Volume 70, Issue 4

This study analyses the meaning spaces of German pitch contours using two modes of melodic movement: continuous or in steps of sustained pitch. Both the continuous and stepped movements are represented by a set of five basic patterns, the latter being derived from the former. Thirty-six German native speakers judged the pattern sets on a 12-scale semantic differential. The semantic profiles confirm that stepped contours can be conceived of as stylized intonation, in a formal as well as in a functional sense. On the one hand, continuous (non-stylized) and stepped (stylized) contours are assigned different overall meanings (especially on the scales astonished - commonplace and interested - not interested). On the other hand, listeners organize the two contour sets in a similar fashion, which speaks in favour of parallel pattern inventories of continuous and stepped movement, respectively. However, the meaning space of the stylized patterns is affected by formal restrictions, for instance in the step transformation of continuous rises.

Apart from Tamil, Malayalam is the only Dravidian language that still retains a fifth liquid in its inventory (denoted as /ʐ/), a sound that can be traced back to Proto-Dravidian. The phonetic and phonological status of this sound has been the subject of considerable debate, both in terms of its rhotic versus lateral status and its phonetic realisation. Studies to date have been impressionistic, so this is the first acoustic study of all five liquids in Malayalam. Production data from 8 male speakers reveal that while /ʐ/ is realised as a central post-alveolar approximant, its ambiguous phonotactic patterning and mixed acoustic profile are contributors to its mixed identity. An understanding of the complex patterning of this sound, which straddles the rhotic/lateral divide, requires an appreciation of the role of secondary acoustic resonance in distinguishing members of a crowded liquid system and in preventing a /ʐ/-/ɭ/ merger.

Elect ropalatographic and acoustic data for the fricative clusters /sʃ/ and /ʃs/ and the single fricatives /s/ and /ʃ/ in Eastern Catalan reveal that, while /sʃ/ undergoes regressive assimilation, /ʃs/ blends into an intermediate fricative or exhibits a twotarget realization. On the other hand, acoustic data for /ʃs/ in Western Catalan show that /s/ may assimilate to preceding /ʃ/ or undergo blending or a two-target realization depending on speaker and prosodic condition (no data are available for /sʃ/ in this dialect). Duration values for the two fricative clusters suggest that the assimilated outcome of /sʃ/ shortens down to the length of /ʃ/. A production-based account of the two sequences /sʃ/ and /ʃs/ in Catalan and other languages is proposed.

Familiarity with a talker or accent can facilitate speech perception, but little is known about the acoustic-phonetic knowledge that is acquired during exposure to an unfamiliar talker, particularly with respect to how a talker realises grammatical and prosodic distinctions. The present experiment used relatively natural stimuli and tasks to investigate listeners' adaptation to a phonetic variant that is systematically associated with a morphological context, specifically the English /ri:/ prefix. Results showed that listeners who had previously been exposed to the atypical variant in /ri:/ prefixes in stories scored higher than control listeners in an intelligibility-in-noise task that included instances of the atypical variant. Perceptual learning was partly specific to the /ri:/ prefix, though there was weak generalisation of learning to word-initial /ri:/ syllables that were not prefixes. These results emphasise the value of developing context-sensitive, probabilistic models of speech perception which include multiple, parallel levels of representation.