JOURNAL

OF COMMCJNICATION

PITCH DEVIANCY

DISORDERS

AND INTELLIGIBILITY SPEECH

NANCY S. McGARR*

237

11 (1978), 237-241

OF DEAF

AND MARY JOE OSBERGER

The Graduate School and University Center, The City University of New York, 33 West 42nd Street, New York, New York 10036

The relationship between pitch deviancy and other aspects of speech production that affect intelligibility of deaf speech was investigated. The speech skills assessed included the following: ratings of pitch deviancy, ratings of overall intelligibility, production of the prosodic features-stress, intonation, and pause, and production of phonemes. Children who could not sustain phonation had speech that was consistently judged unintelligible. This group of children also had pure tone averages greater than 90 dB. For the remaining children, the relationship among pitch deviancy, intelligibility, and hearing level was variable. The highest intercorrelations were among prosodic feature production, phoneme production, and intelligibility.

Introduction Many investigators (Green, 1956; Angelocci et al., 1964; Martony, 1968) have noted that disorders of phonatory control are common in deaf speech. The pitch of deaf children is frequently reported as inappropriately high or low with respect to age and sex. In addition, the speech may be characterized by sudden pitch breaks or other evidence of inability to control voicing. Since stress and intonation are conveyed, at least in part, by pitch, and voicing information requires the control of laryngeal abduction and adduction, it is clear that problems of laryngeal control may have serious consequences for overall intelligibility. In addition, a grossly deviant pitch pattern may be sufficiently distracting in a communication situation to have indirect effects on intelligibility. The purpose of this article is to investigate the interrelationships of pitch deviancy and other factors that may influence intelligibility. Review of the Literature A relatively small amount of quantitative data is available on the effect that suprasegmental errors have on intelligibility. Levitt et al. (1972) noted that children with approximately the same frequency of segmental errors had speech intelligibility scores differing by as much as 30%. Errors (such as intermittent

* Address correspondence e Elsevier North-Holland,

to: Nancy McGarr, Inc , 1978

at the above address. 0021-9924/78/0011-0237$01.75

238

NANCY

S. MCGARR

and MARY

JOE OSBERGER

phonation, spasmodic variation in pitch and loudness, and excessive variability of loudness) correlated with differences in intelligibility. Other aspects of poor phonatory control present in deaf speech may include excessively high fundamental frequency (Green, 1956; Angelocci et al., 1964; Martony, 1968) and less than normal variation in fundamental frequency resuhing in flat and monotonous sounding speech (Voelker, 1938; Green, 1956; Hood, 1966; Smith, 1972). Prosodic features, such as stress and intonation, may be conveyed in part by changes in pitch. Several investigators (Stark and Levitt, 1974; Gold, 1975; McGarr, 1976) have shown that deaf children have difficulty in producing such features as stress, pausal juncture, and intonation. The contribution of deviant pitch patterns to errors in prosodic feature production, and the extent to which these errors reduce intelligibility, have not been considered. By far, the greatest amount of effort has been devoted to identifying segmental errors in deaf speech. In brief, error types include incorrect production of palatal and alveolar fricatives, affricates, and velar nasals. Errors of omission and substitution are frequent (Hudgins and Numbers, 1942; Smith, 1972). One common substitution, the voiced-voiceless error, is dependent on laryngeal control. The purpose of this article is to examine the effect of pitch deviancy on intelligibility and the relationship of poor phonatory control on segmental and suprasegmental errors in deaf children’s speech. Procedure Subjects The speech production of fifty-seven 1 I- to 12-year-old deaf children was analyzed. The study included 25 females and 32 males. These 57 children were part of a group of 125 children who participated in a longitudinal project designed to assess the language and communication skills of deaf children (Levitt et al., 1975). All of the children in this study were enrolled in state-supported schools for the deaf in New York State. The 57 children, who comprise the present study, were not preselected but were those children for whom the constraints of scheduling permitted completion of the test battery. Measurements

of Speech Production

Each child was required to describe two Speech production-pitch register. or more picture sequences. Recordings were made of the child’s speech, which was then rated by three listeners familiar with the speech of the deaf. The child’s pitch register rating was an average of ratings of three listeners who each rated several speech samples for a total of 6-9 observations for each child. The rating system for pitch register is part of the Speech and Voice Diagnostic Evaluation

PITCH DEVIANCY

AND INTELLIGIBILITY

239

Form developed at the National Technical Institute for the Deaf (Subtelny, 1975b). Table 1 shows the profile rating and functional descriptors for evaluating the child’s pitch register. Ratings 2, 3, and 4 show progressive degrees of pitch deviancy. For example, if a child was given a rating of 3 + , this would indicate that the child’s pitch was judged to be moderately inappropriate and too high. A child with a pitch that is moderately inappropriate and too low would be given a rating of 3 - . A test of prosodic feature production Speech production-prosodic features. was developed in order to examine production of three prosodic features of English. The general format was suggested by Bronstein’s (1960) description of American English. A speaker recorded sentences with contrasting stress, intonation, and juncture or pause. Extreme examples of each (in terms of acoustic cues) were chosen. Six sentences were constructed for each variable. Three were two syllables in length; three were five syllables in length. The sentences in the test of prosodic feature production are shown in Table 2A. Examples of these features and the methodology of representation for the children are shown in Table 2B. The feature of pause was indicated by three dots. The feature of stress was indicated by underlining and capitalization. The feature of intonation was indicated by punctuation. Each child practiced several nontest items in order to familiarize himself with the written cues for stress, pause, and intonation. When it was felt that the child understood the concept, he was asked to read the sentences and his utterances were recorded. Raters experienced in listening to the speech of the deaf evaluated the recordings. The listeners were asked to indicate whether the intended feature was correct, and if not, what error was made. Their choice of error types consisted of pause in the wrong location, stress in the wrong location, inappropriate intonation (e.g., a statement if a question was intended), staccato (equal stress and equal pause on all syllables), or unintelligible utterance. In addition to an analysis of error types, the average score in percent was computed for each child. An articulation test was used to Speech production-segmental features. measure the child’s production of vowel and consonant sounds in isolated words. The test words were selected from sentences developed by Smith (1972). The words were constructed to include all the vowels, as well as the consonants in the initial, medial, and final position of words. Each child read the list of test words. An examiner, familiar with the speech of the deaf, evaluated the child’s production as the test was administered. If the target phoneme was not produced correctly, the examiner indicated whether an error of omission, substitution, or distortion had occurred. If a substitution error occurred, the examiner indicated which phoneme was substituted for the target. The addition of an adventitious phoneme was also noted.

240

NANCY S. MCGARR

and MARY JOE OSBERGER

TABLE 1 Pitch Rating Profile rating

Functional

1. 2. 3. 4. 5.

Cannot sustain phonation Much above (+) or much below (-) optimal level Moderately above (+) or below (-) optimal level Slightly above (+) or below (-) optimal level Appropriate for age and sex

Sentences

descriptor

TABLE 2A for the Test of Prosodic

Feature Production

Come here. Oh boy. Thank you. He has one big dog. I want to see it. My new hat is blue.

TABLE 2B Sample of Test Sentences Come here. COME here. Come HERE. Come here?

Each of the speech samples that Speech production-overall intelligibility. had been recorded and rated for pitch register was subsequently rated for overall intelligibility. The intelligibility rating is comparable in form to the rating for pitch register. The rating scale was based on the Communication Profile, developed at the National Technical Institute for the Deaf (Johnson, 1975). Table 3 shows the profile rating with the functional descriptors for defining the child’s speech intelligibility. Following the procedure that had been established for rating pitch, the overall intelligibility rating was also an average of three listeners who each rated several speech samples for a total of 6-9 observations for each child. Audiological information. Pure tone thresholds and other related audiological data were obtained from each child’s school record. The pure tone average of 500, 1000, and 2000 Hz for the group of children was 98 dB in the better ear.

PITCH DEVIANCY

AND INTELLIGIBILITY

241

TABLE 3 Intelligibility Rating Profile rating

Functional

descriptor

1.

Speech cannot be understood

2.

Speech is very difficult ligible

3.

Speech is difficult stood

to understand-only

to understand;

isolated words or phrases are intel-

however,

4.

Speech is intelligible

with the exception

5.

Speech is completely

intelligible

the gist of the content can be under-

of a few words or phrases

Results Pitch Register A striking feature of the data on pitch register was the large number of children who received pitch ratings that were either appropriate for their age and sex (a rating of 5) or differed only slightly from optimal level (a rating of 4). Thirty-two of the children received ratings higher than 4.00. Figure 1 shows the distribution of scores for pitch register. The vertical axis shows the number of children in each group; the horizontal axis shows the average rating of pitch register of the 57 children. The children have been divided into two groups: those with pure tone averages less than or equal to 90 dB in the better ear (represented by the black bars), and those children with pure tone averages greater than 90 dB in the better ear (illustrated by the white bars). A level of 90 dB was chosen in order to conform with discussion of speech intelligibility data. Studies by Boothroyd (1970) and Smith (1972) have shown that hearing levels and speech intelligibility are correlated until below 90 dB, at which point speech intelligibility is uniformly low. Although Fig. 1 shows that the children with better hearing, on the average, received higher ratings than children with hearing loss greater than 90 dB, it would seem that good hearing is not of itself a prerequisite for developing appropriate pitch register. A large number of children received ratings of 4.0 or better. However, no child with good hearing had difficulty sustaining phonation and fewer children with good hearing received pitch register ratings that deviated from optimal level (either a rating of 2 or 3).

NANCY

242

S. MCGARR

and MARY

JOE OSBERGER

20-

&$ ,“g rn> =i! so

m

s 9OdB

0

j

pure

17

tone overage - 12

90dB

IO7

1.0

I1.99 AVERAGE

22.99 RATING FIG.

Prosodic

Feature

33.99 OF

PITCH

44.99

5

REGISTER

1

Production

The results of the Prosodic Feature Production test are shown in Table 4. Four children were not given this test. While the children with poorer hearing have a lower average score, the difference between the two groups is not great, and the range is large. The results suggest that the children were most successful at producing the intended feature of pause. The children did less well on production of stress. The children seemed to be slightly better in producing a stressed form that occurred near the end of a sentence than one that occurred near the beginning. By far the most difficult form was the question. Only a small proportion of the children were able to control the intonation contour sufficiently well enough to produce a recognizable question. The most common error type was production of the sentence with equal stress followed by equal pause for each syllable (i.e., staccato). Articulation

The results of the articulation test are also shown in Table 4. Eleven children were not given the test. Again, children with poor hearing have a lower average score, but overlap is substantial. Data obtained on the test of segmental features revealed similar articulatory patterns that have been reported for speech of deaf children (Hudgins and Numbers, 1942; Smith, 1972). The children were most successful in producing consonants with a place of articulation at the front of the mouth and least successful in producing consonants with a place of articulation either in the middle or back of the mouth, respectively. The most frequent error type was omission of the target phoneme, which occurred 29% of the time. Substitution of consonants occurred 17% of the time. A very common error was the voiced-voiceless confusion. Errors involving the distortion of a phoneme, or the addition of an adventitious phoneme, were

PITCH DEVIANCY

243

AND INTELLIGIBILITY

TABLE 4 Articulation

Prosody 90dB

14

18

39

> 90 dB 28

0.55

0.42

0.54

0.48

0.25-0.80

O-0.80

0.14-0.75

0.09-0.92

TABLE 5 Pitch Pitch

-

prosody

io.370

Intel1

+0.36O

Prosody i-o.31 u +0.64b

Intell.

Phoneme

Pure tone average

+0.36’

+0.29’

-0.22

+0.64b

+0.65”

-0.29”

-

+0.63’

-0.38O

Phoneme

+0.29’

+0.65 b

+0.63b

-

-0.43”

Pure tone average

-0.220

-0.29~

-0.38O

-0.430

-

“P < 0.005. bP < 0.001.

relatively infrequent. These error patterns are essentially the same as those obtained by Smith (1972) for deaf children’s articulation of sentences. Overall Intelligibility

Figure 2 shows the intelligibility ratings. The vertical axis shows the number of children in each group; the horizontal axis shows the rating of overall intelligibility. Here again, the children have been divided into two groups with respect to a hearing level of 90 dB. The black bars represent those children with pure tone averages of less than or equal to 90 dB in the better ear; the white bars represent those children with pure tone averages greater than 90 dB in the better ear. The data on overall speech intelligibility showed a large number of children who received very low ratings. Twenty-six of the children received a rating of 1.99 or less; that is, the children’s speech could not be understood by listeners familiar with the deaf. Only two of the children had speech that was rated completely intelligible. In general, Fig. 2 shows that those children with better hearing (i.e., hearing level 2 90 dB pure tone average in the better ear) achieved higher intelligibility

244

NANCY S. MCGARR

20-

and MARY JOE OSBERGER

m

$ 9OdB pure

0

‘9OdB

tone

overoge

15 kz

-

w

$2 :I’ 53= t*

-

IO-

IO

7

6

6

n 1.0

I-

1.99 AVERAGE

22.99 RATINGS

33.99 OF

44.99

5

INTELLIGIBILITY

FIG. 2

ratings. However, several children with relatively good hearing received poor intelligibility ratings. The data suggest that this group of children are capable of better speech production and that speech training is necessary to maximize their potential. Another aspect of the data showed that no child with a hearing loss greater than 90 dB achieved a rating of 5 (i.e., speech is completely intelligible). However, several children with poor hearing achieved ratings of 4 (i.e., speech is intelligible with the exception of a few words or phrases). In each of these cases, although the pure tone average was greater than 90 dB, the children had some measurable hearing through at least 1000 Hz. Pitch Deviancy

and Intelligibility

Figure 3 shows the relationship between pitch register ratings and intelligibility. Ratings of pitch deviancy are shown on the horizontal axis; ratings of intelligibility are shown on the vertical axis. Each of the three listeners’ ratings for the 57 children in the study are contained in the plot. The number of observations for pitch register ratings and intelligibility ratings is represented by the width of each box. The children who were given pitch ratings of 1 represent the subgroup who cannot sustain phonation. These children have evidence of pitch breaks or large fluctuations in pitch. Because of these characteristics, these children have been separated from the rest of the population on the graph. This group of children had intelligibility ratings that were very low. In fact, the highest intelligibility rating given was a 2 (i.e., only isolated words or phrases were understood). Children who were given pitch ratings of 2 through 5 showed a somewhat different pattern with respect to intelligibility. The data for this group suggest the following.

PITCH DEVIANCY

I-

AND INTELLIGIBILITY

n

I

PITCH

n

245

I

RATING

FIG. 3

First, there appears to be two groups of children who received a pitch rating of 5. A pitch rating of 5 indicates that the pitch is judged appropriate for the child’s age and sex. Within the group, speech is judged intelligible (rating of 5) for some of the children. For other children, speech was completely unintelligible (rating of 1). Second, for children with pitch ratings of 2, 3, or less, the grouping was pronounced. There was an overall trend for intelligibility to decrease as the degree of rated pitch deviancy increased. Although there are more children whose pitch was considered too high than too low, there is no evidence that the direction of the deviancy in pitch affects intelligibility. These findings suggest that for the majority of the children, there appears to be no simple relationship between pitch deviancy, as rated here, and intelligibility of speech. However, in children who are unable to sustain phonation, and who have pitch breaks, speech is consistently judged to be unintelligible. The effect of faulty pitch control in reducing the intelligibility has also been noted by Subtelny (1975b) in working with a population of deaf adults. Correlations The correlations of all factors with pitch deviancy are significant, but not particularly high (Table 5). The correlations between pitch deviancy and prosodic feature production were highest as might be expected. The correlation coefficients show two other interesting areas. First, high correlations were obtained for factors related to speech production sills, namely, intelligibility and the production of prosodic features (0.64), and the production of phonemic features and prosodic features (0.65). A predictor of intelligibility would be the number of articulatory errors and the ability to produce prosodic features. Second, the pure tone average is significantly correlated with all factors but correlations are higher with intelligibility and phoneme production than with pitch and prosody.

246

NANCY S. MCGARR

and MARY JOE OSBERGER

Discussion Overall, there is a relationship between poor phonatory control, on the one hand, and hearing level and intelligibility, on the other. However, this relationship is more complex than some general comments in the literature might lead us to suspect. On the one hand, the group of children who cannot sustain phonation are all in the group with greater than 90 dB loss and are all very unintelligible. On the other hand, children with simple deviant overall pitch levels are quite variable with respect both to hearing loss and to intelligibility. Looking at the data in terms of the pattern of correlations, we see that pitch rating has significant correlations with all other variables; however, the strongest intercorrelations are among prosody, intelligibility, and phoneme production. This leads to a belief that even a severely deviant pitch level may coexist with intelligible speech, and that control of prosody must depend in part on factors other than pitch level control. A final comment should be made with reference to the training of the children in the study. Some children with relatively good hearing had relatively poor pitch register ratings and poor ratings of overall intelligibility. Since none of these children had other known handicaps, which would interfere with speech development, it would seem that their potential is not being fully realized. We would like to acknowledge the assistance of our colleagues Harry Levitt, Katherine S. Harris, Harvey Stromberg, and Toni Gold in preparing this study. The research has been generously supported through grants from the Committee on Cooperative Research Endeavors in Education of the Deaf (CREED), New York State Education Department, Bureau for Special School Supervision, Ofice for Education of Children with Handicap Conditions, and the National Institute of Health (#NH 09252). References Angelocci, A., Kopp, G., Holbrook, A. The vowel formants of deaf and normal hearing eleven to fourteen year old boys. J, Speech Hear. Dis., 1964, 29, 156170. Boothroyd, A. Distribution of hearing levels in the student population of Clarke School for the Deaf. (SARP Report #3). Northampton: Clarke School for the Deaf, 1970. Bronstein, A. J. The pronunciation of American English. New York: Appleton-Century-Crofts, 1960. Gold, T. Perception and production of prosodic features by hearing impaired children. Paper presented at the Convention of the American Speech and Hearing Association, Washington, D. C., November 1975. Green, D. S. Fundamental frequency of the speech of profoundly deaf individuals. Ph.D. thesis, Purdue University, 1956. Hood, R. B. Some physical concomitants of the perception of speech rhythm of the deaf. Ph.D. thesis, Stanford University, 1966.

PITCH DEVIANCY

AND INTELLIGIBILITY

247

Hudgins, C. V., and Numbers,G. C. An investigation of the intelligibility of the speech of the deaf. Genet. Psychol. Monogr.. 1942, 25, 28%392. Johnson, D. D. Communication characteristics of NTID students. J. Acad. Rehab. Audiol., 1975, 8 (1, 2). 17-32. Levitt, H., Smith, C. R., Stromberg, H. Acoustic, articulatory and perceptual characteristics of the speech of,deaf children. Speech Communication Seminar, Stockholm, 1974. Levitt, H., Stark, R., McGarr, N. S., Carp, J., Stromberg, H., Gaffney; R. S., Barry, C., Velez, A., Osberger, M. J., Leiter, E., Freeman, L. Language and communication sills of deaf children: 197>1975, unpublished report. McGarr, N. S. The production and reception of prosodic features. Paper presented at the annual meeting of the A. G. Bell Association, Boston, Mass., June 1976. Markides, A. The speech of deaf and partially hearing children with special reference to factors affecting intelligibility. Br. J. Dis. Commun., 1970, 5, 126140. Martony, J. On correction of voice pitch level for severely hard of hearing subjects. Am. Am. Deaf, 1968, 113, 195202. Nober, E. H. Articulation of the deaf. Except. Child., 1967, 33, 611-621. Smith, C. R. Residual hearing and speech production in deaf children. Ph.D. thesis, City Univ. of New York, 1972. Stark, R. E. and Levitt, H. Prosodic feature perception and production in deaf children. J. Acous. Sot. Amer., 1974, 55, (A). Subtelny, J.D. An overview of the communication skills of NTID students with implications for planning of rehabilitation. J. Acad. Rehab. Audiol.. 1975a, 8 (l&2), 35-50. Subtelny, J. D. Speech assessment of the deaf adult. .I. Acad. Rehab. Audiol., 1975b, 8 (l&2), 11&116. Voelker, C. An experimental study of the comparative rate of utterance of deaf and normal hearing speakers. Amer. Ann. Deaf,, 1938, 83, 274284.

Pitch deviancy and intelligibility of deaf speech.

JOURNAL OF COMMCJNICATION PITCH DEVIANCY DISORDERS AND INTELLIGIBILITY SPEECH NANCY S. McGARR* 237 11 (1978), 237-241 OF DEAF AND MARY JOE OS...
622KB Sizes 0 Downloads 0 Views