Volume 39, Issue 2-3

In order to investigate articulatory cues for the perception of bilabial consonants, a production model was set up whose control parameters were converted into acoustic parameters by which a terminal analogue speech synthesizer was driven to synthesize test sounds for the perception experiments. Experiments concerning discrimination between /b/ and /w/ and between /m/ and /w/ were then performed with these test sounds. Statistical testing of the results showed that discrimination among the bilabials in an intervocalic context entirely depended on the minimum value of lip opening and that, contrary to the facts hitherto known, it was independent of the rate of lip opening/closure. At the same time, the features [ + anterior, – coronal] and [ + nasal] turned out to be perceptually independent phonological attributes of bilabial and nasal consonants, respectively.

Many phonetics textbooks state that, in sequences of two stop consonants in English, the first stop is commonly unreleased. For nonhomorganic stop consonant sequences, this statement may be taken to imply that the (necessary) articulatory release of the first stop has no observable acoustic consequences. To examine this claim, we recorded sentences, produced by several native speakers of American English at a conversational rate, containing word-internal sequences of two nonhomorganic stops, either across a syllable boundary (e.g., cactus, pigpen ), or in word-final position (e.g., act, sobbed ). Oscillograms of the critical words revealed that release bursts of the first stop occurred in the majority of tokens, except in those where the second stop was bilabial. The bursts were acoustically rather weak and difficult to detect by ear, which may account for their having been neglected in the literature. Instead of a simple ‘released’-‘unreleased’ distinction, we propose five classification categories which make use of articulatory, acoustic, perceptua and contrastive phonetic criteria.

(Right to Left) and (Left to Right) coarticulation of vowels with stop consonants in VGV logotemes was studied before and during oral anesthetization. Correlational analysis revealed that neither LR nor RL coarticulation was markedly reduced in extent, suggesting that orosensorv feedback was not crucial to control of coarticulation. This lends support to the notion of central control of coarticulation.

Linear prediction analysis-synthesis techniques are commonly implemented in low-bit rate (2,400–9,600 bauds) speech transmission systems. We know that these systems do not work properly if speech is noisy. With the aid of a diagnostic rhyme test in French, we could evaluate the degradation of the intelligibility of synthesized speech. This test was applied to original speech and to linear-prediction synthesized speech, for different signal-to-noise ratios (white noise). The results obtained show clearly that intelligibility is degraded by the linear prediction analysis-svnthesis in a high ambient noise environment.

An experiment was conducted to test two predictions entailed by the hypothesis that Japanese morae have constant durations. The first prediction is that a segment’s duration will conform to its moraic status in an utterance; a consonant will be longer when it is syllabic than when it occurs in a GV mora, and phonemically long consonants will have durations that reflect their moraic structure. The second prediction is that segment lengths will vary to compensate for the intrinsic durations of adjacent segments in order to equalize mora lengths. Neither of these predictions was borne out.

The presupposition that listeners tend to estimate speech of an unfamiliar language as being faster than speech in their mother tongue was tested by means of a perception test consisting of five different languages (Dutch, English, French, Spanish and Moroccan Arabic) and three different speech rates (normal, fast and extra fast). The perception test was offered to two listening panels, differing in level of familiarity with one language (English) and in general listening ability (trained versus untrained listeners). The results indicate that the presupposition has to be rejected, as both panels were able to estimate the speech rate of items of different languages and intended speech rates with great accuracy and agreement. The results also indicate that the listeners’ main cue for tempo detection is not their knowledge of the lexical information of the signal, but rather some temporal feature or features in the speech signal.