Paper presented at the 4th International Conference on the Evolution of Language. Cambridge, MA.

The Case of Cases and Word Order: The Role of Syntactic Cues in the Evolution and Acquisition of Language

Gary Lupyan & Morten H. Christiansen

Abstract

While long-standing theories describe evolution of language via the evolution of a Chomskyan universal grammar, an alternative approach that is gaining ground is the adaptation of linguistic structures to the human brain rather than vice versa (e.g., Christiansen, 1994; Kirby, 1998). On this account, language universals may reflect non-linguistic, cognitive constraints on learning and processing of sequential structure, rather than constraints prescribed by an innate universal grammar. Previous work has shown that some linguistic universals can be learned by sequential-learning devices with no language-specific biases (e.g., Ellefson & Christiansen, 2000; Christiansen & Devlin, 1997; Van Everbroeck, 1999). Here, we examine the ways in which case markings and word order may function as cues for a sequential learning device acquiring syntactic structures. If language evolution occurred through linguistic adaptation, we should expect that typologically rare languages should be more difficult to learn by a sequential-learning device than the more common ones.

We hypothesize that languages with free word-order (FWO) are not significantly harder or take longer to learn than languages with strict word-order (SWO). What is important is that there exist some set of cues to indicate syntactic relationships--there is nothing inherently special about word order or case markings. This hypothesis contrasts with the view espoused by proponents of an innate Language Acquisition Device. According to the Subset Principle of generative linguistics, children default to strict word-order when acquiring a language (Pinker, 1995) and therefore FWO languages are predicted to be more difficult to learn. The alternative hypothesis that we are advocating argues that children need not have innate linguistic predispositions to aid them in acquiring this aspect of language.

We trained simple recurrent networks on four artificial grammars, reflecting the possible combinations of FWO or SWO, and the presence or absence of case markings. A two-way ANOVA revealed a highly significant interaction (p<.001) between the presence of case markings and flexibility of word order. FWO grammars _without_ case markings were significantly more difficult to learn (p's<.001) than grammars involving either SWO or FWO with case markings. There was no difference between the latter two (p>.9). The difficulty associated with learning a FWO language without case markings is underscored by typological evidence, suggesting that FWO languages predominately use case markings to signal grammatical relationships (Payne 1992). Our results further show that typologically rare languages such as SOV without cases (Greenberg, 1963) are more difficult to learn by the networks than more common SWO languages such as SVO or VSO. These simulation results support our hypothesis that the precise nature of syntactic cues (word order versus cases) may be unimportant.

We are currently working on experiments in which human subjects are trained on artificial languages similar to the ones used in the simulations. If successful, these results will further argue for the adaptation of language to non-linguistic constraints of a sequential learning device.

Home \| People \| Research \| Links \| Contact \| Publications \| Presentations

Please email suggestions/errors to mhc27@cornell.edu