Volume 79, Issue 4

This study investigated how predictability and prosodic phrasing interact in accounting for the variability of syllable duration in Taiwan Southern Min. Speech data were extracted from 8 hours of spontaneous speech. Three predictability measurements were examined: bigram surprisal, bigram informativity, and lexical frequency. Results showed that higher informativity and surprisal led to longer syllables. As for the interaction with prosodic positions, there was a general weakening of predictability effects for syllables closer to the boundary, especially in the pre-boundary position, where pre-boundary lengthening was the strongest. However, the effect of word informativity appeared to be least modulated by this effect of boundary marking. These findings are consistent with a hypothesis that prosodic structure modulates the predictability effects on phonetic variability. The robustness of informativity in predicting syllable duration also suggests a possibility of stored phonetic variants associated with a word's usual contextual predictability.

Glottalized sonorants are a rare sound type that has been under scrutiny for a number of reasons of general relevance to the phonetic theory. It has been claimed that the timing of glottalization of glottalized sonorants may shift in accordance with the position in the syllable onset (pre-glottalization) or coda (post-glottalization), to provide a cue for its place of articulation; other studies argued against this claim. The paper investigates acoustic properties of the glottalized lateral in Rikvani Andi, a one-village dialect of Andi (East Caucasian). Based on the data from elicitations and free narratives, we consider the acoustic correlates that have been argued in the literature to differentiate glottalized sonorants from their modal counterparts, including aperiodicity, intensity, duration and spectral tilt. In Rikvani Andi, all of the correlates prove to be statistically significant in recordings of isolated words, but the differences tend to decrease in free narratives. The timing of glottalization does not support the existing generalizations – while the glottalized lateral only occurs in Rikvani Andi in the syllable onset, it tends to be mid- to post-glottalized. We discuss two possible explanations of why the Rikvani Andi glottalized sonorant fails to comply with typological expectations.

The paper investigated possible perceptual insensitivity effects in the perception of lexical pitch accents by native and non-native listeners, that is, by Serbian and English listeners, respectively. The objective of the study was to explore which word-prosodic categories listeners used when they were required to contrast and recall sequences of lexical pitch accents. To that effect, Serbian and English listeners performed a Sequence Recall Task (SRT) in which they contrasted pairs of non-words with different Serbian lexical pitch accent types, and recalled the sequences of these non-words under different memory load conditions. Listeners’ answers were coded correct and incorrect and the accuracy scores between the groups were compared and analyzed. Both groups had almost identical levels of accuracy and they performed well above chance level on each contrast. Neither group exhibited any effects of perceptual insensitivity to lexical pitch accents. English (non-native) listeners did not differ in their performance from native Serbian listeners, which, contrary to what previous research suggested, implied that one’s native language word-prosodic category inventory did not preclude the encoding of non-native word-prosodic categories. Instead, non-native listeners were capable of deploying different prosodic resources such as post-lexical pitch accents to recall the sequences.