prop-
erties from moment to moment would make the analysis process especially
challenging. Indeed, unstable noise may completely obliterate any possibility
of sorting SAPs using any of the measures mentioned here.
(Paraphrased from Lewicki, 1998)

The Futile Search for Connectivity Maps

Spike action potential sorting or clustering (assigning a SAP to a particular neuron)
is not the only important task when one is attempting to study neuronal networks.
Another major analytical goal is determining the connectivity or topological
organization of a neural network. In other words, we must also ask—what are
functional connections that permit one neuron’s activities to drive or cause
another’s activity? Specifically, it is to determine how activity in Neuron X
98 Microneuronal Network Theories

stimulates or causes activity in Neuron Y and ultimately to represent the orga-

nization of a network of many neurons that may account for a cognitive process.
This is a far more demanding task than sorting SAPs to particular neurons simply
because it requires a robust solution to the sorting problem before the connectiv-
ity issue can be attacked and solved. However, we must be repeatedly forewarned
that beyond a very few neurons (a number that may be as small as three) deter-
mining the organization of a neuronal network or characterizing its neuronal
connections may not only be difficult but impossible.
The problem is exacerbated by a vague, indeed, sloppy use of the word con-
nectivity. It is not entirely clear that different authors mean the same thing when
they use this ambiguous term. In an article that deserves much more attention
than it has so far received, Horwitz (2003) pointed out that brain “connectivity”
has a variety of different meanings. He distinguished between three main
meanings—anatomical, functional, and effective connectivity. Anatomical con-
nectivity is conceptually clear enough (and, when available, it is the “truth”
standard against which the other two must be judged). Anatomical connectivity
is what Sporns, Tononi, and Kotter (2005) referred to as the “connectome”—“a
comprehensive structural description of the network of elements and connections
forming the human brain” (p. 245).3
Horwitz defined functional connectivity as “the temporal coherence among
the activity of different neurons . . . measured by cross-correlating their spike
trains” (p. 467). The cross correlation measure was supposed to indicate the
strength of an observed relationship between neurons.
Horwitz went on to assign a different meaning of connectivity in the third
context—effectiveness. Effective connectedness is “the simplest neuron-like circuit
that would produce the same temporal relationship as observed experimentally”
(p. 467). From one point of view, this is the goal of all of the analytical methods
that are described in this section—the actual map of the effective interconnections.
He also made an exceedingly important statement at this point in his discussion.
He argued that assaying the effective connectivity is severely limited because the
neuronal network responses are underdetermined by the kinds of measurements
we are able to make. That is, there are many possible neuronal networks that can
produce the same functional neuronal activity pattern, and available experimental
data can neither validate plausible nor reject improbable ones.
A similar vagueness permeates the nature of the word “complexity.” Like its
partner “connectivity,” this word has different meanings to different authors and
lacks a precise definition in the cognitive neurosciences. This makes it difficult
to compare networks (although precise numerical values at the level of complex-
ity with which microneuronal brain researchers must deal probably do not make
any practical difference). Nevertheless, like the huge numbers associated with
factorials or Bell’s numbers, it would be worthwhile if such a numerical value
for complexity existed. Unfortunately, mathematicians also disagree on a defini-
tion of complexity. It is often approximated by neuroscientists as a simple estimate
 Microneuronal Network Theories 99

of the number of neurons and synaptic connections in neural tissue, but this
misses the whole point.
Network complexity, therefore, is viewed as a property of systems with numer-
ous parts and inter-part connections. Measures such as the factorial can provide
estimates of “complexity,” but, in general, the utility of these numbers collapses
as the number of components and interactions increases and numbers such as
the factorial soar to hypercosmological levels. For computer scientists, the com-
plexity of a problem has to be computable in order to lead to criteria for the
solvability or convergence of certain algorithms. For cognitive neuroscientists
operating at the microneuronal network level, almost everything beyond three
neurons is, for all practical purposes, “infinitely” complex.
The reason for this uncertainty forewarns us about one of the most funda-
mental limitations of any technique that seeks to provide an effective map of a
neuronal network. It is something that is likely to raise the hackles of those who
seek to represent the structure of a neuronal network but it is essential that those
who seek to use MEAs to disentangle neuronal networks understand its con-
notation, if not its denotation. Horwitz (2003) went on to say:

A key experimental issue that shaped these definitions originated in the

near-impossibility of knowing the exact anatomical relationship between
a neuronal pair whose functional interactivity was being accessed (at least
in mammalian experimental preparations).
(p. 467)

Should this be true for a pair of neurons, how much less likely is achieving
the goal of mapping the “functional interactivity” of a suitably sized collection
of cognitively significant neurons? Thus, the quest for a unique effective map
may always be a search for a chimaera; the best we will be able to do will be
to develop a functional description—effective maps being denied to us.
Despite these musings, work continues to determine the organization of
realistic neuronal networks—mostly with invertebrate model preparations. It is
worthwhile reiterating that although some progress has been made in response
clustering of simple preparations, the grand promise of producing a solution to
the general network problem (mapping or decoding the specific neuronal net-
work for a particular cognitive process) is still unfulfilled and nothing on the
horizon is promising.
An interesting place to start a discussion of this kind is with a work that can
only be described as a heroic example of a class project. This work is presented
here because of its pedagogical simplicity and conceptual clarity. It is unlikely
that this approach could prove to be an efficient procedure for mapping real
neuronal networks, but nowhere are the methodological and logical assumptions
behind neuronal network construction made clearer and the inevitable failure
of the “holy grail quest” highlighted.
100 Microneuronal Network Theories

Two students at MIT, Jonathan Karr and Daniel Herman, authored4 a labora-
tory semester project in which indications of microneuronal network connectivity
were determined by the following method. Their experiment5 used data recorded
from an in vitro culture of hippocampal neurons using an 8 × 8 MEA. Each
of the 64 electrodes was used in turn as a stimulator and the responses induced
in all of the other 63 electrodes measured. Correlations were then individually
computed between the stimulus applied to the single stimulator and the responses
measured at all of the other responding electrodes—a selection that might be
fewer than 63 because not all of the electrodes picked up a neuronal response.
This produced a maximum of 4,032 correlations (64 × 64 minus the 64 auto-
correlations of the stimulator with itself). The amplitudes of the correlations
were interpreted to be measures of the degree of interconnectivity among the
electrodes; high correlations, driven by high amplitudes and short latencies, being
tantamount to intimate causal interconnections and spatial propinquity between
the stimulating and the responding neurons.
Karr and Herman’s brief report is exceptional not only because of its successes
but also because of the way it highlighted some of the limitations of the whole
idea of using MEAs to determine the connectivity of a network of neurons.
On the positive side, their technique embodied the ability to provide initial
estimates of a number of the connectivity parameters of the hippocampal neu-
ronal network they were studying. Based on the assumption that the probability
of an inferred connection would be higher for a correlated pair of electrodes
than for an unstimulated pair and such other factors as the delay between the
stimulus and the response, preliminary estimates could, in principle, be made of
factors such as the geometry of connection (e.g., how far apart the stimulated
and the responding neurons were either spatially or in terms of intervening
neurons) and the interconnection strength (i.e., the magnitude of the correlation
between the single stimulating and the other 63 responding electrodes).
Despite the unusual accomplishments of these students, their work also dem-
onstrated the extreme difficulty of the task of determining the connectivity of
a neural network with MEAs and the most powerful analytical techniques avail-
able. For example, there is a cryptic assumption in experiments of this kind that
an ideal, pseudoregular neural network underlies these correlations. Neuroana-
tomically, however, the irregular geometry of a real neuronal network does not
even come close to that of the regular 8 × 8 array of the microelectrodes used
to collect these data. Collectively, the real neuronal networks are much more
irregular. Whereas a realistically complex neuronal network may have dozens, if
not hundreds, of synaptic connections between pairs of neurons and many
thousands on each neuron, the analytically derived network implicitly assumes
a much simpler kind of network based on a much simpler concept of connec-
tivity than is actually present.
To exacerbate this situation even further, as I noted earlier, each extracellular
SAP may be detected by more than one electrode and that one electrode may
 Microneuronal Network Theories 101

FIGURE 3.9 The effect of arbitrary thresholds on the connectivity pattern of a neural
network.
From Karr and Herman (2004) with permission.

record activity from many neurons. Therefore, even under the best conditions,
analyses with even the most limited goals are constrained to deal with small
numbers of neurons and unverifiable ambiguities.
Another issue is the arbitrariness of the interconnectivity networks generated
by methods like those reported by Karr and Herman. The main measure used
in their analysis—correlations—as a measure of interconnectivity can be instanti-
ated in a large number of different maps depending on arbitrary decisions. If a
low threshold for a correlation is used as a criterion for interconnectivity, the
resulting pattern would suggest a large number of putative links. On the other
hand, if a high threshold was used, only a few links between neurons would be
generated, as shown in Figure 3.9. The essential problem is the arbitrariness of
the threshold for what would be acceptable as a connection.
Nevertheless, the importance of Karr and Herman’s laboratory exercise went
far beyond the preliminary nature of their accomplishment. Others (e.g., Maccione
et al., 2012) also subsequently reported the use of correlational techniques similar
to those used by these two students. Whereas Karr and Herman stimulated the
electrodes to define correlated responses, Maccione and his colleagues allowed
spontaneous activity to define causal functional relationships and thus to imply
interneuronal connections. In addition, Maccione et al. also used filtering techniques
to reduce the number of spurious connections—combinations of distance and
latency being used to eliminate at least some of the implausible connections.
In recent years a number of other more sophisticated schemes have been devel-
oped that purport to determine the connectivity pattern of a neuronal network
by means of MEAs and novel analytical techniques. One of the most comprehensive
is the work of Stevenson et al. (2009). This group applied Bayesian methods to
determine the statistical properties of neuronal networks in both toy and in vivo
preparations of monkey brain neurons. Their goal was to characterize the functional
connectivity of networks in the form of a matrix in which causal relations between
pairs of neurons were either present or not. Their data were presented in the form
of a matrix that showed which neurons were interacting with which others. Scores
were binary values of interactions between all possible pairs of neurons in a man-
ner comparable to the correlational scores reported by Karr and Herman. However,
neither group of researchers was able to produce a connectivity map of the general
type desired nor could they distinguish which interconnections were monosynaptic
and which were the result of indirect pathways.
102 Microneuronal Network Theories

To accomplish what they did, Stevenson and his colleagues had to make
assumptions about the nature of the neuronal networks they were studying.
These simplifying assumptions (“priors” in Bayesian terms) made it possible to
avoid a combinatorial explosion, but also limited how detailed the inferred
networks could be. Their priors included, first, the assumption that neuronal
interconnections are sparse, a generalization that almost everyone agrees is not
anatomically true. Second, they also assumed that neuronal interactions are tem-
porally smooth, that is, that the interactions occur without abrupt discontinuities
in time. This assumption is also probably incorrect.
In summary, we now list some of the difficulties and challenges in reported
attempts to determine the connectivity of a neuronal network:

1. At the base of many of the other problems faced by those who attempt to
map out a realistic neuronal network is the fundamental underdetermina-
tion of the data obtained with an MEA. That is, it cannot be definitively
determined by any analytical method yet proposed which of several (many?)
equally plausible network configurations is responsible for a functional
response. Neuron X may be connected to Neuron Y by a single direct syn-
apse or by an indirect pathway that may involve a number of neurons. Any
pair of neurons may be interconnected by a multitude of alternative network
connections. Simply demonstrating a correlation does not adequately define
a specific anatomical connection. Indeed, there is a strong possibility that it
cannot, in principle, do so.
2. There is a discrepancy between neuronal networks determined electrophysi-
ologically and those produced by analytical methods. It is often necessary
to assume that there is a sparse pattern of connectivity (often approaching
one-to-one connections) to avoid generating intractable combinatorial prob-
lems. In anatomical fact, however, all of our microscopic evidence suggests
that each neuron is complexly and multiply interconnected with many other
neurons, by as many as 1,000 synapses.
3. Unlike the neat grid-like arrangements of MEAs, neurons are irregular and
idiosyncratic in both their shape and connectivity.
4. In many cases, real neuronal networks are influenced by neuronal activity
from outside the network under study. Thus, both centrifugal and centripetal
sensory or motor activity may be influencing what was originally thought to
be a network that could be studied in isolation.
5. Almost all of the current methods are applied in experiments that, of neces-
sity, report the use of relatively small numbers of neurons in their analyses.
The availability of the MEA suggests, erroneously I believe, that this is no
longer an issue since the activity of large numbers of neurons can be recorded
simultaneously. However, a careful scrutiny of the literature suggests that the
interconnectivity pattern of only a few neurons has actually been determined.
In short, 1,000 electrodes in an MEA does not mean that the interactions
 Microneuronal Network Theories 103

among 1,000 neurons can be determined. In practice, the neuronal networks

reported in all research I have encountered so far are sparse and few. In gen-
eral, the MEA technique has largely been applied to a few cultured, but disas-
sociated, neurons in vitro; in far fewer experiments have any attempts been
made to map neuronal networks. The basic reason for this constraint is the
influence of biological combinatorial complexity, a topic to be expanded
upon in the next section. For all practical purposes we are still constrained
to work at the level of a very few neurons, usually in the nervous systems of
invertebrate model preparations. This approach is discussed in Chapter 4.

3.5 Combinatorics of Complexity

Given the idiosyncratic nature of its interconnections and the large number of
its components, an argument can be made that the brain is the most complex
bounded entity in the universe. Regardless of which level of analysis one is
considering, the combinatorial complexity of the brain stresses any putative
microneuronal explanation of its function far beyond what we may predict are
the limits of any modern or prospective science. The basic conclusion we must
infer from experiments such as those just discussed is that it takes very few
interacting objects or events to pose an intractable combinatoric problem. How-
ever, the nature, extent, and influence of neural network complexity are not a
part of the intuition of many cognitive neuroscientists. The goal of this section
is to fill that gap by introducing complexity and the limiting role it plays in
cognitive neuroscience.
Early on, Stockmeyer and Chandra (1979), among many others, showed how
relatively simple network problems (e.g., minimizing a traveling salesman’s itiner-
ary) could require hyper-astronomical processing times. For example, estimates
suggest that if the salesman had as few as 21 stops, an exhaustive solution to the
problem would require 77,000 years of computer calculation time.6 Meyer (1974)
proposes a general theorem that forcibly makes this point:

THEOREM. If we choose sentences [binary strings] of length 616 . . .

and code these sentences into 6 × 16 = 3696 binary digits, than any logi-
cal network with 3696 inputs which decides truth of these sentences
contains at least 10123 operations. . . . We remind the reader that the radius
of a proton is approximately 10− 13 cm, and the radius of the known uni-
verse is approximately 1028 cm. Thus, for sentences of length 616, a network
whose atomic operations were performed by transistors the size of a proton
connected by infinitely thin wires would densely fill the entire universe.
(p. 481)

Similar calculations can be made to show that there are many other problems
that cannot be solved in the lifetime of the universe. Therefore, neither time
104 Microneuronal Network Theories

nor space is adequate to deal with the complexity of even relatively (superficially)
simple problems in the lifetime of the universe.
Karp (1986), a prominent complexity theorist, also reported how many other
superficially simple combinatorial problems, many of which were analogs of
those faced in cognitive neuroscience, were actually known to be intractable in
their demands on time and space. Karp’s (1986) work on the intractability of
many other superficially simple combinatorial problems is especially relevant in
this context. Specifically, he pointed out that many, if not most, real-world
problems are said to be NP-complete, that is, they cannot be solved in real
amounts of time. Although the work of Karp and others is well known in
mathematics and computer theory, cognitive neuroscientists seem to be assuming
that some of their greatest challenges are in principle solvable when it has already
been proven that they are intractable.
Despite these reminders, there has been an implicit acceptance of the idea
that the mind–brain problem will ultimately be resolvable at the microneuronal
network level. (See the discussion in Chapter 5.) In this section, I consider some
basic arithmetical background suggesting that any attempt to develop a con-
nectionist model of anything approaching a cognitively significant neuronal
network is not a feasible experimental goal.
By far the most profound challenge facing those who seek to determine the
functional topology of real neuronal networks is what can be designated as
“combinatorial or computational complexity.” The general term “complexity”—
a phrase thrown about by cognitive neurophysiologists with total abandon and
freedom from any precise definition of the term—is heard in virtually every
conversation about the brain. The term, regardless of how deeply we understand
its numerical implications, certainly has profoundly affected our research by both
explicitly and implicitly limiting our research efforts to toy networks with only
a few neurons or to macroneural networks.
The constraint that inhibits progress in understanding network theories of all
kinds is how large the number of alternative organizations of even a very small
number of interconnected components can be. The basic arithmetical terms that
are germane to the topic of complexity are combinations, factorials, and permuta-
tions. These terms are relevant to neuronal networks because the number of
alternative, yet equally plausible, organizations of a very few components or nodes
can very quickly rise to huge numbers.
Factorials (n!) are particularly relevant to neural networks because they cor-
respond to the number of different ways that n objects (e.g., neurons) can be
arranged (e.g., interconnected) into different groupings. The sum of all of the
possible maps that are necessary to fully characterize a network quickly becomes
enormous. The factorial n! of a number n is defined as the multiplicative product
of all of the positive integers that are less than or equal to n. Even for small
numbers, surprisingly large factorials are generated. For example, although 3! is
equal to 6, 10! is equal to 3,628,800. Larger factorials (for example, 25! is equal
 Microneuronal Network Theories 105

to 1.551121004 × 1025) very quickly approximate and then exceed the number
of basic particles in the universe. This numerical explosion provides a good idea
of what happens even when one studies relatively simple networks.
Combination (C) is a more specific term that is defined in terms of the
number of ways that a particular subset of objects can be arranged. The formula
for combinations of n things taken r at a time is:

n!
C=
(n − r )!(r !)  Eqn. 3.1

C can also get very large very quickly as n grows. Consider the classical example
of playing ordinary 5-card poker for which n = 52 and r = 5. Applying Eqn. 3.1
specifies that there are 2,598,960 such combinations if one ignores the order in
which the items are selected. Considering how much larger n and r can be in
the brain, the number of coded neuronal combinations is also enormous.
If the order of the ways that a set of things are combined matters, then the
word “combination” has a more restricted meaning—it becomes what is called
a permutation. Because the particular way in which a set of neurons is function-
ally arranged would presumably lead to different cognitive processes, it is probably
more correct to speak about changes in brain states as permutations rather than
combinations.
Furthermore, this is not the worst case. There are other mathematical series
that grow even faster than factorials or combinations—e.g., superfactorials—that
although not a priori irrelevant to this discussion help to give us an intuition
about brain complexity. A superfactorial of n is the product of the first n fac-
torials. All such “super” functions make the same point: exhaustively studying
networks or seeking to identify the unique arrangement that encodes a cognitive
process is probably not currently a feasible or plausible approach to generally
solving the riddle of how the brain makes the mind. Although we may be able
to make some statements about the general properties of such networks, it is
obviously impractical to attempt exhaustive experiments to understand their
function. This idea of factorials as measures of possible organizations exceeds by
far the already serious problems generated by the simple numerousness of the
neurons in the brain.
The link of these esoteric mathematical concepts to neuronal networks is
clear. Even if we were able to represent the brain as a system of billions of
uniform and simply interconnected “billiard balls” it would underestimate the
true complexity of the brain. Neurons are not simple entities as are billiard balls,
but are themselves complex entities with many idiosyncratic properties and a
variety of interconnections of varying weight.
Koch (2012) recently put this problem into perspective when he calculated
that the time required to “exhaustively describe a [neural] system” (and this is
what we assume we have to do to explain how the brain generates the mind)
106 Microneuronal Network Theories

would grow faster than exponentially. As an example, Koch calculated that the
2 million neurons of the visual cortex of a mouse would require a computer
system running for 10 million years even if the computer’s powers were simul-
taneously growing at the modern rate suggested by Moore’s Law (“computer
component density and thus computational speed doubles roughly every 2 years”).
Koch did go on to note that there were some simplifying conditions (compa-
rable to those used to speed up a standard Fourier transform to a “fast” FFT
version) that could speed up the process by a few orders of magnitude. One
way to do this is by grouping neurons into clusters. However, whenever one
regularizes a procedure in this manner, there is the ever-present possibility that
the cognitively relevant microneuronal organizational properties one is seeking
may be tossed out.
Koch (2012) also reminded us of another measure of complexity—Bell’s
number—defined as the number of different partitions into which a system
of n objects can be divided. For example Bell’s number for n = 10 is 115,975.
Playing out its calculations serves as another reminder of the complex ways
in which the components of a network can interact or be divided. A Bell’s
number of 115,975 compares to factorial 10 which is equal to 3,628,800.
Thus, factorials grow faster than Bell’s numbers (and both scale faster than
exponentials) but the effect is the same—huge eventual demands for analytical
speed if we are ever to exhaustively study the brain as an example of a neu-
ronal network.
Whatever the best measure, they all pale into insignificance compared with
the actual number of neurons in the human brain, and much of current think-
ing is that almost all of them participate in encoding cognition. The only
hope, according to Koch, of ameliorating this constraint of “colossal” numbers
of calculations is to find some way to reduce the number of subunits so that
the number of possible partitions can be reduced. However, the irregular nature
of most brain tissue (notwithstanding the discovery of such prototypes as
repetitive columnar organization) makes it likely that we will join such alter-
native mathematical treatments as “graph theory” in attending not to the
specific problems of cognitively realistic networks but rather to the general
properties such as reachability, directedness, and degree of toy networks. As
interesting as the work on graph theory of Harary (1969) or Christofides
(1975) may be, little attention is paid by graph theorists to predicting the
function of complex graphs or the differential effects of different partitions.
Their attention is drawn to general properties of very simple graphs that may
extrapolate in some ill-defined way to larger networks. This decision by graph
theorists to concentrate on simplified abstract models is well chosen since, as
noted earlier, even the three-body (or, in this case, the three-neuron problem)
is probably not generally solvable.
“Complexity” has other ramifications beyond simply the raw numerousness
of the nodes and connections on which I have concentrated so far in this
 Microneuronal Network Theories 107

section. Another aspect of complexity may be summed up in the term “noise.”

Noise, in the context of neuronal networks, refers to all of the sources of uncon-
trolled variability that tend to distort or destabilize response measures. This adds
a level of temporal variability to the spatial kind of studies described in earlier
sections of this chapter. It is difficult enough to deal with the topological (i.e.,
spatial) complexity of stable neuronal networks; how much more difficult it
would be to deal with functional networks that are constantly changing their
states and yet sometimes emitting the same behavior.
Obviously, for a host of procedural and instrumentation reasons, the experi-
ments necessary to establish the cognitive role of a specific neuronal network
are probably impossible to carry out. Experiments of this general kind, studying
particular cognitively significant networks, make impossible demands on our
technology, our mathematics, and our computer capability, not to mention the
conceptual foundations underlying this approach. It is likely that we will never
be able either to manipulate neuronal network states or measure the subtle dif-
ferences between the aggregate neuronal states representing different cognitive
states. Yet, this is the kind of experiment that would have to be carried out to
“decode” or “map” brain representations of cognitive states. Proposed experi-
ments at this level are at best Gedanken experiments that can be discussed ad
infinitum but never consummated.
Others, such as Lichtman and Denk (2011), have also argued that any proposed
attempt to study the brain at the microscopic level of size and, thus, at the
enormous numbers of neurons, is likely to be beyond both our instrumentation
technology and the necessary analytical capability of current mathematics. Their
reasons include (1) the immense diversity of cell types in the brain; (2) the
diversity of cellular responses in both space and time; (3) the scale difference
between microscopic neurons and their extent over vast macroscopic regions of
the brain (the brain must be studied “over sizes that span six orders of magni-
tude” [p. 620]); and (4) the overwhelming flood of data at the microneuronal
level. Lichtman and Denk concluded that:

There is also no other organ system for which the complexity of the
structure is so great that earnest arguments can be made against delving
into structural analysis because such an effort might well provide an
unprecedentedly gigantic, yet totally incomprehensible, mass of data.
(p. 618)

Although Lichtman and Denk express optimism that the task of mapping
neuronal circuits may be tractable sometime in the future, as of this date I have
not been able to find any report of a cognitive (i.e., something beyond a periph-
eral motor or sensory mechanism) linkage to vertebrate microneuronal networks
that meets the criteria expressed by Martin, Grimwood, and Morris for robust
empirical proof at the microneuronal level.
108 Microneuronal Network Theories

3.6 Interim Conclusions

The major conclusion of this chapter is that there exists additional profoundly
compelling, if not overwhelming, evidence that suggests that the search for the
specific microneuronal networks that represent or encode cognitive processes in
the vertebrate brain cannot be fulfilled with current technology or concepts.
This does not mean, however, that there will not be wonderful and amazing
progress in determining how individual neurons or particular parts of the brain
may be involved in particular activities or unraveling the mysteries of the anatomy
and physiology. The extraordinary new work of Nabavi and his colleagues being
a wonderful example of how we may be able to learn some things (e.g., the
general proposition that changes in synaptic conductivity encode memories)
while others (e.g., the specific synaptic pattern differences that encode different
memories) remain obscure. There are many other examples of what would have
seemed to be impossible only a few years ago; the accomplishments in under-
standing the ionic forces that explain how individual neurons work and interact
at the cellular and molecular levels are among the older crown jewels of neu-
rophysiology. So, too, are the gross anatomical studies of the macroconnections
within the brain that have been made possible by the development of devices
such as the diffusion tensor MRI (Jones, Simmons, Williams, and Horsfield, 1999).
However, the specific mapping of the microneuronal networks that most cogni-
tive neuroscientists would argue logically are the “psychoneural equivalents” of
cognition remain elusive, if not absolutely inaccessible.
In this chapter, I have concentrated on the role of networks in current think-
ing, both in terms of available technology and our conceptual underpinnings.
The most advanced technological development that may be of use in studying
networks has been progress in multiple electrodes arrays. Unfortunately, there
remain many obstacles to the use of arrays of MEAs to record or stimulate the
activity of many neurons. These include:

1. Microelectrode arrays of this kind characteristically consist of electrode tips

that are larger (60 μm) than single microelectrodes (1 μm). Therefore, they are
limited to recording only extracellular responses.
2. There are two tasks to be accomplished for which the MEA initially seemed
well suited: (1) the assignment of recorded responses to their neuron of
origin, that is, the clustering or sorting problem, and (2) the determination of
the neuronal map indicating which neurons are involved and what are their
interconnections. However, there are serious impediments to accomplishing
either of these analytical tasks.
3. The amplitude and other properties of recorded neuronal responses will
vary depending on the fortuitous spatial relationships between neurons and
electrodes—relationships that cannot be predicted in advance.
4. It is, therefore, extremely difficult to definitively determine which responses
are coming from which neurons with current technology. Different neurons
 Microneuronal Network Theories 109

may display the same shapes and amplitudes and the same neuron may be
characteristically different from one trial to the next. No single discriminat-
ing criterion for solving the clustering problem works in a general way.
5. Recording is fortuitous: Some electrodes will record nothing and some will
pick up the responses from multiple neurons in an unpredictable manner.
6. The regular pattern of an array of 100 such microelectrodes (e.g., as shown
in Figure 3.2B) does not necessarily correspond to the spatial arrangement
of the neurons in the neural tissue under investigation. That is, beyond the
sensory and motor representations, they may not be topologically consistent.
Therefore, it may not be possible to map directly from electrode recordings to
the irregular pattern of a cognitively significant neuronal network.
7. The irregular distribution of synaptic connection also contributes to the lack
of repeatability of experiments.
8. The biological and chemical responses to the foreign matter from which the
electrodes are constructed can mitigate their use in long-term applications—a
requirement for clinical use. Similarly, long-term implantation of electrodes
remains a major challenge. A good discussion of the difficulties encountered
when one attempts to record or stimulate for long periods can be found in
the review by Moxon et al. (2009).
9. Even the most minute array of electrodes will do damage to neurons and
interconnections. Efforts to develop flexible and compliant microelectrode
arrays have met with only moderate success.
10. Much more devastating to the hope that the organization of realistic neural
networks will be uncovered using MEAs is the enormous amount of data that
is obtained from even a relatively few irregularly spaced neurons. If the goal
is to understand how multiple neurons interact, then the persistent problem
of combinatorial computability produced by the complexity of neuronal net-
works cannot be overlooked.

The thesis of this chapter is that partly because of technical and procedural
limitations, but mostly because of the intrinsic nature (i.e., the complexity) of the
problem, the direct, exhaustive, brute force strategy of stimulating and recording
the individual components of a microneuronal network that worked so well at the
level of single neurons is not likely to ever be implemented at the microneuronal
level for ensemble processes. Microneuronal network theory building in such a
many-neuron environment is severely inhibited because of the improbability of
ever simultaneously and controllably stimulating and measuring the activity of the
number of neurons involved in even a “simple” cognitive process. Thus, the
possibility of developing an authentic, valid, overarching microneuronal theory of
cognition based on robust empirical evidence appears to be remote. This is so
despite the fact that there are no plausible alternatives to Hebb’s conjecture and the
speculative theories that flow from it. Hebb was right, most cognitive neuroscientists
seem to agree, but there is no way to prove it in detail. Although we may all agree
110 Microneuronal Network Theories

that, in principle, cognition is encoded by synaptic mechanisms, the specific

relation of mental processes to the detailed states of the microneuronal network
that represent our thoughts is likely to remain permanently elusive.
In conclusion, despite some remarkable technical accomplishments, such as
development of MEAs, technology has not yet begun to contribute answers to
the great question—how do neural networks encode or represent cognitive
processes? There appear to be fundamental limits to what can be accomplished
with this type of multiple electrode device. Difficulties, including the lack of
selectivity of the individual electrodes of an MEA, the irregularity of neuronal
network organization, and the usual problems with poor signal-to-noise ratios
will perpetually obstruct progress toward a comprehensive neuronal network
theory. Many other technical difficulties exist, some of which may be solved
in the future whereas others will remain to perpetually frustrate the desire to
study the action of microneuronal networks. Looming over everything are the
problems of computational explosions and data storage and manipulation because
of the complexity of neuronal networks.
An additional conclusion to be drawn from this discussion is that naïve hopes
of building a brain with as many synthetic “neuromimes” is probably also
impossible. Not just because we do not have suitably sized supercomputers (we
are close to that goal) but because the logistics of stimulating and recording from
them are beyond us. These huge numbers of real neurons speak directly to any
hopes that we might have had for general or direct computational demonstra-
tions in support of Hebb’s conjecture and, as noted, restrict us to theories and
experiments involving only a few neurons—abstractions known as “toy networks.”
This pessimistic, but realistic, argument is sound because it is almost certainly
true given that immense numbers of neurons are involved in even the simplest
cognitive process.
On the immediate scene, complexity as we have presented it here has profound
practical effects. First, those who toil in the laboratory seeking the neural codes
underlying cognitive processes are inevitably and unavoidably going to be disap-
pointed barring some extraordinary reconceptualization of the problem. Second
(as suggested to me in a personal communication from the Arizona State Uni-
versity neurophysiologist Steven Helms Tillery), many originally optimistic new-
comers to the field have now begun to withdraw from efforts to map neuronal
network topologies except in the context of sensory and motor systems. These
peripheral transmission systems may help us to understand certain general prin-
ciples of network organization but cannot penetrate the specific codes for cogni-
tive processes.
A concise summary of the review presented in this chapter is that, however
plausible and unchallenged Hebb’s conjecture may be, there is almost no empiri-
cal evidence to accept its basic postulate—that cognitive activity is encoded by
specific patterns of activity in microneuronal networks. Despite much published
speculation revolving around this question and the development of ingenious
 Microneuronal Network Theories 111

TABLE 3.1 Relationships between Theories and Technologies

Theories Technologies
Mind is in the heart Dissection and trauma
Mind is in the head Behavioral observations
Mind is distributed EEG
Mind is localized fMRI
Single neuron theory Microelectrodes
Neuronal networks Computers and MEAs
Hebb’s conjecture Plausible speculation
Source unknown.

devices and complex analytical techniques, the goal of mapping a cognitive

process onto a neuronal network remains an unfulfilled hope. Experiments may
optimistically start with MEAs and elaborate statistical techniques, but they
quickly founder in a sea of complex interactions and numbers far too large to
be processed by even the most powerful computers of which we can conceive.
Although success with the relatively simple task of sorting SAPs provides an
illusion of success, the gap between spike sorting and mapping the neuronal
equivalent of a thought remains unbridged.
If this view of the neuroreductionist approach to cognition turns out to be cor-
rect and realistic, albeit pessimistic, how, then, can we account for the continued
effort to try to crack a code that may be uncrackable? The answer to this query
must lie in one of the premises of this book—namely that both our experimental
protocols and our prevailing theories of mind–brain relationships are in large part
determined by the technology that is available rather than robust empirical find-
ings. To highlight this relationship, consider Table 3.1.
In each case, theory is driven by the kind of instruments available, producing
a particular kind of data. The major exception to this generality is that the most
widely accepted theory—Hebb’s conjecture—is also the least substantiated by
solid evidence.

Notes
1. The optogenetic technique is a major step forward in neurophysiological research. By
various methods (such as using a virus that carries a light-sensitive trait) genetic material
can be introduced into a neuron (or, more usually, a class of neurons) that makes it
respond differentially to light stimuli of different wavelengths or temporal patterns.
Manipulations of opsins, the light-sensitive materials of the eye, are particularly useful in
this regard since their optical properties are well known. Because of the different optical
sensitivities of the genetically manipulated materials, specific neuronal or synaptic types
can be turned on or off by stimulating the neurons with different kinds of light. In the
Nabavi et al. study the virus used to vary the behavior of the synapses produced neurons
that were selectively sensitive to low-frequency trains of light. Specifically, the optical
112 Microneuronal Network Theories

stimulus–producing LTD was 900 pulses of light, each 2 msec, at 1HZ (Nabavi et al.,
2014). The stimuli-producing LTP was 5 trains of light (each train 100 pulses, 100Hz.
In this manner, the degree of synaptic activity could be manipulated and direct control
over their function achieved for the first time.
2. The three-body problem has been most extensively studied in physics in the context
of planetary motion. It has been known since the 19th century that the behavior of
three gravitational interacting bodies cannot be solved in general unless one of the
bodies is much smaller than the other two. The dynamic behavior of three equal-sized
objects is, as far as we know, nonrepetitive and unpredictable.
3. Sporns’ (2011) book is an outstanding discussion of the nature of both anatomical
connectivity and connectivity in general. It is a necessary part of the education of any
cognitive neuroscientist. It tells us much about networks, but it says very little about
the relation between the anatomical connectome and cognition.
4. Unfortunately, this work seems not to have been published (it is dated March 31, 2004);
more complete details can be found in a single entry to it here: http://web.mit.edu/9.29/
www/neville_jen/jkarr/mea/. It was, however, so interesting in the present context that
I could not ignore it.
5. It was not entirely clear from their report whether they collected the data they analyzed
or whether prerecorded data were provided to them for analysis. In an April 2014
personal communication from Karr, he indicated to me that they used prepackaged
data as the raw material for their analysis. For the present discussion, it does not really
matter since the goal of their paper (and of the present discussion) was to infer the
connectivity of the microneuronal network described by the data. Having prepackaged
data made the class project possible.
6. Of course, this number will change depending on whatever computer is available at
any given time. However, the numbers are such that the Stockmeyer theorem argues
that no conceivable computer could perform the necessary calculations in the lifetime
of the universe. Whatever the specific numbers, this theorem does provide an intuitive
foundation for the gap between computational needs and computational capabilities.
4
INTERMEDIATE LEVEL
NEURONAL NETWORKS—
THE ROLE OF MODELS

4.1 Introduction
A major goal of modern cognitive neurophysiology is to establish connections
between cognitive processes and the pattern of neural activity embodied in
networks of neurons. Obviously, however, this is going to be an extremely chal-
lenging goal for systems as complex as the human brain where innumerable
neurons are presumably involved in the simplest of thoughts. When dealing with
complex networks potentially involving millions and billions of neurons, the
technical and conceptual problems inhibiting this kind of research are profound
as was pointed out in the previous chapter. For all practical purposes, combina-
torics and numerousness probably will not permit experiments of the necessary
theoretical kind—for the present, at least, no empirical neurophysiological studies
of the detailed arrangement of cognitively realistic neuronal networks have yet
been carried out. An alternative strategy, of increasing importance in cognitive
neuroscience, is the use of relatively simple, naturally occurring model systems
that mimic or model some aspect of the behavior of much more complicated
organisms—hopefully with a simple analyzable neural network. The immediate
goal of this line of research is to at least understand the basic principles that
may help us understand how very complex neural networks operate by studying
the accessible and tractable neural networks of a simple creature.
Lying between the extremes of single neurons and full-blown neuronal net-
works, however, is an intermediary region in which there is a body of empirical
neurophysiological evidence and for which the networks are simple enough to
permit at least approximate descriptions of their functionality. This approach
depends on the discovery of relatively simple neuronal networks in “model”
organisms that are capable of mimicking what are almost certainly more complex
114 Intermediate Level Neuronal Networks

analogs of vertebrate neuronal networks. Invertebrate creatures have a long his-

tory of use as simplified surrogates for vertebrate nervous system functions. I
discussed one—the squid Loligo vulgaris—in Chapter 2’s discussion of single
neuron technology where anatomical freaks (e.g., large size neurons) permitted
breakthroughs in our understanding of the basic physiology of neurons. This
chapter takes the discussion of invertebrate model organisms that have served us
well in understanding how relative small networks can be related to behavioral
functions. The most important reason for the role model preparations play in
current cognitive neuroscience is that they provide a mechanism for the study
of neuronal networks that are otherwise too complicated to provide any hope
of unraveling the network structure.
To be used as a model of a vertebrate system, a simple invertebrate system
must meet certain criteria. These include the following:

1. The basic assumption underlying this kind of research is that behavior is

controlled by neuronal networks—that is, interactions among a number of
neurons.
2. The immediate goal of the research described here is that specific neural net-
works can be associated with specific behaviors.
3. The long-term goal is to understand how neuronal networks represent
cognitive activity. However, in most vertebrates, the number of neurons
controlling mental activity is so great as to currently pose an intractable
problem. Therefore, cognitive neuroscientists have turned to invertebrate
model organisms whose salient neural networks consist of only a few
neurons.
4. The number of neurons in a model neuronal network necessary to induce a
particular behavior must fall into an intermediate range bounded at one end
by a single neuron (see Chapter 2) and at the other end by practical and com-
putational limits on the analysis of cognitively significant neuronal networks.
Typically, in this kind of research, the upper end of this range varies from a
few to as many as a dozen neurons. Beyond that lies the analytical intractabil-
ity typical of even slightly more complicated networks.
5. Simplifying regularities (e.g., orderly and regular networks), however, may
occasionally increase the accessibility of networks consisting of as many as
1,000 neurons (see the section on the modeling of the Mach Band by the
horseshoe crab L. polyphemus later in this chapter).
6. The model organism must have analogous behavioral properties to make it
relevant to the modeler. However, these properties are not constant from
preparation to preparation. Size and accessibility of the modeled neurons usu-
ally help but so, too, does a quick turnover in the life cycle of the model
animal.
7. If the neurons of the model animal are large and their responses easily
recorded, either intracellularly or extracellularly, this is advantageous.
 Intermediate Level Neuronal Networks 115

8. It also helps if the nervous system of the model is stable from specimen to
specimen so that the involved neurons can be individually identified and
experiments be repeated.
9. Because many experimental procedures are now based on genetic manipula-
tions, having a more complete knowledge of the model animal’s genome is
also advantageous.
10. The design of an experiment must be able to distinguish between a necessary
and a sufficient theory of network organization (Olsen and Wilson, 2008).
11. Analogous behaviors must be carried out by both the modeled and the model
system. “Analogous” in this case means that the behavior accomplishes the
same function (such as learning) in both the modeled and the model organ-
isms. It does not mean that exactly the same kinds of networks are used to
implement both. Indeed, we typically do not know what mechanisms exist in
the modeled organism to produce the behavior—that is why we have turned
to simpler models for enlightenment.
12. The behavior of the model organism must be measurable and the network
configurations controllable in order to carry out controlled experiments. A
variety of methods, including genetic and neurophysiological techniques, are
now available to control behavior.
13. The a priori judgment that the structure of the model is exactly the same
as the structure of the modeled network is almost never provable. The rela-
tionship between the underlying mechanisms of each is one of analogy, not
homology, and though behaviorally similar, there remains uncertainty about
the salient networks of the model and the modeled, respectively.

If an experiment meets these criteria in a way that permits suitable experi-

ments to be carried out, then the relationship between behavior and simple
neuronal networks observed in the model can be extrapolated by analogy to
explain comparable behavior in the modeled vertebrate. Inferences may then be
drawn about the possible similarities in the underlying neuronal networks of the
modeled and the model, respectively. However, it is essential to appreciate that,
however robust the data, this is not proof that the kind of simple neuronal
networks observed in the model exist in the real system. Modeling experiments
of this kind are suggestive but not definitive. The problem is that such specific
behaviors as classical conditioning could be instantiated in vastly different neu-
ronal networks in the invertebrate model and the vertebrate system.
By judicious care in carrying out truly relevant experiments, a neuroreduc-
tionist theory of cognition may be generated. If the number of involved neurons
is sufficiently small, then it is possible that a mathematical model of the observed
neurophysiology can also be constructed to describe how a modest-sized network
of a few neuronal components in an invertebrate can produce what is hoped is
the same behavior as that produced by the millions of neurons in the vertebrate
nervous system. (A comprehensive review of the application of model systems
116 Intermediate Level Neuronal Networks

to a wide variety of fields in biological and medical research can be found in

Wilson-Sanders, 2011.)
This strategy of using models to understand the neuronal networks of much
more complex organisms is confounded by one great uncertainty—are the
analogous behaviors exhibited by a simple neuronal network in a model inver-
tebrate and the complex response mediated by a vertebrate organism really the
same? Or are they just functional analogues mediated by different neuronal
networks serving different behavioral needs? If they are just functionally similar
behaviors mediated by vastly different neuronal systems, what conclusions can
be drawn about the relationship between the simple responses of a model prepa-
ration and the highly complicated psychological processes of a human (e.g., in
the former case, habituation of neuronal responses in A. californica and, in the
latter case, habituation to pain in humans)?
Nevertheless, invertebrate model preparations consisting of a few neurons by
virtue of their very simplicity, if nothing more, permit a kind of plausibility test
of neuroreductionist theories that is not possible in a vertebrate brain. They can
demonstrate that basic principles of neuronal organization, such as lateral inhibi-
tion, habituation, accommodation, adaptation, feedback, or learning (i.e., behav-
ioral change as a function of experience), actually exist in microcosm as well as
the macrocosm. Furthermore, to the degree that these functions are common
to both the modeled and the model, they might well resolve some of questions
of how these processes might be represented in the macrocosm of the brain.
Theoretical inferences of this kind also permit the execution of mathematical
analyses that are difficult or impossible to carry out in higher-level nervous
systems. Thus, they can also serve as heuristics, stimulating theories that are
totally intractable to analysis in the complexity of higher organisms. Not the
least, they can be used as surrogates for human studies that are prohibited for
ethical reasons—none of us are likely to insert electrodes or excise tissue in a
human unless there is an overwhelming medical need.
It is in this spirit of a partial, incomplete, and approximate approach to the
problems of neuronal representations based on model preparations and interme-
diate levels of coding complexity that I review in this chapter four relatively
simple model systems. Studies that are based on hypothetical neuronal networks,
rather than real biological observations, will receive only secondary attention.
The problem is that not all intermediate scale network studies or theories are
based on even simple neurophysiological data. An entire field—“neurocomputing”
or “computational neuroscience” (e.g., see the work of Hertz, Krogh, and
Palmer, 1991; Anderson, Pellionisz, and Rosenfeld, 1993; Ham and Kostanic,
2000)—has grown up in which simplified neuronal networks are constructed
(rather than observed) in order to study the general properties of networks.
Many have no real neural data supporting them. The goal of these “dry” neural
network theorists is to develop the mathematical foundations of cognitively
significant networks in a world of artificial neuronal networks rather than to
 Intermediate Level Neuronal Networks 117

link them to real neurophysiological systems. Exercises in neurocomputing are

often based on highly simplified hypothetical neuronal networks that, more often
than not, draw only limited inspiration from real biological systems. Indeed, once
past the general idea of a network and parallel processing, it is rare to find an
analysis of a system based on real biological findings.
Why is this true? The answer to this question lies in the nature of real neural
networks. They are complicated to the point of computational intractability by
their irregularity of function, idiosyncratic interconnectivity, and sheer numer-
ousness. The result of these intrinsic impediments to dealing with real biology
has been to shift the emphasis from the study of understanding biological net-
works to the pursuit of strictly engineering topics (e.g., parallel processing)
although retaining the “neurocomputing” nomenclature. Some of the most
significant progress in the field of computational neuroscience has, thus, come
from research that was only distantly related to real neural networks. The major
works of such notable pioneers in the field of neurocomputing as McCulloch
and Pitts (1943), Rosenblatt (1958), and Hopfield (1982) were all carried out
with only perfunctory allusion to their original biological influences. These
distinguished pioneers often simplified their models to achieve mathematical
tractability to the point of irrelevance to biology as they made efforts to explain
how simplified networks might plausibly imitate biological functions.
Another issue should be considered at this point. The importance of model
systems is indisputable. However, there is an argument that even the advantageous
opportunities they represent may have the ability to misdirect our research and
theoretical activities in the same way that technology does. Just as the availability
of a measuring instrument may influence the kind of thinking that goes into
planning a research program, so, too, do the accidents of the availability of a
particular model system guide and constrain our thinking about the neuronal
substrates of a behavior. It seems quite clear, for example, that despite the fact
that it seems to provide a compelling foundation for a theory of perceived edge
enhancements, L. polyphemus did not evolve as a convenience for Hartline’s
group. It was available and an analysis of its network properties was sufficient
to analogize a particular visual phenomenon. However, proving its unique neces-
sity as an explanation for the perceptual phenomenon of edge enhancement
reported by human observers requires some strategies that may not be possible—
for example, the manipulation of the spatial stimulus and recording from many
neurons simultaneously. This may be considered to be a quibble, but the accident
of availability will always have to be a consideration—particularly when choosing
between competing ideas. Proving sufficiency is easy; proving the necessity of
a particular configuration is far more difficult.
The goal of the rest of this chapter is to show how a sampling of model
invertebrate systems allows us to examine real neurophysiological networks in a
way that permits us to provide rudimentary theories of well-defined cognitive
processes. The work I shall discuss collectively defines an intermediate level of
118 Intermediate Level Neuronal Networks

analysis in which actual neuronal networks are simple enough to be associated

with some well-defined behavior. We start with what is both a classical study
and a persistent example of perhaps the most successful and comprehensive neu-
ronal model of a well-known perceptual phenomenon yet developed—the edge
enhancement illusion known as the Mach Band emerging as modeled by the
horseshoe crab.

4.2 Limulus polyphemus and the Mach Band

Since the time1 of the great physicist, philosopher, and early cognitive neurosci-
entist Ernst Mach (1838–1916), scientists have known that edges between two
different luminous regions of different intensity would be perceived as being
enhanced on both sides of the edge. That is, the contour on the side of the
brighter region would appear to an observer to be brighter than simple radio-
metric measurements suggested it should be; and the contour on the dimmer
side would appear to be darker than the same physical measurement would
suggest it should be. Figure 4.1 shows the physical stimulus distribution and the
perceptual experience. These perceived edge enhancements are illusory; they
cannot be detected by the most precise measurements of physical stimulus inten-
sity and, therefore, must be the products of neuronal transformations carried out
in the nervous system.
As long ago as 1865, Mach presciently appreciated that this phenomenon was
associated with some kind of interaction between different points in the visual
scene. He was probably the first to have stated that the Mach Band was produced
Amplitude

Physical

Subjective

Distance

FIGURE 4.1 The shape of the physical and subjective responses to a gradient of
illumination.
Reproduced from Uttal (1973).
 Intermediate Level Neuronal Networks 119

by reciprocal interactions between different points on the retina and/or the rest
of the perceptual nervous system. Of course, in the 19th century, it was not pos-
sible for researchers to carry out the necessary experiments to confirm this
hypothesis, but in the latter half of the 20th century, it became eminently feasible,
if not yet on the vertebrate visual system, then on a simpler model preparation—
the visual system of Limulus polyphemus—the North American horseshoe crab.
The horseshoe crab is a vestige of organic evolution that has survived from
roughly 400 million years ago. It is an arthropod but actually unrelated to crabs
and other crustaceans; it is more closely related to spiders and other arachnids.
Fortunately for cognitive neuroscience it is also graced, like spiders, with multiple
eyes on the top of its carapace. Two large compound eyes, in particular, are of
special interest because each consists of approximately 1,024 relatively large cell
complexes called ommatidia. Each ommatidium contains a number of secondary
cells, but lateral neuronal connectives from the eccentric cell (the main light-
sensitive neuron in the ommatidium) are the main links to other cells of the
complex ommatidium. By this means, light-induced neural responses interact
with surrounding ommatidia through lateral connections that can be seen in
optical microscopes (as shown in Figure 4.2—a drawing of the neural network
in one ommatidium).

Corneal lens Glassy cells B C

ret. c. Pigment

rhab.
C

Eccentric cell

Neuropile
A sh. cell nuc.

FIGURE 4.2 The anatomy of an ommatidium in the L. polyphemus compound eye.

Analogs of the lateral connections at the base of the figure are believed to account
for the Mach Bands.
Reproduced from MacNichol (1956) with permission.
120 Intermediate Level Neuronal Networks

The relatively huge size of the cells in the ommatidia first led researchers
(Hartline and Graham, 1932) to use them as a model preparation for the study
of single neurons. However, an inspired epiphany stimulated the group led by
Hartline (1940) to consider the L. polyphemus eye as a model of interneuronal
interactions. The basic paradigm that Hartline and his coworkers used was to
stimulate a narrow region of the crab’s eye with a spot of light, measure its
neural activity with electrodes, and then stimulate it with another spot in a dif-
ferent location. The typical result was that the addition of the second light
stimulus would reduce the response from the ommatidium from which the
recording was made. Although only a single electrode was used to make this
measurement (and the response of only a single ommatidium recorded at a time),
both the electrode and the two spots of stimulating light could be moved across
the compound eye to track out the spatial pattern of the response as a function
of such variables as the distance between the two light spots. By this reiterative
procedure, it was thus possible to deal with this model preparation as if multiple
electrodes were recording from a spatially distributed response to a spatially
distributed stimulus—sequential recordings taking the place of a simultaneous
spatial distribution of responses.
This technique may have been the first experiment to deal neurophysiologi-
cally with neurons interacting in a microscopically defined real neuronal network.
Previously only recordings had been made from single neurons. The work carried
out by the Hartline group was, therefore, a major milestone in the study of
neuronal network theory thanks to the unusual nature of an evolutionary
anomaly—L. polyphemus. Although the number of neurons in the creature’s
eye was larger than what I have designated here as an intermediate level, the
extreme regularity of the anatomy made it possible to deal experimentally with
the network properties of this unusual eye. It is problematic whether this feat
has been repeated in any other model preparation.
The specific question asked in this pioneering series of experiments was—does
the presence of the second stimulus affect the first region’s response (Hartline,
1949)? The answer to this questions was that yes, it certainty does and in a very
specific manner whose properties could be measured—the sine qua non of a
model preparation. The interactions between ommatidia were solely inhibitory,
diminished in strength with the distance between stimulated regions, and were
also mutual (Hartline, Wagner, and Ratliff, 1956)—the effect of one on another
was the same as the other’s on the first. The main implication of this research
was to open the door to the study of the simultaneous interactions of many
neurons in the form of a network. It provided a compelling explanation of the
Mach Band phenomenon by demonstrating a neurophysiological analog of edge
enhancement, thus validating what had only been an inspired guess by Mach.
This simple model invertebrate system thus also provided a compelling
explanation of the similar neural interactions that presumably were going on in
the human visual system. It was determined that these lateral inhibitory
 Intermediate Level Neuronal Networks 121

interconnections in the eye of L. polyphemus produce contour enhancements

in the animal’s compound eye that were at least analogous to the psychophysi-
cally measured Mach Bands in humans, as shown in Figure 4.1, if not homologous.
Because of the regularity and simplicity of the neural mechanisms, it immediately
became obvious that a formal mathematical model of the system was possible.
The formularization of the appropriate mathematical model was carried out by
Hartline and Ratliff (1958) in the form of a system of simultaneous equations,
shown in Equation 4.1. Each of the equations in this system calculated the
response (rp) of each ommatidium as positively influenced by the stimulus excita-
tion (ep) and negatively by the sum of all inhibitory influences from its neigh-
boring ommatidia. In Equation 4.1 the sum of the inhibitory forces from n cells
is modified by a constant (kpj) specific to the interaction between any pair of
ommatidia. The term (rj − rp,j) refers to the threshold of response for each
ommatidium:
n
rp = e p − ∑k pj (r j − rpj )  Eqn. 4.1
j =1

There are several important aspects of Hartline and Ratliff ’s mathematical

model, not the least of which is the fact that it was one of the first such theories
to take real neurophysiological data (not just some speculations about hypotheti-
cal mechanisms) and express them in a form that could be numerically evaluated.
In many ways it was nearly an ideal model preparation. The number of ommatidia
was small (1,024), they were very large, and they were interconnected in virtually
the simplest possible manner—mutual lateral inhibitory connections. The arrange-
ment of these visual units of the L. polyphemus eye was also highly regular;
indeed, the orderly arrangement of the ommatidia is almost crystalline in the
way they are distributed and interconnected within the compound eye. Thus, it
was possible to manipulate a real neuronal network by varying some of the key
properties in order to determine some of the values that characterize the real
system but might not be directly measurable neurophysiologically.
Most important of all is that this model preparation responds in a way that
is comparable to the behavior of the system being modeled. Comparable in this
case means that it—the model—follows the same laws and tracks the same
trajectory that the real world process does producing the same phenomenological
outcome. In other words, the model’s and the real system’s behavior are functional
analogs of each other. In short, they are doing the same thing—enhancing edges
that are not present in the physical stimulus! But are they really the same?
This question brings us back to the caveat expressed earlier—is model system
research producing the same behavior (i.e., edge enhancement) because of the same
underlying mechanism—lateral inhibitory interaction—or are the common behaviors
simply analogous response outcomes of what are actually different mechanisms?
The point is that it is entirely possible that there is more than the one explanatory
122 Intermediate Level Neuronal Networks

model on which this explanation may be based. There is no way to determine

if this is the case in the L. polyphemus–human comparison situation because
there is no way to determine if the human perceptual Mach Band is caused by
the same kind of lateral inhibitory interaction used by L. polyphemus’ visual
systems. Our ability to use the model system to examine the actual neurophysiol-
ogy of the network in L. polyphemus is not matched by our ability to examine
human neurophysiological processes. There are alternative possible systems that
involve non-lateral inhibitory interaction, perhaps in the form of symbolic2 rather
than the geometrical interactions epitomized by Mach’s famous theory. That
alternative explanations may exist as plausible, if undemonstrated, alternatives is
suggested by some discrepancies between the way L. polyphemus produces the
neurophysiological analogs and the nature of related human illusions. Specifically,
I refer to the equally well-known simultaneous contrast phenomenon, which dis-
tributes the enhancement effects equally over the entire spatial extent of an
illuminated region surrounded by a light of different intensity rather than just at
the edges; there is no edge enhancement in this illusion yet it is typically “explained”
by the same lateral inhibitory interaction mechanism as the Mach Band. Both
cannot be correct.
What does result from Hartline’s analysis is, therefore, a highly plausible theo-
retical network description of a complex visual process based on certain functional
analogies and robust neurophysiological data. Given the current state of knowledge
of the origins of our perceptual experience it, therefore, becomes defensible, if
not in a definitive way, then in a highly plausible and possible way to say that
we have “explained” the Mach Band phenomenon.
The advantage of using L. polyphemus as a model of a human perceptual phe-
nomenon is that the anatomy and physiology of the model animal is so simple
and stable from one animal to another that it can be used to study one possible
neural explanation of the perceived Mach Band in a direct way that would not
be amenable to direct experimentation in humans. To have such compelling forms
of analogous behavior in both the recordings from L. polyphemus and the spatial
dimensions of perceptual experience reported by humans was a major part of the
success of this project. The specific parameters of the mathematical model are
based on specific anatomical and neurophysiological measures obtained from
experimentation. No a priori assumptions about the underlying mechanisms were
necessary although in this simple system, a genius such as Mach was able to pre-
sciently infer an explanation based on reciprocal inhibitory neural interactions de
novo. In more complex networks, preliminary assumptions may be misleading and
the possibility of alternative explanations in the form of other kinds of brain func-
tions cannot be completely rejected. For example, the illusion may also be explained
in terms of the high band pass properties known to exist in the visual pathway—an
entirely different process than the reciprocal lateral inhibitory interaction on which
the Hartline model is based. None of this diminishes one of the great success
stories in cognitive neuroscience.
 Intermediate Level Neuronal Networks 123

This does not mean that the final explanation of this phenomenon in humans
is at hand. That the neurophysiological properties underlying edge enhancement
in the L. polyphemus and the human perceptual phenomenon are mediated by
exactly the same simple neuronal networks seems intuitively unlikely. The Hartline-
Ratliff “theory” that the neural and perceptual phenomena are alike in principle
is based on a very simple sensory mechanism for which there is ample neuro-
physiological supporting data and in which the neuronal network is relatively
simple in structure. In other words, although the L. polyphemus eye effect is based
on local geometric interactions, we still have not established that this is also the
case in the human. Nevertheless, what we do know about the neurophysiology
and neuroanatomy of L. polyphemus is probably still the best model we have
today in which the neural and psychophysical data are so mutually consistent.
In sum, the Hartline et al. research program was one of the first, if not the
first, to develop a mathematical theory of observed neurophysiological network
mechanisms that might underlay a psychophysically reported illusion—a discrep-
ancy between the stimulus and the resulting experience based on the properties
of an interactive network. It did so by providing a body of relevant experimental
data that built on direct measurements of neurophysiological responses. The
model has been widely accepted because it reproduced the perceptual experience
in a simple model neurophysiological preparation. It is however, based on another
unproven assumption, namely that the spatial mapping of the neuronal response
in L. polyphemus is recapitulated in the perceptual experience of the human
observer. This gap in the logical chain remains unbridged.
The key to this theoretical success was the susceptibility to mathematical
modeling provided by the simplicity of the L. polyphemus eye. An important
aspect that makes this analysis so compelling is that the mechanisms used to
explain the overall system responses are from microneuronal interactions—that
is, network properties. The edge enhancement phenomenon would not be
observed in the activity of a single ommatidium. It was only when multiple
components were shown to interact with one another that the neuronal con-
nections and interactions of this compound eye exhibited the edge enhancement
phenomenon. The presumption is that similar (in principle) but probably much
more complex neuronal networks account for the Mach Band perceptual illusion
reported by humans. If nothing else, this work was one of the earliest to provide
an empirical foundation for a fundamental postulate of cognitive neuroscience—
that it is from the information-processing capabilities of neuronal networks that
a preliminary understanding of cognition is most likely to come.

4.3 Aplysia californica and Nonassociative Learning

If L. polyphemus provided the first and still most famous vehicle for the devel-
opment of an intermediate level network theory of sensory or perceptual func-
tion, it was another quite different animal that provided a model for a kind of
124 Intermediate Level Neuronal Networks

behavioral change referred to as Nonassociative Learning (NAL). NAL is a

change in behavior caused by simple experience with stimuli. Unlike classical
or instrumental learning in which some additional stimulus must be associated
with the one under study to reinforce or modify behavior, NAL depends purely
on the number of times a stimulus is experienced. Thus, an experiment studying
NAL may involve nothing more than simple repetition of a stimulus and mea-
surement of changes in a neuronal response.
It is well established that NAL occurs in humans—engaging in repetitious
tasks leads to a reduction in performance. Furthermore, we learn not to be aware
of the constant stimuli coming from our clothing or the repetitious ringing of
a bell in the background of our environment. In general, this reduction in
response strength resulting from a repetitive or constant stimulus is a form of
NAL behavior called habituation. Habituation is a relatively easy protocol to
study experimentally. All one has to do is repeatedly stimulate a preparation and
observe the strength of the evoked response.
A related phenomenon is the rapid recovery from habituation referred to as
dishabituation. Dishabituation occurs when the quality or locale of a stimulus
changes, that is, when a novel stimulus is presented. Habituation and dishabituation
are supposedly the result of a reduction in the behavioral salience of the stimulus
but until recently, where in a neuronal network this kind of learning resided could
only be speculative. The behavioral significance of habituation is that repeated
stimulation that does not harm the animal tends to diminish in behavioral salience.
Habitation is thus neatly and concisely defined as a response decrement as a result
of repeated stimuli that is not due to sensory or motor fatigue.
Experimentally distinguishing habituation from sensory or motor fatigue is
not easy; both processes represent decrements in performance as a result of
stimulus repetition. For habituation, the best indicator is that after habituation,
when a novel or a nonspecific stimulus is presented, the original response imme-
diately recurs with a strength equal or greater to that obtained before the habitu-
ation stimulus was initially presented. To the contrary, if the response is due to
fatigue, the response produced by a novel stimulus would still be attenuated and
would remain so until the sensory or motor mechanism recovered from whatever
resource exhaustion had diminished its response strength. In sum, habituation is
a decrease in responsiveness with repeated stimuli not due to fatigue. Other
suggested criteria for distinguishing habituation from fatigue, such as the expo-
nential shape of the decrease of the response as a stimulus is repeated, are not
definitive since fatigue and habituation generally produce similarly shaped response
decrement curves.
A closely related, but inverse, phenomenon to habituation is sensitization—an
increase in response strength with repeated stimuli. Sensitization is considered to be
a defensive response to stimuli that are noxious. Sensitization, however well docu-
mented, is more elusive of definition, but to use a parallel with disinhibition, we
may define sensitization as a response increment as a result of repeated stimuli.
 Intermediate Level Neuronal Networks 125

Because of the conceptual simplicity of NAL as a highly reduced form of

learning, explorations of its neuronal origins raised the possibility that an animal
model of habituation might illuminate some of the basic principles of learning
in general. Gerard and Forbes (1928), for example, in an early article, had studied
what they called “fatigue” in decerebrate cats. It now seems that their methods
suggest that the response they observed may have been better defined as a form
of habituation. The state of psychological studies of NAL has been periodically
reviewed by investigators such as Harris (1943) and Thompson and Spencer
(1966) and, although currently less of an active research topic than previously,
its simplicity as a neurophysiological model of learning has rejuvenated interest
in the field.
In the 1960s and 1970s a team led by Eric Kandel, who was subsequently
to receive the Nobel Prize for this work, began to study habituation and dis-
habituation in the sea hare (Aplysia californica, a gastropod mollusk) as a model
of simple forms of learning. A. californica has some highly desirable features
that made it a superb model preparation. It has a relatively simple nervous system
and yet has been shown to exhibit a variety of different kinds of learning,
including both NAL and associative types of conditioning. The neurons in its
nervous system are large and relatively few—approximately 20,000. Furthermore,
many of the neurons are both individually identifiable and constant in position
and shape from one animal to the next.
A. californica’s natural behavior included two defensive reflexes to protect the
sensitive tissue of its siphon and gills. These reflexes are mediated by different
neurons in the abdominal ganglion and elicited by what turned out to be very
simple neuronal networks. Some motor neurons (e.g., LDs1) in this ganglion
controlled contraction of the siphon while others (e.g., L7) controlled gill retrac-
tions when stimulated by what could be dangerous or toxic stimuli or even
slight touch. Because of what turned out to be a behavioral similarity to human
NAL and the relative simplicity of the neuronal network controlling this behavior,
A. californica became a model animal for the study of habituation and disha-
bituation in humans. This relevance could be enhanced because the animal
proved to be robust enough to have the germane portions of its sensory and
motor systems surgically isolatable, further simplifying the model system down
to a minimum number of neurons and synapses—a number that surprisingly
turned out to be mediated by a single synapse.
The early experimental designs to study NAL in A. californica were straight-
forward. Stimuli were typically mechanical (droplets of water) although in some
other work flashes of light or small electrical stimuli were used. Motor responses
were measured in various experiments by examining neuronal responses with
both intracellular and extracellular electrodes or by photocells placed in the
pathway of the retracted tissues to study motor functions such as gill withdrawal.
The system proved to be so simple that both electrophysiological and motor
responses could be measured and compared in identified neurons.
126 Intermediate Level Neuronal Networks

The pioneering experimental paradigm used by Kandel and his colleagues to

study habituation and dishabituation in A. californica was described in what are
now considered to be three classical papers (Castellucci, Pinsker, Kupfermann,
and Kandel, 1970; Kupfermann, Castellucci, Pinsker, and Kandel, 1970; Pinsker,
Kupfermann, Castellucci, and Kandel, 1970). The first of these three reports
(Pinsker et al., 1970) documented a purely behavioral study in which an immo-
bilized but intact A. californica was stimulated with a “brief jet of sea water”
delivered to a sensitive spot on the animal’s skin. The response measured was a
physical retraction (gill withdrawal) and was measured with a photocell that was
tucked under the gill. Repetitive stimulation with a series of constant-amplitude
stimuli showed a progressive decrement in the magnitude of the gill withdrawal
response as the stimulus repeated, that is, habituation. Even after the A. califor-
nica’s gill withdrawal response was completely habituated, the response magnitude
would promptly recover to prehabituation levels if a relatively strong stimulus
was delivered to another sensitive region of the skin, that is, dishabituation. This
observation and the presence of spontaneous gill withdrawals at a high amplitude
level during the experiment established that the responses were not due to sensory
or motor fatigue, but could be considered to be a legitimate form of NAL.
Based on this behavioral study, the next two articles in this series were aimed
at determining the underlying neuronal mechanisms of this simple form of
learning by establishing the neural network that mediated the gill withdrawal
behavior. In the second article in this series (Kupfermann et al., 1970), intracel-
lular microelectrodes were inserted into neurons of the abdominal ganglion.
Because of their large size, individually identified neurons could be examined
from one animal to the next. The stimuli used to activate the gill withdrawal
response in this second study were both electrical and tactile, the former of
which were delivered through a double-barreled microelectrode.
Because of the large size and anatomical repeatability of the neurons in the
abdominal ganglion, Kupfermann and his colleagues were able to identify addi-
tional motor neurons that individually could produce the gill withdrawal response
to stimuli. They were also able to determine that there were no changes in the
amplitude of the sensory neuron response or in the neuromuscular junctions
that activated the muscles controlling the gill withdrawal during the repeated
series of habituating stimuli. The main surprising result of their experiment was
that the electrophysiological changes mediating the gill withdrawal reflex seemed
to occur in the single synapse between a sensory and a motor neuron. The
neural network they developed of the A. californica gill withdrawal is shown in
Figure 4.3. This is not an abstraction or a diagrammatic representation but a
more or less complete network made up of the actual neurons and synapses
involved in A. californica’s NAL.
The final article in this series (Castellucci et al., 1970) carried the Kandel
group’s analysis of A. californica habituation to the next level. In the two earlier
papers, the gill and siphon withdrawal reflexes were examined first by behavioral
 (a) (b)

Rec stim LRQG L2 L3 R5 R4

R3
R1
L1
L4 L5
R6 R
7
R8 RRQG
Lucite rod LCQG L9
L6
R2

L7
and stage L8
LD LD RCQG
R14
L10 R13

(c)

Siphon

Mantle shelf
Gill

INT

Polysynaptic
L7 LD L9 L9
G 1 2

SENSORY
N.

Monosynaptic

FIGURE 4.3 The anatomy of the gill and siphon neural mechanisms in Aplysia cali-
fornica. (a) and (b) show the gross anatomy of the system, and (c) shows the six
neurons involved in withdrawal from noxious stimuli, some of which are
redundant.
Reproduced from Kupfermann et al. (1970) with permission.
128 Intermediate Level Neuronal Networks

studies and then by electrophysiological studies on the relatively intact animal,

respectively. Because the anatomy of the A. californica nervous system was not
completely known, there still remained the possibility that more complicated,
but unknown, networks were redundantly involved in habituation of the response.
Recall that at least four motor neurons had been demonstrated that could elicit
the withdrawal reflex. Furthermore, it was subsequently determined (Zecevic
et al., 1989) that as many as 300 neurons might actually be involved in the gill
withdrawal reflex.
What was needed was an even more reduced model to determine exactly what
was happening in this simple nervous system. The goal was precisely to determine
the locus of the changes that produced habituation by eliminating other possibili-
ties, and what better way to accomplish this than by dissecting away other possible
neuronal mechanisms. To accomplish this, Castellucci et al. produced a dissected
preparation that was reduced to its most fundamental simplicity—a minimal set
of components that could sustain the habituation process. This minimal prepara-
tion included a patch of skin from the sensitive receptive field (the stimulation
of which produced the gill withdrawal response), a single nerve “strand” passing
centrally from this sensitive region, and an abdominal ganglion. In this highly
reduced preparation, electrical or tactile stimuli applied to the sensitive patch of
excised skin would produce synaptic responses in the motor neurons when
stimulated with an impaled microelectrode. (The motor neuron responses were
monitored as a surrogate for the gill withdrawal response as determined in the
first two experiments.) Although there were several motor neurons, 85% of the
withdrawal response was mediated by a single one—LDG1—and much of their
subsequent research utilized an isolated segment of this neuron. The responses of
LDG1 in the abdominal ganglion were measured as the series of the habituating
stimuli were presented.
After carrying out a number of control experiments, Castellucci and his col-
leagues were able to show that both the habituation and dishabituation processes
seemed to result from changes in the conductivity of the single synapse interposed
between the sensory and motor neurons. As the stimulus was repeated, the
synaptic conductivity was reduced and the response in the motor neurons pro-
gressively decreased. Similarly, when a dishabituation stimulus was applied, the
synaptic response increased to its original level. Later on, Kandel’s group (Cohen,
Kaplan, Kandel, and Hawkins, 1997) showed that the synaptic effects were dif-
ferent for habituation, on the one hand, and dishabituation and sensitization, on
the other. This follow-up work suggested that habituation might be produced
by reduction in the sensory synapses, whereas the other two behaviors might be
due to more complex networks involving such additional mechanisms as excita-
tion of neuromuscular junctions.
In short, this research showed that habituation of the gill withdrawal reflex
occurred in the simplest of possible neuronal networks—a sensory and a motor
component as regulated by modifications in the conductivity of a single
 Intermediate Level Neuronal Networks 129

(monosynaptic) junction whose postsynaptic properties mirrored the behavior

of the intact animal. Fatigue was ruled out in both the motor and sensory
mechanisms. Whatever the details of the specific mechanisms involved, the fact
that this reduced form of NAL could be attributed to a specific synaptic mecha-
nism was a major breakthrough. A form of learning—NAL—was shown to be
explained (in a very strong sense of the word) by synaptic conductivity changes.
In other words, the relation between a stimulus and a behavior was caused by a
very specific change in the properties of this simple neuronal network.
This was an extraordinarily important finding despite the fact that it came
from a very simple model preparation. For what was perhaps the first time, a
well-known form of learning, however primitive this form of NAL may be,
could be attributed directly to synaptic changes. Although there had been much
speculation (see the discussion of Hebb’s prescient ideas in Chapter 3) that such
synaptic changes lay at the foundation of learning, this may have been the first
time that such a strong correlation between a behavioral change and changes in
synaptic conductivity had been demonstrated. Of course, this particular mecha-
nism for habituation and dishabituation—modifications in a single synaptic
junction between sensory and motor neurons—might be present only in such
simple invertebrates as A. californica; more complex networks may be necessary
to explain analogous processes for habituation, dishabituation, and sensitization
in higher animals. Nevertheless, the principle that changes in synaptic conduc-
tivity might account for behavioral changes reverberates down through all of
modern cognitive neuroscience.
Kandel and his colleagues went on to elaborate the properties of this model
preparation. Carew, Pinsker, and Kandel (1972) demonstrated that the habituation
process might have very long-term effects lasting as long as three weeks, raising
the possibility that not only short-term memory (in the form of NAL) could
be explained, but perhaps also, a longer-term form of memory. Castellucci and
Kandel (1976) subsequently showed that habituation results from synaptic con-
ductivity reductions, sensitization is likely to be associated with presynaptic
facilitation, but dishabituation is more likely mediated by post-synaptic changes.
An important expansion of their work was reported by Carew, Hawkins, and
Kandel (1983) in which they showed that classical as well as Nonassociative
Learning could be demonstrated in A. californica. Nevertheless, it was the initial
trio of publications that set the stage for decades of work to follow.
The most important aspect of their contribution was the confirmation of the
idea that the changes in synaptic conductivity could be associated with changes
in molar behavior. It must also be appreciated, however, that this example of
how a model animal might trigger theoretical insights into more complex learn-
ing processes is made possible by the fact that the relevant portion of A. cali-
fornica’s nervous system is virtually as simple as it can be in terms of the neuronal
interconnections. In A. californica, only a single sensory neuron and a single
motor neuron interconnected by a single synapse are sufficient to provide a
130 Intermediate Level Neuronal Networks

compelling explanation of simple behavioral changes. As usual, the standard

caveat must be expressed; it is not likely that the simple monosynaptic junction
to which they attribute NAL in this model animal is the same neuronal mecha-
nism as that mediating NAL in higher vertebrates. The gill withdrawal response
and human habituation are at best analogs of each other. The utter simplicity
of the neuronal interconnections in this preparation make any network theory
trivial—the A. californica neuronal mechanism for NAL could be modeled at
the synapse level simply by assigning graded valences to the operation of a single
synapse as information is passed from a single sensory neuron to a single motor
neuron. As a result, based as it is on such robust findings and relatively simple
recording techniques, the simple A. californica nervous system has become an
excellent choice for studying the biochemistry of synaptic functions, especially
the role of certain ions in distinguishing between pre- and postsynaptic changes
in synaptic conductivity. Research on this model preparation has mainly shifted
from the nearly complete study of the synaptic information processing to the
biochemistry that can explain the macromolecular changes in the synapse. For
example, Gingrich and Byrne (1985) and Gingrich, Baxter, and Byrne (1988)
explored the role of calcium and potassium ions in regulating presynaptic facili-
tation. An even more up-to-date simulation by Zhang et al. (2012) dealing with
the enhancement of long-term learning protocols in A. californica concentrates
on the role of such chemicals as protein kinase A and signal-related kinase in
responding to the replication rate of stimuli in prolonging long-term effects.
The point in considering these few examples of what is certainly a large
research literature is to illustrate the value of a highly simplified model prepara-
tion not only as a model of learning but also to the degree it opens the door
to understanding the molecular details of synaptic action. What we must ten-
tatively conclude from this tale of model nervous systems and behavioral changes
is that it is more likely than not that synaptic plasticity is the neurophysiological
basis of NAL and probably all other kinds of learning.

4.4 Caenorhabditis elegans and Learned Chemoaversion

Although it may be stretching the meaning of the term “intermediate level of
analysis,” no discussion of model nervous systems would be complete without
considering current knowledge concerning the detailed system organization and
anatomy of the microscopic (its body is approximately 1 mm long) nematode
Caenorhabditis elegans. This tiny nematode has some of the same advantageous
properties exhibited by the A. californica and L. polyphemus nervous systems.
Most of all, compared with vertebrates, it is very simple. The number of neurons
in the nervous system of C. elegans is also relatively small (302) and the axons
interconnecting neurons unusually do not branch very much as they run between
ganglia. As a result, only a few thousand synapses are known to interconnect
the 302 neurons. Furthermore, not only are the same neurons present in one
 Intermediate Level Neuronal Networks 131

sample of C. elegans to the next, but also each neuron has been identified and
both its shape and its location mapped. According to Sporns (2011), this is the
only model creature for which such a complete nervous system map currently
exists. The mapping was an enormous task carried out by White, Southgate,
Thomson, and Brenner (1986) based on an exhaustive reconstruction of a series
of electron micrographs. As a result of this complete mapping and some inge-
nious research techniques, C. elegans has proven to be a useful model animal
for studying the relations between behavioral and neurophysiological
processes.
However, this approach has not been without difficulties and challenges. Why
should this be the case given the detailed nature of all of the neurons in the
brain map we currently possess? The main answer to this question is that the
C. elegans nervous system, despite its small size, is considerably more complex
than the other two model nervous systems just discussed. The word “complex-
ity,” so easy to articulate and so difficult to explain, is an expression of a general
principle permeating this entire book—the complexity and numerousness of the
neuronal network of even these simple model animals pose combinatorial prob-
lems that quickly explode to demand unachievable computational powers.
The situation concerning the use of C. elegans as a model nervous system,
therefore, is quite different than the ones characterizing A. californica or L.
polyphemus, both of which have simplifying attributes. In the former, it was pos-
sible to distill the germane interactive network function down to a single repeti-
tive functional process (mutual lateral inhibitory interaction) in which the
arrangements of individual nodes, the ommatidia, were almost crystalline in their
regularity. This made iterative computational algorithms simple enough to be
executed. In A. californica, it turned out that the germane functional process
(habituation) was actually ascribable to mechanisms that could be localized to a
single interneuronal junction.
Furthermore, both A. californica and L. polyphemus are relatively large animals
compared with C. elegans. Thus, they could be manipulated and probed with
stimulating and recording instruments in a way that the smaller neurons of C.
elegans could not. The ommatidia of L. polyphemus and the axons and ganglia
of A. californica were large enough to be easily manipulated in an electrophysi-
ological experiment. The small size of the animal made simple mechanical
manipulation of C. elegans difficult—special techniques had to be developed to
immobilize the creature’s miniature worm-like body. C. elegans neurons were
not only small but had the disconcerting property of bursting when punctured
with a microelectrode. Nevertheless, the task of electrophysiological recording
from this tiny creature was accomplished for neuromuscular junctions by Raizen
and Avery (1994) and then for neurons by Lockery and Goodman (1998) and
Goodman, Hall, Avery, and Lockery (1998). A useful compendium of practical
advice detailing how neurophysiological recordings can be made from C. elegans
can be found in an article by Richmond (2006).
132 Intermediate Level Neuronal Networks

Despite its small size, the resulting practical problems, and the relatively large
number of neurons in this “simple” nervous system, C. elegans has been success-
fully used as a model for olfactory aversive learning. Ha et al. (2010) were able
to compare changes in behavior as a result of experience with the operation of
a relatively small neuronal network consisting of 13 neurons located in a ganglion
in the anterior end of this nematode. The experimental procedure they used
was based on an innate behavior exhibited by C. elegans; this tiny nematode
responds to the activation of chemosensitive neurons in its head region to a large
number of chemicals secreted by the bacteria that form its food (Bargmann,
Hartwieg, and Horvitz, 1993). Interestingly, although many of these chemicals
are detrimental to the animal, instigating diseases or metabolic failures, naïve
animals that have not been exposed to these chemicals are attracted to rather
than repelled by the chemical odorants. It is only after some experience with
the negative effects of these chemicals that C. elegans avoids them. The task faced
by Ha and his group was to determine how and where this model nervous
system was organized to alter this behavior from attracted to aversive.
One of the first problems encountered was to measure the change in behavior
from the naïve (attractive) to the mature (aversive) state. Fortuitously, the pattern
of movement executed by C. elegans toward attractive chemicals differed from
its pattern of movement away from aversive chemicals. There was a much smaller
proportion of a kind of behavior Ha and his colleagues designated as Ω turns
(“sharp turns in which the animal’s body shape resembles the Greek letter omega
Ω”), when the animal behaved as if there was an attractive odorant present, than
when the behavior suggested that the chemical was aversive. Populations of C.
elegans could then be trained to be used as controls for those that had been
“taught” to avoid detrimental odorants and those that had not been. The pro-
portion of Ω turns during attractive behavior could then be compared with the
proportion during aversive behavior to determine whether the chemical was
attracting or repelling C. elegans.
To explore the questions of the neural correlates of this learned behavior
(from attraction to avoidance), it was then necessary to manipulate the C. elegans
nervous system to determine the role of individual neurons in the learning
process. This was accomplished by what was also an ingenious technique. Ha
and his colleagues used a laser whose beam size had been demagnified to the
point that it could target individual neurons. The purpose of the laser was to
destroy targeted neurons. Therefore, any changes in C. elegans’ behavior could
be attributed to the absence of that particular neuron. The results of their
experiments with this kind of selective destruction allowed them to map out
the details of a network of 13 neurons that accounted for the naïve attractive-
ness of the odorants, the learned avoidance of them, and the experienced avoid-
ance of the odorants. This analysis allowed Ha and his colleagues to uncover
the fact that two different configurations of a relatively simple neuronal network
 Intermediate Level Neuronal Networks 133

Synapses:
AWB AWC
1 - 10
11 - 20
AIY
ADP 21 - 30
31 +
AIZ

Gap junctions:

RIA AIB
1-3
4+

RIV RIM RIB

RMD Sensory neurons
Interneurons
SMD SAA Motor neurons

FIGURE 4.4 The neural network mediating the attractive–aversive behavior of

Caenorhabditis elegans to odorants. Only 13 neurons control a complex behavioral
pattern.
Reproduced from Ha et al. (2010) with permission.

mediated the naïve attractiveness and the trained aversion responses to the odor-
ants. The network of neurons mediating the two forms of behavior as well as
the learning process that changed the behavior is shown in Figure 4.4. Each of
the blocks identifies one of the 13 neurons that are interconnected by the number
of synapses indicated in the side bar.
Ha and his colleagues summarized the role of destroying various neurons on
the naïve (attractive) and learned (aversive) chemosensitive behavior as well as
the role individual neurons had on the learning process itself. With the pattern
of observed neuronal interventions and behavioral changes they were able to
construct the neuronal network shown in Figure 4.4—a remarkable tour de
force made possible by the evolutionary accident of C. elegans’ anatomy. The
following list abstracts a few of the conclusions they drew from their experi-
ments that allowed them to construct the neuronal network of C. elegans:

1. Learned aversive olfactory behavior requires the presence of the AWB and
AWC olfactory neurons. The network shows one candidate network that
regulates this naïve and learned behavior.
134 Intermediate Level Neuronal Networks

2. Two different neuronal networks regulate the attractive (naïve) and aversive
behavior (learned), respectively.
3. The destruction of the AWB sensory neuron in one of these networks abol-
ished the naïve attractiveness and learning.
4. The destruction of RIA interneuron in the other network abolished the
learned preference and learning but did not affect the naïve attractiveness.
(Abstracted and paraphrased from Ha et al. 2010)

To cognitive neuroscientists, this work is of special interest because it is one

of the relatively rare instances in which a model neuronal network has been
directly manipulated and then changes in an animal’s behavior measured. Of
course, it cannot be a complete story since changing behavior from attraction
to aversion is only one of the functions that must be carried out by even as
simple a creature as C. elegans to guarantee the survival of its species. Other of
the 320 neurons in its tiny body have to be involved in other behaviors. Fur-
thermore, there are about 24 chemosensory neurons that are not involved in the
particular behavior Ha and his colleagues have highlighted in their important
and ingenious research. How far toward cognitive neuroscience we can search
for a full decoding of even this intermediate example of a neuronal network is
yet to be determined.

4.5 Drosophila melanogaster and Courtship Singing

No discussion of model organisms can be made without at least a brief mention
of Drosophila melanogaster, the common fruit fly. This tiny creature was probably
the first intentional example of the use of a simple organism to serve as a sur-
rogate of more complicated creatures. Its first use as a model is attributed to
Thomas Hunt Morgan (1866–1945) who showed that chromosomes were the
vehicles for the transference of genetic information from generation to genera-
tion. He was able to accomplish this feat (for which he, too, won the Nobel
Prize) because the genetic structure of D. melanogaster was much simpler than
that of higher organisms—it consisted of only four chromosomes compared with
the currently accepted human count of 46. Many of the genes on these four
chromosomes have been associated with specific behavioral, chemical, and ana-
tomical aspects of this simple creature. By manipulating the genotypes it has also
been possible to associate genetic properties with observed behavior, thus making
it relevant to cognitive neuroscience.
However, our interest is in neural theories and the literature describing the
use of D. melanogaster in neuroscience is sparse. The reason for this lacunae is
that the nervous system of this fly contains many more and much smaller neurons
than are found in the L. polyphemus eye (~1000), A. californica (perhaps as
many as 300 are involved in some way in the gill withdrawal response), or C.
elegans (320). Estimates are that the D. melanogaster nervous system contains
 Intermediate Level Neuronal Networks 135

about 135,000 neurons compared with these other model animals. This number
currently exceeds the threshold for the kind of neuron-by-neuron analysis that
was possible in the other animals. Current efforts to map out the nervous system
are limited to a few “compartments” and only 16,000 of the 135,000 neurons
of the fruit fly’s nervous system have been identified (Kohl and Jefferis, 2010;
Chiang et al., 2011). To accomplish even this “coarse” and partial wiring diagram
required that Chiang’s group dissect more than a million fly brains! Clearly, this
type of complete neuronal wiring diagram is probably not going to be possible
for an organism as complicated as that of D. melanogaster, much less that of the
human brain in the foreseeable future. Furthermore, even if we had a complete
wiring diagram, the technical difficulties involved in repeatedly stimulating and
recording from a specified neuron (or group of neurons) in the fly are formi-
dable. The neuronal network of D. melanogaster clearly establishes a high limit
for what we have designated as intermediate levels of complexity.
It is also important to remember that in the long history of using D. mela-
nogaster as a means of studying genetic effects on anatomy and physiology,
cognitive neuroscientists are doing something quite different when comparing
behavior and genetic information. Olsen and Wilson (2008) clearly express this
caveat in the following terms:

Many classical Drosophila behavioral paradigms were designed to screen

many flies simultaneously for profound [and easily observable] effects. But
neuroscientists are increasingly interested in fly behavior for its own sake,
rather than simply viewing it as a tool for isolating genetic mutations.
(p. 512)

In other words, behavior is not nowadays being used as a tool for studying
chromosomal effects; rather the goal is to explain the neural foundations of overt
behavior. There is an important difference between studying the genotype and
the neural or behavioral phenotype; the success that has been forthcoming in
the genetic studies sometimes leads us to overstate our accomplishments in the
neurophysiological domain. Furthermore, many of the properties that were
advantageous in the other model preparations (e.g., small number of neurons,
simple coding schemes, isolatable and identifiable neurons, or simplifying con-
siderations like the regularity of the ommatidia anatomy in L. polyphemus) no
longer apply.
Despite these disadvantages, some impressive neuroscientific work relevant to
behavior has been done using D. melanogaster. The fruit fly has some advantages;
for example, it has a wide variety of easily observed behaviors—eating, navigat-
ing, reproduction, etc. A major advantage (countering the increased number of
neurons and anatomical uncertainties about its brain) is that the fruit fly has a
much wider array of behaviors for which neurophysiological correlates can be
sought than is available in some simpler animals. The price one pays for this
136 Intermediate Level Neuronal Networks

expanded behavioral array is that the fly’s nervous system is also more compli-
cated and the difficulty in determining which neurons mediate which behavior
has proven to be a serious constraint on the use of D. melanogaster as a model
preparation in neural and behavioral comparisons.
The initial task, therefore, in planning an experiment to study a neural behav-
ioral relationship is to determine which neurons of D. melanogaster’s 135,000
neurons participate in a given behavior. Olsen and Wilson (2008) proposed three
general methods to link particular neurons to particular networks:

1. Anatomical Method: Using a genetic manipulation to make a small number

of neurons fluoresce when stimulated with light. The fluorescing neurons
are assumed to be associated with an otherwise unknown set of interacting
neurons.
2. Behavioral Method: Genetic manipulations may be used to selectively
“silence” certain neuronal networks. If an associated behavior is “defective,”
the silenced neurons are assumed to be associated with the behavior. (Con-
versely, Flood, Gorczyca, White, Ito, and Yoshihara, 2013, have also developed
methods for activating specific interacting neuronal networks. In this case,
neuronal networks are activated rather than suppressed.)
3. Physiological Method: Presynaptic neurons are stimulated. The idea is that
only synaptically connected neurons respond. The response in this case is
activation of a calcium indicator (G-CaMP) in neurons that are functionally
interconnected with the stimulated presynaptic neuron.
(Paraphrased from Olsen and Wilson 2008, pp. 513–515)

Using combinations of techniques such as these, it has been possible to make

some limited progress in linking specific neurons with specific behavioral com-
ponents in D. melanogaster. One of the most interesting experiments is a study
of the control of the male courtship song of D. melanogaster carried out by
von Philipsborn et al. (2011). Not surprising to any of us that have heard a fly
buzzing around, this “song” is a well-developed behavior. It is, however, surpris-
ing that fly singing is controlled by a relatively small interconnected group of
neurons that deals with different components of the song. One of the components
necessary to evoke the male’s song is a sensory signal from a female. In the
absence of such a signal, it proved possible for von Philipsborn and her colleagues
to induce a fly to sing by thermal stimulation.
The fly’s song is a complex of a number of different motor responses. Two
components of a “pulse song” include a sequence of from 2 to 50 pulses and
a “sine song” consisting of a 140–170 HZ tone that together make up the
neuronal foundation of a “decision” to sing (von Philipsborn et al., p. 509).
Five classes of neurons, all of which are genetically controlled by a single gene
designated as frum, have been associated with the expression of the male phenotype
of singing. Although these are not individually identified neurons (as many as 2,000
 Intermediate Level Neuronal Networks 137

neurons may be involved in controlling this behavior), it has been possible to make
some associations between groups of neurons and specific components of the song.
The five types of neurons and the role each plays in the courtship song are discussed
by von Philipsborn et al. and I briefly summarize the findings here.
Two of the five classes of neurons (P1 and plP10) are not involved directly
in motor control of the singing behavior; instead, they seem to control the deci-
sion whether or not to sing. Neuron class P1, von Philipsborn and her colleagues
asserted, was the key to triggering the pulse song—an analog of “decision mak-
ing.” It possessed a rather short axon and was limited anatomically to the region
of the brain. Similarly, plP10 neurons were also involved in the decision to sing.
The suggestion was that P1 was an input to plP10, conveying information from
a distant ganglion since it (plP10) had a very long axon that terminated in a
distant region of the brain.
The three other neuron classes that were involved in the fruit fly’s courtship
song—dPR1, vPR6, and vMS11—coexist in the thoracic ganglion of D. melano-
gaster and may, according to von Philipsborn and her colleagues, collectively
encode the program controlling the specific properties of the song. Figure 4.5
is a theoretical diagram proposed by them for the connectivity pattern of the
salient neurons in a network of neurons for the control of courtship singing.
The array of challenges involved in carrying out an experiment such as this is
impressive.
It is also important to appreciate that the theory shown in Figure 4.5 is of a
much more conjectural nature than are the three models proposed in the earlier
sections of this chapter.

4.6 Interim Summary

This chapter has reviewed an intermediate level of neuronal networks based
on real neurophysiological data. Because of a variety of evolutionary accidents,
it has proven possible to use the nervous systems of simple animals (usually
invertebrates) as surrogates or models for more complex animals (usually ver-
tebrates). Models, thus, serve as simple accessible idealizations of more complex
systems. The individual utility of each model preparation is based on one
simplifying condition or another that occurred in the course of organic evolu-
tion, the reasons for which we are only barely aware. Because of the relative
simplicity of the observed neuronal networks, it has been possible to dissect
(both functionally and anatomically) the simple neural systems of the model
animals discussed here. It has also been possible to determine a modicum of
the relationship between behavior and what are usually very simple neuronal
networks. Indeed, all four of these applications of real biological model systems
end up discussing networks that are almost trivial, thus suggesting that the
direct study of real neurophysiological networks is not likely to be a promising
strategy.
138 Intermediate Level Neuronal Networks

FIGURE 4.5 The neural network mediating singing behavior in Drosophila melano-
gaster. Five classes of neurons (not individual neurons) are involved. P1 and plP10
neurons control the decision to sing and dPR1, vPR6, and vMS11 control the tem-
poral characteristics of the song.
Reproduced from von Philipsborn et al. (2011) with permission.

A prevailing assumption guiding this work is that these model systems are
similar enough behaviorally to the behavior of higher-level organisms to at least
justify drawing functional inferential analogies between the model and the mod-
eled and, at best, assert that we have found the actual neuronal mechanism of
what are common behaviors.
This line of research, however, is constrained by the fact that although the
observable behaviors may seem to be the same, they may be actually instantiated
by very different mechanisms in the model and modeled organisms. For all
practical purposes these are bottom-up approaches to the role of networks and
this type of research using conventional neurophysiological techniques fails for
networks consisting of more than a few neurons.
Throughout this discussion, the limits of this direct neurophysiological
approach have been apparent because the equivalence of the mechanisms under-
lying the model and the modeled behavior, respectively, remains uncertain.
Although philosophers may blanch at the idea (referring to the logical fallacy
of assuming something is true because it is possible), neurophysiologists will
appreciate that there are innumerable, quite distinct mechanisms that could
produce the same behavior. This harks back to the problem of underdetermina-
tion of the underlying mechanisms if we can observe only the behavior.
Among many other conceptual problems with this direct approach to study-
ing neuronal networks, it appears that many investigators who used the model
preparation approach seem to assume that not only do behavioral analogies
exist between higher and lower animals, but also mechanistic identities in the
form of neurophysiological homologies. Unfortunately, even the most robust
 Intermediate Level Neuronal Networks 139

data says nothing about the comparable neuronal network in the higher, mod-
eled animal. That there are alternative possible and plausible neuronal network
mechanisms that could produce responses in different animals based on totally
different neuronal networks seems likely; indeed, our intuition tells us that
there must be such alternative explanations. The advantage of this bottom-up
approach is that it is anchored to neurobiological reality—the mechanisms
described to explain behavior are based on real, tangible, and observable
mechanisms.
An alternative and sometimes useful strategy to understand relatively simple
intermediate neural networks is to work from the top down—to infer the possible
nature of the neural network from an organism’s behavior. Unfortunately, most
such theories are functional inventions comparable to those operating at the psy-
chological level. Models such as these are exemplified by the “biologically inspired
cognitive architectures” (Goertzel, Lian, Arel, de Garis, and Chen, 2010, p. 30).
The important thing to remember about such top-down contributions is that
they are not initially based on neural or neuronally based empirical data; they
are functional mathematical descriptions in which the starting point is behavior
and the end points are plausible mathematical and neural constructs—in other
words, hypothetical constructs. The mathematical representation is invented to
subsequently complete the description by showing a plausible manifestation of
the behavior in neural terms—hypothetical neural mechanisms that could imple-
ment the functions described by the mathematics. These theories are not neu-
rophysiological evidence-based but are top-down theoretical statements based on
behavioral observations rather than neurophysiological ones.
A final and very important finding from the direct neuronal approach is that
none of the methods or approaches discussed here was really successful at unrav-
eling the details of a network. Limulus polyphemus permitted us to examine
results distributed over a few receptors but that had to be simplified by the
regular, almost crystalline, structure of the animal’s eye. In the other experiments,
it was not the network interaction between neurons that was studied but rather
the role of specific and individually identified neurons. The “networks” studied
were, at best, interactions between a few neurons. Interactions were just too
complicated for our instruments, our analytical techniques, and our basic assump-
tions to tease out the details of neuron networks whose number exceeded the
fingers on our hand. In short, only trivial “model” networks could be examined.
Multiple electrode arrays could not reach our goal, nor would limiting ourselves
to what were supposedly simplified (to the point of triviality) model prepara-
tions. Nor, for that matter would the construction of hypothetical entities inferred
from top-down methodologies.
In conclusion, the use of model preparations with simple nervous systems has
allowed us to unravel systems with a few neurons. However, this direct neuro-
physiological approach is severely limited to networks made up of very few
neurons and cannot be expected to solve problems involving realistic networks.
140 Intermediate Level Neuronal Networks

What seems clear is that a direct neurophysiological approach to unraveling

neuronal networks is not a promising approach. Although it is likely that we
will learn much about neurons and synapses, neural networks will remain inscru-
table from this point of view.
Looming on the horizon, however, was a new development that promised
to break through the conceptual and technological limits I have discussed in
this chapter. The development of the digital computer seemed to provide an
ideal means of dealing with complex networks. The next chapter reviews the
impact of these extraordinary devices on theory development in cognitive
neurosciences.

Notes
1. It is likely that the phenomenon had been known years before. Edge enhancement
occurs in the work of artists who antedated Mach by many years.
2. By symbolic, I am suggesting nothing more than that there might be neuronal mecha-
nisms that are much more complex than the simple lateral inhibitory interactions
accounting for the phenomenon in the human brain. We cannot exclude alternative
explanations with the kind of experiments we are able to carry out in L. polyphemus—
nor can we for any of the other model preparations discussed later in this chapter.
5
LARGE-SCALE COMPUTER
SIMULATIONS (THEORIES)
OF COGNITION

5.1 Introduction
In the previous chapter, I discussed an approach to studying neuronal networks
that emphasized the examination of real, but small, neuronal networks using
direct neurophysiological methods. The history of this bottom-up method makes
one point very clearly: the complexities of even a very small network (consisting
of from one to perhaps a dozen neurons) were such that any hope of under-
standing how a network worked with a conventional neurophysiological approach
was not likely to be fulfilled with the tools that were available. It was only special
cases (such as the Hartline Lab’s work on Limulus polyphemus) in which certain
simplifying constraints (repetitive regularity) could be invoked that permitted
larger numbers of neurons to be considered. Even then, traditional neurophysi-
ological methods do not seem promising—there are just too many neurons for
standard analytical techniques to work.
Currently, however, new tools are becoming available that promise to change
not only our experimental methodology but the concepts and assumptions that
are certain to guide theory in particular in the years to come. Some of the most
influential are certain to be modern versions of the digital computer. In principle
and a priori, the tool “par excellence” for studying microneuronal networks is
the digital computer, but not just the simpler machines of the class sitting on our
desks. They may be useful for studying the intermediate level of neural networks
of the kind described in Chapter 4. Instead, computers are needed that have the
ability to manipulate huge numbers of computational elements, each one of which
is capable of representing the activity of a single discrete element—an element
such as a neuron—and to store astronomical amounts of data.
142 Computer Simulations of Cognition

The potential solution to the challenges of numerousness and complexity in

the microworld is a new kind of computer—the massively parallel system in
which very large numbers of basic computing elements are physical or programmed
mimics of neurons. The goal is to use a system such as a giant simulation device
to study the collective information-processing properties of many neurons and
to determine if such a simulation compares reasonably well with some aspect of
cognitive behavior. Indeed, the microneuronal approach is often not even a
mathematical or computational one—the “computer” is used as a simulator or
model of a complex network instead of a formal mathematical engine. The details
of this approach will become obvious as we review the relevant literature con-
cerned with “supercomputer-based” theories and note the paucity of ordinary
mathematical formulations used by their authors. Instead, the reports are filled
with discussions of the properties of the networks that are constructed on them.
The hope is that by understanding how simulations or emulations of large net-
works work, we may understand comparably sized networks of neurons.
Let’s not get carried away, however, at this point. The goal of understanding
the microneuronal network basis of the human mind is extremely ambitious;
recall that the number of neurons in the brain is estimated to be about 86 bil-
lion. It would not be possible to pursue the goal of decoding the brain without
the recent development of supercomputers that have the capacity to represent
very large numbers of “neuromimes” and the operation of the even larger
numbers of synapses estimated to participate in cognitive processes. Fortunately,
as we see shortly, considerable progress has been made in engineering supercom-
puters of the kind that may be useful in Micro Connectionist Theory (μct)
research. While it is certain that we will develop new computer technologies
and programming techniques, it is no way nearly as certain that a compelling
theory of mind–brain relations will emerge from the new kind of research that
has been enabled by these primarily engineering achievements. As we see at the
dawn of supercomputer research aimed at cognitive neuroscience goals, there is
considerable controversy concerning the eventual outcome of this new approach.
For example, philosophers (e.g., Searle, 1997) express considerable doubt that a
large-scale μct simulation, however behaviorally successful, would actually solve the
most challenging problems of all—problems such as “can a simulation be conscious?”
The problem is that the meaning of a “successful” simulation or explanatory theory
of mind–brain relations is intertwined with some of the most important questions
of human existence. It is for this reason that the goal of solving the mind–brain
problem is of such widespread interest, even before we are able to determine what
a “solution” would actually look like in a world in which we do not yet have a
good definition of what mind or consciousness or cognition really are.
Thus, we have several goals in this chapter. First, I discuss the digital computer
technology. I placed this brief introduction to this kind of technology at the
beginning of this chapter because unlike some of the others discussed in previous
chapters the digital computer plays a slightly different role than, for example, does
 Computer Simulations of Cognition 143

microelectrode technology. Microelectrode technology provides a means of col-

lecting the necessary data on which to build a bottom-up theory. The digital
computer technology that is used in the theoretical applications to be discussed,
however, is more like a top-down approach in which we seek to replicate a cogni-
tive process with computer program and circuit tools. Understanding how modern
supercomputers are used as theoretical engines in cognitive neuroscience is not
directly tantamount to understanding many of the operations of the brain or solv-
ing the mind–brain problem. Indeed, as we see, most current μct supercomputer-
based “theories” of the brain are actually descriptions of the computers on which
they run rather than explicit theories of how the brain works. I begin, therefore,
with a review of the technology of these extraordinary new supercomputers.

5.2 From Computer to Supercomputer

Serial Computers
The modern electronic digital computer represents one of the most influential
technological developments in human history. Its importance ranks with the
bow and arrow, animal domestication, the wheel, the microscope, and the auto-
mobile as vehicles for not only technological but also social change. Increasingly,
in modern times, computers have affected the way we live and promise to have
even a greater impact in the future. Especially in the past half-century, computer
power seems to be evolving at a rate that could not have been imagined only
a few years ago. Computer capability has grown so much that it now possible
to dream, if not to achieve an actual simulation of human cognitive processes
at the microneuronal level—thus activating a sea-change in cognitive neurosci-
ence theory development. This is not being done at a symbolic or indirect level
of analysis at which psychology and macroscopic theories operate, but at what
many agree is the level of operation at which the brain actually accomplishes
its wonderful, albeit mysterious, conversion of the neurophysiological to the
mental. To understand the modern supercomputer fully, it is useful to give a
brief history of some of the key points in its development.
Although computing engine-like devices of one kind or another have been
used for many years (e.g., the Roman and Asian digital abacus, Napier’s Bones,
or the 17th century slide rule—an analog device) and a few mechanical calculat-
ing engines (e.g., the Jacquard loom) had been invented over the years, the
modern electronic digital computer that will dominate the following discussion
was first conceptualized and built by Atanasoff and Berry at Iowa State University—
a usually overlooked history told by Burks and Burks (1989). The Atanasoff-Berry
machine was an all-electronic, special-purpose, binary device capable of solving
linear equations by numerical methods—the first of its kind. However, it had one
critical missing feature: It could not be rapidly reprogrammed as it was hardwired
to perform only a single function.
144 Computer Simulations of Cognition

The critical idea—the storage of a sequence of instructions in the same

memory as the data being processed so that both could be accessed and manipu-
lated by the program—is attributed to both von Neumann (1945) and Turing
(1950, 2004). Turing’s design, although subsequent to von Neumann’s, was
considerably more detailed. Their joint breakthrough meant that a computer
could rapidly (even during the course of a computation) alter its function. In
this manner the computer became “general purpose” and metamorphosed into
the universal information-processing machine that now underlies so much of
modern science. The extraordinary idea underlying the great success of the
digital computer is that decisions can be made by instructions that enable the
programmer to choose the next instruction to be executed. Thus, what happens
next could be contingent on what happened in the past. From this basic concept
of contingent program executions ultimately came enormous powers of com-
putation, simulation, and control. The von Neumann “computer,” as it came to
be known, thus became a general purpose device that belies its common name.
As we see later, the supercomputer is only partially used as a computer; its role
in cognitive neuroscience theory is overwhelmingly noncomputational, simulating
neuronal networks and functions rather than arithmetical operations.
The von Neumann architecture, as it came to be known, has been the standard
of computer design for many years. It was typically represented in the form
shown in Figure 5.1.

CPU

Arithmetic-Logical Unit

INPUT OUTPUT

Sequence Controller

MEMORY

FIGURE 5.1 The organization of the single-core von Neumann computer composed
of input device, output devices, memory, a central processing unit (CPU) that consists
of an arithmetical-logical unit, and a program sequence step controller. Parallel
processing computers may have many of each of these units connected in a variety
of ways.
 Computer Simulations of Cognition 145

The block labeled INPUT consisted of any number of different devices that
are able to take the information stored on some physical medium and enter them
into the memory of the computer. Punched cards, punched paper tapes, teletypes,
keyboards, cameras, and, nowadays, such devices as high-capacity “thumb” memories
or highly sensitive cameras have been used as INPUT devices. Output devices
such as mechanical printers or cathode ray tubes were among the early means of
communicating the results of computations to the outside world. Nowadays such
exotica as three-dimensional printers have begun to be used as output devices.
The central processing unit (CPU) was the computational device that accessed
input data, manipulated it, and prepared it for presentation to the output devices.
The CPU is often depicted as composed of two parts—an arithmetical-logical
unit that actually executes the stored instructions and a control unit that tells
the computer what instructions should occur next as it sequences the various
steps of a program. Another name for the CPU was the “core,” a term that
rose in popularity as the older notion of a simple, single CPU began to be
replaced by parallel processing systems that had multiple CPUs or specialized
data manipulation devices that were quite different than the conventional single
CPU of the classical von Neumann computer.
Memory technology evolved over the years from a novel use of Crookes
tubes, to small magnetic spots on a drum, to vacuum tubes, to discrete solid-state
circuits, to a huge variety of highly miniaturized and integrated electronic circuits,
capable of rapid reading and writing of vast amounts of stored information. The
development of fast printed memory circuits that could stably preserve informa-
tion with or without power has been one of the grand successes of computer
memory technology. Where originally a couple of thousand volatile memory
locations constrained the size of programs, modern memories consist of gigabits
of stable and rapidly accessible memory locations.
Early computers based on vacuum tube technology or early solid-state circuitry
were ponderously slow by today’s standards. They were all serial machines in
which each instruction or computer access had to be completed before the next
one could be executed. The main practical strategy to speeding up the execution
of a computer program in the early years was to improve the speed at which
memories could be accessed and information processed by the CPU. This led
to an emphasis on engineering development of faster and faster technology (i.e.,
the electronic components); however, the basic organization plan of the von
Neumann machine remained more or less constant, in general, if not in detail.
Eventually, physical limits of heat generation and disposal and the basic physical
limits on electrical conduction time (i.e., speed of light considerations) placed
limits on how fast circuits in this kind of computer could function.
Refrigeration helped in part to overcome heat problems and new semicon-
ductor materials helped to speed up computers, but eventually the physical limits
encountered with very small circuits seemed to place limits on how fast a serial,
single CPU computer could operate. Nevertheless, the first few generations of
146 Computer Simulations of Cognition

computers were all serial computers with a single CPU that was driven to the
physical limits of available circuit design speeds. This did not mean serial com-
puters were paralyzingly slow for their times; very large systems (e.g., the Cray-1
of 1976 with only a single CPU-dominated computer system speed until the
1970s) could accomplish much by concentrating on speeding up a single CPU.
We can sum up the extraordinary properties of digital computers by listing some
of the most significant ideas that emerged during the past half-century that influ-
enced the history, design, and applications of modern computer engineering:

1. Both program and data were stored in a common memory.

2. Random addressing of data and instructions was possible.
3. The program was accessible during its execution.
4. Programs and data could be modified during program execution.
5. This resulted in the ability to make decisions and modify the thread of the
program sequence.
6. Both logical and arithmetical operations could be carried out.
7. Vacuum tubes and then solid electronic technologies permitted increasingly
fast operation—but only up to a point determined by physical laws.

The physical limits on how fast a serial computer could process information
were encountered just as the real world was demanding solutions to problems
that required much faster computing speeds. Weather forecasting, molecular
interactions, and a host of other civilian and military applications arose that could
not be satisfied by the fastest available conventional single-CPU computer. The
engineering solution that finessed the physical limit barrier was based on the
concept of parallel processing—an idea that leapt into prominence in the 1980s.
This is the topic of the next section.

Parallel Processing Computers

Many of the limits of serial, single-CPU computers were overcome to a con-
siderable degree with the advent of multiple core or parallel processing computers.
When it became obvious that the intrinsic speed of a circuit could not be pushed
beyond fundamental physical limits, an appreciation emerged for how the effec-
tive speed of a computer with multiple CPUs operating in parallel might over-
come these limits.

A parallel processing computer is a device with multiple CPUs or cores so

that multiple parts of a complex task can be carried out nearly simultane-
ously. Success with parallel processing depends on the task being partition-
able into subcomponents.
 Computer Simulations of Cognition 147

The idea of parallel processing as a means of speeding up computer process-

ing had been around almost as long as the idea of the programmable electronic
digital computer. Surprisingly, one of the earliest computers, the Eniac (Goldstein
and Goldstein, 1946), had some parallel processing capability in the form of
several adders that could function simultaneously.
The idea of parallel processing, however promising it seemed to be, remained
of secondary interest to computer engineers for many of the succeeding years,
with the search for ever faster serial circuity dominating engineering thinking.
The circuit design technology was advancing so rapidly that there was only an
occasional need for seeking alternative approaches to improving computational
speed. Even more inhibiting was a kind of psychological barrier among engineers—
the notion that the ultimate speed limit was not going to be hit for many years.
Increasing speed was associated with increasing circuit density and it had been
predicted that circuit density was going to double every two years (Moore, 1965);
empirically this seemed to be true. “Speed up,” therefore, seemed unlimited into
the foreseeable future as new materials and exotic physical principles were
explored on which to base faster and faster circuitry. Occasionally (as noted by
Gill, 1958, in one of the first articles specifically dedicated to a discussion of
parallel computation), a primitive form of parallelization was built into a current
computer design in the form of auxiliary devices that carried out some special
computations at the same time as the single core was carrying out its computa-
tions. But, in the main, parallelization was not of high interest until the 1960s
when several companies began to develop computers with a few duplicate cores
or auxiliary arithmetical processing units. In addition, considerable concern was
raised about the predicted difficulty in programming these multiple core systems.
The goal of writing software to automatically partition a problem so that its
parts are suitable for simultaneous execution and designing software specialized
for parallel processing remains highly challenging.
By the end of the 1960s development of minimally parallel computers was
off and running, with one system after another being offered commercially. It
was in 1971 when the first dual-core processor was constructed on a single chip
(by the Intel Corporation) that the massively parallel computer with many cores
became technically feasible. Progress in printed circuit technology was rapid, but
not instantaneous, and, as a result, even as late as 1984 the fastest computers (e.g.,
the Cray X-MP) still had only four cores. Whereas early supercomputers such
as the Cray-1 sought to improve computing power by increasing the speed of
the electronic circuitry of which they were constructed, the potential advantages
of multiple core systems started to become evident. From that time till now,
however, progress in parallelization has been extraordinary, with massively parallel
computers now dominating the international speed race.
The first efforts to develop commercial multi-core and distributed memory
parallel processing computers were relatively straightforward: two CPUs were
mounted on a single “chip” and the computational workload distributed between
148 Computer Simulations of Cognition

the two. Special computational units in the earliest supercomputers were added
to what were otherwise serial computers in an effort to speed up computational
throughput. Later attempts to build ultrafast supercomputers made a dramatic
change when it became possible to “print” large numbers of CPUs (the number
now approaching millions of these no longer “central” but now better described
as “distributed and parallel” processing system processing units). In recent years
it has also been possible to build virtual supercomputers (e.g., the Beowulf system
described in Becker, 1995) by taking advantage of world-wide communication
links to employ distributed networks of ordinary desk computers to collectively
solve the kind of problem that could be partitioned into separate tasks. The
fastest supercomputers, however, still strive to package many memory and pro-
cessing units in a single location; otherwise, communication times can become
troublesome. The best solution to this difficulty is to concentrate on problems
in which the partitioned components are totally independent. Unfortunately,
nature does not often serve up its problems so neatly.
During the past decades of the 20th century, the idea of designing parallel
processing computers with very large numbers of multiple cores and huge arrays
of memory storage units became increasingly popular, and substantial investments
were made by corporations and governments to engineer ultrafast computers.
Parallel processing computers have evolved into a wide variety of system types;
most fundamental differences are involved in the manner in which the multiple
cores are interconnected with one another and with the memory storage units.
Cores can be interconnected so that they are clustered together but still operate
quasi-independently. Parallel processing computers also may have a variety of
different means of accessing the memory. In the simplest cases, in which the
memory is shared by the two CPUs, both might have full access to all of the
memory, while in other designs only limited access to the common memory is
available to each core. In a system in which two or more memories are con-
nected separately to their respective CPUs, a common communication “bus” is
necessary to let the two memories connect to each other. From a primitive
two-processor, shared memory, the parallel processor has evolved a substantial
variety of computer architectures in which multiple memories and CPUs are
interconnected in complex ways by ever more complicated buses. Although it
is impossible to provide a complete list of all possible parallel processing archi-
tectures, a simple and inclusive taxonomy was presented by Flynn (1972) that
characterizes the main types. He categorized the main types but acknowledged
that there were many variations on these main architectural types:

• Single Instruction Single Data (SISD)—The typical desktop sequential

computer
• Single Instruction Multiple Data (SIMD)—One instruction affects all data
• Multiple Instruction Single Data (MISD)—Many instructions affect all data
• Multiple Instruction Multiple Data (MIMD)—The most common parallel
processor
 Computer Simulations of Cognition 149

• Single Program Multiple Data (SPMD)—Multiple programs operating

independently
• Multiple Program Multiple Data (MPMD)—Multiple programs under con-
trol of one core
(After Flynn 1972)

Speed is achieved in a parallel processing computer by adding and optimally

organizing hardware—CPUs and memory—so that separable parts of appropri-
ately parsed programs may be run as concurrently as possible rather than simply
trying to speed up the operation of individual logical components. All other
things being constant, a parallel processing computer should run faster on many
types of problems than a single-CPU system as a straightforward function of the
number of CPUs wired together in parallel.
However, there is a major problem built into the previous statement—all
other things are not constant; in fact, parallelizing a computer virtually never
speeds up program executions at the rate expected by simply counting the
number of cores. The problem emerges from a limit on how the basic advantage
of parallelicity can be exploited; namely the expectation that a program can be
broken up into parts that can be run separately and the outcome patched together
after the individual components (program parts) are all completed. The problem
has nothing to do with the hardware or even the software of the parallel pro-
cessing computer; the difficulty lies in the nature of the real-world problems
being processed. Some tasks are intrinsically parallel: each step can be independent
of previous or successive computations, that is, for some problems the order in
which the various component parts are carried out does not matter.
An example of such an intrinsically parallel task is a computer vision system (such
as a face recognition system) in which the various features can be processed inde-
pendently and then combined to recognize a face. In tasks of this kind, each pixel
(or larger area) can be examined by a parallel processing core independently to
determine the respective parameters of a “face.” The reason underlying the intrinsic
parallelicity in this case is that the order in which each region is examined is incon-
sequential; each core in a parallel processor can be assigned to process a separate part
of the image and the computational steps required for each part are independent of
the others. When all parts of the image have been processed, the entire picture can
be patched together, with only minor corrections of edge effects.
On the other hand, some problems are intrinsically serial and each successive
part of the program depends on preceding ones. It is not possible to assign partial
tasks to another CPU because the second part may depend on the evaluation
of the first. Serial tasks have to wait until the necessary precursor program steps
are completed before they can be executed. In short, each subsequent step depends
on the result of preceding steps. An example of an intrinsically serial task is the
computation of a Bayes theorem problem in which the prior values must be
known before the post values can be evaluated. There is no way to jump ahead
because subsequent steps are dependent on previous ones.
150 Computer Simulations of Cognition

These idealized examples do not tell the full story because some tasks are only
partially parallelizable. That is, in most problems, only a portion of the program
can be broken down in a way that would permit additional CPUs to speed up
the solution of the problem by bringing more raw computer power to the task.
Indeed, there is a classical relationship known as Amdahl’s (1967) law between
the maximum possible speed increase and the proportion of the program that is
parallelizable. A much simplified version of Amdhal’s law asserts that the maximum
possible speedup is less than 1/1-P for a given number of processors, where P is
the proportion of the problem that can be parallelized. Thus, the maximum pos-
sible speedup of a task is not only a function of how many processors might be
present in the computer (which in a perfectly parallel problem would be a linear
function of the number of available cores), but also of the nature of the posed
problem. As an example, if the problem was determined to be only about 50%
parallelizable, then this law would predict that the fastest parallel computer would
only be able to speed up program execution by a maximum factor of 2 over that
achievable with a single ultrafast CPU computer. This would hold no matter
how many CPUs were assigned to work on the problem.
In addition, the execution of a parallel program typically involves a number
of additional housekeeping steps that were not required in a serial computer
(e.g., assigning parts of the problem to particular subsets of the hardware and
communicating data to multiple locations). In sum, one cannot expect to reach
the theoretical maximum improvement suggested by Amdahl’s law; there will
also be some housekeeping costs to pay.
Nevertheless, with the extraordinary number of highly miniaturized cores
that are now being combined into today’s supercomputers, an enormous amount
of computing power is becoming available that could not be achieved with
simply technological (i.e., individual circuit speedup) changes. Some of the most
powerful systems claim that they are composed of multiple millions of cores
simulating tens of thousands of neurons. In addition to the economic, military,
and scientific implications of high-speed computing, there are now quasi-political
overtones and international competitions. Over the past few decades there has
been an international race to build the “fastest” computer and for each competi-
tor the strategy of choice has been to instantiate the contending computer in
the form of a massively parallel system and then to test its performance with a
standardized program, an example of which is Linpack, a program designed to
solve a “dense system of linear equations.”1
The metric of computer speed used to compare the contenders in this race
is the unit referred to as Floating Point Operations per Second (FLOPS). As a
raw measure of computer speed, the number of FLOPS is preferred to the number
of machine language program instructions carried out per second because it
represents a more realistic estimate of problem-solving ability, especially in float-
ing point’s ability to represent very small and very large numbers and fractional
values (integer arithmetic deals only with whole numbers).
 Computer Simulations of Cognition 151

TABLE 5.1 Computer Performance Metrics

Name FLOPS
YettaFLOPS 1024
ZetaFLOPS 1021
ExaFLOPS 1018
PetaFLOPS 1015
TeraFLOPS 1012
GigaFLOPS 109
MegaFLOPS 106
KiloFLOPS 103

A typical desk computer today is capable of a few gigaFLOPS (or 109,

or 10 billion floating point operations per second). Current state-of-the-art
supercomputer speeds are astonishingly fast and getting faster yearly. The top
reported speeds of the fastest supercomputers now available (e.g., the Chinese
Tianhe-2) are reported to be in the petaFLOPS range (or about 1015, or about
a quadrillion FLOPS). A next goal of participants in the international competition
is the construction of a computer capable of exaFLOPS, or 1018 floating point
operations a second (a number for which we do not yet have a name, but which
adds a certain informal substance to the word “zillion”). Table 5.1 summarizes
the vocabulary used in quantifying computer speeds. The exponential growth
in computer speeds in the past few years suggests that we might see a yetta-
FLOPS computer in our lifetime, probably instantiated in some new hardware
technology.
Whatever computer emerges that might serve as a test bed of an overarching
theory of mind–brain relations, it is clear at this stage that even these high speeds
will be stressed to their limits.

The Role of Parallel Processing in Cognitive Neuroscience

As modern neurology, neurophysiology, and many other theoretical and clinical
investigators gathered more and more knowledge about brain activity, it became
clear that mental activity is the result of brain activity or, in some more subtle
manner, is brain activity. The correlations, if not the identity of brain activity
and structure, on the one hand, and cognitive activity, on the other, have become
increasingly obvious as decades of research and clinical experience provided a
wealth of correlations between the two domains. Nevertheless, efforts to explain
the specific relationship between brain and mind (in all of its ramifications)
continue to fail for reasons that have occupied philosophers and theologians
for millennia as well as psychologists and physiologists for centuries. The evo-
lutionary emergence of human cognitive information processing, much less the
152 Computer Simulations of Cognition

inscrutable and inaccessible process we call consciousness or sentience, remain

mysterious and covert.
Why should this be the case? One answer to this question lies in the com-
plexity of both the brain and the mind. As noted in the preface to this book,
somewhere in the neighborhood of 86 billion neurons and a thousand times as
many synaptic interconnections make up the great neuronal networks of the
brain. Today, most cognitive neuroscientists agree that it is in the actions of these
networks—at the level of the neuron—that mind is encoded. That is, network
operations executed at the neuronal level result in or become mental activity.
However, the details of these network operations remain recalcitrant to theoretical
explanation—there are just too many neurons and too few tools that seem to
offer any hope of bridging the gap between the mental and the neuronal. As a
result of this discrepancy between neuron numbers and analytical and techno-
logical tools, there was and continues to be a very sparse sample of satisfying
mind–brain explanatory theories. Those that have been offered, such as the
widely accepted Hebb conjecture, consist of some highly plausible conjectures
that cannot be tested and remain speculative (with a few recent exceptions).
However, as just presented, the past few decades have seen an extraordinary
technological development—massively parallel, multi-core supercomputers with
enormous numbers of cores and memory storage units, which were described
in the previous section. The parallel processing supercomputer represents the
most recent technological development in the string of influential technologies
that have successively guided cognitive neuroscience thinking. For the first time,
however, there is the possibility that we might be able to study the operation
of the brain or brain-like mechanisms at the microneuronal level of analysis at
which cognitive processes are thought to be actually encoded. It is by no means
certain that the progress we hope for will actually happen, but for the first time
computer technology and brain science seem to be converging on a common,
hitherto inaccessible level of analysis. For this moment in scientific history at
least, there is a possibility of progress in mind–brain theory. Indeed it may not
just be progress, but actual solutions.
However, this is still only a possibility and formidable obstacles lie between
today’s conjectures and the ultimate answer to how the brain produces mental
activity. However, there are no guarantees that the “problem” will be “solved”
by this new approach. It is not even clear what would be required to declare
the “problem” solved. It is a new frontier of mind–brain science on which we
have embarked in the past few years and where it will ultimately go is uncertain.
Of one thing, we can be assured—this will be an exciting adventure.
The main reason that optimism is widespread these days is that there is a
natural conceptual relationship between the modern supercomputer and the brain
that did not exist with previous technologies. Both are made up of extraordinarily
large numbers of basic units whose parallel processing is the main source of their
joint power to manipulate information. Although the number of cores in even
 Computer Simulations of Cognition 153

the largest parallel computer still does not come close to the 86 billion neurons
of the human brain, it has been possible to use a smaller number to simulate
numbers approaching that of the brain by means of iterative programming.
Therefore, there is at least a superficial comparability that suggests that the basic
functionality of the brain might scale down to that of the best modern comput-
ers or the computer might scale up to the level of complexity of the human
brain. The current world speed holder—the Chinese Tianhe-2 supercomputer—
consists of approximately 3 million separate cores of several different types. IBM’s
Sequoia supercomputer, the record holder in 2012, was built from 1.5 million
cores.
It is important to distinguish among the number of electronic neurons, cores,
and electronic circuits in a supercomputer simulation of a neuronal network. At
the bottom of the hierarchy is the number of electronic circuits or gates. The
number of these electronic circuits can be very large. Modern printed circuity
has reduced the size of a binary flip-flop (the basic logical circuit) to molecular
levels. Millions of these circuits are present in the most advanced supercomput-
ers. On the other hand, a much smaller number of cores or CPUs are constructed
from these basic circuits. Several million equivalents of cores seem to be the
frontier currently. In turn, a number of cores must work together to represent
a neuron so that a neuronal network can be simulated. About a million synthetic
neurons is the best so far achieved. What is clear is that the number of actual
physical neurons in the synthetic network is still significantly smaller than in the
real biological brain and thus, repeated cycles of program execution would be
required for the computer to run as fast as the brain.
Thus, there still remains a scale difference between the number of neurons
in the brain and the number of cores in the most powerful supercomputers. It
is not certain whether this difference will be significant or whether computers
can be scaled up to the count of biological neurons. The important thing is
that the past decade is the first period in cognitive neuroscience history that the
basic organization properties of the technological tools available to study the
brain and those of the brain itself are commensurable. The von Neumann serial
computer under control of a programmable sequencer and a single computational
element is fundamentally different in this regard from how the parallel processing
computer and the brain operate. It was a poor model on which to base theory
development, not only because of its serial nature, but also because of the dif-
ference in how the fundamental serial concept is embodied in the hardware.
Although the degree of parallelicity is still far greater in the brain than the best
computer, at least the basic principles of operation are now apparently converg-
ing. The basic concept defining the organization of a parallel processing computer
and of the brain match much better now than with any previous technology.
Computers and the brain share another basic similarity that makes the parallel
processing computer nearly the ideal tool for studying the brain. Unlike the serial
computer in which each instruction is controlled in sequence by the commands
154 Computer Simulations of Cognition

of a “program,” the parallel processing computer can, in principle, be far more

autonomous. It is not necessary for a “program” of instructions to exist once the
system has been set up; the momentary initial state of each of the simulated
neurons in a parallel system of interacting cores can act as an initiator of activity.
Thus, all the system needs to start running is an impetus—–turning on the power
or providing a metabolic source of energy might be all that is needed. Then, in
accord with a set of rules of interneuronal interaction, it will continue to run
out its normal course of action until it arrives at some final outcome determined
by the interconnection pattern. The brain must be acting very much like this in
basic principle; an initiating condition or state or stimulus triggers the action of
a vast network of neurons that runs continuously until a behavior is selected and
executed or the metabolic power source is discontinued. In neither case need
there be an external program guiding the activity. Instead, there is just a sequence
of complex parallel states that are subject to rules of interaction.
The main objection to the universality of the autonomous “free running”
parallel processing computer is that the electronic parallel processing computer
must be set up with the rules of interaction and the initial states of the neurons.
However, this difference is secondary—once the states of the simulated neurons
are established, the parallel processing computer and the brain operate in very
much the same way, as a network of nodes interconnected by communication
links whose activity is driven by the momentary state of the system rather than
by a sequence of commands from some external source.
Another similarity between parallel processing supercomputers (and not the
basic von Neumann computer) and the brain is that the information is processed
in both systems by being passed simultaneously from one neuron, electronic or
neurobiological, to another by multiple tracts. Feedback, feed forward, and lateral
excitatory and inhibitory processing are all found in computer simulations as
well as in natural brain organization. Regardless of the specific organization of
a particular parallel processing computer, the modern supercomputer system is
designed to achieve its speed or power by the simultaneous executions of many
operations rather than by speeding up the action of one CPU.
On the dissimilarity side of this thesis, there is a vast difference between the
instruction execution speed of a neuron and that of a typical solid-state core.
Individual neurons are outrageously slow in the present context, operating on a
time scale of milliseconds. Communication speeds along an axon are measured
in tens of meters per second and there are additional delays inserted by synaptic
connections. On the other hand, computers—even the most modest desk
computer—operate at clock speeds (to choose one way of quantifying computer
speed) measured in gigahertz (billions of cycles per second). Even so, the key to
this kind of speed enhancement in the brain as well as in the most powerful
supercomputers is massive parallel organization of multiple components.
In sum, parallel processing is a foundation organization principle of both
modern computer engineering and how the brain has evolved. Parallelicity is
 Computer Simulations of Cognition 155

the basic concept that permits the brain to carry out operations that would have
been impossible with the intrinsic slow speed of the neuron. Single cell theories
of cognition are, from this point of view, simply incorrect at the most basic
conceptual level. It is not possible for any single neuron to carry out the
information-processing speed represented by our amazing behavioral abilities
without massive parallelicity. The hope is that the new technology will open
the door to future understanding, perhaps even crack the mind–brain problem.
However, there are formidable obstacles to the construction of an overarching
theory of how the brain produces the mind.
One such obstacle is that even the most advanced supercomputers are cur-
rently limited in the number of individual CPUs that have been joined together;
the 2 or 3 million that have so far been combined are still far short of the
number of neurons in the brain. Complexity increases with numerousness,
therefore, there is no assurance that the emergent properties we designate as
cognitive will appear with the supercomputers that are available at the moment.
It is likely, if not certain, that we may still not be at the appropriate level of
analysis and may still have to depend on future supercomputers with many bil-
lions of synthetic neurons before we can claim a solution.
Another obstacle is the problem of interpreting the results—when will we
know that we have built the desired overarching theory of mind–brain equiva-
lence? The output of any system, brain or computer, is presented in some form
of behavior—for example, a successful pattern recognition. It is possible, however,
to produce behaviors by several different means (where the word “several” may
indicate very large numbers; see the discussion of Hilgetag, O’Neill, and Young,
1996). We call these behaviors that are produced by different mechanisms and
processes “analogs” of each other. The problem is that the externally observed
behavior of a person or a computer is underdetermined; there are not enough
data in the behavior to determine which of “several” possible mechanisms
accounted for that behavior. Both computers and people can be directed to
produce particular behaviors in innumerable ways.
How then do we determine that a particular underlying mechanism (a reduc-
tive theory) is actually explaining the behavior in humans and not just simulating?
The answer to this rhetorical question is that there is no way of choosing among
the analogs. However accurate a theory may be in producing a particular behavior,
it may do so by mechanisms that are not only completely different than those
used by the brain but may also be inaccessible to the most powerful of our tools.
Recognizing a face, for example, can be accomplished by a computer by measur-
ing the connecting lines between a set of key points (e.g., the tip of the nose)
and then creating a numerical vector associated with that particular face. Humans
also recognize faces but it seems extremely unlikely that we do so by comparable
numerical calculations and measurements. Rather, cryptic (so far), organizational
processes involving molar spatial relationships of the kind studied by the
nearly forgotten Gestalt psychologists seem more likely mechanisms of
156 Computer Simulations of Cognition

face recognition. In principle, therefore, the actual neuronal underlying mecha-

nisms of face recognition, like many other cognitive processes, may remain
uncertain if all we have is behavioral data and knowledge of how large-scale
parallel processing computers work.
Again, it cannot be stated too often that the availability of a new tool such
as parallel processing supercomputers does not guarantee success in the search
for understanding; however, it does offer an attractive promise that was not
available until now. How far we can go in solving the mind-body problem using
this powerful new tool remains to be seen.
Now let us briefly consider some of the parallel processing theories of cogni-
tive processes.

5.3 Parallel Processing Theories of Cognitive Processes

Some History and Background

The foundation concept that the cognitive functions of the brain are embodied
in a parallel processing network predated the availability or understanding of
parallel processing computers, and indeed of computers in general. What was
probably the first explicit model of a parallel processing network in the nervous
system was the work of Pitts and McCulloch (1947) reported in their classical
article on “universals.” Among the earliest reports in which parallel processing
networks were studied neurophysiologically was that by Hartline, Wagner, and
Ratliff (1956) concerning the fortuitously simple organization of the horseshoe
crab eye. (See the discussion of this research in Chapter 4.) As we saw there,
because of the large size of the receptor cells in this compound eye, they were
able to neurophysiologically examine the lateral inhibitory interactions between
the neurons located in the horseshoe crab’s ommatidia and explicitly evaluate a
real neurophysiological parallel processing system. Hartline and his colleagues
demonstrated that this primitive form of parallel processing could neurophysi-
ologically mimic the human perceptual phenomena of Mach Bands—the
enhanced edges that appeared despite the absence of any corresponding luminance
differences in the stimulus. Although a few computer scientists who had access
to computers subsequently modeled this process, to the best of my knowledge,
no reports of a computer simulation of the horseshoe crab eye were published
during this early period. Nevertheless, the fundamental concept of parallel pro-
cessing was established by these pioneering studies; parallel processing became a
part of the theoretical culture that followed.
The idea of parallel computations as the basis for cognitive information
processes such as learning took a major leap forward with the work of Rosen-
blatt (1958). He invented a device designated as a “perceptron” that was
essentially a parallel processing system consisting of a small number of elements
that could learn to discriminate between different stimulus patterns. Rosenblatt’s
 Computer Simulations of Cognition 157

theory set off a tsunami of activity in which parallel processing, perceptron-

like devices were studied by a large number of theoreticians, including Selfridge
(1958), Kohonen (1977), Hopfield (1982), and Hinton and Sejnowski (1986).
Unfortunately, this work was largely, albeit temporarily, suspended when it was
shown by Minsky and Papert (1969) that simple perceptron-like networks were
intrinsically incapable of certain problems that were basic to machine learn-
ing.2 The study of these simple networks was rejuvenated when Rumelhart,
Hinton, and Williams (1986) defined the back propagation technique—an
iterative parallel processing procedure in which parameters were constantly
adjusted, thus overcoming some of the limitations raised by Minsky and Papert
(1969). Most of the perceptron-like devices consisted of simulations of less
than 100 neurons, beyond which many of the systems seemed to saturate, that
is, all of the neurons became active. Study of these toy systems today is essen-
tially the study of parallel interactions within networks composed of modest
numbers of linear components, but has drifted far from its neurophysiological
roots.
One branch in the historical development of the parallel network approach
to theories of cognitive processes emphasized a network of functional cognitive
units that were quite unlike the neurophysiological ideas that characterized some
earlier models. The nodes in the networks proposed by the PDP Research Group
(McClelland, Rumelhart, and the PDP Research Group, 1988; Rumelhart, McClel-
land, and the PDP Research Group, 1988), for example, were macroscopic
behavioral units rather than either macroneural or microneuronal components.
This approach came to be called “connectionism” and is still a powerful force
in cognitive theory. Although it emphasizes parallelism, connectionism is almost
totally free of neurophysiological concepts and concentrates on symbolic, func-
tional, or behavioral nodes rather than explicit macroneural or microneuronal
networks. Progress in this line of theory has been reviewed by Goertzel et al.
(2010) and is discussed in detail in my previous books (Uttal, 2016).
Although the concept of parallel processing expanded to permeate the most
advanced neuroreductionist theories of cognition, there was a major missing
factor inhibiting progress in theory development at the end of the 1980s. Most
of the theories and prototheories that I have discussed so far and will discuss in
this section had to be executed on modest-sized serial computers—parallel pro-
cessing supercomputers did not yet exist or were not available to researchers
interested in modeling cognitive processes. For those few models developed and
carried out on serial computers, parallelism was simulated by reiterative methods
in which each component in a network was evaluated in serial order while the
processing of every other component simply stopped and awaited its turn for
attention by the single CPU. Obviously, however, this kind of iterative seriality
is a time-consuming process even if the fastest available high-speed computer
was used. The result was that whatever models were tested had a relatively small
number (<100) of components in the networks they simulated.
158 Computer Simulations of Cognition

As the end of the 20th century approached and passed, this situation began
to rapidly change. Massively parallel supercomputers became available, and sci-
entists and engineers of all persuasions quickly took advantage of their presence
to execute large-scale simulations.
As promising as the parallel processing supercomputer is, in principle, as an
engine of theory building, it is also possible to foresee some of the difficulties
that are being faced and will continue to be faced in their application to the
task of building cognitive neuroscience theories. Some of these difficulties are
subtle and some are naïve confusions between the search for an explanatory
theory, on the one hand, and the engineering of bigger, faster computers, on
the other. Merely noting that a computer engineering project is “inspired by”
a brain property does not a mind–brain theory make.
An initial problem facing theoreticians who use massively parallel processing
computers as their tool of choice to theorize about cognitive or brain activity
is that it is difficult to know what would constitute an acceptable explanatory
theory. There are two source of this difficulty. First is the fact that both the
overall internal states of a computer and mental activity are largely inaccessible.
We have only the most indirect indications—mainly interpersonally observable
behavior or introspective reports of the cognitive concomitants of a brain activity.
However, neither of these measures permits access to the lower-level mechanisms
that constitute the “meat and potatoes” of any reductive theory.
Thus, the most abstruse entity of all—consciousness—remains poorly defined
and immeasurable as well as inaccessible; neither overt behavior nor covert
thought provide any direct indication of what are the underlying mechanisms,
either neurophysiological or hypothetical cognitive constructs that correspond
to our awareness. As noted repeatedly in the course of my writing, behavior
is underdetermined—it can be produced by what may be an innumerable
number of possible internal mechanisms and there is insufficient information
in a behavioral observation to determine the actual nature of those mechanisms.
Therefore, we may arrive at a point where we have a sufficient answer to the
basic query but no way of providing the necessary answer. If plausible sufficiency
is acceptable to some theoreticians as the criterion for acceptance of a (or the)
theoretical explanation, then so be it. It must be remembered, however, that
plausible sufficiency is not too distant from the pejorative term “folk psychologi-
cal theory”—it is just one of a possible multitude of mechanisms that can
produce analogous behavior. One implication of such an interpretation is that
if some kind of “necessity” or “truth” is desired, our search for a theory may
be a long one.
Another major difficulty in basing theories on computers is that the very
processes in which we are interested may emerge at a level of system complexity
that exceeds even the most powerful computer’s ability to manipulate the relevant
information. That is, whatever the cognitive process under study, it simply may
not be present at the levels of analysis at which we are capable of problem
 Computer Simulations of Cognition 159

solving. To make this subtle point more clearly, whatever consciousness is, it may
not exist in a network (neural or otherwise) until a sufficiently high level of
network complexity exists and then, for computational reasons, it may be impos-
sible to either produce or evaluate an overarching theory of the mind. Only
when network complexity rises to an unknown, but extreme level, may a cogni-
tive process actually become manifest. One has only to compute the factorial
of a relatively modest number to appreciate that technological and combinatorial
inadequacy in the face of realistic complexity may persist as a problem for
cognitive neuroscience into the foreseeable future. (For a more complete discus-
sion of this issue, I refer my readers to Chapter 3.)
I am most interested in the critique of the strategy of trying to draw overall
conclusions of the individual’s nature from studying the social algorithms.

. . . looking at the algorithms [components] will not yield much insight,

because the interplay of social and behaviors yields patterns that are fun-
damentally emergent. These patterns cannot be gleaned from reading code.
(Lazer, 2015, p. 1090)

Although extracted from a completely different field of investigation, the point

is that it may be infeasible to use the components of a system to predict its overall
behavior because the characterization one is looking for only occurs when all of
the pieces have been combined into the whole system. Although the “sum of
the parts” may, in principle, be no greater than the whole, the whole may not
become manifest until the parts have been integrated into a whole. Thus, the
idea of trying to produce an overarching theory or explanation of something like
the mind by emphasizing the nature of the parts may be a false hope. The mind
(or whatever else one wishes to call it) may simply not exist in the context of
its separate parts but only when all of the parts have been accounted for. This
paradox may make any effort to analyze the mind unachievable.
This discussion clarifies one point. No matter how powerful a parallel pro-
cessing supercomputer may be nor how plausible the parallel processing approach
may seem, it is not certain at this point that raw brute force computing power
will be the research vehicle that ultimately solves the mind–brain problem. Nor,
for that matter, can we yet say for certain that the mind–brain problem is solv-
able. The residual question is—how would we know a complete, overarching,
neuroreductionist theory of mind should we ever encounter it? Simply simulating
a behavior will not produce a convincing theory because of the multiple ways
in which a behavioral response might be implemented neurophysiologically.
Presumably, such a theory would at the least require that we were able to deter-
mine the exact state of an ensemble of neurons and to demonstrate the necessity
of that state for producing a behavior. It would also require that we are able to
produce changes in the behavior by manipulating that state of the ensemble of
individual neurons.
160 Computer Simulations of Cognition

The question of the criteria for acceptance of a computer simulation as a

theory of mind remains salient. Martin, Grimwood, and Morris (2000), in addi-
tion to their list of experimental strategies that would be necessary to support a
neuronal theory of learning (which I paraphrased in Chapter 3), also provided
us with a list of the empirical outcomes that should be expected if we are to
support the basic hypothesis that patterns of network activity are the psychoneural
equivalents of cognitive processes. Once again, I paraphrase their very important
suggestions about the possible general findings (originally expressed by them in
terms of learning) in a more general way to emphasize their relevance to other
cognitive topics:

1. Correlation: Activity in neuronal networks must correlate with specific cog-

nitive processes in at least some ways.
2. Induction: Changes in cognitive processes must produce changes in neuronal
activity and vice versa.
3. Occlusion: Saturation of neuronal network activity should block cognitive
processes and prevent learning new ones.
4. Intervention: Blocking relevant neuronal activity by experimental means
should block related cognitive activity.
5. Erasure: Erasing a particular pattern in a neuronal network should eliminate
the associated cognitive process.
(Paraphrased from Martin, Grimwood,
and Morris, 2000, p. 652)

The point of this list in the present context is that not all of these expected
outcomes may be available to researchers for both practical and conceptual
reasons. Nevertheless, it is at this level (that of the individual states of the under-
lying individual neurons and synapses) that the essence of our cognitive activity
is most likely to reside. It is not clear, for example, how one could actually carry
out an experiment in which the experimenter had sufficient control over the
independent variables to erase or produce a particular pattern. Experiments
calling for “induction” or “saturation” sound very much as if they were mac-
roneural manipulations rather than the microneuronal network manipulations
that characterize parallel processing networks of the kind with which we are
concerned in this chapter.
There is another very important question that has to be considered about
each parallel processing theory—what are the respective goals of the various
supercomputer “brain” models considered here? Although this book is aimed at
considering those that specifically link to cognitive and neurophysiological stud-
ies, this is not necessarily the basic motivation of all of the models discussed
here. Some, as we shall see, are primarily interested in large-scale simulations of
neuronal networks, some with the production of particular useful behaviors, and
some in the engineering of massively parallel computer systems themselves. The
 Computer Simulations of Cognition 161

varied motivations cover a wide range of network interactions. Sometimes, the

links are limited to just borrowing a few words from psychology or the neuro-
sciences. Thus, computers were early on referred to as electronic “brains” despite
a complete absence of any neural ideas beyond the fact that computers and
brains were both “information-processing” machines. This was sufficient to
create a metaphorical relation between the two (“computers are brains”), but
insufficient to establish that an explanatory model of the transformation from
brain processes to cognitive processes had been produced. Later, as it became
obvious that the brain was built on quite different principles than the von
Neumann computer, the serial computer metaphor slowly fell out of flavor.
Computer engineers, in their never-ending search for more powerful and
faster computers, often looked to a popular computer metaphor for the brain—
the parallel processor—for insights and suggestions about the organization of
such systems. To these computer engineers, their efforts were “inspired” by the
brain, but they were not intended to be explicit simulations of it. In some cases,
this allusion to the brain often went no further than an acknowledgement of
its most superficial properties (e.g., parallelism itself).
The point is that engineering parallel processing computers is a complex task
in its own right, with its own problems and goals, many of which are irrelevant
to the cognitive neuroscience task. It may be reasonably argued that the relation
between parallel processing computers and the nervous system is still metaphorical
rather than based on any robust empirical or conceptual links. Clearly, the materi-
als from which the two devices are constructed are different and these material
differences may produce different information-processing strategies. According
to Hertz, Krogh, and Palmer (1991) the engineering problems associated with
computer parallelicity can be summed up by the following questions:

1. What is the best architecture?

2. How can a network be programmed?
3. What can the various types of networks do?
4. How can a network be built in hardware?
(From Hertz, Krogh, and Palmer, 1991, pp. 8–9)

None of which has anything necessarily to do with the brain or cognition

beyond the initial metaphor of their joint parallelicity. The overarching goal
in this context is to build the fastest, and thus the most powerful general-
purpose computer system—regardless of how the brain has evolved. Beyond
the patently computer engineering projects, there are also a number of studies
using parallel computer systems that are aimed at a neuron-by-neuron repre-
sentation of known neuroanatomical structures. In this category, the goal is to
observe the structure and neurophysiological responses of a network that
exhibits comparable neurophysiological behavior to a structure such as a “neo-
cortical column”—a basic anatomical unit now known to characterize the
162 Computer Simulations of Cognition

cerebral cortex. Others have had as their goal simulation of the interactions
that may occur between other anatomical structures such as the thalamus and
the cortex. So far, however, in this listing, explaining cognitive processes has
been almost completely absent.
Another class of parallel processing computer research, however, is aimed spe-
cifically at comparisons between computer architecture and cognitive processes.
Nevertheless, in this class of research, there is less interest in neurobiological fidelity
than in producing comparable behavioral outcomes. In this context, the goal is
to make computers behave like people, but not to restrict or constrain programs
or hardware algorithms to the same computational processes used in the brain.
Needless to say, this imitation, simulation, or emulation (and often enhancement)
of human behavioral processes has been the main thrust of modern applications—to
supplement or replace many of the tasks that previously had been carried out
exclusively by humans without necessarily reproducing the same logical
“programs.”
As one considers this spectrum of motivations and despite the enormous
interest in linking brain anatomy and neurophysiology to cognitive processes, it
is important to appreciate that no cognitive neuroscience problem of mind and
brain relationship has been resolved so far by the expensive and complex research
that has been forthcoming; most theories of computer organization that have
any relevance to cognition in computer science are still “toy problems.” Clearly,
there is a multitude of ways that computers can simulate human behavior that
do not require the construction of exactly the same computer operations and
procedures that the brain presumably uses. Simulations and imitations, no matter
how perfect, cannot by themselves provide a necessary explanation of how the
brain works to produce cognitive activity—merely a sufficient imitation of
behavior. To the contrary, the ultimate theoretical goal of cognitive neurosciences
is to understand the brain well enough to mimic behavior to a degree of detail
that permits us to assert that we have solved the mind–brain problem.
This leads to another strategy in modeling cognitive processes with large-scale
supercomputers—what we may refer to as the “hopeful” approach. The technique
embodied in the hopeful strategy is to develop a simulation of a prototypical
neuron and then connect large numbers of them together and just see what
happens. It is probably the case that more of the parallel processing prototheories
currently being considered fit into this category than is admitted by investigators
in this field. Why this should be the case is obvious—although we know a lot
about neurons and synapses, we know very little of the rules governing their
transition to behavior. On the positive side of this “hopeful” scenario is the fact
that an enormous amount has been learned about the engineering organization
and performance of large-scale, parallel processing supercomputers.
There can be no dispute, however, that the most likely tool currently available
to achieve any of these goals is likely to be some future descendent of the parallel
processing supercomputer we know today. In this cognitive neuroscience context,
 Computer Simulations of Cognition 163

however, the primary goal is to solve the problem of mind–brain correspondence,

not to engineer the hardware itself. The use of supercomputers to achieve this
goal is the current frontier. Projects are just starting that have the specific goal
of providing an answer to the ancient conundrum—how does the brain make
the mind? Only time and extensive research will determine if this is a qualitative
paradigm shift or just another false alarm.
Obviously there is an increase in conceptual difficulty as the goals of research
become more relevant to cognitive neuroscience—particularly, in light of the
stated difficulty of determining what constitutes an acceptable theory of mind–
brain interaction. Indeed, it is often not clear where each of these alternative
strategies begins and another ends. It is equally possible that some engineering
development will inadvertently “crack the code” for cognitive neuroscience as
an “unintended consequence” of simply being active in the field as it is that
some neurobiological discovery will suggest a new way of engineering higher-
speed computers. Nevertheless, whatever the nature of the symbiosis between
engineering and neuroscience, it must be appreciated that not all computer
engineering projects, no matter how inspired by the brain, are intended to be
theories of the mind–brain relation.
In sum, it is important to remind my readers once again that not all parallel
processing computer projects to be discussed here are relevant to theories of
the mind–brain relationship motivating this book. Simply using neurobiologi-
cal or cognitive vocabularies does not automatically make either the hardware
or the software into a theory nor does claiming that a parallel processing
computer was “inspired” by the brain. There is, however, an optimistic note.
Regardless of the goals of one of these parallel processing research projects,
whether neurobiological or engineering, there is a common thread of activi-
ties, actions, and accomplishments. Each of the projects now to be discussed
are based on a day-to-day technological foundation of parallel processing and
each, to various degrees, does attempt to link their work to either neurophysi-
ological or psychological parameters, albeit in what may be only the use of
a few words.

Parallel Processing Computer Theories of the Brain

In the following sections of this chapter, I review some of the most significant
of the recently reported large-scale computer applications to brain science.
The criterion that has guided my selection of these exemplar studies is that
they are based, in principle, on two fundamental assumptions—namely, massive
parallelicity at the microscopic level and the idea that this mode of organiza-
tion is common to the brain and the supercomputer. As we progress through
this mini-review, it will become clear what some of the advantages and dis-
advantages are of this bifurcated approach to theory building in cognitive
neuroscience.
164 Computer Simulations of Cognition

An Early Model of the Thalamocortical System

From the point of view of neuroscience (as opposed to that of large-scale com-
puter engineering), the neural network simulation project carried out by Lumer,
Edelman, and Tononi (1997a, 1997b) was one of the earliest to attempt the
challenging task of simulating neural networks on a digital computer. The
immediate goal of these investigators was to develop a network model of
the thalamocortical system—a group of brain regions previously having been
shown to be involved in visual perception. Their long-term goal was to study
how manipulation of the properties of a simulated neuronal network affected
the behavior of the network and thus might provide some insights into the
perceptual processes that were assumed to be carried out by neuronal interactions
in this portion of the brain.
Specifically, Lumer, Edelman, and Tononi observed that there were two
regions of the visual cortex (a primary visual area and a secondary region,
each of which was organized into a three-layer column) and corresponding
primary and secondary regions of the thalamus involved in their simulation.
The cortex was subdivided into three horizontal neuronal levels and the
thalamus into two layers in accord with the known anatomy of these regions.
Both inhibitory and excitatory connections between the regions were simu-
lated as well as both ascending (centripetal) and feedback (centrifugal)
connections.
The Lumer, Edelman, and Tononi simulation fleshed out the detailed micro-
neuronal structure of the various regions by simulating the activity of 65,000
neurons and 5 million synaptic connections. The model was run on a “general-
purpose object-oriented neural simulator” (1997a, p. 211), programmed on an
unpublished software system, and was executed on an otherwise unidentified,
but, given its date, presumably serial computer. The primary method was to
create an initial state of the network by stimulating it with grating patterns to
initiate the network activity and then, by pooling the responses of the involved
neurons, describe system-wide activity.
Lumer, Edelman, and Tononi’s main goal was to examine the effect of
manipulating various parameters of the network on the generation of network-
wide oscillations. The parameters (which serve as the independent variables of
their computer experiments) included such factors as synaptic strength, transmis-
sion delays, inhibitory time constants, and the structure of the connecting loops
between the 10 regions (six cortical and four thalamic) being simulated. To
observe the cumulative response of the system they averaged the individual
neuronal activities at each of the regions of their model, seeking information
about how each of these parameters affected the oscillatory behavior. Lumer,
Edelman, and Tononi also carried out experiments in which “lesions” were made
by removing the connections between the simulated brain regions represented
in their model. The results of their experiments identified important features at
 Computer Simulations of Cognition 165

both the individual cellular and “macrocircuit” levels that determined the kinds
of oscillatory behavior resulting from these interventions.
Lumer, Edelman, and Tononi’s report is especially interesting because it is
unusually complete in spelling out the details of their goals, methods, results, and
conclusions. It is clear how they go about defining the initial conditions—the
stimuli—and the cumulative or population-averaged response measure they used
to examine a very specific property of networks—its oscillatory behavior. From
the neuroscience point of view it is anchored to well-established neurophysiologi-
cal data concerning the structure and function of the early portions of the visual
system. Thus, it added a significant amount to our understanding of how neural
networks might be affected by measurable properties of the nervous system. The
anatomical properties of the model correspond to well-known properties of the
visual system such as layering and feedback.
On the other hand, their model is limited, as they themselves point out, in
that it does not link directly to many of the problems of perception—instead it
is a test bed to study biologically plausible networks of a well-defined portion
of the nervous system. But they have not correlated these plausible visual neu-
rophysiological properties with psychophysical data. In their words:

While the simulations presented in this paper do not yet address func-
tional questions . . . they provide a powerful tool to begin investigating
these [perceptual] phenomena in closer contact with experimental
observations.
(Lumer, Edelman, and Tononi, 1997a, p. 224, italics added)

From a computer science point of view, despite what must be considered by

today’s standards to be a relatively small simulation in terms of the number of
simulated neurons and synaptic connections, this pioneering work was extremely
fruitful in establishing a methodology that guided later investigators to study
and manipulate simulated neuronal networks.
In sum, the Lumer, Edelman, and Tononi neuronal network model was a
project aimed at simulation of the neurophysiological foundations of a perceptual
mechanism by a system of 65,000 computer-generated neurons.

Spiking Neurons
Another early project in which the stage was set for the more recent and much
larger computer simulations was reported by Izhikevich (2003). Izhikevich
programmed a relatively simple network model in which each simulated neuron
(of a set variously mentioned as including 1,000, 10, 000, and “tens of thousands
of simulations”) was represented by a set of simple differential equations. The
simulated neurons were interconnected by up to a million simulated synaptic
connections, 80% of which were excitatory and 20% of which were inhibitory.
166 Computer Simulations of Cognition

Izhikevich’s goal was to simulate the spiking behavior of a slab of neurons

comparable to the number in the rat’s motor cortex. This was one of the first
models that attempted to start at the level of the Hodgkin-Huxley membrane
equations. The measure of neuron activity was an integrated value plotted on
a time-spike occurrence chart from which the pattern of spiking activity could
be used as a measure of the frequency components of the EEG (e.g., alpha and
gamma rhythms) as a response measure. The activity in this slab of simulated
interconnected neurons was triggered by initiating activity in a small set of the
excitatory synapses.
Although all of the neurons were identical, the pattern of synaptic intercon-
nections was random. The fact that there was no “learning” or “adaptive”
responses of the synapses programmed into the model meant that the emerging
waves of activity in the network led to orderly patterns of dynamic neuronal
activity in the simulated tissue. Izhikevich argued that his simple model repro-
duced complex wave-like activity typical of the organic brain that included
“spiking, bursting, and mixed modes” (p. 1572). Another interesting aspect of
his model was his claim that, because of the relatively simple properties of the
simulation (the structure of the model’s neurons was minimally represented, each
by a common differential equation), a system of “tens of thousands” of neurons
and a relatively large number of synapses could be run on a small PC-class
computer, albeit presumably not in real time. Nevertheless, the activity of a
simulated slab of brain tissue was complex enough, even with the constraints of
the simplified parameters of the model, to reproduce complex wave-like patterns
of coordinated brain activity.
Izhikevich’s pioneering work set the stage for some of the simulations that
were to follow based on his idea that patterns of spiking activity would
spontaneously arise as a function of the initial arrangement of the synaptic
junctions. Subsequently, the Izhikevich model was expanded by Izhikevich
and Edelman (2008) to a much larger-scale simulation of the thalamocortical
system, with a greatly enlarged and detailed simulation of 22 different kinds
of neurons, as well as more specific connections among a larger number of
simulated brain and brain stem regions. A number of other properties of the
network were discovered, not the least of which was its sensitivity to the
response of a single neuron and its tendency to turn off rather than to satu-
rate. (This latter property was probably a result of the proportion of excitatory
synaptic junctions.) Izhikevich and Edelman also developed novel display
techniques that helped investigators represent the overall picture of the net-
work’s simulated neuronal responses.
In sum, the Izhikevich and Izhikevich and Edelman simulations involved “a
million neurons, tens of millions of neuronal compartments, and nearly half a
billion synapses” (p. 17) whose goal was to simulate the activity of a large neu-
ronal network. The model did not run in real time even on the 60 core parallel
processor available to them.
 Computer Simulations of Cognition 167

The Neurogrid Project

Another example of a network research project in which a large number of parallel
organized neurons was simulated was the Neurogrid project (Boahen, 2006; Ben-
jamin et al., 2014). Avowedly drawing inspiration from brain circuitry, this group
used a combined analog and digital circuit technology with the goal of building
a power-efficient, parallel processing computer that simulated the activity of a
million neurons and 6 billion synaptic junctions. The Neurogrid project was
unusual in that it attempted to carry out the simulation in real time, an advantage
then made possible by the partial analog technology used by its originators.
The end product of this engineering development project was intended to
be a computer hardware system that could be easily programmed without any
specific knowledge of the networks that instantiated the hardware. The intended
applications of this hardware were primarily in specific engineering applications
to robotics and sensory prostheses. Although paying some initial lip service to
the goal “understand[ing] how cognition arises from neuronal properties”
(Boahen, 2006, p. 9), all of the attention of this project was directed at engineer-
ing application. Nevertheless, as I noted earlier, it is sometimes difficult to dis-
tinguish between engineering and cognitive neuroscience applications when one
examines research in parallel processing networks.
In sum, the Neurogrid project was a hardware development endeavor aimed
at producing an easily programmed, power-efficient, parallel processing system
consisting of a million simulated neurons that involved both analog and digital
circuitry.

The Spaun Project

An ambitious attempt to specifically solve some of the most challenging of
cognitive neuroscience problems by building a parallel processing computer
model of a cognitively competent system has been presented by Eliasmith et al.
(2012). This group’s goals were to show how a large-scale simulation of a brain
system could link to behavior—a connection that these investigators believed
had generally not been accomplished in most of the earlier work. In their words,
the goal was “to relate the incredibly complex behavior of animals to the equally
complex activity of their brains” (p. 1202).
To achieve this goal, Eliasmith and his group programmed a software system
called Nengo to simulate 2.5 million neurons to carry out one of eight kinds
of cognitive tasks, examples of which were “copy” “to recognize” an input
alphabetical character. Stimuli were either handwritten or typed characters and
responses were “the movements of a physically modeled arm” (i.e., a simulation
of an arm rather than an actual physical implementation).
Nengo is not adequately described in their publications for us to make an
evaluation of it. As a software system being run on what was reported to be an
168 Computer Simulations of Cognition

“eight core Xeon conventional computer,”3 it is certain that the execution of

Eliasmith’s program did not run in real time. Because their goal was primarily
conceptual and cognitive neuroscientific, this is hardly an objection to their
work. However, the Eliasmith project can be criticized in the context that they
set for themselves—linking the “incredible” complexities of brain and behavior.
It was not made clear in their publication how this work actually accomplishes
that task.
In sum, Spaun is another programmed (as opposed to hardware) instantiation
of a system of 2.5 million simulated neurons that its authors claim has the ability
to carry out what are analogs of cognitive activity. It is of special interest in the
present context because its motivation arises from the cognitive neuroscience
task rather than from engineering or neurophysiological considerations.

The European Human Brain Project

Perhaps the most ambitious project currently aimed at developing a supercom-
puter model of the brain is the European Human Brain Project (HBP). This
project has, as its ultimate goal, a computer simulation of the entire human
brain (Markram, 2006), ranging from the biochemical level of ionic distribu-
tions to the whole brain with all of its neurons and intricate synaptic con-
nections. A complete simulation of this magnitude would require the parallel
interaction of almost 100 billion neurons and would probably require at least
an ExaFLOPS-capable computer—a device that does not yet exist. Current
work is being carried out on an IBM Blue Gene computer capable of simulat-
ing about 200 million neurons.
The strategy used by the HBP to achieve this goal is to develop a succession
of ever more complex substructures of the brain, including a standard model of
a single neuron that closely approximates neurophysiological reality. Some may
argue that the details of the molecular properties of neurons are not particularly
relevant to explication of the information-processing network properties germane
to the process of cognitive neuroscience. However, in principle, this is a brain
project, not one aimed at the mind–brain question. Currently, it has been reported
that the HBP has been successful in developing a model for a cortical column
in the cerebral cortex of a rat—a structure that contains about 10,000 neurons—
and an aggregation of 100 of these simulated columns into a single program for
a total of about a million simulated neurons. However, it has proven difficult to
find fully vetted publications describing the accomplishments of the HBP in
detail; the most recent serious formal publication seems to have been that one
written in 2006. Other reports are essentially ambitious projections, proposals,
or incomplete statements of long-term goals.
The HBP is not intended at this stage of the game to be a model of cogni-
tive processes; it is an effort whose goal is to simulate a complex system—the
brain—at all levels of neural activity starting from the basic biochemistry of
 Computer Simulations of Cognition 169

neurons, moving on to interacting neurons and synapses, and, in the long-term

future, arriving at some connections to cognition. Therefore, it does not address
the mind-body problem directly; nevertheless, if it should succeed, ambitious
simulation of this level would massively contribute to our understanding of how
cognitive processes might be encoded at all neural levels.
Unfortunately, despite this promise, large-scale simulations such as the HBP
have come under severe criticisms in recent years by neuroscientists for several
reasons. Among the most important is the problem of controlling and display-
ing the input stimuli and the output response of such a system. Determining
the initial conditions (the starting states of each and every one of the multitude
of neurons simulated in the supercomputer) would be a formidable job unto
itself. Obviously, some novel display methodologies similar to those developed
by Izhikevich and Edelman would be necessary but those plots of neuronal
activity versus time are very close to the raw data and more imaginative display
techniques will be needed. Determining the output (i.e., the final states) of
millions, not to say the billions of simulated neurons becoming possible with
today’s very large-scale computers would also strain current display technology
unless in some unexpected way the simulation communicated the musings of
an independent sentience.4
In 2014, an open letter signed by 776 mainly European neuroscientists was
sent to the European Commission, the organization that funded the HBP (to
the prodigious and unprecedented promised amount of a billion euros) in
which it was stated that “HBP is not a well-conceived or implemented project
and that it is ill suited to be the center piece of European neuroscience”
(from Summary of Open Message to the European Commission concerning
the Human Brain Project, 18 July 2014). The project has been characterized
as being “opaque,” “poorly managed,” and “premature” and, although intended
to be a collaboration of more than a hundred universities, has suffered from
a loss of many of the institutions that were originally involved in planning
the project. Others have challenged its goals, arguing that it is actually a
computer development project hiding within the “sheep’s clothing” of brain
research.5
The HBP is an extremely controversial approach whose goals are, to a sig-
nificant degree, more hyperbole than scientific accomplishment as evidenced in
the absence of any refereed research publications. The latest development in this
ongoing tale is that the billion-euro funding from the European Union has been
redirected and “a re-evaluation of the scientific goals that relegate modeling of
the whole brain to the back burner” (Enserink, 2015).
In sum, the HBP is aimed at a complete simulation of neurophysiological
networks from the level of membrane dynamics to cognitive processes. It is
currently being run on an IBM Blue Gene computer. The project is reported
to have programmed a neuronal network that is the equivalent of a cortical
column consisting of 10,000 simulated neurons and 108 synapses.
170 Computer Simulations of Cognition

This is an ambitious, perhaps overly ambitious, project that it is hoped will

offer insights into network organization. Currently the organizers of this project
seem to have little interest in bridging the gap to cognition.

The United States BRAIN Project

A comparably ambitious, but hopefully more modest in its goals, research program
is now being financed by the United States government along with a number
of private contributors. The goal of this effort (Brain Research through Advanc-
ing Innovative Neurotechnologies [BRAIN]) is to develop methodologies for
simultaneously recording and stimulating activity in multiple neurons. Although
not as extravagantly funded and much more diffuse in its goals than the HBP,
BRAIN has also been criticized as putting the “cart before the horse,” that is,
developing special equipment before the need for it has been demonstrated.

The DARPA SyNAPSE Project

In 2008, the Defense Advanced Research Projects Agency (DARPA) initiated
and funded a program to develop a new computer architecture based on the
parallelicity idea. The motivating assumption was that the goals of this project
could be best achieved by imitating some of the features of the organic brain.
This project, which continues to this day, is referred to as the Systems of Neu-
romorphic Adaptive Plastic Scalable Electronics (SyNAPSE) program. The goals
of this project were to develop new “neuromorphic” logical circuits based on
the properties of biological neurons as well as to develop parallel processing
computer systems consisting of novel hardware, programming, and system archi-
tecture. Secondary goals included engineering goals such as power minimization
and redundant architecture that were resistant to local faults. The original par-
ticipants in the SyNAPSE program included a large number of universities
divided into two teams headed by IBM and HRL (an off-shoot of Hughes
Research Laboratory), respectively. Funding now approaches $100 million.
The two laboratories have taken different directions in the latter years of the
SyNAPSE funding cycle. HRL’s goal has been to concentrate on system architectures
that behave very much as do neurons; specifically self-programming, learning, and
what we would call today habituation. IBM has concentrated on building very
powerful circuit boards that can be grouped together to build ever larger parallel
processing computers. HRL’s part of the project has emphasized biologically realistic
properties of synapses. Thus, their circuits are smaller and designed to simulate pro-
cesses that affect the strength of synaptic conductivity such as those suspected to be
involved in learning and habituation. HRL concentrated on developing practical
applications of devices that can learn to improve their performance. An autonomous
drone helicopter is one of their current successes—the most recent version is
controlled by a single chip consisting of 576 simulated neurons.
 Computer Simulations of Cognition 171

Over the past decade, the IBM part of the project has gone in a different
direction—they have aimed at developing novel computer architectures with
ever higher numbers of cores with faster and faster processing capability. Using
the Sequoia supercomputer, one of the most powerful parallel processing super-
computers available in 2012, the IBM group was able to execute a program that
they claim simulated a system of 500 billion neurons and 100 trillion synapses
(as described in the article by Service, 2014). This had to be done with an itera-
tive process since the number of actual cores was only 1.5 million and the logic
was built on relatively slow conventional integrated circuits. As a result, the
program ran much more slowly than did estimates of the parallel processing
speed of the human brain. The Sequoia project was not designed to simulate
the human brain; it was mainly intended to develop an efficient computer
hardware system that was “inspired” by the brain.
The next step in the development of such a system was to engineer a unit-
ized “neuromorphic” large-scale integrated circuit chip that incorporated the
processing features of many neurons by means of specialized processing cores.
The IBM TrueNorth device that emerged from these primarily engineering
goals was a single chip that contains 5.4 billion transistors and 256 million simu-
lated synapses simulating the behavior of a million simulated neurons. The logic
design was organized in a manner that was assumed to be comparable to the
complexities of the human brain. Current research reports suggest that it has
been possible to link “more than a dozen” of these integrated circuit chips
together, but the developers feel that this is only the beginning and that the
system is capable of unlimited upwards scalability by simply attaching more and
more of the TrueNorth chips to one another.
One of the most interesting aspects of the TrueNorth project is that the cores
on these chips are not based on the serial von Neumann prototype. That clas-
sical design placed the core and the memory in different locations, resulting in
high-energy demand whenever information was transmitted between the memory
and the cores. This limitation of the von Neumann–type core could be abetted
by placing memory and cores on the same chip, thus reducing transmission times
and resulting in a remarkably low power demand. The IBM group (Merolla
et al., 2014) reports that the TrueNorth chip drew only 63 milliwatts.
The development of the TrueNorth chip was both an important milestone in
packing circuits onto a single chip and novel logic design. However, it consisted
of only a relatively small number of simulated cores and neurons in comparison
to the 2012 Sequoia-based project. What it accomplished should not be confused
with the kind of progress made in the 2012 project. The 2012 project carried out
programs on a supercomputer that simulated numbers that actually exceeded the
biological brain’s neuron counts (530 billion) but had to do so by repetitive use
of the circuitry in the Sequoia computer they utilized and thus not in real time.
The ultimate goal of the TrueNorth project, on the other hand, is to design
circuitry that mimics these large numbers of neurons in the brain and pack it
172 Computer Simulations of Cognition

onto a single chip—a chip that could be scaled upwards by interconnection with
other TrueNorth chips to represent billions of neurons. The TrueNorth chip has
properties that may ultimately achieve this goal, but it has not yet been expanded
to approximate the number of neurons in the brain—86 billion.
The IBM TrueNorth project, although motivated mainly by engineering goals,
also has been used to simulate some relatively simple cognitive processes. Inputs were
exemplars of a set of moving objects. The behavioral goal of the system was correct
detection and recognition by the TrueNorth chip of these different objects; however,
no performance data were provided in the brief articles describing the TrueNorth-
based simulations nor were the methods used to input the initial conditions or to
develop program modification (i.e., learning) as a result of the activity in the chip.
In summary, the HBP and the TrueNorth projects had many similarities in
their goals; however, they also were quite different in some ways. Specifically,
the main goal of the HBP is to program whatever computer system is available
to simulate billions of neurons in the human brain and to do this at multiple
levels of analysis. As each level is simulated, the intent is to concatenate these
various simulations until a neural “model” of the brain emerges. The target in
this case is the neuroanatomy and neurophysiology of the brain itself.
The IBM work is, on the other hand and quite to the contrary, aimed at the
development of hardware, inspired to a degree by the biology of the brain, but
more concerned with such functional properties as power efficiency, fault toler-
ance, packing density, and scalability, as well as programmability. The brain, to
the IBM group, is only a distant benchmark to be implemented someday by
complex circuitry—the simple behavioral functions are merely a means of dem-
onstrating the capabilities of the system. Should there be any insights about the
brain and specifically about what artificial cognitive functions are obtained during
this engineering-motivated development project, that would be “icing on the
cake.” In the meantime, allusions to the brain-like features of all of these super-
computers might well be taken with extreme caution.

5.4 Interim Summary

There is little question that the large-scale supercomputer represents the most
promising potential breakthrough tool in the search for a solution to the mind–
brain problem. The foundation premises of the two systems—the human brain
and the modern computer—are the same. Each is based on the general principle
of a network of nodes, cores, or neurons that is able to carry out many com-
putations simultaneously by virtue of its ability to parallel process. Furthermore,
the supercomputers that are available today (or will be soon) are approaching
the number of neurons in the organic brain. Although the promise is great,
none of these large-scale projects, regardless of their biological or engineering
goals, have yet achieved anything close to functional equality with the brain.
Although liberally sown with the vocabulary of psychology and neurophysiology,
there is nothing that appears to be even the harbinger of a theory of mind–brain
 Computer Simulations of Cognition 173

relations in the published literature. Indeed, we have not yet learned how to carry
out a relevant experiment on these supercomputers. A major problem inhibiting
such accomplishments is the difficulty of programming these computational behe-
moths. In actuality, the task is comparable to the analogous development process
that occurs at least over the lifetime of an individual human and possibly over eons
of organic evolution. Indeed, it is not even sure what the word “programming”
would mean in the context of a computer that operates in parallel on many com-
ponents of the input and responds in the form of a distributed network.
In large part, whatever the actual motivations and goals, many supercomputer
engineering projects are being promoted on the basis of promises and hopes in
the fields of neurophysiology and psychology. Although the power of the brain
is clearly a function of the myriad synaptic connections, we still have no idea
what are the specific patterns of these connections that correspond to behavior
selection, much less consciousness. Without question, newer and faster computer
systems are going to emerge from this activity, but we are still in need of a
conceptual revolution to determine how the parallel processing supercomputer
will contribute to cognitive neuroscience.
Surveying this chapter, we may draw a number of tentative and preliminary
conclusions about the promising role of these new computers as theory engines:

1. The development in recent years of the computer from a simple single-

core system to one with billions of functional components has been
amazing.
2. The simple concept of being able to store both data and instructions in the
computer memory was a key factor in the development of the modern stored
program computer.
3. Another key factor was the ability to modify the program during its
execution.
4. Speeding up a computer simply by adding processing nodes is not always pos-
sible. Some programs are not inherently decomposable; in this case Amdhal’s
law holds wherein the speedup in processing is equal to 1/1–P where P is the
proportion of the program that can be parallelized.
5. Conceptually, the parallel processing supercomputer promises to come closer
than any other available tool or methodology to understanding the information-
processing aspects of the brain.
6. A challenging conceptual and technological barrier to theory development is
the fact that we do not yet know how to interpret the information processes
that may be carried out by these systems. Specifically it is not yet clear how we
would program them to set up initial states, nor do we know how to interpret
the patterns of activity that might correspond to organic mental activity.
7. A related issue is that we do not have a satisfactory way of defining and mea-
suring cognitive processes given their inaccessibility. How would we know
that a powerful computer is conscious? The well-known Turing test, despite
popular opinion, does not distinguish the conscious from the automatic.
174 Computer Simulations of Cognition

8. We do not yet have a solid idea of how many neurons in the brain are required
to carry out cognitive processes or how many cores are necessary to solve a
computer problem.
9. Large-scale parallel processing computer research is being carried out with
multiple motivations and goals. Most are aimed at engineering development
or simulation of particular neuronal networks. Often, research in this field is
embellished with words such as “brain” or “neuron,” but the cognitive and
neurophysiological meanings of these words is long lost. Many such “brain”
projects have little to do with the goals of cognitive neuroscience.
10. Looming on the horizon is what is probably the most significant practi-
cal obstacle of all—the combinatorial explosion. The enormity of combina-
tions and possibilities is not yet a part of the thinking of most psychologists
and physiologists. There may well be things we cannot do because it would
require computers as large as the knowable universe.
11. Similarly, the criteria for demonstrating equivalence of brain function and
cognitive processes are far more stringent than generally understood. How
will we know when a simulation provides a necessary explanation and not
just a sufficient one?
12. A sampling of the large-scale parallel processing simulations being carried
out in recent years displays the diversity of their goals. Some are strictly engi-
neering, some are neurophysiological, and a few aspire to being models of
cognition. Most end up being examples of cumulative neuronal activity that
are not linked to cognition. Some end up being restricted to sensory pro-
cesses rather than anything that might be called cognitive. Some, after years of
research and huge amounts of money expended, end up with nothing of rel-
evance to cognitive neuroscience. The words used to describe one’s favorite
project often do not discriminate between these disparate goals and findings.

In conclusion, the advent of large-scale parallel processing supercomputers

has opened up a whole new approach to the problem of how we deal with the
huge number of neurons and synapses populating the organic brain. Many
imponderables and challenges still lie ahead before we can assert that we have
determined the codes for cognitive process; however, this approach is in its
infancy and only the future can tell if the Hebbian concept of interacting neu-
rons represents a tractable or intractable path to solving the most important
problem of human history.

Notes
1. It is important to remember that not all “neural” computers engage ideas from neu-
rophysiology. The mathematics specific to neural-like networks is characterized by a
goal of dealing with networks of linear threshold units. Engineers are concerned with
increasing the speed and power of computers using strategies that may range from
simple metaphors of “brain-like” networks without any specific interest in building a
 Computer Simulations of Cognition 175

theory of cognition. In this book, I have striven, not always successfully, to find examples
of projects that have specific cognitive neuroscience implications.
2. About this time, I (Uttal, 1975) proposed an autocorrelation theory of form detection
that was also a parallel processing system. Calculations of the model were successful in
predicting the detectability of dotted forms hidden in dotted noise.
3. The Xeon computer CPU is a modestly powered core used in a system that is just
barely above a simple desk computer. It is not possible to estimate the speed at which
such a computer might run a 2.5 million-neuron simulation. However, it is very clear
that, even with eight cores, it would not be comparable in execution time to any of
the high-level parallel processors discussed elsewhere.
4. Some glimmering of hope for the development of super displays is on the horizon.
Ahrens et al. (2013), for example, have developed a technique combining optogenetic
and optical techniques to image the entire body of a larval zebrafish. This permitted
them to simultaneously show 80% of neurons in this model preparation at a resolution
that allowed them to scrutinize individual neurons. Although promising a considerable
technical accomplishment, this technique does not solve the problem of numerousness
(the data explosion) nor of complexity (the combinatorial explosion).
5. Although the same criticism may be made of the United States–sponsored BRAIN
project, that very large project has much more modest goals—namely to provide the
instruments necessary to “reverse engineer” the brain at some future time.
This page intentionally left blank
6
EMERGING CONCLUSIONS

This book has been aimed at partially answering one practical question and one
theoretical question. The practical question is—how does the available technol-
ogy affect microneuronal theory development in cognitive neuroscience? To
answer the first question, it has been deemed to be appropriate to survey the
nature of the various kinds of instruments that are available, the kinds of empiri-
cal results that have been forthcoming from their application, and the relevant
theories. It is clear that remarkable progress has been made in both instrumenta-
tion and empirical neurophysiological research and that the impact on contem-
porary theory has been substantial.
The main answer to the second question—the theoretical one—that we must
draw from the review carried out in this book is that despite substantial progress
in understanding the brain, we still have no idea of how the mental emerges
from the physiological. What we have, instead, is an unsubstantiated but increas-
ing appreciation that the key to solving the theoretical question is going to be
found at the microneuronal level, that is, at the level at which the properties of
the individual neuron or of networks of neurons are preserved. It is not to be
found, as I argued in my companion book (Uttal, 2016), at the macroneural
level that characterizes so much of current brain-imaging research. It is not, I
argued there, that the cumulative, statistically summarized, pooled macro-responses
that purportedly represent the codes for cognition, but, instead, the nature of
individual neurons and their combinations and interactions.
This book has surveyed the technology of microneuronal research and the
kinds of theories that are driven by observations that are obtained at this micro-
neuronal level. It is a truism that none of these methods is perfect and that
each is flawed in some fundamental manner or is still in an early stage of
development. Nevertheless, much of what we obtained at the level of single
178 Emerging Conclusions

neurons is based on some outstanding accomplishments in measurement of the

activity of these microscopic entities. The fortuitous availability of model
preparations has in some cases permitted us to study some relatively simple
neuronal networks. Unfortunately, baldly extrapolating these essentially single-
neuron neurophysiological accomplishments to theory leads one to one of the
great dead ends of modern mind–brain theory—the idea that single neurons
can encode or represent complex cognitive concepts. Equally unfortunately, our
ability to examine even the simplest of networks is limited to structures involv-
ing only a few neurons.
The most promising new technology capable of influencing theory now
appearing on the scene is the large-scale parallel processing “supercomputer.”
This kind of information-processing giant has now been engineered to a point
that, in principle at least, it should be capable of enlisting a vast number of
computing elements, a number that is comparable to the number of neurons in
the human brain. As powerful as it is, however, we still have only the most
primitive ideas concerning how to run a computer experiment involving so
many components—how, for example, do we initialize such a behemoth and
how would we know what it is doing.
The overarching theoretical question, therefore, remains unanswered. Worst
of all, we do not yet even know if the mind–brain problem is solvable. Over
the years, I have struggled with this general question without resolution. The
differences between the properties of the tangible brain and the intangible
mind are so great that the whole approach to seeking a neuroreductionist
answer to the question of the nature of the mind may be an impossible ques-
tion or a bad question.
Regardless whether we will ultimately be able to answer the grand question,
there are many less earth-shaking issues that can be studied and from which
general understanding and formal, if specialized, theories will eventually emerge.
Whether it be potential progress in understanding networks or better descrip-
tions of cognitive processes, there is much to be learned about the nature of
mind–brain relations before we abandon the chase. The techniques and theories
described here go hand-in-hand and the mutual impact that each has on the
other should not be underestimated. No one who is interested in the field of
cognitive neuroscience can survive professionally in the future if unacquainted
with these techniques and their implications. If this book introduces both
practice and theory to those who have aspirations in this field, then it will have
served its purpose.
In this last chapter, I present a list of emerging observations and principles
that summarize the extended discussions of the preceding chapters. Many of
these statements, admittedly, are subject to controversy and varied interpreta-
tion. However, I believe collectively that they represent a comprehensive state-
ment about the current state of the microneural approach to cognitive
neuroscience.
 Emerging Conclusions 179

1. The most fundamental postulate in cognitive neuroscience is what we may

call the ontological assumption. This postulate asserts that all cognitive pro-
cesses in some unknown way are accounted for in terms of the neurophysical
processes and organization of the brain. Whatever it is that will ultimately
explain the mind–brain relation, it cannot in any way countervail the laws of
physics. Without this postulated bridge between mind and matter the enter-
prise of cognitive neuroscience could not exist.
2. The ultimate theoretical goal of this science is to determine the “psychoneu-
ral equivalent” of cognition, that is, to determine which neural mechanisms
embody which cognitive processes.
3. The mind–brain problem is enormously complex, and it is not yet certain
whether it is scientifically tractable.
4. Despite this difficulty, theories abound in cognitive neuroscience in which
mental activity is associated with brain components and functions. These
theories exist at many different levels of analysis.
5. This book has concentrated on the microlevel—the level of neurons and
synapses—and of networks of them.
6. A basic assumption in writing this book was that the methodology and equip-
ment used at any time in history strongly determined the type of theory that
will emerge despite what may be ample countervailing evidence.
7. Theories can be categorized as reductive and nonreductive. Reductive theo-
ries may be neural or cognitive depending on the elements from which they
are supposed to be constructed. Nonreductive or descriptive theories track
the trajectory of a process but do not attempt to analyze the underlying
mechanisms.
8. Theories can also be categorized as top-down or bottom-up. The former is
the strategy used by psychologists and other behavioral scientists. The latter is
the approach taken by neurophysiologists.
9. Currently, the most promising path to an overarching theory of mind–brain
relationships lies at the microscopic network level in which neuronal orga-
nization is the key. Top-down theories can never be definitive because of the
underdetermined nature of critical evidence.
10. Single cell theories are deeply flawed and implausible at their conceptual
roots. Although there is no question that some neurons in the brain respond
highly selectively to specific stimuli, the converse is questionable. That is,
activation of a single neuron to determine what its cognitive effects are is an
undoable experiment. Single neuron theories will always be based on Gedan-
ken experiments and analogies rather than actual ones.
11. Even if we could carry out an experiment, we are confronted with the
“exclusivity” problem. That is, it is impossible to know what effect an indi-
vidual neuron is having on mental processes in the cacophony of all of the
other responding neurons. In other words, how does one of many neurons
produce or dominate the cognitive outcome to the exclusion of all others?
180 Emerging Conclusions

12. The idea of a network of many interacting neurons has begun to dominate
and replace theoretical thinking in cognitive neuroscience. This approach
was stimulated by the pioneering work of Donald Hebb, who postulated
networks of neurons organized into functional groups whose configurations
varied with the state of the involved synaptic junctions.
13. Unfortunately, the Hebb conjecture is probably untestable since it requires
enormous amounts of information about the detailed structure of the net-
work. No conceivable neurophysiological methodology permits experiments
of this kind in which either adequate data are accumulated or stimulus con-
figurations are properly controlled.
14. Some (how many can only be guessed at) cognitive neuroscientists seem not
to appreciate the basic principles of combinatorial arithmetic. A very small
sample of nodes (neurons) can produce astronomical numbers of combi-
nations—numbers that often turn out to exceed the space available in the
known universe or the durations that exceed the known age of the universe.
The possibility of an exhaustive analysis of a system like the brain is remote,
perhaps even demonstrably impossible. Even such superficially simple tasks
such as spike action potential clustering or the determination of functional
connectivity maps may be impossible because of combinatoric constraints.
15. The direct neurophysiological approach in which the arrangement and
states of the neurons involved in a real neuronal network is also unlikely
to be achieved. It is not possible in even the most reduced model prepara-
tion to track out the functions of a network consisting of more than a few
neurons.
16. Furthermore, it is never certain whether a mechanism observed in a model
nervous system is implementing the same function in the same way as the
ones underlying human thought. Indeed, it is more likely in a simplified
model preparation that we are looking at analogs rather than homologs of the
processes of interest.
17. Redundancy is a basic fact of life in the nervous system. An observed neu-
rophysiological mechanism may be only one of many alternative (and con-
stantly shifting) structures.
18. The parallel processing supercomputer is achieving numbers of synthetic or
programmed neurons that are so great that it is feasible to begin to think of
the possibility of simulating human brains. However, loading a supercom-
puter with initial conditions and recovering the myriad details of the resulting
response remain major difficulties. However, this supercomputer approach is
only a few years old and much is to be learned about what they can and can-
not offer to the problems of cognitive neuroscience theory development.
19. Even a successful simulation of a cognitive process on a supercomputer would
not solve the problem of distinguishing between a sufficient and clever ana-
log, on the one hand, and a correct and necessary explanatory explanation
 Emerging Conclusions 181

of how the magical transformation of tangible brain into intangible mind

occurs, on the other. A functional simulation is not an explanation.
20. Despite the many obstacles and uncertainties confronting research in cogni-
tive neuroscience, there is no more exciting and significant scientific problem
than the mind–brain one. I have no regrets about spending my career on it,
no matter how impenetrable the task seems at the moment.
This page intentionally left blank
BIBLIOGRAPHY

Adamos, D. A., Kosmidis, E. K., & Neophilidis, G. (2008). Performance evaluation of

PCA-based spike sorting algorithms. Computer Methods and Programs in Biomedicine,
91(3), 232–244.
Ahrens, M. B., Orger, M. B., Robson, D. N., Li, J. M., & Keller, P. J. (2013). Whole-
brain functional imaging at cellular resolution using light-sheet microscopy. Nature
Methods, 10, 413–420.
Alivisatos, A. P., Chun, M. Y., Church, G. M., Deisseroth, K., Donoghue, J. P., Greenspan,
R. J., . . . Yuste, R. (2013). The brain activity map. Science, 339(6125), 1284–1285.
Amdahl, G. M. (1967). Validity of the single processor approach to achieving large-scale computing
capabilities. Paper presented at the AFIPS Spring Joint Computer Conference, April
1967, Atlantic City, NJ.
Ananthanarayanan, R., Esser, S. K., Simon, H. D., & Modha, D. S. (2009). The cat is out
of the bag: Cortical simulations with 10^9 neurons, 10^13 synapses. Paper presented at the
Proceedings of the Conference on High Performance Computing Networks, Storage
and Analysis, SC ’09, New York, NY.
Anderson, J. A., Pellionisz, A., & Rosenfeld, E. (1993). Neurocomputing II. Cambridge,
MA: MIT Press.
Anderson, M. L. (2010). Neural reuse: A fundamental organizational principle of the
brain. Behavioral and brain sciences, 33(04), 245–266.
Argonov, V. Y. (2012). Neural correlate of consciousness in a single electron: Radical
answer to “quantum theories of consciousness.” Neuroquantology, 10(2), 276–285.
Aue, T., Lavelle, L. A., Cacioppo J. T. (2009). Great expectations: What can fMRI research
tell us about psychological phenomena? International Journal of Psychophysiology, 73,
10–16.
Azevedo, F. A. C., Carvalho, L. R. B., Grinberg, L. T., Farfel, J. M., Ferretti, R. E. L.,
Leite, R. E. P., . . . Herculano-Houzel, S. (2009). Equal numbers of neuronal and
nonneuronal cells make the human brain an isometrically scaled-up primate brain.
Journal of Comparative Neurology, 513(5), 532–541.
184 Bibliography

Bakker, M., & Wicherts, J. M. (2011). The (mis)reporting of statistical results in psychol-
ogy journals. Behavior Research Methods, 43, 666–676.
Bandettini, P. (2007). Functional MRI today. International Journal of Psychophysiology, 63(2),
138–145.
Bargmann, C. I., Hartwieg, E., & Horvitz, H. R. (1993). Odorant-selective genes and
neurons mediate olfaction in C. elegans. Cell, 74(3), 515–527.
Barlow, H. B. (1953). Summation and inhibition in the frog’s retina. Journal of Physiology-
London, 119(1), 69–88.
Barlow, H. B. (1961). Possible principles underlying the transformations of sensory mes-
sages. In W. A. Rosenblith (Ed.), Sensory Communication (pp. 217–234). Cambridge,
MA: MIT Press and New York: John Wiley.
Barlow, H. B. (1972). Single units and sensation: A neuron doctrine for perceptual psy-
chology. Perception, 1, 371–394.
Barrett, L. F. (2011). Bridging token identity theory and supervenience theory through
psychological construction. Psychological Inquiry, 22, 115–127.
Barrett, L. F. “What Emotions Are (and Aren’t).” The New York Times, August 2, 2015.
Barrett, L. F., & Satpute, A. (2013). Large-scale brain networks in affective and social
neuroscience: Towards an integrative architecture of the human brain. Current Opinion
in Neurobiology, 23, 361–372.
Bashir, S., Perez, J. M., Horvath, J. C., Pena-Gomez, C., Vernet, M., Capia, A., . . . Pascual-
Leone, A. (2014). Differential effects of motor cortical excitability and plasticity in
young and old individuals: A Transcranial Magnetic Stimulation (TMS) study. Frontiers
in Aging Neuroscience, 6(111), 1–13.
Becker, D. J. (1995). Beowulf: A parallel workstation for scientific computation. Paper presented
at the International Conference for Parallel Processing.
Belliveau, J. W., Kennedy, D. N., Mckinstry, R. C., Buchbinder, B. R., Weisskoff, R. M.,
Cohen, M. S., . . . Rosen, B. R. (1991). Functional mapping of the human visual-
cortex by magnetic-resonance-imaging. Science, 254(5032), 716–719.
Belykh, I., de Lange, E., & Hasler, M. (2005). Synchronization of bursting neurons: What
matters in the network topology. Physical Review Letters, 94(18).
Benjamin, B., Gao, P., McQuinn, E., Choudhary, S., Chandrasekaran, A. R., Bussat, J. M.,
. . . Boahen, K. (2014). Neurogrid: A mixed-analog-digital multichip system for
large-scale neural simulations. Proceedings of the IEEE, 102(5), 699–716.
Bennett, M. R. (2001). History of the Synapse. Boca Raton, FL: CRC Press.
Berger, H. (1929). Uber das electroenkephalogramm de menschem. Archiv fur psychiatric
und Nervenkrankheiten, 97, 527–570.
Binet, A. (1907). The Mind and the Brain. London: K. Paul, Trench, Trübner & Co. Ltd.
Blake, A. J., Rodgers, F. C., Bassuener, A., Hippensteel, J. A., Pearce, T. M., Pearce,
T. R., . . . Williams, J. C. (2010). A microfluidic brain slice perfusion chamber for
multisite recording using penetrating electrodes. Journal of Neuroscience Methods, 189(1),
5–13.
Boahen, K. (2006). Neurogrid: Emulating a million neurons in the cortex. Conference
Proceeding of the IEEE Engineering in Medicine and Biology Society, Supplement, 6702.
Bowers, J. S. (2009). On the biological plausibility of grandmother cells: Implications for
neural network theories in psychology and neuroscience. Psychological Review, 116(1),
220–251.
Breasted, J. H. (1991). The Edwin Smith Surgical Papyrus (Vol. 3–4). Chicago: University
of Chicago Oriental Institute Publications.
 Bibliography 185

Brindley, G. S. & Merton, P. A. (1960). The absence of position sense in the human eye.
The Journal of Physiology, 153, 127–130.
Brown, E. N., Kass, R. E., & Mitra, P. P. (2004). Multiple neural spike train data analysis:
State-of-the-art and future challenges. Nature Neuroscience, 7(5), 456–461.
Burks, A. R., & Burks, A. W. (1989). The First Electronic Computer: The Atanasoff Story.
Ann Arbor: University of Michigan Press.
Button, K. S., Ioannidis, J. P. A., Mokrysz, C., Nosek, B. A., Flint, J., Robinson, E. S. J. &
Munafò, M. (2013). Power failure: Why small sample size undermines the reliability
of neuroscience. Nature Reviews Neuroscience, 14, 365–376.
Cajal, S. R. Y. (1906, December 12). The structure and functions of neurons. Paper presented
at the Nobel Foundation Award Ceremony. Stockholm, Sweden.
Carew, T. J., Hawkins, R. D., & Kandel, E. R. (1983). Differential classical-conditioning
of a defensive withdrawal reflex in Aplysia-Californica. Science, 219(4583), 397–400.
Carew, T. J., Pinsker, H. M., & Kandel, E. R. (1972). Long-term habituation of a defensive
withdrawal reflex in Aplysia. Science, 175(4020), 451–454.
Cariani, P. A. (2004). Temporal codes and computations for sensory representation and
scene analysis. IEEE Transactions on Neural Networks, 15(5), 1100–1111.
Carp, J. (2012a). The secret lives of experiments: Methods reporting in the fMRI literature.
NeuroImage, 63(1), 289–300.
Carp, J. (2012b). On the plurality of (methodological) worlds: Estimating the analytic
flexibility of fMRI experiments. Frontiers in Neuroscience, 6, 149.
Carpenter, G. A., Grossberg, S., & Reynolds, J. H. (1991). Artmap—Supervised real-time
learning and classification of nonstationary data by a self-organizing neural network.
Neural Networks, 4(5), 565–588.
Castellucci, V., & Kandel, E. R. (1976). Presynaptic facilitation as a mechanism for behav-
ioral sensitization in Aplysia. Science, 194(4270), 1176–1178.
Castellucci, V., Pinsker, H., Kupferma, I., & Kandel, E. R. (1970). Neuronal mechanisms
of habituation and dishabituation of gill-withdrawal reflex in Aplysia. Science, 167(3926),
1745–1748.
Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental Theory. New York:
Oxford University Press.
Charkhkar, H., Knaack, G. L., Gnade, B. E., Keefer, E. W., & Pancrazio, J. J. (2012).
Development and demonstration of a disposable low-cost microelectrode array for
cultured neuronal network recording. Sensors and Actuators B-Chemical, 161(1),
655–660.
Chiang, A. S., Lin, C. Y., Chuang, C. C., Chang, H. M., Hsieh, C. H., Yeh, C. W., . . .
Hwang, J. K. (2011). Three-dimensional reconstruction of brain-wide wiring networks
in Drosophila at single-cell resolution. Current Biology, 21(1), 1–11.
Christofides, N. (1975). Graph Theory: An Algorithmic Approach. New York: Academic
Press.
Cogan, S. F. (2008). Neural stimulation and recording electrodes. Annual Review of Bio-
medical Engineering, 10, 275–309.
Cohen, T. E., Kaplan, S. W., Kandel, E. R., & Hawkins, R. D. (1997). A simplified prepa-
ration for relating cellular events to behavior: Mechanisms contributing to habituation,
dishabituation, and sensitization of the Aplysia gill-withdrawal reflex. Journal of Neu-
roscience, 17(8), 2886–2899.
Coltheart, M. (2006). What has functional neuroimaging told us about the mind (so far)?
Cortex, 42(3), 323–331.
186 Bibliography

Cox, D. R., & Smith, W. L. (1953). The superposition of several strictly periodic sequences
of events. Biometrika, 40, 1–11.
Cox, D. R., & Smith, W. L. (1954). On the superposition of renewal processes. Biometrika,
41(1–2), 91–99.
Crick, F., & Koch, C. (1995). Are we aware of neural activity in primary visual-cortex?
Nature, 375(6527), 121–123.
Crick, F., & Koch, C. (1998). Consciousness and neuroscience. Cerebral Cortex, 8(2),
97–107.
Cullington, H. E., & Zeng, F. G. (2008). Speech recognition with varying numbers and
types of competing talkers by normal-hearing, cochlear-implant, and implant simula-
tion subjects. Journal of the Acoustical Society of America, 123(1), 450–461.
Damadian, R. (1971). Tumor detection by nuclear magnetic resonance. Science, 171(3976),
1151–1153.
Dawson, G. D., & Scott, J. W. (1949). The recording of nerve action potentials through
skin in man. Journal of Neurology, Neurosurgery, and Psychiatry, 12, 259–267.
de Garis, H., Chen, S., Goertzel, B., & Lian, R. T. (2010). A world survey of artificial
brain projects, Part I: Large-scale brain simulations. Neurocomputing, 74(1–3), 3–29.
DeMonasterio, F. M., & Gouras, P. (1975). Functional properties of ganglion-cells of
rhesus-monkey retina. Journal of Physiology-London, 251(1), 167–195.
Desai, S., Stickney, G., & Zeng, F. G. (2008). Auditory-visual speech perception in normal-
hearing and cochlear-implant listeners. Journal of the Acoustical Society of America, 123(1),
428–440.
DeWeerth, S. P., Meacham, K. W., Giuly, R. J., Guo, L., & Hochman, S. (2008). A
lithographically-patterned, elastic multi-electrode array for surface stimulation of the
spinal cord. Biomedical Microdevices, 10(2), 259–269.
di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V., & Rizzolatti, G. (1992). Understanding
motor events—A neurophysiological study. Experimental Brain Research, 91(1),
176–180.
Djourno, A., & Eyries, C. (1957). Prothese auditive par excitation electrique a distance
du nerf sensoriel a laide dun bobinage inclus a demeure. Presse Medicale, 65(63),
1417–1417.
Duhem, P. (1914). La Theorie Physique: Son Objet et Sa Structure. Paris: Marcel Riviera &
Cie.
Ekanadham, C., Tranchina, D., & Simoncelli, E. P. (2014). A unified framework and
method for automatic neural spike identification. Journal of Neuroscience Methods, 222,
47–55.
Eliasmith, C., Stewart, T. C., Choo, X., Bekolay, T., DeWolf, T., Tang, Y., Rasmussen, D.
(2012). A large-scale model of the functioning brain. Science, 338(6111):1202–5.
doi:10.1126/science.1225266
Enroth-Cugell, C., & Robson, J. G. (1966). The contrast sensitivity of retinal ganglion
cells of the cat. Journal of Physiology-London, 187(3), 517–552.
Enserink, M. (2015). A [Euro] billion brain reboot. Science, 347(6229), 1406–1407.
Erickson, J., Tooker, A., Tai, Y. C., & Pine, J. (2008). Caged neuron MEA: A system for
long-term investigation of cultured neural network connectivity. Journal of Neuroscience
Methods, 175(1), 1–16.
Eversmann, B., Jenkner, M., Hofmann, F., Paulus, C., Brederlow, R., Holzapfl, B., . . .
Thewes, R. (2003). A 128x128 CMOS biosensor array for extracellular recording of
neural activity. IEEE Journal of Solid-State Circuits, 38(12), 2306–2317.
 Bibliography 187

Farah, M. J. (2014). Brain images, babies, and bathwater: Critiquing critiques of functional
neuroimaging. Hastings Center Report, 44, S19–S30.
Fatt, P., & Katz, B. (1952). Spontaneous subthreshold activity at motor nerve endings.
Journal of Physiology-London, 117(1), 109–128.
Feigl, H. (1958). The mental and the physical. Minnesota Studies in the Philosophy of
Science, 2, 370–497.
Field, D. J. (1994). What is the goal of sensory coding? Neural Computation, 6(4), 559–601.
Flood, T. F., Gorczyca, M., White, B. H., Ito, K., & Yoshihara, M. (2013). A large-scale
behavioral screen to identify neurons controlling motor programs in the Drosophila
brain. G3-Genes Genomes Genetics, 3(10), 1629–1637.
Flynn, M. J. (1972). Some computer organizations and their effectiveness. IEEE Transac-
tions on Computers, C-21(9), 948–960.
Fodor, J. A. & Pylyshyn, Z. W. (1988). Connectionism and cognitive architecture: A
critical analysis. Cognition, 28, 3–71.
Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., & Raichle,
M. E. (2005). The human brain is intrinsically organized into dynamic, anticorrelated
functional networks. Proceedings of the National Academy of Sciences of the United States
of America, 102(27), 9673–9678.
Fox, P. T., & Friston, K. J. (2012). Distributed processing: Distributed functions? Neuro-
Image, 61(2), 407–426.
Fried, I., MacDonald, K. A., & Wilson, C. L. (1997). Single neuron activity in human
hippocampus and amygdala during recognition of faces and objects. Neuron, 18(5),
753–765.
Galambos, R. (1996). Autobiography. In L. Squires (Ed.), The History of Neuroscience in
Autobiography (Vol. 1, pp. 178–221). Washington, DC: Society for Neuroscience.
Galambos, R. (2007). Reflections on the conceptual origins of neuron-glia interactions.
Neuron Glia Biology, 3, 89–91.
Gall, F. J., & Spurzheim, J. C. (1808). Recherches sur le systeme nerveux en general et sur
celui du cerveau en particulier. Paris: Memoirs de Academie de Sciences.
Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Action recognition in the pre-
motor cortex. Brain, 119, 593–609.
Gerard, R. W., & Forbes, A. (1928). “Fatigue” of the flexion reflex. American Journal of
Physiology, 86(1), 186–205.
Gerstein, G. L., & Clark, W. A. (1964). Simultaneous studies of firing patterns in several
neurons. Science, 143(361), 1325–1357.
Gill, S. (1958). Parallel programming. Computer Journal, 1(1), 2–10.
Gingrich, K. J., Baxter, D. A., & Byrne, J. H. (1988). Mathematical-model of cellular
mechanisms contributing to presynaptic facilitation. Brain Research Bulletin, 21(3),
513–520.
Gingrich, K. J., & Byrne, J. H. (1985). Simulation of synaptic depression, posttetanic poten-
tiation, and presynaptic facilitation of synaptic potentials from sensory neurons mediating
gill-withdrawal reflex in Aplysia. Journal of Neurophysiology, 53(3), 652–669.
Goertzel, B., Lian, R. T., Arel, I., de Garis, H., & Chen, S. (2010). A world survey of
artificial brain projects, Part II: Biologically inspired cognitive architectures. Neurocom-
puting, 74(1–3), 30–49.
Goldstein, H. H., & Goldstein, A. (1946). The ENIAC. M.T.A.C., 2, 97–110.
Golgi, C. (1906). The neuron doctrine—theory and facts. Paper presented at the 1906 Nobel
Foundation Awards Ceremony. Stockholm, Sweden.
188 Bibliography

Gonzalez Castillo, J., Saad, Z. S., Handwerker, D. A., Inati, S. J., Brenowitz, N., & Ban-
dettini, P. A. (2012). Whole-brain, time-locked activation with simple tasks revealed
with massive averaging and model-free analysis. Proceedings of the National Academy of
Sciences of the United States of America, 109, 5487–5492.
Goodman, M. B., Hall, D. H., Avery, L., & Lockery, S. R. (1998). Active currents regulate
sensitivity and dynamic range in C. elegans neurons. Neuron, 20(4), 763–772.
Graham, J., & Gerard, R. W. (1946). Journal of Cellular and Comparative Physiology, 28, 99.
Gross, C. G. (2002). Genealogy of the “grandmother cell.” Neuroscientist, 8(5),
512–518.
Gross, G. W., Rieske, E., Kreutzberg, G. W., & Meyer, A. (1977). New fixed-array multi-
microelectrode system designed for long-term monitoring of extracellular single unit
neuronal-activity in vitro. Neuroscience Letters, 6(2–3), 101–105.
Gross, G. W., Williams, A. N., & Lucas, J. H. (1982). Recording of spontaneous activity
with photoetched microelectrode surfaces from mouse spinal neurons in culture.
Journal of Neuroscience Methods, 5(1–2), 13–22.
Grossberg, S., & Mingolla, E. (1985). Neural dynamics of perceptual grouping: Textures,
boundaries, and emergent segmentations. Perception & Psychophysics, 38(2), 141–171.
Grundfest, H., Sengstaken, R. W., Oettinger, W. H., & Gurry, R. W. (1950). Stainless steel
micro-needle electrodes made by electrolytic pointing. Review of Scientific Instruments,
21(4), 360–361.
Ha, H. I., Hendricks, M., Shen, Y., Gabel, C. V., Fang-Yen, C., Qin, Y. Q., . . . Zhang, Y.
(2010). Functional organization of a neural network for aversive olfactory learning in
Caenorhabditis elegans. Neuron, 68(6), 1173–1186.
Ham, F. M., & Kostanic, I. (2000). Principles of Neurocomputing in Science and Engineering.
New York: McGraw-Hill.
Harary, F. (1969). Graph Theory. Reading, MA: Addison-Wesley Pub. Co.
Harris, J. D. (1943). Habituatory response decrement in the intact organism. Psychological
Bulletin, 40(6), 385–422.
Hartline, H. K. (1938). The response of single optic nerve fibers of the vertebrate eye to
illumination of the retina. American Journal of Physiology, 121(2), 400–415.
Hartline, H. K. (1940). The effects of spatial summation in the retina on the excitation
of the fibers of the optic nerve. American Journal of Physiology, 130(4), 700–711.
Hartline, H. K. (1949). Inhibition of activity of visual receptors by illuminating nearby
retinal areas in the Limulus eye. Federation Proceedings, 8(1), 69–69.
Hartline, H. K., & Graham, C. H. (1932). Nerve impulses from single receptors in the
eye. Journal of Cellular and Comparative Physiology, 1(2), 277–295.
Hartline, H. K., & Ratliff, F. (1957). Inhibitory interaction of receptor units in the eye
of Limulus. Journal of General Physiology, 40, 357–376.
Hartline, H. K., & Ratliff, F. (1958). Spatial summation of inhibitory influences in the
eye of Limulus, and the mutual interaction of receptor units. Journal of General Physiol-
ogy, 41, 1049–1066.
Hartline, H. K., Wagner, H. G., & Ratliff, F. (1956). Inhibition in the eye of Limulus.
Journal of General Physiology, 39(5), 651–673.
Hebb, D. O. (1949). The Organization of Behavior: A Neuropsychological Theory. New York:
Wiley.
Heer, F., Franks, W., Blau, A., Taschini, S., Ziegler, C., Hierlemann, A., & Baltes, H. (2004).
CMOS microelectrode array for the monitoring of electrogenic cells. Biosensors &
Bioelectronics, 20(2), 358–366.
 Bibliography 189

Hell, S. W., & Wichmann, J. (1994). Breaking the diffraction resolution limit by stimulated-
emission—Stimulated-emission-depletion fluorescence microscopy. Optics Letters,
19(11), 780–782.
Henson, R. (2006). Forward inference using functional neuroimaging: Dissociations versus
associations. TRENDS in Cognitive Sciences, 10(2), 64–69.
Hertz, J., Krogh, A., & Palmer, R. (1991). Introduction to the Theory of Neural Computation.
Redwood City, CA: Addison-Wesley.
Hickok, G. (2009). Eight problems for the mirror neuron theory of action understanding
in monkeys and humans. Journal of Cognitive Neuroscience, 21(7), 1229–1243.
Hilgetag, C. C., O’Neill, M. A., & Young, M. P. (1996). Indeterminate organization of
the visual system—Rejoinder. Science, 271(5250), 777–777.
Hinton, G. E., & Sejnowski, T. J. (1986). Learning and relearning in Boltzmann machines.
In D. E. Rumelhart & J. L. McClelland (Eds.), Parallel Distributed Processing. Explora-
tions in the Microstructure of Cognition. Vol. 1: Foundations. Cambridge, MA: MIT Press.
Hodgkin, A. L., & Huxley, A. F. (1952). Currents carried by sodium and potassium ions
through the membrane of the giant axon of Loligo. Journal of Physiology, 116(4),
449–472.
Hodgkin, A. L., Huxley, A. F., & Katz, B. (1952). Measurement of current-voltage rela-
tions in the membrane of the giant axon of Loligo. Journal of Physiology, 116(4),
424–448.
Hodgkin, A. L., & Katz, B. (1949). The effect of sodium ions on the electrical activity
of giant axon of the squid. Journal of Physiology, 108(1), 37–77.
Hoogerwerf, A. C., & Wise, K. D. (1994). A 3-dimensional microelectrode array for
chronic neural recording. IEEE Transactions on Biomedical Engineering, 41(12),
1136–1146.
Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective
computational abilities. Proceedings of the National Academy of Sciences of the United States
of America-Biological Sciences, 79(8), 2554–2558.
Horvath, J. C., Carter, O., & Forte, J. D. (2014). Transcranial direct current stimulation:
Five important issues we aren’t discussing (but probably should be). [Review]. Frontiers
in Systems Neuroscience, 8, 2.
Horvath, J. C., Forte, J. D., & Carter, O. (2015a). Quantitative review finds no evidence
of cognitive effects in healthy populations from single-session transcranial; direct cur-
rent stimulation (tDCS). Brain Stimulation, 8(3), 535–550.
Horvath, J. C., Forte, J. D., & Carter, O. (2015b). Evidence that transcranial direct cur-
rent stimulation (tDCS) generates little-to-no reliable neurophysiologic effect beyond
MEP amplitude modulation in healthy human subjects: A systematic review. Neuro-
psychologia, 66, 213–236.
Horwitz, B. (2003). The elusive concept of brain connectivity. NeuroImage, 19(2),
466–470.
Hubel, D. H. (1957). Tungsten microelectrode for recording from single units. Science,
125(3247), 549–550.
Hubel, D. H., & Wiesel, T. N. (1959). Receptive fields of single neurones in the cat’s
striate cortex. The Journal of Physiology, 148, 574–571.
Hubel, D. H., & Wiesel, T. N. (1965). Receptive fields and functional architecture in 2
nonstriate visual areas (18 and 19) of cat. Journal of Neurophysiology, 28(2), 229–289.
Hubel, D. H., & Wiesel, T. N. (1968). Receptive fields and functional architecture of
monkey striate cortex. Journal of Physiology-London, 195(1), 215–243.
190 Bibliography

Izhikevich, E. M. (2003). Simple model of spiking neurons. IEEE Transactions on Neural

Networks, 14(6), 1569–1572.
Izhikevich, E. M., & Edelman, G. M. (2008). Large-scale model of the mammalian thalamo-
cortical system. PNAS, 105, 3593–3598.
James, W., & Rouben Mamoulian Collection (Library of Congress). (1890). The Principles
of Psychology. New York: H. Holt & Co.
Jenison, R. L., Rangel, A., Oya, H., Kawasaki, H., & Howard, M. A. (2011). Value encod-
ing in single neurons in the human amygdala during decision making. Journal of
Neuroscience, 31(1), 331–338.
Johnson, M. H. (2011). Interactive specialization: A domain-general framework for human
functional brain development? Developmental Cognitive Science 1(1):7–21. doi:10.1016/j.
dcn.2010.07.003. Epub 2010 Jul 30.
Johnstone, A. F. M., Gross, G. W., Weiss, D. G., Schroeder, O. H. U., Gramowski, A., &
Shafer, T. J. (2010). Microelectrode arrays: A physiologically based neurotoxicity test-
ing platform for the 21st century. Neurotoxicology, 31(4), 331–350.
Jones, D. K., Simmons, A., Williams, S. C. R., & Horsfield, M. A. (1999). Non-invasive
assessment of axonal fiber connectivity in the human brain via diffusion tensor MRI.
Magnetic Resonance in Medicine, 42(1), 37–41.
Jones, K. E., Campbell, P. K., & Normann, R. A. (1992). A glass silicon composite
intracortical electrode array. Annals of Biomedical Engineering, 20(4), 423–437.
Jung, R., & Spillmann, L. (1970). Receptive-field estimation and perceptual integration
in human vision. In F. A. Young & D. B. Lindsley (Eds.), Early Experience and Visual
Information Processing Perceptual and Reading Disorders (pp. 181–197). Washington, DC:
National Academy of Sciences.
Kandel, E. R. (1991). Cellular mechanisms of learning and the biological basis of individu-
ality. In E. R. Kandel & J. H. Schwartz (Eds.), Principles of Neural Science (3rd Edition)
(pp. 1009–1032). New York: Elsevier.
Kandel, E. R., & Tauc, L. (1964). Mechanism of prolonged heterosynaptic facilitation.
Nature, 202(492), 145–147.
Karp, R. M. (1986). Combinatorics, complexity, and randomness. Communications of the
Acm, 29(2), 98–109.
Karr, J. & Herman, D. (2004). Using a MEA to analyze a dissociated hipposcampal neural
network. http://web.mit.edu/9.29/www/neville_jen/jkarr/mea/
Keefer, E. W., Botterman, B. R., Romero, M. I., Rossi, A. F., & Gross, G. W. (2008).
Carbon nanotube coating improves neuronal recordings. Nature Nanotechnology, 3(7),
434–439.
Kennard, M. A. (1955). Effect of bilateral ablation of cingulate area on behavior. Journal
of Neurophysiology, 18, 159–169.
Kibler, A. B., Jamieson, B. G., & Durand, D. M. (2012). A high aspect ratio microelec-
trode array for mapping neural activity in vitro. Journal of Neuroscience Methods, 204(2),
296–305.
Killeen, P. R. (2013). The structure of scientific evolution. Behavior Analyst, 36(2),
325–344.
Ko, K. I. (1984). On the computational complexity of ordinary differential equations.
Information and Control, 58, 157–194.
Koch, C. (2012). Modular biological complexity. Science, 337, 531–532.
Kohl, J., & Jefferis, G. S. X. E. (2010). Neuroranatomy. Current Biology, 21 ,
R19–R20.
 Bibliography 191

Kohonen, T. (1977). Associative Memory: A System-Theoretical Approach. Berlin: Springer

Verlag.
Konorski, J. (1967). Integrative Activity of the Brain: An Interdisciplinary Approach. Chicago:
University of Chicago Press.
Kreiman, G., Koch, C., & Fried, I. (2000). Category-specific visual responses of single
neurons in the human medial temporal lobe. Nature Neuroscience, 3(9), 946–953.
Kruger, J., & Bach, M. (1981). Simultaneous recording with 30 microelectrodes in monkey
visual-cortex. Experimental Brain Research, 41(2), 191–194.
Kupfermann, I., Castellucci, V., Pinsker, H., & Kandel, E. (1970). Neuronal correlates of
habituation and dishabituation of gill-withdrawal reflex in Aplysia. Science, 167(3926),
1743–1745.
Kwong, K. K. (2012). Record of a single fMRI experiment in May of 1991. NeuroImage,
62(2), 610–612.
Lazer, D. (2015). The rise of the social algorithm. Science, 348, 1090–1091.
Leopold, D. A., Bondar, I. V., & Giese, M. A. (2006). Norm-based face encoding by
single neurons in the monkey inferotemporal cortex. Nature, 442(7102), 572–575.
Lettvin, J. Y., Maturana, H. R., Mcculloch, W. S., & Pitts, W. H. (1959). What the frogs
eye tells the frogs brain. Proceedings of the Institute of Radio Engineers, 47(11),
1940–1951.
Lewicki, M. S. (1998). A review of methods for spike sorting: The detection and clas-
sification of neural action potentials. Network-Computation in Neural Systems, 9(4),
R53–R78.
Lichtman, J. W. & Denk, W. (2011). The big and the small: challenges of imaging the
brain's circuits. Science, 334, 618–623.
Lindquist, K. A., Wager, T. D., Kober, H., Bliss-Moreau, E., & Barrett, L. F. (2012). The
brain basis of emotion: A meta-analytic review. Behavioral and Brain Sciences, 35(3),
121–143.
Ling, G., & Gerard, R. W. (1949). The normal membrane potential of frog sartorius
fibers. Journal of Cellular and Comparative Physiology, 34, 383–396.
Litke, A. M., Bezayiff, N., Chichilnisky, E. J., Cunningham, W., Dabrowski, W., Grillo,
A. A., . . . Sher, A. (2004). What does the eye tell the brain? Development of a
system for the large-scale recording of retinal output activity. IEEE Transactions on
Nuclear Science, 51(4), 1434–1440.
Lockery, S. R., & Goodman, M. B. (1998). Tight-seal whole-cell patch clamping of
Caenorhabditis elegans neurons. Ion Channels, Pt B, 293, 201–217.
Logothetis, N. K., Pauls, J., Augath, M., Trinath, T., & Oeltermann, A. (2001). Neuro-
physiological investigation of the basis of the fMRI signal. [Research Support, Non-U.S.
Gov’t]. Nature, 412(6843), 150–157.
Lorente de No, R. (1938). Synaptic stimulation of motoneurons as a local process. Journal
of Neurophysiology, 1, 195–206.
Lumer, E. D., Edelman, G. M., & Tononi, G. (1997a). Neural dynamics in a model of
the thalamocortical system. 1. Layers, loops and the emergence of fast synchronous
rhythms. Cerebral Cortex, 7(3), 207–227.
Lumer, E. D., Edelman, G. M., & Tononi, G. (1997b). Neural dynamics in a model of
the thalamocortical system. 2. The role of neural synchrony tested through perturba-
tions of spike timing. Cerebral Cortex, 7(3), 228–236.
Luo, Y. H., & da Cruz, L. (2014). A review and update on the current status of retinal
prostheses (bionic eye). British Medical Bulletin, 109, 31–44.
192 Bibliography

Maccione, A., Garofalo, M., Nieus, T., Tedesco, M., Berdondini, L., & Martinoia, S. (2012).
Multiscale functional connectivity estimation on low-density neuronal cultures recorded
by high-density CMOS Micro Electrode Arrays. Journal of Neuroscience Methods, 207(2),
161–171.
MacCorquodale, K., & Meehl, P. E. (1948). On a distinction between hypothetical con-
structs and intervening variables. Psychological Review, 55(2), 95–107.
MacNichol, E. F., Jr. (1956). Visual receptors as biological transducers. In R. G. Grenell &
L. J. Mullins (Eds.), Molecular Structure and Functional Activity of Nerve Cells (Publication
1, 34). Washington, D.C.: American Institute of Biological Sciences.
Margolis, D. J., Gartland, A. J., Singer, J. H., & Detwiler, P. B. (2014). Network oscil-
lations drive correlated spiking of ON and OFF ganglion cells in the rd1 mouse model
of retinal degeneration. PLoS One, 9(1), e86253. doi:10.1371/journal.pone.0086253
Markram, H. (2006). The Blue Brain Project. Nature Reviews Neuroscience, 7(2), 153–160.
Martin, S. J., Grimwood, P. D., & Morris, R. G. (2000). Synaptic plasticity and memory:
An evolution of the hypothesis. Annual Review Neuroscience, 23, 649–711.
McClelland, J. L., & Rumelhart, D. E. (1988). A simulation-based tutorial system for
exploring parallel distributed-processing. Behavior Research Methods Instruments & Com-
puters, 20(2), 263–275.
McClelland, J. L., Rumelhart, D. E., & The PDP Research Group (1988). Parallel Dis-
tributed Processing: Explorations in the Microstructure of Cognition. Vol 2 Psychological and
Biological Models. Cambridge, MA: MIT Press.
McCulloch, W. S. (1944). The functional organization of the cerebral cortex. Physiological
Reviews, 24(3), 390–407.
McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous
activity. Bulletin of Mathematical Biophysics, 5, 115–133.
Merolla, P. A., Arthur, J. V., Alvarez-Icaza, R., Cassidy, A. S., Sawada, J., Akopyan, F., . . .
Modha, D. S. (2014). Artificial brains: A million spiking-neuron integrated circuit
with a scalable communication network and interface. Science, 345(6197), 668–673.
Meyer, A. R. (1974). The inherent computational complexity of theories of ordered sets. Paper
presented at the Proceedings of the International Congress of Mathematicians, Van-
couver, Canada.
Minsky, M. L., & Papert, S. (1969). Perceptrons: An Introduction to Computational Geometry.
Cambridge, MA: MIT Press.
Moore, E. F. (1956). Gedanken—experiments on sequential machines. In C. E. Shannon &
J. McCarthy (Eds.), Automata Studies (pp. 129–153). Princeton, NJ: Princeton Uni-
versity Press.
Moore, G. E. (1965). Cramming more components onto integrated circuits. Electronics,
38(8), 114–117.
Moxon, K. A., Hallman, S., Sundarakrishnon, A., Wheatley, M., Nissanov, J., & Barbee,
K. A. (2009). Long-term recordings of multiple single-neurons for clinical applications:
The emerging role of the bioactive microelectrode. Materials, 2, 1762–1794.
Mukamel, D., Ruffo, S., & Schreiber, N. (2005). Breaking of ergodicity and long relaxation
times in systems with long-range interactions. Physical Review Letters, 95(24), 240604.
Murphy, T. H., Blatter, L. A., Wier, W. G., & Baraban, J. M. (1992). Spontaneous syn-
chronous synaptic calcium transients in cultured cortical-neurons. Journal of Neuroscience,
12(12), 4834–4845.
Nabavi, S., Fox, R., Perolux, C. D., Lin, J. Y., Tsien, R. Y., & Malinow, R. (2014). Engi-
neering a memory with LTD and LTP. Nature, 511, 348–352.
 Bibliography 193

Neri, P., & Levi, D. M. (2006). Receptive versus perceptive fields from the reverse-correlation
viewpoint. Vision Res, 46(16), 2465–2474.
Nicolelis, M. A. L., & Chapin, J. K. (1994). Spatiotemporal structure of somatosensory
responses of many-neuron ensembles in the rat ventral posterior medial nucleus of
the thalamus. Journal of Neuroscience, 14(6), 3511–3532.
Nordhausen, C. T., Maynard, E. M., & Normann, R. A. (1996). Single unit recording
capabilities of a 100 microelectrode array. Brain Research, 726(1–2), 129–140.
Ogawa, S., Lee, T. M., Kay, A. R., & Tank, D. W. (1990). Brain magnetic-resonance-
imaging with contrast dependent on blood oxygenation. Proceedings of the National
Academy of Sciences of the United States of America, 87(24), 9868–9872.
Olsen, S. R., & Wilson, R. I. (2008). Cracking neural circuits in a tiny brain: New
approaches for understanding the neural circuitry of Drosophila. Trends in Neurosciences,
31(10), 512–520.
Oosterwijk, S., Lindquist, K. A., Anderson, E., Dautoff, R., Moriguchi, Y., & Bennett, L.
F. (2012). States of mind: Emotions, body feelings, and thoughts share distributed
neural networks. NeuroImage 62, 2110–2128.
Palmer, S. E. (1999). Vision science: Photons to phenomenology. Cambridge, MA: MIT.
Papadimitriou, C. H., & Tsitsiklis, J. (1986). Intractable problems in control-theory. Siam
Journal on Control and Optimization, 24(4), 639–654.
Perrett, D. I., Rolls, E. T., & Caan, W. (1982). Visual neurones responsive to faces in the
monkey temporal cortex. Experimental Brain Research, 47(3), 329–342.
Pickel, V. M., & Segal, M. (Eds.). (2013). The Synapse: Structure and Function. Oxford, UK:
Academic Press.
Pillow, J. W., Shlens, J., Chichilnisky, E. J., & Simoncelli, E. P. (2013). A model-based
spike sorting algorithm for removing correlation artifacts in multi-neuron recordings.
PLoS One, 8(5), e62123. doi:10.1371/journal.pone.0062123
Pine, J. (1980). Recording action-potentials from cultured neurons with extracellular
micro-circuit electrodes. Journal of Neuroscience Methods, 2(1), 19–31.
Pinsker, H., Kupfermann, I., Castellucci, V., & Kandel, E. (1970). Habituation and disha-
bituation of the gill withdrawal reflex in Aplysia. Science, 167, 1740–1732.
Pitts, W., & McCulloch, W. S. (1947). How we know universals: The perception of audi-
tory and visual forms. Bulletin of Mathematical Biophysics, 9, 127–147.
Place, U. T. (1956). Is consciousness a brain process? British Journal of Psychology, 47(1),
44–50.
Poldrack, R. A. (2006). Can cognitive processes be inferred from neuroimaging data?
Trends in Cognitive Sciences, 10(2), 59–63.
Poldrack, R. A. (2010). Mapping mental function to brain structure: How can cognitive
neuroimaging succeed? Perspectives on Psychological Science, 5(6), 753–761.
Posner, M. I., Petersen, S. E., Fox, P. T., & Raichle, M. E. (1998). Localization of cognitive
operations in the human brain. Science, 240, 1627–1631.
Quine, W. V. O. (1951). Main trends in recent philosophy. Philosophical Review, 60, 20–43.
Quiroga, R. Q. (2012). Concept cells: The building blocks of declarative memory func-
tions. Nature Reviews Neuroscience, 13(8), 587–597.
Quiroga, R. Q., Nadasdy, Z., & Ben-Shaul, Y. (2004). Unsupervised spike detection and
sorting with wavelets and superparamagnetic clustering. Neural Computation, 16(8),
1661–1687.
Quiroga, R. Q., Reddy, L., Kreiman, G., Koch, C., & Fried, I. (2005). Invariant visual
representation by single neurons in the human brain. Nature, 435(7045), 1102–1107.
194 Bibliography

Rabi, I. I., Zacharias, J. R., Millman, S., & Kusch, P. (1937). A new method for measuring
nuclear magnetic moments. Physical Review, 53, 318.
Rabi, I. I., Zacharias, J. R., Millman, S., & Kusch, P. (1939). The molecular beam resonance
method for measuring nuclear magnetic moments. The magnetic moments of 3LI6,
3LI7, and 9F19. Physical Review, 55, 526–535.
Raizen, D. M., & Avery, L. (1994). Electrical-activity and behavior in the pharynx of C.
elegans. Neuron, 12(3), 483–495.
Rakover, S. S. (2011). A plea for methodological dualism and a multi-explanation approach
in psychology. Behavior and Philosophy, 39, 17–43.
Ratliff, F., Hartline, H. K., & Miller, W. H. (1963). Spatial and temporal aspects of retinal
inhibitory interaction. Journal of the Optical Society of America, 53, 110–120.
Richmond, J. E. (2006). Electrophysiological recordings from the neuromuscular junction
of C. elegans. In The C. elegans Research Community (Ed.), WormBook, 1–8.
Rosenblatt, F. (1958). The perceptron—A probabilistic model for information-storage and
organization in the brain. Psychological Review, 65(6), 386–408.
Roskies, A. L. (2010). How does neuroscience affect our conception of volition? Annual
Review of Neuroscience, 33, 190–130.
Rousche, P. J., & Normann, R. A. (1998). Chronic recording capability of the Utah
intracortical electrode array in cat sensory cortex. Journal of Neuroscience Methods, 82(1),
1–15.
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by
back-propagating errors. Nature, 323(6088), 533–536.
Rumelhart, D. E., McClelland, J. L., & The PDP Research Group (1988). Parallel Dis-
tributed Processing. Explorations in the Microstructure of Cognition. Vol 1: Foundations.
Cambridge, MA: MIT Press.
Schwarz, D. A., Lebedev, M. A., Hanson, T. L., Dimitrov, D. F., Lehew, G., Meloy, J., . . .
Nicolelis, M. A. (2014). Chronic, wireless recordings of large-scale brain activity in
freely moving rhesus monkeys. Nature Methods, 11, 670–676.
Searle, J. R. (1990). The Mystery of Consciousness. New York: New York Review of
Books.
Searle, J. R. (1997, March 6). Consciousness and the Philosophers. New York: New York
Review of Books.
Sejnowski, T. J., Kienker, P. K., & Hinton, G. E. (1986). Learning symmetry groups with
hidden units—Beyond the perceptron. Physica D-Nonlinear Phenomena, 22(1–3),
260–275.
Seker, E., Berdichevsky, Y., Begley, M. R., Reed, M. L., Staley, K. J., & Yarmush, M. L.
(2010). The fabrication of low-impedance nanoporous gold multiple-electrode arrays
for neural electrophysiology studies. Nanotechnology, 21(12). doi:10.1088/0957-4484/
21/12/125504
Selfridge, O. G. (1958). Pandemonium: A paradigm for learning. Paper presented at the
Proceeding of a Symposium the Mechanization of Thought Processes, London.
Service, R. F. (2014). The brain chip. Science, 345(6197), 614–616.
Sevush, S. (2006). Single-neuron theory of consciousness. Journal of Theoretical Biology,
238(3), 704–725.
Sherrington, C. S. (1940). Man on His Nature. Cambridge, UK: The University Press.
Shlens, J., Field, G. D., Gauthier, J. L., Greschner, M., Sher, A., Litke, A. M., & Chich-
ilnisky, E. J. (2009). The structure of large-scale synchronized firing in primate retina.
Journal of Neuroscience, 29(15), 5022–5031.
 Bibliography 195

Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-positive psychology: Undis-
closed flexibility in data collection and analysis allows presenting anything as significant.
Psychological Science, 22(11), 1359–1366.
Skinner, B. F. (1950). Are theories of learning necessary? Psychological Review, 57,
193–216.
Spillmann, L. (2014). Receptive fields of visual neurons: The early years. [Research Sup-
port, Non-U.S. Gov’t]. Perception, 43(11), 1145–1176.
Spillmann, L., Dresp-Langley, B., & Tseng, C.-H. (2015). Beyond the classical receptive
field: The effect of contextural stimuli. Journal of Vision, 15(9), 7, 1–23.
Spillmann, L., & Werner, J. S. (Eds.). (1990). Visual Perception: The Neurophysiological
Foundations. San Diego: Academic Press.
Spira, M. E., & Hai, A. (2013). Multi-electrode array technologies for neuroscience and
cardiology. Nature Nanotechnology, 8(2), 83–94.
Sporns, O. (2011). Networks of the Brain. Cambridge, MA: MIT Press.
Sporns, O., Tononi, G., & Kotter, R. (2005). The human connectome: A structural descrip-
tion of the human brain. PLoS Computational Biology, 1(4), 245–251.
Steidl, E. M., Neveu, E., Bertrand, D., & Buisson, B. (2006). The adult rat hippocampal
slice revisited with multi-electrode arrays. Brain Research, 1096, 70–84.
Stevenson, I. H., Rebesco, J. M., Hatsopoulos, N. G., Haga, Z., Miller, L. E., & Kording,
K. P. (2009). Bayesian inference of functional connectivity and network structure from
spikes. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 17(3),
203–213.
Stockmeyer L. J., & Chandra, A. K. (1979). Provably difficult combinatorial games, SIAM
Journal on Computing, 8(2), 151–174.
Takeuchi, T., Duszkiewicz, A. J., & Morris, R. G. (2013). The synaptic plasticity and
memory hypothesis: Encoding, storage and persistence. [Research Support, Non-U.S.
Gov’t]. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences,
369(1633), 20130288.
Tanaka, K. (1993). Neuronal mechanisms of object recognition. Science, 262(5134),
685–688.
Thomas, C. A., Jr., Springer, P. A., Loeb, G. E., Berwald-Netter, Y., & Okun, L. M.
(1972). A miniature microelectrode array to monitor the bioelectric activity of cultured
cells. Experimental Cell Research, 74(1), 61–66.
Thompson, R. F., & Spencer, W. A. (1966). Habituation—A model phenomenon for
study of neuronal substrates of behavior. Psychological Review, 73(1), 16–43.
Thyreau, B., Schwartz, Y., Thirionm, B., Frouin, V., Loth, E., Vollstädt-Klein, S., . . .
IMAGEN Consortium. (2012). Very large fMRI study using the IMAGEN database:
Sensitivity-specificity and population effect modeling in relation to the underlying
anatomy. NeuroImage, 61(1), 295–303.
Turella, L., Pierno, A. C., Tubaldi, F., & Castiello, U. (2009). Mirror neurons in humans:
Consisting or confounding evidence? Brain and Language, 108(1), 10–21.
Turing, A. (1950). Computing machinery and intelligence. Mind, 59, 433–460.
Turing, A. (2004). Lecture on the automatic computing engine. In B. J. Copeland (Ed.),
The Essential Turing (pp. 362–394). Oxford: Oxford University Press.
Turkeltaub, P. E., Eden, G. F., Jones, K. M., & Zeffiro, T. A. (2002). Meta-analysis of the
functional neuroanatomy of single-word reading: Method and validation. NeuroImage,
16(3), 765–780.
Underwood, E. (2014). Brain Project meets physics. Science, 344, 954–955.
196 Bibliography

Uttal, W. R. (1959). A comparison of neural and psychophysical responses in the som-

esthetic system. Journal of Comparative and Physiological Psychology, 52(4), 485–490.
Uttal, W. R. (1973). The Psychobiology of Sensory Coding. New York: Harper & Row.
Uttal, W. R. (1975). An Autocorrelation Theory of Form Detection. Hillsdale, NJ and New
York: L. Erlbaum Associates; Distributed by Halsted Press.
Uttal, W. R. (1978). The Psychobiology of Mind. Hillsdale, NJ: Erlbaum.
Uttal, W. R. (1990). On some 2-way barriers between models and mechanisms. Perception
& Psychophysics, 48(2), 188–203.
Uttal, W. R. (2001). The New Phrenology: The Limits of Localizing Cognitive Processes in the
Brain. Cambridge, MA: MIT Press.
Uttal, W. R. (2004). Dualism: The Original Sin of Cognitivism. Mahwah, NJ: Erlbaum.
Uttal, W. R. (2005). Neural Theories of Mind. Mahwah, NJ: Erlbaum.
Uttal, W. R. (2009). Distributed Neural Systems: Beyond the New Phrenology. Cornwall-on-
Hudson, NY: Sloan Pub.
Uttal, W. R. (2011). Mind and Brain: A Critical Appraisal of Cognitive Neuroscience. Cam-
bridge, MA: MIT Press.
Uttal, W. R. (2013). Reliability in Cognitive Neuroscience: A Meta-Meta-Analysis. Cambridge,
MA: MIT Press.
Uttal, W. R. (2016). Macroneural Theories in Cognitive Neuroscience. New York: Psychology
Press.
Uttal, W. R., & Krissoff, M. (1966). Effect of stimulus pattern on temporal acuity in
somatosensory system. Journal of Experimental Psychology, 71(6), 878–883.
Vacariu, G. (2012). Cognitive Neuroscience Versus Epistemologically Different Worlds. Bucharest,
Romania: University of Bucharest Press.
Vacariu, G., & Vacariu, M. (2013). Troubles with cognitive neuroscience. Philosophia
Scientiae, 17, 151–170.
Valbo, A. B., & Hagbarth, K. E. (1968). Activity from skin mechanoreceptors recorded
percutaneously in a wake human subjects. Experimental Neurology, 21, 270–289.
Van Essen, D. C., Anderson, C. H., & Felleman, D. J. (1992). Information processing in
the primate visual system: An integrated systems perspective. Science, 255, 419–423.
Van Orden, G. C., & Papp, K. R. (1997). Functional neuroimages fail to discover pieces
of the mind in the parts of the brain. Philosophy of Science Proceedings, 64, 585–594.
Vimal, R. L. P. (2009). Meanings Attributed to the Term “Consciousness”: An overview.
Journal of Consciousness Studies, 16(5), 9–27.
Vogelstein, J. T., Park, Y., Ohyama, T., Kerr, R. A., Truman, J. W., Priebe, C. E., & Zlatic,
M. (2014). Discovery of brainwide neural-behavioral maps via multiscale unsupervised
structure learning. [Research Support, Non-U.S. Gov’t, U.S. Gov’t, Non-P.H.S.]. Science,
344(6182), 386–392.
von Neumann, J. (1945). First Draft of a Report on the EDVAC. Philadelphia: University
of Pennsylvania.
von Philipsborn, A. C., Liu, T., Yu, J. Y., Masser, C., Bidaye, S. S., & Dickson, B. J.
(2011). Neuronal control of Drosophila courtship song. [Comparative Study Research
Support, Non-U.S. Gov’t]. Neuron, 69(3), 509–522.
Wagenaar, D. A., Madhavan, R., Pine, J., & Potter, S. M. (2005). Controlling bursting in
cortical cultures with closed-loop multi-electrode stimulation. Journal of Neuroscience,
25(3), 680–688.
Wagenaar, D. A., Pine, J., & Potter, S. M. (2006). Searching for plasticity in dissociated
cortical cultures on multi-electrode arrays. [Research Support, N.I.H., Extramural
Research Support, Non-U.S. Gov’t]. Journal of Negative Results Biomedicine, 5, 16.
 Bibliography 197

Ward, A. A., & Thomas, L. B. (1955). The electrical activity of single units in the cerebral
cortex of man. Electroencephalography and Clinical Neurophysiology, 7(1), 135–136.
Wark, H. A., Sharma, R., Mathews, K. S., Fernandez, E., Yoo, J., Christensen, B., Tresco,
P. . . . Tathireddy, P. (2013). A new high-density (25 electrodes/mm²) penetrating
microelectrode array for recording and stimulating sub-millimeter neuroanatomical
structures. Journal of Neural Engineering, 10(4):045003. doi:10.1088/1741-2560/
10/4/045003.
Watson, J. D., & Crick, F. H. C. (1953). Molecular structure of nucleic acids—A structure
for deoxyribose nucleic acid. Nature, 171(4356), 737–738.
Westheimer, G. (1990). Relating neural mechanisms to visual perception. In L. Spillman &
J. S. Werner (Eds.), Visual Perception: The Neurophysiological Foundations. San Diego:
Academic Press.
White, J. G., Southgate, E., Thomson, J. N., & Brenner, S. (1986). The structure of the
nervous-system of the nematode C. elegans. Philosophical Transactions of the Royal Society
of London Series B-Biological Sciences, 314(1165), 1–340.
Wild, J., Prekopcsak, Z., Sieger, T., Novak, D., & Jech, R. (2012). Performance comparison
of extracellular spike sorting algorithms for single-channel recordings. Journal of Neu-
roscience Methods, 203(2), 369–376.
Wilson-Sanders, S. E. (2011). Invertebrate models for biomedical research, testing, and
education. ILAR Journal, 52(2), 126–152.
Yoshimura, Y., Dantzker, J. L. M., & Callaway, E. M. (2005). Excitatory cortical neurons
form fine-scale functional networks. Nature, 433(7028), 868–873.
Young, J. Z. (1936). The giant nerve fibres and epistellar body of cephalopods. Quarterly
Journal of Microscopical Science, 78, 367–386.
Young, M. P., & Yamane, S. (1992). Sparse population coding of faces in the inferotem-
poral cortex. Science, 256(5061), 1327–1331.
Yule, G. U. (1926). Why do we sometimes get nonsense-correlations between time-series?—
A study in sampling and the nature of time-series. Journal of the Royal Statistical Society,
89, 1–69.
Zecevic, D., Wu, J. Y., Cohen, L. B., London, J. A., Hopp, H. P., & Falk, C. X. (1989).
Hundreds of neurons in the Aplysia abdominal-ganglion are active during the gill-
withdrawal reflex. Journal of Neuroscience, 9(10), 3681–3689.
Zeng, F. G., Rebscher, S., Harrison, W., Sun, X., & Feng, H. (2008). Cochlear implants:
System design, integration, and evaluation. [Research Support, N.I.H., Extramural
Research Support, Non-U.S. Gov’t Review]. IEEE Reviews in Biomedical Engineering,
1, 115–142.
Zhang, Y. L., Liu, R. Y., Heberton, G. A., Smolen, P., Baxter, D. A., Cleary, L. J., & Byrne,
J. H. (2012). Computational design of enhanced learning protocols. Nature Neurosci-
ence, 15(2), 294–297.
Zrenner, E. (2002). Will retinal implants restore vision? [Research Support, Non-U.S.
Gov’t Review]. Science, 295(5557), 1022–1025.
This page intentionally left blank
INDEX

Note: Figures and tables are denoted with italicized page numbers; endnote information is
denoted with an n and note number following the page number.

acetylcholine 38 as 12–13; search for objectivity as 14–15;

Amdahl’s law 150, 173 underdetermination as 24–5
amplifiers 46, 47–8 behavioral or descriptive theories 2–3
analysis techniques, microneuronal Bell’s number 106
network: amplitude measurements from Blue Gene computer 168, 169
94–5; complexity considerations with BRAIN (Brain Research through
98–9; connectivity maps from 91–3, 92, Advancing Innovative
97–103, 101; interval periodicity from Neurotechnologies) project 170, 175n5
94; with multiple electrode arrays Braun, Karl Ferdinand 46
91–103; shape recognition from 95; bursting activity, neuronal 87–8, 131
spike action potential clustering from
93–7 Caenorhabditis elegans (nematode)
Aplysia californica (sea hare) 116, 125–30, 130–4, 133
127, 131, 134 Cajal, Santiago Ramón y 37
Atanasoff-Berry computer design 143 cathode ray oscilloscope 46
autocorrelation theory of form detection cell assemblies 70, 71
175n2 clamps, patch and voltage 48–9
axons 40–1, 43 cognitive neuroscience: assumptions
in analysis of 5–8, 25; barriers to
back propagation technique 157 theory development in 10–16, 23–5;
barriers to theory development: macroneural theory in 2, 3, 9–10, 16–32,
complexity as 11–12; false analogies 34–5, 177; microneuronal theory in (see
between sensory, motor, and cognitive microneuronal theory and practice);
processes as 15–16; inaccessibility search for explanation of mind–
as 13–14; levels of analysis as 11; brain connection in 1–8; taxonomy
macroneural 23–5; microneuronal of theories in 2–5; technological
10–16; neurologizing of psychological influences on theories of 8–10 (see also
language as 16; pooling process as 23–4; technological advances); see also mind–
poor definition of cognitive constructs brain connection
200 Index

cognitive reductive theories 3 connectivity: anatomical 98; arbitrary

combinatorics of complexity 11–12, 33, thresholds for 101, 101; connectionism
103–7, 174, 180 on 157; connectivity maps on 91–3,
complexity: as barrier to theory 92, 97–103, 101, 110–11; effective 98;
development 11–12; Bell’s number functional 98; Micro Connectionist
measuring 106; combinations related Theory on 142–3; synaptic (see synapses
to 33, 105, 180; combinatorial and synaptic connectivity)
11–12, 33, 103–7, 174, 180; consciousness 12, 58, 158–9, 173
definition of 98–9; factorials and convergence/progressive convergence
superfactorials related to 33, 104–5, 50–1, 53, 59
159; graph theory applied to 106; Courtship Singing 136–7, 138
intermediate level neuronal network Cray-1 146, 147
research consideration of 131; Cray X-MP 147
large-scale computer simulations Crookes, William/Crookes tube 46, 145
of cognition challenges of 158–9,
174; microneuronal network theory Dale, Henry H. 38
consideration of 33, 69, 75, 77, 98–9, DARPA (Defense Advanced Research
103–7, 110–11, 180; noise as element Projects Agency) SyNAPSE Project
of 107; permutations related to 105 170–2
computational neuroscience 116–17 decision-making, value metrics in 57
computer simulations/calculations: De Forest, Lee 46
Amdahl’s law on 150, 173; Atanasoff- dendrites 40
Berry computer design for 143; Blue Drosophila melanogaster (fruit fly)
Gene computer for 168, 169; central 134–7, 138
processing unit for 144, 145–50,
153, 155; cognitive neuroscience edge enhancement 55, 117, 118, 118–23,
application of parallel processing in 119, 140n1, 156
151–6; combinatorial complexity of EEG techniques 9–10, 11, 17, 23, 49, 56,
microneuronal networks and speed of 111, 166
106; Cray-1 for 146, 147; Cray X-MP Eintoven, Willem 45
for 147; Eniac for 147; facial recognition electrodes: high-impedance 47–8;
in 149, 155–6; Floating Point Operations micro- 44–5, 47–8, 73, 76–103, 108–9,
per Second metric for 150–1, 151; 110, 126, 128, 143; recording 41–3;
input and output devices for 144, stimulating 43
145; intrinsically parallel vs. serial electron microscopes 40
tasks in 149–50; large-scale computer Eniac 147
simulations of cognition as 33, 141–75, Erlanger, Joseph 46
180–1; memory technology for 144, European Human Brain Project
145, 148–9; Moore’s Law on 33, 106, 168–70, 172
147; Nengo software system for 167–8; exclusivity question 63, 65, 179
Neurogrid project to develop hardware
for 167; parallel processing in 145–74; facial recognition: computer vs. cognitive
parallel processing theories of cognitive process of 149, 155–6; single neuron
processes for 156–72; perceptron as theory on 39, 51, 57, 58, 60, 64
precursor to 156–7; Sequoia for 153, Floating Point Operations per Second
171; serial computers for 143–6, 153–4, (FLOPS) 150–1, 151
157; Tianhe-2 for 153; TrueNorth fMRI (functional Magnetic Resonance
device for 171–2; Turing computer Imaging) techniques 9–10, 11, 17, 18–
design for 144; virtual supercomputers 19, 21–2, 23–4, 27, 30–1, 49, 61, 65, 111
for 148; von Neumann computer design
for 144, 144, 153, 154, 171; Xeon Galvani, Luigi 9
computer for 168, 175n3 Gasser, Herbert 46
connectionism 157 Gerlach, Joseph von 37
 Index 201

glial cells 41 neuronal networks 113–40, 156, 180;

Golgi, Camillo 37 of learned chemoaversion 132–4, 133;
“grandmother” neurons 53, 65 Limulus polyphemus (horseshoe crab) as
graph theory 106 117, 118–23, 119, 131, 134, 139, 156;
Loligo vulgaris (squid) as 43, 114; of
habituation/dishabituation 116, 124, 126, Mach Band 117, 118, 118–23, 119, 156;
128–30, 131, 170; see also Nonassociative mathematical models of 121, 122, 123,
Learning 139; of Nonassociative Learning 116,
Hebb, Donald: Hebb’s conjecture 70–6, 123–30, 127, 131; plausibility tests using
109, 110, 111, 180 116, 122, 139; of single neurons 43, 114;
Herman, Daniel 100–1 as surrogates for human studies 116
HRL 170
Human Brain Project (HBP) 168–70, 172 Karr, Jonathan 100–1

IBM 153, 168, 169, 170–2 large-scale computer simulations of

impedances 46, 47–8 cognition: autocorrelation theory of
Intel Corporation 147 form detection as 175n2; cognitive
intermediate level neuronal networks: neuroscience application of parallel
analogous behavior extrapolated for processing in 151–6; complexity
115–16, 121–2, 130, 138–9; Aplysia challenges in 158–9, 174; computer
californica and Nonassociative Learning advances leading to 143–56; DARPA
research on 116, 123–30, 127, 131, SyNAPSE Project as 170–2; European
134; Caenorhabditis elegans and Human Brain Project as 168–70,
Learned Chemoaversion research on 172; facial recognition in 149, 155–6;
130–4, 133; complexity considered Floating Point Operations per Second
in 131; criteria for models of 114–15; metric for 150–1, 151; goals or
Drosophila melanogaster and Courtship motivations for 160–3, 164, 167, 168,
Singing research on 134–7, 138; 170–2, 174; “hopeful” approach to 162;
heuristics or theory stimulation interim summary on 172–4; Micro
about 116; interim summary of Connectionist Theory research use of
research on 137–40; invertebrate 142–3; in microneuronal theory and
system as model of 113–40, 156, 180; practice 33, 141–75, 180–1; necessary vs.
Limulus polyphemus and Mach Band sufficient thresholds in 158; Neurogrid
research on 117, 118, 118–23, 119, project as 167; overarching conclusions
131, 134, 139, 156; mathematical drawn from 159; overview of 141–3,
models of 121, 122, 123, 139; in 172–4; parallel processing in 145–74;
microneuronal theory and practice parallel processing theories of cognitive
113–40, 156, 180; neurocomputing processes for 156–72; perceptron as
or computational neuroscience study precursor to 156–7; serial computers
of 116–17; overview of 113–18, 137, in 143–6, 153–4, 157; Spaun project
139; plausibility tests of 116, 122, 139; as 167–8; spiking neuron simulation as
surrogates for human studies of 116 165–6; thalamocortical system model
invertebrate system models: analogous as 164–5, 166; United States BRAIN
behavior extrapolated from 115–16, project as 170, 175n5; see also computer
121–2, 130, 138–9; Aplysia californica simulations/calculations
(sea hare) as 116, 125–30, 127, 131, learning: learned chemoaversion as 132–4,
134; Caenorhabditis elegans (nematode) as 133; meaning of 12; Nonassociative
130–4, 133; of Courtship Singing 136–7, Learning as 116, 123–30, 127, 131
138; criteria for 114–15; Drosophila Leeuwenhoek, Anton van 8
melanogaster (fruit fly) as 134–7, 138; Limulus polyphemus (horseshoe crab) 117,
heuristics or theory stimulation from 118–23, 119, 131, 134, 139, 156
116; interim summary of research Linpack 150
using 137–40; of intermediate level Loligo vulgaris (squid) 43, 114
202 Index

Long-Term Depression (LTD) 74–5, simulations of cognition in 33, 141–75,

111–12n1 180–1; macroneural vs. 2, 3, 9–10, 16–
Long-Term Potentiation (LTP) 71, 72, 32, 34–5, 177; microneuronal network
74–5, 111–12n1 theories in 33, 67–112, 180; single
Lowei, Otto 38 neuron practice and theory in 37–66,
114, 155, 179; technological influences
Mach Band 55, 117, 118, 118–23, 119, on 8–10 (see also technological advances)
140n1, 156 microscopes 8, 40
Mach, Ernst 118 mind–brain connection: barriers to
macroneural theory 2, 3, 9–10, 16–32, theory development in 10–16, 23–5;
34–5, 177 macroneural theory on 2, 3, 9–10,
membranes, neuron: definition and 16–32, 34–5, 177; microneuronal theory
description of 38; large molecules on (see microneuronal theory and
penetrating 41; membrane potential of practice); psychoneural equivalent in 5,
48; microelectrodes penetrating 44–5; 49–56, 58–9, 61, 62, 63–5, 179; search
patch and voltage clamps on 48–9 for explanation of 1–8; single neuron
Micro Connectionist Theory (μct) 142–3 practice and theory addressing 49–59;
microelectrodes: bottom-up theory built technological influences on theories of
on data from 143; invertebrate system 8–10 (see also technological advances);
model research using 126, 128; multiple see also cognitive neuroscience
electrode arrays of 73, 76–103, 108–9, mirror neurons 57–8, 61–2
110; single neuron practice and theory models see invertebrate system models
informed by 44–5, 47–8 Moore’s Law 33, 106, 147
microneuronal network theories: Morgan, Thomas Hunt 134
analysis techniques for 91–103; cell multiple electrode arrays (MEAs): analysis
assemblies in 70, 71; combinatorics techniques using 91–103; applications
applied to 33, 103–7, 180; complexity of 86–90; for cochlear implants 88–9,
considerations in 33, 69, 75, 77, 98–9, 89; complexity considerations with
103–7, 110–11, 180; connectivity maps 98–9; connectivity maps from 91–3, 92,
on 91–3, 92, 97–103, 101, 110–11; 97–103, 101; limitations of 77–8, 83–6,
criteria to authenticate 73; definition 93, 96–103, 108–9, 110; microneuronal
and description of 67; empirical testing network theories explored using 73,
of 72–6; Hebb’s conjecture on 70–6, 76–103, 108–9, 110; for neuronal
109, 110, 111, 180; interim conclusions bursting activity research 87–8; for
on 108–11; in microneuronal theory neurotoxicity research 87; penetrating
and practice 33, 67–112, 180; multiple microelectrode arrays as 78–81, 79, 80;
electrode array use on 73, 76–103, planar microelectrode arrays as 81–6,
108–9, 110; optogenetic techniques in 82, 83; for retinal network/prosthetics
research on 74, 111–12n1; overview of research 86–7, 89–90, 90; spike action
67–70, 108–11; phase sequences in 70; potential clustering from 93–7; types of
principles of 67–8; register metaphor 78–86
for 68, 68–9; spike action potential myelin coating 40
clustering in 93–7; synaptic conductivity
or efficacy in 70, 71, 72, 74–5; toy necessary vs. sufficient thresholds: in
network models of 69, 92, 92, 101, 104, intermediate level neuronal network
106, 110 research 115, 117; in large-scale
microneuronal theory and practice: computer simulations of cognition 158;
assumptions in analysis of 5–8; barriers in microneuronal network theories
to development of 10–16; conclusions 67; top-down vs. bottom-up theory
about 34–5, 177–81; definition and approaches to 6–7
description of 4; intermediate level Nengo software system 167–8
neuronal networks in 113–40, 156, 180; neurocomputing 116–17
introduction to 33; large-scale computer Neurogrid project 167
 Index 203

neuronal networks: intermediate level Schwann cells 40

neuronal networks on 113–40, 156, 180; selective stimulus sensitivity 58–9
microneuronal network theories on 33, sensitization 124, 128, 129
67–112, 180; single neuron practice and Sequoia 153, 171
theory vs. 49, 51–2, 64–5 Sherrington, Charles S. 38, 52
Neuron Doctrine 37–8, 49 single neuron practice and theory:
neurons: “grandmother” 53, 65; conflated vs. causal correlations in
microneuronal theory and practice 54, 58, 60–1; on consciousness 58;
on (see microneuronal theory critique of 58–64; on decision-making
and practice); mirror 57–8, 61–2; 57; definition and description of 49;
number of, in human brain 142, 152; exclusivity question on 63, 65, 179; on
structure and function of 39–41 (see facial recognition 39, 51, 57, 58, 60,
also membranes, neuron); synapses 64; factors engendering 59; “gnostic”
connecting (see synapses and synaptic neuron interpretations of 60–1; interim
connectivity); transmitter substances comments on 64–5; invertebrate system
with 38, 40, 41 as model of 43, 114; limitations of/
neuroreductive theories 3–4, 7 problems with 50–1, 60–4, 155, 179;
neurotransmitters 38, 40, 41 in microneuronal theory and practice
“Nodes of Ranvier” 40 37–66, 114, 155, 179; mind–brain
noise: amplifiers addressing 46, 48; as problem addressed by 49–59; on mirror
element of complexity 107; multiple neurons 57–8, 61–2; neuronal network
electrode arrays picking up 79, 84, 94–5, theory vs. 49, 51–2, 64–5; on Perceptive
97, 110; signal-to-noise ratios 22, 84, Field/Receptive Field equivalents
94–5, 97, 110 55–6; progressive convergence theory vs.
Nonassociative Learning (NAL) 116, 50–1, 53, 59; psychoneural equivalent
123–30, 127, 131 in 49–56, 58–9, 61, 62, 63–5; on
nuclear magnetic resonance (NMR) 18 selective stimulus sensitivity 58–9; on
sensory processing 56–7; single neurons,
olfactory aversive learning 132–4, 133 generally 37–9; single neuron theories,
ontological assumption 179 specifically 52–8; sparse vs. distributive
optogenetic and optical techniques 74, encoding theories applied to 64–5; on
111–12n1, 175n4 spatial frequency patterns 55; structure
and function of neurons in 39–41;
patch clamps 48–9 technological basis of 39–49
penetrating microelectrode arrays 78–81, spatial frequency patterns 55
79, 80 Spaun project 167–8
Perceptive Field/Receptive Field spike action potential (SAP) clustering:
equivalents 55–6 amplitude measurements with 94–5;
perceptron 156–7 interval periodicity analysis with 94;
planar microelectrode arrays 81–6, 82, 83 microneuronal network analysis of 93–7;
progressive convergence 50–1, 53, 59 shape recognition with 95
psychoneural equivalent: definition and spiking neuron simulation 165–6
description of 5; single neuron practice stimulating electrodes 43
and theory based on 49–56, 58–69, 61, string galvanometer 45–6
62, 63–5; ultimate theoretical goal to sufficient threshold see necessary vs.
find 179 sufficient thresholds
supercomputer simulations see large-scale
recording devices 45–6 computer simulations of cognition
recording electrodes 41–3 SyNAPSE (Systems of Neuromorphic
reductive vs. nonreductive theories 3–4, 6, Adaptive Plastic Scalable Electronics)
7, 179 Project 170–2
register metaphor 68, 68–9 synapses and synaptic connectivity:
Ruska, Ernst 38 chemical vs. electrical 40; discovery
204 Index

of 38; Hebb’s conjecture on 70, 71, recording electrodes as 41–3; stimulating

72, 74–5; intermediate level neuronal electrodes as 43; string galvanometer as
network research on 128–30; large- 45–6; tDCS as 29; technological basis of
scale computer simulations of 165–72; macroneural theories identified via
number of, in human brain 152; 17–19; technological basis of single
optogenetic techniques in research on neuron theories identified via 39–49;
74, 111–12n1 triode vacuum tube amplifiers as 46;
voltage clamps as 48–9
tDCS (transcranial Direct Current thalamocortical system model 164–5, 166
Stimulation) 29 theories: behavioral or descriptive 2–3;
technological advances: amplifiers cognitive reductive 3; macroneural 2, 3,
and impedances 46, 47–8; cathode 9–10, 16–32, 34–5, 177; methodological
ray oscilloscope as 46; computer origins of 5; microneuronal (see
simulations/calculations as 33, 106, microneuronal theory and practice);
141–75, 180–1; Crookes tube as 46, neuroreductive 3–4, 7; reductive vs.
145; EEG techniques as 9–10, 11, 17, nonreductive 3–4, 6, 7, 179; taxonomy
23, 49, 56, 111, 166; electricity and of types of 2–5; top-down vs. bottom-
electrical measuring instruments as up 4, 6–7, 179
8–9; electron microscopes as 40; fMRI Tianhe-2 153
techniques as 9–10, 11, 17, 18–19, 21–2, transmitter substances 38, 40, 41
23–4, 27, 30–1, 49, 61, 65, 111; high- TrueNorth device 171–2
input impedance preamplifiers as 46, Turing computer design 144
48; high-magnification microscopes
as 8; microelectrodes as 44–5, 47–8, United States BRAIN project 170, 175n5
73, 76–103, 108–9, 110, 126, 128,
143; microneuronal network theories voltage clamps 48–9
driven by 111, 111; microneuronal von Bekesy, Georg 89
theory development influenced by von Gerlach, Joseph 37
8–10; multiple electrode arrays as 73, von Neumann computer design 144, 144,
76–103, 108–9, 110; nuclear magnetic 153, 154, 171
resonance as 18; optogenetic and optical
techniques as 74, 111–12n1, 175n4; Waldmeyer, Wilhelm 37
patch clamps as 48–9; perceptron as
156–7; recording devices as 45–6; Xeon computer 168, 175n3