Abstract
In the experiments reported here, listeners categorized and discriminated speech and non-speech analogue stimuli in which the durations of a vowel and a following consonant or their analogues were varied orthogonally. The listeners' native languages differed in how these durations covary in speakers' productions of such sequences. Because auditorist and autonomous models of speech perception hypothesize that the auditory qualities evoked by both kinds of stimuli determine their initial perceptual evaluation, they both predict that listeners from all the languages will respond similarly to non-speech analogues as they do to speech in both tasks. Because neither direct realist nor interactive models hypothesize such a processing stage, they predict instead that in the way in which vowel and consonant duration covary in the listeners' native languages will determine how they categorize and discriminate the speech stimuli, and that all listeners will categorize and discriminate the non-speech differently from the speech stimuli. Listeners' categorization of the speech stimuli did differ as a function of how these durations covary in their native languages, but all listeners discriminated the speech stimuli in the same way, and they all categorized and discriminated the non-speech stimuli in the same way, too. These similarities could arise from listeners adding the durations of the vowel and consonant intervals (or their analogues) in these tasks with these stimuli; they do so when linguistic experience does not influence them to perceive these durations otherwise. These results support an autonomous rather than interactive model in which listeners either add or apply their linguistic experience at a post-perceptual stage of processing. They do not however support an auditorist over a direct realist model because they provide no evidence that the signal's acoustic properties are transformed during the hypothesized prior perceptual stage.