Restrictions on phonemes in affixes: A crosslinguistic test of a popular hypothesis

The hypothesis that inflectional affixes use a restricted set of phonemes and that these are the less marked phonemes of the language is discussed and tested on the verbal affixes in a sample of twenty-three maximally unrelated languages. The results show that the tendency for languages to use only a smaller subset of their phonemes in verbal inflection than would be predicted by chance is only a weak trend and not by any means a universal of language. In addition, the tendency to use less marked or less complex segments in affixes is also only a trend and not a universal. However, some generalizations can be made about languages that have patterned exclusions from affixes. It is argued that no one explanation covers all the facts and that multiple diachronic trends, such as phonological reduction in grammaticization and the re-use of old affixes in creating new grammaticized affixes, produce the weak tendency evidenced by the data.