UNIVERSITY OF 
 
 ILLINOIS LIBRARY 
 
 AT URBANA-CHAMPAIQN 
 
 BOOKSTACKS 
 
CENTRAL CIRCULATION BOOKSTACKS 
 
 The person charging this material is re- 
 sponsible for its renewal or its return to 
 the library from which it was borrowed 
 on or before the Latest Date stamped 
 below. The Minimum Fee for each Lost 
 Book is $50.00. 
 
 Theft, mutilation, and underlining of books are reasons 
 for disciplinary action and may result in dismissal from 
 the University. 
 TO RENEW CALL TELEPHONE CENTER, 333-8400 
 
 UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN 
 
 
 When renewing by phone, write new due date below 
 previous due date. L162 
 
Digitized by the Internet Archive 
 
 in 2012 with funding from 
 
 University of Illinois Urbana-Champaign 
 
 http://www.archive.org/details/attributingcause184staw 
 
Faculty Working Papers 
 
 ATTRIBUTING THE "CAUSES" OF PERFORMANCE: 
 AN ALTERNATIVE INTERPRETATION OF CROSS- 
 SECTIONAL RESEARCH ON ORGANIZATIONS 
 
 Barry M. Staw 
 
 #184 
 
 College of Commerce and Business Administration 
 
 University of Illinois at Urbana-Champaign 
 
FACULTY WORKING PAPERS 
 College of Commerce and Business Administration 
 University of Illinois at Urbana-Champaign 
 May 30, 1974 
 
 ATTRIBUTING THE "CAUSES" OF PERFORMANCE: 
 AN ALTERNATIVE INTERPRETATION OF CROSS- 
 SECTIONAL RESEARCH ON ORGANIZATIONS 
 
 Barry M. Staw 
 
 #184 
 
Attributing the "Causes" of Performance: 
 An Alternative Interpretation of 
 Cross-sectional Research on 
 Organizations 
 
 Barry M. Staw 
 Organizational Behavior Group 
 Department of Business Administration 
 University of Illinois at Urbana -Champaign 
 
Abstract 
 This paper presents a general alternative interpretation of 
 correlational findings vihich link perceptual or questionnaire measures 
 to data on performance. Essentially, it is posited that organizational 
 participants possess theories of performance just as do organizational 
 researchers, and that respondents will use knowledge of performance 
 as a cue by which they attribute characteristics to themselves, their 
 work groups, and organizations. According to this attribution 
 hypothesis, self-report data on organizational characteristics may 
 actually represent the consequences rather than the determinants of 
 performance. To test this alternative interpretation of correlational 
 findings, an experiment was conducted in which knowledge of group 
 performance (positive vs. negative) was a manipulated independent 
 variable. The results showed that knowledge of performance affected 
 the levels of influence, cohesion, communication, motivation, and 
 openness to change attributed by members to their work groups. These 
 findings were also cross-validated by an interpersonal simulation. 
 The data of the true experiment and the interpersonal simulation, 
 together, provided strong evidence for the attribution hypothesis. 
 
Although much cf the research in organizational behavior is 
 devoted to understanding the causes of performance, the findings in 
 the field are still largely based u on correlational data in which the 
 direction of causation is unknown. At present, the research supporting 
 most organizational theories contains hypothesized independent variables 
 which can either be the causes of performance, the effects of performance, 
 cc-variates of third variables, or the results of a network of reciprocal 
 causation. Therefore, it could be argued strongly that, in terms of 
 both theory and application, resolving ambiguity in causal inference 
 is one of the field's most pressing issues. 
 
 Previously, there have been two empirical studies specifically 
 designed to demonstrate problems in interpreting cross-sectional 
 (correlational) findings. In the first of these studies, Lowin and 
 Craig (1963) eKperimentally manipulated the performance of subordinates 
 and measured the leadership styieof persons hired to perform a real 
 supervisory role. The results of this study showed that closeness of 
 supervision may be a function of subordinate performance rather than a 
 causal dettrminrnfc of performance, as previously believed. In a some- 
 what parallel study, Farris & Lim (1969) compared the leadership style 
 of work group supervisors after knowledge of subordinate performance had 
 been experimentally manipulated. This research involved a role playing 
 eicerciee in which one student was designated as a foreman and three 
 other students acted «s a three-person work group in an industrial conflict 
 situation. Each group worked with it6 foreman for 20 minutes toward the 
 solution of the "Change in Work Procedure Case" (Maier, Solem, & Maier, 1957), 
 

 
 
 
 
 
 
 
2 
 and then completed a post-experimental questionnaire on the foreman's 
 behavior. Knoxjledge of performance was manipulated by providing 
 information to the foreman (before the work session) that his group was 
 one of the highest or lowest groups in terms of previous performance. 
 The results showed that, for high performing groups, the foreman was 
 perceived to be more supportive of the workers, higher in goal emphasis, 
 and more facilitative of interaction than was the foreman of low 
 performing groups . 
 
 By showing that changes in performance can cause changes in other 
 behavioral variables, both the Lowin & Craig (1968) and Farris & Lim 
 (1969) studies represent efforts to stimulate more causal research on 
 organizations. The approach represented by their research is a step- 
 by-step demonstration of the plausibility of reversals in causal order. 
 In fact, from this approach, one might advocate measuring the effects 
 of performance upon an array of individual, group, and organiztional 
 variables, and the construction of a thorough inventory of likely causal 
 reversals. With this information, researchers eventually would know where 
 to invest substantial resources on research with methods mere conducive 
 to causal inference (i.e. field experimentation, longitudinal analysis, 
 and laboratory simulations of organizational processes). 
 
 The step-by-step demonstration of causal reversals is no doubt a 
 worthwhile procedure to help budge the field of organizational behavior 
 from its near total reliance on cross-sectional (correlational) data. 
 However, it is believed that this procedure is neither sufficiently speedy 
 nor now necessary to encourage a significant increase in causal research. 
 

3 
 The reason for this conjecture is a new alternative interpretation of 
 cross-sectional data which is both parsimonious and of general 
 applicability to correlational findings linking performance data to 
 self-report measures of individual; group, and organizational characteristics. 
 This alternative interpretation of correlational findings is derived from 
 previous work on attribution theory. 
 
 Attribution theory is specifically concerned with how individuals 
 assign enduring traits or dispositions to themselves and other persons (Heider, 
 1957; Jones and Davis, 1965; Kelley, 1971, 1973; Nisbett and Valins, 1971). 
 It assumes that individuals have a need to understand and explain the 
 events around them, and that based upon this need, individuals will 
 develop a lay or "naive" psychology of behavior (Heider, 1958). To date, 
 most of the research in attribution theory has studied the perception of 
 personal characteristics under varied environmental conditions (e.g., 
 Bern, 1965; Calder and Staw, 1974a, 1974b; Deci, 1971; Jones, Davis, and 
 Gergen, 1961; Jones and Harris, 1967; Schachter and Singer, 1962; Staw, 
 1974a, 1974b; Strickland, 1958). However, in its broadest context, 
 attribution theory is concerned with the ascription of characteristics to 
 any entity. As Kelley (1973) has noted, all of the judgments of the 
 type, "Property X characterizes Entity Y" can be viewed as causal 
 attributions. Thus, it seems reasonable to assume that the organizational 
 participant, in a desire to understand and control his particular 
 environment, may develop a lay psychology of individual, group, and 
 organizational functioning. Just as individuals may possess an implicit 
 personality theory to guide their impressions of others (Bruner and Tagiuri, 
 

 
 
 
1954), the organizational participant may possess a theory of the relation- 
 ships between organizational characteristics and subsequent performance. 
 
 The specific attribution hypothesis posited here is that 
 individuals utilize knowledge of performance as a cue by which they 
 ascribe characteristics to an individual, group, or organizational unit. 
 The attribution hypothesis posits that performance is a potent independent 
 variable, and that many of the correlations between performance and 
 self-report data may be accounted for by the following causal sequence: 
 
 Level of Performance ^ Attribution of Characteristics ^ 
 
 Self-report of Characteristics. That is, performance data may cause 
 persons to assign an entire set of characteristics (i.e. a stereo- 
 type) to individuals, groups, and organizations, and this attributed 
 
 set of characteristics may underlie many of the correlations derived 
 
 2 
 from cross-sectional studies of organizational processes . 
 
 The attribution hypothesis can be illustrated by a questionnaire 
 
 developed by Likert (1967) to support his System 4 theory of management. 
 
 Likert askec several hundred manage. 3 to "think of the most productive 
 
 department, division, or organization (they) have known well." The 
 
 managers were then asked to rate this entity in terms of organizational 
 
 processes such as motivation, influence, communication and cooperation. 
 
 Subsequently, these same managers were also asked to rate their least 
 
 productive department, division, or organization on each of these 
 
 dimensions. As expected, a high degree of motivation, mutual influence, 
 

 
 
 
 
 
 
 
 
5 
 cooperation, and communication were associated with the highest producing 
 units. Although it is not yet clear whether the processes seen by 
 managers as being associated with high performance actually contribute 
 to performance, Likert's data do illustrate that, perceptually , individuals 
 will distinguish between high and low producing units. Moreover, the 
 existence of distinct stereotypes of successful versus unsuccessful 
 organizations points to the very possibility that significant correlations 
 between performance and self-report data may only be reflecting the 
 respondents' "theories" of organizational performance rather than 
 actual events. And as Heider (1958) has noted in his now classic 
 analysis of interpersonal perception, a lay or "naive" psychology of 
 behavior may or may not be correct. 
 
 Clearly, if knowledge of peformance causes one to attribute 
 particular characteristics to individuals, groups, or organizations, 
 it may therefore be risky (and certainly unscientific) to posit that 
 self-report data on these characteristics accurately represent the 
 causal determinants of performance. In essence, questionnaire measures 
 considered by organizational researchers to be indicators of the 
 determinants of performance, may actually constitute the consequences 
 of performance. This possibility is of substantial importance to 
 organizational research since individual, group, and organizational 
 characteristics are rarely observed directly, but are generally measured by 
 respondents' perceptions within a field setting. 
 
 A laboratory experiment was conducted to test the relevance of the 
 attribution interpretation to some important correlational findings. Specifically 
 it seemed desirable to test whether this alternative interpretation 
 is applicable to Tannenba urn's (1968) replicated finding that 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6 
 
 high mutual influence is associated with high performance, Likert's 
 (1961) finding that group cohesiven.ss is associated with high performance, 
 and Evan's (1965) finding that interpersonal conflict (but not task 
 conflict) is related to performance. In addition, the relationships of 
 performance to motivation (Galbraith and Cummings , 1967), communication, 
 and openness to change (Likert, 1961) were investigated by this research. 
 
 METHOD 
 Subjects Subjects for this experiment were undergraduate students 
 enrolled in the College of Commerce and Business Administration at the 
 University of Illinois, Urbana -Champaign. Sixty students were randomly 
 assigned to three-man groups and each group was asked to participate 
 in a "Financial Puzzle Task." Group members were given copies of the 
 1969 annual report of a medium-sized (but not well known) electronics 
 company. The report contained a description of the company, a letter 
 from the president on the firm's prospects, and five preceding years 
 of financial data. The group membe 3 were tola that their task was to 
 estimate company sales and earnings per share for 1970, taking into 
 consideration any knowledge they might have of the electronics industry 
 or state of the economy at that time. Each group was given thirty 
 minutes to discuss the issue and make any necessary calculations in 
 formulating a group estimate of sales and earnings per share. Subjects 
 were told that the purpose of the experiment was to evaluate the 
 performance of groups of various sizes and that previous research had 
 been conducted on three, four, and five-man groups. 
 

 
7 
 Manipulation of Perf o rmance After each group presented ics estimates 
 of sales and earnings per share, th ■> experimenter staged that "it would 
 be interesting to see how well this group had performed relative to 
 previous three-man groups." The experimenter then took the group's 
 estimates of sales and earnings per share and searched through several 
 file cabinets in the next room. On returning to the (randomly 
 assigned) High Performance groups, the experimenter announced that the 
 group had "done quite well," that their sales figure was off by only 
 $10,000, earnings per share was accurate within $.05 a share, and 
 that the group's overall performance was clearly in the top 20% of 
 three-man groups. On returning to the (randomly assigned) Low Perforraanc o 
 groups, the experimenter announced that they had "not done too well ," 
 that, their estimate for sales was off by $10,000,000, their estimate 
 for earnings per share was off by $1.00, and that the group's overall 
 performance was in the lowest 20% of previous three-man groups. No 
 subjects expressed strong doubts about their grotp'j performance. 
 However, it should be noted that th>- annual report us<,d in ^hr'.s 
 experiment was selected specifically on the basis of its ambiguity cno 
 could be interpreted in either a positive or negative manner. 
 Dependent Variables After being cold of their group's performance, 
 subjects were led to separate rooms and asked to complete a short question- 
 naire about, "what went on in the group." On the ques tionno ire were it 
 to measure group cohesiveness , influence, covnmunica tion, task conflict, 
 openness to change, motivation, ability, and clar5.ty of instructions. 
 Although the questions were randomly ordered on the questionnaire, they 
 are listed below under the appropriate variable headings. 
 

 
 
 
 
Cohesiveness 
 
 a. To what extent did you enj' y working with your teammates? 
 (11 point: scale from "not at all" to "to a great extent") 
 
 b. In working on the financial puzzle cask, what were your 
 personal feelings toward your teammates? 
 
 (11 point, scale from "I disliked them" to "I liked them") 
 
 c. How would you rate the cohesiveness or group spirit of 
 your team? 
 (11 point scale from "extremely low" to "extremely high") 
 
 II . Influence 
 
 a. How much influence did you have on final solution of the task? 
 (11 point scale from ''very little" to "a great amount") 
 
 b. How much influence did your teammates have on the final 
 solution of the task? 
 
 (11 point scale from "very little" to "a great amount") 
 
 III . Communication 
 
 a. How would you rate the quantity of communication between 
 you and your teammates? 
 
 (11 point scale from "very low" to "very high") 
 
 b. How would you rats the quality of communication between 
 you and your teammates? 
 
 (11 point scale from "very low" to "very high") 
 
 IV. Task Conflict 
 
 a. To what extent: did you and your teammates each have differeni" 
 ideas about methods to solve the financial puzzle task? 
 
 (11 point scale from "not at all" to "to a great extent") 
 
 b. If you and your teammates had different ideas about solving 
 the task, to what extent did you have an open confrontation 
 of ideas? 
 
 (11 point scale from "not at ail" to "to a great extent") 
 
 V . Openness to Chang e 
 
 a. How open were your teammates to your ideas and suggestions 
 about solving the financial puzzle task? 
 
 (11 point scale from "not open at all" to "extremely open") 
 
 b. In solving the task, to what extent did your teammates ever 
 attempt to impose or force their position(s) on you? 
 
 (11 point scale from "not at all" to "to a great extent") 
 
VI. Satisfaction 
 
 a. To what extent did you enjoy working on the Financial Puzzle 
 Task? 
 (11 point scale from "not at all" to "to a great extent") 
 
 VII. Motivation 
 
 a. To what extent were you interested in performing vrell on 
 the financial puzzle task? 
 
 (11 point scale from "not at all" to "to a great extent") 
 
 b. To what extent were your teammates interested in performing 
 well on the financial puzzle task? 
 
 (11 point scale from "not at all" to "to a great extent") 
 
 VIII. Ability 
 
 a. In general, how would you rate your ability in solving 
 financial puzzles? 
 
 (11 point scale from "very low" to "very high") 
 
 b. In general, how would you rate your teammates 5 ability 
 in solving financial puzzles? 
 
 (11 point scale from "very low" to "very high") 
 
 IX. Role Clarity 
 
 a. Were the instructions for solving the financial puzzle 
 made clear to you? 
 (11 point scale from "not at all" to "very clear") 
 
 RESULTS 
 
 Check on the performan c e manipulation 
 
 Suhjects randomly assigned to High Performance groups rated their 
 ability in solving financial puzzles as higher than did subjects in 
 Low Performance groups (t = 5.64, d.f. = 58, p<.001). Subjects in the 
 High Performance groups also rated their teammates' ability as higher 
 than did those in Low Performance groups (t = 2.60, d.f. = 58, p<.01) . 
 These data support the hypothesis that subjects believed the information 
 provided by the experimenter on their group's performance. 
 

 
 
10 
 It should be noted thac, in actuality, the groups assigned to 
 the High Performance condition per.ormed no better than those assigned 
 to the low Performance Condition. In fact, in terms of predicting corporate 
 sales and earnings, groups told chat they had performed well actually 
 performed slightly worse than those told they had performed poorly (For 
 sales: t = -.48, N.S.; for earnings: t = -.23, N.S.). Thus, any 
 reported differences in the perception of group characteristics are 
 likely to be due to manipulated knowledge of perf ormance rather than 
 to any actual differences in the behavior of the groups. Again, it 
 should be stressed that the financial data comprizing the group task 
 was specifically selected (in terms of ambiguity) so as to allow a 
 credible manipulation of knowledge of performance. 
 Effect of knowledge of performance on. perceptions of interpersonal beh avi o r 
 
 The perceptions of several dimensions of interpersonal behavior 
 for subjects in both High and Low Performance groups are shown in 
 Table 1. Where more than one item was used to measure a particular 
 variable, and where these items we. e significantly ir tercorrelated, a 
 combined score and resulting t value is also reported. 
 
 Ijij^e_rt_Ta_bl_e_l_«iboat here 
 As shown in Table 1, individuals who were randomly assigned to High 
 Performance groups rated their groups as more cohesive (t ~ 1.68, c.f. ~ 53 
 p<.05) and enjoyed working with their r.eammaces to a greater extent 
 (t - 1.81, d.f. = 58, p<-05) than did individuals assigned to Low 
 Performance groups. Persons in High Performance groups also rated 
 their groups higher in quality and quantity of communication (t = 1.77, 
 

11 
 
 d.f. = 58, p<.C5), higher in total influence (t = 1.86, d.f. = 58, 
 
 p<.05) , and marginally higher in or ^nness to change (t = 1.49, d.f. = 53, 
 
 p<.10). It is interesting to note that the effect of performance on 
 
 total influence was due primarily to the large effect of performance 
 
 on the perception of one's own influence (t = 2.47, d.f. = 58, p<.01) 3 
 
 2nd that there was no effect of performance on the perception of 
 
 teammates' influence on the group task. No clear relationship to 
 
 performance was shown by the two indicators of task, conflict and these 
 
 two scales were not significantly intercorrelated, 
 
 Effects of k nowled ge of pe rformance on satisfaction, motivation, a bill .ty 
 and role clarity 
 
 Table 2 shows that subjects assigned to High Performance groups 
 
 enjoyed working on the experimental task to a greater extent than did 
 
 subjects assigned to Low Performance groups (t = 5 . 54 , d.f. = 58, p<.001) 
 
 In addition, subjects in High Performance groups rated cheir own 
 
 interest in performing well on the task as greater than subjects assigned 
 
 to Low Performance groups (t - 5.33 d.f. - 58, p<.001). Similarly., 
 
 these same subjects rated their teammaces' interest in performing well 
 
 on the task higher than did subjects m Low Performance groups. Finally 
 
 as previously reported, feedback on performance affected the subjects 
 
 rated ability (t = 5.64, d.f. = 58, p<.001), his perception of his 
 
 teammates' ability (t = 2.60, d.f. = 58, p<.01), and also the rated 
 
 clarity of instructions for the task (t = 2.20. d.f. = 58, p <.05) . 
 
 Insert Table 2 about here 
 
12 
 
 DISCUSSION 
 
 As illustrated by the data of Tables 1 and 2, knowledge cf performs, 
 had a marked affect on the self-report measures of intragroup processes. 
 As expected, individuals who were told that they had participated in 
 a high-performing group rated their group higher in conesiveness 
 influence, communication, openness to change (marginally significant) 
 and motivation as compared to individuals who were told that they had 
 participated in a low performing group. As a whole, these data provide 
 support for the notion that individuals attribute one set of character- 
 istics to a work group they believe is effective and another, different, 
 set of characteristics to an ineffective work group. As a whole, 
 these data also offer support for an attrioutional interpretation of 
 correlations between self-report data and measures of group performance. 
 
 The data on cohesiveness and task conflict provide, a particalarly 
 interesting test of the attribution hypothesis. Previously, Evan 
 (1965) had hypothesized that the impact of intragroup conflict upon 
 performance may not necessarily be negative, and that the effects of 
 conflict might depend on the type of conflict involved. Specifically, 
 Evan postulated that interpersonal conflict should have a negative 
 effect on work group performance, while task conflict night prove 
 beneficial. By correlating self-report measures of conflict to the 
 performance of R & D groups, Evan's data showed a significant negative 
 relationship between interpersonal conflict and performance, but no 
 clear relationship between task conflict and performance. As shown in 
 Table 1, quite similar results were obtained in this study .jhen knowledge 
 

13 
 
 of performance was the manipulated independent variable. Knowledge of 
 high performance caused subjects to perceive less interpersonal conflict 
 (greater group cohesiveness) , while there was a tendency (but not 
 totally consistent) to rate a high performing group as being higher 
 in task conflict. Evan's relatively complex relationship between 
 conflict and performance was thus replicated when knowledge of 
 performance was the manipulated independent variable. 
 
 A second test of the attribution hypothesis is provided by the 
 data on intragroup . influence . Within several organizational settings, 
 Tannenbaum (1968) has found that the amount of total control or influence 
 was significantly related to organizational effectiveness. In each of 
 these studies (Smith and Tannenbaum, 1963; Tannenbaum, 1962, Tannenbaum, 
 1968), self-report measures of influence are correlated with objective 
 measures of organizational performance. Although Tannenbaum has 
 interpreted these findings as indicating that greater total influence 
 causes improved performance, an attribution interpretation is also 
 plausible. In fact, the hypothesis that individuals attribute greater 
 influence to high rather than low producing groups is generally supported 
 by the data of this experiment . 
 
 The data en quality and quantity of communication also provide 
 support for the attribution hypothesis. Although communication has 
 previously been found to correlate with organizational effectiveness (see 
 Price, 1967), the direction of causation has not been clear. In this 
 experiment, however, members of high producing groups inferred higher 
 quality communication to their groups and tended also to infer a greater 
 

 
14 
 quantity of communication. In addition, persons with knowledge of 
 high performance tended to rate their teammates as being more open to 
 change (see Likert, 1961, 1967, Tor concomitant correlation), and 
 perceived both themselves and their teammates as being higher in 
 motivation (see Galbraith and Cummings , 1967, for concomitant 
 correlation) . 
 
 Although the data of this experiment are generally supportive 
 of the attribution hypothesis, it should be noted that some of the 
 data can be explained by alternative processes. For example, one 
 indicator of group cohesiveness (enjoyed working with teammates) may 
 have been higher among persons assigned to High Performance groups 
 due to the reinforcement associated with task success. Although 
 this explanation would also clearly apply to the measure of task 
 satisfaction, it would not, however, be as applicable uo other 
 intragroup processes measured on the questionnaire (e.g. influence, 
 conflict, communication, motivation, and openness to change). 
 
 A second alternative interpre. ition is suggested by the data on 
 intragroup influence and motivation. Because persons assigned to 
 Low Performance groups attributed less influence to themselves and 
 rated themselves as lower in task motivation than persons in High 
 Performance groups, an ego-defensive process is suggested (Weiner, 1971). 
 One problem with the ego-defensive expiaxna tion , however, is that 
 subjects also rated their teammates' motivation as lower under the Low 
 Performance condition, and this result would not be predicted by an 
 ego-defensive process. A second problem with the ego-defensive explain 
 

 
 
 
15 
 is that subjects rated their own ability under Low Performance 
 conditions as significantly lower tMn chat of their teammates. 
 Clearly, if an ego-defensive process were operating, one would 
 expect subjects to depreciate their teammate's abi mder low groi 
 
 performance, while keeping their own rated ability intact. 
 
 In sum, the results of this experiment support the contention that 
 knowledge of performance is a relatively potent independent variable. 
 Moreover j the overall pattern of results can be more parsimoniously 
 explained by an attribution theory than by either a reinforcement or 
 ego-defensive process. The attribution process posited here is that 
 individuals hold distinct stereotypes of high versus low performing 
 groups, and that persons will attribute these characteristics to a 
 group based upon mere knowledge of its performance. So as to provide 
 cross-validation of this attribution process, r.n "interpersonal simula- 
 tion" (Bern, 1965) was also performed. 
 A Cross-valida ting Jlnterpers ona 1 $ imula ticn 
 
 In order to provide specific da ,a on the stereotypes individuals 
 hold and the attachment of these stereotypes to nigh arid low 
 performing groups, an ''interpersonal simulation" (Ben, 1965) was 
 conducted. As described below, th ;dy providea direct data on the 
 
 attribution process in addition to important cross-validation of the 
 experimental findings. 
 
 For the interpersonal simulation, sixty students */ere asked to 
 participate in a study on perceptual accuracy. They were told that a 
 large number of undergraduate business students had previously participa 
 
16 
 in a group problem-solving study in which measurements were taken of 
 intragroup processes and performance. Subjects were told that the 
 researchers were interested in seeing how accurately individuals could 
 assess intragroup processes based upon a minimal amount of information, 
 and that their assessments would be compared to "true" observational 
 measures of group processes collected over the past year. The 
 "Financial Puzzle Task" (as used in the above experiment) was then 
 thoroughly described to the subjects in both written and oral form. 
 Subsequently, subjects were asked to rate a typical group of business 
 undergraduates who had performed in the lowest (or highest) 20% of all 
 three-man groups. Via random assignment, thirty subjects were asked 
 to rate a high performing group anc thirty a low performing group. 
 Efforts were made to keep the rating scales as similar as possible to 
 those used in the previous experiment, 
 
 I_ns_e r t_Tab le_3__a bou t her e 
 As shown in table 3 the results of the "interpersonal simulation" 
 followed closely those of the previous study. High performing groups 
 were perceived to be higher in cohesiveness , total influence, quality 
 and quantity of communication, motivation, and openness to change than 
 low performing groups. As in the previous experiment, interpersonal 
 conflict (i.e. low group cohesiveness) was negatively related to 
 performance, while task conflict tended to be positively associated 
 with performance. Likewise, total influence was perceived to be 
 greater in high rather than low performing groups. However, because 
 persons in the interpersonal simulation did not actually participate 
 

 
 
17 
 a problem-solving group, total influence was not measured by a combinat 
 of the rateo influence of self and .. ie's teammates. Iistead, total 
 influence was measured by. 1) combining the perceived influence scores 
 for the "most influential" and "least inf luen.-H.al" persons in the 
 group, and, 2) by simply asking subjects to rate the influence of eac^i 
 group member. By either of these methods, cotal influence appeared 
 to be positively associated with group performance. 
 Conclusions 
 
 The data of the true experiment and the interpersonal simulation, 
 together, provide strong evidence for the attribution effect. The 
 similiarity of results from these two studies demonstrate that mere 
 knowledge of performance may cause an indivudual to attribute one set 
 of characteristics to a high performing group and a different set of 
 characteristics to a low performing group. Supported by these data, 
 the attribution effect thus constitutes a very plausible interpretation 
 of correlations linking perceived group characteristics to work group 
 performance. Moreover, though not yet specifically tested, this same 
 attribution orocess ma^ underlie many correlations between self-report 
 data on individual characteristics (e.g. attitudes, perceived role 
 conflict and ambiguity, perceived effort) and individual performance 
 data, as well as many correlations between self-re] jn organi- 
 
 zational variables (e.g. openness, conflict, goal orientation, climate) 
 and organizational performance data. In sum, the orocess by which 
 individuals attribute the ''causes" of performance may have important 
 implications for the conduct of organizational research. 
 
18 
 
 From the data presented here, the attribution effect can be viewed 
 as potentially more threatening to t he interpretation of correlational 
 findings than the simple reversal of causal sequences. As noted by 
 Lowin and Craig (1968) and Farris and Lim (1969), an assumed direction 
 of causation may be incorrect since performance can affect actual 
 interpersonal behavior. However, actual reversals in causation do 
 not always occur and often it is possible for the researcher to 
 discount the probability of their occurrence on logical and theoretical 
 grounds. In essence, the more intuitively obvious or plausible is a 
 particular causal sequence, the safer it is for researchers to discount: 
 its actual reversal. In direct contrast, the attribution interpretation 
 posits that organizational participants possess theories of performance 
 just as do organizational researchers. Thus, the more intuitively 
 obvious or plausible is a theory of organizational behavior, the more 
 likely is a correlation between self-report data and performance zo 
 be threatened by an attribution interpretation. Since there are no 
 doubt a greater number of obvious than ncn-obvious finaings in 
 organizational research, the attribution effect may therefore be a 
 greater threat to cross-sectional findings than ectual reversals in 
 causal order. 
 
 Clearly, a major problem still facing the field of organizational 
 behavior is a dearth of firm causal findings. The results of this 
 study, together with previous experiments on the effects of performance, 
 underscore the need for organizational research with methods more 
 conducive to causal inference. Three primary solutions to this dilemma 
 

 
 
 
 
19 
 
 have already been posited, but not yet widely adopted. First, by 
 conducting longitudinal studies us^ng cross-lag correlation procedures 
 (Pelz and Andrews, 1964; Vroom, 1967) there can be an improvement in 
 our knowledge of causal order. (It should be noted, however, that the 
 use of cross-lag correlational techniques implies equal time lag in 
 
 the causal links X > Y _ and Y . , ■ - ■■■ > X „) . Second, by 
 
 conducting true and (strong) quas i-experiments within organizations, 
 we may be able to increase the internal validity of our findings without 
 unduly sacrificing external validity (Campbell and Stanley, 1963; Cook 
 and Campbell, 1974). Both as consultants to planned organizational 
 changes and as documenters of naturally occurring organizational 
 changes (Staw, 1974), there are many opportunities to obtain data from 
 which causal inferences may be drawn. Third, it may be possible tc 
 constructively combine the advantages of laboratory and field methods 
 in the investigation of organizational processes (McGrath, .1964; 
 Evan, 1971). By coordinating laboratory and field studies (e.g. 
 terms of chosen variables and measurement instruments) the resultant 
 findings could be high in both internal and external validity. 
 
20 
 
 Footnotes 
 The author is indebted to Grej R. Oldham for his comments on an 
 earlier version of this paper, and to Ramamoorthi Narayan for serving 
 as an experimenter in this research. 
 
 Farris & Lim (1969) interpreted their data as knowledge of 
 performance affecting actual supervisory behavior. However, these 
 data can also be alternatively interpreted by an attribution 
 effect. Persons playing subordinate roles in the study may 
 have learned from their leaders that they were members of a 
 high or low performing work group, attributed this past performance 
 to the foreman's leadership capabilities, and then reported these 
 characteristics on the post-experimental measures of perceived 
 leadership behavior. It is therefore possible that knowledge 
 of performance did not affect actual supervisory behavior but 
 only subordinates' perceptions of it. 
 

 
 
 
 
21 
 
 References 
 
 Bern, D. J. An experiments 1 analys s of self -persuas^ on . Journal of 
 Experimental Social Psychology , 1965, 1, 199-218. 
 
 Bern, D. J. Self -perception: the dependent variable of human performance. 
 Organizational Behavior and Human Performanc e, 1967, _2, 105-121. 
 
 Bern, D. J. Se If -perception theory. In L. Berkowitz (Ed.), Advances in 
 Experimental Social Psychology . Vol. 6. New York: Academic 
 Press, 1972, pp. 1-62. " 
 
 Bruner, J. S. and Taguiri, A. The perception of people. In G. 
 Lindzey (Ed.) Handbook of Social Psychology , 1954. 
 
 Galder, B. J. and Staw, B. M. The interaction of intrinsic and 
 
 extrinsic motivation: Some methodological notes. Journal of 
 Personality and Social Psychology , in press. 
 
 Calder, B. J. ana Staw, B. M. The self-perception of intrinsic and 
 
 extrinsic motivation. Journal of Personality and Social P syc holog y, in press, 
 
 Cook, T. D. and Campbell, D. T. The design and conduct of quasi- 
 experiments and true experiments in field settings. In M.D. 
 Dunnette (Ed.), Handbook of Industrial and O rganizational Research , 
 in press . 
 
 Deci, E. L. Effects of externally mediaced rewards on intrinsic 
 
 motivation. Journal of Personality and Social Psychol ogy , 1971, 
 JL8, 105-115. 
 
 Evan, W. M. Conflict and perfcrma ce in R 6c D organizations. Industrial 
 Management Review , 1965, 7_, 37-45. 
 
 Farris, G. F. and Lim, F. G. Effects of performance on leadership, 
 
 cohesiveness , influence, satisfaction, and subsequent performance. 
 Journal of Applied Ps y chology , 1969, 53, 490-497. 
 
 Galbraith, J. and Cummings , L. L. An empirical investigation of the 
 
 motivational determinants of task performance: Interactive effects 
 between instrumentality-valence and motivation-ability. Organiz a tiona l 
 Behavior and H uman Perfor mance, 1967, 2, 237-257. 
 
 Heider, F. The Psychology of Interpersonal Relations . New York: Wiley, 
 1968. 
 
 Jones, E. E. and Davis, K. E. From acts to dispositions: The attribution 
 process in person perception. In L. Berkowitz (Ed.) Advance s in 
 Experimenta l Social Psychol ogy. Vol. 2. New York: Academic Tress, 1965. 
 
22 
 
 Jones, E. E., Davis, K. E. and Gergen, K. E. Role playing variations 
 and their informational value for person perception. Journal of 
 Abnormal and Social Psycholog y, 1961, _6_3, 302-310. 
 
 Jones, E. E. and Harris, V. A. The attribution of attitudes. Journal 
 of Experimental Soc ial Psychology, 1967, 3_, 1-24. 
 
 Kelley, H. H. Attribution in social interaction . New York: General 
 Learning Press, 1971. 
 
 Kelley, H. H. The processes of causal attribution. American Psychologist . 
 1973, 28, 107-128. 
 
 Likert, R. Measuring organizational performance. Harvard Business 
 Review , March-April, 1958. 
 
 Likert, R. New Patterns of Management . New York: McGraw-Hill, 1961. 
 
 Likert, R. Human Organization: Its Management and Value . New York: 
 McGaw-Hill, 1967. 
 
 Lowin, A. and Craig, J. R. The influence of level of performance on 
 
 managerial style: An experimental object-lesson in the ambiguity 
 
 of correlational data. Organizational Behavior and Human Performan ce , 
 
 1968, 2, 440-458. 
 
 Maier, N. R. F., Solem, A. R. and Maier, A. Supervisory and executive 
 development: A manual for role playing. New York: Wiley, 1957. 
 
 McGrath, J. Toward a "theory of method" for research on organizations. 
 In W. W. Cooper, H. J. Leavitt, and M. W. Shelby II (Eds.), New 
 Perspectives in Organizational Resea rch. New York: Wiley, 1964 
 
 Nisbett, R. E. and Valins, S. Perceiving the causes of one's own 
 behavior . New York. General Learning Press, 1971. 
 
 Pelz, E. C. and Andrews, F. M. Detecting causal priorities in panel 
 study data. Amer i can Socio l ogical Review , 1964, 2_9, 836-848. 
 
 Schachter, S. and Singer, J. E. Cognitive, social and physiological 
 
 determinants of emotional state. Psychological Review , 1962, 69 , 
 379-399. 
 
 Smith, C. G. and Tannenbaum, A. S. Organizational control structure: A 
 comparative analysis. Human Relation s , 1963, JL6, 299-316. 
 
 Staw, B. M. Notes toward a theory of intrinsic and extrinsic motivation, 
 paper presented at Eastern Psychological Association , 1974. 
 
23 
 
 Staw, B. M. The attitudinal and behavioral consequences of changing a 
 major organizational reward, Journal of Personality and Social 
 Psychology , In press . 
 
 Staw, B. M. The psychology of intrinsic and extrinsic behavior . New 
 York: General Learning Press, in progress. 
 
 Strickland, L. H. Surveillance and trust. Journal of Personality , 
 1958, _2&> 200-215. 
 
 Tannenbaum, A. S. Control in organizations: Individual and Organizational 
 performance. Administrative Science Quarterly , 1962, 7, 236-257. 
 
 Tannenbaum, A. S. Control in organizations . New York: McGraw-Hill, 1968. 
 
 Vroom, F. H. A comparison of static and dynamic correlational methods 
 in the study of organizations. Organizational Behavior and Human 
 Performance , 1966, _1, 55-70. 
 
 Weiner, B., Frieze, I., Kukla, A., Reed, L., Rest, S., and Rosenbaum, 
 R. M. Perceiving the c a uses of success and failure . New York: 
 General Learning Press, 1971. 
 

24 
 
 Table 1 
 
 Effect of Knowledge of Performance 
 on Individual Percept! ns of Intragroup Processes 
 
 Low High t 
 Performance Performance Value 
 
 Cohesiveness 
 
 Cohesiveness of group 
 Enjoy working with teammates 
 Liking for teammates 
 Combined cohesiveness score 
 
 6.70 
 
 7.83 
 
 1.68** 
 
 7.23 
 
 8.2: 
 
 1.81** 
 
 8.77 
 
 9.23 
 
 1.04 
 
 7.57 
 
 8.43 
 
 1.72** 
 
 Influence 
 
 Teammates influence on task 
 
 solution 7.57 7.43 -.24 
 
 Own influence on task solu- 
 tion 
 
 Combined influence score 
 
 Communication 
 
 6.00 
 
 7.73 
 
 2.47*** 
 
 6.78 
 
 7.58 
 
 1.86** 
 
 6.77 
 
 7.93 
 
 1.75** 
 
 6.47 
 
 7.30 
 
 1.33 
 
 6.62 
 
 7.61 
 
 1.77** 
 
 Quality of communication 
 Quantity of communication 
 Combined communication score 
 
 Task Conflict 
 
 Differences in ideas about 
 
 methods to solve problem 4.83 4.93 .17 
 Confrontation of ideas with 
 
 teammates 5.34 7.03 1.97** 
 
 Openness to Chan ge 
 
 Openness of teammate to ideas 
 and suggestions about sol- 
 ving problem 7.73 8.55 1.52* 
 
 Extent teammate attempted to 
 force his position on you 
 (scale reversed) 8.53 9.21 1.02 
 
 Combined openness score 8.14 8.88 1.49* 
 
 * p< .10, one- tailed test 
 
 ** p< .05, one-tailed test 
 
 *** p< .01, one-tailed test 
 

 
25 
 
 Table 2 
 
 Effect of Knowledge of Performance on 
 Satisfaction, Motivation, Ability, & Role Clarify 
 
 Low High t 
 Performance Performance Value 
 
 Motivation 
 
 Teammates' interest in 
 
 
 
 
 performing well 
 
 4.67 
 
 7.90 
 
 3.87*** 
 
 Own interest in per- 
 
 
 
 
 forming well 
 
 4.73 
 
 7.47 
 
 5.33*** 
 
 Combined motivation score 
 
 4.70 
 
 7.68 
 
 5.24*** 
 
 Ability 
 
 
 
 
 Teammates' ability 
 
 Own ability 
 
 Combined ability score 
 
 5.50 
 
 7.13 
 
 2.60*** 
 
 3.57 
 
 6.80 
 
 5.64*** 
 
 4.54 
 
 6.96 
 
 5.00*** 
 
 Satisfaction 
 
 Enjoyed working on financial 
 task 
 
 3.47 
 
 7.20 
 
 5.93*** 
 
 Role Clarity 
 
 Clarity of instructions for 
 the task 
 
 7.23 
 
 8.70 
 
 2.20** 
 
 * p4 .10, one-tailed test 
 
 ** p<< .05, one-tailed test 
 
 *** p< .01, one-tailed test 
 

26 
 
 Table 3 
 
 Effects of Knowledge of 
 Performance for Int< rpersonal Simulation 
 
 Low High 
 Performance Performance 
 
 Cohesiveness 
 
 Cohesiveness of group 
 
 Enjoyed working with, teammates 
 
 Liking for teammates 
 
 Combines cohesiveness score 
 
 Influence 
 
 Influence of each member 
 
 Influence of "most inf luencial" 
 member 
 
 Influence of "least influencial" 
 member 
 
 Combined influence score 
 
 Communication 
 
 Quality of communication 
 Quantity of communication 
 Combined communication score 
 
 Task Conflict 
 
 Difference in ideas about 
 
 methods to solve problem 
 
 Confrontation of ideas with 
 teammates 
 
 Combined task conflict score 
 
 5.17 
 
 8.90 
 
 2.60 
 5.75 
 
 6.80 
 
 5.30 
 6.05 
 
 6.97 
 
 8.90 
 
 4.03 
 6.47 
 
 6.50 
 
 7.03 
 6.77 
 
 t 
 Value 
 
 3.00 
 
 8.67 
 
 17.18*** 
 
 4.10 
 
 8.10 
 
 10.62*** 
 
 4.93 
 
 7.50 
 
 7 . 1 7*** 
 
 4.01 
 
 8.09 
 
 15.46*** 
 
 3.49*** 
 
 .00 
 
 2 . 74** 
 2.21* 
 
 2.93 
 
 8.80 
 
 17.08*** 
 
 4.50 
 
 8.37 
 
 8.22*** 
 
 3.72 
 
 8.58 
 
 14.47*** 
 
 -.53 
 
 2 . 98** 
 
 + 
 1.53 
 
27 
 
 Table 3 
 (Continued) 
 
 Low High 
 Performance Performance 
 
 Openness to Change 
 
 Openness to ideas & suggestions 
 
 about solving problem 4.27 
 
 Extent group members ever attempted 
 
 to force their positions 4.03 
 
 (scale reversed) 
 
 Combined openness score 4.65 
 
 Motiva tion 
 
 Group members' interest in 
 
 performing well 3.33 
 
 Ability 
 
 Rated ability of group on task 2.90 
 
 Role Clarity 
 
 Clarity of instructions for the 
 
 task 6.50 
 
 8.07 
 
 4.30 
 
 6.68 
 
 8.30 
 
 8.93 
 
 t 
 Value 
 
 7.27*** 
 
 .40 
 
 5 . 04*** 
 
 11.84*** 
 
 19.13*** 
 
 3.37 
 
 4 . 2 7*** 
 
 + 
 
 p<.10, one-tailed test 
 *p<.05, one- tailed test 
 **p<.0l, one-tailed test 
 ***p<.001, one-tailed test 
 

ond eT* 
 
 1-9»