Modularity in the Channel: The Link between Separability of Features and Learnability of Dependencies between Them

Moreton [10] argued for a distinction between analytic bias and channel bias in language learning. Analytic bias is defined as a set of cognitive predispositions for certain types of generalizations that constrains the learner but does not influence perception and production. Channel bias is defined as ‘phonetically systematic errors in transmission between speaker and hearer’. Recent studies [6, 10] document "modularity bias", whereby dependencies between consonant features and dependencies between vowel features are easier to learn than dependencies involving a vowel feature and a consonant feature. Moreton [10, 11] claims that modularity bias is an analytic bias. In this paper, I employ the Garner interference paradigm [4] to show that the modularity bias can be caused by how acoustic cues are parsed into cognitive representations in perception [2] p.151-153, and therefore should be regarded as a channel bias.