Home HomeABSTRACT
Two production experiments investigating possible factors influencing the domain of final lengthening are described. Results indicate that final lengthening is generally confined to the final syllable, except when its rhyme contains only a schwa, in which case the penultimate rhyme is lengthened as well. Apparently, only the weight of the final syllable influences the size of the domain which is lengthened. Next, a perceptual acceptability experiment was run. Results indicate that while listeners are sensitive to differences in the amount of final lengthening, they are not very sensitive to the way this is distributed over the preboundary segments. Apparently, the specific distribution of final lengthening in production has no com-municative function, but is the result of the human speech mechanism, together with restrictions on the expandability of segments.
A0559.pdf
TOPABSTRACT
It is well known now that speech chain is not constituted by discrete units. Speech sounds have an influence on other sounds directly in contact with them. We hypothesize that this influence is not noise but plays an important role for perception. An experiment is managed to evaluate the relative importance of two kinds of cues: those of phonetic distinctive features (voiced and unvoiced) and those of voicing assimilation (for liquids). Our results confirm that voicing assimilation of liquids plays an important role to identifie clusters: a/ the absence of assimilation cues increases reaction times; b/ subjects use assimilation cues in preference to distinctive features.
A0880.pdf
TOPABSTRACT
The aim of this study is to determine whether the phonological process of degemination, in which one of two adjacent and identical consonants is deleted, is perceptually complete when it applies over word boundaries. Measurements on the duration of the boundary consonant have shown durational differences between two-word phrases with underlying single and double consonants, even at fast speech rates. Results of a pseudo-gating experiment using a binary forced choice task show that correct segmentation of two-word phrases with underlying single or double consonants, spoken at a fast speech rate, does not exceed chance level. We conclude, therefore, that degemination actually occurs in Dutch and that this process is perceptually complete. Implications for word recognition will also be discussed.
A0955.pdf
TOPABSTRACT
This study explores the hypothesis that relatively invariant properties characterizing lexical items include non contrastive phonetic details such as the amount of linguopalatal contact, or aspects of inter-gestural timing. We show that, in French, a sequence of consonants resulting from the loss of schwa maintains some of the fine articulatory characteristics of the lexical form containing schwa. Such characteristics distinguish this sequence of consonants from an underlying cluster. Thus, we show that "d'r™le" 'some role', with the apostrophe indicating schwa loss, remains articulatorily distinct from "dr™le" 'funny'. A perception experiment shows that the two types of sequences (CC and C'C) are only marginally discriminable by French listeners. However, when the subjects identify correctly the two types of sequences, the distinct characteristics identified in production correlate with the listeners' judgments.
A0992.pdf
TOPABSTRACT
Catalan has been studied by several authors, who gave phonologic as well as phonetic descriptions of the language. In most studies, authors have nevertheless focussed on categories of sounds, considered each in turn, rather than on possible associations of sounds. Implementations of this knowing may therefore raise problems, since acoustical patterns of dvnamical evolution are not well known. The paper addresses the particular case of the []-[lj] and [n]-[nj] acoustical differences. The data show that the differences to be found are mainly related to reorganisations of the time function.
A1104.pdf
TOPABSTRACT
Speech timing at different speaking rates was studied for the Slovenian language and the results were applied in the two level duration prediction model in the Slovenian text-to-speech system S5 [1]. In order to provide the synthesiser with the possibility to pronounce input text with several speaking rates, tests were made to study the impact of speaking rate on syllable duration and duration of individual phonemes and phoneme groups for the Slovenian language.
A1350.pdf
TOP
Recordings