Leigh Lisker

Some effects of context on voice onset time in English stops.

Recent work has led us to the conclusion that the English stop categories /bdg/ and /ptk/ are distinguished by the timing of changes in glottal aperture relative to supra-glottal ariculation. In word-initial position, the environment of current interest to us, this is manifested acoustically by voice onset time, that is, the time interval between the burst that marks release of the stop closure and the onset of quasi-periodicity which reflects laryngeal vibration. For citation forms of words this measure of voice onset time completely separates the two phonemic categories. In running speech, however, the separation is less sharp ; there is some overlap along the dimension of voice onset time. We have examined running speech in some detail to discover the extent to which certain contextual features are responsible for this overlap. It is clear that the presence of a voiceless stop in a stressed syllable makes for a greater lag in the onset of voicing. In unstressed syllables, an environment of high contextual redundancy as well as low functional yield for the phonemic contrast, there is considerable reduction of the distinction along the dimension. A further increase in voicing lag is noted in syllables bearing the final sentence stress. In addition, the farther such a syllable is from the end of the sentence, the less the effect is likely to be. The importance of voice onset time continues to be apparent, even in running speech, although other effects of context remain to be explored.

“Voicing” in English: A Catalogue of Acoustic Features Signaling /b/ Versus /p/ in Trochees

The English category sets /b, d, g/ and /p, t, k/ are now usually referred to as voiced and voiceless stops respectively, although it is recognized that membership in these sets is not entirely determined by whether, according to commonly accepted definitions, a given phonetic element is voiced or voiceless; nor need it even be described as a stop. What is true is that if a phonetic element is phonetically a voiced stop then it will be assigned to the /b, d, g/ set, and if it is a voiceless stop it may, but need not be, assigned to /p, t, k/. A context in which the stop members of the two phonological sets may be distinguished simply on the basis of voicing (as narrowly defined with respect to stop consonants) is between vowels, as for example in the pair rabid-rapid. Acoustically, however, as many as 16 pattern properties can be counted that may play a role in determining whether a listener reports hearing one of these words rather than the other. In purely acoustic terms these properties are rather disparate, although most of them show variations that can plausibly be considered to be primarily the diverse effects of a relatively simple difference in the management of the larynx together with the closing and opening of the mouth. This diversity makes it difficult to rationalize a purely acoustic account of the rabid-rapid opposition, — i.e., one that makes no reference to the articulatory mechanisms and maneuvers by which the common linguistic effect of varying these acoustic properties might be explained.

Closure Duration and the Intervocalic Voiced-Voiceless Distinction in English

Thanks to the means available for speech analysis and synthesis, there is at present considerable knowledge of the acoustic differences whereby we distinguish among the various sounds of English. We know, for example, what the more important acoustic cues are by which listeners decide whether a stop sound is velar, alveolar, or bilabial.1 In the case of the stops, moreover, we know by and large what the acoustic differences are between the members of each of the homorganic pairs p-b, t-d and k-g2 (in other words, between the so-called voiced and voiceless stops3). But we are as yet unable to state with any exactness the extent to which each of these acoustic differences separately affects our ability to discriminate between the two kinds of stops. Moreover, where progress has been made in this direction it is certainly in some measure due to the fact that we have dealt chiefly with the stops in those positions where the voiced-voiceless contrast is phonetically maximal, that is, in initial and final positions. In the first context, aspiration is a prominent feature of the voiceless as against the voiced stops; in the second, the duration of a preceding vowel varies, depending upon whether the stop is of the one kind or the other. For these two positions it is difficult to separate features which contribute to what the phonetician would call voicing as distinct from aspiration and vowel duration, for there is some evidence suggesting that the acoustic correlates of these latter features may suffice as cues to the voicing distinction in the absence of any other differentia.4 There is, however, at least one position in English, namely the post-stressed intervocalic, where the voiced-voiceless distinction is observed even though,

Is it VOT or a first‐formant transition detector?

Discussion of voicing as a distinctive property of English stop consonants in initial position has centered on the measure of ’’VOT,’’ the time of onset of laryngeal signal relative to the noise pulse generated by the stop release, but it has been shown that listeners’ selection of ‖b,d,g‖ vs ‖p,t,k‖ responses to synthetic stop+vowel stimuli is not determined entirely by VOT. Significant effects have been reported to depend on the behavior of the first formant (F1) frequency immediately following voice onset, and on this basis it has been suggested that a ’’feature detector’’ responsive to a rapidly shifting F1 better explains the infant’s discrimination of the two stop categories than some mechanism which measures VOT directly. The relative importance of VOT as against the presence versus absence of F1 frequency shift after voice onset is assayed in several synthesis experiments in which VOT and F1 configurations are systematically varied. Labeling data obtained indicate that varying VOT regularly effect...

DISTINCTIVE FEATURES AND LARYNGEAL CONTROL

Physiological, acoustic, and perceptual data indicate that the timing of events at the glottis relative to articulation differentiates homorganic stops in many languages. Such categories are variously described in terms of voicing, aspiration, and force of articulation. Chomsky & Halle 1968 have proposed a universal set of phonetic features, four of which-voice, tensity, glottal constriction, and heightened subglottal pressure-allegedly operate to control the onset timing of laryngeal pulsing. But the observational basis for their analysis is flimsy, and Chomsky & Halle have no substantive argument for rejecting the possibility of temporal control of laryngeal function. Until fairly recently the non-historical study of language was, at least in the United States, pretty much the province of two groups of people: the grammarians and the phoneticians. Each group paid little if any serious attention to the problems and findings of the other, even in the area of phonology, where their interests would seem to converge. In the case of the phoneticians, their ignorance of linguistics was not normally elevated to a matter of principle. Some grammarians, however, refused to consider phonetic research an integral part of linguistics. Such work was consigned to physiology and physics at the very time that the primacy of the spoken over the written forms of language was being asserted most emphatically. The dichotomy drawn between langue and parole may have served as an excuse for minimizing the attention given to language in its most directly observable manifestation. Moreover, from the principle that only message-differentiating phonetic features are relevant to language description, linguists proceeded to the practice of knowing only as much about the processes of speech production and perception as sufficed to provide a set of labels by which to spell different messages distinctively.2 In the linguists concern with the components of sentences and their arrangements, his primary interest may not be in 1 This point has been discussed at length, with reference to various linguists, by Haugne 1951. 2 Phoneticians, often enough scolded for doing research not immediately relatable to the linguists own interests, have generally tried to remedy this situation; but sometimes this seems to take the form of renouncing research in any area not directly relevant to linguistics as most narrowly defined. Thus a phonetician with some training in linguistics can write, in connection with a study of mechanical pressures developed in the articulation of certain consonants, that the nasals are still another matter, as they do not enter into the lenis/fortis opposition, and calculating percentages of overlapping of their values with those of the stops would be meaningless (Mal6cot 1966a:176). Phoneticians have failed to exploit research possibilities that closer attention to linguists discussions would have made them aware of; but this does not imply that areas of phonetic research with which linguists have not concerned themselves are without relevance to linguistics. Recent discussion by Mattingly & Liberman 1969 suggests that linguists have been sometimes too ready to deny linguistic relevance to language and speech studies which threatened to yield findings not readily expressible in the current mode of linguistic description.

F0 gives voicing information even with unambiguous voice onset times

The voiced/voiceless distinction for English utterance-initial stop consonants is primarily realized as differences in the voice onset time (VOT), which is largely signaled by the time between the stop burst and the onset of voicing. The voicing of stops has also been shown to affect the vowels FO after release, with voiceless stops being associated with higher FO. When the VOT is ambiguous, these FO perturbations have been shown to affect voicing judgments. This is to be expected of what can be considered a redundant feature, that is, that it should carry a distinction in cases where the primary feature is neutralized. However, when the voicing judgments were made as quickly as possible, an inappropriate FO was found to slow response time even for unambiguous VOTs. This was true both of FO contours and level FO differences. These results reinforce the plausibility of tonogenesis, and they add further weight to the claim that listeners make full use of the signal given to them, even when overt labeling would seem to indicate otherwise.

Minimal Rules for Synthesizing Speech

In attempting to synthesize speech by rule, one must take account of the fact that the perceptually discrete phonemes are typically encoded at the acoustic level into segments of approximately syllabic length. It is, therefore, not possible to synthesize speech by stringing together prefabricated phonemes. By taking advantage of knowledge about the acoustic cues for speech perception, however, one can write rules for synthesis in terms of the constituent phonemes plus a few rules of combination. Thus, the number of rules can approximate the number of phonemes rather than the number of syllables. Indeed, one can reduce the number of rules still further by writing them in terms of subphonemic dimensions, viz., place and manner of articulation and voicing. Several complicating factors make it impossible to achieve an ideal minimum. First, rules must be added to take care of certain prosodic and positional variations. Failure to do so not only affects naturalness, but also impairs intelligibility, even at the...

Transillumination of the Larynx in Running Speech

A fundamental distinction between speech sounds depends on whether the excitation is a noise source or a quasiperiodic one. The noisy or voiceless sounds are presumably produced with an opened and quiescent larynx, while for voiced sounds the larynx is closed down and in rapid oscillation. Direct evidence of this has come from motion pictures recorded through the open mouth, a method obviously limited to particular sounds. Running speech requires different methods, and is being studied by a transillumination technique. A fiber‐optics bundle introduced into the laryngeal vestibule through the nose is used to illuminate the glottis, while a photocell placed below the thyroid cartilage registers the variable light transmitted through the glottis and the tissues of the neck. The “glottograms” so obtained are compared with acoustic waveforms simultaneously recorded with air and throat microphones to determine how the voiced‐voiceless distinction correlates with closed versus open states of the larynx.

On Pushing the Voice-Onset-Time (Vot) Boundary About:

There is voluminous evidence that homorganic stop consonants are distinguishable on the basis of voice onset time relative to their supraglottal articulation. For initial stops a convenient acoustic reference point is the onset of the release burst, and VOT has been defined as the interval between this point and onset of glottal signal. VOT boundary values between voiced and voiceless initial stops of English have been established by spectrographic measurements of naturally produced isolated words and by perception testing of synthesized CV syllables. The close match between the two kinds of boundary values suggests that fairly natural values were chosen for the invariant features of the synthetic speech patterns tested. It is known, however, that certain of these affect voicing perception. New data from synthesis experiments show that VOT boundaries shift with changes in transition duration, and that it is the first formant and not higher ones which are responsible.

Supraglottal Air Pressure in the Production of English Stops

Generally higher levels of intra-oral air pressure during occlusion are regularly reported for English /p,t,k/ as compared with /b,d,g/, and considerable significance has been attached to such observations, particularly by phonologists. Many of the published bodies of data, however, represent quite limited speech samples, consisting in the main of short nonsense utterances which contain stops in only a few contextual situations. In the present study ten readings of a list of twenty short English phrases were recorded for a single speaker. The phrases contained stops in initial, medial and final positions, and under various conditions of stress. A pressure transducer mounted at the end of a nasal catheter was introduced into the pharynx, and variations in intra-oral pressure were recorded on a multi-channel penwriter. From the pressure records obtained it appeared that /p,t,k/ and /b,d,g/ differ in their mean peak pressures, but that this difference is not equally significant in all contexts. Both initially and medially before a stressed vowel the two categories show peak pressures that overlap very extensively. This makes questionable the view that such differences more regularly co-exist with the /p,t,k/:/b,d,g/ distinction than any difference in voicing or mode of laryngeal operation, and that they demonstrate the fortis : lenis nature of the distinction.