Cohort effects and asymmetrical word-level sound change
The cohort model of lexical retrieval (Marslen-Wilson & Welsh 1978; Marslen-Wilson & Tyler 1980; Marslen-Wilson 1984, 1987), a theory of speech perception not traditionally applied to questions of historical linguistics, offers evidence that the point at which spoken words become unique within their lexicon is cognitively significant. This uniqueness point divides a
word into two regions which appear to be processed by different neural machinery. Historical linguistic research suggests that material which maintains meaningful contrasts is more resistant to reduction than material which does not (Blevins 2005; Blevins & Wedel 2009; Wedel, Jackson, & Kaplan 2013). Perhaps, then, the sublexical regions outlined by the cohort model, which differ in their ability to contrast words, are similarly affected by sound change unevenly distributed across the word. To examine the extent to which this difference in processing could relate to a difference in language change, this study uses the concept of the cohort model to bifurcate genetically-related words selected from eight Indo-European languages which feature uniqueness points within their respective lexicons. Through the comparison of phonological distance among these forms, operationalized as the Levenshtein distance (Levenshtein 1965) between phonemic representations, this analysis finds evidence for an asymmetrical distribution of sound change among cognates: within a word, sound change is more likely to occur after the point at which the word is distinct from all other words in the lexicon.