JOURNAL
OF COMMUNICATION
DISORDERS
9 (1976), 167-173
EFFECTS OF TASK DIFFICULTY ON SENTENCE COMPREHENSION PERFORMANCE OF APHASIC SUBJECTS ROBERT Director,
Aphasia Section,
H. BROOKSHIRE
Veterans Administration
Hospital.
Minneapolis,
Minnesota
55417
Twenty aphasic subjects participated in an auditory comprehension task in which they were asked to point to pictures named by the examiner. The spoken commands delivered to each subject were either “easy” or “difficult,” as determined by the subject’s performance in a preexperimental session. Difficulty was determined by the number of items that the subject was asked to point to in sequence. Each subject performed in two conditions. In the Low Error Condition, a small number of difficult commands (as determined in the preexperimental session) was interspersed in a larger number of easy commands (as determined in the preexperimental session). In the High Error Condition, a small number of easy commands was interspersed in a larger number of difficult commands. Analysis of the results indicated that presentation of a series of difficult commands interfered with subject’s subsequent performance to commands that would ordinarily be very easy.
Introduction In a previous study, Brookshire (1972) reported the presence of a “task difficulty effect” when aphasic patients were asked to name pictures of common objects. To demonstrate the effect, he chose a set of easy-to-name pictures and a set of difficult-to-name pictures for each subject, based upon the subject’s performance in naming those pictures in a preexperimental session. Each subject then attempted to name the pictures in two conditions. In one condition, a small number of trials with difficult-to-name items was interspersed within a larger number of easy-to-name items. In a second condition, a small number of easy-to-name items was interspersed in a larger number of difficult-to-name items. The results of the study indicated that error responses to difficult-to-name items appeared to degrade subjects’ performance on subsequent items. The inverse was also true; that is, correct responses to easy-to-name items appeared to facilitate performance on subsequent items. Brookshire concluded that the occurrence of error responses generates effects that affect subsequent performance. He also concluded that correct responses may generate effects that can facilitate subsequent performance. The effects described by Brookshire appear to be related to effects described in a number of experiments in which the effects of errors on performance in learning tasks have been evaluated (Terrace, 1963a, 1963b; Stoddard and Sidman, 1967; Moore and Goldiamond, 1964; Touchette, 1968). In general, these experiments have demonstrated that the occurrence of error responses in learning tasks frequently interferes with the progress of learning in the task. This interference may 0 American
Elsevier Publishing
Company,
Inc.,
1976
167
168
ROBERT
H. BROOKSHIRE
be generated by reinforcement of responses that are incompatible with the correct responses (Sidman and Stoddard, 1967), by intrusion of responses that have been accidentally reinforced during training (Hively, 1962), or by introducing “extinction trials, ’ ’ in which responding is not reinforced, causing the overall rate of responding to decline (Moore and Goldiamond, 1964). Brookshire demonstrated that errors affected subsequent performance in a word-retrieval and production task. Several other investigators have demonstrated similar effects in other tasks with other kinds of subjects. It seems reasonable to expect that the effect should not be limited to retrieval and production by aphasic subjects, but should also be found in other kinds of speech and language tasks. The study reported here was designed to determine whether errors generate similar effects in tasks requiring responses to spoken commands. Responding to spoken messages was selected because of its importance in general language abilities, and because of the emphasis that has been placed upon responses to auditory messages in the rehabilitation of aphasic individuals (Schuell et al., 1965). The experimental questions considered were (a) If aphasic subjects are placed in a task in which they are required to respond to complex spoken commands, will error responses to these commands interfere with subsequent performance on simpler spoken commands? (b) If aphasic patients are placed in a task in which they are required to respond to simple spoken commands, will correct responses to these commands facilitate performance on more complex commands?
Method Subjects were 20 patients diagnosed as aphasic by the Porch Index of Communicative Ability (Porch, 1967), the Token Test (DeRenzi and Vignolo, 1962), and other measures. The subjects represented a number of types and degrees of aphasic involvements. They ranged in age from 23 to 69 years. Time since onset of aphasia ranged from 3 months to 5 years. Seventeen were aphasic as a result of cerebrovascular accident; three were aphasic subsequent to trauma. Subjects were selected from the treatment roster of the Aphasia Section, Minneapolis Veterans Administration Hospital, on the basis of their performance in a preexperimental task. In this preexperimental task (henceforth called Baserate), the subject was seated at a table in a sound treated booth. A sheet containing 10 pictures of common objects (within the first 4,000 most frequently occurring words according to Thomdike and Lorge, 1944) was placed on the table in front of the subject. The same set of 10 pictures was used with all subjects. The experimenter sat to the subject’s right and operated a tape recorder that delivered a series of spoken ” The directions differed in directions, of the form “Point to the length. Level I was the easiest level, and consisted of single unit commands, such as “Point to the shoe.” Level II contained two-unit commands, such as “Point to
TASK DIFFICULTY
EFFECTS
ON SENTENCE
COMPREHENSION
169
the dog and the man.” Level III contained commands such as “Point to the shoe, the boy, and the door.” Likewise, Level IV contained four-unit commands, and Level V, five-unit commands, in the manner of the lower levels. The tape played in the Baserate session contained 50 commands (10 from each of Levels I through V), randomly distributed on the tape. After the subject was instructed concerning what he was to do in the task, and had performed several simple examples, the Baserate tape was played, and his response to each command was recorded. The subject was allowed 15 set to respond to each command in the Baserate session. Following each 15-see interval, the pictures were covered for 5 set; then the pictures were uncovered and the next command was given. When the Baserate tape was completed, the subject was dismissed and given an appointment for the next session. Following the Baserate session, each subject’s responses to the items on the Baserate tape were tabulated, in order to establish two task levels for each patient: a Basal Level and a Ceiling Level. A subject’s Basal Level was defined as the highest level at which no errors (on items or order of items) were made. If the subject did not perform errorlessly at any level, then his Basal Level was defined as the highest level at which he made S/IO correct responses. A subject’s Ceiling Level was defined as that level at which he made all errors. If he did not make all errors at any level, his Ceiling Level was defined as the lowest level at which he committed S/IO errors. If he failed to generate either Basal or Ceiling Levels according to the above criteria, he was not included in the experiment. Each subject who generated Basal and Ceiling Levels in the Baserate session subsequently performed in two experimental conditions: Low Error Condition and High Error Condition. In the Low Error Condition, each subject heard a tape that contained 40 Basal items (as determined by his performance on the Baserate tape) within which were randomly’ interspersed 10 Ceiling items. In the High Error Condition, each subject heard a tape that contained 40 Ceiling items (determined by his performance on the Baserate tape) within which were randomly interspersed 10 Basal items. In this manner, experimental tapes were assigned to each subject, based on his performance in Baserate. In the experimental sessions, half the subjects received the Low Error Condition first and half received the High Error Condition first. Time between successive sessions ranged from 1 day to 1 week, and depended on the subject’s clinical treatment schedule. All subjects less than 6 months post onset were seen on successive days. Only neurologically stable subjects were seen at greater between-session intervals. Stimulus presentation and response recording for Easy and Difficult conditions were the same as in the Baserate session. ‘The distribution items intervened
was not completely random, because items were distributed so that at least 3 of the 40 between successive members of the 10 items.
9.30
7.25
Treatment.
2.05
41
9 145
186
u N = 10 for Basal items in Baserate, Treatment, and Ceiling items in Baserate, h A = Low Error Condition first, B = High Error Condition first.
Mean
Total
2 1 4 3 3 1 1 1 2 1 5 2 1 3 2 7 1 2 -1 0
Difference
8 8 6 I 6 9 9 9 8 8 1 6 8 7 8 3 8 8 9
II I I II II 0 II I III I III II I II II III II II II III
A B A B A B A B A B A B B B A B A B A A
1 2 3 4 5 6 7 8 9 10 11 12 13 14 1.5 16 17 18 19 20
High error condition V IV III V V IV V III V IV V V III V V V V V V V
Level
0.60 __~
1.35
27
0 9 1 1 1 0 3 0 0 0 0 0 5 1 0 0 1 0 2 3
1 0 0 2 0 0 1 0 0 1 0 0 0 2 0 1 1 1 0 2 12
Low error condition
Baserate
~__
TABLE 1 to Basal Items and Ceiling Items by Aphasic Subjects in Baserate and Treatmenta
10 9 10 10 9 10 10 10 10 9 6 8 9 10 10 10 9 10 8 9
Baserate
Level
Order h
Subj.
Number of Correct Responses
1
1
-0.75
-15
0 -2 0 1 1 0 0 -5 1 0 1 0 1 -2 -1
-1
-9 -1
_
Difference
TASK DIFFICULTY
EFFECTS
ON SENTENCE
COMPREHENSION
171
Results The results are summarized in Table 1 and Fig. 1. Figure 1 suggests that subjects’ performance on Basal Level (easy) items was considerably better in Baserate than it was in Treatment (within experimental conditions), when those Basal Level items were interspersed among more difficult items that generated high error rates. Figure 1 also suggests that subjects’ performance on Ceiling Level items in Treatment was not substantially different from their performance on those items in Baserate. In order to determine whether the differences in performance between Baserate and Treatment on Basal Level and Ceiling Level items were significant, Wilcoxen Matched-Pairs Signed-Rank tests were computed on the data. The results of these tests indicated that performance on Basal Level items was significantly poorer in Treatment than it was in Baserate (d = 19, T = 15.5, p < O.OOl), and that performance on Ceiling Level items in Baserate was not significantly different from performance on those items in Treatment (d = 11, T = 15,~ > 0.05). Table 1 shows that the decrement in performance on Basal Level items from Baserate to Treatment was generally small for each subject, but consistent across subjects. Three subjects (Ss 3, 11,16) demonstrated decrements of 40% or more in correct responding between Baserate and Treatment. Changes in performance on Ceiling Level items from Baserate to Treatment were not so consistent across subjects. Although 16 of 20 subjects demonstrated differences of 10% or less
o,fasal
Items
Cai I i ng Items / .-
Base-rata Treatment Session Fig.
1. Mean number of correct responses to easy and difficult items in Baserate and Treatment sessions.
172
ROBERT
H. BROOKSHIRE
between Baserate and Treatment, two subjects (Ss 2, 13) showed substantial inprovement on Ceiling Level items when they were interspersed within easier items. Neither of these latter subjects had demonstrated substantial differences in performance between Baserate and Treatment on Basal Level items. Discussion The results of this study suggest that errors committed by aphasic subjects in following spoken directions tend to generate additional errors on subsequent items. This was true both for subjects as a group and for individual subjects. These results replicate a similar effect found in a previous study of the effects of errors in a picture-naming task (Brookshire, 1972). These results suggest that programs utilized in treatment of aphasic patients should endeavor to keep error rates low, in the manner suggested by a number of investigators (Hively, 1962; Terrace, 1963b; Stoddard and Sidman, 1967). We did not find, in this study, that presentation of a series of items which generate low error rates facilitated performance on a subsequent more difficult item, at least for the group of aphasic subjects as a whole. We did find that two of the subjects demonstrated remarkable improvement in their performance on items that were ordinarily quite difficult for them, when those items were interspersed in a series of extremely easy items. Interestingly, neither of these subjects was a subject who had exhibited large decrements in performance on easy items that were interspersed in more difficult items. However, in the absence of a significant group effect for Ceiling items between Baserate and Treatment, we shall not discuss the performance of these two subjects at length, except to note that it may be that some aphasic subjects may be more amenable to facilitation of performance by easy items than others are. In any case, our studies suggest that errors committed by aphasic patients in speech and language tasks generate effects that are likely to interfere with performance on trials that follow the errors. Because errors generate more errors, it behooves the clinician to keep error rates low, and when errors do occur, to provide time for the effects of the errors to dissipate. This investigation was supported by the Research and Education Committee, Minneapolis Veterans Administration Hospital. Jean Dale assisted in the collection of data. References Brookshire, R. H. Effects of task difficulty on naming performance of aphasic subjects. J. Speech Hearing Res., 1972, 15, 551-558. DeRenzi, E., Vignolo, L. A. The Token Test: A sensitive test to detect receptive disturbances in aphasics. Brain, 1962, 85, 65S678.
TASK DIFFICULTY
EFFECTS
ON SENTENCE
COMPREHENSION
173
Hively, W. Programming stimuli in matching to sample. J. Exp. Anal. Behav. 1962, 5, 279-288. Moore, R., Goldiamond, I. Errorless establishment of visual discrimination using fading procedures. J. Exp. Anal. Behm., 1964, 7, 269272. Porch, B. Porch index ofcommunicative ability. Palo Alto, Calif.: Consulting Psychologists Press, 1967. SchuelI, H., Jenkins, J., Jimenez-Pabon, E. Aphasia in adults: diagnosis, prognosis, and treatment. New York: Harper and Row, 1965. Sidman, M., Stoddard, L. T. The effectiveness of fading in programming a simultaneous form discrimination for retarded children. J. Exp. Anal. Behav., 1967, 10, 3-15. Stoddard, L. T., Sidman, M. The effects of errors on childrens’ performance on a circle-ellipse discrimination. J. Exp. Anal. Behav., 1967, 10, 261-270. Terrace, H. S. Discrimination learning with and without “errors.” J. Exp. Anal. Behav., 1963a, 6, l-27. Terrace, H. S. Errorless transfer of a discrimination across two continua. J. Exp. Anal. Behav., 1963b, 11, 3948. Thomdike, E. L., Large, I. The teacher’s word book of 30,000 words. New York: Columbia University Press, 1944. Touchette, P. E. The effects of graduatedstimulus change on the acquisitionof asimplediscrimination in severely retarded boys. J. Exp. Anal. Behav., 1968, 11, 3948.
OF COMMUNICATION
DISORDERS
9 (1976), 167-173
EFFECTS OF TASK DIFFICULTY ON SENTENCE COMPREHENSION PERFORMANCE OF APHASIC SUBJECTS ROBERT Director,
Aphasia Section,
H. BROOKSHIRE
Veterans Administration
Hospital.
Minneapolis,
Minnesota
55417
Twenty aphasic subjects participated in an auditory comprehension task in which they were asked to point to pictures named by the examiner. The spoken commands delivered to each subject were either “easy” or “difficult,” as determined by the subject’s performance in a preexperimental session. Difficulty was determined by the number of items that the subject was asked to point to in sequence. Each subject performed in two conditions. In the Low Error Condition, a small number of difficult commands (as determined in the preexperimental session) was interspersed in a larger number of easy commands (as determined in the preexperimental session). In the High Error Condition, a small number of easy commands was interspersed in a larger number of difficult commands. Analysis of the results indicated that presentation of a series of difficult commands interfered with subject’s subsequent performance to commands that would ordinarily be very easy.
Introduction In a previous study, Brookshire (1972) reported the presence of a “task difficulty effect” when aphasic patients were asked to name pictures of common objects. To demonstrate the effect, he chose a set of easy-to-name pictures and a set of difficult-to-name pictures for each subject, based upon the subject’s performance in naming those pictures in a preexperimental session. Each subject then attempted to name the pictures in two conditions. In one condition, a small number of trials with difficult-to-name items was interspersed within a larger number of easy-to-name items. In a second condition, a small number of easy-to-name items was interspersed in a larger number of difficult-to-name items. The results of the study indicated that error responses to difficult-to-name items appeared to degrade subjects’ performance on subsequent items. The inverse was also true; that is, correct responses to easy-to-name items appeared to facilitate performance on subsequent items. Brookshire concluded that the occurrence of error responses generates effects that affect subsequent performance. He also concluded that correct responses may generate effects that can facilitate subsequent performance. The effects described by Brookshire appear to be related to effects described in a number of experiments in which the effects of errors on performance in learning tasks have been evaluated (Terrace, 1963a, 1963b; Stoddard and Sidman, 1967; Moore and Goldiamond, 1964; Touchette, 1968). In general, these experiments have demonstrated that the occurrence of error responses in learning tasks frequently interferes with the progress of learning in the task. This interference may 0 American
Elsevier Publishing
Company,
Inc.,
1976
167
168
ROBERT
H. BROOKSHIRE
be generated by reinforcement of responses that are incompatible with the correct responses (Sidman and Stoddard, 1967), by intrusion of responses that have been accidentally reinforced during training (Hively, 1962), or by introducing “extinction trials, ’ ’ in which responding is not reinforced, causing the overall rate of responding to decline (Moore and Goldiamond, 1964). Brookshire demonstrated that errors affected subsequent performance in a word-retrieval and production task. Several other investigators have demonstrated similar effects in other tasks with other kinds of subjects. It seems reasonable to expect that the effect should not be limited to retrieval and production by aphasic subjects, but should also be found in other kinds of speech and language tasks. The study reported here was designed to determine whether errors generate similar effects in tasks requiring responses to spoken commands. Responding to spoken messages was selected because of its importance in general language abilities, and because of the emphasis that has been placed upon responses to auditory messages in the rehabilitation of aphasic individuals (Schuell et al., 1965). The experimental questions considered were (a) If aphasic subjects are placed in a task in which they are required to respond to complex spoken commands, will error responses to these commands interfere with subsequent performance on simpler spoken commands? (b) If aphasic patients are placed in a task in which they are required to respond to simple spoken commands, will correct responses to these commands facilitate performance on more complex commands?
Method Subjects were 20 patients diagnosed as aphasic by the Porch Index of Communicative Ability (Porch, 1967), the Token Test (DeRenzi and Vignolo, 1962), and other measures. The subjects represented a number of types and degrees of aphasic involvements. They ranged in age from 23 to 69 years. Time since onset of aphasia ranged from 3 months to 5 years. Seventeen were aphasic as a result of cerebrovascular accident; three were aphasic subsequent to trauma. Subjects were selected from the treatment roster of the Aphasia Section, Minneapolis Veterans Administration Hospital, on the basis of their performance in a preexperimental task. In this preexperimental task (henceforth called Baserate), the subject was seated at a table in a sound treated booth. A sheet containing 10 pictures of common objects (within the first 4,000 most frequently occurring words according to Thomdike and Lorge, 1944) was placed on the table in front of the subject. The same set of 10 pictures was used with all subjects. The experimenter sat to the subject’s right and operated a tape recorder that delivered a series of spoken ” The directions differed in directions, of the form “Point to the length. Level I was the easiest level, and consisted of single unit commands, such as “Point to the shoe.” Level II contained two-unit commands, such as “Point to
TASK DIFFICULTY
EFFECTS
ON SENTENCE
COMPREHENSION
169
the dog and the man.” Level III contained commands such as “Point to the shoe, the boy, and the door.” Likewise, Level IV contained four-unit commands, and Level V, five-unit commands, in the manner of the lower levels. The tape played in the Baserate session contained 50 commands (10 from each of Levels I through V), randomly distributed on the tape. After the subject was instructed concerning what he was to do in the task, and had performed several simple examples, the Baserate tape was played, and his response to each command was recorded. The subject was allowed 15 set to respond to each command in the Baserate session. Following each 15-see interval, the pictures were covered for 5 set; then the pictures were uncovered and the next command was given. When the Baserate tape was completed, the subject was dismissed and given an appointment for the next session. Following the Baserate session, each subject’s responses to the items on the Baserate tape were tabulated, in order to establish two task levels for each patient: a Basal Level and a Ceiling Level. A subject’s Basal Level was defined as the highest level at which no errors (on items or order of items) were made. If the subject did not perform errorlessly at any level, then his Basal Level was defined as the highest level at which he made S/IO correct responses. A subject’s Ceiling Level was defined as that level at which he made all errors. If he did not make all errors at any level, his Ceiling Level was defined as the lowest level at which he committed S/IO errors. If he failed to generate either Basal or Ceiling Levels according to the above criteria, he was not included in the experiment. Each subject who generated Basal and Ceiling Levels in the Baserate session subsequently performed in two experimental conditions: Low Error Condition and High Error Condition. In the Low Error Condition, each subject heard a tape that contained 40 Basal items (as determined by his performance on the Baserate tape) within which were randomly’ interspersed 10 Ceiling items. In the High Error Condition, each subject heard a tape that contained 40 Ceiling items (determined by his performance on the Baserate tape) within which were randomly interspersed 10 Basal items. In this manner, experimental tapes were assigned to each subject, based on his performance in Baserate. In the experimental sessions, half the subjects received the Low Error Condition first and half received the High Error Condition first. Time between successive sessions ranged from 1 day to 1 week, and depended on the subject’s clinical treatment schedule. All subjects less than 6 months post onset were seen on successive days. Only neurologically stable subjects were seen at greater between-session intervals. Stimulus presentation and response recording for Easy and Difficult conditions were the same as in the Baserate session. ‘The distribution items intervened
was not completely random, because items were distributed so that at least 3 of the 40 between successive members of the 10 items.
9.30
7.25
Treatment.
2.05
41
9 145
186
u N = 10 for Basal items in Baserate, Treatment, and Ceiling items in Baserate, h A = Low Error Condition first, B = High Error Condition first.
Mean
Total
2 1 4 3 3 1 1 1 2 1 5 2 1 3 2 7 1 2 -1 0
Difference
8 8 6 I 6 9 9 9 8 8 1 6 8 7 8 3 8 8 9
II I I II II 0 II I III I III II I II II III II II II III
A B A B A B A B A B A B B B A B A B A A
1 2 3 4 5 6 7 8 9 10 11 12 13 14 1.5 16 17 18 19 20
High error condition V IV III V V IV V III V IV V V III V V V V V V V
Level
0.60 __~
1.35
27
0 9 1 1 1 0 3 0 0 0 0 0 5 1 0 0 1 0 2 3
1 0 0 2 0 0 1 0 0 1 0 0 0 2 0 1 1 1 0 2 12
Low error condition
Baserate
~__
TABLE 1 to Basal Items and Ceiling Items by Aphasic Subjects in Baserate and Treatmenta
10 9 10 10 9 10 10 10 10 9 6 8 9 10 10 10 9 10 8 9
Baserate
Level
Order h
Subj.
Number of Correct Responses
1
1
-0.75
-15
0 -2 0 1 1 0 0 -5 1 0 1 0 1 -2 -1
-1
-9 -1
_
Difference
TASK DIFFICULTY
EFFECTS
ON SENTENCE
COMPREHENSION
171
Results The results are summarized in Table 1 and Fig. 1. Figure 1 suggests that subjects’ performance on Basal Level (easy) items was considerably better in Baserate than it was in Treatment (within experimental conditions), when those Basal Level items were interspersed among more difficult items that generated high error rates. Figure 1 also suggests that subjects’ performance on Ceiling Level items in Treatment was not substantially different from their performance on those items in Baserate. In order to determine whether the differences in performance between Baserate and Treatment on Basal Level and Ceiling Level items were significant, Wilcoxen Matched-Pairs Signed-Rank tests were computed on the data. The results of these tests indicated that performance on Basal Level items was significantly poorer in Treatment than it was in Baserate (d = 19, T = 15.5, p < O.OOl), and that performance on Ceiling Level items in Baserate was not significantly different from performance on those items in Treatment (d = 11, T = 15,~ > 0.05). Table 1 shows that the decrement in performance on Basal Level items from Baserate to Treatment was generally small for each subject, but consistent across subjects. Three subjects (Ss 3, 11,16) demonstrated decrements of 40% or more in correct responding between Baserate and Treatment. Changes in performance on Ceiling Level items from Baserate to Treatment were not so consistent across subjects. Although 16 of 20 subjects demonstrated differences of 10% or less
o,fasal
Items
Cai I i ng Items / .-
Base-rata Treatment Session Fig.
1. Mean number of correct responses to easy and difficult items in Baserate and Treatment sessions.
172
ROBERT
H. BROOKSHIRE
between Baserate and Treatment, two subjects (Ss 2, 13) showed substantial inprovement on Ceiling Level items when they were interspersed within easier items. Neither of these latter subjects had demonstrated substantial differences in performance between Baserate and Treatment on Basal Level items. Discussion The results of this study suggest that errors committed by aphasic subjects in following spoken directions tend to generate additional errors on subsequent items. This was true both for subjects as a group and for individual subjects. These results replicate a similar effect found in a previous study of the effects of errors in a picture-naming task (Brookshire, 1972). These results suggest that programs utilized in treatment of aphasic patients should endeavor to keep error rates low, in the manner suggested by a number of investigators (Hively, 1962; Terrace, 1963b; Stoddard and Sidman, 1967). We did not find, in this study, that presentation of a series of items which generate low error rates facilitated performance on a subsequent more difficult item, at least for the group of aphasic subjects as a whole. We did find that two of the subjects demonstrated remarkable improvement in their performance on items that were ordinarily quite difficult for them, when those items were interspersed in a series of extremely easy items. Interestingly, neither of these subjects was a subject who had exhibited large decrements in performance on easy items that were interspersed in more difficult items. However, in the absence of a significant group effect for Ceiling items between Baserate and Treatment, we shall not discuss the performance of these two subjects at length, except to note that it may be that some aphasic subjects may be more amenable to facilitation of performance by easy items than others are. In any case, our studies suggest that errors committed by aphasic patients in speech and language tasks generate effects that are likely to interfere with performance on trials that follow the errors. Because errors generate more errors, it behooves the clinician to keep error rates low, and when errors do occur, to provide time for the effects of the errors to dissipate. This investigation was supported by the Research and Education Committee, Minneapolis Veterans Administration Hospital. Jean Dale assisted in the collection of data. References Brookshire, R. H. Effects of task difficulty on naming performance of aphasic subjects. J. Speech Hearing Res., 1972, 15, 551-558. DeRenzi, E., Vignolo, L. A. The Token Test: A sensitive test to detect receptive disturbances in aphasics. Brain, 1962, 85, 65S678.
TASK DIFFICULTY
EFFECTS
ON SENTENCE
COMPREHENSION
173
Hively, W. Programming stimuli in matching to sample. J. Exp. Anal. Behav. 1962, 5, 279-288. Moore, R., Goldiamond, I. Errorless establishment of visual discrimination using fading procedures. J. Exp. Anal. Behm., 1964, 7, 269272. Porch, B. Porch index ofcommunicative ability. Palo Alto, Calif.: Consulting Psychologists Press, 1967. SchuelI, H., Jenkins, J., Jimenez-Pabon, E. Aphasia in adults: diagnosis, prognosis, and treatment. New York: Harper and Row, 1965. Sidman, M., Stoddard, L. T. The effectiveness of fading in programming a simultaneous form discrimination for retarded children. J. Exp. Anal. Behav., 1967, 10, 3-15. Stoddard, L. T., Sidman, M. The effects of errors on childrens’ performance on a circle-ellipse discrimination. J. Exp. Anal. Behav., 1967, 10, 261-270. Terrace, H. S. Discrimination learning with and without “errors.” J. Exp. Anal. Behav., 1963a, 6, l-27. Terrace, H. S. Errorless transfer of a discrimination across two continua. J. Exp. Anal. Behav., 1963b, 11, 3948. Thomdike, E. L., Large, I. The teacher’s word book of 30,000 words. New York: Columbia University Press, 1944. Touchette, P. E. The effects of graduatedstimulus change on the acquisitionof asimplediscrimination in severely retarded boys. J. Exp. Anal. Behav., 1968, 11, 3948.
