Phoneme Perception Research Articles

Modeling closed-set phoneme and open-set word recognition by multi-channel cochlear implant users

There has been phenomenal growth in research on speech perception by cochlear implant (CI) users since the printing of the ‘‘Bilger Report’’ [Bilger et al., Ann. Otol. Rhinol. Laryngol. 86 (S38), 1–176 (1977)]. Undoubtedly, average speech perception performance with these devices has improved dramatically. However, despite advances in implant technology, CI users continue to demonstrate a wide range in the ability to perceive speech. Little progress has also been made in understanding how CI users actually perceive speech. Although many correlational analyses have been carried out, little research has focused on mechanisms of speech perception, and many clinical decisions are made on a trial-and-error basis. Using a new approach, we have developed a quantitative, psychophysically based model (Multidimensional Phoneme Identification, MPI) of phoneme perception by CI users [M. A. Svirsky and T. A. Meyer, J. Acoust. Soc. Am. 103, 2977 (1998)]. The MPI model generates phoneme confusion matrices from performance on psychophysical tasks. In a complimentary line of work, we aim to predict open-set spoken word recognition from phoneme performance [S. Frisch and D. B. Pisoni, Res. Spoken Lang. Proc. 261–288 (1998)]. In both lines of research, we have found predictable relations between perception of phonetic features, phonemes, and words by CI users. [Work supported by NIH, AAO-HNS, DRF, NOHR.]

Contribution of articulatory kinematics to visual perception of stop consonants

Little is known about visual cues that are critical to the accurate categorization, and perceived quality, of individual phonemes. The current study provided a preliminary evaluation of the relationship between orofacial movements and visual perception of phonemes distinguished by place of articulation. A female talker produced VCV utterances for /b/, /d/, and /g/ consonants; a natural range of variation was obtained by producing each consonant in several vowel contexts. Peak displacement, peak closing velocity, and peak opening velocity were measured from the upper lip, lower lip, and jaw. Observers identified place and rated the perceived quality, or goodness, of each token as an example of the consonant category (1=poor,7=good). As expected, observers identified bilabials with near-perfect accuracy, whereas they often confused alveolars and velars. More importantly, kinematic variation affected category goodness, particularly for /b/. Linear regression analyses indicated that goodness ratings within the /b/ category were highly correlated with lower lip and jaw displacement. In fact, articulatory displacement predicted these goodness ratings more effectively than did vowel context. Implications for understanding visual place cues are discussed. [Work supported by NICHD.]

Spoken word recognition by non-native listeners

While the ability to perceive novel phoneme contrasts is necessary for accurate spoken word recognition by non-native listeners, it is likely that novel phoneme perception functions in a linguistically meaningful manner only once the contrast in question signals a known lexical contrast. In other words, knowledge of the sound-based structure of the target language lexicon is just as important for non-native speech perception as experience with the structure of the target language phoneme inventory. In order to understand non-native speech perception fully, there is a need to investigate recognition of word-sized units by non-native listeners using stimulus materials that are well controlled in terms of the sound-based structure of the target language lexicon. Accordingly, how non-native listeners recognize English words under conditions known to affect spoken word recognition by native English listeners was investigated. Results showed that the ability to take advantage of consistent surface phonetic information, such as a consistent talker across items, is a perceptual skill that transfers easily from first to second language perception. However, non-native listeners had particular difficulty with words that are easily confused with other similar sounding words, suggesting that non-native word recognition may be compromised when fine phonetic discrimation at the segmental level is required.

Conduction aphasia-11 classic cases

Eleven cases of conduction aphasia derived from a survey of the literature of the classic period of German aphasiology are reviewed. They were published between 1885 and 1934. Only those case reports were included that presented sufficient data concerning spontaneous speech, repetition and comprehension and gave examples of the patients' verbal behaviour. In summary, conduction aphasia was characterized by; (1) phonemic paraphasia with conduite d'approche, (2) preserved auditory and reading comprehension, (3) a repetition impairment that depended upon word length, (4) paragraphia with writing to dictation, and (5) a tendency in some patients to produce the unusual symptom of form-related semantic paraphasias. As anatomical explanations, it was proposed that there existed two routes, a phonological and a semantic, for repetition and that the right hemisphere may contribute to comprehension performance. Functionally, a dissociation between phonemic components of words, a disorder of the generation of the temporal sequence of speech sounds or deficient transcription of speech sound images into speech motor images, an impaired production of the single phoneme, a disorder of the associative basis for phonemes, impaired phoneme perception (incomplete sound deafness), an impairment of word sound control and the contribution of a memory deficit were assumed as relevant preconditions. The theories of the classical writers are compared with modern accounts of the pathogenesis of conduction aphasia.

Normal and disturbed phoneme perception as reflected by the mismatch negativity: Do event-related potentials help to understand dyslexia?

Normal and disturbed phoneme perception as reflected by the mismatch negativity: Do event-related potentials help to understand dyslexia?

The relationship between non-native phoneme perception and the MMN

The mismatch negativity (MMN) auditory-evoked potential reflects the detection of minimal stimulus differences at a preattentive level. It has been reported that the MMN reflects acoustic, rather than phonetic processing of speech stimuli. The purpose of the present study was to examine the relationship between discrimination at a neurophysiologic level, as reflected by the MMN, and the behavioral ability of native speakers of American English to discriminate and label pairs of Hindi stop consonants that are contrasted by a place of articulation feature that is not phonemic in English. These pairs were the voiceless unaspirated Hindi stop consonants, dental and retroflex /t/, and the voiced unaspirated Hindi stops, dental and retroflex /d/. The present study addressed two questions: (1) whether the MMN response would be evoked by acoustic differences resulting from differences in place of articulation, even though behavioral discrimination ability was poor; and (2) whether there would be a difference in the presence/absence or other parameters (latency, duration, amplitude, and area) of the MMN response depending on which voicing contrast was presented.

Listener variability and multiple perception processes

The relationship between the perception of individual phonemes and the perception of words is investigated to determine if these are similar tasks for listeners. Since listeners may be able to use top-down processes in full words, it is argued that, at least for some, the task of identifying isolated phonemes may be very different from the task of identifying a full word. Because greater variability in performance is expected from non-native speakers, 15 native speakers of Sudanese Arabic (as well as a control group of 12 native speakers of American English) were tested on identification of American English phonemes in word context and the same instance of these phonemes extracted from the words. The overall agreement between corresponding items on the two tests was significantly better than chance, but scores of individual subjects varied substantially, with phoneme-word test agreement falling below chance for many subjects. Additionally, not all non-native subjects performed better on the full word task, suggesting that they do not all use higher-level information. It is therefore argued that studies of phoneme-level perception are more relevant to some listeners’ word perception than to others’.

The Phoneme Perception Task (PPT): A Measure of Phoneme Perception in Orally Read Sentences

Analysis of scores on the Phoneme Perception Task (PPT), a new measure of perception of phonemes in sentences, revealed an overall 75% accuracy of classification of students' group membership (56 students who never received special education services in grades 2 through 5 (non LD group) and 40 students of corresponding age, attending a school for students with learning disabilities (LD group). Concurrent validity was determined using the SCAN: A Screening Test for Auditory Processing Disorders (r =.39) (Keith 1986) and the Test of Written Language-2 (TOWL-2) Spontaneous Writing Quotient (Hammill & Larsen, 1988) ( r =.48). Interjudge reliability (.97) indicated strong observer agreement. Internal consistency (alpha = .74) was considered to be reduced by specific items that need revision prior to further study.

Perceptual development of the categorization of the /r-w/ contrast in normal children

The present study examined the course of the perceptual development of the /r-w/ contrast. In an attempt to extend earlier research, which focused primarily on isolated age groups, data from four age-groups were gathered in the study: adults, and three, four, and five year-old children. The experimental paradigm involved a two-alternative, forced-choice identification task. The stimuli were members of a 7-step synthetic [r-w] continuum modeled after an adult male voice, varying in onset and duration of F2and F3formant transitions. Percent /w/ responses were computed and transformed to yield measures of phonemic boundaries. The significant shift in the phonemic boundary with age, as well as the increase in both the slope-steepness of the identification curve and the consistency of the responses as a function of age, indicated progressive development in phonemic perception. Results are discussed in terms of age-related changes in the auditory system and cognition.

Categorical perception effects induced by category learning.

The authors report a series of studies designed to determine whether effects similar to those observed in the innate categorical perception of color and phonemes are induced during the learning of simple unidimensional categories and more complex multidimensional ones. In Experiment 1 no evidence was found for such effects when stimuli varied on 1 dimension. Experiments 2 and 3 demonstrated a within-category compression effect but no between category expansion effect for stimuli varying in 2 dimensions. Compression only was also shown in Experiment 4, which used pictures of actual objects. Multidimensional scaling analyses illustrate how within-category compression without expansion was sufficient to produce categorical clustering of items in the similarity space. These analyses also show that learning changed the dimensional structure of similarity space. Results are compared with those from other studies exploring similar phenomena and with neural network simulations.

Chairs’ introduction to special session on phonetic perception and word recognition

A classic statement of the speech perception problem is how to account for the perception of phonemes based on the acoustic stimulus. A classic statement of the word recognition problem is how to account for the access and selection of lexical representations, with little concern for phoneme/phonetic perception. Over the past two decades, research in spoken-word recognition has begun to explicitly model the stages of word recognition in terms of phonetic/phonemic perception and later stages involving lexical knowledge. These different models incorporate very specific proposals for the interaction of information from the bottom-up stimulus and the lexicon. They also raise issues concerning the appropriate model for phoneme perception, that is, for example, does it precede or follow the lexical processing stage? It would be convenient if theories/models of phoneme perception could ignore effects of the lexicon, and therefore, autonomously account for the ‘‘front-end’’ of word recognition. But can they, and be adequate theories/models of either phoneme perception or the front-end to word recognition? Alternatively, are there phoneme identification effects that are irrelevant to word recognition? In this session, researchers representing a variety of perspectives on the true relationship between lexical processes and phoneme/phonetic perception are brought together.

Lexical neighborhood and phoneme frequency effects in phonetic processing

The phonetic coding of speech reflects both a stimulus driven process and influences from the mental lexicon. There are at least three distinct forms of influence of the mental lexicon on phonetic processing: lexical status (is the item a word), lexical neighborhood (how many words is the item similar to), and phoneme frequency (how often does the phoneme occur in the language). Using natural speech series, the identification of an ambiguous phoneme is influenced by both the lexical neighborhood of the stimulus (syllable), and the probability of occurrence of the target phonemes in the language. Listeners consistently identify ambiguous phonemes with the label that corresponds to the end of the series with a larger lexical neighborhood, and listeners are biased to report more probable phonemes. Neighborhood and phoneme frequency effects are also found in a lexical decision task for clear, unambiguous tokens. Together with other work on higher-level influences on phoneme perception, these results indicate that phoneme perception is the result of an interaction between acoustic–phonetic information and the lexical knowledge of the individual, regardless of whether the target item is a word or not. [Work supported by NIDCD Grant No. R01 DC00219 to SUNY at Buffalo.]

Predicting the perception of words from the perception of phonemes

The present paper explores the degree to which the perception of CVC syllables can be factored into the perception of its constituent phonemes. Categorization experiments with synthetic speech typically manifest such factorability to a remarkable degree (regardless of lexical status of the syllables.) Such factorability is also compatible with intelligibility studies of English CVCs [ e.g., A. Boothroyd and S. Nittrouer, J. Acoust. Soc. Am. 84, 101–114 (1988)]. These show that the identification of nonsense syllables can be predicted as the product of the probabilities of identification of its phoneme parts, while word identification is systematically higher. Simulation studies are reported here involving a ‘‘factored’’ perceptual model that consists of a set of phoneme-likelihood estimators whose outputs are modulated by prior probabilities related to lexical status. This model can approximate the patterns observed in human perception. Simulations were also run with nonfactorable models, where syllables and words involve information about unique stimulus properties that cannot be predicted from their constituent phonemes. Consistent with a conjecture of Allen [J. Allen, IEEE Trans. Speech Audio Process. 2, 567–577 (1994)], such syllable template models do not produce behavior compatible with human performance. [Work supported by SSHRC.]

Follow-up Study of a Right- and a Left-Hemispherectomized Child: Implications for Localization and Impairment of Language in Children

Two hemispherectomized girls, one operated on the right, the other on the left, were followed from time of surgery until 9 and 10 years of age and compared with respect to course of language acquisition following surgery. At conclusion of follow-up, receptive and expressive language, phoneme perception and production, and sentence processing of the two hemispherectomized children were compared with those of two control groups of similar age, one developing language normally, the other language-impaired. The left-hemispherectomized child's abilities were similar to those of the language-impaired children; the right-hemispherectomized child's abilities resembled those of the language-normal children. Implications for localization of developmental anomalies in language-impaired children are discussed.

Similarity Mapping in Spoken Word Recognition

Five phoneme monitoring experiments are reported that investigate the relationship between the degree of lexical activation (based on similarity of the input to a real word) and phoneme perception. Experiment 1 showed that phoneme monitoring detection times increased as similarity of the carrier to a real word decreased. Experiment 2 replicated these results with a set of two syllable stimuli. Experiment 4 extended the results of Experiments 1 and 2 to a new phoneme. Two additional control experiments (Experiments 3 and 5) were conducted with truncated stimuli where lexical contributions were removed. The results are discussed in terms of the architectural relationship between the acoustic-phonetic input, form-based lexical levels, and semantic levels of representation.

Categorical perception of Japanese moraic phonemes

This study found that segment duration is perceived categorically in Japanese. Using a speech time-scale modification algorithm, PICOLA plus2, the nonword/kesonato/pronounced by a native Japanese male was manipulated to generate stimulus continua. The durations of /s/, /n/, /a/, and /t/ closure interval were manipulated in four continua. In the first experiment, three stimulus pairs, which had the same disparity in duration were selected from the short, long, and middle region of each continuum. Forty-eight randomized trials were presented to 20 native Japanese, who were asked individually to distinguish the difference of sounds by the ABX method. Pairs from the short and long regions of the continua were not accurately discriminated while pairs from the middle region were. In the second experiment, 44 stimuli were selected from every continuum. They were presented to the subjects for identification in Hiragana (Japanese characters). The second result demonstrated the existence of a moraic phoneme boundary paralleling the discrimination peak in the first experiment. This research concluded that native Japanese perceive the duration of relatively steady-static sounds categorically. And also, it is suggested that native Japanese utilize an automatic duration detection routine.

Variability in /s/ and /S−P174n/ productions within and across talkers

A primary issue in speech perception research is the apparent ‘‘lack of invariance’’ between the acoustic information in a signal and the listener’s phonemic perception. The same intended phoneme can be produced with a wide range of acoustic values, and different intended phonemes may be produced with the same values. Although the existence of this variability is not in question, the degree to which listeners have to account for it is unclear. Some studies have attempted to measure the extent of this variability, but few have done so for fricatives, making it difficult to determine the degree to which individuals need to normalize fricative input for the individual talker. The current experiment attempts to rectify this hole in the literature by examining the fricative centroids for over 100 utterances beginning with /s/ and /■/ from each of 20 different speakers. Although there was variability in the centroid values, the categories showed virtually no overlap within a talker. However, there was substantial overlap across talkers. This suggests that if listeners use centroids to cue the /s/–/■/ distinction, they would first need to normalize the signal for the individual talker’s frequency range. [Work supported by NIDCD Grant No. R01 DC00219 to SUNY at Buffalo.]

Lexical Inhibition and Attentional Allocation during Speech Perception: Evidence from Phoneme Monitoring

The TRACE model (McClelland & Elman, 1986) predicts lexically-driven inhibition at the phonemic level. This is due to the combination of top-down excitatory connections from the lexical to the phonemic level, and inhibitory connections between competing units within the phonemic level. Frauenfelder, Segui, and Dijkstra (1990, Experiment 2) tested this prediction in French and found no evidence of such inhibition. Experiment 1 of the current study replicated their results with English stimuli: Instead of having longer reaction times (RTs), targets in Inhibiting Nonwords (INWs) were detected just as fast as targets in control nonwords. Our Experiment 2 improved the design of the original experiment by adding a more appropriate control condition, increasing the number of critical items, and employing balanced target locations and conditional target probabilities. Under these conditions, RTs to INWs were significantlyfasterthan baseline RTs, an effect opposite in direction to the hypothesized inhibition. Experiment 3 used a dual-task paradigm to examine the attentional demands of processing different types of nonwords. In addition to performing the phoneme monitoring task as before, subjects also monitored a pure tone for frequency modulations. The RT advantage for INWs was replicated in this experiment for both phoneme and modulation targets. In Experiment 4 we replicated the INW advantage for both phoneme and modulation targets, and found that the advantage disappeared for stimuli that carriedbothtypes of targets. The results suggest that both lexical inhibition and attentional allocation affect phoneme perception; their interaction can mask the effect of each.

Perception of stop consonants in children with expressive and receptive-expressive language impairments.

The performance of 32 children with language impairment-11 with expressive language impairment only (LI-E subgroup) and 21 with both receptive and expressive language impairment (LI-ER subgroup)-and of 22 children without language impairment (LN subgroup) was examined in a study of perception and imitation of synthesized /ba/ and /da/ syllables. Formant transition duration and task difficulty were varied in the perceptual tasks. The LI-E children were able to identify the syllables as well as the LN; the LI-ER were not. Of the children who succeeded on an identification task and proceeded to a serial ordering task incorporating the same stimuli, the LI-E children were the least successful on the second task. The ability to label the stimuli perceptually was highly correlated with absence of speech articulation errors in the LI children and with performance on the imitation task in all subjects. The findings are examined in relation to the hypotheses that rapid-rate perceptual processing is the sole basis of language impairment in children and that, in these children, production skill may predict phoneme perception rather than the reverse.

Perception of the voicing contrast by Dutch children and adults

A phoneme identification experiment manipulating closure duration was carried out to investigate perception of the word-medial voicing contrast by Dutch four-year-old, six-year-old, and 12-year-old children, and Dutch adults. Pretest results show that all children responded correctly to real-word minimal pairs differing in stop voicing. Subsequently, subjects were tested in an identification task using nonsense words, in which closure duration for labial and alveolar stops was manipulated through waveform editing. The two older age groups consistently distinguished voiced and voiceless stops. The younger children displayed a relatively high percentage of ambiguous responses, indicating that they had difficulties with the categorization of voiced and voiceless stops. Results are discussed in relation to adult perception, the perception-production relationship, and early infant research. It is argued that phoneme perception in children from the age of four evolves synchronously with their phoneme production, and that the perceptual processes develop gradually with age.