versity David Ross Research Fellowship; Kay Deaux, principal investigator. 2Requests for reprints should be sent to Dr. Richard L. Patten, Department of ...
Motivation and Emotion, Volume 1, Number 1, 1977
Independent Effects of Achievement Motivation and Overt Attribution on Achievement Behavior 1 Richard L. Patten 2 University of Minnesota
Leonard A . W h i t e Purdue University
The cognitive attribution theory o f achievement motivation was examined in a series o f three studies. The persistence o f performance under failure feedback was investigated, with improvement over trials on a digit-symbol substitution task constituting the measure o f persistence. Although evidence was obtained relating overt causal attributions of failure to persistence, partial correlation statistics indicated that this relationship was independent of achievement motivation. Partial correlation statistics also revealed that the relationship between achievement motivation and persistence was independent of attributional variables. In addition, greater persistence was exhibited by high- than by low-achievement subjects in the absence o f attribution differences between these groups. Additional evidence was collected suggesting that the act o f overt attribution was an ego-involving operation that led to these achievement differences in improvement. In general, achievement classification and overt causal attributions had separate effects on persistence, precluding the interpretation of achievement-related behavior in terms o f attribution theory. Overt causal attributions were found to be, in part, a function of the ego-involving character of the task instructions, as was improvement.
Weiner (1972) has proposed a cognitive attribution theory of achievement motivation. The present study investigated applications of the theory to 'Portions of this research were conducted while L. A. White was supported by a Purdue University David Ross Research Fellowship; Kay Deaux, principal investigator. 2Requests for reprints should be sent to Dr. Richard L. Patten, Department of Psychology, University of Minnesota, Minneapolis, Minnesota 55455. 39 © 1977 Plenum Publishing Corporation, 227 West 17th Street, New Y o r k , N.Y. 1 0 0 t l . No part of this p u b l i c a t i o n may be reproduced, stored in a retrieval system, or transmitted, in any f o r m or by any means, electronic, mechanical, p h o t o c o p y i n g , m i c r o f i l m i n g , recording, or otherwise, w i t h o u t w r i t t e n permission of the publisher.
40
Patten and White
achievement behavior in the face of failure, and examined a source of possible confounding in previous tests of the theory. The achievement motivation test was constructed by McClelland, Atkinson, Clark, and Lowell (1953) so that high achievers would perform differently than low achievers on tasks administered under ego-involving conditions. The achievement motivation test was designed to detect subjects (high achievers) whose achievement behavior functioned according to the hunger model of human motivation: just as food deprivation leads to the arousal of food-getting behavior and imagery (Atkinson and McClelland, 1948), ego-involving instructions that describe a task as a test of important abilities would lead to achievement-related imagery and behaviors, such as greater persistence in the face of failure by high- than by low-achievement subjects (e.g., Atkinson, 1964). Weiner (1972) has suggested an application of Heider's (1958) cognitive theorizing to explain the persistence differences of high- and lowachievement subjects..Weiner's theory regards the differential persistence of high- and low-achievement subjects as due to the different attributions they make regarding the causes of their failure. Specifically, high-achievement subjects are thought to attribute failure at achievement-related tasks to lack of effort and bad luck (unstable factors). These attributions hold open the possibility of future success, and therefore high achievers persist in spite of prior failure. Low-achievement subjects, in contrast, attribute failure more to lack of ability and task difficulty (stable causes). Weiner suggests that attribution of failure to stable factors diminishes the prospect of success on future trials, and therefore low-achievement subjects tend to quit after failure. In support of this cognitive theory, Weiner (1972) described a study by Meyer (1973) in which high-achievement subjects did, when asked, attribute failure to unstable factors to a greater extent than did low-achievement subjects. More specifically, the Meyer study employed a digit-symbol substitution task and gave the subjects 1 min to complete four rows of symbol substitutions. The task instructions were neutral in that no attempts were made to arouse achievement motivation by stressing that important abilities were being measured and compared to standards of excellence. On each of four trials, subjects were given failure feedback and were asked to attribute the causes of their failure to lack of ability, lack of effort, bad luck, or task difficulty. A significant positive correlation was obtained between unstable attributions of failure after Trial 1 and improvement on Trial 2. The typical finding of greater persistence (improvement) by high-achievement subjects than by low-achievement subjects was also obtained and was explained as being due to their relative greater attribution of failure to unstable factors.
Achievement Motivation, Attribution, and Persistence
41
However, Meyer's data present a puzzle, for the persistence differences between achievement groups were obtained under instructional conditions that were ostensibly neutral rather than ego-involving. In view of the widespread attention currently given to Weiner's attribution theory of achievement motivation (e.g., Heckhausen and Weiner, 1972; Weiner, 1974a,b), and the critical position of the Meyer study in providing support for the theory, it seemed important to determine the manner in which achievement behaviors were produced in this study under apparently nonego-involving conditions. Examination of the methods employed by Meyer focused on the attribution procedure. This procedure appeared to be a possible difference between this study and previous studies of achievement motivation effects. It appeared that the attribution procedure used by Meyer may have introduced unintended ego-involving properties, which would provide conditions for obtaining persistence differences between high- and low-achievement subjects. That is, asking subjects to overtly attribute failure to factors such as ability, task difficulty, and effort may be ego-involving by indicating to the subjects that their abilities and industriousness were being evaluated and compared to standards. According to this analysis, the instigation of achievement motivation by introducing overt attribution of failure may have been the stimulus that led to greater persistence by high- than by low-achievement subjects in the Meyer study. To test this hypothesis, the first experiment used four groups of subjects in basically the same experimental design employed by Meyer. The possible performance effects of the overt attribution procedure were examined by comparing the performance of groups of high-achievement and lowachievement subjects who either were or were not asked to attribute the causes of failure after each trial. In accordance with the hypothesis that overt attribution leads to greater persistence in high- and low-achievement subjects by arousing achievement motivation, the persistence differences between achievement groups should be obtained only under conditions of overt attribution. In contrast, according to Weiner (1972), who regards the overt attribution procedure as simply assessing ongoing covert processes (Weiner, 1972; Weiner and Kukla, 1970), achievement differences in persistence should occur even when overt attributions are not requested. That is, it should not be necessary to assess attributions overtly in order to observe their effects on persistence. A second issue considered was the possibility that the important correlation between attribution to unstable causes on Trial 1 and increased improvement on Trial 2 obtained by Meyer was confounded with speed of prior performance. It is possible that subjects who worked slowly in Trial 1, and truthfully attributed their failure to lack of effort, exhibited the most
42
Palten and While
improvement on Trial 2 simply because of their relatively slow speed on Trial 1. Such a state of affairs would produce a statistical correlation between attribution and improvement on Trial 2 even if attributions had no actual effect on Trial 2 performance. This hypothesis was tested using the technique of partial correlation to control the possible mediating effects of Trial 1 speed while assessing relationships between attribution and improvement. Two additional groups o f high- and low-achievement subjects were employed who were not given failure feedback. These control groups were included to test the hypothesis (Weiner, 1972) that low-achievement subjects give up after failure.
EXPERIMENTI
Method Subjects. The 117 male subjects were sophomore and freshmen volunteers at Iowa State University who received extra credit in their introductory psychology course for their research participation. These subjects were selected f r o m a larger group of subjects given the Mehrabian (1969) paper and pencil test of resultant achievement motivation? Three weeks prior to the experiment, the subjects who scored in the top and b o t t o m quartiles o f the distribution were individually contacted by phone and asked to participate in the present experiment. Subjects were randomly assigned within each level of achievement motivation to one of three treatment groups: failure-attribute, f a i l u r e - n o attribute, no f a i l - n o attribute. The study was conducted in two parts. Forty-eight subjects were used in the first replication and 66 in the second? The experimenters in this study were both males and were unaware of each subject's achievement score. Procedure. The procedure followed closely that used by Meyer, as reported in part by Weiner et al. (1972). Apart from the possible ego-involving properties of the attribution procedure, the instructional condition was ~This measure of achievement motivation has been used by Weiner and his associates (Weiner, Johnson, & Mehrabian, 1968; Weiner & Kukla, 1970; Kukla, 1972; Weiner & Sierad, 1975) and is highly recommended (Mehrabian, 1969; Cohen, Reid, & Boothroyd, 1973). The Mehrabian (1969) test was designed to measure resultant achievement motivation (Atkinson & Litwin, 1960), which provides for each individual a measure of the tendency to approach success minus the tendencyto avoid failure. Thus, all references to achievement motivation in this paper are to resultant achievement motivation. ~Threeof the subjects were subsequently discarded from consideration when a postexperimental questionnaire and interview revealed their suspicion of th~ experimental manipulations.
Achievement Motivation, Attribution, and Persistence
43
neutral, as in Meyer's study. Each subject was administered six trials of a digit-symbol substitution task, composed of the symbols ~_, -~, -r, U, L , _A, and the digits 1-6. On each trial, the subject was given a mimeographed sheet containing four rows of 16 numerals randomly selected from the digits 1-6, with a space provided below each digit for the symbol substitution. After an initial untimed practice trial in which the subject completed four rows of substitutions, he was told that his task was to complete all four rows on a fresh page within a l-rain time period. Failure was introduced after each of the five trials by stopping the subject after he had completed a few substitutions in the fourth row and informing him that the time was up and that he had failed because he had not completed all the rows of substitutions. The performance measure recorded on each trial was the time required to complete the first three rows of substitutions. Subjects in the failure-attribute condition were asked to attribute their failure to ability, task difficulty, effort, and luck, following closely the instructions used by Meyer and described by Weiner, Heckhausen, Meyer, and Cook (1972). The instructions described each of the four causal factors and the scale to be used in making attributions. Subjects were instructed to make their attributions in percentage terms by completing four statements such as "Ability was ...... % a cause," with the total percentage for the four attributions required to equal 100%. Subjects in another condition were given failure feedback after each trial, but were n o t asked to indicate the causes of their failure. A third no-fail control group did not receive failure information or provide overt attributions. Half the subjects in each of the three treatment conditions were high-achievement subjects, and half were low-achievement subjects. The same six conditions were utilized in a complete replication of the study the following semester. The second replication differed from the first in only two respects. The groups of high- and low-achievement subjects that made overt attributions contained 17 subjects each, rather than eight. These additional subjects were included to provide increased reliability for the correlation and regression analysis planned for these data. A 9-point attribution scale designed for more specific measurement of attributions to the four causal factors was also used? On each scale, there were two opposing -'One reason for preferring the authors' scale, which measures attributions independently of each other, is that measurement with the scale used by Meyer (1973) does not permit a determination o f specific attribution differences. For example, Weiner (1972, p. 384) reported attribution differences a m o n g achievement groups in the Meyer study that he described as indicating that high achievers attribute failure more to unstable causes than do low achievers, and that low achievers attribute failure more to stable factors than do high achievers. However, since attributions were measured with an interdependent scale, the relationship described as one o f low achievers attributing fhilure less to unstable causes than high achievers m a y actually be one o f high achievers attributing failure to stable causes to a lesser extent than low achievers, etc.
44
Patten and White
statements, such as " L a c k o f ability was a cause" and " L a c k of ability was not a cause," arranged symmetrically above the scale on opposite sides of the midpoint. Four points ranging from "very strong agreement" to "slight agreement" were designated on the scale beneath each statement. To complete the rating, each subject was instructed to first decide which o f the two opposing statements he agreed with and then to indicate the extent of his agreement by marking the scale at one of the four points under the statement he chose. If the subject was unable to decide, or did not agree with either statement, he was to check "neither agree nor disagree" at the midpoint. Agreement with a statement indicating a factor was not a cause was scored from "very strong agreement" (I) to "slight agreement" (4). Responses at the midpoint received a score of (5). Agreement with a statement indicating a factor was a cause was scored from "slight agreement" (6) to "very strong agreement" (9). All subjects in the groups that did not overtly attribute were given a 30-sec intertrial interval that corresponded to the mean time required to complete the attribution measure after each trial. Following the last trial, subjects evaluated their performance on a 7-point scale with endpoints "very p o o r " and "very g o o d . " All subjects were then debriefed and encouraged to discuss their impressions and thoughts about the study. Results Persistence. The individual trial performance scores were converted to improvement scores for each subject by subtracting the time required on each trial from the time required on Trial 1. Group means for the four improvement scores are presented in Figure 1 in blocks of two trials. The 22 20
NFC: NO-FAIL CONTROL GPS NA: NO-ATTRIBUTE A:ATTRIBUTE
.....,A
18 t~
16 14
/2 t0
Q=
.--------
8
_ ~
6
~"
4
~
2~
I
/
~
N
A
:
NFC
- '
H I - A CH.
-- -- --LO-ACH. I
TRIAL BLOCKS
2
Fig. l. Group mean block improvement scores as a function of the experimental conditions.
Achievement Motivation, Attribution, and Persistence
45
block improvement scores were calculated by adding the two improvement scores within each block. An initial analysis of variance indicated that the six treatment groups did not have different mean times on the first trial (all Fs < 1). Another preliminary analysis of variance indicated that these groups did not differ on the total number of incorrect substitutions made on the task (all Fs < 1). The subjects receiving failure feedback rated their performance as significantly poorer than did the no-fail control subjects [F (1,102) = 9.87, p < .01]. This difference supports the idea that the failure feedback was an effective manipulation. A two (measure) by two (achievement classification) by three (conditions) by two (block) unweighted means analysis of variance was performed on improvement scores. The measure factor had two levels corresponding to the two replications of the study employing different attribution measures. The achievement classification factor had two levels, highvs. low-achievement. The conditions factor had three levels: failure-attribute, no fail-attribute, and no fail-no attribute. The trial block factor had two levels, one for each block of two trials. A significant interaction between achievement classification and condition was obtained [F (2,102) = 6.03, p < .01]. Simple-effect comparisons among groups receiving failure feedback showed significantly greater improvement for subjects who overtly attributed than for subjects who did not overtly attribute. This effect was found for both high achievers [F (1,102) = 35.42, p < .01] and low achievers IF (1,102) = 8.55, p
CAUSAL ATTRIBUTION c.
R - A C H .----,> IMPROVEMENT
IMPROVEMENT b.
--~ IMPROVEMENT
~--
UNSTABLE ATTRIBUTION
d.
Fig. 2. Four causal hypotheses considered for relationships presented in Tables I and II.
vidual regression coefficients representing the linear change in performance across trials were computed for each subject. Analyses revealed a significant correlation between subjects' attribution of failure to unstable causes on Trial 1 and the linear change in performance across the five trials [r (32) = .46, p < .01]. In other words, subjects who attributed initial failure more to unstable causes exhibited the greatest improvement. This correlation was not reduced when achievement motivation and T P T were controlled. The correlations between achievement motivation and improvement, with attributions and T P T controlled, are presented in Table II. Significant correlations between achievement motivation and improvement on Trials 4 and 5, and linear change over all trials were obtained. These correlations were not decreased by controlling relationships between attributions and improvement or by controlling TPT. In sum, a correlation between initial attribution to unstable causes and improvement was obtained. This correlaTable IL Correlations Between Resultant Achievement Motivation and Improvement Controlled variable None Lack o f effort Bad luck Unstable causes Lack of ability Task difficulty Stable causes TPT
ap < .05. bp < .01.
Trial 2
Trial 3
Trial 4
Trial 5
All trials
-.17 -.17 -.20 ..... -.17 -,16 -,17 -.07
.0t -.0t .00 .00 -.01 -.01 -.02 ,06
,36 a .37 a ,37 a .37 a .37 a .36 a .37 a .39 a
.37 a .37 a .37 a .37 a ,37 a .36 a .35 a .38 a
.47 b .54 b .48b .48 b .49 b .47 b .48b .51 b
Achievement Motivation, Attribution, and Persistence
49
tion was independent of both TPT and achievement motivation. A correlation between achievement motivation and improvement was also obtained, and was independent of both TPT and attributions. These independent relationships are diagrammed in Figure 2d. Because controlling for attributions did not reduce the relationship between achievement classification and improvement, the causal relationship diagrammed in Figure 2c, Weiner's attribution theory, was not supported.
Discussion The results of the present study present a number of problems for Weiner's attribution theory of achievement motivation. First, high-achievement subjects were more persistent than were low-achievement subjects, but only when overt attributions of failure were obtained. Second, even though Meyer's significant positive correlation between attribution to unstable causes after Trial 1 failure and improvement on Trial 2 was confirmed in the present study, evidence suggests that the correlation was due to confounding by a third variable, TPT. Third, although evidence was presented indicating the development over trials of a relationship between attribution to unstable causes and improvement, which is consistent with attribution theorizing, the correlation between achievement motivation and improvement was independent of these attributional relationships--contrary to Weiner's attribution theory of achievement motivation. Fourth, the finding that high achievers exhibited greater improvement than low achievers in the absence of attribution differences between these groups indicates that attribution differences are not necessary for the appearance of persistence differences between high- and low-achievement subjects. Based on the results of the present study, a new model was proposed to describe the relationships among attributions, achievement motivation, and improvement. This model, diagrammed in Figure 2d, suggests independent effects of attribution and achievement motivation on improvement. In general, the results of the present research provide support for the rationale that led to the conclusion that the procedure of overt causal attribution of failure was a source of ego-involvement leading to persistence differences between high- and low-achievement subjects. This formulation is consistent with the McClelland et al. (1953) conceptualization of achievement motivation, which holds that persistence differences between high and low achievers are obtained only under ego-involving conditions. The following experiments were conducted to provide additional evidence supporting the proposed ego-involving properties of the overt causal attribution procedure.
50
Patten and White EXPERIMENT2
The purpose of this experiment was to further establish the proposed functional similarity o f the overt attribution procedure and explicit egoinvolving instructions by showing that they have similar effects on improvement. All subjects received failure feedback on each trial of a digit-symbol task. The first trial was conducted under neutral conditions for all subjects, as in Experiment 1. Prior to Trial 2, subjects were administered either egoinvolving or relaxed instructional conditions, as designed by McClelland et al. (1953) to operationally establish different levels of achievement motivation. Introducing ego-involving instructions after Trial 1 was expected to produce greater improvement on Trial 2 relative to the neutral or relaxed conditions--just as overt attribution did in the preceding experiment. That is, ego-involving instructions were expected to facilitate improvement by arousing achievement motivation, just as these instructions facilitated the appearance of achievement-related themes in thematic apperception stories. After Trial 2, the overt attribution procedure was introduced for some subjects within each o f the three instructional conditions. As in Experiment 1, greater improvement was expected when subjects overtly attributed than when they did not overtly attribute. Subjects working under conditions that combined both ego-involvement and overt attribution were expected to show the most improvement.
Method Subjects. The subjects, 84 male college student volunteers recruited from the same population and under the same conditions used in the first experiment, were each randomly assigned to one of six groups (n = 14 each) defined by the possible combinations of three task conditions (ego-involved, neutral, relaxed) and two attribution conditions (presence or absence of overt attribution). Procedure. A digit-symbol substitution task was used to obtain measures of persistence, as in Experiment 1. The instructional treatment conditions employed were as follows: In the ego-involving (achievement motivation) condition, subjects were told that the task was a test of leadership potential and general intelligence, using the following instructions designed by McClelland et al. (1953): The task which you are working on today was taken from the performance section of the WechslerAdult IntelligenceScale (WAIS). The Wechslerhas been frequently used as a measure of adult intelligence... The task in conjunction with the eleven other sub-scales providesa relativelyaccurate estimate of one's general intelligence
Achievement Motivation~ Attribution, and Persistence
51
and general intellectual capacity. The performance section of the WAIS of which the digit-symbol task is also a part has also been shown in a series of longitudinal studies to be a relatively accurate indicator of leadership potential,
To reinforce this impression, each subject was shown the title page and the digit-symbol substitution task on the back page of the WAIS. In the neutral condition, nothing was deliberately done to stimulate or suppress achievement motivation. The experimenter gave a straightforward description o f the task, identical to that used in Experiment I. In the relaxed condition, in contrast, the experimenter behaved in a generally relaxed, casual manner and told the subject that we were interested in trying out the digit-symbol task for possible use later in a real experiment rather than in how well any subject did on the task. All subjects, run individually, were given five timed trials after an untimed practice trial. Failure feedback was given after each timed trial. Trial 1 was conducted under neutral conditions for all 84 subjects. The remaining procedures used in administering the digit-symbol task were the same as in Experiment 1, with the following exceptions: Prior to Trial 2, one third (28) o f the subjects were given the ego-involving instructions, and a second third (28) were given the relaxed instructions. The remaining third of the subjects were continued in neutral conditions on Trial 2 and on each of the remaining three trials. Between Trials 2 and 3, a random one half (14) of the subjects within each o f the three instructional conditions were asked to make overt attributions o f their failure after Trial 2 and after each o f the remaining trials using the independent measure from Experiment 1. The remaining subjects did not make overt attributions. Following the last trial, subjects who did not overtly attribute were asked to write down any factors that may have caused or influenced their performance on the task. After the session, each subject was fully debriefed and given time for questions and answers.
Results Persistence. Persistence was measured using improvement scores, as in Experiment 1. Preliminary analyses of variance showed that the six groups did not differ on the time taken to complete the first trial, or on the total number o f incorrect substitutions during the first five trials (all Fs < 1). An initial analysis o f variance indicated that corresponding groups o f subjects run under neutral and relaxed instructional conditions exhibited almost identical improvement scores on all trials. Thus, the group means for neutral and relaxed conditions were combined for ease of graphic and statistical presentation. The group mean improvement scores for each condition are presented in Figure 3 as a function of trials.
52
Patten and White
I6 /4
~b.j 10
-----NEUTRAL-RELAXED EGO INVOL VED NA:NO-ATTRIBUTE A." ATTRIBUTE~ A ~
N
N
A
~ oA
/
2
ol
2
3
4
5
TRIALS
Fig. 3. Individual group mean improvement scores as a function of the experimental conditions and trials.
The immediate and persisting effects of shifts from one condition to another were examined separately. The immediate effect o f introducing ego-involving instructions between Trials 1 and 2 was a much larger improvement score on Trial 2 by subjects under ego-involving conditions than by subjects in neutral or relaxed conditions IF (1,82) = 27.60, p < .01]. The immediate effect o f introducing requests for overt attributions between Trials 2 and 3 for subjects working in neutral and relaxed conditions was greater improvement on Trial 3 by subjects who overtly attributed than by subjects who did not IF (1,52) --- 9.19, p < .01]. The shift to overt attribution did not further increase the improvement scores across Trials 2 and 3 o f subjects working under ego-involved conditions. Analyses of performance across Trials 3, 4, and 5 revealed the continuation of greater improvement in the neutral and relaxed conditions when attributions were requested than when they were not [F (1,54) = 24.02, p < .01]. However, the apparent facilitation of improvement by overt attribution in the ego-involving condition did not attain statistical significance. A significant trials effect IF (2,52) = 5.48, p < .01] indicated an increase in the improvement scores over Trials 3, 4, and 5 for both groups of subjects working under ego-involving conditions. The subjects run under neutral and relaxed conditions who overtly attributed also continued to improve over the last three trials IF (2,108) = 11.05, p < . 0 1 ] . No other effects were significant. Attributions. The subjects run under neutral and relaxed conditions provided virtually identical attribution patterns, and so have been combined into a single group for purposes of statistical analysis. Attributions of
Achievement Motivation, Attribution, and Persistence
53
failure to lack of ability and stable causes were greater under neutral and relaxed conditions than under ego-involved conditions [F (1,40) = 4.81, p