Sonja Eisenbeiss

Dieser Workshop gibt einen Überblick über Korpora, lexikalische Datenbanken und Tools, die man zur Materialentwicklung für Unterricht und Forschung bzw. für die Datenauswertung verwenden kann. Dabei werden wir diskutieren, welche Daten und Tools sich für welche Zwecke eignen und wie man sie für das eigene Projekt kombiniert.
U.a. werden vorgestellt: das Deutsche Referenzkorpus und die darauf basierenden Wortlisten, die Leipzig Corpora Collection, die childLex-Datenbank, die CHILDES-Kindersprachkorpora, Sketchengine-Tools zur Korpuserstellung und Auswertung, der Multimedia-Annotator ELAN zur Annotation von Video/Audiodateien, Toolbox und das Natural Language Processing Toolkit für Korpusstudien sowie verschiedene Tagger, mit denen man Wörtern in Korpora ihre jeweilige grammatische Kategorie oder grammatische Merkmale zuweisen kann (z.B. TreeTagger). Wir werden auch kurz diskutieren, wie auch Anfänger R und Python bei der Arbeit mit Korpora und Datenbanken einsetzen können.

This talk gives an overview of research on Early Language Development in the new Centre for Language Development through the Lifespan (LaDeLi), University of Essex. It was part of the launch ceremony for LaDeLi and covers three research areas: (i) very early language development, (ii),  cross-cultural studies on children’s grammatical development, and (iii)  language games and child-directed speech.

The majority of experimental studies in linguistics, psychology, and the social sciences involve participants who are undergraduate students in research-active universities or children of educated families in societies that are WEIRD (Western, Educated, Industrialised, Rich, Democratic; see Henrichs et al. 2010). This leads to claims about universals of human language and behaviour that are not based on an appropriate empirical basis. It also violates the Principle of Justice as many populations are excluded from such studies and their benefits, for instance the development of appropriate materials for teaching or speech and language therapy. Hence, more and more experimental linguists have started to study previously under-researched languages and populations; and we are providing resources and information to support such projects (http://experimentalfieldlinguistics.wordpress.com/).
These projects pose a broad range of ethical challenges. Some of them are challenges that any “traditional” linguistic fieldworker has to face, for instance avoiding coercion and guaranteeing informed consent when dealing with communities that are poor and characterised by low levels of education. However, the introduction of experimental methods into fieldwork contexts also gives rise to new ethical problems. In particular, we will discuss ethical issues that arise when standardized tests of performance (e.g. IQ-tests or tests of working memory) are carried out in small, close-knit communities where native speakers from the same community may become involved in the analysis of such data. We will also discuss (i) problems caused by data sharing in cross-linguistic and cross-cultural studies and (ii) conflicts that can occur when inter-disciplinary studies require ethical approval from boards with members from different disciplines (e.g. medical sciences, linguistics, psychology, and anthropology).

Keywords: linguistics, justice, psycholinguistics, experiments, ethics

The Role of Morphological Structure in the Processing of Complex Forms: Evidence from Setswana Deverbative Nouns

Naledi Kgolo1, Sonja Eisenbeiss1, Nancy Kula1

1. Department of Language and Linguistics, University of Essex, UK, [email protected] 

Keywords: Derivational processing; Setswana; visual word recognition; masked priming.

Current research on morphological processing focuses on the question of whether the morphological structure of a complex word like neat-ness plays a role in processing or whether morphological effects can be reduced to the combined effects of shared forms and meanings (e.g. Feldman, 2000). This paper contributes to this debate by investigating deverbative nouns in Setswana, a Bantu language. Setswana Class 9 derivations are not readily segmentable into an obvious stem and affix (tsheko-seka ‘court case-stand trial’) in contrast to other noun classes such as Class 1, where the prefix mo- and suffix -i transparently indicate deverbal nouns (e.g. moseki-seka ‘the accused- stand trial’). Moreover, Setswana has "pseudo-derived nouns", which look as if they are derived from a verb, but are in fact not morphologically or semantically related to a verb (kgabo-gaba ‘large fire-pull stomach in’). 
We present results of (i) a frequency analysis for deverbative nouns in an existing corpus of Setswana (Otlogetswe, 2010), (ii) a subjective frequency rating survey with 25 participants, (iii) a visual word-non-word lexical decision experiment with 83 participants, and (iv) a masked priming experiment with 53 participants. The findings show that frequency measures derived from the existing Setswana corpus correlate significantly with the subjective frequency ratings from the survey, suggesting that they reflect speakers’ intuitions despite being based on more formal written texts. Moreover, the lexical decision experiment shows significant word-form frequency effects for both Class 1 and Class 9 derivations: the more frequently speakers encounter a word form, the faster they recognise it as a word. This suggests that stored complex word forms play a role in the processing of both noun classes under investigation. In the masked priming task, prior presentation of a morphologically related prime-form (kitso-ITSE ‘knowledge-KNOW’) speeded recognition times for both deverbative classes significantly, compared to a baseline with unrelated primes (bobi-ITSE ‘spider web-KNOW’). However, reaction times for the morphologically related conditions were significantly slower than the reaction times for the identity priming conditions. Such partial priming effects did not occur for pseudo derivations (morafe-RAFA ‘tribe-PILE ON’) or for forms that only overlap in form (moroba-ROBA ‘fun-BREAK’). This suggests that stored word forms may play a role in morphological processing, but that morphological structure affects the processing of complex word forms independently of pure form overlap. This supports models of morphological processing in which both whole-word representations and morphological structure play a role.
References
Feldman, L. B. (2000). Are Morphological Effects Distinguishable from the Effects of Shared Meaning and Shared Form? Journal of Experimental Psychology: Learning, Memory and Cognition, 26, 1431-1444.
Otlogetswe, T.J. (2010). Setswana Sketch Engine Corpus. http://www.sketchengine.co.uk/

Many parents, teachers, and speech therapists employ games to encourage communication and make language learning fun and effective. We will give an overview of different types of language games and show that such games play a crucial role in research on language acquisition and communication. We will also discuss which properties of learners’ input can support language learning - and how one can incorporate these properties into the design of language games.

Using SharePoint to Link Research and Teaching
In this session, I will discuss how I used SharePoint (a Microsoft product that provides a platform for online collaboration and document management) to develop a Virtual Research Environment (VRE) for the Psycholinguistic Research Group (PRG) in the Department of Language and Linguistics, University of Essex. This hour-long session will touch on the process of developing a VRE, explore the advantages of implementing such a system and demonstrate how SharePoint can be used for group work in a postgraduate research-skills module.

The Acquisition of German Adnominal Possessive Constructions



Previous studies on the acquisition of adnominal possessive constructions (APCs) have mainly focused on the availability of possessive markers and constraints on their use (e.g. Armon-Lotem et al. 2005, Eisenbeiss 2000, Marinis 2002, Radford/Galasso 1998). In this presentation, we will investigate incremental extension both in the range of constructions and in the range of possessive relations
We analysed 64 recordings from 7 monolingual German children (1;11-3;6), assigned to stages of noun phrase development by Eisenbeiss (2000): The rate of overt D-elements (determiners, possessive pronouns and quantifiers) is initially low in stage I, rising to 60-64% at the end of this stage. In II, the overt-D rate drops to 4-42%, increasing gradually in III and reaching target-like values in IV. This U-shaped development suggests reanalysis, which is supported by the observation that in I, D-elements occur in formulaic predicate+D-combinations (e.g. das-is-ein-X ‘that-is-a-X’, ≤74% of overt D) or in a few D+noun-combination types (<10 per file). Earlier analyses of these recordings demonstrated that possessive markers only appear in stages II and III and show initial lexical restrictions to individual nouns (Eisenbeiss 2000). Our new analysis has shown the following:
• Not all children use APCs early on: Hannah does not produce any APCs in I/II, but only precursors, such as single-word utterances that consist of the Possessor’s name or a possessive pronoun.
• APCs emerge incrementally: In I, Leonie only uses kinship term or proper name possessives (papas hose ‘daddy’s trousers’), in II/III, she starts using possessive pronouns (meine mama ‘my mummy’). Annelie, and Mathias produce both possessive pronouns and kinship term or proper name possessives in stage I/II; and so does Andreas, for whom we only have datat from stage III. Prepositional constructions (die pelle von der wurst ‘the skin of the sausage’) only appear in stage IV data from Carsten, Hannah, and Svenja.
• The percentage of pronominal Possessors increases over time: I: 30%, II:33%, III:77%, IV: 86%. Initially, possessive pronouns only appear with a few noun types (meine Mama ‘my mommy’,…).
• Children extend the range of possessive relations they encode in APCs - from ownership and kinship relations with human possessors (from stage I) via body part relations (from stage III) to part-whole relations for inanimate objects (stage IV).
• In stages III and IV, we found 10 utterances where a legal or habitual ownership relation is encoded noun-phrase internally and a temporary ownership or physical control relation is encoded at the sentential level (e.g. Mathias 3;4: der hat deine uhr ‘this-one has your clock’). This suggests that children start to distinguish between these types of possessive relations.
Taken together, our analysis shows an incremental extension both in the range of constructions and in the range of possessive relations that are encoded by these constructions. Specifically, we found that types of possession that involve physical control and proximity are acquired earlier than more abstract notions of possession. Finally, we observed a preference to position the Possessor before the Possessum even when this results in a highly marked word order pattern (Carsten 3;6: von wurst die pelle ‘of sausage the skin’). We will interpret our results on the basis of typological studies on possessive constructions (Heine 1997, Seiler 1983); arguing that children encode more prototypical possessive relations earlier than less prototypical ones.

In this presentation, I will
(i) challenge some views about children’s linguistic input and the way we respond to children’s utterances.
(ii) give an introduction to (our) current research on child-directed speech.
(iii) present some resources for student projects on child-directed speech and children’s own language.