SENTENTIAL AND ACOUSTIC FACTORS IN THE RECOGNITION OF OPEN-AND CLOSED-CLASS WORDS

Closed-class words are highly frequent yet relatively difficult to per ceive; although this ought to impair communication, we communicate eas ily under normal conditions. Modular and interactive architectures off er differing explanations of this paradox, with different assumptions about how the acoustic and grammatical properties of those words are c ombined. The interaction of these factors was investigated by having s ubjects listen for and repeat open-and closed-class homophones (spoken by a male) that were spliced into three female-voice sentences: (a) t he same sentence, (b) a neutral sentence, and (c) the ''swapped'' sent ence (e.g., open-target in a closed-class context). Results show that: (a) under neutral conditions, it is harder to identify closed- than o pen-class tokens; but (b) they differ little in their original context s; (c) open-class tokens are very easy to identify in a closed-class c ontext; (d) recognizability of closed-class tokens in the swapped cont ext was generally poor; and (e) these interactions are influenced by s entence prosody but not by target length. It is argued that these resu lts indicate a relatively early interaction between perceptual and con textual processing. (C) 1997 Academic Press.