Yael Greenberg
Hi! I am a generative linguist working at the department for English literature and Linguistics at Bar Ilan University (Israel). My research area is semantics and its interfaces, especially with syntax, pragmatics and prosody. Some of my main research interests are:- Focus and alternative sensitive operators, Scalarity and polarity, Gradability, Speech acts and metalinguistic operations, Genericty and modality, Event semantics.
less
InterestsView All (14)
Uploads
(1) (Context: We are arriving late to a committee meeting evaluating John’s and Bill’s achievements, and manage to hear what one of the committee members says about Bill):
a.. …and during the last 3 years he only wrote [3]F papers. > 5 is ‘a little’
b. …and during the last 3 years he even wrote [5]F papers . > 5 is ‘a lot’
I examine three schematic suggestions which did try to capture the scalar antonymy of only and even, namely that only vs. even presuppose that their prejacent, p is lower vs. higher (respectively), (a) than what is expected / what is the default standard in the context (the ‘mirative / evaluative antonymy’ view), (b) than some (salient) alternative in the set of contextually relevant focus alternatives, C, (the ‘existential antonymy’ view), or (c) than all alternatives in C (the ‘superlative antonymy’ view). The differences between these views are often subtle, and in many contexts the predictions they make empirically overlap. One goal of the paper, then, is to examine the felicity and interpretation of only and even in a wider range of contexts than has been done so far, so we end up with better diagnostics for teasing apart the predictions of each of these view. A second goal is to argue that this examination supports the preferability of the ‘superlative’ view over the ‘existential’ and the ‘mirative / evaluative’ views, and to discuss ways to handle issues which were raised against this view. Finally, I hope to discuss potential implications and open questions raised by the examination of the scalar antonymy of only and even regarding the way contextual factors affect the construction of sets of alternatives, and compare them to parallel implications and questions raised in experimental studies on activation / processing of alternatives and in research on Scalar Implicatures.
In light of such parameters we first examine the family of even-like particles in Modern Hebrew (which hasn’t been done so far), mainly afilu, ve-lu, af and bixlal. We show that accounting for the full range of differences between these particles requires (a) adding a new parameter to the existing typologies, namely the (in)ability to operate over domain-based/degree-based alternatives (cf. Greenberg 2014, Chierchia 2013) and (b) a fined-grained characterization of existing ‘context/discourse’-based parameters, e.g. the ability to function as ‘discourse-even’ on corrections/denials, besides questions. This last parameter may be more generally characterized as (in)ability to operate over speech act alternatives.
We then compare the resulting even-like typology with the (as yet much more limited) typologies developed for only-like particles in e.g. Beaver & Clark,2008, Coppock & Beaver,2014 (English), Tomaszewicz,2012B (Polish) and Orenstein & Greenberg,2012, Orenstein,2016 (Hebrew). This comparison reveals striking parallels in the parameters along which the Hebrew particles within both the even-family and the only-family vary. Such ‘shared’ parameters include (in)ability to operate over degree/domain-based alternatives, operation over logically-based vs. contextually-supplied scales and (in)ability to operate on questions and corrections/denials (i.e. over speech acts of different sorts).
We suggest that, together with reported ‘flipped’ readings of some even-like particles as only-like (cf. Gast & Van-der-Awuera 2011, Tomaschevish,2012B, Grubic & Zimmermann,2011), such parallels further motivate the need for a unified and precise typology for the family of scalar particles cross-linguistically, and we take first steps in identifying some core properties of such a typology.
The main goal of this paper is to examine cases where a focus sensitive expression can also operate on another type of alternatives, which I will call ‘internal’ alternatives. Intuitively, such alternatives are not alternatives to p, but rather alternative interpretational versions of p. That is, with internal alternatives, the focused element is not replaced by a distinct element (of the same semantic type), but rather with an identical element with a distinct interpretation. Thus, the overt material in p and the propositions in the set of internal alternatives is identical, and the difference only lies in the interpretation given to covert material in p.
More formally, I will propose that in the case of ‘internal’ altenatives the alternatives to the prejacnet are derived by assigning a covert variable in the focused element a distinct value. Depending on the semantics of the associate, this covert variable can be a standard variable (with gradable expressions), a covert quantifier, a domain restriction variable, a precision standard variable, etc.
The main operator I will analyze in this respect is the Hebrew 'bixlal' (cf. Migron 2003, Greenberg & Khrizman 2012). I will claim that it is an 'even-like operator, which unlike the standard 'even' (and the Hebrew 'afilu') can operate on such 'internal alternatives.
Integrating these two views we suggest that assertion speech acts should be themselves modeled as gradable, and more specifically, that the covert speech act operator ASSERT denotes a (credence) degree relation, and is modifiable by overt and covert degree modifiers, manipulating this degree. We show that such a view enables capturing novel parallels between overt and covert degree modification of adjectives at the propositional level and some overt and covert modifiers of assertion speech acts.
(1) (Context: We are arriving late to a committee meeting evaluating John’s and Bill’s achievements, and manage to hear what one of the committee members says about Bill):
a.. …and during the last 3 years he only wrote [3]F papers. > 5 is ‘a little’
b. …and during the last 3 years he even wrote [5]F papers . > 5 is ‘a lot’
I examine three schematic suggestions which did try to capture the scalar antonymy of only and even, namely that only vs. even presuppose that their prejacent, p is lower vs. higher (respectively), (a) than what is expected / what is the default standard in the context (the ‘mirative / evaluative antonymy’ view), (b) than some (salient) alternative in the set of contextually relevant focus alternatives, C, (the ‘existential antonymy’ view), or (c) than all alternatives in C (the ‘superlative antonymy’ view). The differences between these views are often subtle, and in many contexts the predictions they make empirically overlap. One goal of the paper, then, is to examine the felicity and interpretation of only and even in a wider range of contexts than has been done so far, so we end up with better diagnostics for teasing apart the predictions of each of these view. A second goal is to argue that this examination supports the preferability of the ‘superlative’ view over the ‘existential’ and the ‘mirative / evaluative’ views, and to discuss ways to handle issues which were raised against this view. Finally, I hope to discuss potential implications and open questions raised by the examination of the scalar antonymy of only and even regarding the way contextual factors affect the construction of sets of alternatives, and compare them to parallel implications and questions raised in experimental studies on activation / processing of alternatives and in research on Scalar Implicatures.
In light of such parameters we first examine the family of even-like particles in Modern Hebrew (which hasn’t been done so far), mainly afilu, ve-lu, af and bixlal. We show that accounting for the full range of differences between these particles requires (a) adding a new parameter to the existing typologies, namely the (in)ability to operate over domain-based/degree-based alternatives (cf. Greenberg 2014, Chierchia 2013) and (b) a fined-grained characterization of existing ‘context/discourse’-based parameters, e.g. the ability to function as ‘discourse-even’ on corrections/denials, besides questions. This last parameter may be more generally characterized as (in)ability to operate over speech act alternatives.
We then compare the resulting even-like typology with the (as yet much more limited) typologies developed for only-like particles in e.g. Beaver & Clark,2008, Coppock & Beaver,2014 (English), Tomaszewicz,2012B (Polish) and Orenstein & Greenberg,2012, Orenstein,2016 (Hebrew). This comparison reveals striking parallels in the parameters along which the Hebrew particles within both the even-family and the only-family vary. Such ‘shared’ parameters include (in)ability to operate over degree/domain-based alternatives, operation over logically-based vs. contextually-supplied scales and (in)ability to operate on questions and corrections/denials (i.e. over speech acts of different sorts).
We suggest that, together with reported ‘flipped’ readings of some even-like particles as only-like (cf. Gast & Van-der-Awuera 2011, Tomaschevish,2012B, Grubic & Zimmermann,2011), such parallels further motivate the need for a unified and precise typology for the family of scalar particles cross-linguistically, and we take first steps in identifying some core properties of such a typology.
The main goal of this paper is to examine cases where a focus sensitive expression can also operate on another type of alternatives, which I will call ‘internal’ alternatives. Intuitively, such alternatives are not alternatives to p, but rather alternative interpretational versions of p. That is, with internal alternatives, the focused element is not replaced by a distinct element (of the same semantic type), but rather with an identical element with a distinct interpretation. Thus, the overt material in p and the propositions in the set of internal alternatives is identical, and the difference only lies in the interpretation given to covert material in p.
More formally, I will propose that in the case of ‘internal’ altenatives the alternatives to the prejacnet are derived by assigning a covert variable in the focused element a distinct value. Depending on the semantics of the associate, this covert variable can be a standard variable (with gradable expressions), a covert quantifier, a domain restriction variable, a precision standard variable, etc.
The main operator I will analyze in this respect is the Hebrew 'bixlal' (cf. Migron 2003, Greenberg & Khrizman 2012). I will claim that it is an 'even-like operator, which unlike the standard 'even' (and the Hebrew 'afilu') can operate on such 'internal alternatives.
Integrating these two views we suggest that assertion speech acts should be themselves modeled as gradable, and more specifically, that the covert speech act operator ASSERT denotes a (credence) degree relation, and is modifiable by overt and covert degree modifiers, manipulating this degree. We show that such a view enables capturing novel parallels between overt and covert degree modification of adjectives at the propositional level and some overt and covert modifiers of assertion speech acts.
on the scalar presupposition of even, according to which the prejacent of even, p, is stronger than its relevant focus alternatives, q.
To that end I first examine both familiar challenges for the popular ‘comparative likelihood’ view of the ‘stronger than’ relation, as well as novel challenges, having to do with the context dependency of even (with entailed and non-entailed alternatives) and with its sensitivity to standards of comparison.
To overcome these challenges and to account for the full range of data I develop a revised, ‘gradability-based’ scalar presupposition for even, which differs from the ‘comparative likelihood’ one in several respects: instead of directly comparing degrees to which propositions (namely p and q) are more or less likely, we compare extents to which non-focus entities x in p and q (in the accessible p worlds and the exhaustified q worlds) exceed the salient standard on a scale associated with a contextually supplied gradable property G.
To capture cases where information about contrastive topics is crucial for fixing two distinct standards on G, I follow theories which view even as a general, two-place alternative-sensitive operator, allowing it to associate with both focus and contrastive topics.
Beyond the ability to account for a large range of intricate felicity variations and inferences found with even, a more general contribution of the paper lies in showing the linguistic relevance of tools originally developed in the literature on gradable predicates to the semantics of scalar alternative–sensitive particles.
three novel problems for this `comparative likelihood' view, having to do with (a) cases where even p is felicitous though p cannot be considered less likely than q, (b) cases where
even p is infelicitous though p asymmetrically entails and is less likely than q, and (c) cases where even interacts with gradable predicates, indicating that merely requiring p to be higher on the scale than q is not enough to make even p felicitous. Instead, both p and q must also yield degrees which are at least as high as the standard of comparison.
In response to these problems I develop a revised scalar presupposition for even which resembles the semantics of comparative conditionals, and which requires that for a salient
x, retrieved from p, and a salient gradable property G, (i) x's degree on G is higher in all accessible p worlds than in all accessible q-and-not-p worlds and that (ii) in the latter
worlds this degree is at least as high as the standard on G. I show how this presupposition accounts for both traditional observations concerning even, as well as for the novel data and
propose that the common presence of `less likely' inferences with even can be indirectly derived from the common use of `distributional' standards of comparison with gradable
properties. A general contribution of the proposal, then, is in attempting to apply tools from research of gradability-based phenomena for a better understanding of scalarity-based
phenomena.
be-sax ha-kol differs from classical exclusives particles like only in that it is more flexible with respect to the set of alternatives to its prejacent. In particular, it can operate not only on "Roothian" alternatives to the prejacent, but also on different interpretational versions of the prejacent. We show how this proposal accounts for the fact that unlike only, be-sax ha-kol can trigger not only a clearly 'exclusive' reading, but also an 'approximative' one. We discuss the projective behavior of the prejacent of be-sax ha-kol in this reading, and the fact that it is
infelicitous with L(ower) –scale adjectives.
Thus, besides the mirror imaged scalar ordering (between p and its all distinct alternatives in C), ‘only’ and ‘even’ should not be seen as ‘mirative / evaluative’ mirror images. Neither should they be taken as mirror images in terms of exclusivity vs. additivity, due to independently made claims indicating that ‘even’ is not a true additive particle. I show that adopting the ‘hybrid’ entry of ‘only’ (but not the ‘non-scalar’ or ‘scalar’ entries), where scalarity and exclusivity are separated components, allows us to capture this limited mirror imaged picture, and take ‘only’ to be a member of the typology of scalar particles cross linguistically. Finally, I discuss parallel challenges for the universal quantification over alternatives in the scalar presuppositions of both ‘only’ and ‘even’, and argue that they can be uniformly solved by appropriately constraining the balance between the contribution of discourse salient material vs. lexicon for constructing alternatives in C, though the precise way to do that still requires research.
To account for the challenging data I develop a revised scalar presupposition for 'even', which integrates previous intuitions concerning the discourse role of 'even', an existing analysis of comparative correlatives as comparative conditionals, and insights from the research on the positive forms of gradable adjectives. I propose to derive the fact that the 'less likely' inference often, though not always, arises with 'even' from the interaction of the revised presupposition with independent claims about the way 'distributional' standards of comparison, are often, though not always, used with (relative) gradable properties. I show that cases where the ‘less likely’ inferences do not arise with even are those where non-distributional standards, e.g. so-called ‘functional’ standards, are at work in the revised presupposition. A more general contribution of the proposal, then, is in applying insights and tools used in the research of gradability to the research of scalarity and scalar operators.
I finish the paper with open questions concerning the applicability and productivity of the revised presupposition in accounting for the behavior of overt even with disjunctive alternatives and in Downward Entailing, as well as that of covert even, argued to be involved in the interpretation of Negative polarity items.