Speech Adaptation Research Articles

Linguistic relativism and selective adaptation for speech: A comparative study of English and Thai

The effect of repeated exposure to prevoiced and voiced stimuli on the voiced/voiceless category boundary was compared for monolingual native English speakers and bilingual native speakers of Thai. The English speakers, for whom the two adaptors were members of the same phoneme class, showed equivalent shifts towards the adapting stimulus under both conditions. The Thai subjects, who perceived the two adaptors as belonging to two distinct phoneme classes, showed a shift of the voiced/voiceless boundary only for the voiced adaptation condition; exposure to the prevoiced adaptor had no effect. Since Eimas and Corbit (1973) first demonstrated selective adaptation for the linguistic feature of voicing, their technique has been used with other categories and classes of speech (Cooper & Blumstein, 1974; Eimas, Cooper, & Corbit, 1973; Morse, Kass, & Turkienicz, 1976). The procedure is simple: repeti­ tion of a token taken from one end of a linguistic continuum leads to a shift in the category boundary, towards the adapting stimulus. In the original demonstration, subjects adapted with a Iba/ found stimuli previously lying on the Iba-pal boundary to be more Ipal-like, while those exposed to Ipa/ re­ ported just the reverse. The first studies of selective adaptation used as adaptors stimuli chosen from the extreme ends of the continuum, stimuli which could be considered good exemplars of their phoneme class. It was, therefore, of both empirical and theoretical interest to test the effect of adaptors closer to the category boundary, which, although they were not as good exemplars acoustically, would still be unambiguously perceived as belonging to one or the other phoneme class. Using the voice onset time (YOT) continuum, ,Anderson (1976) and Miller (Note 1) found that stimuli close to the preadaptation category boundary produced less adaptation than more extreme stimuli. Miller (Note 1) interpreted these results to support the existence of feature detectors selectively tuned to overlapping distributions of YOT values. The closer the stimulus to the optimal values, the more strongly the feature detector fires, and the greater the This research was funded by Grant R03 MH 26168 from NIMH and FRAP 11184 from the Research Foundation of the City University of New York. The author wishes to thank Drs. F. S. Cooper and A. M. Liberman for making available the facilities of Haskins Laboratories, Ralph Bianca for gathering the data, and the Thai Student Association of Columbia University for providing volunteers. Above all, thanks are due to the Thai students themselves, who endured not only the experimental sessions, but a 3-h round-trip subway ride as well. 347

Properties of friction analyzers for [j

Listeners were presented with stimuli from a nine syllable [?a]–[da] continuum (constructed using a natural speech [?a]) in which syllables differed in the duration of frication from 52 to 8 msec. In each experimental session, an initial identification test was conducted to determine the category boundary. This base-line test was followed by one of three adaptation tests in which subjects were presented with repeated syllables. The adapting stimuli consisted of (a) a single [?a] adaptor with 52 msec of frication, (b) a single [?a] adaptor with 32 msec of frication, (c) a ’’variable’’ adapting sequence that included the first six members of the test series and had a mean frication duration of 37 msec. The variable adapting sequence produced an adaptation effect equal in magnitude to the effect produced by the single adaptor with 32 msec of frication. The adapting syllable containing 52 msec of frication produced a significantly larger effect. The results demonstrate that (a) selective adaptation of speech did not require presentation of a constant adapting syllable and (b) the amount of adaptation depends upon the mean duration of frication contained in the adapting material at least up to 52 msec. A second experiment was performed to determine whether adaptation along a [?a]–[da] continuum was linearly dependent upon friction duration at values beyond 52 msec by comparing adaption produced by [?i], [?i], and [di] adaptors, containing friction duration of 120, 60, and 12 msec, respectively. The results showed that [?i] and [?i] produced highly similar adaptation effects on a [?a]–[da] series. The results suggest that an analyzer underlying perception of [?] is neither linearly nor symmetrically tuned to a particular range of friction duration.

Coarticulatory effects in speech adaptation

Experimental prosthodontic modification of alveolar contour induces a compensatory form of articulation. The initial effects on speech, noticeable particularly during the first few days of adaptation, are heard as lisping and a general indistinctness in articulation. After two weeks of adaptation, when articulatory compensations have been learned, the speech of most subjects sounds essentially normal. It would seem that the reorganization in this motor task must involve a reestablishment of subtilties in articulatory coordination. Data on jaw motions and tongue/palate contacts were analyzed to see the extent to which coarticulatory phenomena in VCV environments would be preserved in initial attempts to speak with a prosthesis and after adaptation to it. [Work supported by NIH.]

Effect of a dental prosthesis on speaker-specific characteristics of voice.

Subjects were fitted with an experimental dental appliance which had the effect of lowering and retracting the alveolar-palatal contour. Recordings of subjects speaking without the prosthesis and with the prosthesis at two stages of speech adaptation were analyzed acoustically and by judgmental evaluation for the effect of the appliance in disguising the voice. Individual differences were found in the type and extent of sibilant anf vowel changes produced by wearing the prosthesis. Listeners made more errors on a speaker authentication task involving recordings obtained immediately after the unfamiliar prosthesis was placed in the mouth than on recordings made after a week of accomodation to the appliance, or recordings made a week apart when no prosthesis was worn.

Persistence of learned motor patterns in speech

It has been noted that “spontaneous” compensations are made during speech for articulatory constraints such as clenching of the jaw. On the other hand, at least several days are required for speech adaptation to a dental prosthesis. A recording of speech and physiological data on jaw activity and tongue contacts was made at the time of initial insertion of an experimental dental prosthesis. Subjects wore the prothesis for two weeks of adaptation, after which a second recording was made, and then waited at least 1 month without experience in speaking with it. A follow‐up recording was then made with the same protocol as when the prosthesis was first inserted. Speech sounded much less disturbed by the insertion of the prothesis after a lapse of a month than it had at initial insertion—knowledge of the compensatory adjustments required had evidently been retained over that period of time. Palatographic data showed a rapid progression toward the compensatory pattern within a few repetitions of the speech sample. [Work supported by NIH.]

Some properties of the auditory component of selective adaptation

Manipulation of the spectral properties of adapter and test syllables in a selective speech adaptation procedure may provide a means to illuminate the microstructure of that component of the adaptation effect attributed to an auditory level of processing. In a pair of experiments, syllables having formant transitions which varied in rate, duration, and spectral location were synthesized to produce syllable series between different places of articulation and between the categories stop and semivowel. The series of test stimuli were identified in both the unadapted condition and after adaptation with a variety of syllables whose formants shared with the test series one or more of the transitional properties rate, duration, and spectral location. The pattern of phoneme boundary shifts produced by these adapters confirms the importance of spectral similarity between adapter and test syllables as a major determinant of the adaptation effect. The results suggest in particular that spectral similarity can be con...

Three models of perceptual adaptation to speech

Experiments were conducted to test the merits of three competing models of speech adaptation. Two models were sensory based and concerned (a) a shift and/or decline in the sensitivity peak of a speech analyzer following repeated listening to an adapting stimulus or (b) a narrowing of the stimulus range to which an analyzer is sensitive following repeated listening. A third model was response based, representing a bias effect of contrast. Three selective adaptation experiments were conducted using series of test syllables ranging from [ǰa] to [da], produced by variations in the duration of frication. In Experiments I and II, listeners were asked to identify by number each of seven test stimuli before and after adaptation with the [ǰa] endpoint member of the test series or after adaptation with either [ǰi] or [di]. A signal detectability analysis revealed highly systematic shifts in the peaks of sensitivity, as measured by values of d′ for adjacent stimuli. No systematic effects of response bias were obtained. In Experiment III, listeners were asked to identify nine test stimuli by consonant category [ǰ] or [d] before adaptation and after adaptation with either a single adapter (as in previous experiments) or with an adapting sequence that included the first six members of the test series. The average duration of frication of the six adapters was equal to the frication duration of the single adapter. The two adaptation tests produced nearly identical effects, suggesting that repeated listening to the single adapter did not reduce an analyzer's sensitivity range. The results provided support for a model in which the sensitivity peak of a speech analyzer is shifted by adaptation.

On trying to disguise the voice with a dental appliance

As part of a beginning study on speech adaptation to dental appliances, four young adults were fitted with an experimental alveolar-palatal acrylic appliance similar to an orthodontic retainer. They wore the appliance for one week in an attempt to learn to speak normally with it. The appliance causes phonetic distortions in the beginning, and a slight voice quality change. The question is whether these speech alterations, both in the beginning and after one week's adaptation time, are sufficient to affect a listener's judgment of the speaker's identity. Paired voice samples representing different persons, the same person with and without the appliance, and the same person with or without the appliance were judged by phonetically naive persons. Judges were asked to indicate whether each pair was spoken by the same person or different persons, and in addition to indicate their confidence in the judgment. Results are reported in terms of the number and types of errors the judges made. [Supported by a University of Maryland Faculty Research Award.]

Use of questions to elicit adaptation in the spontaneous speech of stutterers and non-stutterers.

Use of questions to elicit adaptation in the spontaneous speech of stutterers and non-stutterers.

Adaptation in mentally retarded stutterers and nonstutterers

Speech adaptation rates in mentally retarded stutterers and nonstutterers were observed to determine whether or not adaptation was consistently present in these groups, and to compare resulting adaptation behavior with that reported for normal and high IQ groups. Adaptation was found consistently in the MR groups and the influence of mental retardation upon this behavior was not evident.