OLA8 Utterances

LabEx Empirical Foundations of Linguistics, Project OLA 8:  Non sentential utterances in child and adult speech

Coordinator: Jonathan Ginzburg (U. Paris Diderot)

Members:  Judit Gervain (U. Paris Descartes), Dimitra Kolliakou  (U. Paris Diderot), Sara Moradlou (U. Paris Diderot)

External members: Shalom Lappin (King's College, London), Eve Clark (Stanford University)

Description

         Until recently formal and computational analysis of language was concentrated on isolated sentences in monologue or text. In recent years there has been growing interest in providing analysis of actually occurring spoken language (see e.g. Asher & Lacsarides, 2003; Poesio & Rieser, 2011; Ginzburg, 2012). This work has shown inter alia that both non-sentential utterances and repair (self-corrections and clarification requests), long assumed to be highly irregular and difficult to describe formally, can be classified in terms of relatively straightforward taxonomies (Heeman & Allen, 1999; Fernández & Ginzburg, 2002; Purver, Ginzburg & Healey, 2003; Schlangen & Rodriguez, 2004). Furthermore, the semantic rules used to determine their meaning are entirely canonical, given a sufficiently rich theory of context (Ginzburg & Sag, 2000; Schlangen, 2003; Fernandez, 2006; Ginzburg, 2012).

         One finds many highly insightful studies of the linguistic competence of infants (see e.g. Gomez & Gerken, 2000; Tomasello, 2001; Nazzi, 2005; Kinzler, Dupoux & Spelke, 2007). Among child language researchers there is increasing insistence on studying acquisition within the context of a natural conversational setting (Clark, 2002; Veneziano, 2010; Morgenstern, Parisse & Sekali, 2011). Nonetheless, hitherto there has been little work on modeling adult/child conversation using the formal and computational tools that have been used to study adult/adult conversation (though see Ginzburg & Kolliakou, 2009; Cooper & Larsson, 2009, for examples of such analyses of child/adult interaction.). Within the current project, we will study dialogue at the pre-syntactic stage, illustrated in (1a-c) from the COLAJE corpus (Parisse & Morgenstern, 2011):

[MOT=mother,CHI=child,OBS=observer]:

(1)

(a) MOT:  dans le bateau .                      CHI (MADEL): ah les poussins .

(b)  MOT:  un escargot ?                        CHI ( ANAE): oh là !             

(c) *OBS: hou, joli petit sac à dos . *CHI: maman . *OBS: c'est `a maman?

         This is actually an intriguing period because many children's productions at this stage that are viewed as communicatively successful by the adult do not transparently reflect adult input. Furthermore, the adult often provides a response that aids the acquisition of sentential utterances:

(2)

(a) MOT:  il n' est pas nécessaire de crier Antoine tu sais  CHI (antoi): cassé . (COLAJE)

(b) MOT:  c'est les barbes à papa !CHI ( ANAE): ça (COLAJE)

(c) [contexte visuel: l'enfant tient une poupée et  montre une saleté du doigt sur sa tête]

 *CHI:  enlever .

*CHI: bébé.

*CHI: laver .

*CHI: l'enlever .

*MOT: tu l'as lavé hier .

*MOT: oui .

*CHI:  bébé

*MOT: oui ah mais on n'a pas réussi à  enlever ça. (COLAJE)

         What is needed is a formal theory of conversational interaction that can describe precisely the range of meanings that children's utterances at this stage can take (from the perspective of an adult interacting with the child), the range of feedback the adult can supply, and how the child's system evolves into a more syntactically/semantically complex system, based on the input and the feedback she receives and various cognitive and grammatical priors.

         Ignoring this stage represents a significant lacune in understanding the acquisition process. Eliminating this lacune will enable us to start resolving a variety of fundamental questions, such as evaluating the extent conversational interaction (and in particular conversational interaction) are crucial for acquisition (Demetras et al 1986, Gathercole and Hoff 2006); discovering how the intrinsically contextual resolution of one/two word utterances is supplanted by more syntactically complex utterances.

         Our overriding aim is the development of a theory of parent/child interaction at the early stages of language emergence. This involves providing both a precise description of the utterances that the child produces on the grammatical and contextual levels, the parent's input and reactions to the child's utterances, and a theory of language emergence based on division of labour between input and priors. Given the importance of visual, as well as linguistic context for this work, our data will be drawn from a number of multimodal corpora in English and French---our current work is on the Providence corpus from CHILDES (Demuth, 2006)

Publications

Ginzburg, J. & Moradlou, S. (submitted). One word utterances: grammar and interaction.

Ginzburg, J. & Kolliakou, D. (in revision). Avoid repetition: the emergence of clarification requests.