Journal ol Personality and Social Psychology 1976, Vol. 33, No. 5, 508-516
Learned Helplessness, Depression, and the Attribution of Failure David C. Klein, Ellen Fencil-Morse, and Martin E. P. Seligman University of Pennsylvania Depressed and nondepressed college students received experience with solvable, unsolvable, or no discrimination problems. When later tested on a series of patterned anagrams, depressed groups performed worse than nondepressed groups, and unsolvable groups performed worse than solvable and control groups. As predicted by the learned helplessness model of depression, nondepressed subjects given unsolvable problems showed anagram deficits parallel to those found in naturally occurring depression. When depressed subjects attributed their failure to the difficulty of the problems rather than to their own incompetence, performance improved strikingly. So, failure in itself is apparently not sufficient to produce helplessness deficits in man, but failure that leads to a decreased belief in personal competence is sufficient.
Dogs given inescapable shock in a hammock subsequently fail to escape shock in a shuttle box. Even if they make an occasional response that turns off shock, they fail to learn. In contrast, dogs given escapable shock or no prior shock escape well (Overmier & Seligman, 1967; Seligman & Maier, 1967). This interference and the process underlying it have been called "learned helplessness" (Maier, Seligman, & Solomon, 1969; Seligman, Maier, & Solomon, 1971). Learned helplessness occurs in a variety of situations, with a variety of uncontrollable events, and across a number of species, including rats, cats, mice, and men (see Seligman, 1975, for a comprehensive review). The two major behavioral symptoms of learned helplessness are (a) motivational— lowered response initiation, and (b) cognitive—a dampened ability to learn that responding produces reinforcement. Seligman et al. (1971) argued that these symptoms
The research reported in this paper was supported by the following National Institute of Mental Health grants: MH19604 to Martin E. P. Seligman, MH08209 to Julius Wishner, and MH19989 to Aaron T. Beck. The authors wish to thank D. R. Hiroto, W. R. Miller, J. Rodin, and S. Roth for their comments at various stages of the preparation of the paper. Requests for reprints should be sent to Martin E. P. Seligman, Department of Psychology, University of Pennsylvania, 3813-1S Walnut Street, Philadelphia, Pennsylvania 19174.
result when the subject learns that his/her responses are independent of reinforcement. The expectancy that responding will be ineffective reduces the incentive to initiate instrumental responses and disrupts later learning of response-reinforcement contingencies. Recently, investigators have attempted to produce learned helplessness in man. Hiroto and Seligman (1975), Miller and Seligman (1975), and Racinskas (Note 1) provided demonstrations of helplessness in man, which are parallel to the deficits that occur in animals. In these studies, subjects were first exposed to uncontrollable situations, that is, responding and reinforcement were independent. Subjects given inescapable shock, inescapable noise, or unsolvable discrimination problems later failed to escape shock or noise, solved anagrams very poorly, and failed to learn from their successes. In addition, Miller and Seligman (in press) showed that inescapable noise led normal subjects to perceive that subsequent responding and reinforcement were independent. Thus, both the response initiation deficit and the cognitive deficit that occur in helpless animals also occur in man and seem to be caused by the perception that responses and reinforcement are independent. Learned helplessness has been proposed as a laboratory model of depression in man (Seligman, 1972, 1973, 1974, 1975; Seligman, Klein, & Miller, in press). Learned helpless-
508
LEARNED HELPLESSNESS AND DEPRESSION
ness and depression may be parallel in their symptoms, etiology, cure, and prevention. The model suggests that depression is caused by uncontrollable situations which lead the individual to perceive that his responses are generally ineffective in obtaining reinforcement. If the model is correct, helplessness induced in the laboratory in nondepressed subjects should produce response initiation and cognitive deficits parallel to those shown by depressed subjects without helplessness training. A growing body of evidence supports the relationship of learned helplessness to depression (Miller & Seligman, 1973, 1975, in press; Miller, Seligman, & Kurlander, in press). The experiment below asks two questions bearing on the learned helplessness model of depression: 1. Do unsolvable discrimination problems produce deficits in anagram performance in nondepressed subjects similar to those found in depressed controls? Depressed subjects given no pretreatment and nondepressed subjects given unsolvable discrimination problems should show response initiation and cognitive deficits relative to nondepressed subjects given no pretreatment. Unsolvable discrimination problems produce anagram deficits in randomly selected college students (Hiroto & Seligman, 1975), and inescapable noise produces anagram deficits in nondepressed subjects similar to the deficits found in depreessed controls (Miller & Seligman, 1975). The learned helplessness model makes no explicit prediction concerning the effect of unsolvability on depressed subjects' performance. 2. Does attribution of failure have a role as a mediator of learned helplessness? Are performance deficits produced by failure per se, or does it matter if the individual blames his own incompetence or the harshness of the environment for failure? In particular, is anagram performance different if failure on discrimination problems is attributed to lack of ability rather than to task difficulty? Do depressed people make different attributions about their failures than nondepressed people? The learned helplessness model makes no explicit prediction concerning these questions,
509
but Wortman and Brehm (1975) predicted that blaming incompetence should intensify helplessness, and Lichtenberg (1957) claimed that depression is an expectation of failure coupled with the belief that one's personal defects are to blame for the failure. We manipulated attributions by showing subjects norms that either indicated the task they failed was usually solved by their peers or usually was not. METHOD Subjects Subjects were 81 University of Pennsylvania undergraduates recruited from the subject pool and from advertisements asking for volunteers for a problem-solving experiment. They were assigned to depressed and nondepressed groups according to their scores on the Beck Depression Inventory (BDI), a test of the affective, cognitive, somatic, and motoric symptoms of depression (Beck, 1967). The validity and rationale for the use of the BDI in this type of experiment have been discussed elsewhere (Miller et al., in press). Subjects were recruited until 20 males and 20 females were obtained for each group. Subjects with scores of 9 or above were assigned to the depressed group, while subjects with scores of 8 or below were assigned to the nondepressed group. One depressed subject was discarded after completing the experiment because of experimental error. The final sample comprised 40 depressed and 40 nondepressed subjects. Within the depressed and nondepressed groups, subjects were assigned to one of the following five treatments; (a) solvable problems, (b) control, (c) unsolvable problems with no attribution of failure instructions, (d) unsolvable problems with internal attribution of failure instructions, and (e) unsolvable problems with external attribution of failure instructions. Assignments were random, with the restriction that each group contain four males and four females. Thus, there were 10 groups of eight subjects, five depressed and five nondepressed groups. Within each of the 10 groups, subjects were assigned to the two experimenters (the first two authors) so that each experimenter was given two males and two females from each group. The means and standard deviations of BDI scores for the groups are summarized in Table 1.
Apparatus Treatment task. The treatment task was a series of 40 four-dimensional stimulus patterns adapted from Levine (1966) by Hiroto and Seligman (197S). Each of the four dimensions had two associated values: (a) letter (A or T) ; (b) letter color (shaded or unshaded); (c) letter size (large or small); (d) type of border surrounding letter (circle or square). Each problem consisted of 10 cards. For each card, the subject pointed to one side of the
D. KLEIN, E. FENCIL-MORSE, AND M. SELIGMAN
510
TABLE 1 MEANS AND STANDARD DEVIATIONS or BECK DEPRESSION INVENTORY SCORES Depressed
Nondepressed
Group
M
SD
M
SD
Solvable Control Unsolvable-no instructions Unsolvable-internal instructions Unsolvable-external instructions
16.00 12.75
7.82 4.27
3.38 3.50
2.20 2.93
13.25
4.80
3.00
1.41
14.88
4.02
3.63
3.07
13.00
3.8!)
3.75
2.96
Note. Each group contains 8 subjects.
card and the experimenter said whether that side contained the correct value. In this manner, the subject was to try to identify the correct value for each problem. Test task. The test task was a series of 20 anagrams devised by Hiroto and Seligman (1975). Each of the anagrams, which were selected from a list of five-letter anagrams (Tresselt & Mayzner, 1966), consisted of five letters arranged in a standard sequence: 3-4-2-5-1, that is, the first letter of the solution word was the fifth letter of the anagram, the second letter of the solution word was the third anagram letter, and so on (e.g., ERLKC, OUHLG). Sliding scales. Three sliding scales were used to roughly monitor sadness, nervousness, and anger. Each scale consisted of a question about the subject's present mood (e.g., How sad are you feeling right now?), an 11-point scale ranging between two extremes (e.g., extremely sad, extremely happy), and a movable pointer.1 The pointers were at the midpoints of the scales whenever the subject was given the scales. Procedure The subject was first seated at a desk and administered the BDI and the sliding scales. Treatment phase. All subjects were shown a sample pattern of the treatment task and were read the instructions for the cognitive discrimination problems. These instructions can be found in Hiroto and Seligman (1975). Subjects in the solvable groups were then given four solvable problems; subjects in the unsolvable groups were shown the same cards as the solvable subjects, but there was no solution for any problem. For each 10-card problem, the experimenter said "right" and "wrong" five times each in a random order, ending with "wrong." 2 To ensure that unsolvable subjects would not be able to deduce systematically that no solution existed, the cards were turned faster for them than for the solvable subjects. Subjects in the unsolvable groups heard three variations of instructions designed to manipulate attribution of failure: (a) Those with no attribu-
tion of failure instructions were shown the problems as just described, (b) Subjects receiving internal attribution of failure instructions were shown a bar graph just before beginning the problems. This graph showed that 55% of subjects succeed on all four problems, 30% on three, 5% on two, 9% on one, and 1% fail on all four. While studying the graph, subjects were read the following: "This graph shows the number of problems solved by students in a previous experiment. As you can see, most people are able to get three or four of the problems correct." (c) Subjects receiving external attribution of failure instructions were shown a different bar graph just before beginning the problems. This graph showed that no subjects solve all four problems, 1% solve three, 2% two, 7% one, and 90% fail on all four. As they were shown the graph, subjects were read the following: "This graph shows the number of problems solved by students in a previous experiment. As you can see, the problems are very difficult and almost no one has been able to solve them." At the end of each problem, subjects were asked to name the correct dimension. After responding, solvable subjects were told, "That's the right answer" and unsolvable subjects were told, "That's the wrong answer." No solvable subject failed any of the problems. After the treatment phase, all subjects were given the sliding scales for a second time. Test phase. Subjects were then presented with the anagrams task. They were allowed up to 100 seconds to solve each anagram. Response latencies were measured by the experimenter with a hand-held stopwatch. The instructions for the anagrams can be found in Hiroto and Seligman (1975). Four dependent measures were obtained for each subject on the anagrams: (a) mean response latency (a failure counted as 100 seconds) ; (b) trials to criterion for solving the anagrams pattern, defined as the trial number of the third successive trial with a response latency of no more than 15 seconds (for subjects who failed to reach criterion, this measure was taken as the earliest trial the subject could have reached criterion if given more ana-
1 The authors wish to thank Aaron T. Beck for the design of the sliding scales. 2 Strictly speaking, unsolvability as used here is not uncontrollability, since the probability of obtaining reinforcement is not the same regardless of what response the subject makes. The experimenter says "right" or "wrong" only if the subject points to one side of the card, so the response of not pointing will lead to a different outcome than the response of pointing. However, the demand characteristics of the situation limit the subject to responses which are task appropriate, all of which have the same probability of obtaining reinforcement. So, for these available responses, reinforcement is uncontrollable (see Miller & Seligman, 1973, for a discussion of this point).
LEARNED HELPLESSNESS AND DEPRESSION »
•
DEPRESSED
•-•
•
NONDEPRESSED
511
22 70z
20-
Learned Helplessness, Depression, and the Attribution of Failure David C. Klein, Ellen Fencil-Morse, and Martin E. P. Seligman University of Pennsylvania Depressed and nondepressed college students received experience with solvable, unsolvable, or no discrimination problems. When later tested on a series of patterned anagrams, depressed groups performed worse than nondepressed groups, and unsolvable groups performed worse than solvable and control groups. As predicted by the learned helplessness model of depression, nondepressed subjects given unsolvable problems showed anagram deficits parallel to those found in naturally occurring depression. When depressed subjects attributed their failure to the difficulty of the problems rather than to their own incompetence, performance improved strikingly. So, failure in itself is apparently not sufficient to produce helplessness deficits in man, but failure that leads to a decreased belief in personal competence is sufficient.
Dogs given inescapable shock in a hammock subsequently fail to escape shock in a shuttle box. Even if they make an occasional response that turns off shock, they fail to learn. In contrast, dogs given escapable shock or no prior shock escape well (Overmier & Seligman, 1967; Seligman & Maier, 1967). This interference and the process underlying it have been called "learned helplessness" (Maier, Seligman, & Solomon, 1969; Seligman, Maier, & Solomon, 1971). Learned helplessness occurs in a variety of situations, with a variety of uncontrollable events, and across a number of species, including rats, cats, mice, and men (see Seligman, 1975, for a comprehensive review). The two major behavioral symptoms of learned helplessness are (a) motivational— lowered response initiation, and (b) cognitive—a dampened ability to learn that responding produces reinforcement. Seligman et al. (1971) argued that these symptoms
The research reported in this paper was supported by the following National Institute of Mental Health grants: MH19604 to Martin E. P. Seligman, MH08209 to Julius Wishner, and MH19989 to Aaron T. Beck. The authors wish to thank D. R. Hiroto, W. R. Miller, J. Rodin, and S. Roth for their comments at various stages of the preparation of the paper. Requests for reprints should be sent to Martin E. P. Seligman, Department of Psychology, University of Pennsylvania, 3813-1S Walnut Street, Philadelphia, Pennsylvania 19174.
result when the subject learns that his/her responses are independent of reinforcement. The expectancy that responding will be ineffective reduces the incentive to initiate instrumental responses and disrupts later learning of response-reinforcement contingencies. Recently, investigators have attempted to produce learned helplessness in man. Hiroto and Seligman (1975), Miller and Seligman (1975), and Racinskas (Note 1) provided demonstrations of helplessness in man, which are parallel to the deficits that occur in animals. In these studies, subjects were first exposed to uncontrollable situations, that is, responding and reinforcement were independent. Subjects given inescapable shock, inescapable noise, or unsolvable discrimination problems later failed to escape shock or noise, solved anagrams very poorly, and failed to learn from their successes. In addition, Miller and Seligman (in press) showed that inescapable noise led normal subjects to perceive that subsequent responding and reinforcement were independent. Thus, both the response initiation deficit and the cognitive deficit that occur in helpless animals also occur in man and seem to be caused by the perception that responses and reinforcement are independent. Learned helplessness has been proposed as a laboratory model of depression in man (Seligman, 1972, 1973, 1974, 1975; Seligman, Klein, & Miller, in press). Learned helpless-
508
LEARNED HELPLESSNESS AND DEPRESSION
ness and depression may be parallel in their symptoms, etiology, cure, and prevention. The model suggests that depression is caused by uncontrollable situations which lead the individual to perceive that his responses are generally ineffective in obtaining reinforcement. If the model is correct, helplessness induced in the laboratory in nondepressed subjects should produce response initiation and cognitive deficits parallel to those shown by depressed subjects without helplessness training. A growing body of evidence supports the relationship of learned helplessness to depression (Miller & Seligman, 1973, 1975, in press; Miller, Seligman, & Kurlander, in press). The experiment below asks two questions bearing on the learned helplessness model of depression: 1. Do unsolvable discrimination problems produce deficits in anagram performance in nondepressed subjects similar to those found in depressed controls? Depressed subjects given no pretreatment and nondepressed subjects given unsolvable discrimination problems should show response initiation and cognitive deficits relative to nondepressed subjects given no pretreatment. Unsolvable discrimination problems produce anagram deficits in randomly selected college students (Hiroto & Seligman, 1975), and inescapable noise produces anagram deficits in nondepressed subjects similar to the deficits found in depreessed controls (Miller & Seligman, 1975). The learned helplessness model makes no explicit prediction concerning the effect of unsolvability on depressed subjects' performance. 2. Does attribution of failure have a role as a mediator of learned helplessness? Are performance deficits produced by failure per se, or does it matter if the individual blames his own incompetence or the harshness of the environment for failure? In particular, is anagram performance different if failure on discrimination problems is attributed to lack of ability rather than to task difficulty? Do depressed people make different attributions about their failures than nondepressed people? The learned helplessness model makes no explicit prediction concerning these questions,
509
but Wortman and Brehm (1975) predicted that blaming incompetence should intensify helplessness, and Lichtenberg (1957) claimed that depression is an expectation of failure coupled with the belief that one's personal defects are to blame for the failure. We manipulated attributions by showing subjects norms that either indicated the task they failed was usually solved by their peers or usually was not. METHOD Subjects Subjects were 81 University of Pennsylvania undergraduates recruited from the subject pool and from advertisements asking for volunteers for a problem-solving experiment. They were assigned to depressed and nondepressed groups according to their scores on the Beck Depression Inventory (BDI), a test of the affective, cognitive, somatic, and motoric symptoms of depression (Beck, 1967). The validity and rationale for the use of the BDI in this type of experiment have been discussed elsewhere (Miller et al., in press). Subjects were recruited until 20 males and 20 females were obtained for each group. Subjects with scores of 9 or above were assigned to the depressed group, while subjects with scores of 8 or below were assigned to the nondepressed group. One depressed subject was discarded after completing the experiment because of experimental error. The final sample comprised 40 depressed and 40 nondepressed subjects. Within the depressed and nondepressed groups, subjects were assigned to one of the following five treatments; (a) solvable problems, (b) control, (c) unsolvable problems with no attribution of failure instructions, (d) unsolvable problems with internal attribution of failure instructions, and (e) unsolvable problems with external attribution of failure instructions. Assignments were random, with the restriction that each group contain four males and four females. Thus, there were 10 groups of eight subjects, five depressed and five nondepressed groups. Within each of the 10 groups, subjects were assigned to the two experimenters (the first two authors) so that each experimenter was given two males and two females from each group. The means and standard deviations of BDI scores for the groups are summarized in Table 1.
Apparatus Treatment task. The treatment task was a series of 40 four-dimensional stimulus patterns adapted from Levine (1966) by Hiroto and Seligman (197S). Each of the four dimensions had two associated values: (a) letter (A or T) ; (b) letter color (shaded or unshaded); (c) letter size (large or small); (d) type of border surrounding letter (circle or square). Each problem consisted of 10 cards. For each card, the subject pointed to one side of the
D. KLEIN, E. FENCIL-MORSE, AND M. SELIGMAN
510
TABLE 1 MEANS AND STANDARD DEVIATIONS or BECK DEPRESSION INVENTORY SCORES Depressed
Nondepressed
Group
M
SD
M
SD
Solvable Control Unsolvable-no instructions Unsolvable-internal instructions Unsolvable-external instructions
16.00 12.75
7.82 4.27
3.38 3.50
2.20 2.93
13.25
4.80
3.00
1.41
14.88
4.02
3.63
3.07
13.00
3.8!)
3.75
2.96
Note. Each group contains 8 subjects.
card and the experimenter said whether that side contained the correct value. In this manner, the subject was to try to identify the correct value for each problem. Test task. The test task was a series of 20 anagrams devised by Hiroto and Seligman (1975). Each of the anagrams, which were selected from a list of five-letter anagrams (Tresselt & Mayzner, 1966), consisted of five letters arranged in a standard sequence: 3-4-2-5-1, that is, the first letter of the solution word was the fifth letter of the anagram, the second letter of the solution word was the third anagram letter, and so on (e.g., ERLKC, OUHLG). Sliding scales. Three sliding scales were used to roughly monitor sadness, nervousness, and anger. Each scale consisted of a question about the subject's present mood (e.g., How sad are you feeling right now?), an 11-point scale ranging between two extremes (e.g., extremely sad, extremely happy), and a movable pointer.1 The pointers were at the midpoints of the scales whenever the subject was given the scales. Procedure The subject was first seated at a desk and administered the BDI and the sliding scales. Treatment phase. All subjects were shown a sample pattern of the treatment task and were read the instructions for the cognitive discrimination problems. These instructions can be found in Hiroto and Seligman (1975). Subjects in the solvable groups were then given four solvable problems; subjects in the unsolvable groups were shown the same cards as the solvable subjects, but there was no solution for any problem. For each 10-card problem, the experimenter said "right" and "wrong" five times each in a random order, ending with "wrong." 2 To ensure that unsolvable subjects would not be able to deduce systematically that no solution existed, the cards were turned faster for them than for the solvable subjects. Subjects in the unsolvable groups heard three variations of instructions designed to manipulate attribution of failure: (a) Those with no attribu-
tion of failure instructions were shown the problems as just described, (b) Subjects receiving internal attribution of failure instructions were shown a bar graph just before beginning the problems. This graph showed that 55% of subjects succeed on all four problems, 30% on three, 5% on two, 9% on one, and 1% fail on all four. While studying the graph, subjects were read the following: "This graph shows the number of problems solved by students in a previous experiment. As you can see, most people are able to get three or four of the problems correct." (c) Subjects receiving external attribution of failure instructions were shown a different bar graph just before beginning the problems. This graph showed that no subjects solve all four problems, 1% solve three, 2% two, 7% one, and 90% fail on all four. As they were shown the graph, subjects were read the following: "This graph shows the number of problems solved by students in a previous experiment. As you can see, the problems are very difficult and almost no one has been able to solve them." At the end of each problem, subjects were asked to name the correct dimension. After responding, solvable subjects were told, "That's the right answer" and unsolvable subjects were told, "That's the wrong answer." No solvable subject failed any of the problems. After the treatment phase, all subjects were given the sliding scales for a second time. Test phase. Subjects were then presented with the anagrams task. They were allowed up to 100 seconds to solve each anagram. Response latencies were measured by the experimenter with a hand-held stopwatch. The instructions for the anagrams can be found in Hiroto and Seligman (1975). Four dependent measures were obtained for each subject on the anagrams: (a) mean response latency (a failure counted as 100 seconds) ; (b) trials to criterion for solving the anagrams pattern, defined as the trial number of the third successive trial with a response latency of no more than 15 seconds (for subjects who failed to reach criterion, this measure was taken as the earliest trial the subject could have reached criterion if given more ana-
1 The authors wish to thank Aaron T. Beck for the design of the sliding scales. 2 Strictly speaking, unsolvability as used here is not uncontrollability, since the probability of obtaining reinforcement is not the same regardless of what response the subject makes. The experimenter says "right" or "wrong" only if the subject points to one side of the card, so the response of not pointing will lead to a different outcome than the response of pointing. However, the demand characteristics of the situation limit the subject to responses which are task appropriate, all of which have the same probability of obtaining reinforcement. So, for these available responses, reinforcement is uncontrollable (see Miller & Seligman, 1973, for a discussion of this point).
LEARNED HELPLESSNESS AND DEPRESSION »
•
DEPRESSED
•-•
•
NONDEPRESSED
511
22 70z
20-
