Journal o f Psycholinguistic Research, VoL 7, No. 3, 1978
Developmental Features of Speech-Sound Production in Language-Impaired Children F. A f f o l t e r , 1 R. B r u b a k e r , 2 a n d W. F r a n k l i n 3 Received August 20, 1976
The speech-sound production o f severely language-impaired children was monitored in a longitudinal study. In order to generalize findings, acoustical, phonemic, and clinical observation data were collecte d from 30 children. Results showed that speech-sound acquisition goes through a hierarchical sequence o f development and that speech-sound production will deteriorate in a predictable manner depending on the pathology. The authors hypothesize a speech-sound acquisition model, SSAM, based on the development observed.
INTRODUCTION In the normal child, development follows a hierarchical sequence of levels. In pathology, development is blocked at one of these levels. Levels 1 and 2 may be developed until a primary disturbance affects performances of level 3. As a consequence, levels 3 and 4 do not develop. Recovery, if it occurs, will follow the same normal hierarchical sequence-level 3 and then level 4. Language is characteristic of a higher stage of development than that of some perceptual mechanisms. With reference to the concept of hierarchical dependency, it can be said that perceptual mechanisms are prerequisite for language acquisition. Clinical case studies based on extended performance This-project was supported by the Swiss National Foundation for Scientific ResearchProjects No. 3.237.69, 3.448,70, and 3.902.72-and by North Carolina State University Research and Development Grant 056. 1Katholische Studentenhans, St. GaUen, Switzerland. 2Department of Speech Communication, Pennsylvania State University, University Park, Pennsylvania 16802. 3Department of Speech Communication, North Carolina State University, Raleigh, North Carolina. 213 0090-6905/78[0500-0213505.00[0 9 1978 Plenum Publishing Corporation
214
Affoltet, Brubaker, and Franklin
profiles (Affolter et al., 1974) showed that children with language deficits also showed deficits in perceptual performances: modality-specific, intermodal, and serial problems. These findings suggest that critical degrees of modalityspecific, of intermodal, and of serial integration are prerequisites for language acquisition. We will not attempt a lengthy discussion here elaborating the differences of two clinical subgroups: the serial group and the intermodality group. The serial group includes children with difficulty in integrating a critical amount of successive stimuli in all three modalities. The intermodality group consists of children characterized by difficulties relating and integrating a critical amount of information of different modalities (auditory, visual, and tactile-kinesthetic). The most severely language-impaired children in this group had been diagnosed as autistic by child psychiatrists. In the following study, speech-sound production of children with serial and intermodal problems was analyzed to investigate several aspects of the language-prerequisites interpretation suggested above. Specifically, it was hypothesized that (1)speech-sound acquisition goes through a hierarchical sequence of development, and in pathology and recovery the sequence of these levels will be the same, and (2)the production of speech sounds once acquired will not always be adequate in children with language problems. Under stress, the production will deteriorate in a predictable manner, with the serial group deteriorating differently from the intermodal group.
METHOD
Subjects Subjects 1971 Initially, in 1971, this research began with 30 children with problems of language development (called the clinical group) and 30 normal children of the same age, 3-10 years old. Because of significant differences of performance profiles within the clinical group, the clinical group was sub-divided into serial and intermodality groups. Serial group 1971 (18 in number, 12 testables on speech) consisted of children who demonstrated disturbances in serial integrations. Intermodality group 1971 (12 in number, three testables on speech) consisted of children who had difficulty in connecting different modalities.
Speech-Sound in Language-ImpairedChildren
215
Subjects 1972 In 1972, another clinical group of 18 children entered the study. This clinical group was subdivided into a serial group 1972 (seven subjects, three testables on speech) and an intermodatity group 1972 (11 subjects, five testables on speech).
Subjects 1971 in 1972 In 1972, five subjects of the original serial group 1971 were available for retesting, all five testable on speech. Intermodality group 1971 children available for retesting were six in number, four of whom were testable in speech-sound production. Procedures The children in the clinical groups were tested for speech-sound production. Observations were made by recording on video or acoustic tape and by written transcription during therapy sessions. Speech-sound production was analyzed in three fashions. Clinical observations within the clinical setting was one of the methods employed in exploring speech-sound production. A phonemic analysis was also undertaken and then selected samples were examined spectrographically.
Spectrographic Analysis Speech-sound production was studied by means of a Kay Electric model 6061B spectrum analyzer. The sonagraph (spectrograph) machine converts spoken sounds into a visual graph (a spectrogram) of frequencies and intensities as a function of time. In general, frequencies above 3000 Hz are associated with consonants, while vowels appear as bands of energy (formants) below 3000 Hz. From the spectrogram, which covers a range of 100-8000 Hz, a measurement of time can be made, so that the rate of speaking can be calculated. A plug-in component, Kay model 6076C scale magnifier, permits the expansion of selected portions of the voice spectrum so that individual harmonics can be identified. From this information, the change in vocal frequency over time can be seen and the intonation contour of the utterance observed. Some of the subjects ha 1971 were recorded live. All others were recorded first on audio tape from which selected samples of speech were used
216
Affolter, Brubaker, and Franklin
for spectrographic anal3/sis. This procedure was necessary because these children did not perform as normal children might. It was extremely difficult to have the children do as one would like. Most of the children had receptive language and could be told to do something. However, whether or not they would respond or were capable of performing on request was doubtful. It was decided to record them in the laboratory or clinical situation. Selected samplings of speech-sound production were taken from the tape recordings of each child and analyzed spectrographically. In most cases, the spectrographic analysis was made from tape recordings. Such spectrograms included any distortions inherent in the tape-recorder system and must be interpreted with caution. While the tape recorder was of sufficient quality for these purposes, no reproduction is as good as the original version. More serious are the recording problems arising in the clinical setting. For example, there is noise of the child's activity, of voices of parents and therapist, and of normal hospital activity. Another problem is that of changing microphone-to-subject distance. It can be shown that increasing this distance will attenuate the high frequencies displayed on the spectrogram. Inasmuch as high-frequency energy is associated with clearly articulated speech, there is a risk of significant error when evaluating articulation on spectrographic evidence alone. All one can do is evaluate a child's articulation, independently of the acoustic records, before and during recording. The spectrographic analysis consisted of five tests of speech-sound production. The five test results were judged to be either "normal" or "inappropriate." The five items of speech-sound production evaluated spectrographically were the following: 1. Consonants: evidence of frequency energy above 3000 Hz. Inasmuch as high-frequency energy is associated with clearly articulated speech, the researchers looked for evidence of energy above 3000 Hz. These high frequencies accompany~ stop and fricative consonant production. The absence of high-frequency energy, in part at least, is attributable to the absence of clear consonant articulation. 2. Vowels: evidence of clear vowel formants, steady after onset, without presence of noise. In the normal speech of adult males, clear vowels show formants which are distinct and smooth in appearance. However, the bandwidth (300 Hz) of the Kay Electric analyzing filter is not ideal for higher-pitched voices. As a result, formant clarity is often reduced in the spectrograms of women and children. For this reason, the observations concerning formant clarity are to be interpreted with caution.
Speech-Sound in Language-lmpaired Children
217
Some children with speech difficulty, although sounding as if they are producing some vowellike phonemes, do not give evidence of same when analyzed spectrographically. Reduced contrast of the formants can result from noise in the vocalic segment and is often caused by voice defects. 3. Intonation: variation in fundamental frequency as a function of time over an utterance. The Kay Electric model 6061B was equipped with a plug-in unit, the 6076C scale magnifier. The magnifying circuit of the sonagraph was used to make an expanded spectrogram of the lower 2000-Hz portion of the child's utterance. Expanded spectrograms show the individual harmonics of the voice. These enable one to determine the intonation contour, as the variation in vocal frequency is called. 4. Rate: in syllables per second (SPS), determined by dividing the number of syllables by the duration in seconds. It had been determined by previous unpublished work of Dr. Robert S. Brubaker that a clinically acceptable rate of speaking for Swiss-German children was 2.1-4.4 SPS. If a SPS count fell outside of that range, it was coded "inappropriate." 5. Consonant clusters: during the phonemic analysis, whether or not a child could produce a string of consonants, such as /str/ or /spr/, in ongoing speech. A child could be judged "inappropriate" in this regard by (a)not having consonant clusters in his or her speech production, (b) simplifying consonant clusters, e.g., substituting /sp/ for /spr/, or (c)not having high-frequency energy above 3000 Hz whenever appropriate.
Phonemic Analysis A search was made to determine what phonemes might not occur in the speech of the clinical groups. The intent was ,to determine if any class of sounds was omitted. Thirty phonemes were selected for the search. It was possible to array the 30 phonemes according to manner of articulation (vertical place) and place of articulation (horizontal place). The selected phonemes were 19 consonants and 11 vowels. The consonants were /pbmwfvt dnsz fljrkgr/x/. tn t971 one phoneme, /3/, was included as one of the 19 consonants, but it was replaced in 1972 by /x/. It seemed that very few clinical-group children had /3/ as part of their repertoire. It became increasingly apparent at that time that /3/ might not occur in Swiss dialects. Consequently, the lingua palatal /3/ was replaced by a lingua velar /x/, a sound quite prominent in Swiss dialects. The 11 vowels were /iaouIeeoeAo~r/. Even though the Swiss-German dialects contain diphthongs, it was determined not to use them in this research. It was concluded that since diphthongs are usually acquired
218
Affolter, Brubaker, and Franklin
later in speech development and since most of the children were just acquiring minimal speech, it would be a meaningless search. In general, labials, postdentals, palatals, and velars were categorized as horizontal place-of-articulation sounds, while stops, nasals, fricatives, semivowels, high vowels, and low vowels were classed as vertical place-of-articulation sounds. To compare the speech-sound ofnission of the clinical groups, expectancies were generated for Tables t, II, IV, and V: for 30 different speech sounds, the probability of one not occurring was 1 in 30.
Clinical Observations The clinical groups' speech-sound production habits were monitored in depth. Video tapes and audio tapes were made of the children so that the researchers might analyze in greater detail the children's performances. The children were observed before, during, and after a therapy session. All tapings were scrutinized by several researchers to see if any particular speech-sound production habits were characteristic for an individual child or for the differing groups.
RESULTS 1971 The following results include phonemic, acoustic, and observational data about serial group 1971 and intermodality group 1971. Phonemic Analysis, Serial Group 1971
Place of Articulation (Horizontal Place) Table I presents an analysis of missing consonants and vowels arranged according to front-back place of articulation. In general, this table rev6als little that cannot be attributed to chance, except in the case of palatals. Palatal sounds produced in the central region of the oral cavity were frequently omitted from the speech of serial children.
Manner of Articulation (Vertical Placej Table II shows that there were about 11% fewer missing phonemes of the stop category than expected. Thus serial children could produce stop-type
Speech-Sound in Language-Impaired Children
219
Table I. Serial Group 1971, Serial Group 1972, and Serial Group 1971 in 1972: Missing Phonemes Analyzed by Place of Articulation (Horizontal Place) I llii
Have Labials /pbmwfv/ Postdentals /tdnsz/ Palatals
Number missing Obtained % Number Expected possible % 1971 1972 1971in1972 1971 1972 1971in1972 6
20
9
8
6
20
21
20
5
17
7
4
6
15
I1
20
10/9a
33/30a
20
13
10
43
34
33
Velars 9/10a /aoukg ^ o'v'~/a
30/33a
10
13
8
21
34
27
100
46
38
30
99
100
100
/f3~jrieIeOe]a
Total
30
i
aln 1971/3/was a sound searched for but was replaced b y / x / i n 1972. This change requires a little adjustment in number possible and expected % figures.
Table II. Serial Group 1971, Serial Group 1972, and Serial Group 1971 in 1972: Missing Phonemes Analyzed by Manner of Articulation (Vertical Place)
Place Stops /pbtdkg] Nasals /mn~/ Fricatives
Numb~ missmg Obtained % Number Expected possible % 1971 1972 1971in 1972 1971 1972 1971in1972 6
20
4
5
4
9
13
13
3
10
4
4
2
9
11
7
6
20
16
10
9
35
26
30
4
13
7
7
4
15
18
13
6
20
8
6
4
17
16
13
5
17
7
6
7
15
16
23
30
100
46
38
30
100
100
99
/fvszf3/a Semivowels
/wgr/ High vowels /iuoIev-/ Low vowels
[aeoeAa/ Total i
i
aln 1971 ['3/was used but replaced by Ix/in 1972.
220
Affolter, Brubaker, and Franklin
articulations. Nasals were omitted within expectancies. The fricatives were expected to contribute about 20% of the omitted phonemes, but Table II shows that this class was the most frequently missed, with 35%. It is reasonable to conclude, therefore, that fricatives were difficult for serially disturbed children. Semivowels, high vowels, and low vowels were omitted about as often as one would expect by chance, so we may conclude they did not present a special problem.
Acoustic Analysis, Serial Group 1971 Table ItI reveals that serial group t97t was inappropriate in the following: (1)high-frequency enelgy associated with stop and fricative consonant articulation, (2)vowel formant clarity, and (3)consonant clusters. Two-thirds of the children had appropriate intonation, while three-fourths had appropriate rate (SPS). An interesting finding emerged at this point. The serial children who had high-frequency components tended not to be appropriate in intonation, while children whose intonation was appropriate tended not to have adequate high-frequency components above 3000 Hz. In short, these children tended to have either clear consonants and poor intonation or poor consonants and good intonation, but not both. Thus the group was successful in one aspect of speech-sound production or another, but not at a more complex level of both.
Clinical Observations, Serial Group 1971 As one observed the children working with a therapist or a parent, or while listening to 30- or 40-min tape recordings of sessions, one was struck with the fact that the children on the whole offered very little ongoing spontaneous speech-sound production. If they responded spontaneously during the session, the response more often ,than not was a one-word response. Now and again, a serially disturbed child responded with a simple short declarative sentence or question, but usually the response was somewhat shorter, perhaps a phrase. A two-way, ongoing conversation did not occur, but the serial child imitated sound and words proferred by the therapist or parent. Serial children had a great deal of difficulty in producing consonant clusters. Some of their errors w e r e / t i m / f o r / f w l m / , / & ~ t o / f o r / f m f t o e / , and /teI/ for /fteI/. The children substituted one sound for another, e,g.,/b-trY/for /f~?/,/fIm/ for /fwlm/, /haus/ for /klaus/, and /pT/ for ~fir~. The children often omitted the endings of words, such as /kroi/ for /kroit/ and /c~ifo/ for /e?~font/.
221
Speech-Sound in Language-Impaired Children
~v--~ C'q t"G',
Developmental Features of Speech-Sound Production in Language-Impaired Children F. A f f o l t e r , 1 R. B r u b a k e r , 2 a n d W. F r a n k l i n 3 Received August 20, 1976
The speech-sound production o f severely language-impaired children was monitored in a longitudinal study. In order to generalize findings, acoustical, phonemic, and clinical observation data were collecte d from 30 children. Results showed that speech-sound acquisition goes through a hierarchical sequence o f development and that speech-sound production will deteriorate in a predictable manner depending on the pathology. The authors hypothesize a speech-sound acquisition model, SSAM, based on the development observed.
INTRODUCTION In the normal child, development follows a hierarchical sequence of levels. In pathology, development is blocked at one of these levels. Levels 1 and 2 may be developed until a primary disturbance affects performances of level 3. As a consequence, levels 3 and 4 do not develop. Recovery, if it occurs, will follow the same normal hierarchical sequence-level 3 and then level 4. Language is characteristic of a higher stage of development than that of some perceptual mechanisms. With reference to the concept of hierarchical dependency, it can be said that perceptual mechanisms are prerequisite for language acquisition. Clinical case studies based on extended performance This-project was supported by the Swiss National Foundation for Scientific ResearchProjects No. 3.237.69, 3.448,70, and 3.902.72-and by North Carolina State University Research and Development Grant 056. 1Katholische Studentenhans, St. GaUen, Switzerland. 2Department of Speech Communication, Pennsylvania State University, University Park, Pennsylvania 16802. 3Department of Speech Communication, North Carolina State University, Raleigh, North Carolina. 213 0090-6905/78[0500-0213505.00[0 9 1978 Plenum Publishing Corporation
214
Affoltet, Brubaker, and Franklin
profiles (Affolter et al., 1974) showed that children with language deficits also showed deficits in perceptual performances: modality-specific, intermodal, and serial problems. These findings suggest that critical degrees of modalityspecific, of intermodal, and of serial integration are prerequisites for language acquisition. We will not attempt a lengthy discussion here elaborating the differences of two clinical subgroups: the serial group and the intermodality group. The serial group includes children with difficulty in integrating a critical amount of successive stimuli in all three modalities. The intermodality group consists of children characterized by difficulties relating and integrating a critical amount of information of different modalities (auditory, visual, and tactile-kinesthetic). The most severely language-impaired children in this group had been diagnosed as autistic by child psychiatrists. In the following study, speech-sound production of children with serial and intermodal problems was analyzed to investigate several aspects of the language-prerequisites interpretation suggested above. Specifically, it was hypothesized that (1)speech-sound acquisition goes through a hierarchical sequence of development, and in pathology and recovery the sequence of these levels will be the same, and (2)the production of speech sounds once acquired will not always be adequate in children with language problems. Under stress, the production will deteriorate in a predictable manner, with the serial group deteriorating differently from the intermodal group.
METHOD
Subjects Subjects 1971 Initially, in 1971, this research began with 30 children with problems of language development (called the clinical group) and 30 normal children of the same age, 3-10 years old. Because of significant differences of performance profiles within the clinical group, the clinical group was sub-divided into serial and intermodality groups. Serial group 1971 (18 in number, 12 testables on speech) consisted of children who demonstrated disturbances in serial integrations. Intermodality group 1971 (12 in number, three testables on speech) consisted of children who had difficulty in connecting different modalities.
Speech-Sound in Language-ImpairedChildren
215
Subjects 1972 In 1972, another clinical group of 18 children entered the study. This clinical group was subdivided into a serial group 1972 (seven subjects, three testables on speech) and an intermodatity group 1972 (11 subjects, five testables on speech).
Subjects 1971 in 1972 In 1972, five subjects of the original serial group 1971 were available for retesting, all five testable on speech. Intermodality group 1971 children available for retesting were six in number, four of whom were testable in speech-sound production. Procedures The children in the clinical groups were tested for speech-sound production. Observations were made by recording on video or acoustic tape and by written transcription during therapy sessions. Speech-sound production was analyzed in three fashions. Clinical observations within the clinical setting was one of the methods employed in exploring speech-sound production. A phonemic analysis was also undertaken and then selected samples were examined spectrographically.
Spectrographic Analysis Speech-sound production was studied by means of a Kay Electric model 6061B spectrum analyzer. The sonagraph (spectrograph) machine converts spoken sounds into a visual graph (a spectrogram) of frequencies and intensities as a function of time. In general, frequencies above 3000 Hz are associated with consonants, while vowels appear as bands of energy (formants) below 3000 Hz. From the spectrogram, which covers a range of 100-8000 Hz, a measurement of time can be made, so that the rate of speaking can be calculated. A plug-in component, Kay model 6076C scale magnifier, permits the expansion of selected portions of the voice spectrum so that individual harmonics can be identified. From this information, the change in vocal frequency over time can be seen and the intonation contour of the utterance observed. Some of the subjects ha 1971 were recorded live. All others were recorded first on audio tape from which selected samples of speech were used
216
Affolter, Brubaker, and Franklin
for spectrographic anal3/sis. This procedure was necessary because these children did not perform as normal children might. It was extremely difficult to have the children do as one would like. Most of the children had receptive language and could be told to do something. However, whether or not they would respond or were capable of performing on request was doubtful. It was decided to record them in the laboratory or clinical situation. Selected samplings of speech-sound production were taken from the tape recordings of each child and analyzed spectrographically. In most cases, the spectrographic analysis was made from tape recordings. Such spectrograms included any distortions inherent in the tape-recorder system and must be interpreted with caution. While the tape recorder was of sufficient quality for these purposes, no reproduction is as good as the original version. More serious are the recording problems arising in the clinical setting. For example, there is noise of the child's activity, of voices of parents and therapist, and of normal hospital activity. Another problem is that of changing microphone-to-subject distance. It can be shown that increasing this distance will attenuate the high frequencies displayed on the spectrogram. Inasmuch as high-frequency energy is associated with clearly articulated speech, there is a risk of significant error when evaluating articulation on spectrographic evidence alone. All one can do is evaluate a child's articulation, independently of the acoustic records, before and during recording. The spectrographic analysis consisted of five tests of speech-sound production. The five test results were judged to be either "normal" or "inappropriate." The five items of speech-sound production evaluated spectrographically were the following: 1. Consonants: evidence of frequency energy above 3000 Hz. Inasmuch as high-frequency energy is associated with clearly articulated speech, the researchers looked for evidence of energy above 3000 Hz. These high frequencies accompany~ stop and fricative consonant production. The absence of high-frequency energy, in part at least, is attributable to the absence of clear consonant articulation. 2. Vowels: evidence of clear vowel formants, steady after onset, without presence of noise. In the normal speech of adult males, clear vowels show formants which are distinct and smooth in appearance. However, the bandwidth (300 Hz) of the Kay Electric analyzing filter is not ideal for higher-pitched voices. As a result, formant clarity is often reduced in the spectrograms of women and children. For this reason, the observations concerning formant clarity are to be interpreted with caution.
Speech-Sound in Language-lmpaired Children
217
Some children with speech difficulty, although sounding as if they are producing some vowellike phonemes, do not give evidence of same when analyzed spectrographically. Reduced contrast of the formants can result from noise in the vocalic segment and is often caused by voice defects. 3. Intonation: variation in fundamental frequency as a function of time over an utterance. The Kay Electric model 6061B was equipped with a plug-in unit, the 6076C scale magnifier. The magnifying circuit of the sonagraph was used to make an expanded spectrogram of the lower 2000-Hz portion of the child's utterance. Expanded spectrograms show the individual harmonics of the voice. These enable one to determine the intonation contour, as the variation in vocal frequency is called. 4. Rate: in syllables per second (SPS), determined by dividing the number of syllables by the duration in seconds. It had been determined by previous unpublished work of Dr. Robert S. Brubaker that a clinically acceptable rate of speaking for Swiss-German children was 2.1-4.4 SPS. If a SPS count fell outside of that range, it was coded "inappropriate." 5. Consonant clusters: during the phonemic analysis, whether or not a child could produce a string of consonants, such as /str/ or /spr/, in ongoing speech. A child could be judged "inappropriate" in this regard by (a)not having consonant clusters in his or her speech production, (b) simplifying consonant clusters, e.g., substituting /sp/ for /spr/, or (c)not having high-frequency energy above 3000 Hz whenever appropriate.
Phonemic Analysis A search was made to determine what phonemes might not occur in the speech of the clinical groups. The intent was ,to determine if any class of sounds was omitted. Thirty phonemes were selected for the search. It was possible to array the 30 phonemes according to manner of articulation (vertical place) and place of articulation (horizontal place). The selected phonemes were 19 consonants and 11 vowels. The consonants were /pbmwfvt dnsz fljrkgr/x/. tn t971 one phoneme, /3/, was included as one of the 19 consonants, but it was replaced in 1972 by /x/. It seemed that very few clinical-group children had /3/ as part of their repertoire. It became increasingly apparent at that time that /3/ might not occur in Swiss dialects. Consequently, the lingua palatal /3/ was replaced by a lingua velar /x/, a sound quite prominent in Swiss dialects. The 11 vowels were /iaouIeeoeAo~r/. Even though the Swiss-German dialects contain diphthongs, it was determined not to use them in this research. It was concluded that since diphthongs are usually acquired
218
Affolter, Brubaker, and Franklin
later in speech development and since most of the children were just acquiring minimal speech, it would be a meaningless search. In general, labials, postdentals, palatals, and velars were categorized as horizontal place-of-articulation sounds, while stops, nasals, fricatives, semivowels, high vowels, and low vowels were classed as vertical place-of-articulation sounds. To compare the speech-sound ofnission of the clinical groups, expectancies were generated for Tables t, II, IV, and V: for 30 different speech sounds, the probability of one not occurring was 1 in 30.
Clinical Observations The clinical groups' speech-sound production habits were monitored in depth. Video tapes and audio tapes were made of the children so that the researchers might analyze in greater detail the children's performances. The children were observed before, during, and after a therapy session. All tapings were scrutinized by several researchers to see if any particular speech-sound production habits were characteristic for an individual child or for the differing groups.
RESULTS 1971 The following results include phonemic, acoustic, and observational data about serial group 1971 and intermodality group 1971. Phonemic Analysis, Serial Group 1971
Place of Articulation (Horizontal Place) Table I presents an analysis of missing consonants and vowels arranged according to front-back place of articulation. In general, this table rev6als little that cannot be attributed to chance, except in the case of palatals. Palatal sounds produced in the central region of the oral cavity were frequently omitted from the speech of serial children.
Manner of Articulation (Vertical Placej Table II shows that there were about 11% fewer missing phonemes of the stop category than expected. Thus serial children could produce stop-type
Speech-Sound in Language-Impaired Children
219
Table I. Serial Group 1971, Serial Group 1972, and Serial Group 1971 in 1972: Missing Phonemes Analyzed by Place of Articulation (Horizontal Place) I llii
Have Labials /pbmwfv/ Postdentals /tdnsz/ Palatals
Number missing Obtained % Number Expected possible % 1971 1972 1971in1972 1971 1972 1971in1972 6
20
9
8
6
20
21
20
5
17
7
4
6
15
I1
20
10/9a
33/30a
20
13
10
43
34
33
Velars 9/10a /aoukg ^ o'v'~/a
30/33a
10
13
8
21
34
27
100
46
38
30
99
100
100
/f3~jrieIeOe]a
Total
30
i
aln 1971/3/was a sound searched for but was replaced b y / x / i n 1972. This change requires a little adjustment in number possible and expected % figures.
Table II. Serial Group 1971, Serial Group 1972, and Serial Group 1971 in 1972: Missing Phonemes Analyzed by Manner of Articulation (Vertical Place)
Place Stops /pbtdkg] Nasals /mn~/ Fricatives
Numb~ missmg Obtained % Number Expected possible % 1971 1972 1971in 1972 1971 1972 1971in1972 6
20
4
5
4
9
13
13
3
10
4
4
2
9
11
7
6
20
16
10
9
35
26
30
4
13
7
7
4
15
18
13
6
20
8
6
4
17
16
13
5
17
7
6
7
15
16
23
30
100
46
38
30
100
100
99
/fvszf3/a Semivowels
/wgr/ High vowels /iuoIev-/ Low vowels
[aeoeAa/ Total i
i
aln 1971 ['3/was used but replaced by Ix/in 1972.
220
Affolter, Brubaker, and Franklin
articulations. Nasals were omitted within expectancies. The fricatives were expected to contribute about 20% of the omitted phonemes, but Table II shows that this class was the most frequently missed, with 35%. It is reasonable to conclude, therefore, that fricatives were difficult for serially disturbed children. Semivowels, high vowels, and low vowels were omitted about as often as one would expect by chance, so we may conclude they did not present a special problem.
Acoustic Analysis, Serial Group 1971 Table ItI reveals that serial group t97t was inappropriate in the following: (1)high-frequency enelgy associated with stop and fricative consonant articulation, (2)vowel formant clarity, and (3)consonant clusters. Two-thirds of the children had appropriate intonation, while three-fourths had appropriate rate (SPS). An interesting finding emerged at this point. The serial children who had high-frequency components tended not to be appropriate in intonation, while children whose intonation was appropriate tended not to have adequate high-frequency components above 3000 Hz. In short, these children tended to have either clear consonants and poor intonation or poor consonants and good intonation, but not both. Thus the group was successful in one aspect of speech-sound production or another, but not at a more complex level of both.
Clinical Observations, Serial Group 1971 As one observed the children working with a therapist or a parent, or while listening to 30- or 40-min tape recordings of sessions, one was struck with the fact that the children on the whole offered very little ongoing spontaneous speech-sound production. If they responded spontaneously during the session, the response more often ,than not was a one-word response. Now and again, a serially disturbed child responded with a simple short declarative sentence or question, but usually the response was somewhat shorter, perhaps a phrase. A two-way, ongoing conversation did not occur, but the serial child imitated sound and words proferred by the therapist or parent. Serial children had a great deal of difficulty in producing consonant clusters. Some of their errors w e r e / t i m / f o r / f w l m / , / & ~ t o / f o r / f m f t o e / , and /teI/ for /fteI/. The children substituted one sound for another, e,g.,/b-trY/for /f~?/,/fIm/ for /fwlm/, /haus/ for /klaus/, and /pT/ for ~fir~. The children often omitted the endings of words, such as /kroi/ for /kroit/ and /c~ifo/ for /e?~font/.
221
Speech-Sound in Language-Impaired Children
~v--~ C'q t"G',
