574 KIPLING D.WILLIAMS AND STEVEN J.KARAL Tablel 0ygGigaRaoraKaeasaAnaono eheo Work condition performance on the task was thought to be related to intell Trust level(quintile) Coactive Collective thatsocialoainewoulloccrorparicipants 袋 Method s at pur SD r participa vorkers to reduce their efforts when workine collectively the only condition in ration task used in experiment which pa icipants though that their co Experime 2,wi地 to 49.1).Wh vere other asked hov ndthowninito in ants fr的 On each desk sa pair of headphones xt to them the floor.In th working coactively and working colectivel Discussion The behavioral data from exp ortfor the hypothesis that levels oft that is thou ors de eorcliminatcsocialloainthisndinedg The exper menter then told the participants that a recent theor artic strate a causal link between factors,it could be that other var th hceinsimia ealso told that once their uses had been counted.the uses rust ated b orker I are i mance Experiment 2 rork The purpose of Experiment 2 was to dir er's efforts.Rather have the work effort as an evaluation of the task's meaningfulness.574 KIPLING D. WILLIAMS AND STEVEN J. KARAU Table 1 Mean Number of Uses Generated for a Knife as a Function of Levels of Trust and Work Condition Work condition Trust level (quintile) Coactive Collective Low M n SD Medium M n SD High M n SD 24.45 20 7.73 25.48 21 9.19 33.20 20 10.04 29.55 22 9.05 21.05 20 3.71 22.61 23 5.86 workers to reduce their efforts when working collectively, the only condition in which participants thought that their co￾workers generated fewer uses than they did (M= 42.91) was the low trust, collective condition (95% confidence interval = 36.7 to 49.1). When asked how much effort participants thought people generally exert on collective tasks compared with indi￾vidual tasks, they indicated (as has often been the case in pre￾vious social loafing research) that there would not be any differ￾ence (M = 50.8, in which 50 represented the same) between working coactively and working collectively. Discussion The behavioral data from Experiment 1 provide strong sup￾port for the hypothesis that levels of trust determine whether social loafing or social compensation will occur. Whereas other research in social loafing has shown that various factors de￾crease or eliminate social loafing, this finding demonstrates a significant increase in effort by individuals on a collective task. Social loafing once again has been demonstrated for the me￾dium trusting person (assuming this typifies the average partici￾pant in past research), and even more so for the high trusting person. Low trusters, however, were actually more productive when working collectively than when working coactively. Of course, as with any study using individual differences to demon￾strate a causal link between factors, it could be that other vari￾ables associated with trust accounted for such a pattern. Fur￾thermore, we did not assess directly our assumption that low trusters compensated because they thought their co-workers were loafing. Although the results from Experiment 1 are in￾triguing and suggestive, we felt that it was necessary to submit the social compensation hypothesis to more direct tests by ma￾nipulating experimentally expectations of co-worker perfor￾mance. Experiment 2 The purpose of Experiment 2 was to directly manipulate the participants' expectations of their co-worker's efforts. Rather than link the co-worker's intended effort with his or her ap￾praisal of the task, as did Jackson and Harkins (1985), we kept the appraisal constant and positive throughout. To ensure that the task was perceived as meaningful, we told participants that performance on the task was thought to be related to intelli￾gence. Two levels of co-worker effort (high or low) were crossed with two work conditions (coactive or collective). We predicted that social loafing would occur for participants working with a high effort co-worker, whereas social compensation would oc￾cur for those working with a low effort co-worker. Method Design and participants. The experiment used a 2 (low vs. high co-worker effort) X 2 (coactive vs. collective work condition) between￾subjects factorial design. Participants were 49 introductory psychology students at Purdue University who fulfilled partial course credit by their participation. Ten participants were eliminated from the final analyses because they indicated possible suspicions regarding loafing (they had read about it in the textbook), the authenticity of the confeder￾ate, or the experimental hypotheses.4 This resulted in a total of 39 participants. Procedure. The idea generation task used in Experiment 1 was used in Experiment 2, with slight modifications. Two participants (one of whom was a male confederate) were greeted by the male experimenter and shown into the laboratory. They were seated next to each other and were separated by a large filing cabinet that prevented the two partici￾pants from seeing one another. On each desk was a pair of headphones, a pen, and a box of blank slips of paper. In the coactive condition, participants had separate boxes placed next to them on the floor. In the collective condition, a common box (with an opening in the top accessi￾ble to both participants but preventing each from seeing inside) was placed between the two tables. The experimenter then read similar instructions as were read in Experiment 1. In addition, participants were told: \bu both will be working on a "brainstorming-type" task. The purpose of this experiment is to examine rapid thinking, a quality that is thought to be highly correlated with intelligence in adults. Your task is to come up with as many uses for an object as you can. \bu will be working side by side, without communicating. The experimenter then told the participants that a recent theory suggested that rapid thinking was highly correlated with intelligence, and that it was therefore extremely important that they come up with as many uses as they possibly could. The experimenter also told partici￾pants that he would count up their individual scores (or group scores, if in the collective condition) at the end of the session and tell them how many uses they had produced. In addition, their scores would be com￾pared with those of other participants who had been in similar re￾search studies at other universities. Participants in the collective condi￾tion were also told that once their uses had been counted, the uses 4 The distribution of suspicious participants was approximately even across three of the four conditions, with no suspicious partici￾pants in the high effort, coactive condition. The suspicious partici￾pants in the high effort, collective condition tended to believe the co￾worker's announcement to be sarcastic, thereby interpreting the mes￾sage to be the opposite, as was intended. In the low effort conditions, suspicious participants found it difficult to believe that their co-worker would not work hard on a task they found interesting. We felt it was necessary to have the co-worker state that the task was interesting so participants would not interpret the co-worker's statement of intended effort as an evaluation of the task's meaningfulness