Morphological boundaries are processing boundaries in spoken production

Goldberg, A.

Dept of Psychology, Tufts University, Medford, USA

One issue that has received relatively little attention in the psycholinguistic literature is the question of how morphologically complex words are processed by the spoken production system. In this talk, I argue that phonological representations must be ‘stitched’ together from their component morphemes and the result is not identical to a monomorphemic representation. Specifically, I argue that phonological processing is more coherent for phonemes within morphemes than phonemes in separate morphemes. One source of evidence comes from the phoneme similarity effect, where speakers are particularly likely to make errors on sequences containing repeated/similar phonemes. This effect has typically been explained in terms of increased competition among phonemes co-activated via shared features. Using mixed-effects regression analyses of oral reading reaction times, I demonstrate that similarity has an inhibitory effect on RT (e.g., kick > sick) and that the similarity effect is weaker for phonemes in separate morphemes (e.g., socks), than for phonemes within the same morpheme (e.g., sex [seks]). This suggests that the spread of activation and the competition that results are stronger among phonemes within the same morpheme than in different morphemes. Another source of evidence comes from an analysis of an aphasic individual’s speech errors. I show that while this individual makes insertion errors that systematically improve the sonority of coda clusters, these ‘repairs’ only occur in multimorphemic environments (walk+ed->[wakit]). This suggests that the phonemes of concatenated morphemes must be bound together and that if this ‘glue’ is weakened (e.g., via brain damage), the grammar may have an opportunity to express itself via phonological repairs. In both cases, processing is more coherent within morphemes than across--morphological boundaries are in effect processing boundaries. Processing theories that can account for this pattern of effects will be discussed.