Learned But Domain-Specific Mechanisms?

Continuing from yesterday's post, to demonstrate that the nativism debate in language is fully distinct from the domain-specificity debate, one needs to demonstrate that a position on one debate does not dictate your position on the other. Yesterday's post examined some theories that might be considered to advocate innate but domain-general capacities that enable language, a perspective not frequently represented in the literature. Below I discuss the flip-side: are there theorists who advocate a learned but domain-general capacity for language?

Thus, the second claim that needs to be substantiated is that domain-specific mechanisms could theoretically be empirically guided. It is tempting to think that empiricist perspectives do not have this “theoretical flexibility,” given that so many prominent empiricists argue for domain-general mechanisms. For example, “dumb attentional mechanisms” are sometimes thought sufficient to account for word learning biases (Booth & Waxman, 2002). Likewise, Bayesian models of syntax learning work partly because they choose “narrow hypotheses over broad ones – but this is not a language-specific constraint” (Regier & Gahl, 2004). Perhaps most tempting is the strong claim to domain-generality made by the connectionist framework (McClelland & Patterson, 2002), a firmly empiricist tradition.

However, empiricist accounts of word learning can remain theoretically coherent and still posit apparently domain-specific mechanisms. For example, the end result of learning in connectionist networks can approximate modularity (Colunga & Smith, 2005), due to the tendency for certain parts of the network to become functionally specialized.

Research on syntactic development illustrates another example of how empiricist theories can be compatible with domain-specificity. Children’s language learning appears to occur by way of imitation, such that words are not initially used based on their part of speech (contrary to what one might expect if they were being placed into a “universal grammar”) but rather based on the contexts in which they are most frequently experienced (Tomasello, 2000). At some critical threshold of linguistic experience, children begin to generalize the constructions of language (on the basis of analogy, according to Tomasello). It is at this point that language becomes a kind of representational modality unto itself, where new linguistic forms are constructed by language-specific rules. Thus, the end result of learning here also approximates domain-specificity.

A third example of domain-specific yet learned mechanisms comes from research on linguistic relativism. For example, the use of grammatical gender in various languages shows that this learned aspect of language can bias adjective generation, “potency” judgments, and a variety of other linguistic tasks by speakers of those languages (Boroditsky, 2003). Although some have interpreted this as evidence that learned aspects of language can have DG effects (i.e., on “thinking-in-general” above and beyond “thinking-for-speaking”), it is difficult to clearly demonstrate this is the case, since language seems to be such a pervasive aspect of cognition.

For example, Boroditsky (2003) reviews evidence from cross-linguistic comparisons of similarity ratings – with an auditory shadowing dual task intended to “disable people’s linguistic faculties.” Although the auditory shadowing task did not interfere with the influence of grammatical gender on similarity ratings, one cannot conclude that this learned aspect of language affects thought more generally, both because similarity ratings are still an arguably linguistic task and because the auditory shadowing task might not have fully occupied the subjects’ linguistic faculties. Likewise, evidence on cross-linguistic differences in time perception clearly demonstrates differences between cultures, but since no articulatory suppression techniques were used, subjects may have covertly recruited language to assist in the task (Casasanto et al., 2004), making this seem like yet another domain-specific effect.

Tomorrow's post will conclude this series with a review of how these two distinct debates have been confounded in prominent cognitive research on language.

Note: This post is part III of a series on how to disentangle the domain-specificity debate from the nativism debate in cognitive studies of language.
Part I: Disentangling Two Debates: Introduction
Part II: Some domain-general mechanisms may be innate
Part III: Some domain-specific mechanisms need not be innate (this post)
Part IV: Dissociations from data and conclusions (coming tomorrow)

References:

Boroditsky, L., Schmidt, L., & Phillips, W. (2003). Sex, Syntax, and Semantics. To appear in Gentner & Goldin-Meadow (Eds.,) Language in Mind: Advances in the study of Language and Cognition.

Booth AE, & Waxman SR. (2002). Word learning is 'smart': evidence that conceptual information affects preschoolers' extension of novel words. Cognition. 2002 May;84(1):B11-22.

Casasanto, D., Boroditsky, L., Phillips, W., Greene, J., Goswami, S., Bocanegra-Thiel, S., Santiago-Diaz, I., Fotokopolou, O., Pita, R., & Gil, D. (2004) How Deep Are Effects of language on Thought? Time Estimation in Speakers of English, Indonesian, Greek and Spanish. Proceedings of the 26th Annual Cognitive Science Society.

Colunga E, & Smith LB. (2005). From the lexicon to expectations about kinds: a role for associative learning. Psychol Rev. 2005 Apr;112(2):347-82.

McClelland, J. L. & Patterson, K. (2002). Rules or Connections in Past-Tense inflections: What does the evidence rule out? Trends in Cognitive Sciences. 6:11 (2002), pp 465-472

Regier T, & Gahl S. (2004). Learning the unlearnable: the role of missing evidence. Cognition. 2004 Sep;93(2):147-55; discussion 157-65.

Tomasello M, Call J, & Hare B. (2003). Chimpanzees understand psychological states - the question is which ones and to what extent.