BRAIN
AND
LANGUAGE
3,28-33
(1976)
Short-Term
Memory
and Aphasia
MARTIN L. ALBERT' Aranne Laboratory for Human Psychophysiology, Department of Neurology, Hadassah University Hospital, Jerusalem, Israel, and the Aphasia Research Center, Department of Neurology, Boston University Medical School and Boston VA Hospital, Boston, Massachusetts. 02130. Short-term retention tests of verbal item information (short-term memory for items) and order information (memory for sequences) were administered to aphasic, nonaphasic brain-damaged, and normal subjects. Memory for both parameters was significantly impaired in aphasics only. One-third of errors made by aphasics resulted from a specific, separable defect in short-term memory for sequences. As information load increased, memory for sequences became critical for linguistic performance of aphasics.
In 1967 Hirsh suggested that there exists within the auditory system a mechanism specifically adapted to place in correct order the elements of the structure of language. Efron (1963) had previously observed that a defect in sequence discrimination was found only when there was some degree of aphasia present. Other authors (Goodglass et al., 1970; Albert, 1972) have since demonstrated a defective capacity for temporal sequencing of auditory stimuli in aphasics, by contrast with nonaphasic brain-damaged subjects. These studies implied that the left hemisphere is dominant for normal human language because of its predominant capacity to maintain and utilize the sequential aspects of acoustic inputs. However, none of these studies adequately isolated memory for sequences from memory for items. An argument could be raised that left hemispheric damage caused a general impairment in auditory verbal short-term memory and not a specific impairment in memory for sequences. To support the hypothesis that the left hemisphere is language-dominant because of its superior auditory sequencing capacity, it is first necessary to demonstrate that retention of item information (short-term memory for items) and retention of order information (short-term memory for sequences of items) are distinct phenomena. Then it must be shown that the damaged left hemisphere is selectively impaired in short-term memory for sequences. I thank Dr. G. Zajicek for advice and assistance regarding statistical analyses; L. K. Obler for theoretical advice and technical assistance: and Drs. J. Nachansohn and I. Schecter for providing subjects through the facilities of Loewenstein Hospital, Tel Aviv. ’ Dr. Albert’s current address is: Aphasia Research Center, Boston VA Hospital, 150 South Huntington Ave., Boston, Mass. 02130. 28 Copyright All rights
0 1976 by Academic Press. Inc. of reproduction in any form reserved.
SHORT-TERM
MEMORY AND APHASIA
29
Warrington and Shallice (1969) presented evidence that auditory verbal short-term memory could be impaired by a specifically-located lesion within the left hemispheric language zone. Tzortzis and Albert (1974) demonstrated with a small group of aphasics that defective short-term memory for sequences could be distinguished from a more general disorder of auditory verbal short-term memory. Their patients could often recall all the items presented, but not in the correct order. The two experiments described in this paper evaluated auditory verbal short-term memory in a large group of normal and brain-damaged subjects in order to consider the following questions: Can memory for sequences be separated from memory for items ? Do language-impaired (aphasic) subjects have a general defect in auditory verbal short-term memory, or a selective defect in short-term memory for sequences? SUBJECTS
AND METHODOLOGY
The tests described below were administered to 82 right-handed subjects; 25 were controls (C); 28 had unilateral cerebral damage with aphasia (A+); 29 had unilateral cerebral damage without aphasia (A-). The controls were hospitalized patients with neurological disease below the level of the neck. Brain-damaged subjects were tested for aphasia by a modified version of the Boston Diagnostic Examination for Aphasia (Goodglass & Kaplan, 1972). Causes of brain damage were vascular 32, trauma 17, tumor 8, equally distributed across groups. Mean ages in years were C, 44.4; A+, 46.5; A-, 42.9. Mean educational level in years were C, 12; A+, 12; A-, 10. Time in months from onset of illness to date of testing was A+, 10.1; A-, 6.5. EXPERIMENT
I
Methodology In the first study, a modified version of the Auditory Sequencing Test was used (Albert, 1972). Subject and examiner sat facing each other. Between them was a table on which 18common objects (comb, pen, etc.) were placed randomly. The examiner ascertained that the subject could correctly indicate each item individually in response to verbal command. The examiner then asked the subject to point to four items in a predetermined, randomized sequence (e.g., “In this order, point to the A, the B, the C, and the D”). Three trials with four different items for each trial were given. AU responses, correct or incorrect, were recorded for analysis.
Results Table 1 shows performance of subject groups.2 For retention of total item information there was no significant difference between nonaphasic 2 Statistical analyses employed z tests of the significance of the difference between two proportions; a standardized, normal variable for an upper tail probability was computed (Wallis & Roberts, 1972). The proportions used in these comparisons are recorded in the tables and represent ratios of items correct to total items; items correct in correct sequences to total items; item errors to total items; sequences incorrect to total sequences; sequences incorrect with sequencing errors to total incorrect sequences; and sequences incorrect with omission errors to total incorrect sequences.
30
MARTIN
L. ALBERT
TABLE 1 FOUR-ITEM SEQUENCINGTESP
Groups
Number
Items correct (Total items)
Items correct in correct sequences (Total items)
Sequences incorrect with sequencing errors (Total incorrect sequences)”
Sequences incorrect with omission errors (Total incorrect sequences)
C
25
282 zii
240 Gii
3 Is
ii
A-
29
318 348
248 348
5 2s
14 2s
A+
28
201 336
24 336
21 62
42 ii
3
a C = controls; A- = nonaphasic brain-damaged subjects; A+ = aphasics. b Responses containing two or more correct items.
brain-damaged subjects and controls (z = 0.05, z cut-off for significance = 1.96). However, aphasics performed significantly worse than did nonaphasic brain-damaged subjects (z = 2.74,~ < 0.05). More striking was the fact that retention of order information (total items correct in correct sequences) was selectively and significantly impaired among aphasics (A+ vs. A-: z = 4.09, p < 0.01). Actual performance of each subject had been noted; and error analyses were carried out. At no time did a subject fail to respond. Responses included from zero to four correct items. Errors were either substitutions (the subject pointed to an item not requested), incorrect sequences of correct items, or omissions (failure to provide an item requested, without substitution). Controls made equal amounts of omission and sequencing errors (20% of errors were of each type). For each brain-damaged group, omission errors were the most prominent type; and there was no significant difference between A+ and A- for this type of error (A+ 54%, A- 56%). Sequencing-type errors, however, in aphasics reached 34%, while in controls and nonaphasic brain-damaged subjects, they accounted for 20% each of all errors. At the level of four-item responses (that is, for responses in which four items were indicated) the difference between A+ and A- for sequencing errors was significant (z = 3.68, p < 0.05, z cut-off for significance = 1.64). For responses in which three items were indicated, the difference between A+ and A- for sequencing errors was not significant, but was nearly so (z = 1.31). The evidence suggests that aphasics, unlike normals or brain-damaged nonaphasics, have significantly impaired auditory short term memories both for total verbal item information and for verbal sequences. In
SHORT-TERM
31
MEMORY AND APHASIA
addition, only for the aphasics, the influence of the defective memory for sequences becomes more pronounced as information load increases. For further consideration of these issues an additional test was administered. EXPERIMENT
II
Methodology The structural format of the preceding test was retained. However, for this test the examiner named two objects. Subjects were required to point to those two items, in the order named, from a group of 18 objects placed before them. Five trials were given, with two different objects for each trial.
Results
Table 2 shows the performance of subject groups. Aphasics made significantly more’total errors (24%) than did the normals (2%) or nonaphasic brain-damaged subjects (2.5%). (A+ vs. A- ,z = 2.29, p < 0.05, z cut-off for significance = 1.64). There was no significant difference between controls and nonaphasic brain-damaged subjects for total errors (z = 0.03). Error analysis revealed that the significant difference in total errors between aphasics and the other two groups was not due to sequencing-type errors, but rather was due primarily to omissions. At the level of two items, there was no significant difference between aphasics and others for sequencing-type errors. Additional analyses, concerned with anatomical localization and clinical correlations, were performed. Evaluation of lesion site was made by neurological examination, electroencephalography, and radioactive gamma encephalography in all cases; and cerebral arteriography in most TABLE 2 TWO-ITEM SEQUENCING TEST RESULTS”
Groups Number
Total item errors (Total items)
Sequences incorrect (Total sequences)
Sequences incorrect with sequencing errors (Total incorrect sequences)
Sequences incorrect with omission errors (Total incorrect sequences)
C
21
4 210
6 los
2 -ii
3 --ii
A-
27
7 270
- II 135
4 11
4 ii
A+
22
52 220
52 ii6
12 52
30 s2
a C = controls; A- = nonaphasic brain-damaged subjects; A+ = aphasics.
32
MARTIN
L. ALBERT
cases. Every aphasic had a lesion in the left hemisphere involving some portion (anterior, posterior, or both) of the perisylvian region, the anatomical region defined by Dejerine (1914) as the zone of language. A lesion anywhere within this anatomical region produced a defective auditory verbal short term memory. The aphasic group was further evaluated according to the clinical classification of anterior vs. posterior aphasia (Geschwind, 1970; Goodglass & Kaplan, 1972). On both the two-item and the four-item tests, posterior aphasics made more total errors (for items) than did anterior aphasics (two-item test total errors: anterior l%, posterior 2%; four-item test total errors: anterior 36%, posterior 45%). However, these differences were not statistically significant. Error analysis revealed no significant differences between anterior and posterior aphasics for sequencing-type errors on either the two-item or four-item test (two-item test sequencing errors: anterior = 2% of total errors, posterior = 17%; four-item test sequencing errors: anterior = 33%, posterior = 34%). Impaired short-term memory for items and impaired short-term memory for sequences were present regardless of clinical type of aphasia. DISCUSSION
Two questions were asked at the beginning of this study: (1) Can memory for sequences be separated from memory for items? (2) Do aphasics have a general defect in auditory verbal short-term memory, or do they have a selective defect in short-term memory for sequences? The answer to the first question is “yes.” Evidence for this was provided by analyses of errors, which were of three types: substitution, omission, and sequencing errors. These results, obtained from a large group of brain-damaged subjects, are consistent with those of Tzortzis and Albert (1974) who made similar observations after studying a small, selected group of aphasics. Both studies demonstrated that aphasic subjects can frequently remember all of the items in a given set, but not the order of the items in the set. As for the second question, the answer is “both.” Aphasics, unlike normals and nonaphasic brain-damaged subjects, have a sign&ant defect in auditory verbal short-term memory for total items regardless of the total information load. Aphasics also have a selective defect in memory for sequences; this selective defect, however, has special characteristics. At low information load levels, the major form of memory defect in aphasics is an “omission” type. As information load increases, defective memory for sequences becomes a critical factor in the linguistic performance of aphasics. REFERENCES Albert,
M. L. 1972. Certains aspects des troubles de la comprehension Langages, 7,37-51.
auditive du langage.
SHORT-TERM
MEMORY AND APHASIA
33
Albert, M. L. 1972. Auditory sequencing and left cerebral dominance for language. Neuropsychologia, 10, 245-248. Dejerine, J. 1914. Semeiologie des affections du systkme nerveux, Paris: Masson. Efron, R. 1%3. Temporal perception, aphasia, and dijti vu. Brain, 86,403-424. Geschwind, N. 1970. Aphasia. Science, 170, 940. Goodglass, H., & Kaplan, E. 1972. The Assessment of Aphasia and Related Disorders, Philadelphia: Lea & Febiger. Goodglass, H., Gleason, J., & Hyde, M. 1970. Some dimensions of auditory language comprehension in aphasia. Journal of Speech and Hearing Research, 13,595-606. Hirsch, I. 1%7. Information processing in input channels for speech and language: The significance of serial order of stimuli. In C. Millikan & F. Darley (Eds.), Brain mechanisms underlying speech and language. New York: Grune and Stratton. Tzortzis, C., & Albert, M. L. 1974. Impairment of memory for sequences in conduction aphasia. Neuropsychologia, 12, 355-366. Wallis, W., & Roberts, H. 1972. Statistics, a new approach, Glencoe, Ill.: Free Press. Warrington, E., & Shallice, T. 1%9. The selective impairment of auditory verbal short-term memory. Brain, 92, 885-896.
AND
LANGUAGE
3,28-33
(1976)
Short-Term
Memory
and Aphasia
MARTIN L. ALBERT' Aranne Laboratory for Human Psychophysiology, Department of Neurology, Hadassah University Hospital, Jerusalem, Israel, and the Aphasia Research Center, Department of Neurology, Boston University Medical School and Boston VA Hospital, Boston, Massachusetts. 02130. Short-term retention tests of verbal item information (short-term memory for items) and order information (memory for sequences) were administered to aphasic, nonaphasic brain-damaged, and normal subjects. Memory for both parameters was significantly impaired in aphasics only. One-third of errors made by aphasics resulted from a specific, separable defect in short-term memory for sequences. As information load increased, memory for sequences became critical for linguistic performance of aphasics.
In 1967 Hirsh suggested that there exists within the auditory system a mechanism specifically adapted to place in correct order the elements of the structure of language. Efron (1963) had previously observed that a defect in sequence discrimination was found only when there was some degree of aphasia present. Other authors (Goodglass et al., 1970; Albert, 1972) have since demonstrated a defective capacity for temporal sequencing of auditory stimuli in aphasics, by contrast with nonaphasic brain-damaged subjects. These studies implied that the left hemisphere is dominant for normal human language because of its predominant capacity to maintain and utilize the sequential aspects of acoustic inputs. However, none of these studies adequately isolated memory for sequences from memory for items. An argument could be raised that left hemispheric damage caused a general impairment in auditory verbal short-term memory and not a specific impairment in memory for sequences. To support the hypothesis that the left hemisphere is language-dominant because of its superior auditory sequencing capacity, it is first necessary to demonstrate that retention of item information (short-term memory for items) and retention of order information (short-term memory for sequences of items) are distinct phenomena. Then it must be shown that the damaged left hemisphere is selectively impaired in short-term memory for sequences. I thank Dr. G. Zajicek for advice and assistance regarding statistical analyses; L. K. Obler for theoretical advice and technical assistance: and Drs. J. Nachansohn and I. Schecter for providing subjects through the facilities of Loewenstein Hospital, Tel Aviv. ’ Dr. Albert’s current address is: Aphasia Research Center, Boston VA Hospital, 150 South Huntington Ave., Boston, Mass. 02130. 28 Copyright All rights
0 1976 by Academic Press. Inc. of reproduction in any form reserved.
SHORT-TERM
MEMORY AND APHASIA
29
Warrington and Shallice (1969) presented evidence that auditory verbal short-term memory could be impaired by a specifically-located lesion within the left hemispheric language zone. Tzortzis and Albert (1974) demonstrated with a small group of aphasics that defective short-term memory for sequences could be distinguished from a more general disorder of auditory verbal short-term memory. Their patients could often recall all the items presented, but not in the correct order. The two experiments described in this paper evaluated auditory verbal short-term memory in a large group of normal and brain-damaged subjects in order to consider the following questions: Can memory for sequences be separated from memory for items ? Do language-impaired (aphasic) subjects have a general defect in auditory verbal short-term memory, or a selective defect in short-term memory for sequences? SUBJECTS
AND METHODOLOGY
The tests described below were administered to 82 right-handed subjects; 25 were controls (C); 28 had unilateral cerebral damage with aphasia (A+); 29 had unilateral cerebral damage without aphasia (A-). The controls were hospitalized patients with neurological disease below the level of the neck. Brain-damaged subjects were tested for aphasia by a modified version of the Boston Diagnostic Examination for Aphasia (Goodglass & Kaplan, 1972). Causes of brain damage were vascular 32, trauma 17, tumor 8, equally distributed across groups. Mean ages in years were C, 44.4; A+, 46.5; A-, 42.9. Mean educational level in years were C, 12; A+, 12; A-, 10. Time in months from onset of illness to date of testing was A+, 10.1; A-, 6.5. EXPERIMENT
I
Methodology In the first study, a modified version of the Auditory Sequencing Test was used (Albert, 1972). Subject and examiner sat facing each other. Between them was a table on which 18common objects (comb, pen, etc.) were placed randomly. The examiner ascertained that the subject could correctly indicate each item individually in response to verbal command. The examiner then asked the subject to point to four items in a predetermined, randomized sequence (e.g., “In this order, point to the A, the B, the C, and the D”). Three trials with four different items for each trial were given. AU responses, correct or incorrect, were recorded for analysis.
Results Table 1 shows performance of subject groups.2 For retention of total item information there was no significant difference between nonaphasic 2 Statistical analyses employed z tests of the significance of the difference between two proportions; a standardized, normal variable for an upper tail probability was computed (Wallis & Roberts, 1972). The proportions used in these comparisons are recorded in the tables and represent ratios of items correct to total items; items correct in correct sequences to total items; item errors to total items; sequences incorrect to total sequences; sequences incorrect with sequencing errors to total incorrect sequences; and sequences incorrect with omission errors to total incorrect sequences.
30
MARTIN
L. ALBERT
TABLE 1 FOUR-ITEM SEQUENCINGTESP
Groups
Number
Items correct (Total items)
Items correct in correct sequences (Total items)
Sequences incorrect with sequencing errors (Total incorrect sequences)”
Sequences incorrect with omission errors (Total incorrect sequences)
C
25
282 zii
240 Gii
3 Is
ii
A-
29
318 348
248 348
5 2s
14 2s
A+
28
201 336
24 336
21 62
42 ii
3
a C = controls; A- = nonaphasic brain-damaged subjects; A+ = aphasics. b Responses containing two or more correct items.
brain-damaged subjects and controls (z = 0.05, z cut-off for significance = 1.96). However, aphasics performed significantly worse than did nonaphasic brain-damaged subjects (z = 2.74,~ < 0.05). More striking was the fact that retention of order information (total items correct in correct sequences) was selectively and significantly impaired among aphasics (A+ vs. A-: z = 4.09, p < 0.01). Actual performance of each subject had been noted; and error analyses were carried out. At no time did a subject fail to respond. Responses included from zero to four correct items. Errors were either substitutions (the subject pointed to an item not requested), incorrect sequences of correct items, or omissions (failure to provide an item requested, without substitution). Controls made equal amounts of omission and sequencing errors (20% of errors were of each type). For each brain-damaged group, omission errors were the most prominent type; and there was no significant difference between A+ and A- for this type of error (A+ 54%, A- 56%). Sequencing-type errors, however, in aphasics reached 34%, while in controls and nonaphasic brain-damaged subjects, they accounted for 20% each of all errors. At the level of four-item responses (that is, for responses in which four items were indicated) the difference between A+ and A- for sequencing errors was significant (z = 3.68, p < 0.05, z cut-off for significance = 1.64). For responses in which three items were indicated, the difference between A+ and A- for sequencing errors was not significant, but was nearly so (z = 1.31). The evidence suggests that aphasics, unlike normals or brain-damaged nonaphasics, have significantly impaired auditory short term memories both for total verbal item information and for verbal sequences. In
SHORT-TERM
31
MEMORY AND APHASIA
addition, only for the aphasics, the influence of the defective memory for sequences becomes more pronounced as information load increases. For further consideration of these issues an additional test was administered. EXPERIMENT
II
Methodology The structural format of the preceding test was retained. However, for this test the examiner named two objects. Subjects were required to point to those two items, in the order named, from a group of 18 objects placed before them. Five trials were given, with two different objects for each trial.
Results
Table 2 shows the performance of subject groups. Aphasics made significantly more’total errors (24%) than did the normals (2%) or nonaphasic brain-damaged subjects (2.5%). (A+ vs. A- ,z = 2.29, p < 0.05, z cut-off for significance = 1.64). There was no significant difference between controls and nonaphasic brain-damaged subjects for total errors (z = 0.03). Error analysis revealed that the significant difference in total errors between aphasics and the other two groups was not due to sequencing-type errors, but rather was due primarily to omissions. At the level of two items, there was no significant difference between aphasics and others for sequencing-type errors. Additional analyses, concerned with anatomical localization and clinical correlations, were performed. Evaluation of lesion site was made by neurological examination, electroencephalography, and radioactive gamma encephalography in all cases; and cerebral arteriography in most TABLE 2 TWO-ITEM SEQUENCING TEST RESULTS”
Groups Number
Total item errors (Total items)
Sequences incorrect (Total sequences)
Sequences incorrect with sequencing errors (Total incorrect sequences)
Sequences incorrect with omission errors (Total incorrect sequences)
C
21
4 210
6 los
2 -ii
3 --ii
A-
27
7 270
- II 135
4 11
4 ii
A+
22
52 220
52 ii6
12 52
30 s2
a C = controls; A- = nonaphasic brain-damaged subjects; A+ = aphasics.
32
MARTIN
L. ALBERT
cases. Every aphasic had a lesion in the left hemisphere involving some portion (anterior, posterior, or both) of the perisylvian region, the anatomical region defined by Dejerine (1914) as the zone of language. A lesion anywhere within this anatomical region produced a defective auditory verbal short term memory. The aphasic group was further evaluated according to the clinical classification of anterior vs. posterior aphasia (Geschwind, 1970; Goodglass & Kaplan, 1972). On both the two-item and the four-item tests, posterior aphasics made more total errors (for items) than did anterior aphasics (two-item test total errors: anterior l%, posterior 2%; four-item test total errors: anterior 36%, posterior 45%). However, these differences were not statistically significant. Error analysis revealed no significant differences between anterior and posterior aphasics for sequencing-type errors on either the two-item or four-item test (two-item test sequencing errors: anterior = 2% of total errors, posterior = 17%; four-item test sequencing errors: anterior = 33%, posterior = 34%). Impaired short-term memory for items and impaired short-term memory for sequences were present regardless of clinical type of aphasia. DISCUSSION
Two questions were asked at the beginning of this study: (1) Can memory for sequences be separated from memory for items? (2) Do aphasics have a general defect in auditory verbal short-term memory, or do they have a selective defect in short-term memory for sequences? The answer to the first question is “yes.” Evidence for this was provided by analyses of errors, which were of three types: substitution, omission, and sequencing errors. These results, obtained from a large group of brain-damaged subjects, are consistent with those of Tzortzis and Albert (1974) who made similar observations after studying a small, selected group of aphasics. Both studies demonstrated that aphasic subjects can frequently remember all of the items in a given set, but not the order of the items in the set. As for the second question, the answer is “both.” Aphasics, unlike normals and nonaphasic brain-damaged subjects, have a sign&ant defect in auditory verbal short-term memory for total items regardless of the total information load. Aphasics also have a selective defect in memory for sequences; this selective defect, however, has special characteristics. At low information load levels, the major form of memory defect in aphasics is an “omission” type. As information load increases, defective memory for sequences becomes a critical factor in the linguistic performance of aphasics. REFERENCES Albert,
M. L. 1972. Certains aspects des troubles de la comprehension Langages, 7,37-51.
auditive du langage.
SHORT-TERM
MEMORY AND APHASIA
33
Albert, M. L. 1972. Auditory sequencing and left cerebral dominance for language. Neuropsychologia, 10, 245-248. Dejerine, J. 1914. Semeiologie des affections du systkme nerveux, Paris: Masson. Efron, R. 1%3. Temporal perception, aphasia, and dijti vu. Brain, 86,403-424. Geschwind, N. 1970. Aphasia. Science, 170, 940. Goodglass, H., & Kaplan, E. 1972. The Assessment of Aphasia and Related Disorders, Philadelphia: Lea & Febiger. Goodglass, H., Gleason, J., & Hyde, M. 1970. Some dimensions of auditory language comprehension in aphasia. Journal of Speech and Hearing Research, 13,595-606. Hirsch, I. 1%7. Information processing in input channels for speech and language: The significance of serial order of stimuli. In C. Millikan & F. Darley (Eds.), Brain mechanisms underlying speech and language. New York: Grune and Stratton. Tzortzis, C., & Albert, M. L. 1974. Impairment of memory for sequences in conduction aphasia. Neuropsychologia, 12, 355-366. Wallis, W., & Roberts, H. 1972. Statistics, a new approach, Glencoe, Ill.: Free Press. Warrington, E., & Shallice, T. 1%9. The selective impairment of auditory verbal short-term memory. Brain, 92, 885-896.
