Neuropsyehologia, 1977, Vol. 15, pp. 19 to 30. Pergamon Press. Printed in England.
P H O N O L O G I C A L FACTORS IN AUDITORY COMPREHENSION IN APHASIA* SHEILA E. BLUMSTEIN, ERROL BAKER and HAROLD GOODGLASS Aphasia Research Center, Department of Neurology, Boston University School of Medicine, and Boston Veterans Administration Hospital, 150 So. Huntington Avenue, Boston, MA 02130, U.S.A.
(Received 20 November 1975) Abstract--It was the object of this study to investigate whether the comprehension deficit of Wernicke aphasics can be attributable to a selective deficit in phonemic hearing or phonemic discrimination as suggested by Luria. The ability to make phonemic discriminations between real words and nonsense syllables in relation to level of auditory comprehension was compared in 25 aphasic patients: 6 Broca's, 6 Mixed Anteriors, 6 Wernicke's, and 7 unclassified Posterior aphasics. The aphasics' ability to discriminate other phonological contrasts such as syllable structure and phoneme order was also tested. The subject's task was to indicate if two auditorily presented stimuli were the same or different. Results indicated that Mixed Anterior aphasics, who had only moderate comprehension deficits, performed worse on all phonological discriminations. Wernicke aphasics, with much more severe comprehension deficit, were less impaired in the present task. However, only the posterior aphasics (including Wernickes) had more difficulty in discriminating place of articulation than voicing contrasts. This deficit was interpreted as reflecting an impairment in linguistic encoding of speech sounds in the auditory association areas. It was concluded that a deficit in phonemic hearing cannot account for the comprehension deficit of Wernicke's aphasics. IN NEARLY all forms o f aphasia, patients evidence some f o r m of language comprehension deficit. Although it is clear that language comprehension is based u p o n the integration o f m a n y linguistic and psychological dimensions, it has only been in recent years that investigators have attempted to determine the contribution o f these dimensions to auditory comprehension. Results o f this research have revealed that dimensions such as vocabulary [1, 2], syntax [3, 4], semantics [5], and information redundancy [6] can be selectively impaired in aphasia. The importance o f phonological perception in language comprehension is obvious. I f the phonological form o f a word or utterance is incorrectly perceived, its contribution to the semantic or syntactic content o f the message m a y be erroneous. It has been suggested by LURIA [7] that the comprehension deficit o f Wernicke's aphasics can be attributed to a selective impairment in phonological perception. Specifically, he argues that the sensory aphasic has a disturbance in 'complex discriminative hearing' [7] manifested by an inability to distinguish minimal phonological attributes. F o r example, phonemes which are distinguished by a minimal n u m b e r o f phonological distinctive features, e.g. /p/-/b/ distinguished by the feature voice are more likely to be confused than phonemes distinguished by several features, e . g . / p / - / n / d i s t i n g u i s h e d by voice, nasality, and place. As Luria states, " W e have every reason to believe that the basic defect in temporal aphasia is the disturbance *This research was supported in part by USPHS Grants NS 07615 to Clark University and NS 06209 to Boston University. Reprint requests should be sent to Sheila E. Blumstein, Brown University, Box E, Providence, R.I. 02912, U.S.A. 19
20
SHEILA E. BLUMSTEIN, ERROL BAKER and HAROLD GOODGLASS
of auditory analysis and synthesis which leads to the loss of phonemic hearing, and, as a secondary result, to the disturbance of all functions which are dependent upon this physiological function"J7]. Luria bases these conclusions upon the Wernicke's aphasics' inability to perform certain production and recognition tasks based on phonemic contrasts. For example, the patient may be required to repeat alternating syllables distinguished by a minimal number of features (/pa/-ba/vs /pa/-/la/). Alternatively, he may be required to distinguish between two phonemes by making a response to one but not the other. Other investigators [8-10] have reported findings consistent with Luria's position. In particular, Wernicke aphasics seem to be selectively impaired in matching an auditorily presented word to its equivalent picture when presented in an array of phonemically similar test items. Although Wernicke aphasics may demonstrate these impairments, it is not clear whether these difficulties reflect a loss in the ability to discriminate speech sounds as suggested by Luria or rather an inability to use these sounds in a linguistically meaningful way. That is, all of the procedures used by Luria to test phonemic hearing demand that the patient make a linguistic response in another modality (repetition or writing) or keep in mind a conceptualized model of two phonemes and respond according to that model. Moreover, the picture identification task used by other investigators tests the relation between the phonological form of a word and its particular semantic representation. None of these procedures directly tests the Wernicke aphasics' ability to discriminate between speech sounds. Even if Wernicke aphasics do have demonstrable difficulty with the discrimination of speech sounds, it seemed unlikely that this impairment alone could be the basis for the profound comprehension deficit typical of these patients. The number of potential confusions among words in running speech based on an inability to discriminate similar speech sounds is quite limited; syntactic and semantic constraints effectively reduce the possible set of items which can belong in a particular context. Moreover, even if the comprehension deficit of these patients has a significant phonological component, it may not be properly defined in terms of a loss of capacity to make phoneme discriminations. The patient may simply be attempting to make a decision on the basis of incomplete use of phonological information, and as a result, may show a general impairment in phonological processing. Consideration of the clinical behavior of Wernicke aphasics suggests that a good deal more than a disorder of phonological processing is responsible for their comprehension deficit. It has often been observed that a patient although unable to choose or point to an item when presented in an array may nonetheless be able to repeat the particular test word. In this case, the patient seems to recognize the sound structure of the test item but not the meaning associated with it. It was the object of this study to investigate phonological processing in aphasia. Specifically we hoped to answer the following questions: (1) Do Wernicke's aphasics have a selective disorder in discriminating minimal phonemic contrasts in comparison to other aphasic groups? (2) Can this impairment account for the comprehension deficit of these patients? (3) Are other types of phonological processing selectively impaired in Wernicke's aphasia? Specifically, are they capable of discriminating differences in syllable structure and phoneme order?
PHONOLOGICAL FACTORS IN AUDITORY COMPREHENSION IN APHASIA
21
METHOD
Subjects Twenty-five aphasic patients from the Boston VA Hospital neurology service served as subjects. All had been prescreened to establish normal hearing in both ears for frequencies in the human speech range. Thresholds were determined for pure tones at 500, 1000, and 2000 Hz, and patients were eliminated whose speech reception threshold indicated a 30 dB mean loss. Patients were assigned to 1 of 4 diagnostic groups on the basis of a composite aphasia examination including psychological, language and neurological tests presented and discussed at the aphasia rounds of the Boston VA Hospital. Patients with anterior left hemisphere lesions were assigned to 2 groups, those with good comprehension (Broca's aphasics, n ----- 6) and those with diminished comprehension (Mixed Anteriors, n ~ 6). Comprehension was assessed by performance on 4 subtests of the Boston Diagnostic Aphasia Battery [11]. Subjects with a mean z-score greater than zero were assigned to the Broca's aphasia group and those with a mean z-score of less than or equal to zero were assigned to the Mixed Anterior group. The remaining population was drawn from patients with posterior left hemisphere lesions. These patients were also divided into 2 groups: patients with predominantly temporal lobe lesions (Wernicke aphasics, n = 6) and patients who represented several diagnostic categories, e.g. anomic, conduction, transcortical sensory (Residual Posteriors, n = 7). PROCEDURE
AND
APPARATUS
Ss were tested individually in a sound-proofed room. Each input to a stereophonic headphone (GrasonStadler) was independently adjusted on an audiometer to + 4 0 dB SRT. Prior to every test session Ss received a training procedure to insure comprehension of the task. This procedure consisted of the presentation of 10 pairs of stimuli, 5 of which were identical pairs and 5 of which were contrasting pairs. The contrasting pairs demanded successively more difficult discriminations, e.g. boy-sheep vs girl-pearl. After each pair S was required to press one button labelled "yes" if both stimuli had been perceived as identical, or a second marked " n o " if perceived as different. The positions of the 'yes' and 'no' buttons were independently counterbalanced across subjects and test sessions. STIMULI
AND
DESIGN
A discrimination test was devised consisting of 2 parts: discrimination between pairs of real words and discrimination between pairs of nonsense words. The nonsense stimuli were constructed to reflect the same phonological relations as the real words. The phonological attributes investigated were: phoneme discrimination; syllable discrimination; and phoneme-order discrimination.
Phoneme discrimination There were 72 pairs of different words constructed for each part of the test. The discriminations required were divided equally among the distinctive feature values of place of articulation, voicing, and voice and place. The English stop consonants ] p t k b d g / were used as the critical target stimuli. Both one and two syllable words were constructed, contrasting equally in initial and final position for the single syllables, and in initial and medial position for the two-syllable stimuli (cf. Table 1).
1 syllable 2 syllables
Table 1. Examples of types of phonological discriminations Phoneme Discrimination Voice Place Voice]place pear-bear pin-tin pen-den rope-robe sheep-sheet shape-shade pony-bony power-tower sopping-sobbing super-suiter Syllable Discrimination describe-prescribe tr~.nsport-mfnsfer Phoneme Order Discrimination main-name tax-task
pardon-garden wiper-wider
Syllable discrimination A total of 8 pairs (per test) of two syllable words were used in which the unstressed syllable was varied
22
SHEILA E. BLUMSTEIN,ERROL BAKER and HAROLD GOODGLASS
while the stressed syllable remained constant. Half of the words varied on the first syllable and half on the second (cf. Table 1).
Phoneme-order discrimination Eight pairs (per list) of one and two syllable words were constructed distinguished by the order of the component phonemes (cf. Table 1). These 88 pairs were presented twice, once with each member appearing in first position. For each of these 176 pairs there was a corresponding "control" pair where the first stimulus was repeated ("same" condition). The resulting 352 pairs were recorded in a different random sequence for the real and the nonsense lists. Each of the 176 individual real and nonsense words was recorded once by a male speaker. Thus for each occurrence of a particular item, the identical recorded segment was utilized. There was a 2 sec delay between each member of a test pair, and approximately 5 sec between each test item. Ss were tested in a minimum of 2 sessions. Real and nonsense stimuli were presented on separate days, counterbalanced for order across Ss.
Scoring
RESULTS
I n a same-different discrimination test, Ss may make two types of incorrect responses. They may respond 'different' to stimuli that are identical (false negatives) a n d they m a y respond 'same' to stimuli that are in fact different (misses). The percent of false negative responses was taken as a measure of Ss guessing rate, i.e. as an index of his ability to perform the task reliably. A n y S that had a guess error rate more than two standard deviations above his group mean was excluded from further analysis. Only 1 subject was eliminated by this criterion. Guessing errors were submitted to a G r o u p x Sense analysis of variance. There were no significant differences in guessing rate a m o n g groups n o r between real a n d nonsense stimuli. I n the statistical analyses to be reported below, scores corrected for guessing were used. This was determined by subtracting the subject's guess error rate from the percent of actual misses. Corrected scores were determined individually for each subject in each test condition.
Phonemic hearing and auditory comprehension A s u m m a r y of the performance of the four aphasic groups in the task of p h o n e m e disc r i m i n a t i o n is presented in Fig. 1. These data were submitted to a G r o u p x P h o n e m e x Discrimination analysis of variance followed by a N e w m a n - K e u l s t-test where appro50/I-
Phoneme
40
~
3(2,
~
I0
drSgrlmlngtlon
Nonsense
? ra a Z- +0.725
Mixed a~terior ~0555
Wern ic ke -I 46~
Res,dual posterior 0 188
FIG. 1. Summary of the performance of the four aphasic groups on the task of phoneme
discrimination.
PHONOLOGICAL FACTORS IN AUDITORY COMPREHENSION IN APHASIA
23
priate. Results of this analysis revealed a significant overall Groups effect (F = 3.25, dr=3, 21 ; P < 0.05). This was due to the fact that the Broca's aphasics made significantly fewer errors than any of the other aphasic groups (P < 0.01 for all groups). The next fewest errors were committed by the Residual Posterior group. Surprisingly, the largest number of errors was made by the Mixed Anterior patients. The difference between the performance of the Mixed Anterior aphasics and Residual Posterior group only approached significance (P = nisse zeigten an, da@ die g e m i s c h t e n \rorderhi r ~ - A ~ h a s i k e r ( p r a e c e n i r a l e A. ) , die nut [,',}[Bi
P H O N O L O G I C A L FACTORS IN AUDITORY COMPREHENSION IN APHASIA* SHEILA E. BLUMSTEIN, ERROL BAKER and HAROLD GOODGLASS Aphasia Research Center, Department of Neurology, Boston University School of Medicine, and Boston Veterans Administration Hospital, 150 So. Huntington Avenue, Boston, MA 02130, U.S.A.
(Received 20 November 1975) Abstract--It was the object of this study to investigate whether the comprehension deficit of Wernicke aphasics can be attributable to a selective deficit in phonemic hearing or phonemic discrimination as suggested by Luria. The ability to make phonemic discriminations between real words and nonsense syllables in relation to level of auditory comprehension was compared in 25 aphasic patients: 6 Broca's, 6 Mixed Anteriors, 6 Wernicke's, and 7 unclassified Posterior aphasics. The aphasics' ability to discriminate other phonological contrasts such as syllable structure and phoneme order was also tested. The subject's task was to indicate if two auditorily presented stimuli were the same or different. Results indicated that Mixed Anterior aphasics, who had only moderate comprehension deficits, performed worse on all phonological discriminations. Wernicke aphasics, with much more severe comprehension deficit, were less impaired in the present task. However, only the posterior aphasics (including Wernickes) had more difficulty in discriminating place of articulation than voicing contrasts. This deficit was interpreted as reflecting an impairment in linguistic encoding of speech sounds in the auditory association areas. It was concluded that a deficit in phonemic hearing cannot account for the comprehension deficit of Wernicke's aphasics. IN NEARLY all forms o f aphasia, patients evidence some f o r m of language comprehension deficit. Although it is clear that language comprehension is based u p o n the integration o f m a n y linguistic and psychological dimensions, it has only been in recent years that investigators have attempted to determine the contribution o f these dimensions to auditory comprehension. Results o f this research have revealed that dimensions such as vocabulary [1, 2], syntax [3, 4], semantics [5], and information redundancy [6] can be selectively impaired in aphasia. The importance o f phonological perception in language comprehension is obvious. I f the phonological form o f a word or utterance is incorrectly perceived, its contribution to the semantic or syntactic content o f the message m a y be erroneous. It has been suggested by LURIA [7] that the comprehension deficit o f Wernicke's aphasics can be attributed to a selective impairment in phonological perception. Specifically, he argues that the sensory aphasic has a disturbance in 'complex discriminative hearing' [7] manifested by an inability to distinguish minimal phonological attributes. F o r example, phonemes which are distinguished by a minimal n u m b e r o f phonological distinctive features, e.g. /p/-/b/ distinguished by the feature voice are more likely to be confused than phonemes distinguished by several features, e . g . / p / - / n / d i s t i n g u i s h e d by voice, nasality, and place. As Luria states, " W e have every reason to believe that the basic defect in temporal aphasia is the disturbance *This research was supported in part by USPHS Grants NS 07615 to Clark University and NS 06209 to Boston University. Reprint requests should be sent to Sheila E. Blumstein, Brown University, Box E, Providence, R.I. 02912, U.S.A. 19
20
SHEILA E. BLUMSTEIN, ERROL BAKER and HAROLD GOODGLASS
of auditory analysis and synthesis which leads to the loss of phonemic hearing, and, as a secondary result, to the disturbance of all functions which are dependent upon this physiological function"J7]. Luria bases these conclusions upon the Wernicke's aphasics' inability to perform certain production and recognition tasks based on phonemic contrasts. For example, the patient may be required to repeat alternating syllables distinguished by a minimal number of features (/pa/-ba/vs /pa/-/la/). Alternatively, he may be required to distinguish between two phonemes by making a response to one but not the other. Other investigators [8-10] have reported findings consistent with Luria's position. In particular, Wernicke aphasics seem to be selectively impaired in matching an auditorily presented word to its equivalent picture when presented in an array of phonemically similar test items. Although Wernicke aphasics may demonstrate these impairments, it is not clear whether these difficulties reflect a loss in the ability to discriminate speech sounds as suggested by Luria or rather an inability to use these sounds in a linguistically meaningful way. That is, all of the procedures used by Luria to test phonemic hearing demand that the patient make a linguistic response in another modality (repetition or writing) or keep in mind a conceptualized model of two phonemes and respond according to that model. Moreover, the picture identification task used by other investigators tests the relation between the phonological form of a word and its particular semantic representation. None of these procedures directly tests the Wernicke aphasics' ability to discriminate between speech sounds. Even if Wernicke aphasics do have demonstrable difficulty with the discrimination of speech sounds, it seemed unlikely that this impairment alone could be the basis for the profound comprehension deficit typical of these patients. The number of potential confusions among words in running speech based on an inability to discriminate similar speech sounds is quite limited; syntactic and semantic constraints effectively reduce the possible set of items which can belong in a particular context. Moreover, even if the comprehension deficit of these patients has a significant phonological component, it may not be properly defined in terms of a loss of capacity to make phoneme discriminations. The patient may simply be attempting to make a decision on the basis of incomplete use of phonological information, and as a result, may show a general impairment in phonological processing. Consideration of the clinical behavior of Wernicke aphasics suggests that a good deal more than a disorder of phonological processing is responsible for their comprehension deficit. It has often been observed that a patient although unable to choose or point to an item when presented in an array may nonetheless be able to repeat the particular test word. In this case, the patient seems to recognize the sound structure of the test item but not the meaning associated with it. It was the object of this study to investigate phonological processing in aphasia. Specifically we hoped to answer the following questions: (1) Do Wernicke's aphasics have a selective disorder in discriminating minimal phonemic contrasts in comparison to other aphasic groups? (2) Can this impairment account for the comprehension deficit of these patients? (3) Are other types of phonological processing selectively impaired in Wernicke's aphasia? Specifically, are they capable of discriminating differences in syllable structure and phoneme order?
PHONOLOGICAL FACTORS IN AUDITORY COMPREHENSION IN APHASIA
21
METHOD
Subjects Twenty-five aphasic patients from the Boston VA Hospital neurology service served as subjects. All had been prescreened to establish normal hearing in both ears for frequencies in the human speech range. Thresholds were determined for pure tones at 500, 1000, and 2000 Hz, and patients were eliminated whose speech reception threshold indicated a 30 dB mean loss. Patients were assigned to 1 of 4 diagnostic groups on the basis of a composite aphasia examination including psychological, language and neurological tests presented and discussed at the aphasia rounds of the Boston VA Hospital. Patients with anterior left hemisphere lesions were assigned to 2 groups, those with good comprehension (Broca's aphasics, n ----- 6) and those with diminished comprehension (Mixed Anteriors, n ~ 6). Comprehension was assessed by performance on 4 subtests of the Boston Diagnostic Aphasia Battery [11]. Subjects with a mean z-score greater than zero were assigned to the Broca's aphasia group and those with a mean z-score of less than or equal to zero were assigned to the Mixed Anterior group. The remaining population was drawn from patients with posterior left hemisphere lesions. These patients were also divided into 2 groups: patients with predominantly temporal lobe lesions (Wernicke aphasics, n = 6) and patients who represented several diagnostic categories, e.g. anomic, conduction, transcortical sensory (Residual Posteriors, n = 7). PROCEDURE
AND
APPARATUS
Ss were tested individually in a sound-proofed room. Each input to a stereophonic headphone (GrasonStadler) was independently adjusted on an audiometer to + 4 0 dB SRT. Prior to every test session Ss received a training procedure to insure comprehension of the task. This procedure consisted of the presentation of 10 pairs of stimuli, 5 of which were identical pairs and 5 of which were contrasting pairs. The contrasting pairs demanded successively more difficult discriminations, e.g. boy-sheep vs girl-pearl. After each pair S was required to press one button labelled "yes" if both stimuli had been perceived as identical, or a second marked " n o " if perceived as different. The positions of the 'yes' and 'no' buttons were independently counterbalanced across subjects and test sessions. STIMULI
AND
DESIGN
A discrimination test was devised consisting of 2 parts: discrimination between pairs of real words and discrimination between pairs of nonsense words. The nonsense stimuli were constructed to reflect the same phonological relations as the real words. The phonological attributes investigated were: phoneme discrimination; syllable discrimination; and phoneme-order discrimination.
Phoneme discrimination There were 72 pairs of different words constructed for each part of the test. The discriminations required were divided equally among the distinctive feature values of place of articulation, voicing, and voice and place. The English stop consonants ] p t k b d g / were used as the critical target stimuli. Both one and two syllable words were constructed, contrasting equally in initial and final position for the single syllables, and in initial and medial position for the two-syllable stimuli (cf. Table 1).
1 syllable 2 syllables
Table 1. Examples of types of phonological discriminations Phoneme Discrimination Voice Place Voice]place pear-bear pin-tin pen-den rope-robe sheep-sheet shape-shade pony-bony power-tower sopping-sobbing super-suiter Syllable Discrimination describe-prescribe tr~.nsport-mfnsfer Phoneme Order Discrimination main-name tax-task
pardon-garden wiper-wider
Syllable discrimination A total of 8 pairs (per test) of two syllable words were used in which the unstressed syllable was varied
22
SHEILA E. BLUMSTEIN,ERROL BAKER and HAROLD GOODGLASS
while the stressed syllable remained constant. Half of the words varied on the first syllable and half on the second (cf. Table 1).
Phoneme-order discrimination Eight pairs (per list) of one and two syllable words were constructed distinguished by the order of the component phonemes (cf. Table 1). These 88 pairs were presented twice, once with each member appearing in first position. For each of these 176 pairs there was a corresponding "control" pair where the first stimulus was repeated ("same" condition). The resulting 352 pairs were recorded in a different random sequence for the real and the nonsense lists. Each of the 176 individual real and nonsense words was recorded once by a male speaker. Thus for each occurrence of a particular item, the identical recorded segment was utilized. There was a 2 sec delay between each member of a test pair, and approximately 5 sec between each test item. Ss were tested in a minimum of 2 sessions. Real and nonsense stimuli were presented on separate days, counterbalanced for order across Ss.
Scoring
RESULTS
I n a same-different discrimination test, Ss may make two types of incorrect responses. They may respond 'different' to stimuli that are identical (false negatives) a n d they m a y respond 'same' to stimuli that are in fact different (misses). The percent of false negative responses was taken as a measure of Ss guessing rate, i.e. as an index of his ability to perform the task reliably. A n y S that had a guess error rate more than two standard deviations above his group mean was excluded from further analysis. Only 1 subject was eliminated by this criterion. Guessing errors were submitted to a G r o u p x Sense analysis of variance. There were no significant differences in guessing rate a m o n g groups n o r between real a n d nonsense stimuli. I n the statistical analyses to be reported below, scores corrected for guessing were used. This was determined by subtracting the subject's guess error rate from the percent of actual misses. Corrected scores were determined individually for each subject in each test condition.
Phonemic hearing and auditory comprehension A s u m m a r y of the performance of the four aphasic groups in the task of p h o n e m e disc r i m i n a t i o n is presented in Fig. 1. These data were submitted to a G r o u p x P h o n e m e x Discrimination analysis of variance followed by a N e w m a n - K e u l s t-test where appro50/I-
Phoneme
40
~
3(2,
~
I0
drSgrlmlngtlon
Nonsense
? ra a Z- +0.725
Mixed a~terior ~0555
Wern ic ke -I 46~
Res,dual posterior 0 188
FIG. 1. Summary of the performance of the four aphasic groups on the task of phoneme
discrimination.
PHONOLOGICAL FACTORS IN AUDITORY COMPREHENSION IN APHASIA
23
priate. Results of this analysis revealed a significant overall Groups effect (F = 3.25, dr=3, 21 ; P < 0.05). This was due to the fact that the Broca's aphasics made significantly fewer errors than any of the other aphasic groups (P < 0.01 for all groups). The next fewest errors were committed by the Residual Posterior group. Surprisingly, the largest number of errors was made by the Mixed Anterior patients. The difference between the performance of the Mixed Anterior aphasics and Residual Posterior group only approached significance (P = nisse zeigten an, da@ die g e m i s c h t e n \rorderhi r ~ - A ~ h a s i k e r ( p r a e c e n i r a l e A. ) , die nut [,',}[Bi
