مسائل متوسط : بررسی شایستگی طولانی تعامل گروه در تولید ایده های خلاق و ناب در یک روش متاآنالیز ادبیات طوفان مغزی گروه الکترونیکی

This meta-analysis examines the influence of electronic communication media on group idea generation tasks. Data from the following three areas of the brainstorming literature are synthesized to assess differences across performance variables and group member satisfaction: (1) electronic brainstorming (EBS) groups versus traditional face-to-face (FTF) interacting groups, (2) EBS groups versus nominal groups, and (3) EBS versus electronic nominal (e-nominal) groups. The results of this integration show that EBS groups are more productive and more satisfied with the interaction process than FTF groups. Additionally, large EBS groups outperformed nominal groups, whereas small nominal groups outperformed EBS groups. These findings have important implications for electronic collaboration and teamwork in both academic and organizational settings, especially given the recent proliferation of virtual teamwork.

Producing creative solutions to problems is an important outcome variable associated with group interactions in the workplace (Kerr and Tindale, 2004 and Vissers and Dankbaar, 2002). Using groups to facilitate creative idea generation has intuitive appeal, because it is widely believed that groups are superior to unaided individuals in such tasks (Faure, 2004, Gallupe et al., 1994 and Valacich et al., 1994). Popularizing this notion, Osborn (1957) formulated specific procedures for interacting groups working on creative idea generation tasks, a process dubbed “group brainstorming.” Group brainstorming is widely used in a variety of settings and derivations of this technique are commonplace in organizations. In brainstorming, participants are encouraged to voice or record as many ideas as possible, regardless of the practicality or feasibility of the ideas. Participants are also encouraged to build upon other members’ ideas (Connolly et al., 1993 and Paulus et al., 1995). According to Osborn, brainstorming groups could be expected to outperform individuals in terms of both the quantity and quality of ideas. However, numerous empirical studies have disconfirmed these assertions (see Kerr & Tindale, 2004 for a review). In traditional laboratory and field studies of group brainstorming, face-to-face (hereafter, FTF) groups of various sizes are compared to nominal groups, which are formed by aggregating the output of a comparable number of individuals working separately. The outcome in nearly all such studies is that nominal groups outperform FTF groups in terms of the production of non-redundant ideas and idea quality (Diehl & Stroebe, 1987). Mullen, Johnson, and Salas (1991) conducted a meta-analysis comparing FTF groups and nominal groups from 1958 to 1990, and found the mean effect sizes from the experimental literature to be large (r = .57 for the number of non-redundant ideas; r = .56 for idea quality). These results contradict the popular, but poorly substantiated notion that communication among individuals will result in “synergistic” effects ( Faure, 2004, Pinsonneault et al., 1999a and Tindale and Larson, 1992). In light of these consistent findings, considerable attention has been paid to electronic brainstorming (EBS), an e-collaboration method that employs networked computer terminals and software designed to allow group members to communicate electronically during idea generation exercises (Connolly et al., 1993 and Dennis and Valacich, 1994). The technology and procedure of electronic group brainstorming are thoroughly described elsewhere (Dennis and Valacich, 1993, Dennis and Valacich, 1999, Dennis and Williams, 2003 and Pinsonneault et al., 1999a), but we shall briefly summarize the process. Participants are presented with a textual description of the brainstorming task on their computer monitors, and when they generate an idea, they simply type it in a designated portion of the screen. Once an idea is entered it becomes available for viewing by other group members in a shared space that appears in a section on the monitor in each workstation. This shared space is also referred to as group memory (Dennis & Valacich, 1993). In some instances, the group space shows all of the cumulative ideas for the session, while a random subset of ideas is displayed in others (Pinsonneault et al., 1999a). Input from each participant is synchronously presented to the group (i.e., in real time or without temporal discontinuity). The most common type of EBS software is a component of GroupSystemsV (Nunamaker, Applegate, & Konsynski, 1987), which is a popular group-support software package. Theoretically, the benefits of electronic brainstorming are related to a cluster of procedural and social psychological factors that are said to underlie weaknesses of traditional brainstorming (Diehl and Stroebe, 1987, Mullen et al., 1991 and Munkes and Diehl, 2003). The first is production blocking, which is based upon the individual’s inability to spontaneously interject ideas without violating group etiquette or breaking the concentration of other members. According to Diehl and Stroebe, 1987 and Diehl and Stroebe, 1991, production blocking has been the main source of observed productivity losses in FTF groups. Electronic groups should have a considerable advantage over their verbal group counterparts, as the sharing of ideas by one person does not prevent simultaneous sharing by other group members ( Dennis and Valacich, 1993 and Gallupe et al., 1991); an explanation supported empirically by Nijstad, Stroebe, and Lodewijkx (2003) who showed that delays between production and articulation of ideas are the major culprit in production blocking. Although some production blocking is likely to occur with EBS (Pinsonneault et al., 1999a), it should be minimal relative to that of FTF groups (Connolly et al., 1993). Next, the synergistic effects of brainstorming should be augmented by the electronic medium (Dennis & Valacich, 1993). Ideas are presented and stored on each participant’s monitor for later reference, hence the likelihood of building upon previous ideas is improved because access to the data is unrestricted by individual recall ( Gallupe et al., 1991 and Gallupe et al., 1994). Moreover, in EBS, the number of ideas that members can build upon should increase as a function of group size. Computer-mediated brainstorming allows for the possibility of maintaining anonymity, which tends to reduce evaluation apprehension and might promote the generation and sharing of more ideas ( Connolly et al., 1990 and Cooper et al., 1998). Finally, computer-mediated brainstorming may occasion additional adaptive responses from the users that then result in increased performance ( DeRosa et al., 2004, Kock, 2000 and Kock, 2001). In sum, EBS ought to ameliorate some of the principal procedural, behavioral, and social psychological constraints on group brainstorming ( Gallupe et al., 1991 and Pinsonneault et al., 1999a).1 1.1. Assessing the benefits of EBS in the current meta-analysis Although electronic brainstorming is intended to improve productivity in interacting groups, its practical benefits must be viewed relative to the types of control conditions to which it is compared. EBS groups are compared to several alternative types of brainstorming groups: (1) traditional FTF groups, (2) traditional nominal groups, or (3) electronic nominal groups (hereafter, e-nominal). The objective of this meta-analysis is to clarify the status of EBS in relation to each of these alternatives, building upon the work of Dennis and Williams (2005) by including a more granular analysis of potential moderators. The first comparison condition is the FTF group, where group members interact face-to-face and, typically, record ideas in a traditional paper-and-pencil manner. The results of experiments comparing EBS to traditional FTF brainstorming teams (e.g., Gallupe et al., 1991, Gallupe et al., 1992 and Valacich et al., 1993) generally illustrate that EBS groups produce greater numbers of ideas. However, the superiority of either group type is unclear for other dependent measures, such as idea quality and group member satisfaction. The second cluster of studies compares EBS to nominal groups. This comparison might be viewed as more important, given that nominal groups are the “gold standard” control condition of traditional brainstorming studies. It would be insufficient for EBS to simply improve upon FTF brainstorming if it were not also demonstrably superior to nominal group brainstorming. Unfortunately, interpretation of results from studies employing these conditions is impossible because the comparison is confounded. Specifically, researchers cannot precisely attribute observed differences in idea quality or number of ideas generated to either the technology or to the use of the brainstorming technique in view of the fact that both factors influence the data. Although such comparisons offer some practical advice, they do not provide information about why the outcomes emerge as they do. Fortunately, researchers have also begun to examine the comparison of EBS groups to an e-nominal condition, representing the third cluster of studies. The e-nominal condition is similar to the traditional nominal group condition, except that members of e-nominal groups generate ideas via computer (with no interpersonal interaction or collaboration). Again, as in nominal conditions, the ideas of e-nominal group members are subsequently pooled to form makeshift groups (Dugosh et al., 2000 and Roy et al., 1996). Thus, the only methodological difference between nominal and e-nominal groups is the use of an electronic medium, making such comparisons correct and interpretable. In a qualitative review of the literature comparing EBS and nominal groups, Pinsonneault et al. (1999a) treated the two types of nominal groups as a single condition and reported turbid results. It is evident that the literature on nominal groups has, to a large degree, ignored the potential influence of the electronic medium on individual idea generation; the assumption of equivalence may be inappropriate. In addition to potential stimulation (synergistic) effects (Gallupe et al., 1991), if EBS groups were found to be more productive than nominal groups, another important factor might be efficiency effects, which refer to the fact that EBS participants are less likely to replicate ideas during brainstorming because they have access to the cumulative output of the group (Pinsonneault et al., 1999a). Also, group size influences both the number and quality of ideas produced (Pinsonneault, Barkhi, Gallupe, & Hoppen, 1999b), though the specific effects of this variable are difficult to discern at present. In some cases (e.g., Dennis & Valacich, 1993), nominal groups outperform EBS when groups were small; however, this trend reverses for large groups. Irrespective of the underlying mechanisms of the procedural differences, reconciling the inconsistency in the available data is the principal objective of this meta-analysis. 1.2. Dependent measures Despite the procedural changes that have occurred in brainstorming research with the use of EBS, the criteria by which brainstorming tasks are evaluated have remained constant. The present meta-analysis focuses on the three most widely used dependent measures: quantity and quality of idea generation and group member satisfaction. 1.2.1. Quantity of non-redundant ideas The primary dependent variable in brainstorming research has been the number of non-redundant ideas that an individual or group can produce. This figure is the summation of the unique ideas produced, rather than a group mean, and thus, high-performing individuals are not controlled for. Though simplistic, this criterion is consistent with Osborn’s (1957) approach to evaluating the effectiveness of brainstorming groups. 1.2.2. Quality of non-redundant ideas The second dependent variable is idea quality, which is operationalized in several ways in the literature (Diehl & Stroebe, 1987). The most common approach to measuring quality is to use external raters to assign numerical values to the ideas according to their originality or relevance to the particular brainstorming task. These idea ratings are then averaged within brainstorming groups to produce a group quality score (see Straus & McGrath, 1994). A less common approach is to assign the ideas to ordinal categories (from high to low quality) and then tally the number of high quality ideas per group (see *Ziegler, Diehl, & Zijlstra, 2000). To date there is little agreement in the EBS literature as to how EBS compares to other types of brainstorming groups in terms of idea quality (Dennis and Valacich, 1994, Gallupe et al., 1992, George et al., 1990 and Straus and McGrath, 1994). It is possible that the different methods for measuring idea quality may partially account for the inconsistent results in this research domain; thus, both are included in the present analysis. 1.2.3. Group member satisfaction Most brainstorming studies include measures of group member satisfaction. Again, the empirical results are mixed, with some researchers reporting that EBS groups were more satisfied with the brainstorming process than FTF groups (Aiken et al., 1994, Fellers, 1989, Gallupe et al., 1991 and Gallupe et al., 1992), and others reaching the opposite conclusion (Adrianson and Hjelmquist, 1991, Cooper et al., 1998 and George et al., 1990). Comparing EBS to nominal conditions, EBS group members were more satisfied than their counterparts (Dennis and Valacich, 1993, Gallupe et al., 1991 and Valacich et al., 1994) while, Roy et al. (1996) found that there was no statistically significant difference in member satisfaction with the process when comparing members of EBS groups and e-nominal group members. The ambiguity surrounding satisfaction may be partially attributed to problems of operationalization and measurement. Some researchers assess satisfaction with the process (Gallupe et al., 1991 and George et al., 1990), others have measured satisfaction with the medium (Straus & McGrath, 1994), and these approaches may reflect important distinctions (DeRosa et al., 2004 and Pawlowicz, 2003). For instance, items assessing process satisfaction may refer to satisfaction with the group members, the medium, or task outcomes, whereas satisfaction with the communication medium may assess ease of use with the medium or whether group members liked the medium. In addition to problems of operational consistently, brainstorming research has no conventional instrumentation for measuring satisfaction. Researchers in this domain often rely on few items to measure satisfaction, and rarely provide evidence of the psychometric properties of their instruments. Collectively, these issues limit the interpretability of existing data.

7.1. EBS versus FTF conditions These results illustrate several advantages of using electronic media for idea generation tasks. First, EBS groups produced more non-redundant ideas than FTF groups (ru = .55). Here, the superiority of EBS over FTF groups is nearly identical to that of nominal groups over FTF groups demonstrated in the Mullen et al. (1991) meta-analysis. Although the test of homogeneity is statistically significant, the 95% confidence interval is narrow and does not cross zero, and thus, we consider this outcome to be quite robust. The box-and-whisker plots of the effect sizes for individual dependent variables in each comparison condition in Fig. 1 show a very narrow range of effect sizes for the number of non-redundant ideas in this condition.2 Gallupe et al. (1991) identified several factors that might account for the findings that EBS groups were superior to FTF groups in producing large numbers of ideas. First, they posited that electronic interacting groups would be less likely to suffer from production blocking than would FTF groups. This explanation has intuitive appeal, and has received indirect support from experimental data (Diehl and Stroebe, 1987, Gallupe et al., 1994 and Nijstad et al., 2003), though it has been contested on theoretical grounds (Pinsonneault et al., 1999b). Unfortunately, it is difficult to evaluate in the context of the current analysis because production blocking cannot be directly measured in the included studies, and thus it remains a hypothetical explanation (Diehl & Stroebe, 1987). Next, the increased levels of anonymity present in electronic interaction might facilitate group members’ contributions and increase participation rates, which would ultimately increase group performance on the task. Third, there might be an inherent memory advantage in EBS given that individuals are able to access the contributions of other group members and can explore these ideas throughout the collaboration. Viewing others’ ideas may serve as a catalyst or cue (Connolly et al., 1993 and Dennis et al., 1996), or competitive stimulus (Munkes & Diehl, 2003) for group members. Furthermore, EBS groups might be more productive due to the presence of visual information, which might simply be processed faster than spoken information. The demands of adapting to the electronic medium may evoke extra effort on the part of EBS group members, which may in turn lead to improved performance (DeRosa et al., 2004). Finally, EBS group members might find the technology unique and engaging, which could result in increased productivity (Gallupe et al., 1992). In addition, members of EBS groups may be exposed to dissenting and minority opinions that may be otherwise quashed in FTF groups; such exposure is known to prompt group members to search for information on all sides of an issue, use alternative strategies and notice solutions that otherwise would have gone undetected (Nemeth & Nemeth-Brown, 2003). Our planned moderator analyses of the (1) relative anonymity of participation and idea contribution and (2) the amount of information that was accessible in the group memory of EBS participants (i.e., cumulative ideas generated versus a random sample of ideas) were intended to address some of these points above. Unfortunately, the methodological reporting in the included studies rarely provided sufficient details on such procedural information. However, based on the limited data available, no consistent trends were evident. The only moderator considered for this condition was the type of incentive used. Previous research shows that monetary reward, or even the perceived probable delivery of a reward, is sufficient to cause dramatically increased rates of creative idea generation (Eisenberger & Shanock, 2003). In brainstorming groups, even in instances where not all participants are rewarded, the simple awareness of the possibility of monetary reinforcement produces large effects compared to control groups (Lorenzi, 1988). Thus, even though the results do not indicate a change in the direction of the effects, they do illustrate an augmentation of the superiority of EBS over FTF brainstorming, in that cash-based incentives lead to the production of more non-redundant ideas than in non-cash conditions. Theoretically, this could be attributed to production blocking in face-to-face groups. Based on existing data, we expect contingent cash to motivate participants to produce more ideas irrespective of other parameters of the experimental condition; however, in FTF groups this additional effort may be dampened by common process losses associated with turn-taking and forgetting ideas. As participants produce a higher volume of ideas, the process losses may increase because there is a ceiling on the number of ideas that can be effectively produced in that medium. In contrast, production blocking effects may be minimized in EBS groups (Dennis & Valacich, 1993), and the electronic medium may accommodate the generation and recording of ideas at higher rates (Nijstad et al., 2003). The second major finding is that EBS groups uniformly outperformed FTF groups (ru = −.46) for both aggregate and specified measures of quality. Again, as with the Mullen et al. meta-analysis, the magnitude of the effect size for quality was large by conventional standards (Cohen, 1988). We make two caveats regarding these results. First, idea quality cannot be assumed to mean the same thing to all judges in all studies, and any such subjective measures may result in inconsistencies in the data. Second, in some cases the higher quality of ideas reflects absolute rather than proportional increases in quality. This was addressed by conducting separate syntheses for the number of high quality ideas and the average quality of ideas respectively. Though the data suggest no major differences between the two definitions, the distinction should be noted. The moderator analysis for type of compensation showed that the superiority of EBS over FTF in producing high quality ideas was less distinctive when participants were compensated with cash, and was more distinctive when non-cash incentives were used. Though the two clusters of effect sizes were both statistically heterogeneous, the mean effect sizes were statistically significant. One possible interpretation of these results is based on participants’ reactions to Osborn’s instruction to focus on producing as many ideas as possible. EBS groups generate more ideas, and produce higher quality ideas than FTF groups; however, the quality, or practicability of ideas is neither discussed nor rewarded. If a cash incentive were added, the participants would have no impetus to focus additional attention on quality, and thus, any improvement in performance would be primarily evident in the number of non-redundant ideas. With attention on producing many ideas, the overall feasibility or utility of the ideas may suffer, thus reducing the discrepancy between EBS and FTF groups in cash-compensated studies. Such effects may also reflect the way in which researchers inform participants about their compensation. If rewards were overtly tied to the number of ideas produced, rather than to idea quality, we would not be surprised to see greater discrepancies in quantity of ideas between EBS and FTF groups (given their assumed levels of production blocking), while in such cases idea quality would go unchanged. Given the rapidly changing landscape in compensation and reward systems (Dixon, Hayes, & Stack, 2003) and in virtual teamwork (DeRosa et al., 2004), this difference is relevant to emerging issues in research and practice. Future research should carefully evaluate the potential effects of tying cash rewards to idea production. Contrary to our hypotheses, EBS groups were more satisfied with the brainstorming process than were FTF groups (ru = −.15). This result is surprising considering that a majority of the literature comparing FTF and electronic teams has shown that electronic group members tend to be less satisfied with their interactions than their face-to-face counterparts ( Adrianson and Hjelmquist, 1991, Barkhi et al., 1999, Hollingshead, 1996, Ocker and Yaverbaum, 1999, Olaniran, 1996, Straus, 1996, Straus and McGrath, 1994 and Warkentin et al., 1999); but more recent research (Pawlowicz, 2003) shows no such difference over time. However, the aforementioned studies employed decision-making, intellective, and other types of tasks rather than idea generation tasks. Thus, it is plausible that some aspect of idea generation tasks makes them distinct in this medium. In addition, these data erode what may be the final possible merit of traditional brainstorming methods, which was that, historically, brainstorming group members at least enjoyed the process. Since the confidence interval associated with the obtained mean ru does not include zero, we feel confident that this result is robust. The accompanying test of homogeneity did reveal rather extreme differences among the effect sizes, but as we indicated previously, such heterogeneity may be based largely on how satisfaction is operationalized. As with the analyses of quantity and quality of ideas, cash-based incentives augment satisfaction in brainstorming groups such that cash incentives lead to increased levels of satisfaction in EBS groups relative to FTF groups. Again, interpretation of these results may be contingent upon how group member satisfaction is operationalized. Specifically, researchers must differentiate between satisfaction with the process (i.e., satisfaction with the team interaction), and satisfaction with the outcome (i.e., satisfaction with the final product, or decision) (Olaniran, 1996). We propose some potential explanations for the observed satisfaction outcomes accounting for the problems of construct operationalization. EBS group members may be more satisfied with the interaction process than FTF group members due to the novelty of the technology. However, the novelty of the technology might decline with increased exposure to computers, which may ultimately lead to decreases in members’ reported satisfaction longitudinally. To address this possibility with the existing data, we calculated the correlation between the year of publication and effect sizes for member satisfaction in FTF and EBS groups, and found that, over time, satisfaction with electronic interaction decreased (r = −.74). This correlation offers preliminary evidence that the novelty of electronic interaction diminishes as individuals become more accustomed to working with computers. Another possibility would have been to consider the moderating influence of participant computer experience. Unfortunately, such information was rarely reported and, in the existing literature, is probably randomly distributed within groups. Another possible explanation of the current findings for satisfaction with outcomes may be attributed to the fact that EBS provides an accessible record of task outcomes for group members, which might lead to feelings of increased productivity, ease of use, and satisfaction. This seems to be especially relevant to idea generation tasks where group members may be more cognizant of being evaluated (e.g., Michinov & Primois, 2005). Moreover, brainstorming tasks involve additive processes where group members can track their discrete inputs and outputs. In contrast, this type of performance monitoring is not often present in intellective and judgment tasks. In this sense, while electronic group members working on idea generation tasks might be more satisfied with the process or the outcome, perhaps group members working on other types of tasks report feeling less satisfied because of the absence of performance feedback, or the lack of emphasis on discrete outcomes. 7.2. EBS versus nominal and e-nominal conditions We included a caveat regarding the emphasis of EBS versus traditional nominal comparisons in the literature due to a confound in the data. Despite this concern, such comparisons are discussed here for several reasons. First, researchers may be interested in the findings despite causal ambiguity. Second, researchers may argue that this confound is inconsequential in light of the empirical evidence indicating that nominal groups always outperform FTF groups. In Table 2 and Table 3, the FTF condition is compared to each type of nominal condition, as well as to a pooled condition termed “combined nominal.” The results of this comparison are not as definitive as those in the FTF versus EBS comparison; however, they do show that EBS is a better method for creative idea generation than nominal group brainstorming under specific conditions. In the primary meta-analysis, for the number of non-redundant ideas, both nominal and e-nominal outperformed EBS groups (ru = −.10). The e-nominal groups performed better relative to FTF (ru = −.11) than did the nominal groups (ru = −.03). However, these differences were comparatively small. The heterogeneity of the effects and the near zero mean effect sizes substantiated the need for moderator analyses. The group size analysis produced fairly profound differences in the mean effect sizes among small groups as compared to large groups. For small group sizes, nominal groups produced more non-redundant ideas (ru = −.25). In contrast, for larger groups, EBS dramatically outperformed nominal groups (ru = .65). Our confidence in these results is further enhanced by the homogeneity of both clusters of effect sizes. For the aggregate measure of idea quality, the inclusion of group size as a moderator dramatically clarified the heterogeneous results of the primary meta-analysis. Combined nominal and e-nominal groups outperformed EBS groups (ru = −.19), but the heterogeneity of the effect sizes warranted moderator analysis. The group size results were analogous to those obtained for the quantity of ideas: in small groups, nominal groups outperformed EBS (ru = −.24), while in large groups, EBS outperformed nominal groups (ru = .64). The striking symmetry of these findings for differences in quantity and quality of ideas is illustrated in Fig. 4. EBS groups can successfully outperform nominal groups, but the groups need to be large, consistent with Dennis and Gallupe’s (1993) assertion. Specifically these data provide quantitative support for the observation that eight members may be a cut-off point for group size effects for both the quantity and the quality of ideas (see Pinsonneault et al., 1999a). Finally, as expected, EBS group members were more satisfied with the brainstorming process than members of either type of nominal group. Again, this may be due to the performance tracking provided by the electronic medium, which might heighten participants’ awareness of their productivity. Moreover, nominal group members may be less satisfied with the brainstorming process because of their social isolation. For this criterion, the obtained effect sizes were homogeneous and the confidence interval shows statistically significant effects, which obviated the need for subsequent moderator analyses. 7.3. Limitations The perennial criticisms of meta-analysis (e.g., the possibility of missing studies, comparing methodologically disparate studies) are addressed elsewhere (see Rosenthal & DiMatteo, 2000); however, we note a number of issues idiosyncratic to this paper that readers should be consider. The first issue is the inclusion of several studies featuring within-subjects designs. Typically, though perhaps erroneously, such studies are excluded from meta-analyses due to their tendency to artificially inflate effect sizes (Hunter and Schmidt, 1990 and Rosenthal, 1991). In this meta-analysis we included several, but not all, of the obtained studies featuring repeated-measures designs. We chose not to include studies with fully-balanced or partially-balanced designs, wherein participants completed brainstorming tasks in multiple conditions. However, we did include studies featuring mixed designs where participants completed more than one brainstorming task within the same communication condition (e.g., Dennis and Valacich, 1993 and Straus and McGrath, 1994). The descriptive statistics reported in these studies were typically the summations of the task outcomes across multiple brainstorming sessions, rather than pre-post difference scores, which are the sources of inflated effect sizes cited in the literature (see Hunter and Schmidt, 1990 and Morris, 2000). In other cases (e.g., Pinsonneault et al., 1999a), the repeated measures studies included only overall test statistics in their results. In the absence of session-specific descriptive data, it was necessary to rely on effect sizes derived from the authors’ significance tests. Inclusion of such data was predicated on the assumption that the results of brainstorming tasks are not closely related to the content of the task, as has been demonstrated in previous research (Cooper et al., 1998 and Dennis and Valacich, 1993). Thus, if participants were to complete more than one brainstorming task, their results would not likely depend on the content of the tasks to which they had been assigned. Finally, if we were to exclude studies on the basis of repeated-measurement designs, we would also need to exclude studies featuring practice tasks, which would have eliminated the majority of the potential studies. Another notable issue is the potential for publication bias. Of the seventeen studies we obtained for this analysis, four authors, Gallupe, Valacich, Dennis, and Cooper, co-authored between three and six studies each. Moreover, many of these studies were conducted in laboratories at either the University of Arizona or at Queens University. With so few studies in this synthesis, readers should not ignore this potential bias; however, it is unclear how such biases might be approximated quantitatively. We do note that results published by these authors show considerable variation in the magnitude of effects, which suggests that such concerns may be unwarranted.3 7.4. Practical implications of the analysis The results of this meta-analysis should be very useful for organizations that employ group brainstorming techniques, especially given the recent proliferation of electronic collaboration and teamwork (Lipnack and Stamps, 2000 and Mohrman, 1999). The finding that EBS groups are superior to FTF groups with respect to creative idea generation tasks supports the assertion of Mullen et al. (1991) that there is little reason to employ traditional FTF brainstorming. Osborn’s promise of more and better ideas can be obtained in interactive groups, but only if idea generation takes place in electronic environments. In addition to the idea quantity and quality benefits, EBS group members are more satisfied with the electronic group interaction than their FTF counterparts. In summary, organizations wishing to use group brainstorming procedures should eschew the traditional face-to-face approach and opt for virtual communication; and further EBS allows for inclusion of more members without loss in quantity or quality of ideas, instead large groups show additional gains. If FTF brainstorming must be used, cash incentives may increase the volume of ideas produced. While the current results indicate that the traditional brainstorming process is obsolete, a better test of the efficacy of EBS may be its comparison to either type of nominal brainstorming group. This synthesis shows that better results in creative idea generation tasks can be expected from groups of more than eight people. In groups with less than eight members, nominal groups should outperform EBS groups. The production of ideas by nominal groups can also be augmented with cash compensation. Though the current findings underscore how to use group size to obtain the best outcomes from brainstorming groups, the practical benefit of the finding may be limited. For example, if electronic brainstorming groups are also asked to perform other tasks, the benefits of brainstorming may be overshadowed by decrements in performance that will likely occur on other criteria. Organizations may be reluctant to devote time and resources to assembling large groups of people simply to attain a large number of high quality ideas. Thus, from a pragmatic standpoint, EBS may represent a middle ground between the two traditional types of brainstorming, such that nominal groups still have a practical edge. However, in instances where it is relatively easy and inexpensive to form large groups or teams, it is reasonable to use very large electronic groups for this type of idea generation task. A further advantage of EBS groups is that cash compensation appears to be unnecessary. The results of this analysis are also pertinent to the literature on virtual teams in organizational and educational settings (DeRosa et al., 2004, Hantula and Pawlowicz, 2004, Michinov and Primois, 2005 and Pinsonneault and Caya, 2005), and in space travel (Brady et al., 2004). Organizations are relying more heavily on virtual communication, which minimizes the need for face-to-face, or same-site communication. This is due, in part, to increased global expansion, geographic dispersion, the need to integrate the work of top employees, and to save time and financial resources from travel expenses (Cascio, 1999, Kock, 2000, Mohrman, 1999 and Townsend et al., 1998). As a result, it is imperative that organizational scholars fully explore the productivity of computer-mediated groups on a variety of tasks (Pawlowicz, 2003). In the context of the virtual teams literature, the current results are noteworthy given the overwhelming empirical evidence demonstrating that computer-mediated groups working on intellective and decision making tasks are less productive, take longer in reaching consensus, and are less satisfied than their FTF counterparts (Adrianson and Hjelmquist, 1991, Barkhi et al., 1999, Hollingshead, 1996, Olaniran, 1996 and Straus and McGrath, 1994). The current results suggest that idea generation tasks are fundamentally unique in some manner, and this distinction must be addressed in future research. Research on electronic group interaction emphasizes the importance of establishing congruence between the type of task and the technological medium. Accordingly, this meta-analysis illustrates the benefits of moderating creative idea generation tasks with electronic communication, and augments the developing taxonomy of group tasks in virtual teams. It seems ironic that Osborn’s (1957) ideas about the benefits of brainstorming find their fruition in a context that was only the stuff of science fiction when they were first developed. As such, this work should serve more as inspiration than precedent, and as is clear in this meta analysis, studies comparing different forms of brainstorming may no longer be very informative. Instead, those that develop and test theory (e.g., Nijstad et al., 2003) in the context of EBS are the field of the future. What we now know is that when it comes to creative idea generation, the medium matters. What we need to learn is why.