Social knowledge contextualizes syntactic representations in development Franklin Chang, Tessei Kobayashi and Shigeaki Amano NTT Communication Sciences Laboratories, NTT Corp., Kyoto, Japan [email protected]

Abstract

Results Experiment 1

Work in syntax acquisition assumes that developmental mechanisms are similar in different social contexts. We examined whether social knowledge could influence the learning of a new structure by adding a social manipulation to a syntax acquisition task (Weird Word Order task, Akhtar, 1999). In the first study, we found that 4-year-old children used a novel structure less in the standard WWO paradigm than in a more socially-oriented paradigm. In the second study, we examined input factors over development in our new paradigm and found that social factors contextualize the roles of input in syntactic development.

Social knowledge • Social knowledge is knowledge that we believe our interlocutor possesses. • Example: Which language(s) does our interlocutor know or prefer?

Found the children often match VERBOBJECT structure with robot, but not when talking to Japanese human (F(1,18) = 43.9, p < 0.001). No difference in OBJECT-VERB (F(1,18) = 0.5, p = 0.5) or OBJECT-TRANSVERB (F(1,18) = 1.2, p = 0.3)

Conclusion: Contrary to the assumption of independence between social factors and syntax, we found a strong influence of social context on the utterances produced.

Syntax acquisition • Syntax acquisition paradigms are not socially oriented. • No interlocutor (e.g., Gomez & Lakusta, 2004) • Scene description (e.g., Akhtar, 1999) • Syntax acquisition mechanisms are independent of social factors • Gradual lexically-based abstraction mechanisms (Abbot-Smith, et al, 2001)

A social syntax acquisition task • Changed social factors in a syntax acquisition task (Weird Word Order, Akhtar, 1999) • Standard weird word order (WWO) teaches a novel structure like “Tamming Big Bird the truck” for a novel action. Test description of novel action with different arguments. Child can say WWO “Tamming Elmo the car” or canonical English “Elmo tamming the car” • Social WWO paradigm - Search for hidden stickers by commanding robot dog with WWO structure to point or look at sticker locations. Child is given experience showing that the robot understand WWO structures. Standard WWO paradigm Interlocutor knows child’s native language Shared knowledge of the scene

Social WWO paradigm Interlocutor only understands WWO Knowledge is not fully shared

Experiment 2 How do social factors interact with input in syntactic development? • Previous work has found that WWO use diminishes… – Over age (Abbot-Smith, et al., 2001; Akhtar, 1999) – Frequency of verbs (Matthews, et al., 2005) • Test input factors in Robot-context WWO task –3 and 4 years –Novel (pate, dote) and familiar verbs (yubisashite, mite) • Factors • Input Strength -> gradual abstraction -> less WWO • Social Strength -> more WWO  Predictions of Two Accounts in Robot WWO task (see Figures)  Independent Account - Input Strength + Social Strength  Contextualize Account - Only use strength factor that is appropriate (Chang, Dell, & Bock, 2006) Design: Between subjects: Novel/Familiar verbs, 3- and 4-years-old, 40 monolingual Japanese children.

Results Experiment 2

Japanese and Weird Word Order Language Japanese language Verb-final, particles (optional) apple point (“ringo yubisashite”, point at the apple) apple look (“ringo mite”, look at the apple) WWO language point apple “yubisashite ringo” look apple “mite ringo”

point

look

Experiment 1 Between Subjects: Robot vs. Human 20 monolingual Japanese children, age 4;6 (range 4;1 - 4;7) Task Demonstrate robot controlled by WWO (e.g., “yubisashite ringo”, point apple) Robot interlocutor Human interlocutor Prompt child to command robot Experimenter commands robot “Tell the robot what to do” “What happened?” “Tell the robot what to do with the apple” “What happened to the apple?” “Is point or look better?” “What did the robot point at?” Prompt with verb three times. Coding “yubisashite X” “mite X” -> VERB-OBJECT “X yubisashite” “X mite” -> OBJECT-VERB “X sagashita” “X mitsuketa” -> OBJECT-TRANSVERB Presented at the 20th Annual CUNY Sentence Processing Conference, San Diego, 2007

• VERB-OBJECT ANOVA (left) main effect of age (F(1,36) = 8.7, p = 0.006), no effect of verb type (F(1,36) = 0.1, p = 0.750), no age by verb type interaction (F(1,36) = 0.1, p = 0.794) • OBJECT-VERB ANOVA (right) no significant effects (age, F(1,36) = 0.1, p = 0.822; verb type, F(1,36) < 0.01, p = 1.00; age by verb type, F(1,36) = 2.5, p = 0.12).

 Contrary to previous studies, WWO use increased over development, canonical use did not increase over development, and Japanese verb-structure regularities were not available in the robot context. Support for Contextualized Account.

Conclusion • Learning of a novel syntactic structure differs in different social contexts • Trajectory over development cannot be explained unless we assume that social knowledge contextualizes the influence of canonical representations.

References Abbot-Smith, K., Lieven, E., & Tomasello, M. (2001). What pre-school children do and do not do with ungrammatical word orders. Cognitive Development, 16, 1-14. Akhtar, N. (1999). Acquiring basic word order: Evidence for data-driven learning of syntactic structure. Journal of Child Language, 26(2), 339-356. Chang, F., Dell, G. S., & Bock, J. K. (2006). Becoming syntactic. Psychological Review, 113(2), 234-272. Matthews, D., Lieven, E., Theakston, A., & Tomasello, M. (2005). The role of frequency in the acquisition of English word order. Cognitive Development, 20(1), 121-136. Gomez, R. L., & Lakusta, L. (2004). A first step in form-based category abstraction by 12month-old infants. Developmental Science. 7(5), 567-580.