Adjacent Vowels Research Articles

The perceptual basis of consonant cluster reduction and final devoicing

Recent studies have indicated that phonological change is a direct function of the acoustic cues for the segments involved. Here it is argued that cues for the voicing characteristic of the final post-vocalic and post-nasal stops and cues for final nasals can be ranked according to relative "strength". That is, cues for final voiced stops are weaker than those for final voiceless stops, which are in turn weaker than those for final nasals. This hypothesis is incorporated into a view of sound change which assumes (1) that consonant cues are carried by adjacent vowels, (2) that vowel reduction and loss weaken and obliterate these cues, and (3) that consonants with weak cues are replaced by homorganic consonants with stronger cues. This theory makes predictions about the direction of change and helps account for changes, such as the following, which have occurred during the history of German: /dIng/ →/ dInk/ → /dIn/. In addition, this theory is applied to a range of devoicing and consonant cluster reduction phenomena, including the widespread tendency among the languages of the world to devoice stops in final position, the loss of final voiceless stops in dialects of Chinese, and the absence of final /mb/ – /ng/ and /lb/ – /lg/ clusters in English.

Articulation of [i]'s

The articulation of the palatal vowel [i] in various phonetic environments was studied in one speaker of English and one speaker of Japanese using the technique of dynamic contact palatography. The utterances examined were of the form C1V1C2V2, where V1 or V2 was [i][u] or [a] and the other vowel always [i]. C2 was any of a variety of consonants proper to the languages involved. Preliminary results suggest that the articulation of [i] is influenced in complex ways by both consonantal and vocalic environment. For example, “front” consonants such as [n] serve to preserve the canonical contact pattern for [i] whereas “back” consonants such as [k] permit greater deviation from the canonical form due to coarticulation with adjacent vowels. Other things being equal, [i] seems to dominate in the coarticulation between vowels in the speech of the Japanese speaker. A momentary reduction in the extent of palatal contact during the consonant in an [ipi] utterance was noted. This is apparently due to passive enlargem...

Acoustic invariance in speech production: evidence from measurements of the spectral characteristics of stop consonants.

On the basis of theoretical considerations and the results of experiments with synthetic consonant-vowel syllables, it has been hypothesized that the short-time spectrum sampled at the onset of a stop consonant should exhibit gross properties that uniquely specify the consonantal place of articulation independent of the following vowel. The aim of this paper is to test this hypothesis by measuring the spectrum sampled at the onsets and offsets of a large number of consonant-vowel (CV) and vowel-consonant (VC) syllables containing both voiced and voiceless stops produced by several speakers. Templates were devised in an attempt to capture three classes of spectral shapes: diffuse-rising, diffuse-falling, and compact, corresponding to alveolar, labial, and velar consonants, respectively. Spectra were derived from the utterances by sampling at the consonantal release of CV syllables and at the implosion and burst release of VC syllables, and these spectra (smoothed by a linear prediction algorithm) were matched against the templates. It was found that about 85% of the spectra at initial consonant release and at final burst release were correctly classified by the templates, although there was some variability across vowel contexts. The spectra sampled at the implosion were not consistently classified. A preliminary examination of spectra sampled at the release of nasal consonants in CV syllables showed a somewhat lower accuracy of classification by the same templates. Overall, the results support an hypothesis that, in natural speech, the acoustic characteristics of stop consonants, specified in terms of the gross spectral shape sampled at the discontinuity in the acoustic signal, show invariant properties independent of the adjacent vowel or of the voicing characteristics of the consonant. The implication is that the auditory system is endowed with detectors that are sensitive to these kinds of gross spectral shapes, and that the existence of these detectors helps the infant to organize the sounds of speech into their natural classes.

A color display system of connected speech for the deaf

The system is composed of a computer for processing speech, digital memories for temporarily storing speech parameters, and a color coder for converting the parameters into composite color video signals. The video signals are transmitted from an antenna and received by conventional color TV sets. In voiced segments of connected speech, three formant frequencies are made to correspond to signals of three primary colors as follows: blue, EB = F2/3F1; the red, ER = 5F1/F3; green, EG = 3F3/5F2. The pitch signal controls the brightness of the colors. In unvoiced segments, as in brightness changes, only black and white stripe patterns appear on the TV screen. The patterns of connected speech flow from the bottom to the top of the TV screen. According to visual tests using connected vowels, the following results caused by simultaneous contrast effects in color can be shown. (1) The transition bands between two vowels judged using colors are much narrower than those determined by acoustical parameters. (2)The colors of adjacent vowels are emphasized with respect to each other so that it is possible that the influence of coarticulation on speech parameters can be compensated visually.

The feature(s) [grave

The relationship between the different proposed acoustic correlates of the feature [Grave] and the varying classes of sounds proposed to be defined by this feature is explored. After it is pointed out that two scales of gravity for consonants, one based on position of concentration of energy in the spectra of burst/fricative hiss of obstruents and the other based on the position of the start of F2 transition to adjacent vowels, are not identical, a means is proposed whereby a single feature [Grave] which applies to both consonants and vowels can be retained. It is further suggested that interchanges in point of articulation of labial and velar obstruents are always due to similarities in the spectral characteristics of the consonants themselves and not to similarities in F2 transitions.

Improvement in vowel articulation of deaf children

Three deaf adolescents who had poorly developed skills in oral communication were given specialized instruction designed to improve their ability to articulate vowels. This instruction was given for 1 hour weekly over a 5-month period, and it led to an improvement in vowel articulation that was substantial enough to be measured acoustically. The instruction emphasized articulatory relations between vowel sounds that are particularly difficult for the deaf to perceive visually or vibrotactually. The formant frequencies of five vowels were measured spectrographically before and after training. The measurements showed that all three children improved in their ability to articulate vowels as indicated by an increased frequency range of the second formant, a reduced amount of overlap in the frequency plots of adjacent vowels, and in more appropriate formant frequency values for individual vowels.

Breton (Treger dialect)

Fairly broad transcription. Stress is strong. Unstressed vowels are short, stressed vowels are half long unless before fortis (voiceless, double) consonants or consonant clusters, where they are short. Adjacent vowels are in hiatus and thus form two syllables, w, j are consonantal except when final or before a consonant where they represent the second element of falling closing diphthongs. , ã = a in W. Treger, ã in E. Treger. Contingent nasality before nasal consonants is not marked, ë, ä are e, a reduced towards ə except in the slowest, clearest forms of speech, θ = rounded ə. Lenis obstruent devoicing in final pausal position and in sandhi is marked .; fortis obstruent voicing in sandhi is marked ˅. h is a lenis, usually unvoiced, with some voicing possible between vowels and next to liquids; in final pausal position or in sandhi = x. m is a fortis. ɲ is a fortis; there is usually a j-glide between it and a preceding vowel, r is a light flap or trill; with some speakers it is ɻ; in some parts of Brittany it is R or B, but not in Treger; when written r it is not usually heard except in slow, clear forms of speech. I may be heard velarized in some districts, but not in Treger. t, d, n may be somewhat advanced towards a dental position, p, t, k may have slight aspiration except after s.

Probabilistic vector model for voicing mode identification of intervocalic stop consonants

A probabilistic vector model was developed to identify the voicing mode of phonemically equivalent intervocalic stop consonants with an accuracy commensurate with trained listeners, such performance to be attained independently of the succeeding vowel's identity or the differences among talkers. Acoustic features known to be salient to the perception and production of stop consonants were studied as random processes whose probability density functions provided models for voicing mode identification. Each acoustic feature present in a stop's production then contributed a vector whose magnitude and direction were determined from the probability distributions resulting from these density functions. Stop recognition was the result of summing these individual vectors, with the resultant vector specifying the voicing mode of the stop under consideration. Acoustic features incorporated into the model were closure duration, stop-consonant duration, fundamental frequency, adjacent vowel amplitudes, burst amplitude, formant transitions, voicing during closure, and voice onset time. When the performance of the probabilistic vector model was compared with that of trained listeners, correct voicing mode identification was approximately 99.0% in both cases.

Andegerebenha Vowel Phonology

The most common vowel system in Australian languages and the system ancestral to other Australian vowel systems is the triangular system /i-u-a/, with or without phonemic length. This paper shows that this is not appropriate for Andegerebenha, considers the suitability of a two-vowel system /e-a/, and rejects this in favour of one vowel plus length /a-a-/. The short vowel is greatly influenced in quality by the adjacent consonants; the longer one much less so. The consonant inventory is augmented by the addition of a number of labialised consonants, which embody the rounding of earlier *u. Conversely, the phenome/V/, a velar glide, was introduced when an earlier *w acquired the feature [-round] of adjacent vowels.

Coarticulation resistance in English /l/

Allophonic variations in the quality and voicelessness of RP /l/ were investigated, largely through their acoustic correlates as revealed by spectrography. The conventional separation of two distributionally-conditioned extrinsic allophones “clear” and “dark” was in general confirmed, although under different right-boundary conditions clear [l] was more widely used, itself varying in degree of clarity, than has been suggested elsewhere. Additionally, and more significantly, in each context /1/ varied further under coarticulatory influence—in its quality attributable to an adjacent vowel, in quality (in terms of overall clarity/darkness) attributable to an adjacent second lateral, and in voicelessness attributable to an adjacent voiceless consonant. Both right-to-left and left-to-right coarticulation effects were observed extensively: surprisingly, both types seemed constrained by a CV -type articulatory syllable. One general principle emerged which better explained the pattern of coarticulatory variation in the data: coarticulation resistance, a property which it is proposed to associate with phonetic specifications for speech segments in the form of values whose magnitude varies, increasing in the case of RP [ł], [ł] and [l] in that order. The validity of this principle for these and perhaps for other published data is independent of the direction (R-L or L-R) of the coarticulatory effect, and of the feature (vowel-quality, lateral-quality or voicelessness) being coarticulated.

Perception of temporal order in vowel sequences with and without formant transitions.

Temporal-order perception of phoneme segments in running speech is much superior to temporal-order perception in repeating vowel sequences. The more rapid rates possible in running speech may be due largely to the presence of formant transitions. In a series of five experiments we observed that many temporal-order misjudgements of repeating vowels can be explained in terms of auditory stream segregation, triggered for the most part by discontinuities in first-formant frequencies of adjacent vowels. Streaming, however, can be suppressed by formant transitions appropriate for the perception of stop consonants and by continuous transitions resembling those in coarticulated vowels. At rapid sequence rates, when the constraints of auditory streaming are removed, correct temporal-order identification is limited by linguistic transformations of vowels into other phoneme segments.

The Nature of Temporal Relationship between Adjacent Segments in Spoken Korean

Abstract In the first experiment, it was found that in Korean the duration of a consonant has a systematic relationship with the duration of the adjacent vowel. However, there is no indication that a syllable boundary affects the degree of influence on the durations of preceding or following vowels, which suggests that the syllable is not the unit of time programming in Korean. In the second experiment, it was argued that all the statistical results used to support the contention that individual segments are not the time programming units, are also compatible with the traditional view that segments are the units of time programming.

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.]

Temporal-Order Perception of Vowel Sequences with Monaural and Binaural Presentation

The ability of subjects to repeat alternated speech depends on the alternation rate. The possibility that alternation interferes with the perception of temporal order was investigated in an experiment in which subjects determined the temporal order of vowel sequences under monaural and binaural conditions of presentation. Test stimuli were formed from segments of the vowels /i, ɛ, a, u/. Each stimulus consisted of four equal segments of the four vowels arranged in a random order. Segment duration was varied from 300 to 50 msec. For half the stimuli, the vowels were contiguous and presented to one ear, while for the remainder of the stimuli adjacent vowels were presented to opposite ears. On a given trial, using either the monaural mode or the binaural modes, the ability of subjects to determine the vowel order decreased monotonically with segment duration. Performance varied from near perfect with 300 msec segments to near chance with 50 msec segments. Performance with binaural stimuli was consistently lower than with monaural stimuli.

Lateral pharyngeal wall movement during speech production.

Movement of one lateral pharyngeal wall during the production of VCVCV trisyllables was monitored via pulsed ultrasound using a time-motion method of display. Three normal speakers produced the six English stop consonants* combined with five vowels. The extent of lateral pharyngeal wall movement was found to be greatest on low vowels and least on high vowels. Pharyngeal wall movement during consonant production appeared to be dependent on the adjacent vowels.

Coarticulation Effects on Pitch

The effects of vowels and consonants on the fundamental frequency of syllables were explored by examining spectrograms of consonant-vowel (C-V) and V-C utterances spoken in isolation. Comparisons of the differences in average fundamental frequencies for these syllables revealed relatively little evidence of consonantal influence on the average fundamental frequencies of adjacent vowels. However, by comparing the degree and amount of rise in fundamental frequency (i.e., pitch contours) in syllable initial and final positions, certain interesting phenomena were observed. In general, syllables that begin with voiceless consonants have a pitch contour characterized by a higher initial Fo than their voiced counterparts and are followed by a sharper fall; syllables with voiceless consonants in final position have a characteristically sudden drop in the pitch contour. An explanation for these facts is offered in terms of changes in transglottal air-pressure drop as a result of variations in supraglottal vocal tract configuration as well as timing of laryngeal muscle gestures relative to consonant closure. It is felt that these predictable mechanical effects are among the variables that should be taken into account when analyzing the role of intonation in speech.

Toward Rules for Natural Prosodic Features in American English

The main body of a set of rules, which assign pitch and timing for the synthesis of reasonably fluent and natural-sounding speech, is described. The rules predict vowel durations with proper regard for the variation of “prevocalic” lengthening with over-all context. Rising, falling and rise-fall intonations are assigned quite plausibly in most cases. Three values of word boundary (juncture) are generated. At present, boundaries for careful but fluent speech are assigned quite well by simple rules. Possibilities for changing the speech rate may be incorporated without great difficulties. The rules are applied in three levels. The first level requires a limited syntax analysis. [Teranishi and Umeda, Int. Congr. Acoust., 6th, Tokyo (1968).] At this level, breath-group boundaries and certain other syntactic features are marked. In the second level, word boundaries not coinciding with breath-group boundaries are marked. In the final level, consonant durations and detailed pitch control are generated. Analysis at the second and last level require limited context only. Word boundaries depend on attributes of adjacent words passed down from the previous level. Vowel and consonant durations depend on information extending only up to the adjacent vowels.

Laterality Effects in the Recall of Steady-State and Transient Speech Sounds

The cues for discriminating fricatives are known to lie both in the spectrum of the friction itself and in the formant transitions into adjacent vowels. Although both these cues can vary acoustically with context, only the formant-transition cue is lateralized in dichotic listening. Steady-state fricatives, whether isolated or followed abruptly by a vowel were recalled equally well from either ear; however, when the appropriate transition was added there was an improvement in recall from the right ear, but no improvement from the left. This advantage for the right ear remained when the friction was removed but the transition left intact to give a stoplike sound. The right-ear advantage for recall of dichotically presented initial stops has been confirmed and a similar advantage found for both released and unreleased final stops. This latter effect is influenced by the relative onset times of the preceding vowels within a dichotic pair. By contrast, no preference for either ear could be found for isolated vowels of the same duration as the stop transitions (40 msec). These results indicate that speech sounds only give an advantage for the right ear in a dichotic listening task when they contain formant transitions.

Acoustic Correlates of Place of Articulation for Constants

Theoretical examination of the generation of sound by a constricted acoustic tube that models vocal-tract shapes for consonants suggests that a distinctive acoustic output is obtained for just four (or possibly five) major places of consonant articulation. The sound output corresponding to each of these places of articulation has well-defined attributes that can be described in terms of frequency positions and temporal changes of spectral energy concentrations relative to the spectral characteristics of the adjacent vowel. Several kinds of experimental data are presented in support of these theoretical considerations. These data are derived from studies of the acoustic characteristics of a variety of consonants (primarily English consonants) and from analysis of the characteristics of the sound generated by turbulence in an idealized model of the vocal tract for fricative consonant production, consisting of an acoustic tube with an adjustable constriction. [This work was supported in part by a contract from the U. S. Air Force Cambridge Research Laboratories, Office of Aerospace Research, and in part by a grant from the National Institutes of Health.]

How accurately must a speaker time his articulations?

Native speakers of a language can detect quite small departures from correct timing in the production of a sentence, which implies that they understand the rules used by the speaker to determine the timing of his articulations. In order to find out what the temporal constraints are on the speaker, an experiment was designed to measure the just noticeable difference (JND) for phoneme duration in natural speech. The stimuli were produced by an electronic equivalent of tape cutting and splicing. The incremental and decremental JND's were measured for the duration of the closure interval of a/p/, in initial position and in intervocalic position, in a stressed and an unstressed syllable, and for the duration of, if necessary, a stressed and an unstressed vowel. JND's were also measured for stimuli in which the duration of a stop and an adjacent vowel were altered simultaneously in opposite directions. 1) The results are not compatible with Weber's law. 2) Subjects were generally more sensitive to changes in vowel duration than to changes in stop-closure duration. 3) Subjects were no more sensitive to changes in the duration of a phoneme when the change was compensated in the following phoneme than when it was not compensated.