5th International Conference on Spoken Language Processing
Sequential models of word perception assigned a very special role to word onsets. This accounted in a natural way for evidence that lexical access is easier from word beginnings than from word endings, a property speech perception shares with reading. Sequential models badly failed on other scores, however. More recent competition models seem to give equal weight to stimulus information, independent of position within the word. The present word recognition experiment aimed at testing the hypothesis that, other things being equal, mismatches are more damaging to word perception at onsets than at offsets of embedded words, both in speech perception and in reading. Results show that word recognition is quite good in all conditions, even when word onsets are mutilated, and mis-timed, thus lending support to competition models. Yet, the results also show that lexical access is modulated by some early-to-late or left-to-right component, as if human word perception displays a mixture of sequential and competition processing.
Bibliographic reference. Nooteboom, Sieb G. / Dijk, Meinou van (1998): "Heads and tails in word perception: evidence for `early-to-late' processing in listening and reading", In ICSLP-1998, paper 0117.