2018 Friday Poster 6473
Friday, November 2, 2018 | Poster Session I, Metcalf Small | 3pm
What input gap is there across socioeconomic status for complex syntax? A quantitative and cognitive modeling analysis of linguistic evidence for learning syntactic islands
A. Bates, L. Pearl
While there are known differences in the quantity and quality of child-directed speech (CDS) across socio-economic status (SES) for some linguistic knowledge (Schwab & Lew-Williams, 2016), it is unknown if these differences also extend to complex syntactic knowledge. We investigate wh-dependency constraints, known as syntactic islands, as a concrete case where the quantity and quality of high-SES CDS has been assessed (Pearl & Sprouse, 2013 (PS2013)). Using quantitative analysis and cognitive modeling to assess low-SES CDS samples, we find that low-SES children’s complex syntactic input, in terms of wh-dependencies, is quantitatively and qualitatively similar to that of high-SES children: the wh-dependencies (i) have similar distributions in the high-SES and low-SES input samples, and (ii) would allow a low-SES child to success- fully acquire knowledge of the same syntactic islands that a high-SES child would from high-SES input. Interestingly, at least one key building block for syntactic island knowledge comes from a different source in low-SES children’s input, but is crucially still present. This suggests that the linguistic evidence for more complex syntactic knowledge like syntactic islands, in contrast with more foundational linguistic knowl- edge, may not differ by SES.
Syntactic islands are thought to constrain which wh-dependencies a language allows. For example, English adults judge What did Jack think that Lily bought what? as more acceptable than What did Jack wonder if Lily bought what?; this has been explained as the second wh-dependency crossing an adjunct island (Ross, 1967). Sprouse, Wagers, and Phillips (2012) identified a behavioral signature of syntactic island knowledge (see Figure 1). PS2013 conducted a quantitative analysis of high-SES CDS to assess its wh-dependency distribution and then constructed a cognitive model implementing a syntactic islands learning strategy; their model demonstrated that high-SES children could successfully learn to have that same behavioral signature.
We assessed low-SES CDS from a subpart of the HSLLD corpus (Dickinson & Tabors, 2001) containing 31K utterances and 4K wh-dependencies, directed at 78 children between ages 3 and 5. Using KL divergence (Kullback & Leibler, 1951), we find that the wh-dependency distribution in low-SES CDS is very similar to that of high-SES CDS.We then assessed qualitative similarity in terms of learning outcomes. Using the same cognitive learning model developed by PS2013, we find that the modeled learners learning from low-SES CDS can internalize knowledge that allows them to generate the behavioral signal of syntactic island knowledge (Figure 1).
One key component that allows this acquisition is the presence of wh-dependencies in the CDS input that involve complementizer that: while the specific dependencies containing this complementizer differ between low-SES and high-SES CDS, both CDS samples contain dependencies of this kind, highlighting another important qualitative similarity across SES.
Our results suggest that once low-SES children have the core linguistic knowledge to leverage the wh– dependency information in their input, those data are sufficient for them to acquire syntactic islands con- straints just as high-SES children do. Future work can investigate other complex syntactic phenomena to see if the necessary linguistic evidence is also similar across SES.
References
Dickinson, D. K., & Tabors, P. O. (2001). Beginning literacy with language: Young children learning at home and school. Paul H Brookes Publishing.
Kullback, S., & Leibler, R. A. (1951). On information and sufficiency. The annals of mathematical statistics, 22(1), 79–86.
Pearl, L., & Sprouse, J. (2013). Computational Models of Acquisition for Islands. In J. Sprouse & N. Hornstein (Eds.), Experimental Syntax and Islands Effects (pp. 109–131). Cambridge: Cambridge University Press.
Ross, J. (1967). Constraints on variables in syntax (Unpublished doctoral dissertation). MIT, Cambridge, MA.
Schwab, J. F., & Lew-Williams, C. (2016). Language learning, socioeconomic status, and child-directed speech. Wiley Interdisci- plinary Reviews: Cognitive Science, 7, 264-275.
Sprouse, J., Wagers, M., & Phillips, C. (2012). A test of the relation between working memory capacity and syntactic island effects. Language, 88(1), 82–124.