Speaking rate, voice-onset time, and quantity: The search for higher-order invariants for two Icelandic speech cues

Abstract

The temporal structure of speech has been shown to be highly variable. Speaking rate, stress, and other factors influence the duration of individual speech sounds. The highly elastic nature of speech would seem to pose a problem for the listener, especially with respect to the perception of temporal speech cues such as voice-onset time (VOT) and quantity: How does the listener disentangle those temporal changes whicqh are linguistically significant from those which are extrinsic to the linguistic message? This paper reports data on the behavior of two Icelandic speech cues at different speaking rates. The results show that manipulations of rate have the effect of slightly blurring the distinction between unaspirated and aspirated stops. Despite great changes in the absolute durations of vowels and consonants, the two categories of syllables-V:C and VC:-are nonetheless kept totally distinct. In two perceptual experiments, it is shown that while the ratio of vowel to rhyme duration is the primary cue to quantity and remains invariant at different rates, no such ratio can be defmed for VOT. These results imply that quantity is the only one of these two speech cues that is selfnormalizing for rate. Models of rate-dependent speech processing need to address this difference. 291Many speech cues are primarily temporal in nature, being defined either by the temporal relationships between acoustic events or by the duration of individual segments (Lehiste, 1970;Lisker, 1974). A noticeable feature of normal speech is the highly variable utterance rate. The fact that listeners do not appear to be overly bothered by such speaking-rate variations poses a challenge for perceptual theory, not least concerning the perception of temporal speech cues. How is the listener able to disentangle those temporal aspects which properly are extrinsic to the phonemic message (i.e., changes of tempo) from those which define the relevant phonemic contrasts, given that both are carried simultaneously in the speech stream?Numerous studies have shown convincingly that ratedependent speech processing does in fact occur and is found to affect numerous speech cues. Miller and Liberman (1979) found that with a longer overall duration of the syllable, listeners required longer formant transition durations to hear /wa/ than to hear /ba/. Summerfield (1981) found that the longer the following vowel, the longer the voice-onset times (VOTs) that were needed

Timing relations in speech and the identification of voice-onset times: A stable perceptual boundary for voicing categories across speaking rates

“…The best-fit regression lines for the aggregate data conform to previous observations showing substantial decreases in VOT for voiceless stops at fast rates of speech. That is, for /t/ VOT, Lane, Wozniak, Matthies, Svirsky, and Perkell (1995, note 1) calculated a regression slope of .14 from the data of Port and Rotunno (1979) and a slope of .17 from the data of Pind (1995) for Icelandic. These estimates approach the .15 slope observed in the present data (on the method used to calculate slopes and intercepts, see Kleinbaum, Kupper, & Muller, 1988).…”

“…For example, some authors see that an intrinsic timing view posits that VOT and syllable duration change with speaking rate "in such a way that this functional relation, or higher-order property, remains invariant" (Wayland et al, 1994(Wayland et al, , p. 2699. Because of the failure to find temporal constants in speech, relativetiming hypotheses have been largely dismissed (see the comments of Pind, 1995). There is substantial evidence that timing relations involving such aspects as VOT or formant transitions are not constant (e.g., Kessinger & Blumstein, 1998;Port, 1981;Volaitis & Miller, 1992).…”

“…The extrinsic account suggests that listeners adjust a criterion value used to make the voicing decision according to information bearing on speaking rate (e.g., Diehl & Walsh, 1989;Dupoux & Green, 1997;Pind, 1995;Pisoni, Carrell, & Gans, 1983;Wayland, Miller, & Volaitis, 1994). Because the VOTs for voiceless stops tend to shorten at fast rates of speech, listeners adjust the perceptualboundary toward shorter values.…”

Rate effects on Southern American English VOT

“…For fortis stops in this study, there was a slope of .05ms VOT/ms word length. However, in the study by Magloire and Green (1999), English had a slope of .27ms VOT/ms word length, Pind (1995) reported a slope of .15ms VOT/ms word length for Icelandic aspirated stops, and Beckman et al (2011) reported a slope of .10 ms VOT/ms word length for aspirated Swedish stops. There was a slope of -.13ms VOT/ms word length for lenis stops in this study, compared to a slope of -.45ms VOT/ms for Spanish prevoiced stops (Magloire & Green 1999) and -.27ms VOT/ms for Swedish prevoiced stops (Beckman et al 2011).…”

“…Rate Effects. Previous research into the effect of speaking rate on VOT durations (e.g., Miller, Green, & Reeves 1986, Pind 1995, Allen & Miller 1999 show that in word-initial stop contrasts, VOT increases in the phonologically specified category as speaking rate decreases, while the unspecified category exhibits little change. Thus, in languages which contrast prevoiced stops with short-lag stops, such as Spanish (Table 3), prevoicing increases in slower speech and decreases in faster speech while short-lag VOT shows little difference.…”