On the Interpretation of Consonant-Vowel Co-occurrence Frequency Biases

Certain recurrent consonant-vowel co-occurrence frequency biases have often been attributed to the biomechanics of the mandible. This paper takes issue with this claim by pointing to some methodological flaws in the literature supporting it. CV co-occurrence data on Spanish and Portuguese are brought to bear on such neglected issues as sampling, effect size, and count type. It is shown that small samples are unrepresentative, effect size is generally low, and type and token counts lead to different results. These drawbacks notwithstanding, the paper supports the need for further research into the biomechanical bases of such biases. Two linguistic contexts have been uncovered where they prove to be statistically robust in both languages: initial unstressed position; and the lexical subset of words with repeated CV pairs. Both contexts have been related to phonetic complexity in the literature. Thus, bias motivation may lie not simply in biomechanics, but, rather, in its interaction with language specific, linguistic proper constraints.