Volume 70, Issue 1-2

Multiple-level tone contrasts are typologically disfavoured because they violate the dispersion principles of maximizing perceptual distance and minimizing articulatory effort. This study investigates the tonal dispersion of a multiple-level tone system by exploring the cues used in producing and perceiving the five level tones of Black Miao. Both production and perception experiments show that non-modal phonations are very important cues for these tonal contrasts. Non-modal phonations significantly contribute to the dispersion of the five level tones in two ways: either by enhancing pitch contrasts or by providing an additional contrastive cue. Benefiting from more than one cue, the level tones T11, T33 and T55 are well distinguished in the tonal space; by contrast, the level tones T22 and T44, only contrasting in pitch, are the most confusable tones. The tonal registers model proposed in this article sheds light on the different uses of non-modal phonations across languages.

This is a proposal in favour of proceeding from communicative function to linguistic form, rather than the reverse, for an insightful account of how humans communicate by speech in languages. A functional framework is developed that encompasses argumentation structures, declarative and interrogative functions, and expressive intensification. Such a function orientation can become a powerful tool in comparative prosodic research across the world's languages. The potential of this approach is shown by comparing the prosodic form of Mandarin Chinese data collected in functionally contextualized scenarios with corresponding data from English and German.

In the light of previous acoustic analyses of Danish stød and Danish intonation, we discuss two different phonological theories. In one, stød is an autonomous laryngeal syllable prosody. In the other, stød is the phonetic manifestation of an HL tonal pattern compressed within one syllable. The tonal representation is found to be contradicted by the phonetic reality, and it cannot account for the structurally determined alternation between non-stød and stød in inflection and derivation, nor for latent stød or stød in compounds. Furthermore, stød patterns are largely constant across regional varieties of Danish, but tonal patterns over the relevant structural domains are highly variable. Thus, stød may occur on any kind of tonal configuration, anywhere in the speaker's pitch range, a variability which is hard to reconcile with a fixed HL representation.

A series of vowel-identification experiments using gated consonant stimuli shows that English listeners are capable of recovering the vocalic context in which a consonant appears from information contained in the consonant alone. This is true for most consonants tested, including liquids, nasals, and stops in onset and coda position. Positional asymmetries in vowel sensitivity go in opposite directions for liquids (coda sensitivity > onset) and stops (onset > coda). Nasals pattern with liquids in terms of vowel sensitivity from consonant steady states alone, but pat- tern more closely with stops when portions outside the steady state are taken into account. It is argued that these asymmetries are related to patterns of cluster-driven vowel compression (also called ‘compensatory shortening') in speech production.

The nonsibilant English fricatives /f/ and /θ / are known to be acoustically nonrobust. Using /f/ and /θ/ stimuli produced in CV, VCV, and VC syllables in /i α u/ contexts spoken by 10 talkers (5 male), we first replicate previous research suggesting that the most robust cues to this contrast are in the formant transitions in adjacent vowels. We also demonstrate vowel and syllable contextual differences that point to the contrast being most robust in /u/ contexts. In a series of perception experiments we go on to demonstrate effects of bias on perception of /f/ and /θ/ that derive from the uninformative nature of the frication noise, making them vulnerable to misperception in general, and especially in low-saliency contexts where the formant transition information is less robust. In experiment 1, listeners' classification of /f/ and /θ/ demonstrated a general bias to respond /f/ for fricatives produced by females and /θ/ for those produced by males. We hypothesize that the perceived concentrations of spectral energy in the fricative are shifted based on the concentration of energy in the vowel, which depend on a talker's gender. In experiment 2, vowel and frication noise portions were cross-spliced to probe this effect, resulting in the same gender-based bias. In a final experiment the vocalic information was removed and only the frication noise was presented to listeners for classification. In this task there was a general bias for /f/, regardless of the talker gender. Overall we demonstrate topdown gender effects in perception that originate in the strong indexical properties of adjacent vowels rather than being present in the frication noise itself.