BASIC AND APPLIED SOCIAL PSYCHOLOGY, 1991, 72(1), 3-23 Copyright �� 1991, Lawrence Erlbaum Associates, Inc. Productivity Loss in Brainstorming Groups: A Meta-Analytic Integration Brian Mullen and Craig Johnson Syracuse University Eduardo Salas Naval Training Systems Center This article reports the results of a meta-analytic integration of previous research on productivity loss in brainstorming groups. The following patterns were observed: Generally, brainstorming groups are significantly less produc- tive than nominal groups, in terms of both quantity and quality. Stronger productivity loss was demonstrated in the context of (a) larger groups, (b) experimenter presence, (c) tape-recorded vocalization of contributions (vs. writing of contributions), and (d) in comparison to a nominal group of truly Alone individuals (vs. a nominal group of Together individuals). These patterns are (a) highly consistent with predictions derived from social psychological explanatory mechanisms, and (b) only marginally consistent with procedural explanatory mechanisms, and (c) highly inconsistent with economic explanatory mechanisms. This article considers the implications of these patterns for the use of, and for future research on, brainstorming. So loud each tongue, so empty was each head, So much they talked, so very little said. (Churchill, 1761) Osborn (1957) proposed brainstorming as an effective means of enhancing the quantity and quality of ideas generated in group settings. Typical brainstorming instructions prompt group members to generate as many ideas as possible, to evaluate uncritically their own ideas before expressing them, to evaluate uncritically other people's ideas when they are expressed, and to improve or combine ideas already suggested. Brainstorming carried the promise (cf. Osborn, 1957, p. 229) of leading the average person to generate twice as many ideas in a group as would be generated individually. Requests for reprints should be sent to Brian Mullen, Department of Psychology, Syracuse University, Syracuse, NY 13210.

4 MULLEN, JOHNSON, SALAS Since Osborn's (1957) formulation of brainstorming, a number of studies have examined the hypothesis that brainstorming groups could outperform individuals. Tests of the effectiveness of brainstorming compare the productivity of people interacting in real groups under brainstorming instructions with the productivity of people in "nominal groups" (i.e., people performing individually with no interaction, whose productivity is then combined). Generally, nominal groups tend to outperform brain- storming groups (e.g., Diehl & Stroebe, 1987 Taylor, Berry, & Block, 1958). This article presents a meta-analytic integration (Mullen, 1989a Mullen & Rosenthal, 1985 Rosenthal, 1984) of research on productivity loss in brainstorming groups. This meta-analytic integration was developed to address a number of specific issues about brainstorming. SIGNIFICANCE, MAGNITUDE, AND CONSISTENCY OF EFFECTS There certainly seems to be a consensus on the part of researchers in this area that brainstorming groups produce fewer ideas than nominal groups (e.g., Casey, Gettys, & Pliske, 1984 Diehl & Stroebe, 1987). We do not have a clear gauge, however, of the extent of productivity loss engendered by the application of the brainstorming technique. Although brainstorming groups may generally produce significantly fewer ideas than nominal groups, the difference could be a small one that accounts for a trivial amount of variance. Moreover, there is the logical possibility (considered, albeit not supported, by Graham, 1977) that decrements in quantitative productivity could be compensated by a corresponding increase in the quality of ideas generated by brainstorming groups. The long-lived, if potentially misguided, popularity of brainstorming techniques highlights the importance of determining a clear and precise summary of the extent of productivity loss in brainstorming groups, for both quantity and quality of productivity. MECHANISMS FOR PRODUCTIVITY LOSS Although systematic theoretical development generally has not character- ized the research literature on brainstorming, we can identify three inde- pendent plausible classes of explanatory mechanisms for productivity loss in brainstorming groups. "Procedural mechanisms" derive from the mun- dane concerns of parsing up a given amount of task performance time amongst a certain number of performers (e.g., production blocking Lamm & Trommsdorf, 1973). "Social psychological mechanisms" are basic under-

A META-ANALYTIC INTEGRATION 5 lying processes engaged by the presence of the other people, and by the individual's membership in the group for example, drive-arousal (Geen & Bushman, 1987 Zajonc, 1965) and self-attention (Carver & Scheier, 1981 Mullen, 1983, 1987 Mullen & Baumeister, 1987). Finally "economic mechanisms" represent a motivated, intentional withdrawal of effort for example, social loafing (Latane, WiUiams, & Harkins, 1979) and free-riding (Kerr & Bruun, 1983). In simpler terms, productivity loss in brainstorming groups has been explained as a result of people interrupting each other (procedural mecha- nisms), the effect of the group on the individual (social psychological mechanisms), or self-interested laziness (economic mechanisms). In a recent examination of productivity loss in brainstorming groups, Diehl & Stroebe (1987) argued that procedural mechanisms were the strongest contributors to productivity losses.^ The relative contributions of these three general mechanisms for productivity loss can be examined using several method- ological facets of the brainstorming research paradigm. These include group size, experimenter presence, response mode, and types of groups. Group Size Each of the three explanatory mechanisms would predict that productivity loss in brainstorming groups should increase as a function of group size. As group size increases, there are more people to interrupt the individual and take up more time talking, representing a productivity loss due to proce- dural mechanisms. As group size increases, there are more people whose presence should arouse the individual, or who can provide a deindividuating background in which the subject becomes perceptually immersed, repre- senting a productivity loss due to social psychological mechanisms. Finally, as group size increases, there are more "suckers" upon whom the individual can "free-ride," or more co-actors among whom the individual can socially loaf, representing a productivity loss due to economic mechanisms. The only study which directly examined the effects of group size on productivity loss in brainstorming groups (Bouchard & Hare, 1970) sug- gested that the difference in quantitative productivity between brain- storming groups and nominal groups increased as a function of group size. However, this study only reported a significant interaction effect, without statistically testing the trend across group sizes. Moreover, this pattern 'Diehl and Stroebe (1987) presented a similar categorization of productivity loss mecha- nisms. For example, "procedural mechanisms" includes Diehl and Stroebe's "production blocking" (as well as cognitive interference). "Social psychological mechanisms" includes Diehl and Stroebe's "evaluation apprehension" (as well as self-attention, and the drive arousal underlying evaluation apprehension). "Economic mechanisms" includes Diehl and Stroebe's "free riding" (as well as social loafing).