Volume 68, Issue 3

Although much has been written on the relative importance of acoustic correlates of linguistic stress for the listener, the role of spectral expansion/reduction has been much understudied. The present article is the first to study the role of spectral expansion/reduction in a two- parameter study together with temporal structure exploiting systematic variation of both parameters in a 7 × 7 stimulus space. We used a single minimal stress pair in Dutch, a language in which all classic acoustic correlates of stress were shown earlier to be relevant in singleparameter studies, i.e. pitch movement, intensity (loudness), temporal organization and spectral expansion/reduction. The results of our study reconfirmed that temporal organization is a strong cue to stress perception when target words are presented out of focus (i.e. without a pitch accent on the target). Spectral expansion/reduction was a very weak stress cue; its effect was noticeable only when temporal structure was ambiguous between initial and final stress. These results suggest that spectral expansion/reduction is indeed the weakest of the four cues traditionally identified in the literature, at least in stress-accent languages such as English and Dutch.

While assimilation was initially regarded as a categorical replacement of phonemes or phonological features, subsequent detailed phonetic analyses showed that assimilation actually generates a wide spectrum of intermediate forms in terms of speech timing and spectrum. However, the focus of these analyses predominantly remained on the assimilated speech sound. In the present study we go one step ahead in two ways. First, we look at acoustic phonetic detail that differs in the French vowels /i, a, u/ preceding single /s/ and /∫/ sibilants as well as /s#∫/ and /∫#s/ sibilant sequences. Second, our vowel measurements include not only F1 and F2 frequencies, but also traditional prosodic parameters like duration, intensity and voice quality. The vowels and sibilants were recorded as the central part of CVC#CVC pseudo-names in a contextualized read-speech paradigm. In the single-sibilant conditions we found that the vowels preceding /∫/ were longer, breathier, less intense, and had more cardinal F2 values than before /s/. For the /s#∫/ and /∫#s/ conditions we found regressive and progressive /s/-to-[∫] assimilation that was complete in terms of spectral centre-of-gravity measurements, although French is said to have only voice assimilation. Moreover, the vowels preceding the /s#∫/ sequences still bear an imprint of /s/ despite the assimilation towards [ ∫∫]. We discuss the implications of these findings for the time window and the completeness of assimilation as well as for the basic units in speech communication.

We present in this article an acoustic study on tone sandhi and tonal coarticulation in Tianjin Chinese. Our results indicate that Tianjin tone sandhi is likely influenced by Standard Chinese and is undergoing a number of changes, causing variations and exceptions to the sandhi patterns, and the majority of the sandhis are non-neutralizing, contra traditional descriptions. Tonal coarticulation in Tianjin exhibits a number of well-known cross-linguistic properties: progressive assimilation, regressive dissimilation, a greater progressive effect, and a number of High/ Low asymmetries. Despite the dissimilatory properties observable from both tone sandhi and tonal coarticulation, they seem to have different characteristics, indicating different sources for the two processes.