Poster session II
Paul D. Allopenna,1 Athanassios Protopapas,2
Steven A. Finney,2 and Peter D. Eimas2
1University of Rochester and 2Brown
University
In order to comprehend spoken utterances, listeners must be capable of reliably segmenting the incoming speech stream in ways that facilitate lexical access. An important general approach to the segmentation problem posits that listeners compare incoming speech to stored representations that are comprised of sublexical units such as phonemes or syllables.
Previous research has found evidence for syllable-sized units of representation in both French (Mehler et al. 1981) and Spanish (Bradley et al. 1993). By contrast, no syllabic effects have been found in monitoring for targets in either British English (Cutler et al. 1986) or Australian English (Bradley et al. 1993). The motivation for the studies reported is two-fold. First, while previous empirical work failed to provide evidence for syllabic processing in English, the results of the various experiments themselves do not provide clear evidence for an alternative type of processsing. Second, syllabic segmentation is compatible with at least one of the major current theories of English segmentation, namely the Metrical Segmentation Strategy (Cutler 1990). Given the already available evidence for syllabic processing in French and Spanish, we argue that syllabic representations provide the segmentation device with a number of useful phonotactic constraints. Additionally, syllabic representations can help to specify exactly how word-initial onsets are segmented in continuous speech.
In the current experiments, all of which test for syllabic effects in American English, we used three basic paradigms to further probe the role of syllables in on-line segmentation. The first series of experiments are simple variations of the standard monitoring experiments, such as those cited above. In these experiments, subjects were asked to monitor for syllabic targets that were embedded in lists of words. The results of these experiments provide evidence for syllabic segmentation in American English under test conditions in which the stress of the carrier word was systematically manipulated. A second set of studies investigated syllabic effects in an induced attention task (e.g., Pallier et al, 1994). Again, evidence for syllabic representations was found, although these effects were largely confined to words with second-syllable stress. Most importantly for research in sentence processing, a monitoring experiment was conducted where syllabic targets were embedded in first-syllable stress words that themselves were embedded in running speech. The results of this study are compatible with the results of our first several experiments. An extension of this study, using second-syllabele stress carrier words, is presently being conducted and will be reported within he general framework outlined above. Taken as a whole, the general conclusion pointed to by the current research is that syllabic representations should play an important role in theories of segmentation of English.