VOT Values Research Articles

Stimulus dominance in dichotic listening.

If, in dichotic listening, vowel onsets of competing plosive-vowel sequences are separated by some 30-60 msec, the signal in the "lagging" ear is more intelligible than in simultaneous presentation (Berlin et al., 1973). The dominance of voiceless over voiced stops in voice-contrasting pairs as found in English (Sepaks and Niccum, 1977) may be due to this lag effect as VOTs of initial voice-contrasting pairs differ radically in English. Therefore, Dutch plosive-vowel sequences, which do not differ much in their VOT values, were presented dichotically. The competing stimuli were aligned with respect to their vocalic boundaries. No dominance of voiceless over voiced plosives could be found for Dutch; on the contrary, there was a slight advantage for voiced over voiceless plosives. When the competing plosives shared the feature voice, a right-ear advantage could be established. Thus, voiceless dominance in English probably results from the noise-burst alignment technique commonly used.

/s/ + stop clusters in the speech of ten language-delayed children

In the acquisition of /s/ + stop clusters, before correct production, children commonly substitute one of the members for the entire cluster, the most common substitute being a stop at the appropriate point of articulation. Furthermore, there is tentative evidence that some children attempt to distinguish the stops substituting for clusters from voiced stops. The purpose of this investigation was to compare the acquisition of /s/ + stop clusters by language-delayed children to normal acquisition, first, in terms of phonemic substitutions and, second, in terms of differential use of VOT for the substituted stops. Ten children diagnosed as language-delayed served as subjects, C.A. 3;0 to 8;7; all used language below expectation for their chronological age. Each child recorded three tokens of near minimal triplets, beginning with a voiced stop, a voiceless stop and an /s/ + stop cluster, (e.g., Kate, gate, skate), for all three points of articulation, bilabial, alveolar, and velar in the course of identifying pictures. The children produced 95% of the intended clusters as singleton stops. VOT measurements for all stops were made from spectrograms. If the criterion for appropriate use of VOT to distinguish voiced from voiceless stops is taken to be +30 ms, then six of the ten children were producing voiceless and voiced stops appropriately over 80% of the time. For these children, 86% of the stops substituting for clusters were produced with VOT values in the long lag range. The other four children employed a unimodal VOT distribution for all types of stops, voiceless, voiced, and substituting for clusters. The data suggest that language delayed and normal children follow similar strategies in the acquisition of /s/ + stop clusters.

The effect of pitch change on rate normalization

In accordance with earlier results of Summerfield, the VOT value for the perceptual boundary between GA and KA was found to vary as a function of the tempo of a precursor phrase. The boundary shifts toward shorter VOTs with increased rate of utterance. Analogous rate dependent results were obtained when the perceived voicing characteristic was cued by closure duration rather than VOT. A sizable downward shift in pitch between the precursor phrase and the test item leads to a significant reduction in rate normalization.

In qualified defense of VOT.

The VOT measure has been said to provide the single most nearly adequate physical basis for separating homorganic stop categories across a variety of languages, granted that other features may also be involved. That transition duration affects perceived voicing of synthesized initial stops of one specific language, English, has suggested the hypothesis by Stevens and Klatt (1974) that a detector responsive to rapid formant-frequency shifts after voice onset better explains the child's acquisition of the contrast than does some mechanism which responds to VOT directly. If such a detector is part of our biological equipment, then it seems remarkably underutilized in language, for the hypothesis asserts that basic to voicing perception is whether laryngcal signal is or is not present during the interval in which the stop-vowel transition occurs. In effect, the “archetypical” voiceless stop is aspirated. Not only do many languages not possess voiceless aspirates, but even in English aspiration is severely restricted. Of course the VOT measure has its limitations - it is inapplicable to prepausal stops. However, there are much more serious difficulties with the posited detector, since even for the English initial stops there is evidence that the presence of a voiced first-formant transition is not required for the perception of /bdg/, nor does the absence of such a transition necessarily yield /ptk/, provided appropriate VOT values are provided.

Hemispheric specialization for VOT perception in the preschool child

Auditory evoked potentials (AEP) were recorded from the left and right hemispheres of six male and six female right handed preschool age children (4 years old) in response to stop consonants with VOT values of 0, +20, +40, and +60 msec. Neuroelectrical correlates of between category phoneme discrimination were found only in the right hemisphere. A second component of the cortical response was sensitive to changes both within and between the phoneme categories. Other AEP components reflected sex and hemisphere differences.

Cross-language study of speech-pattern learning

In order to investigate the nature of some processes in speech acquisition, synthetic speechlike stimuli were played to groups of English and French children between two and fourteen years of age. The acoustic parameters varied were voice onset time and first-formant transition. Three stages were observed in the development of children’s labeling behavior. These were called scattered labeling, progressive labeling, and categorical labeling, respectively. Individual response patterns were examined. The first stage (scattered labeling) includes mostly children of two to three years of age for the English and up to about four for the French. Children label most confidently those stimuli closest in physical terms to those of their natural speech environment. All stimuli with intermediate VOT values are labeled quasirandomly. Progressive labeling behavior is found mostly amongst children aged three and four for the English, up to about seven for the French. Children’s response curves go progressively—almost linearly—from one type of label (voiced) to the other (voiceless): response follows stimulus continuum. Categorical labeling becomes the dominant pattern only at the age of five to six for the English, one or two years later for the French. This development was found to be highly significant (p smaller than 0.003 for both English and French, using Kendall’s tau measure of correlation). English children learn to make use of the F1 transition feature around five years, whereas French children never use it as a voicing cue. French children will have fewer features than English children at their disposal: This may account for the later age at which French children, as a group, reach the various labeling behavior stages, and for labeling curves being less sharply categorical for French than for English children. These findings indicate that categorical labeling for speech sounds is not innate but learned through a relatively slow process which is to a certain extent language specific. The implications of the results are discussed in the light of previous work in the field.

Voicing distinction in Polish word-initial stop consonants

Voice onset time measurements were made of word-initial stop consonants produced by native speakers of Polish in order to determine how the values for the Polish voiced and voiceless categories are distributed along the VOT continuum. According to traditional phonetic descriptions, the Polish language utilizes the voicing lead and short-lag categories along the VOT continuum to contrast its voiced and voiceless stops. VOT measurements were made of stops produced in a number of different speaking conditions: minimal pairs read in citation form, words embedded in controlled sentence contexts, and words excised from both formal and informal discourse. Preliminary results indicate that VOT values for productions of the Polish voiced stop are distributed almost exclusively within the “lead” region of the continuum. Values for the voiceless stop, however, in certain conditions do not cluster in the predicted, contiguous short-lag region of the continuum. In certain speaking conditions Polish voiceless stops cluster within the noncontiguous, long-lag region and tend to be aspirated. Acoustic characteristics of stop consonants as a function of speaking style will be discussed, and results of Polish listeners' identification of synthetic CV syllables along a VOT continuum will be provided.

Properties of feature detectors for VOT: the voiceless channel of analysis.

Binaural adaptation was combined with a dichotic ear-monitoring task to estimate the output function of the voiceless detector sensitive to VOT and to determine how this function is modified by selective adaptation. On each test trial, a voiced stop was paired with one of a set of voiceless stops, which spanned a range of VOT values. The percentage of voiceless responses per pair, which reflects both the number of correct voicing judgments on voiceless targets and the number of voiceless intrusions on voiced targets, was measured before and after adaptation with a voiced stimulus and a variety of voiceless stimuli. Before adaptation, the percentage of voiceless responses per pair systematically increased with an increase in the VOT value of the voiceless stimulus of the pair. Adaptation resulted in a shift in the percentage of voiceless responses per pair in the predicted direction, with progressively larger shifts occurring for voiceless adapting stimuli with higher VOT values. These results indicate (1) that the output of the detector is graded signal whose magnitude is an increasing function of VOT value and (2) that adaptation depresses the entire output function from its preadaptation level, with the amount of depression determined by the extent to which the adapting stimulus excites the detector.

Mothers' simplification of phonetic input to their children in English and Polish

Voice onset time measurements were made of three mothers' productions of word-initial spiral stops to their 18 month old children in order to determine whether mothers play a significant role in their child's acquisition of the voiced—voiceless distinction. Two of the mothers spoke native English to their children, and one spoke native Polish. Measurements were also made of each mother's productions of word-initial stops to another adult. Results show that mothers speak differently to their 18 month old children than they do to other adults. All three mothers employed certain clarification strategies for reducing the overlap of VOT values between voiced and voiceless categories which exists in normal adult adult discourse. These strategies included an increase in the median VOT values for productions of the voiceless stop, and/or a reduction in the range of VOT values for the voiceless category. Mothers, therefore, seem to be facilitating the task of phonological acquisition for their children by presenting them with a maximally distinct, tailored phonetic input. [Work supported by NICHD.]

Voicing in the speech of language‐delayed children

Voicing as a phonological contrast emerges relatively early in the speech of children learning English. However, the acoustic realization of this contrast is quite complex. The purpose of this study was to investigate the details of the control of the voicing contrast in the speech of 10 children diagnosed as language‐delayed in comparison with the speech of 10 children developing language normally. Each child was given a short test to determine if he could respond appropriately to minimal pairs differing in the voicing of stop consonants, and then recorded repeating 12 test words. Spectrograms of the three best productions of each word were analyzed, examining VOT differences between voiced and voiceless stops in initial position and the duration of the vowel preceding the stop pairs in final position. For all children, the control of VOT seemed to be in one of three stages of development: (1) overlap, (2) little overlap but great variability, (3) little overlap and decreased variability. A similar pattern of development, involving a decrease in overlap with a subsequent decrease in variability, was observed for vowel duration. Neither VOT nor vowel duration values comparable to adult values were achieved by the children.

Voice onset time for voiceless stop consonants in the fluent reading of stutterers and nonstutterers.

Ten stutterers and ten nonstutterers read ’’The Rainbow Passage.’’ Intervocalic voiceless stop consonant segments were selected and displayed on wide-band spectrograms. Results of VOT measurements obtained indicated that (1) the stutterers displayed longer VOT values than the nonstutterers, and (2) VOT values increased in duration as place of articulation moved back in the oral cavity.

Speech perception by the chinchilla: Categorical perception of synthetic alveolar plosive consonants

Kuhl and Miller [Science 190, 69–72 (1975)] recently demonstrated that in an identification task the chinchilla responds to (“labels”) an alveolar synthetic-speech continuum as though an abrupt qualitative change occurs at the place where English speakers hear a change from /da/ to /ta/. Discrimination functions for these same stimuli have now been obtained. The animal's task is to detect a change from a repeating “standard” stimulus to a “comparison” stimulus and report that change by crossing a midline barrier to avoid shock. The standard stimuli are 0, +10, +20, +30, +40, +50, +60, +70 and +80 msec VOT. The VOT value of the comparison stimulus is gradually changed until it approaches the VOT value of the standard. The animal's threshold (the minimum Δt detected with 75% accuracy) is measured for each standard stimulus and is best when the standard is +30 msec VOT, that is, at the animal's perceptual boundary as defined in the labeling experiments. This agreement between the labeling and discrimination data demonstrates categorical perception for a voiced-voiceless continuum by an animal listener. [Supported by NIH grant NS 03856 to CID.]

Properties of feature detectors for VOT

Certain properties of a feature detector system sensitive to VOT were investigated with a binaural adaptation-dichotic testing paradigm. On each dichotic test trial, a nonboundary voiced stop was paired with one of a set of voiceless stops. The voiceless stops varied in VOT, from values close to the phonetic boundary to values well within the voiceless category. The relative effectiveness of each of the voiceless stimuli in competing for processing with the voiced stimulus was assessed before adaptation and after adaptation with voiceless stops with a range of VOT values. During both pre- and postadaptation, the number of correct voicing responses when targeting for the voiceless stop, and the number of voicing intrusions when targeting for the voiced stop, systematically varied as a function of the VOT value of the voiceless stimulus. In addition, the VOT value of the adapting stimulus determined the amount of adaptation obtained. These results indicate first, that the output of the detector is a graded signal, and second, that the magnitude of decrease in detector sensitivity during adaptation is a function of the VOT value of the adapting stimulus. [Supported by NIH grants.]

Voice onset time of spontaneously spoken Spanish voiceless stops

Spectrographic measurements were made of voice onset time for 161 tokens of Spanish voiceless stops from one individual's conversational speech in an attempt to replicate laboratory findings on VOT. In running Spanish, occlusive allophones of voiced-stop phohemes occur only in absolute initial position and after nasal consonants. In other environments, voiced-stop phonemes are phonetically voiced fricatives. Therefore the contrast between voiced- and voiceless-stop phoneme categories is maintained not only by the presence or absence of voicing, but also by the presence of frication (voiced phonemes) or its absence (i.e., closure for voiceless phonemes). Due to the limited distribution of voiced-stop allophones, these are scarcely represented in the present sample, and their VOT values are not reported. In spite of VOT's “reduced” work load, however, the VOT values for voiceless stops in this study conform quite closely to earlier observations of production and perception of Spanish stop phonemes. VOT values for voiceless stops in utterance-initial and post-nasal positions are not significantly different from those for stops in other positions.

VOT Discrimination by four to six and a half month old infants from Spanish environments

Thirty 4–6.5 mo-old-infants from monolingual Spanish speaking homes were tested for discrimination of bilabial stop consonants differing in VOT. Discrimination was determined by the habituation and dishabituation of the orienting reflex as indexed by cardiac deceleration. Subjects were placed in one of three groups of ten subjects each. The three groups were used to determine whether bilabial stops with VOT values of −60 and −20 msec, −20 and +20 msec or +20 and +60 msec could be discriminated. Subjects were presented eight trial blocks of ten repetitions of a CV syllable; these eight trial blocks were followed by two trial blocks of ten repetitions of a different CV syllable. The only difference between the CV syllables presented during the first eight and the last two trial blocks was the VOT value of the bilabial stop consonants. Subjects in the −20 and +20 msec group showed no signs of being able to discriminate these stimuli while the subjects in the −60 and −20 msec group and the +20 and +60 msec group did. These results correspond to those reported for infants from English speaking families but do not correspond to those reported for Spanish speaking adults. Thus, the failure to find evidence of discrimination in the −20 and +20 msec group suggests that experience has little effect in determining bilabial stop phonetic categories in 4–6.5 mo-old infants.

Tape Cutting Experiments with Stop Consonants

In Danish words with /p,t,k,b,d,g/+/i,a,u/, stop bursts, aspirations, and formant transitions have been removed and exchanged. [Danish /p,t,k/ are strongly aspirated (/t/ affricated), /b,d,g/ are voiceless.] 500 stimuli were presented to 21 listeners, who had to identify the words. Main results were: For the distinction between /p,t,k/ and /b,d,g/, the small difference in stop burst is insignificant, whereas the aspiration is decisive (a simple pause between burst and vowel is not sufficient). The duration of the aspiration is important, but the start of the vowel (noisy start of higher formants) may be sufficient for the identification of /p,t,k/, particularly before /a/, even with VOT values of 20–30 msec. For the identification of place of articulation the aspiration (affrication) of /p,t,k/ is sufficient except for /k/+/a/, but the burst is also sufficient except for t. In /b,d,g/ the transition is sufficient if it is extensive (e.g., /ba/, /da/, /ga/, /bi/, /du/); it is not enough that it points to the locus. The burst is necessary not only in velars before rounded vowels, as previously assumed, but also, for example, in /bu, /di/, and /gi/.

Some Perceptual Mechanisms in Speech Processing

Enough is known about the acoustical cues in speech to enable highly intelligible synthesis from simple algorithms, but, in most cases, the most valid expressions of cues from the point of view of perception and the mechanism of detection, weighting, and decision are not known. Experiments that manipulate several cues and the phonetic context offer some hope of charting information flow through the human perceptual process and providing detailed block diagram models. Several examples are given. Perception of voicing in initial stops depends upon the place of articulation as well as the traditional voicing cues such as VOT—i.e., the boundary VOT value differs with place. Two separate experiments show that this effect is carried not by acoustical place information but by a perceptual decision about place, logically prior to the: voicing decision. But, in another context effect, dependence of consonant place upon the adjacent vowel, the effect appears conditioned by the acoustical vowel information and not by the decision as to linguistic identity. Implications for perceptual theory and speech recognition are pointed out. [Work supported by J.S.R.U.—U.K.]