Modularity in the Channel : A Response to Moreton ( 2008 ) 1

Moreton (2008) argues 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, which constrains the learner but does not influence perception and production. Channel bias is defined as „phonetically systematic errors in transmission between speaker and hearer‟. Moreton proposes a new type of analytic bias, the modularity bias, whereby dependencies between consonant dimensions, such as voicing features of different consonants within the same word, and dependencies between vowel dimensions, such as height features, are easier to learn than dependencies involving a vowel dimension and a consonant dimension, e.g., vowel height and consonant voicing. Moreton argues that the modularity bias does not come from either perception or articulation because the voicing of a consonant influences the height of the preceding vowel as much as the identity of the following vowel does, and the acoustic consequences of the voicing of a consonant on the voicing of a non-adjacent consonant in the same word are relatively minor. In this paper, I employ the Garner interference paradigm (Garner, 1974) to show that the modularity bias does in fact arise in perception and thus can be seen as a type of channel bias, arising from how acoustic cues are parsed into cognitive representations (Blevins, 2004: 151-153).