Medial Consonants Research Articles

An acoustic analysis of the bidirectionality of coarticulation in VCV utterances

This study investigated the extent of vowel-to-vowel anticipatory and carry-over coarticulation in VCV utterances. Five speakers of American English produced open (V.CV) and closed (VC.V) syllable shapes with /i/, /ɔ/, /e/ and /u/ as “fixed vowel” contexts and /i/ and /ɔ/ as a “changing vowel” context. Medial consonants were /bdgptk/. Anticipatory coarticulation was indexed by measuring the difference in F 2 offset frequencies for V 1, with a given fixed vowel context, when V 2=/i/ vs. /ɔ/ (e.g., [ibi] vs. [ibɔ]); carry-over coarticulation was indexed by measuring the difference in F 2 onset frequencies for V 2, with a given fixed vowel context, when V 1=/i/ vs. /ɔ/ (e.g., [ibi] vs. [ɔbi]). In open syllables, anticipatory exceeded carry-over coarticulation for labial and velar stops, but carry-over effects predominated in alveolar contexts. Alveolars maximally resist anticipatory effects of V 2 and thus may allow greater V 1 carry-over effects. In closed syllables, carry-over exceeded anticipatory effects across all stops. This was attributed to a short V-to-V temporal interval indexing carry-over effects relative to a disproportionately longer interval separating vowels for anticipatory effects.

Phoneme monitoring in internal and external speech

Four experiments examine the time course of phoneme monitoring in internally and externally generated speech. The aim of this research was to replicate and extend previous findings of Wheeldon and Levelt (1995), who required their Dutch participants to monitor their own prearticulatory speech in order to investigate the generation of an abstract phonological code. Experiment 1 required British participants to silently generate English words with a CVCCVC structure and to monitor these words for the four constituent consonants. Similar to the Dutch study, a clear left-to-right monitoring effect was observed and a significant difference was seen between latencies for the consonants separated by the initial vowel and for the consecutive word medial consonants. Internal speech monitoring was also found to speed up across the word. A perception version of the task (Experiment 2) yielded a different pattern of monitoring latencies. In Experiments 3 and 4, all phonemes in the initial CVC syllable of bisyllabic words were monitored in internal and external speech respectively. A significant left-to-right pattern of monitoring latencies was again observed in the internal speech task. However, in the external speech task, monitoring latencies for the vowel and final consonant of the syllable did not differ. The conclusions of Wheeldon and Levelt are discussed and refined in light of the present results.

Segmental and coarticulatory timing in Ikalanga prenasalized stops

The temporal characteristics of prenasalized stops in Ikalanga, a Bantu language of Botswana, are investigated. Eight speakers were recorded producing Ikalanga words with VNCV (NC=/mb nd ŋg), VNV (N=/mnŋ), and VCV (C=/b d g/) sequences. Acoustic measures of 3 speakers’ productions showed that oral closure durations (from vowel offset to closure release) were only slightly longer for NC than for N or C (94, 83, and 83 ms, respectively). When analyzed separately, the N and C portions of VNCV were somewhat shorter than their counterparts in VNV and VCV, but nearly all of the shortening occurred in the C portion (in VNCV, N=78 ms and C closure=16 ms). The durations of vowels preceding the medial consonants did not differ across the three consonant types. Thus, for segmental durations, the main difference between Ikalanga VNCV and VNV was the presence of a brief oral occlusion in the former. However, measures of coarticulatory nasalization point to another systematic difference: while both sequences exhibited comparable anticipatory vowel nasalization, VNV (but not VNCV) sequences showed carryover nasalization extending through more than 70% of the post-nasal vowel. Findings will be compared with previous work on Bantu prenasalized stops. [Work supported by NSF.]

Preliminary Results with the AVR ImpaCt Frequency-Transposing Hearing Aid

Currently, the only behind-the-ear hearing aid that provides a frequency transposition function is the ImpaCt DSR675, recently introduced by AVR Communications Ltd. of Israel. In tests with three hearing-impaired adults, the performance of the ImpaCt aid(s) was compared with that of each subject's own (nontransposing) hearing aids. Recognition of monosyllabic words and medial consonants did not differ significantly between the two types of aids. This suggests that the transposition function of the ImpaCt was not effective at providing these subjects with increased high-frequency speech information, at least for the programmable parameters applied in the experiments. However, the subjects' understanding of sentences in a competing noise was significantly poorer with the ImpaCt than with the subjects' own aids. In that test, the ImpaCt aids were programmed to attenuate parts of the noise. The decreased sentence recognition may have resulted from this program, which effectively reduced the bandwidth of the ImpaCt aids.

The acquisition of Arabic consonants.

This normative study of the acquisition of consonants of Arabic as spoken in Jordan answered 4 questions: (1) What percentage of children at each of 9 age levels produced each consonant correctly? (2) What are the ages of customary production, mastery, and acquisition for each phoneme? (3) Does accuracy of consonants within sound classes vary by position in the word? (4) What are the differences in ages of acquisition between Arabic and English? Samples were collected from 180 normally developing children between the ages of 2:0 and 6:4. The percentages of accuracy of both standard and acceptable consonants were plotted and showed clear developmental trends. Medial consonants were significantly more accurate than initial and final consonants. The ages of customary production, acquisition, and mastery of Arabic consonants were similar to those for English but with notable exceptions that have implications for description of phonological acquisition. Support for previously proposed universal sound acquisition sequences was found, but some language-specific effects were also seen.

Auditory-visual speech recognition by hearing-impaired subjects: consonant recognition, sentence recognition, and auditory-visual integration.

Factors leading to variability in auditory-visual (AV) speech recognition include the subject's ability to extract auditory (A) and visual (V) signal-related cues, the integration of A and V cues, and the use of phonological, syntactic, and semantic context. In this study, measures of A, V, and AV recognition of medial consonants in isolated nonsense syllables and of words in sentences were obtained in a group of 29 hearing-impaired subjects. The test materials were presented in a background of speech-shaped noise at 0-dB signal-to-noise ratio. Most subjects achieved substantial AV benefit for both sets of materials relative to A-alone recognition performance. However, there was considerable variability in AV speech recognition both in terms of the overall recognition score achieved and in the amount of audiovisual gain. To account for this variability, consonant confusions were analyzed in terms of phonetic features to determine the degree of redundancy between A and V sources of information. In addition, a measure of integration ability was derived for each subject using recently developed models of AV integration. The results indicated that (1) AV feature reception was determined primarily by visual place cues and auditory voicing + manner cues, (2) the ability to integrate A and V consonant cues varied significantly across subjects, with better integrators achieving more AV benefit, and (3) significant intra-modality correlations were found between consonant measures and sentence measures, with AV consonant scores accounting for approximately 54% of the variability observed for AV sentence recognition. Integration modeling results suggested that speechreading and AV integration training could be useful for some individuals, potentially providing as much as 26% improvement in AV consonant recognition.

Speech recognition as a function of the number of electrodes used in the SPEAK cochlear implant speech processor.

Speech recognition was measured in listeners with the Nucleus-22 SPEAK speech processing strategy as a function of the number of electrodes. Speech stimuli were analyzed into 20 frequency bands and processed according to the usual SPEAK processing strategy. In the normal clinical processor each electrode is assigned to represent the output of one filter. To create reduced-electrode processors the output of several adjacent filters were directed to a single electrode, resulting in processors with 1, 2, 4, 7, 10, and 20 electrodes. The overall spectral bandwidth was preserved, but the number of active electrodes was progressively reduced. After a 2-day period of adjustment to each processor, speech recognition performance was measured on medial consonants, vowels, monosyllabic words, and sentences. Performance with a single electrode processor was poor in all listeners, and average performance increased dramatically on all test materials as the number of electrodes was increased from 1 to 4. No differences in average performance were observed on any test in the 7-, 10-, and 20-electrode conditions. On sentence and consonant tests there was no difference between average performance with the 4-electrode and 20-electrode processors. This pattern of results suggests that cochlear implant listeners are not able to make full use of the spectral information on all 20 electrodes. Further research is necessary to understand the reasons for this limitation and to understand how to increase the amount of spectral information in speech received by implanted listeners.

Acoustic and kinematic evidence for difference in language timing structure

Proposed differences in the timing organization of Italian and Finnish are examined from the perspectives of simultaneous acoustic duration and bilabial movement (selspot tracked) collected from four native speakers of each language. All utterances were disyllabic, shaped CVC(C)V for Italian and CV(V)C(C)V(V) for Finnish. Italian, generally believed to be syllable-timed, shows regular acoustic intervals between vowels implemented by shortening of the initial vowel in closed syllables. Movement data shows that shortening is achieved by early initiation of clusters and geminate consonants, truncating the initial vowel. Finnish, generally described as mora-timed, shows increments of acoustic duration of disyllabic utterances with the addition of moras, either to vowels or to medial consonants. An apparent constraint on syllable duration leads to acoustic shortening of both initial vowels and intervocalic consonants in the longest utterances. Movement data show that increases in quantity affect both medial consonant and preceding vowel portions of utterances. It is proposed that syllable-timing works directly from vowel to vowel, while mora-timing allows increase in duration of disyllabic utterances, adjusting both vowel and consonant durations to fill out or limit the possible intervals between vowels. [Work supported by NIH Grant DC 00121 to Haskins Laboratories.]

Variable word initial and medial consonant relationships in children's speech sound articulation.

The heterogeneous nature of consonants in the medial position of words may have prevented earlier studies from depicting a clear phonetic relationship between them and consonants in the initial and final positions. In this study, auditory perceptual judgments of children's consonant productions in the initial position of words were compared with those on medial consonants varying in number and order. Neither the perceived frequency of correct articulation nor the error pattern differentiated single consonants in the initial and medial positions of words at any age. Performance on consonant blends in the initial position of words differed from that on both types of medial sequences, but it was more similar to medial consonant sequences that corresponded with permissible word initial blends than to those that did not. It was concluded that initial and medial consonants may be regarded as performance equivalents only under certain conditions.

Perceptual features for normal‐hearing listeners' confusions of initial and medici consonants in a reverberant lecture hall

Twelve normal‐hearing listeners responded to a nonsense syllabic test consisting of 25 consonant‐vowel‐consonant‐vowel (CVCV) items while seated at a 0‐deg azimuth to and at three distances (i.e., 1 m, 4 m, and 8 m) from a loudspeaker within a 415‐m3 lecture hall with T = 1.9 s, S/N > + 17 dB, and an estimated critical distance of 3 m. Consonant scores were significantly better within the critical distance (1 m) than at distances beyond the critical distance (4 and 8 m), which were not significantly different from each other. Subjects' confusions of initial and medial consonants were submitted to a sequential information analysis (SINFA) to document a priori perceptual features. More information was transmitted within than beyond the critical distance for both consonant positions. However, more information was transmitted for initial consonants than for medial consonants. Further, the relative effects of masking by reverberation on perceptual features was more evident for medial consonants than for initial consonants. Beyond the critical distance, high‐frequency features (i.e., sibilancy) were transmitted somewhat better than those based more on low‐frequency energy (i.e., voicing). [Work supported by UC General Research Grant ♯8587529‐19900‐7.]

An articulatory study of consonant‐vowel overlap

Fowler [Phonetica (1981); J. Exp. Psychol. (1983)] has proposed that speech production is a series of global articulatory movements associated with the vowels of an utterance, upon which smaller, consonantal movements are superimposed. A possible articulatory test of this theory models consonants and vowels as alternations between a relatively closed and open vocal tract. Using the Wisconsin x‐ray microbeam system, articulatory records were obtained of English phrases with varied accent placement and alveolar consonant sequences of varied duration. Analysis of one speaker's productions is now complete. When the data are pooled over all conditions, the total duration of the bisyllabic sequence and the time difference between sonority peaks is strongly related to the length of the medial consonants; there is no evidence for articulatory overlap—i.e., vowel shortening in the neighborhood of long consonants or shifting of sonority peaks toward longer consonants. However, correlations within accent types uncover a strong shortening effect for accented syllables preceding long consonants, along with a reduced displacement in the following vowel. Thus any comprehensive overlapping consonant‐vowel theory of speech production must take into account the general prosodic structure of the utterance. [Work supported by the National Science Foundation under grant No. IRI‐8617852 to Mary Beckman.]

A cross-linguistic contrast in consonant and vowel timing

Two opposing explanations of the timing of consonants and vowels are that (1) they are timed relative to each other, or (2) vowels are produced in a continuous rhythm with consonants superimposed on them [Fowler, J. Exp. Psychol. 112, 386–412 (1983)]. Languages with different temporal properties, such as Japanese, a mora-timed language, and Italian, a syllable-timed language, might be expected to exemplify these different relationships. Speakers of each of these languages were recorded producing pairs of disyllabic words with medial consonants of contrasting length (single or geminate). Evidence from measures of acoustic duration and vowel-to-vowel coarticulation, shown by contextual changes in F2, suggest that Fowler's model (2) predicts the timing patterns found in Italian, while the other model (1) predicts the behavior of Japanese. [Work supported by NSF.]

Durational changes of apraxic speakers

Durations of phrases, words, and phones were obtained for apraxic and normal speakers during normal and fast speaking conditions. Utterances were a word (permit) and a nonsense disyllable (perpit) embedded in a carrier phrase. Each test utterance was spoken with two syllabic stress patterns. Durations of the phrase, target word, / ɑ ̢ / (/ ə ̢ /), /I/, and the medial consonant of the target word were measured from oscillograms. Absolute segment durations were longer for apraxic than for normal speakers. During the fast rate condition, normal subjects decreased the duration of three of the four words. There was no evidence that apraxic speakers changed their speaking rates between the two rate conditions. Further investigation of temporal control and segment duration is suggested.

The perception of plosive consonants by hearing‐impaired listeners

The reduced speech perception by hearing‐impaired subjects has often been studied. In order to investigate the role of coarticulation we did some experiments on a group of 20 hearing‐impaired and 10 normal‐hearing subjects. From 45 embedded CVCVC words we deleted either the plosive bursts, or the vocalic CV or VC transitions, or the stationary vowel portions, or any combination of these four. The listeners were asked to identify the three plosive consonants in the resulting stimuli. The results for the normal‐hearing subjects agree with results of Schouten and Pols [Proc. 10th ICPS, Utrecht (1983), p. 464]: identification is affected more by deletion of plosive bursts than of vocalic transitions, and in medial consonants the CV transition contains more information about the consonant than does the VC transition. For the hearing‐impaired listeners the vocalic transitions play a more important role: identification with the plosive bursts only results in guessing; in contrast with the results for the normal‐hearing subjects more errors occur for the final consonant than for the initial or the medial consonant. We also found some relations with other auditory functions measured. [Work supported by ZWO.]

Perception of intervocalic stop consonants: the contributions of closure duration and formant transitions.

Acoustic analyses of vowel-consonant-vowel (VCV) utterances indicate that they generally include formant transitions from the first vowel into a period of closure (VC transitions), and transitions out of the closure into the second vowel (CV transitions). Three experiments investigated the perceptual importance of the VC transitions, the CV transitions, and the closure period in identification of medial stop consonants varying in place of articulation. Experiment 1 compared identification of members of synthetic VC and CV continua with those from VCV series made by concatenating corresponding VC and CV stimuli using various closure durations. Experiment 2 examined identification of VCV stimuli constructed with only VC, only CV, or both VC and CV transitions; again closure duration was systematically varied. Experiment 3 correlated CV and VC identification with identification of VCV stimuli. Neither closure duration nor formant transition structure (i.e., only VC, only CV, or both) had an independent effect on identification. Instead, the formant structure and closure duration together strongly affected stop identification. When both VC and CV transitions were present, the CV transitions contributed somewhat more to identification of medial stops with short closures, than the VC transitions did. With longer closure durations, neither set of transitions appeared to determine perceived place of articulation in any simple way. Overall, the data indicate that the perception of a medial consonant is more than simply a (weighted) sum of its parts.

Children's Articulation of Medial Consonant Clusters: Implications for Syllabification

Among solutions to syllabification of medial clusters are those which yield syllable assignment for medial consonants that conform to constraints on word initial and final segments. In this study, selected speaker/listener performances on medial clusters were compared with response trends for word initial/final consonants. Thirty preschool children (ten at each of three articulation proficiency levels on /s/ based on deep test scores) imitated 40 nonsense bisyllables of equal stress. The stimuli included intervocalic unclustered /s/, as well as consonant clusters that conformed to word initial /s/ clusters and those that did not. Syllable assignment was imposed on the medial consonants using boundary insertion rules proposed by Hooper (1972, 1973). Performance differences between medial clusters with assigned tautosyllabic and heterosyllabic features in both the frequency and pattern of articulation errors appeared to reflect constraints on word initial/final segments. The results suggest that medial contiguous consonants that conform to word initial clusters and those that do not should be differentially treated in syllable boundary assignment.

How the components of speaking rate influence perception of phonetic segments.

In two studies we investigated the way in which the components of speaking rate, articulation rate and pause rate, combine to influence processing of the silence-duration cue for the voicing distinction in medial stop consonants. First, we replicated the finding that the articulation rate of a carrier sentence, that is, the rate at which the speech itself is produced, influences how the duration information is used to assign voicing values. Second, and more importantly, the assignment of voicing values was also influenced by the pause rate of the sentence. Thus, the listener adjusts for both articulation rate and pause rate when processing the phonetically relevant information. Finally, the two rate components did not function in an equivalent manner, since changes in articulation rate had considerably more effect on phonetic judgments than did changes in pause rate. Alternative explanations fo the relative weighting of the two variables are discussed.

Enkele Resultaten Van Een Kontrastief Experimenteel Onderzoek Naar Uitspraakgedrag In De Vreemde Taal Voor Het Nederlands En Het Duits

The point of departure for this investigation were some generally accepted impressionistic statements about durational differences in Dutch and German vowels. An acoustic analysis of utterances spoken by various groups of speakers was carried out. Differences between native speakers and second-language learners were manifest to a varying degree. Long training does not guarantee the best approximation to native performance. Furthermore it is apparent that current ideas about vowel duration in the two languages should be modified: the differences are not so great as generally thought. Phonological factors rather than specific durations are relevant. Unlike Dutch, German vowels are considerably longer before voiced than before voiceless medial consonants.

The Contextual Effect of Intermediate Consonants of the Perception of Final Plosives

Context dependency in the perception of final Hindi plosive consonants and distinctive features have been investigated as a function of the nature of adjacent sounds. VCC utterances comprising of 3 short vowels (/I/,/A/,/U/), two medial consonants (/s/,/I/) and 16 plosive consonants in the final position were recorded by 6 male speakers and heard by 10 listeners. Confusion matrices for final consonants were made separately for each medial consonantal context and analysed according to various feature systems using co-variance method of information transmission. It bas been found that voicing and aspiration are very much affected by the nature of preceding consonants. The results have been compared with earlier observations and discussed.

Context-determined effects of varying closure duration

Much evidence exists to show that /p/ is produced with longer closure than /b/, when they occur in words like rabid and rapid. If /b/ closures are silenced and lengthened, listeners tend to report /p/. There is the temptation to draw from this the generalization that closure duration serves to distinguish /ptk/ from /bdg/, i.e., that it is a cue to voicing. There is a context, however, in which the manipulation of closure duration has a rather different phonetic outcome; where in fact a longer duration is reported as /d/ and a shorter as /t/. The word pair center-sender, where the medial consonants are described for many varieties of American English as a nasalized flap in the one case and as a sequence of nasal plus stop in the other, is such a context: shortening the oral voiced closure of sender yields center when the closure is reduced to 40 ms or less. [The support of the National Institute of Child Health and Human Development is gratefully acknowledged.]