Persistence of Brainstorming Groups: How Do People ... - Science Direct

Journal of Experimental Social Psychology 35, 165–185 (1999) Article ID jesp.1998.1374, available online at http://www.idealibrary.com on

Persistence of Brainstorming Groups: How Do People Know When to Stop? Bernard A. Nijstad, Wolfgang Stroebe, and Hein F. M. Lodewijkx Utrecht University, Utrecht, The Netherlands Received March 31, 1998; revised August 31, 1998; accepted September 7, 1998 Two experiments compared task persistence and productivity for brainstorming individuals and groups of various sizes under conditions without a time limit. We expected that the stop rules people employ in deciding to end a session would have important consequences for the relation between group size, persistence, and productivity. In Experiment 1, conducted with individuals, dyads, and four- and six-person groups, it was found that persistence increased linearly with group size. As a result, the productivity loss usual in brainstorming groups was substantially reduced. This finding was replicated in Experiment 2, but only under conditions without time pressure. The degree to which a brainstorming topic was perceived as enjoyable did not affect persistence. Findings were consistent with the hypothesis that people stop neither when they are satisfied with their performance nor when they no longer enjoy the task, but when the expected relation between effort and performance falls below a certain cut-off point. The theoretical and practical implications of these findings are discussed. r 1999 Academic Press

The generation of ideas is a first stage of many group tasks, such as decision making and problem solving. In the 1950s, Osborn (1957) suggested brainstorming as a method to improve group idea generation. Brainstorming is characterized by the use of four rules, which serve to minimize inhibiting effects of evaluation and maximize possibilities for mutual stimulation: people are instructed to generate many ideas, to think of uncommon ideas, to combine and improve ideas, and to refrain from criticism. However, Osborn’s (1957) claim that group interaction facilitated performance has not been supported by empirical research. There is a great deal of evidence that nominal groups, consisting of a number of individuals who work alone, outproduce interactive groups of the same size by a large margin (for reviews, see Diehl & Stroebe, 1987; Mullen, Johnson, & Salas, The help of D. Bijleveld, T. Eikelboom, S. Gomperts, D. de Graaff, and M. de Kleuver during data collection and preparation is gratefully acknowledged. This research has been funded by grant 575-31.007 of the Netherlands Organization for Scientific Research (NWO) awarded to W. Stroebe. Address correspondence and reprint requests to B. Nijstad, Department of Social and Organizational Psychology, Utrecht University, P.O. Box 80140, 3508 TC Utrecht, The Netherlands. 165 0022-1031/99 $30.00 Copyright r 1999 by Academic Press All rights of reproduction in any form reserved.

jesp 1374 @xyserv1/disk3/CLS_jrnl/GRP_jesp/JOB_jesp35-2/DIV_326a02 marg No. of Pages—21 First page no.—165 Last page no.—185

166

NIJSTAD, STROEBE, AND LODEWIJKX

1991). This productivity loss of interactive brainstorming groups is already present in three-member groups and increases with group size (Mullen et al., 1991). In the last decade, significant progress has been made in explaining the productivity loss of interactive brainstorming groups. Diehl and Stroebe (1987, 1991) studied the effect of mutual production blocking, which arises from the constraint that in a group only one person can speak at a given time. As a consequence, group members, unlike individuals, cannot always verbalize ideas as they occur, but instead have to wait to make their contribution. Diehl and Stroebe (1987, 1991) found empirical support for blocking as an important cause of the productivity loss of interactive brainstorming groups. An alternative explanation of the productivity loss has been suggested by Paulus and Dzindolet (1993), who argued that group members will tend to compare and match their performance to that of others. The resulting group standard will be quite low, because typically there are no incentives to perform well in brainstorming experiments. In addition, this production standard is set at the beginning of the session, when productivity suffers most from the blocking effect and evaluation apprehension. Paulus and Dzindolet (1993; see also Camacho & Paulus, 1995) found some support for this line of reasoning. One issue which has not been addressed in previous research is the question of how long groups of various sizes and individuals continue brainstorming without externally imposed constraints. The issue of brainstorming persistence is interesting because there is evidence that longer sessions are associated with higher productivity (e.g., Diehl & Stroebe, 1991; Kanekar & Rosenbaum, 1972). Although the rate of production generally declines in the course of a session, spending more time on the task will eventually result in a greater number of ideas.1 If brainstorming groups proved to be more persistent than individuals, this would provide an opportunity to reduce or even reverse their usual productivity loss. The issue of brainstorming persistence cannot be adequately addressed within the traditional experimental brainstorming paradigm where the same time limit is given to individuals and groups. This so-called Equal Man-Hour-Comparison (Stroebe & Diehl, 1994) severely constrains the variation in brainstorming persistence between individuals and groups. Persistence therefore needs to be studied with groups and individuals who brainstorm without a time limit set by the experimenter. The study of brainstorming persistence under these conditions raises the question of how people decide when to stop and whether this decision is based upon a certain criterion or stop rule. Although there has been some interest in the effects of stop rules on persistence in the area of mood research (e.g., Hirt, Melton, McDonald, & Harackiewicz, 1996; Martin, Ward, Achee, & Wyer, 1993; 1 There is, however, some evidence that time abundance leads to lower rates of production (e.g., Karau & Kelly, 1992). If the absence of time constraints induces an experience of time abundance, it may lead to lower rates of production, counteracting productivity gains due to increased persistence.

jesp 1374 @xyserv1/disk3/CLS_jrnl/GRP_jesp/JOB_jesp35-2/DIV_326a02

donn

167

PERSISTENCE

Sanna, Turley, & Mark, 1996), the question of how people decide to stop has never been investigated in group brainstorming. The reason for this lack of curiosity is probably that group brainstorming research has always been conducted with set time limits. However, the fact that brainstorming subjects typically stop generating ideas some minutes before the end of their allotted time (Diehl & Stroebe, 1987; Paulus, Dzindolet, Poletes, & Camacho, 1993) suggests that persistence may affect productivity even under these conditions. In the following section four stop rules will be discussed and their potential impact on brainstorming persistence examined. Stop Rules and Persistence Depletion. The simplest stop rule and one implicitly adopted by many brainstorming researchers is that individuals approach the brainstorming task with a limited pool of ideas and end the session when this pool has been depleted. According to the depletion stop rule, the productivity loss of brainstorming groups is due to the fact that, because of production blocking, group members need more time than individuals to express their ideas. The removal of time constraints should therefore result in greater increases in persistence of group than of individual brainstormers. Since blocking effects are exacerbated by group size, the discrepancy in persistence between individuals and groups should increase with group size. It needs little reflection to realize that the assumption of a limited pool of ideas is implausible both on theoretical and empirical grounds. It implies that ideas are somehow stored in memory and only need to be retrieved during the brainstorming session. This is inconsistent with the conception of brainstorming as a creative task (e.g., McGrath, 1984), in which new ideas are generated. In addition, there are empirical results that suggest an unlimited, or at least very large, pool of ideas. Most striking are the results from goal setting experiments (Hyams & Graham, 1984; Paulus & Dzindolet, 1993). When participants are provided with challenging and specific goals a dramatic increase in productivity results. From this result it appears that people can always generate more ideas when appropriately motivated. Expectancy. However, even though the depletion hypothesis is implausible as a scientific theory, it seems a plausible lay theory to brainstormers who often report that they have ‘‘run out of ideas’’ (e.g., Diehl & Stroebe, 1987). But if the potential pool of ideas is not limited, how can subjects come to believe that their ideas have been exhausted? In analogy to the availability heuristics (Kahneman & Tversky, 1973), we would argue that people base their estimate of the subjective probability of being able to produce further ideas on a given topic on the perceived ease or difficulty of producing ideas at a given point in time. When this probability falls below some criterion the session is terminated. This stop rule will be called the expectancy rule, because of its similarity to the construct of expectancy in expectancy valence theory (Vroom, 1964). Productivity generally declines in the course of a session (e.g., Diehl &


donn

168


Stroebe, 1991; Kanekar & Rosenbaum, 1972), because when more ideas have been suggested, it will be increasingly hard to come up with new ideas. This decline is likely to be less marked in groups because of the blocking effect. Since the competition for speaking time is especially severe in the early stages of the session when most group members are still able to generate ideas, per person production rate will be rather low. As a consequence, production will be more evenly distributed across the session and periods of silence are less prevalent in group than individual sessions. Because the frequency and length of periods of silence are likely to be salient indicators of the ease or difficulty of idea production, expectancy should be higher in groups than in individual settings. To derive a hypothesis regarding the effect of group size from the expectancy stop rule, several lines of reasoning are possible. First, larger groups suffer from an increased blocking effect, because the competition for speaking time is more severe. As a consequence production should be more evenly distributed across the session in larger groups. The difficulty of generating additional ideas will therefore rise at a less steep rate in larger groups and expectancy will remain higher. Second, pauses in which no group member is able to come up with an idea are less prevalent in larger groups, since, in absolute numbers, larger groups have a higher rate of production than smaller groups. These considerations lead to the prediction that expectancy remains higher in larger groups and that persistence will increase with group size. Satisfaction. A further stop rule is described in the mood literature (e.g., Martin et al., 1993) and will be called the satisfaction rule. This rule implies that people stop performing a task when they think that they have accomplished their goal or have attained a sufficient degree of task progress. How can group members or individuals know that they have done well with their brainstorming task? Group members may compare the ideas produced by the group with some individual standard such as the number of ideas which occurred to them at the beginning of the brainstorming session or the number of ideas they have been able to generate individually for similar problems in the past. In comparison to this individual standard, the group product is likely to appear quite substantial and to produce feelings of accomplishment. Once the discrepancy between their individual standard and the group product exceeds a certain level, group members may feel that they have accomplished their task and that it is time to stop. This type of comparison has been called the ‘‘baseline fallacy’’ (Stroebe, Diehl, & Abakoumkin, 1992), because it disregards group size in evaluating productivity. Brainstorming individuals, on the other hand, only see the ideas they are producing themselves. Their performance may therefore appear much less impressive and less satisfactory, which should motivate them to continue brainstorming. Thus, according to this hypothesis, we would expect individuals to be more persistent than groups. This discrepancy should increase with increasing group size, because larger groups are likely to produce more ideas than smaller groups in a given period of time, and according to the baseline fallacy production should appear more impressive in large groups.


donn

169

PERSISTENCE

Enjoyment. According to a fourth stop rule, the enjoyment rule (cf. Martin et al., 1993), participants ask themselves whether they still enjoy the task and end it when they become bored. Task enjoyment is partly related to productivity, since brainstorming will not be enjoyable when one is unable to come up with ideas. However, task enjoyment is likely to be related to other aspects of the social situation as well, most markedly to the interaction with fellow group members. If task enjoyment is used as a criterion to end the task, we expect groups to be more persistent than individuals, simply because group members can interact with each other and enjoy the session more than individuals. The effect of group size on task enjoyment and persistence is more difficult to predict. It is likely to depend on the extent to which participants are task motivated. Because the competition for speaking time increases with group size and the discussion is often dominated by a few group members (see Levine & Moreland, 1990), task enjoyment may be negatively related to group size for highly task motivated individuals. On the other hand, larger groups also offer greater anonymity and the opportunity to free ride (Stroebe & Frey, 1982) and could therefore be more enjoyable for individuals whose task motivation is low. This would lead to the prediction that groups are more persistent than individuals but that this discrepancy either decreases or increases with increasing group size. Overview The present article examines the relationship between brainstorming productivity and persistence in order to examine whether differential persistence compensated or aggravated the productivity loss of brainstorming groups. Two experiments are reported in which subjects who worked either in groups of various sizes or individually were allowed to make the decision to end the brainstorming session themselves. These are the first experiments that compare the persistence of brainstorming groups to that of individuals. The operation of stop rules was explored on the basis of subjects’ self-reports as well as the impact of type of setting (individual vs group) and group size on task persistence. EXPERIMENT 1 Ideally, to assess task persistence in an unbiased way, no time constraints should be imposed upon participants. However, it is not possible to sign up participants for an experiment without specifying the time needed. Such a time limit may act as a demand characteristic and induce participants to fill their allotted time. To solve this problem, participants have to be persuaded that persistence on the brainstorming task is unrelated to the total time spent in the laboratory. One way to do this is to tell them that a second task will follow after the brainstorming task and will take up the remaining time. An additional advantage of this procedure is that the effects of time costs will be minimal, and persistence will be unaffected by such considerations. Obviously, ample time has to be given to avoid ceiling effects. In the first experiment this procedure was used to examine the persistence of


donn

170


individuals and groups of various sizes. The major goals of this experiment were to see whether groups or individuals are more persistent and to assess which stop rule participants apply in deciding to end the task. The first question can be addressed with a simple comparison of the time devoted to the brainstorming task by individuals and groups; the second question requires, in addition, analyzing the relation between group size and persistence. It is further possible to compare the idea production of real groups of a given size to that of nominal groups, to establish whether groups can compensate or even overcompensate their productivity loss by showing higher levels of persistence. This last procedure can also be used to test the depletion hypothesis that, although there will be a positive relation between group size and persistence, per person productivity will be constant across groups of various sizes. As an additional check on depletion, participants were asked to generate ideas on the same problem for a second time, immediately after they had terminated the session. If individuals had exhausted their pool of ideas during the first session, they should have been unable to generate a substantial number of new ideas at this later time. Method Participants and Task Participants were 122 students (56 male and 66 female) from Utrecht University, who volunteered to take part in the experiment and were paid DFL 20 (at that time approximately $12) for their participation. They were told that the experiment would take 2 h, which, according to a pretest, was sufficient to avoid ceiling effects. Participants had to generate ideas on the topic of how the number of tourists visiting Utrecht could be increased. Similar topics have been used in previous research (e.g., Paulus & Dzindolet, 1993).

Independent Variables Group size. Participants brainstormed either alone (n 5 26), as member of a dyad (n 5 9 dyads), or as member of a 4-person group (n 5 9 groups) or a 6-person group (n 5 7 groups). Gender. All groups were homogeneous with regard to gender. In total there were 14 males and 12 females who brainstormed individually, 4 male and 5 female dyads, 4 male and 5 female 4-person groups and 3 male and 4 female 6-person groups.

Procedure The experiment was carried out in four rooms. One was the control room, and the others were used to run three parallel sessions. For each session up to nine participants were enrolled, who were either all male or all female. Participants were randomly assigned to conditions, but if one or more participants did not show up they were sometimes run in a different condition. Upon arrival, participants were led to a large room, where instructions were given collectively. First, all participants were asked to hand in their watches. It was explained that this was done to reduce feelings of time pressure. Then the instruction was read to them. Participants were told that the experiment would take 2 h and would consist of two tasks, the first being a brainstorming task. They were next informed about the four brainstorming rules. They were then told that they had to decide themselves when to end the first task. They were instructed to end the session when they felt it was a good time to stop. They would even be free to use the whole 2 h for the first task. When there were no more questions regarding the first task, the participants were, if necessary, divided into several groups. Participants who knew each other were separated. Following this, they


donn

171

PERSISTENCE

were provided with the topic of the brainstorming session and were led to their rooms. In every room there was a table, with a microphone in the middle, which was connected to a recorder in the control room and was used to record the brainstorming session. Further there was a signaling device with a light and a button that was connected to the control room. Participants were requested to start brainstorming when the light was switched on. When they felt it was a good time to stop, they had to push the button. The experimenter then recorded the time the participants had spent at the task. After they had finished the first task, participants were requested to come to the control room, where they individually filled out the postexperimental questionnaire. After everyone had completed the questionnaire, participants had to brainstorm individually for 5 min on the same topic that had been used in the first session. They were instructed to generate ideas which had not been mentioned in the previous session. A bonus (DFL 50; approximately $30) was promised for the best performance. Finally, participants were paid for their participation, debriefed, and dismissed.

Dependent Variables The two major dependent variables were persistence (time spent on the first task) and productivity (number of nonredundant ideas generated during the first task). In addition, the number of new ideas generated during the second task was determined. The questionnaire administered after the first task contained an open-ended question about why the participant had ended the task. In addition, there were two 9-point scales, with only the extremes marked, that were related to different stop rules: participants were asked how much they enjoyed the task (very much–not at all) and how satisfied they were with their performance (not satisfied–very satisfied). Finally, participants were asked to indicate what they thought of the topic of the session (very boring–very interesting).

Data Analytic Strategy In the brainstorming literature, productivity of interactive groups is usually compared to productivity of nominal groups. This procedure is used because the sum of the nonredundant ideas produced in a nominal group is a good estimate of the potential productivity of an interacting group of the same size (Steiner, 1972). For the persistence data, however, this procedure is not appropriate, because one is interested in a comparison between the average persistence of individuals and the average persistence of groups. To create nominal groups in this case would involve averaging persistence twice, first at the level of the nominal groups and then within conditions. Therefore, persistence data are analyzed using a 4 (group size: individuals, dyads, four- and six-person groups) 3 2 (gender) ANOVA. In the case of productivity, however, the comparison between nominal and real groups allows us to test whether the real groups of different sizes suffer a productivity loss even after time constraints have been removed. Nominal dyads and four- and six-person groups were created by randomly selecting individuals, pooling their production, and removing duplicates. Since, for reasons of (subject) economy, the nominal groups of different sizes were created using the same pool of individual brainstormers, the productivity data of the nominal groups are not independent. Therefore, the type of group (real or nominal) cannot be entered as a separate factor into an ANOVA. Instead, the mean production of the nominal groups of a given size was subtracted from the productivity of the real groups of the same size, thus creating a variable reflecting the productivity gains or losses of real as compared to nominal groups. This variable was subjected to a 3 (group size: dyads, four- and six-person groups) 3 2 (gender) ANOVA. Finally, the questionnaire items and the productivity of the second session were analyzed using 4 (group size) 3 2 (gender) ANOVAs. For participants who brainstormed individually the data are independent, but this is not true for participants who worked in groups, who may have influenced each other. Therefore, the questionnaire items were averaged within groups and the resulting group average was subjected to the ANOVAs. Intraclass correlations were computed for all these items and are reported when significant.

Scoring of Productivity Data Tapes were transcribed by two research assistants, who marked ideas in the text of the transcriptions and counted the number of ideas on each tape. To determine interrater reliability this procedure was


donn

172

NIJSTAD, STROEBE, AND LODEWIJKX TABLE 1 EXPERIMENT 1: MEAN PERSISTENCE AND PRODUCTIVITY LOSSES COLLAPSED OVER GENDER

Persistence (minutes) Productivity loss (No. ideas)

Individuals (n 5 26)

Dyads (n 5 9)

Four-person groups (n 5 9)

Six-person groups (n 5 6)

30.20 (18.32) —

34.48 (14.30) 2.43 (43.45)

39.76 (12.07) 238.74 (77.20)

44.14 (6.48) 292.67 (88.47)

Note. Standard deviations are in parentheses.

repeated by a second rater for a random subset of 12 tapes. The correlation between these two counts was .97. The agreement on separate ideas scored by both raters was 84%. (The latter index is an underestimation of reliability, because instances where both raters agreed that a certain excerpt did not contain ideas are not incorporated.) Next, nominal groups consisting of two, four, and six individuals were created, by randomly selecting individuals, pooling their ideas, and removing duplicates. For a subset of eight transcriptions a second rater independently rescored the duplicates, which, using the Diehl and Stroebe (1987) formula, resulted in an agreement of 99.60%. Finally, the additional ideas generated in the second session were compared to the ones generated in the first session and duplicates were removed, to assess whether participants were able to generate new ideas in the second session. Again a second rater repeated this for a random subset of eight transcriptions, resulting in almost perfect agreement (99.99%).

Results Persistence Data Preliminary analyses of the persistence data revealed that one of the male six-person groups was an outlier, scoring about three standard deviations below the mean of the six-person groups. This group was removed from all the analyses.2 We then conducted a 4 (group size: individuals, dyads, four- and six-person groups) 3 2 (gender) analysis of variance on the persistence data. This ANOVA failed to yield significant effects: there was no main effect of group size, F(3, 42) 5 1.82, p . .10; no main effect of gender; and no significant interaction, Fs , 1. However, the observed power for the group size effect was only .38, and the lack of significant results may have been due to the small number of observations. For further analysis we collapsed the results over gender and conducted a planned contrast between individuals and groups of all sizes. This contrast was significant, t(46) 5 2.06, p , .05, showing that groups were more persistent than individuals (see Table 1). Further, the one-way ANOVA performed on the persistence of individuals, dyads, and four- and six-person groups showed a 2 When this group, with a persistence score of 22 min, is included in the analyses, the linear trend of persistence over group size is less strong, F(1, 47) 5 3.96, p 5 .05, and some effects on the questionnaire items are stronger.


donn

173

PERSISTENCE

significant positive linear relation between group size and persistence, F(1, 46) 5 5.18, p , .05. Productivity Data To compare the productivity of individuals to that of groups of different sizes, we created nominal dyads and four- and six-person groups. Since an analysis of the productivity of nominal groups revealed no gender effects (Fs , 1), the results were collapsed over gender and subtracted from the productivity of the corresponding real groups.3 The resulting productivity losses were analyzed using a 3 (group size: dyads, four- and six-person groups) 3 2 (gender) ANOVA (see Table 1). This analysis revealed a marginally significant effect of group size, F(2, 18) 5 2.80, p , .10. Post hoc tests (LSD; p , .05) indicated that this was due to a marginally lower productivity loss of the dyads as compared to the six-person groups, while four-person groups fell in between and did not differ significantly from the other conditions. The main effect of gender and the interaction were not significant (Fs , 1). Finally, the constant was significant, F(1, 18) 5 7.42, p 5 .01, indicating that real groups overall still suffered a productivity loss as compared to nominal groups. The analysis of the ideas produced in the second session indicated that participants generated on average seven new ideas, which is about 1.4 ideas per minute (Table 2). To check whether the data of subjects who worked in groups were statistically dependent we computed the intraclass correlation, which was .13 ( p , .25). There was also no indication of the production of new ideas being influenced by group size. The 4 (group size) 3 2 (gender) ANOVA, performed on the average group scores, yielded no effects of group size, gender, or an interaction (all Fs , 1). Postexperimental Questionnaire Participants had to indicate why they had ended the session at a certain point in time. Most participants said that they had run out of ideas (since more than one reason could be given percentages exceed 100%). The reasons that were mentioned most often were ‘‘running out of ideas’’ (76%), ‘‘repeating myself’’ (36%), ‘‘dealt with everything’’ (19%), ‘‘long silences’’ (15%), and ‘‘bored with the topic’’ (12%). Two 4 (group size) 3 2 (gender) analyses of variance conducted on task enjoyment and satisfaction revealed main effects of group size on task enjoyment, F(3, 42) 5 2.30, p , .10, and satisfaction (F(3, 42) 5 3.72, p , .05. There were 3 The 4 (group size: individuals, dyads, 4 and 6-person groups) 3 2 (gender) ANOVA performed on the absolute number of ideas generated during the first task resulted in a main effect of group size, F(3, 42) 5 16.87, p , .001, showing that individuals (M 5 55.85) were less productive than groups of all sizes, and dyads (M 5 107.89) were less productive than four-person groups (M 5 171.44) and six-person groups (M 5 212.50; LSD post hoc test; p , .05). Neither the effect of gender nor the interaction was significant (both Fs , 1). The group size effect, however, is trivial, because groups of different sizes are directly compared to individuals.


donn

174


TABLE 2 EXPERIMENT 1: PRODUCTIVITY ON SECOND TASK AND RESULTS OF POSTEXPERIMENTAL QUESTIONNAIRE, COLLAPSED OVER GENDER

Productivity second task (No. ideas) Enjoyed first task Satisfied with production Topic interesting

Individuals (n 5 26)

Dyads (n 5 9)

Four-person groups (n 5 9)

Six-person groups (n 5 7)

6.27 (3.61) 6.50a (1.45) 5.15a (2.33) 6.58 (1.81)

6.00 (1.56) 7.33b (0.75) 7.17b (0.75) 6.44 (1.26)

8.03 (2.57) 7.36b (0.67) 6.47b (1.19) 6.50 (0.94)

6.45 (1.73) 7.28b (0.41) 6.50b (0.71) 6.81 (0.79)

Note. Different superscripts indicate a significant difference ( p , .05) on the post hoc tests (LSD), except task enjoyment ( p , .10). Scores of group members are averaged within the groups. Standard deviations are in parentheses.

no significant main effects of gender, nor significant interactions. Post hoc tests (LSD) revealed that the groups of different sizes did not differ, whereas individuals differed from groups of all sizes (see Table 2). This indicates that the group size main effect can be attributed to the fact that compared to individuals, groups of all sizes enjoyed the session more and were more satisfied with their performance. The same ANOVA conducted on topic interest revealed no significant main effects of group size or gender on this variable and no significant interaction (all Fs , 1). With a mean score of 6.5 on a 9-point scale, subjects appeared to have found the topic reasonably interesting. Discussion Experiment 1 demonstrates that individuals who brainstorm in groups continue longer than individuals who brainstorm in individual settings and that the persistence of group members increases monotonically with group size. This finding contradicts the satisfaction hypothesis which predicted that, due to the baseline fallacy, group members would be prematurely satisfied with their performance and terminate the session earlier than individuals. The finding that groups were more persistent than individuals is consistent with the three other stop rules, which we will now discuss in turn. The depletion hypothesis attributes the greater persistence of groups to the fact that group members need more time than individuals to exhaust their pool of ideas. Inconsistent with this interpretation, real groups overall showed a productivity loss, while this hypothesis predicted that real and nominal groups would not differ. In addition we found that people were still quite capable of producing ideas after they had ended the session, generating an average of 1.4 ideas per minute.


donn

175

PERSISTENCE

When we compare this to the rate of production for individuals in the first session (approximately two ideas per minute), this is quite substantial, taking into account that productivity declines in the course of a session. These findings are inconsistent with the assumption that people can only generate a limited number of ideas and render the depletion hypothesis implausible. The expectancy and enjoyment hypotheses differ mainly in their predictions about the association between group size and persistence. The expectancy hypothesis predicted that persistence would increase monotonically with group size, because the rate at which the perceived difficulty of idea generation rises in the course of a brainstorming session should be lower in larger groups. Predictions from the enjoyment stop rule are less clear. Either a negative or positive association between group size and persistence would be compatible with this hypothesis. Thus, the positive association between persistence and group size observed in Experiment 1 would be consistent with both stop rules. However, there are some additional data which are more supportive of the expectancy than the enjoyment stop rule. Thus, according to the enjoyment stop rule a positive association between group size and persistence would presuppose a positive relationship between group size and enjoyment. As we reported earlier, enjoyment was unrelated to group size. Furthermore, the finding that the majority of participants reported that they had stopped brainstorming because they were running out of ideas is also more consistent with the expectancy than the enjoyment rule. Thus, overall the data are somewhat more supportive of the expectancy than the enjoyment stop rule. The second major finding of Experiment 1 is that the greater persistence of group over individual brainstormers resulted in a reduction of the productivity loss of interactive groups. The interactive dyads were exactly as productive as the nominal dyads, and the productivity loss of the six-person groups differed only marginally from the loss of dyads. Thus, in contrast to the usual findings of brainstorming studies, where four-person groups who work for a predetermined period of time are only half as productive as four individuals (e.g., Diehl & Stroebe, 1987), our four-person groups showed a comparatively small productivity loss of 18% when compared to nominal groups. This difference did not even reach acceptable levels of significance. However, since the power of our analysis was rather low (.42 for the group size effect), the loss of four-person groups may have been significant with more observations. Although there is clear evidence for a persistence–productivity trade-off in our data, there is no indication of an overcompensation effect. Thus, our groups were not persistent enough to overcompensate the blocking effects and outperform individual brainstormers. This pattern may be representative for brainstorming groups in everyday life. It is also possible, however, that the experimental procedure designed to reduce time constraints may not have been completely successful. Despite our emphasis that subjects could spend as much of the 2-h period on the first task as they wished, the knowledge that they had to perform two tasks may have led to a tendency to use half the available time for the first task and


donn

176


the other half for the second. There is some evidence that groups performing a task show transitions in their approach to a task at the midpoint of their available time. Gersick (1988, 1989) found that participants were likely to switch from a first phase of task performance to a second phase when exactly half their time was up. However, participants spent on average from 30 (individuals) to about 45 min (six-person groups) on the first task, which is less than the 60 min that would have constituted the midpoint of the 2-h sessions. This does not necessarily conflict with the findings of Gersick, since our participants had been asked to hand in their watches and therefore could only guess how long they had been working on the first task. Moreover, participants in fact had less than 2 h available for the tasks, since the instruction was read to them within this 2-h period. Therefore, it is possible that our procedure has reduced the differences between conditions, and a different procedure may produce different results. Second, it is possible that the difference in persistence between individuals and groups was small because task enjoyment does play a role in the decision to terminate the brainstorming session. The topic of the brainstorming session was perceived to be quite stimulating. This may have had more impact on the persistence of individuals than on that of groups, because individuals are more dependent on the stimulation derived from the task than group members. Group members can be stimulated by social interaction as well and are therefore to a lesser degree affected by the characteristics of the topic. If this line of reasoning were valid, one would expect the difference between groups and individuals to be larger when the topic is less stimulating, because the persistence of individuals would be lower, while the persistence of groups would hardly be affected. Therefore, in Experiment 2 task enjoyment was manipulated through the topic of the brainstorming session. Further, to assess the validity of the experimental procedure employed in Experiment 1 to reduce time constraints, this procedure was manipulated as well in Experiment 2. EXPERIMENT 2 One way to manipulate task enjoyment directly is to manipulate the topic of the brainstorming session, because topics vary in the degree to which they are perceived as stimulating. Therefore, if task enjoyment were an important determinant of persistence, then choice of topic would have a major impact on persistence. Manipulating topic has the additional advantage of testing the line of reasoning put forward above, that the interest value of the brainstorming topic may affect the persistence of individuals more than that of groups. Therefore, in Experiment 2 we chose to vary the topic of the brainstorming session. We had argued that instructing participants that they had to perform two tasks and that everybody would eventually spend 2 h in the laboratory may have led to a tendency to divide the available time between the two tasks. One way to address this problem is to inform participants that they have to perform several tasks in the available time, without specifying the number of tasks. This should make it


donn

177

PERSISTENCE

impossible to divide time among the different tasks. Therefore, in Experiment 2 half the participants were told that the experiment consisted of several task, without specifying the number of tasks, and the other half were told that they had to perform two tasks. As a third factor, group size (individuals, dyads, and four-person groups) was manipulated again in this experiment. The experiment thus employed a 3 (group size) 3 2 (topic) 3 2 (tasks) design. Method Participants Participants were 112 students from Utrecht University, who were paid DFL 20 (approximately $12) for their participation. Since Experiment 1 had revealed no important gender differences, only female students were used.

Independent Variables Group size. Participants brainstormed either alone (n 5 32), as members of a dyad (n 5 16 dyads), or as members of a four-person group (n 5 12 groups). Topic. Participants generated ideas either on how the number of tourists coming to Utrecht may be increased (enjoyable topic), which is the same topic used in Experiment 1, or on how the education at their university may be improved (less enjoyable topic). Earlier research (Nijstad, 1995) indicated that the education problem is perceived to be less interesting than the tourist problem. Tasks. Participants were either told that the experiment would consist of two tasks or that the experiment would consist of several tasks.

Procedure Each session up to eight female participants were enrolled. Participants were randomly assigned to conditions. However, if one or more did not show up, participants were sometimes run in a different condition. Procedures were similar to those of Experiment 1, with three exceptions. First, half the participants were told that the experiment consisted of two different tasks, while the other half were told that the experiment consisted of several different tasks. Second, two different brainstorming topics were used. Third, in contrast to Experiment 1, there was no second brainstorming session. Participants were paid, debriefed, and dismissed immediately after filling out the postexperimental questionnaire.

Dependent Variables The two major dependent variables were persistence (how much time was spent at the task) and the number of nonredundant ideas. The postexperimental questionnaire was identical to the one used in Experiment 1, but one item was added to check the experimental manipulation. Participants rated on a 9-point scale whether they had ended the task earlier or later because they knew that there would be either another or several other tasks (much earlier–much later). The data analytic strategy was similar to the one employed in Experiment 1.

Scoring of Productivity Data Tapes were transcribed by two research assistants, who marked ideas in the text of the transcription. For a random subset of 10 tapes this was repeated by a second rater, resulting in a correlation between counts of .97. The agreement on separate ideas was 89%. The ideas of participants who had worked individually were pooled and duplicates were removed to create nominal dyads and four-person groups. This task was repeated by an independent rater for a random subset of 8 tapes, resulting in an agreement on duplicates of 99.80%.


donn

178


Results Manipulation Check A 3 (group size) 3 2 (tasks) 3 2 (topic) ANOVA was conducted on ratings of task enjoyment and on the ratings of the topic. For group members, the average group scores were entered in these analyses to take possible statistical dependency into account. The ANOVA on task enjoyment revealed the expected effect: the tourism problem led to greater task enjoyment (M 5 7.01) than the education problem (M 5 5.87), F(1, 48) 5 10.36, p , .01. In addition there was a group size main effect, F(2, 48) 5 3.83, p , .05, showing that members of four-person groups (M 5 7.19) enjoyed the session significantly more than individuals (M 5 6.00), while members of dyads were in between (M 5 6.75) and did not differ from the other conditions (LSD post hoc test; p , .05). For the task factor (e.g., two or several tasks) the main effect was not significant and there were no significant interactions.4 Second, with regard to topic interest, the intraclass correlation was .24, marginally significant ( p , .10). There was a main effect of topic, F(1, 48) 5 4.17, p , .05, with the tourism problem (M 5 6.20) being judged to be more interesting than the education problem (M 5 5.15). There were no significant effects of group size or of tasks or any significant interactions. In general, these results confirm that the manipulation of task enjoyment through topic has been successful. Persistence Data Table 3 lists the results for the persistence and productivity of Experiment 2. The 3 (group size) 3 2 (tasks) 3 2 (topic) ANOVA conducted on the persistence data yielded a main effect of group size, F(2, 48) 5 8.96, p , .001, with persistence increasing with group size. However, an LSD post hoc test ( p , .05) indicated that this effect was due to individuals being less persistent than groups. Thus, individuals were significantly less persistent (M 5 15.91) than both dyads (M 5 26.05) and four-person groups (M 5 26.13), while dyads and four-person groups did not differ from each other. There was also a significant main effect of tasks, F(1, 48) 5 7.55, p , .01. However, contrary to our expectations, participants were more persistent in the condition with two tasks (M 5 23.96) than in the condition with several tasks (M 5 17.37). There was no main effect of topic and no significant interactions with topic (Fs , 1), which is inconsistent with the assumption that persistence is affected by enjoyment. The main effects of group size and tasks are qualified by a marginally significant interaction between group size and tasks, F(2,48) 5 2.60, 4 The interaction between group size and topic was not significant, F(2, 48) 5 1.13, p . .30, which is inconsistent with the hypothesis that task enjoyment of individuals is more affected by topic than that of group members. However, the tendency is in the right direction, with the difference between topics largest for individuals (d 5 1.62) and smaller for dyads (d 5 0.75) and four-person groups (d 5 0.37).


donn

179

PERSISTENCE TABLE 3 EXPERIMENT 2: PERSISTENCE IN MINUTES AND PRODUCTIVITY LOSSES IN NUMBER OF IDEAS Persistence

Two tasks Tourism

Education

Several tasks Tourism

Education

Productivity losses

Ind.

Dyads

Four-person groups

Ind.

Dyads

Four-person groups

17.54 (5.41) n58 16.88 (9.47) n58

24.70 (8.48) n54 34.23 (6.77) n54

36.93 (12.35) n53 32.26 (12.08) n53

—

114.62 (52.81) n54 14.87 (34.89) n54

215.08 (23.97) n53 233.42 (31.37) n53

14.01 (12.46) n58 15.22 (6.79) n58

24.08 (12.98) n54 21.20 (10.34) n54

15.60 (7.08) n53 12.75 (9.22) n53

—

27.13 (47.37) n54 222.38 (13.67) n54

296.08 (36.67) n53 243.08 (7.77) n53

—

—

Note. Standard deviations are in parentheses.

p , .10. Subsequent tests showed that the group size main effect occurred only in the conditions with two tasks, F(2, 27) 5 11.10, p , .001. Post hoc tests (LSD; p , .05) performed on these conditions showed that the group size main effect was due to a significant difference between individuals (M 5 17.21) and both dyads (M 5 29.46) and four-person groups (M 5 34.59), whereas dyads and four-person groups did not differ. Furthermore, in line with the results of Experiment 1, the linear trend of group size on persistence is strongly significant in this condition, F(1, 27) 5 21.34, p , .001. In the condition with several tasks, on the other hand, the group size effect is not significant, F(2, 27) 5 1.75, p . .20. These findings contradict our hypothesis the differences between individuals and groups will be enhanced in the condition with more tasks. Productivity Data To compare the productivity of individuals to that of groups of different sizes, we created again nominal dyads and four-person groups for each of the two topics.5 As in Experiment 1, the means of these nominal groups were subtracted 5 A 2 (group size: nominal dyads and four-person groups) 3 2 (topic) 3 2 (tasks) ANOVA performed on the productivity of nominal groups indicated that nominal dyads (M 5 65.75) were less productive than nominal four-person groups (M 5 128.75), F(1, 16) 5 30.39, p , .001, and participants brainstorming on the tourism problem (M 5 107.33) were more productive than those brainstorming on the education problem (M 5 66.17), F(1, 16) 5 14.60, p , .01. Because no further effects were obtained, results were collapsed over tasks, and the means of the remaining four types of nominal groups (dyads/tourism: M 5 81.63; dyads/education: M 5 49.88; four-person groups/


donn

180


from the corresponding productivity scores of the real groups to compute productivity losses (Table 3).6 Productivity losses were analyzed using a 2 (group size: dyads and four-person groups) 3 2 (topic) 3 2 (tasks) ANOVA, which yielded main effects of group size, F(1, 20) 5 10.72, p , .001, showing that four-person groups suffered a larger productivity loss (M 5 246.92) than dyads (M 5 22.50), and of tasks, F(1, 20) 5 6.20, p , .05, showing that the productivity loss in the condition with two tasks (M 5 24.82) was smaller than the loss in the condition with several tasks (M 5 238.25). Neither the main effect of topic (F , 1) nor the interactions were significant (all Fs , 2, p . .15). Finally, the constant was significant, F(1, 27) 5 6.65, p , .05, showing that the overall productivity loss was significant. Postexperimental Questionnaire As in Experiment 1, participants were asked to indicate why they had ended the session at a certain point in time. Answers to this question were quite similar to those of Experiment 1: running out of ideas (73%), repeating myself (38%), dealt with everything (30%), silences (14%), and bored with the topic (18%). The other questionnaire items were analyzed using 3 (group size) 3 2 (tasks) 3 2 (topic) ANOVAs. We first analyzed the answers to the question whether the fact that there would be more tasks had any effects on the persistence of the participants. The ANOVA only resulted in a main effect of group size, F(2, 48) 5 7.44, p , .01. Post hoc tests indicated that the difference was due to the individuals, for whom this had been a reason to end the task earlier (M 5 4.66) than dyads (M 5 5.59) or four-person groups (M 5 5.35; LSD; p , .05). No other effects were significant. Second, participants were asked how satisfied they tourism: M 5 158.75; four-person groups/education: M 5 98.75) were subtracted from the corresponding productivity scores of the real groups to compute productivity losses. 6 The 3 (group size) 3 2 (topic) 3 2 (tasks) ANOVA performed on the absolute number of ideas yielded a main effect of group size, F(2, 48) 5 14.57, p , .001, which is due to a significant lower production of individuals (M 5 33.25) as compared to both dyads (M 5 63.25) and four-person groups (M 5 81.83), whereas dyads and four-person groups did not differ (LSD; p , .05). Second, the effect of topic was significant, F(1, 48) 5 14.90, p , .001, showing that participants generated more ideas on the tourism problem (M 5 65.23) than on the education problem (M 5 36.70). The effect of tasks was significant as well, F(1, 48) 5 5.41, p 5 .02, showing that participants in the condition with two tasks (M 5 59.57) generated more ideas than participants in the condition with several tasks (M 5 42.37). In addition, the interaction between group size and tasks was marginally significant, F(2, 48) 5 2.56, p 5 .09, which reflects the persistence findings. Thus, in the condition with two tasks a main effect of group size was obtained, F(2, 24) 5 16.78, p , .001, showing that individuals (M 5 34.75) generated significantly fewer ideas than dyads (M 5 75.50) and four-person groups (M 5 104.50), whereas dyads and four-person groups did not differ (LSD; p , .05). In this condition the main effect of topic was significant as well (tourism M 5 75.60, education M 5 43.53), F(1, 24) 5 10.78, p , .01, as was the interaction between group size and topic, F(2, 24) 5 3.73, p , .05, due to an increasing difference between the topics with group size. In the condition with several tasks only a main effect of topic was obtained, F(1, 24) 5 5.07, p , .05, showing that the tourism problem (M 5 54.87) led to more ideas than the education problem (M 5 29.87). Neither the main effect of group size, F(2, 24) 5 2.21, p 5 .13, nor the interactions (F , 1) were significant in this condition.


donn

181

PERSISTENCE

were with their performance. The ANOVA showed that dyads (M 5 6.13) were more satisfied with their performance than individuals (M 5 4.22), while the four-person groups were in between (M 5 5.35), F(2, 48) 5 5.82, p , .01. Post hoc tests (LSD; p , .05) showed that only the difference between individuals and dyads was significant. Second, participants who had worked on the tourist problem (M 5 5.37) were marginally more satisfied than participants who had worked on the education problem (M 5 4.54), F(1, 48) 5 2.89, p , .10. For this item there was no main effects of tasks, and no significant interactions (Fs , 1). Discussion Experiment 2 was designed as a further test of the expectancy hypothesis against a revised version of the enjoyment hypothesis. In general, the revised version of the enjoyment hypothesis could not be supported. According to the hypothesis that people stop brainstorming when they no longer enjoy it, persistence should be higher for more enjoyable topics and this effect should be stronger for individuals because they cannot derive stimulation from social interaction with fellow group members. However, task enjoyment did not affect persistence, since neither the main effect of topic on persistence nor the expected interaction between group size and topic was found. Second, the findings of Experiment 1 were replicated when a similar procedure was used. When participants were told that they had to perform two tasks in 2 h, the linear trend of persistence across group size was replicated, which is consistent with the expectancy hypothesis. Also consistent with this hypothesis, most subjects in this experiment indicated that they had stopped because they were running out of ideas. The results in the condition where participants were told that they had to perform several tasks in 2 h were quite different: no effects of group size on task persistence were found. Apparently, this instruction reduced the differences in persistence between conditions and has induced groups to be less persistent, which is opposite to our prediction that the differences in this condition would be enhanced. An explanation for these findings may be that participants wanted to participate in all the tasks of the experiment and therefore ended the session at a point in time where this would still be possible. According to the questionnaire data, however, the fact that there would be several tasks did not affect their persistence. In addition, most participants indicated that they had stopped because they were running out of ideas, and not because they wanted to participate in the other tasks. We can only assume that participants were unwilling to report that they had stopped because they wanted to start the next task. Somehow, this may have seemed a less valid reason to participants than to stop because one was running out of ideas. Due to minimal blocking effects, productivity losses have rarely been observed for two-person groups even in experiments with set time limits (Diehl & Stroebe, 1987). In contrast, with variation in persistence constrained by set time limits,


donn

182


large losses practically always occur in four-person groups. However, in the condition with two tasks without predetermined time constraints the overall loss was quite small. The loss of four-person groups is limited to only 9.5% (15.08/ 143.67 1 15.08) for the tourism problem and 33.8% (33.42/65.33 1 33.42) for the education problem, while a loss of about 50% is usually obtained (Mullen et al., 1991). Thus, their greater persistence could partly compensate the productivity loss of these groups, in particular for the tourism problem. In the condition with several tasks, where groups showed lower levels of persistence, the productivity loss of four-person groups was much larger (for the tourism problem, 96.08/62.67 1 96.08 5 60.5%; for the education problem, 43.08/55.67 1 43.08 5 43.6%). These findings demonstrate that, at least under conditions where time costs are minimal, groups can and do compensate through greater persistence for the productivity loss caused by blocking. GENERAL DISCUSSION In both experiments, we have found brainstorming groups to be more persistent than brainstorming individuals. Moreover, in Experiment 1 and in Experiment 2 in the condition with two tasks we found that persistence increased with group size. We would argue that this pattern of findings is most consistent with the expectancy stop rule. According to this rule, individuals stop brainstorming when they feel it is no longer worth the effort. They base their estimate of the subjective probability that they will be able to produce further ideas on a given topic, on the ease or difficulty of producing ideas at a given point in time. Because, due to blocking, perceived difficulty should increase more slowly over time in group than in individual sessions, and in larger rather than small groups, application of this stop rule would account for the differences in persistence observed in our experiments. Further support for this interpretation comes from the reasons subjects give themselves for having stopped brainstorming. The findings that the majority of participants in both experiments reported that they had stopped because they ran out of ideas would also be consistent with the expectancy stop rule. In trying to evaluate the generality of these findings, we have to consider the possibility that the choice of stop rules may be influenced by situational factors and that our research paradigm may have favored the expectancy stop rule. One could argue that since our participants expected that they had to perform a second task, they had to choose whether they wanted to stay with the present task or shift to the next. This requires a cost/benefit analysis, and clearly the costs of staying with the task increase as it gets harder to generate ideas. In conditions where there is only one task, other stop rules (satisfaction, enjoyment) may be more important. However, as the problems we encountered in reducing time costs illustrate, such conditions are rare, if not nonexistent. There are always alternatives to which one could devote one’s time, even if they consist of chatting with a colleague or leaving work early. By being more persistent, small groups were able to compensate part of their


donn

183

PERSISTENCE

usual productivity loss. Thus, no productivity loss was found in dyads, and for four-person groups the loss was quite low, whereas it is usually found that four-person groups are only half as productive as four individuals (e.g., Diehl & Stroebe, 1987). However, the group size effect did only occur in the conditions where participants were told that they had to perform two tasks. When participants believed that they had to perform several tasks, no differences between individuals and groups and no linear trend of persistence across group size were found. This may be attributable to a tendency of participants to want (or feel obliged) to participate in all tasks. A cost/benefit analysis in that case may result in higher benefits for switching to the next task, because when this task has an unfavorable cost/benefit ratio, it is always possible to switch to the next. In view of the fact that brainstorming groups often work without explicit time limits in applied settings, our persistence findings may go some way toward reconciling the discrepancy between researchers and practitioners in their evaluation of the effectiveness of group brainstorming. Despite overwhelming evidence to the contrary, practitioners are still convinced that group brainstorming is very effective, and group brainstorming is still widely applied in organizations of all kinds. Although researchers have found this belief to be so unrealistic as to be labeled the ‘‘illusion of group effectivity’’ (Paulus et al., 1993; Stroebe et al., 1992), it may reflect the fact that the greater persistence of groups over individuals may have compensated (or sometimes even overcompensated) for their productivity loss. Although this assumption goes some way toward explaining the illusion of effectivity in applied settings where groups typically brainstorm without set time limits, it cannot account for the fact that this illusion even affects experimental subjects who work under the Equal Man-Hour-Comparison. Even under these conditions, group members are more satisfied with their performance than individuals (e.g., Diehl & Stroebe, 1987). The line of reasoning suggested earlier may help account for the illusion of these subjects. Idea generation in groups may be less effortful, because group members can sit back and listen to others at times that they themselves are unable to come up with new ideas. Individuals, on the other hand, are forced to fill the complete session themselves, and it may happen to them more often that their efforts are unsuccessful. The greater ease of idea generation in groups will influence not only task enjoyment but also satisfaction with brainstorming productivity, since the session will appear to have proceeded very smoothly. Consistent with this interpretation, we found that the easier topic (tourism) led to greater task satisfaction than the more difficult topic (education). Finally, in the light of findings reported in this paper, we would like to expand on the practical advice we derived from our earlier studies. Thus, Stroebe and Diehl (1994) suggested that since people preferred generating ideas in groups rather than individually because brainstorming was more enjoyable in group settings, they should at least try to keep the size of these brainstorming groups small. This advice is further supported by the findings reported in this article that groups of no more than four members, who worked without time pressure, were


donn

184


able to compensate for the productivity loss due to blocking by being more persistent. It is important to note, however, that when time is of the essence, it might be safer to reduce group size to two persons. As enjoyment does not seem to vary with group size, it would already have reached a maximum while blocking effects would be kept at a minimum. REFERENCES Camacho, L. M., & Paulus, P. B. (1995). The role of social anxiousness in group brainstorming. Journal of Personality and Social Psychology, 68, 1071–1080. Diehl, M., & Stroebe, W. (1987). Productivity loss in brainstorming groups: Toward the solution of a riddle. Journal of Personality and Social Psychology, 53, 497–509. Diehl, M., & Stroebe, W. (1991). Productivity loss in idea-generating groups: Tracking down the blocking effect. Journal of Personality and Social Psychology, 61, 392–403. Gersick, C. J. P. (1988). Time and transition in work teams: Toward a new model of group development. Academy of Management Journal, 31, 9–41. Gersick, C. J. P. (1989). Marking time: Predictable transitions in task groups. Academy of Management Journal, 32, 274–309. Hirt, E. R., Melton, R. J., McDonald, H. E., & Harackiewicz, J. M. (1996). Processing goals, task interest, and the mood-performance relationship: A mediational analysis. Journal of Personality and Social Psychology, 71, 245–261. Hyams, N. B., & Graham, W. K. (1984). Effects of goal setting and initiative on individual brainstorming. Journal of Social Psychology, 123, 283–284. Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological Review, 80, 237–251. Kanekar, S., & Rosenbaum, M. E. (1972). Group performance on a multiple-solution task as a function of available time. Psychonomic Science, 27, 331–332. Karau, S. J., & Kelly, J. R. (1992). The effects of time scarcity and time abundance on group performance quality and interaction process. Journal of Experimental Social Psychology, 28, 542–571. Levine, J. M., & Moreland, R. L. (1990). Progress in small group research. Annual Review of Psychology, 41, 585–634. Martin, L. L., Ward, D. W., Achee, J. W., & Wyer, R. S. Jr. (1993). Mood as input: People have to interpret the motivational implications of their moods. Journal of Personality and Social Psychology, 64, 317–326. McGrath, J. E. (1984). Groups: Interaction and performance. Englewood Cliffs, NJ: Prentice Hall. Mullen, B., Johnson, C., & Salas, E. (1991). Productivity loss in brainstorming groups: A metaanalytic integration. Basic and Applied Social Psychology, 12, 3–24. Nijstad, B. A. (1995). Het geheel groter dan de som der delen? Produktiviteitswinst in brainstormgroepen (Productivity gains in brainstorming groups). Unpublished master’s thesis, Utrecht University. Osborn, A. F. (1957). Applied imagination: Principles and procedures of creative problem solving. New York: Scribner. Paulus, P. B., & Dzindolet, M. T. (1993). Social influence processes in group brainstorming. Journal of Personality and Social Psychology, 64, 575–586. Paulus, P. B., Dzindolet, M. T., Poletes, G., & Camacho, L. M. (1993). Perception of performance in group brainstorming: The illusion of productivity. Personality and Social Psychology Bulletin, 19, 78–89. Sanna, L. J., Turley, K. J., & Mark, M. M. (1996). Expected evaluation, goals, and performance: Mood as input. Personality and Social Psychology Bulletin, 22, 323–335. Steiner, I. D. (1972). Group process and productivity. New York: Academic Press.


donn

185

PERSISTENCE

Stroebe, W., & Diehl, M. (1994). Why groups are less effective than their members: On productivity losses in idea-generating groups. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology: Vol. 5 (pp. 271–303). London: Wiley. Stroebe, W., Diehl, M., & Abakoumkin, G. (1992). The illusion of group effectivity. Personality and Social Psychology Bulletin, 18, 643–650. Stroebe, W., & Frey, B. (1982). Self-interest and collective action: The economics and psychology of public goods. British Journal of Social Psychology, 21, 121–137. Vroom, V. H. (1964). Work and motivation. New York: Wiley.


donn