Abstract
There is substantial variation in language experience, yet there is surprising similarity in the language structure acquired. Constraints on language structure may be external modulators that result in this canalization of language structure, or else they may derive from the broader, communicative environment in which language is acquired. In this paper, the latter perspective is tested for its adequacy in explaining robustness of language learning to environmental variation. A computational model of word learning from cross‐situational, multimodal information was constructed and tested. Key to the model's robustness was the presence of multiple, individually unreliable information sources to support learning. This “degeneracy” in the language system has a detrimental effect on learning, compared to a noise‐free environment, but has a critically important effect on acquisition of a canalized system that is resistant to environmental noise in communication.
Highlights
A key question in the cognitive sciences is how, despite the enormous variation in linguistic experience, the language learner acquires broadly the same language structure, “within a fairly narrow range” (Chomsky, 2005). This perspective has led to proposals for mechanisms that ensure canalization of language structure
Canalization was once considered as a consequence of the natural selection of mechanisms that operate to minimize phenotypic variation (Waddington, 1942)
Wagner (1996) demonstrated that selecting for canalizing regulators required a rate of mutation that is higher than that observed in biological evolution, inconsistent with Gallistel’s (2000) suggestion of modular, domain-specific learning constraints within the individual
Summary
A key question in the cognitive sciences is how, despite the enormous variation in linguistic experience, the language learner acquires broadly the same language structure, “within a fairly narrow range” (Chomsky, 2005). Consistent with views in biological evolution, is that the key function of multiple cues for language learning is their interactivity, resulting in a system stable to variation in the environment This property of language is its “degeneracy,” defined as “the ability of elements that are structurally different to perform the same function or yield the same output” (Edelman & Gally, 2001). From a single learning trial, there may be several potential referents within the child’s environment given an utterance, over multiple situations if, whenever a particular word is heard in speech, the target referent in the environment is present, the learner can increase his or her association between the target word and target object (McMurray, Horst, & Samuelson, 2012) Such cross-situational learning (Yu & Smith, 2012) can further be supplemented by information that the speaker uses to indicate the field of reference. A model trained with a degenerate environment should be able to effectively map between words and referents even when environmental cues that support this mapping are no longer available
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
