Learning, bottlenecks and the evolution of recursive syntax 论文

摘要

this paper. The language learning device clearly does impose constraints directly in a similar fashion --- there are certain types of language that the learner simply cannot acquire --- however these constraints are far less severe than those imposed by the LAD. As can be seen in the initial stages of the simulation, very un-language like systems can be acquired by this learner. The constraints on variation are not built into the learner, but are instead emergent properties of the social dynamics of learned communication systems and the structure of the semantic space that the individuals wish to express. The theory presented here gives us a neat explanation of why human languages use syntactic structure to compositionally derive semantics, have recursive subordination to express infinite distinctions in a digital way, have words with major syntactic categories such as noun and verb, and use structural relations (such as word order) to encode meaning distinctions. However, it does not seem to allow us to understand more specific universals. For example, why particular constituent orders are far more frequent than others across the languages of the world (Hawkins 1983; Dryer 1992). Perhaps the best explanation for these types of universal should look at the effect of parsing and generation in the transmission of replicators (see Kirby 1998a; Kirby 1997 for details). On the other hand, at least some of these word order constraints may eventually be explained in terms of linguistic adaptation without appealing to processing (see, Christiansen 1994; Christiansen & Devlin 1997 for some suggestions along these lines). X-bar theory --- a sub part of UG which constrains the structure of syntactic trees cross categorially (Jackendoff 1977) --- has been implicated in various word...