SOCIAL LOAFING AND SOCIAL COMPENSATION 573 Wee would carry ther the group task.Participants cause th would tend to expect others to loaf and on to writ harder so as to maintainr e uses Method and de iocenterreente ded out a rietquestioanai plit-hal s of 68 and5 on tw eted the question and dismissed. Result Performance measures.Overall,there was a significant loaf ing more when wor (high tru 120)=5.51,p<.03.More m the esults ort the cdtoptrto The the task nt to the pa Table1).A prio contrasts rev led tha pro t in the time prov ded.Inou and 21 (39) (de Ccondtionl co of 4 to 8w greeted by the expe e led to a re iddle of the chairs w 20)=4.03,p< .03,in that medi other.On ped pen uld b s that varied a ad th along Participants in eareinterese meani chec hty (0 e f an ct an tas be to 49 tifa ipants was not significant (20).Debrief and su ses you c tant t as mar eve th An other question asked participar 00-in nted the Supporting the notion that low trusters expected their co- ing: 'lnibothEpcrimeatland3,thereappearedtobeomehct me log tra uses as your partners I want to sce how many useach iormationrcvealedidenticalpaternsofmeansandsignih cant SOCIAL LOAFING AND SOCIAL COMPENSATION 573 We expected participants scoring high or medium on trust to loaf collectively, because they would tend to expect that others would carry their own weight on the group task. Participants scoring low in trust were hypothesized to compensate collec￾tively because they would tend to expect others to loaf and would feel compelled to work harder so as to maintain a favor￾able evaluation of themselves. Method Participants and design. In a pretest session at the beginning of the quarter at the University of Washington, 1,085 introductory psy￾chology students responded to Rotter's (1967) Interpersonal Trust Scale. Rotter reported test-retest coefficients of .68 and .56 on two samples, p < .01, and a significant split-half reliability coefficient of .76 on a sample of 547 college students, p < .001. From the resulting normal distribution of this scale, 126 students were called back to participate in the experiment but were not told their participation was related to their test scores. Forty-two of these participants scored in the bottom quintile (low trusters) of the pretest sample, 41 scored in the middle quintile (medium trusters), and 43 scored in the top quintile (high trusters). Experimental task. The additive task of idea generation was chosen for two reasons. First, effort would be directly related to performance. Second, the task could be presented in a meaningful way so that perfor￾mance on the task would be important to the participants. The idea generation task requires participants to come up with as many uses as possible for a given object in the time provided. In our study, participants were asked to come up with as many uses as possible for a knife in a 12-min period. They were instructed to write each use on a separate slip of paper and insert the slip into the appropriate box (depending on the condition). Procedure. Upon entering the waiting area, participants in groups of 4 to 8 were greeted by the experimenter and were led to a room where they were seated in one of eight chairs, arranged in a circle. Between each set of chairs and directly in the middle of the chairs were black cloth partitions that prevented the participants from seeing each other. On each desk was a felt-tipped pen (so that it would be impossi￾ble for participants to monitor each other's performances). Partici￾pants were then handed instruction booklets that varied according to the coactive/collective work condition. The experimenter read the in￾structions aloud while the participants followed along. Participants in both conditions read the following instructions: We are interested in studying the performance of groups and indi￾viduals on what is called a "brainstorming" task. 'You will be given the name of an object and your task will be to come up with as many uses for this object as you can. Don't be concerned about the quality of the uses you come up with. The uses can be ordinary or unusual. It is, however, important that you write down as many uses as you can in the time allotted. We are interested in the num￾ber of uses that can be generated for a single object, so all of you will be generating uses for the same object. The next portion of the written instructions varied depending on the work condition. Participants in the coactive condition read the follow￾ing: I will ask you to write one use on each slip of paper, and then put that slip of paper into the box at your side before you write down the next use. It is not important if you happen to come up with the same uses as your partners. I want to see how many uses each of you can come up with. Thus, each of you bears the full responsibil￾ity for generating as many uses as you can. Participants in the collective condition read the following: I will ask you to write one use on each slip of paper, and then insert that slip of paper into the common box in the middle. It is not important if you happen to come up with the same uses as your partners. I want to see how many uses your group can come up with. Thus, you share the responsibility with your partners for coming up with as many uses as you can. The experimenter told the participants to begin, started his stop￾watch, and then left the room. When the task was completed, the ex￾perimenter reentered the room and handed out a brief questionnaire with items assessing various participant perceptions as well as manipu￾lation checks. Participants then completed the questionnaire while the experimenter was out of the room. Participants were then debriefed and dismissed. Results Performance measures. Overall, there was a significant loaf￾ing effect, with participants generating more uses when work￾ing coactively (27.7 uses) than collectively (24.5 uses), F(l, 120) = 5.51, p < .03. More important, the results support the social compensation hypothesis, with a significant Trust X Work Condition interaction, F(2,120) = 10.59, p < .0001 (see Table 1). A priori contrasts revealed that medium trusters dem￾onstrated a significant social loafing effect, producing 25.5 uses in the coactive condition and 21.1 in the collective condition, r(39) = 2.00, p < .03. The high trusters showed an even stronger loafing effect, producing 33.2 uses coactively and 22.6 uses col￾lectively, t(41) = 4.29, p < .001. As predicted by the social com￾pensation hypothesis, low trusters actually worked harder col￾lectively (29.6 uses) than coactively (24.5 uses), r(40) = -1.95, p < .03. There was also a significant main effect for trust, F(2, 120) = 4.03, p < .03, in that medium trusters (23.3 uses) were less productive than either low (27.1) or high trusters (27.5).3 There were no significant effects for gender. Ancillary data. All questions used 100-point scales. As ex￾pected, the task was regarded as being relatively easy (M = 19.9). All participants regarded the task as somewhat fun (A/= 54.2), and felt they experienced about an average amount of anxiety (M = 46.9). These data suggest that the participants found the task to be at least moderately meaningful. A process check on the differences in perceived identifiability (0 = not at all and 100 = very) of coactive (M = 49.3) versus collective (M = 46.3) participants was not significant (p > .20). Debriefing and sub￾sequent research, however, has led us to believe that partici￾pants had difficulty interpreting this particular question. An￾other question asked participants if they thought the others in their session generated fewer, the same, or more uses than they did (on a 100-point scale in which 50 represented the same). Supporting the notion that low trusters expected their co- 3 In both Experiments 1 and 3, there appeared to be some heterosce￾dasticity in the number of uses data. Reanalyses of these data sets using a log transformation revealed identical patterns of means and signifi￾cant effects