Theories Of Speech Perception Research Articles

Stimulus variability and processing dependencies in speech perception

Processing dependencies in speech perception between voice and phoneme were investigated using the Garner (1974) speeded classification procedure. Variability in the voice of the talker and in the cues to word-initial consonants were manipulated. The results showed that the processing of a talker's voice and the perception of voicing are asymmetrically dependent. In addition, when stimulus variability was increased in each dimension, the amount of orthogonal interference obtained for each dimension became significantly larger. The processing asymmetry between voice and phoneme was interpreted in terms of a parallel-contingent relationship of talker normalization processes to auditory-to-phonetic coding processes. The processing of voice information appears to be qualitatively different from the encoding of segmental phonetic information, although they are not independent. Implications of these results for current theories of speech perception are discussed.

The role of peak amplitude in vowel recognition

Many theories of speech perception rely on the loci of spectral peaks as at least one factor upon which pattern recognition is based. However, when a peak is lower in amplitude than its neighbors, it may not be used in phonetic recognition. In the first experiment, a [u]-[i] series was constructed by manipulating the amplitude of a spectral peak (851 Hz). Subjects readily identified an item from the series with a low-amplitude, 851-Hz spectral peak as an [i]. It would appear that this peak, at a low amplitude, is not used at a phonetic level of processing. Further experiments test the perceptual locus of the use (or nonuse) of this low-amplitude peak information. Selective adaptation experiments were run in which the adaptors, including the [i] from the first experiment, varied in spectral overlap with a [u]-[u] test series in order to determine the degree to which the low-amplitude, 851-Hz peak is utilized in processing. The results will be discussed in terms of how peaks are analyzed at different levels of processing and how this relates to various theories of speech perception. [Work supported by NIDCD DC00219.]

Perception of vowels by budgerigars (Melopsittacus undulatus)

Budgerigars (parakeets) were trained using operant conditioning techniques to respond to differences between speech stimuli. Response latencies were used to construct similarity matrices and multidimensional scaling procedures were then used to produce perceptual (spatial) maps of these stimuli. For natural vowel tokens, budgerigars showed evidence of perceiving phonetic categories in spite of variation in talkers including talker gender and talker age. Experiments with synthetic vowel tokens generally confirmed and extended these findings. Multiple regression techniques revealed that the perceptual dimensions obtained from MDS were highly correlated with the frequencies of the first and second formants. These results suggest that both natural and synthetic vowels are probably perceived in similar ways by budgerigars and humans. These results have relevance for theories of speech perception and language development in humans. [Work supported by NIH.]

When the Objects of Perception Are Spoken

An analysis of the auditory enhancement theory of speech perception reveals its liabilities despite its plausibility. Four points are featured in a critique of its main premise: that the objects of speech perception are sounds, and, therefore, the perception of speech relies primarily on the categorization of auditory qualities. First, this critique notes that auditory enhancement theory is conventional, despite professed differences from the accounts which it disputes, in adopting consonant and vowel inventories as the objects of perception. Second, the critique argues that auditory enhancement theory fails three empirical tests of this equa- tion of phonetic attributes and auditory qualities. Third, the critique alleges that the auditory enhancement theory seeks support from equivocal experiments on animal training, in which the limited similarity of human and nonhuman test performance remains unexplored empirically and unexplained. Last, the critique concludes that the theory of speech perception by au...

Vocal-auditory functions in the chimpanzee: Vowel perception

The perception of vowels was studied in chimpanzees and humans, using a reaction time task in which reaction times for discrimination of vowels were taken as an index of similarity between vowels. Vowels used were five synthetic and natural Japanese vowels and eight natural French vowels. The chimpanzees required long reaction times for discrimination of synthetic [i] from [u] and [e] from [o], that is, they need long latencies for discrimination between vowels based on differences in frequency of the second formant. A similar tendency was observed for discrimination of natural [i] from [u]. The human subject required long reaction times for discrimination between vowels along the first formant axis. These differences can be explained by differences in auditory sensitivity between the two species and the motor theory of speech perception. A vowel, which is pronounced by different speakers, has different acoustic properties. However, humans can perceive these speech sounds as the same vowel. The phenomenon of perceptual constancy in speech perception was studied in chimpanzees using natural vowels and a synthetic [o]- [a] continuum. The chimpanzees ignored the difference in the sex of the speakers and showed a capacity for vocal tract normalization.

Human neuroethology of emotion

Ploog, Detlev: Human Neuroethology of Emotion. Prog. Neuro-Psychopharma-col. & Biol. Psychiat. 1989, 13 (Suppl):S15–S22 1. 1. Based on ethological theory, the question of what is the difference between human and nonhuman primate emotionality is investigated. 2. 2. The anatomical basis for this difference is the greater number of neurons in the anterior thalamic nuclei in humans than in monkeys and apes. This may represent an increased differentiation of the limbic message being sent to the cortex. 3. 3. Only humans can report about experiences and subjective feelings in certain motivational states. The two most general states are wakefulness and sleep. The subjective aspect of (desynchronized) sleep is dreaming. The causal relationship between dreaming and certain lower brain stem mechanisms is analysed. 4. 4. Whereas the motor system is usually blocked during desynchronized sleep, there are individuals who voice their emotions and speak while sleeping. As there are essential differences in the substrates for the voluntary control of the voice in the human and nonhuman primates there are essential differences in the voluntary control of emotions. 5. 5. Similar to the motor matching theory of speech perception a motor matching process of affect perception is suggested. 6. 6. The evolutionary change in the human motivational system is thought to be one of several prerequisites for the evolution of language.

Learning to identify phonemic orders

Several laboratories have reported that the order of items within repeating vowel sequences cannot be identified at item durations below 100 ms. This is puzzling, since phonemic sequences in speech consist of items that are considerably briefer. The present study demonstrates that listeners can, without feedback or knowledge of results, readily learn to name phonemic orders at item durations even briefer than those of normal speech. The listeners first identified different orders of components in repeating three‐item vowel sequences at a few hundred ms per item (an easy task), and were then required to identify orders of these sequences at regularly decreasing item durations. By a series of successive generalizations to ever shorter items, order was named down to about two glottal pulses/vowel (the shortest duration used). Within the durational range of vowels in speech, our listeners distinguished different arrangements by their resemblances to particular words. Below this range, listeners employed qualitative differences of a nonverbal nature. Implications for theories of speech perception will be discussed. [Work supported by NIH.]

Perception and production of the final /s-ts/ contrast in Dutch by misarticulating children.

This study investigates the specific relationship between speech perception and speech production by comparing 6- and 7-year-old children who misarticulated the final consonant cluster /-ts/ with three control groups. One control group of children misarticulated other phonemes than the specific /-ts/ cluster, another group of children had no articulation problems at all, and a third control group consisted of normally articulating adults. Perception was studied by assessing identification functions on a /moes/-to-/moets/ speech continuum in which the silence period had been manipulated. Production was analyzed by measuring the silence periods in productions of the Dutch word muts. The results show a clear hierarchical trend: The poorer the articulation proficiency of a group, the more variability there is in both production and perception. In perception, this variability is not restricted to the problematic /-ts/ cluster but includes the whole /-s/-to-/-ts/ contrast. Furthermore, it appears that a general as well as a specific relationship exists between perception and production and that the character of this relationship does not support the motor theory of speech perception.

Integration and segregation in speech perception.

This paper reviews the various applications of the concepts of integration and segregation in speech perception research. These applications are illustrated with selected examples from the literature. After laying out some conceptual issues and basic assumptions, the review discusses auditory temporal and spectral integration, integration of phonetic information in its various forms, auditory temporal, spatial, and spectral segregation, segregation of linguistic from paralinguistic information, and segregation of intertwined linguistic information. It is concluded that the concepts of integration and segregation are necessary ingredients of a theory of speech perception.

The problem of serial order in auditory word recognition

Recently, there has been a resurgence of interest in parallel distributed processing models of human perception. When speech perception is modeled in this type of spatially distributed network, a problem arises in coding the temporal order of perceptual units such as phonemes or words. In general, three solutions to this problem have been proposed: First, perceptual units may be context coded such as in context‐sensitive allophones. The order of units presented at different points in time can be determined by matching the context “edges” of each activated unit. Second, different portions of the network may represent different time frames. By this approach, the recognition of each successive perceptual unit activates representations in successive segments of the network. Finally, temporal order may be represented in the computational dynamics of the network. In this case, expectations about serial order are used to shift the focus of processing attention within the network. Thus, while the first two approaches recode temporal order into a spatial or spatial‐like representation, the third uses a temporal representation. Each of these approaches has positive and negative attributes, the implications of which will be discussed for a neuromorphic theory of speech perception. [Work supported, in part, by NIH.]

Perception of synthetic speech sounds by the budgerigar (Melopsittacus undulatus)

Two budgerigars were trained to respond differently to two synthetic speech tokens, /da/ and /ta/. Voice onset times (VOTs) of these synthetic stimuli were 0 and +70 msec, respectively. After reaching criterion level performance on these stimuli, the budgerigars were tested with synthetic stimuli having VOTs between 0 and 4–70 msec. The categorization functions derived from each bird’s responses to these intermediate stimuli were similar in shape but differed in their boundary locations from the functions typically obtained from humans and other mammals. In addition, the birds found this speech discrimination task extraordinarily difficult. These results define some of the limitations of the budgerigar’s auditory system for the processing of complex acoustic stimuli. Since the budgerigar has a distinctly nonmammalian auditory system, these results also have relevance for current theories of speech perception.

Speech sound identification influenced by adjacent “restored” phonemes

When listeners' identifications of speech sounds are influenced by adjacent sounds, do they operate only on the physical phonetic characteristics of there sounds or could they also use just their linguistic identity? We tested this by leading subjects to restore or induce the noise‐obliterated medial consonant in VCV utterances by first presenting them with several prior utterances where this medial consonant was consistently the same, either a /b/ or /d/. Included as V1 were synthetic vowels from the /i‐u/ continuum. As expected, more /u/s were identified out of this continuum in the environment of physically present /d/s than /b/s [Ohala et al., J. Acoust. Soc. Am. Suppl. 1 64, S18 (1978)]. But restored /d/s and /b/s had the same effect (though with slightly less intensity) thus indicating that the influence of context need not operate only via physical phonetic features. We discuss the implications of this finding for “direct realist” theories of speech perception and for claims of the existence of inv...

Effect of the spectral model order in automatic speech recognition

It has been observed that in speaker‐independent multi‐template digit recognition, the 5th‐order perceptually based LP (PLP) analysis method yields about 40% lower error rates than does the standard 14th‐order LP. In speaker‐dependent recognition, the 5th‐order PLP model yields essentially the same recognition accuracy as do high‐order PLP or standard LP models. In order to clarify the reasons for this result, RPS distances for LP and PLP models of single‐frame phoneme‐like spectra of male and female speakers were computed and arranged into distance matrices. Single‐speaker 10th‐order model matrices were adopted as the reference matrices for both LP and PLP methods. Cross‐speaker PLP matrices become more similar to PLP reference matrices as the model order of cross‐speaker matrices decreases to the 4th order. LP cross‐speaker matrices remain relatively similar to the 5th order. Single‐speaker PLP matrices remain relatively similar to the 5th order of the model while the single‐speaker LP matrices of lower order differ significantly from the high‐order ones. These results agree well with recognition rates obtained in the cross‐speaker alpha‐digit recognition for the same speakers. Results support the F1, F2′ theory of speech perception.

Timbre: A better musical analogy to speech

A recurring claim in speech research has been that “speech is special”—different perceptual processing is involved for speech than for other acoustic stimuli. In response to this claim, several demonstrations of speech‐like results have been reported for nonspeech stimuli, including musical intervals (e.g., major versus minor chords), and manner of playing a string (e.g., pluck versus bow). The research to be reported explores another, possibly more appropriate, musical analog: timbre. The notion is that families of instruments (e.g., horns, strings) may be analogous to families of phonemes (e.g., liquids, stops, fricatives). To test this notion, stimuli were synthesized to sound like various instruments. Standard speech experiments were then conducted, including forced‐choice identification, ABX discrimination, duplex perception, selective adaptation, and paired contrast. The similarity of the obtained results to those for speech will be reported, and the implications of these findings for theories of speech perception will be discussed. [Work supported by AFOSR.]

Effects of filter configuration on categorical perception of tactually presented speech

The effect of filter configuration on the relationship between auditory and tactual perception of speech was investigated via a 32-channel computer controlled electrocutaneous display and normal audition. Two 11-step synthetic speech continua, /a/ to /ə/ and /sta/ to /sa/, served as stimuli for study of three filter configurations: logarithmic, linear, and average (arithmetic mean of log and linear). Four well-practiced subjects performed two tasks with each stimulus continuum and each filter configuration: (1) a standard identification procedure where all 11 steps are categorized as either endpoint 1 or 11, and (2) a pairwise discrimination task (where equal interval stimuli were discriminated across the continuum). The same tasks were presented in the auditory modality for comparison purposes. Results indicate (1) a close correspondence between tactual and auditory perception, (2) categorical perception of the consonontal continuum and more continuous perception of the vowel continuum in both modalities, and (3) an interaction between filter configuration and speech syllable type that affects the degree of similarity between auditory and tactual functions. Implications in terms of theories of speech perception will be discussed. [Work supported by NIH.]

Spectral edge orientation as a discriminator of fricatives

Both “physical” and “auditorily transformed” spectra of English sibilant fricatives were investigated for evidence that might: (a) discriminate between /s/ and /∫/; (b) do so independently of vowel context; and (c) normalize the substantial speaker‐sex differences found. Measurements included the frequency location of the first peak of fricative noise, the location of the low‐frequency edge descending from this peak, the amplitude range of this descent, and its gradient or orientation. Analyses of variance on our data demonstrated a marked superiority for the orientation of the spectral edge, which differentiated the phonemes reliability, while remaining independent of vowel context or speaker sex. This was true. however, only of the edge as transformed in auditory space. Given the analogous ability of the eye‐to‐track edges, the notation appears to have some plausibility for speech perception theory and an obvious implication for automatic speech recognition. Perceptual confirmation of the importance of spectral edges in fricatives is now being sought.

Comparison of tactual and auditory discrimination of speech

The relationship between tactual and auditory perception of speech was investigated by presenting two speech continua to subjects via a 32‐channel computer controlled electrocutaneous display and via normal audition. The stimuli were synthetic /a/ to /ə/ and /sta/ to /sa/ with 9 equally spaced intermediate stimuli between each endpoint yielding two 11‐step continua. Five well‐practiced adults performed three tasks: (1) an adaptive discrimination task using each of the endpoints as target stimuli; (2) a standard identification procedure (all 11 steps categorized as either endpoint 1 or 11); and (3) a same‐different task (equal interval stimuli were discriminated across the continuum). For tactual discrimination, channel information was presented in three spectral configurations conforming to (1) logarithmic, (2) linear, and (3) average (arithmetic mean of log and linear) filtering. Results indicate (a) a close correspondence between auditory and tactual perception, (b) subphonemic discrimination in both modalities, (c) categorical perception in both modalities, and (d) a more auditorylike pattern with log and average filter configurations than with linear. Implications in terms of theories of speech perception will be discussed.

The motor theory of speech perception revised

A motor theory of speech perception, initially proposed to account for results of early experiments with synthetic speech, is now extensively revised to accommodate recent findings, and to relate the assumptions of the theory to those that might be made about other perceptual modes. According to the revised theory, phonetic information is perceived in a biologically distinct system, a ‘module’ specialized to detect the intended gestures of the speaker that are the basis for phonetic categories. Built into the structure of this module is the unique but lawful relationship between the gestures and the acoustic patterns in which they are variously overlapped. In consequence, the module causes perception of phonetic structure without translation from preliminary auditory impressions. Thus, it is comparable to such other modules as the one that enables an animal to localize sound. Peculiar to the phonetic module are the relation between perception and production it incorporates and the fact that it must compete with other modules for the same stimulus variations.

Shadows, echoes and auditory analysis of speech

A review of research on speech-motor perceptual and auditory processing, based particularly on studies at the I.P. Pavlov Institute of Physiology in Leningrad by V. Kozhevnikov and L. Chistovich. The use of spoken responses to speech stimuli, i.e. shadowing and mimicking, is given particular emphasis. Early versions of a motor theory of speech perception are also presented.

Auditory responses in avian vocal motor neurons: a motor theory for song perception in birds.

The hypoglossal motor neurons that innervate the vocal organ (syrinx) of the male zebra finch show a selective, long-latency (50-millisecond) response to sound. This response is eliminated by lesions to forebrain song-control nuclei. Different song syllables elicit a response from different syringeal motor neurons. Conspecific vocalizations may therefore be perceived as members of a set of vocal gestures and thus distinct from other environmental sounds. This hypothesis is an avian parallel to the motor theory of speech perception in humans.