DQ
Group Dynamics: Theory, Research, and Practice 1999, Vol. 3, No. 4, 243-256
Copyright 1999 by the Educational Publishing Foundation 1089-2699/99/S3.00
Group Cohesiveness and Creative Performance
Traci Y. Craig and Janice R. Kelly Purdue University
This study examined the effects of task cohesiveness and interpersonal cohesiveness on group performance on a novel group creativity task. S. Zaccaro and M. C. McCoy (1988) suggested that cohesiveness is 2-dimensional and that the effects of task cohesiveness and interpersonal cohesiveness depend on the task type and the way in which performance is evaluated. The creativity task used here was more interactive than classic idea-generation tasks. Three-person groups were asked to draw a single projected image created by superimposing the images from 3 separate overhead projectors. Drawing performance was judged for creativity and technical quality. The predicted interaction between task and interpersonal cohesiveness was found such that groups in the high task-cohesive and high interpersonal-cohesive conditions produced the most creative group drawings for 1 of the 2 objects presented. Also as predicted, high task cohesiveness alone improved technical quality of the drawings.
People count on groups in many ways to complete a variety of important tasks. For example, in academic, organizational, and gov- ernment settings, groups are asked to solve important problems and to make recommenda- tions about important decisions. Although re- search identifying variables that are important to high-quality group problem solving and deci- sion making is amassing, relatively little is known about another important task type assigned to groups—creativity.
Examples of real-world creativity groups are fairly easy to generate, from advertising creativ- ity teams, to groups of imagineers at Disney, to organizational research and development teams. The types of tasks that these groups are asked to perform, however, may only partially overlap with the most common type of laboratory task used to study creativity—idea generation. Fur- thermore, simple idea-generation tasks do not allow for complex interaction of group members such as would be found in real groups.
The purpose of this study was therefore twofold. First, we wanted to introduce a novel
Trad Y. Craig and Janice R. Kelly, Department of Psychological Sciences, Purdue University.
We are grateful to Rebecca Henry for her comments on an earlier version of this article.
Correspondence concerning this article should be ad- dressed to Janice R. Kelly, Department of Psychological Sciences, Purdue University, 1364 Psychological Sciences Building, West Lafayette, Indiana 47907. Electronic mail may be sent to kelly@psych.purdue.edu.
creativity task that, in addition to an idea- generation component, contained other compo- nents that more closely mirror some additional creative processes that might occur in real creativity groups. Second, we wanted to exam- ine the effects of an important process variable, cohesiveness, on performance on this novel creativity task.
Cohesiveness as a Unitary or a Multidimensional Construct
Small-group researchers have long tried to identify the variables that lead groups to positive outcomes. The group-cohesiveness construct in particular has received renewed interest in recent years. However, a consistent or standard definition of the construct has eluded many researchers, and the varied ways in which it is thought to influence performance may reflect this lack of a singular definition.
In the past, cohesiveness has been defined in a number of ways. Festinger (1950) defined cohesiveness as "the resultant of all the forces acting on the members to remain in the group" (p. 274). This general and vague definition has proven to be difficult to translate into concrete measures. However, other definitions have been equally unwieldy and difficult to operationalize. Carron (1982) defined cohesiveness as a process that reflects a group's tendency to stick together and remain united to reach a common goal. Goodman, Ravlin, and Schminke (1987) defined
243
244 CRAIG AND KELLY
cohesiveness as the commitment of members to the group's task. Evans and Jarvis (1980) concluded that "member attraction to the group" (p. 360) is the most common definition of cohesiveness. Note that some of these definitions make particular note of group tasks or goals, whereas others describe a more general staying-together tendency.
The relationship between cohesiveness and performance also has yielded sometimes contra- dictory results. For example, a review by Evans and Dion (1991) concluded that the relationship between cohesiveness and performance is posi- tive. However, studies also have reached the opposite conclusion. Several other reviews (Lott & Lott, 1965; Shaw, 1976; Stogdill, 1972) revealed similarly mixed effects of cohesiveness on performance. In fact, Kelly and Duran (1985) found that very high levels, as well as low levels, of cohesiveness were associated with poor group performance and thus concluded that the cohesiveness-performance relationship is curvilinear.
It is also possible that the contradictory findings about the cohesiveness-performance relationship are due to the influence of several important moderating variables. Gully, Devine, and Whitney (1995) suggested that the effect of interpersonal cohesiveness, in particular, im- pacts performance to the extent that it allows for the adoption of specific group performance norms. That is, interpersonal cohesiveness might enhance performance to the extent that it leads to the adoption of norms that encourage openness, playfulness, and disagreement. How- ever, if the group adopts norms that lead to conformity, agreement, or engagement in non- task activity, then interpersonal cohesiveness might in fact negatively impact group perfor- mance. For example, Seashore (1954) found that cohesive groups had more uniformly productive members, such that cohesive groups with high performance goals were uniformly more produc- tive and groups with low performance goals were uniformly less productive. Noncohesive groups tended to have more within-group variation in productivity. This finding was attributed to the cohesive groups not being committed to their productivity task. Task interdependence also might influence the extent to which group cohesiveness impacts perfor- mance. For example, a cohesive group may perform well on an interdependent task because
it can more easily coordinate group efforts, whereas such coordination may not produce a benefit on tasks that are more independent (Gully etal., 1995).
The confusion about the cohesiveness- performance relationship has led some research- ers to suggest that cohesiveness may be multidimensional. For example, Gross and Martin (1952) described cohesiveness in terms of two underlying dimensions: task cohesive- ness and interpersonal cohesiveness. Task cohe- siveness is defined as a group's shared commit- ment or attraction to the group task or goal (Hackman, 1976). Interpersonal cohesiveness is defined as the group members' attraction to or liking of the group (Evans & Jarvis, 1980). Task cohesiveness is thought to increase commitment to the task and to increase individual effort by group members on the task. Interpersonal cohesiveness, on the other hand, allows groups to have less inhibited communication and to effectively coordinate their efforts. These two dimensions are therefore likely to have different effects on performance outcomes.
Other researchers also have identified these two similar types of cohesiveness (Bollen & Hoyle, 1990; Carron, 1982; Hackman, 1976; Hogg, 1992; Tziner, 1982). Mullen and Copper (1994) distinguished between three types of cohesiveness: interpersonal cohesiveness, task cohesiveness, and group pride (which has received relatively little attention). Their meta- analysis concluded that task cohesiveness is the "critical element of group cohesiveness" (p. 224) when the cohesiveness-performance rela- tionship is considered. They also concluded that interpersonal cohesiveness might do little more than cause members to exert only as much effort as is required to remain in the group.
The theoretical existence of different types of cohesiveness is fairly well established. How- ever, research that demonstrates the independent operation of cohesiveness constructs is more scarce. Zaccaro and Lowe (1986) compared unidimensional and multidimensional conceptu- alizations of cohesiveness in terms of their effect on performance. Specifically, they proposed that different types of cohesiveness would differen- tially affect performance on specific types of tasks. Zaccaro ijnd colleagues tested this idea using both additive and disjunctive tasks (Zac- caro & Lowe, 1986; Zaccaro & McCoy, 1988). Additive tasks are tasks on which the group's
GROUP COHESIVENESS AND CREATIVITY 245
performance is measured by the sum of each individual member's input. A brainstorming task, in which the group generates as many uses for a common object as possible, is one example. Additive tasks do not require group members to cooperate with one another or to coordinate actions. Rather, the group should produce more ideas as long as each member stays committed to the task (producing solu- tions). Zaccaro and Lowe thus hypothesized that task cohesiveness would aid those groups performing additive tasks through increases in commitment to task and individual effort. Interpersonal cohesiveness, in contrast, which is thought to increase communication and enable the group to better coordinate its efforts, would not be important for a good group performance, because additive tasks do not require communi- cation within the group nor coordination of effort.
Zaccaro and Lowe (1986) conducted a study in which groups of 4 participants were assigned to either a high task-cohesiveness (HTC) or a low task-cohesiveness (LTC) condition and also to either a high interpersonal-cohesiveness (HIC) or a low interpersonal-cohesiveness (LIC) condition. Groups were then asked to complete an additive task. This task involved the group members folding paper moon tents, with a common goal of producing as many moon tents as possible. As predicted, the findings revealed a main effect only of task cohesiveness. Interper- sonal cohesiveness had no effect on perfor- mance. It should be noted, however, that Lott and Lott (1965) found that interpersonal cohe- siveness could result in lower group perfor- mance on additive tasks by causing groups to become distracted and stray from the task. Similarly, Stogdill (1972) suggested that effort that is devoted to the social development of the group is subtracted from effort devoted to the task.
Zaccaro and McCoy (1988) also investigated the effects of interpersonal and task cohesive- ness on performance on a disjunctive task. Disjunctive tasks are tasks in which group performance is determined by the performance of the group's best member. Eureka problems, for which there is an intuitively correct and obvious answer, exemplify the disjunctive task type. For the best member to benefit the group, two conditions must be met: (a) The group member must be committed to the task, and (b)
the group member must feel that the group will be accepting of Jhe new idea. Therefore, task cohesiveness should benefit performance through increased task commitment and increased indi- vidual effort. However, interpersonal cohesive- ness also should benefit the group through open communication and acceptance of ideas.
Thus, Zaccaro and McCoy (1988) hypoth- esized that both task cohesiveness and interper- sonal cohesiveness would be necessary for the group to perform well. The task involved solving a "subarctic survival" problem in which groups ranked a list of items that would be important to survival in a subarctic region in Canada. Supporting their predictions, the best group performance occurred when groups had both high levels of task cohesiveness and high levels of interpersonal cohesiveness compared with all other combinations. It is then apparent that cohesiveness can be informatively con- strued as a multidimensional construct with different effects of both task and interpersonal cohesiveness on performance. However, a group's creative performance presents an espe- cially intriguing case for this two-dimensional construct.
Group Creativity and Cohesiveness
Group creativity often has been studied using the standard brainstorming tasks popularized by Osborn (1957). The introduction of this additive creativity task led to a great deal of research directed primarily at comparing face-to-face groups and nominal groups, with nominal groups typically outperforming interacting groups (Lamm & Trommsdorff, 1973; McGrath, 1984). Recent reviews of the group brainstorm- ing literature suggest that there are three reasons why real groups, as compared with nominal groups, experience lower idea-generation rates: production blocking, free riding, and evaluation apprehension (Diehl & Stroebe, 1987; Paulus & Dzindolet, 1993). Production blocking entails group members being logistically prohibited from expressing their ideas because of other group members' ideas interfering with their own ability to think of, verbalize, and otherwise generate their own ideas. Free riding is the idea that group members experience a diffusion of responsibility because of no explicit or obvious measure of individual effort, leading them to allow other group members to be responsible for
246 CRAIG AND KELLY
the group outcome. Evaluation apprehension, or fear of being negatively evaluated by other group members, may cause group members to choose not to express the ideas that they have successfully generated. Evaluation apprehen- sion is thought to cause individual members to censor their performances so as to adhere to group norms. Although support has been found for the operation of all three variables, evalua- tion apprehension is particularly relevant to cohesiveness and its subsequent effects on performance. Specifically, cohesiveness is thought to impact many group processes that should alleviate feelings of evaluation apprehen- sion and the subsequent production loss experi- enced by groups.
Interpersonal cohesiveness has many positive influences on group interaction and group creativity. For example, Hogg (1992) suggested that interpersonal liking can increase creativity by allowing communication to be less inhibited. Firestein (1990) found that groups who per- formed well creatively also laughed and smiled more, as well as made more verbal support statements and fewer critical statements. Fi- nally, Nystrom (1979) suggested that cohe- siveness should enhance creativity through in- creasing feelings of psychological safety and self-actualization. All of these processes that result from interpersonal cohesiveness also may alleviate, at least to some extent, feelings of evaluation apprehension and thus improve creative performance. In addition to the positive influence of interpersonal cohesiveness on group creativity, task cohesiveness should en- hance creativity by increasing group members' willingness to contribute to the group effort (Hackman, 1976).
However, Zaccaro and Lowe's (1986) conclu- sion that additive tasks are influenced only by task cohesiveness and are not influenced by interpersonal cohesiveness raises an interesting issue for creative group performance. Much of the research that has examined group creativity has used brainstorming tasks. The general form of a brainstorming task requires the group to generate as many uses as it can for a specific object (knife, brick, etc.). Thus, brainstorming tasks are additive. Given Zaccaro and Lowe's findings, one might hypothesize that group creativity is not enhanced by interpersonal cohesiveness and that only task cohesiveness is
necessary for a group to perform well on a creativity task.
In fact, though, the handful of studies that have directly investigated the effects of interper- sonal cohesiveness on creative group perfor- mance have generally found benefits. On the one hand, Paulus, Larey, and Ortega (1995), in their examination of nominal and interacting brain- storming groups from a team-oriented organiza- tion (in which cohesiveness was thought to be high), found that nominal groups generated more ideas than did interacting groups. Paulus et al. concluded that although there were benefits of cohesiveness to group creativity, these benefits were not great enough to overcome productivity loss in this particular study.
On the other hand, Cohen, Whitmyre, and Funk (1960) found that groups who were interpersonally cohesive (allowed to choose a partner) and trained in creative problem solving generated more ideas than similarly trained noncohesive groups (assigned to a nonchosen partner). Similarly, Woodman, Sawyer, and Griffin (1993), in a review of organizational creativity, suggested that a group's creative performance is enhanced by a high-commitment work system (task cohesiveness) and that creative performance has a curvilinear relation- ship with group cohesiveness (interpersonal cohesiveness), with only extremely high levels of interpersonal cohesiveness leading to perfor- mance decrements. Given these general find- ings, it seems likely that both task and interpersonal cohesiveness should have a posi- tive effect on creative performance on brain- storming tasks.
But real-world groups who are asked to perform creativity tasks do not simply have the task of brainstorming an idea. Most of these groups are responsible for the project from conceptualization through implementation. Real- world creativity tasks are thus more complex than brainstorming tasks. Lamm and Tromms- dorff (1973), in their review of brainstorming research, encouraged researchers to keep in mind that the generation of ideas is only one of many activities required of real groups. So, although one might conclude that both task and interpersonal cbhesiveness may improve cre- ative idea generation, there is less knowledge about how these variables affect other aspects of creative performance.
GROUP COHESIVENESS AND CREATIVITY 247
Overview and Hypothesis
The purpose of this study was to investigate the effects of task and interpersonal cohesive- ness on group performance on a more complex and interactive creativity task. The interactive task introduced here required each person of a three-person group to draw on a separate overhead projector. Each of the three projectors was focused onto a single screen, so that the image on each projector overlapped with the images on the other two projectors. Groups were then asked to draw an object (i.e., a mansion or a bridge). The drawings were photographed and then rated by independent judges for creativity and technical quality.
This task can be thought of as occurring in stages. According to Steiner's (1972) suggestion that task type be determined by the dimension on which the group performance was evaluated, this task had multiple, distinct task types within it. For example, to complete the task, the group first had to generate ideas about what should be drawn or included in the picture. However, from among the solutions generated, a single creative idea had to be chosen for further development. Thus, this first stage most closely mirrors a disjunctive task. According to the previously reviewed research (Zaccaro & McCoy, 1988), the selection goal in this generation stage should be accomplished most effectively by groups in which there are both high levels of interpersonal cohesiveness and high levels of task cohesive- ness. Interpersonal cohesiveness allows for less inhibited communication, and task cohesiveness provides a commitment to a common goal of selecting the most creative idea. Good perfor- mance on this first stage of the task should be reflected in ratings of novelty or creativity of the group's drawing.
The second stage of the task involved the actual drawing of the object. Performance here depended on the group's ability to allow each member to contribute to the composition by adding detail and complexity. The execution stage is therefore more additive in nature and thus should be facilitated by task cohesiveness (Zaccaro & Lowe, 1986). The group's commit- ment to the task should allow the group to execute whatever ideas were generated in the initial generation stage of the task. The perfor- mance of the group as the members execute the
drawings should be reflected in ratings of the technical quality_pf the composition.
Therefore, because the generation and execu- tion stages had different requirements and measures of successful performance, we ex- pected that each type of cohesiveness would operate differently for each stage. Specifically, generation, as reflected by ratings of creativity, should be facilitated by a combination of HTC and HIC. However, execution, as reflected by ratings of technical quality, should be facilitated by task cohesiveness alone. Interpersonal cohe- siveness was not predicted to affect technical performance because the increase in communi- cation does not aid in the adding of detail or other additive aspects of the task.
Because the creativity task used in this study was new, we also were interested in the degree to which individual measures of creativity were related to performance. Specifically, we explored whether individual levels of creativity were cor- related with the group's overall creative perfor- mance. In addition, we used this measure as a covariate to explore the unique contributions of cohesiveness to creative group performance.
Method
Participants and Design
College students enrolled in introductory psychology classes participated for partial course credit. Participants were recruited for a study entitled "Creativity and Projected Images." A total of 189 students (94 men and 89 women) participated in groups of 3.1 All possible gender compositions of a 3-member group occurred naturally through the randomization process. The design was a 2 X 2 X 2 between-groups design, with two levels (high or low) of task cohesiveness, two levels (high or low) of interpersonal cohesiveness, and two objects (mansion or bridge).
Task
The task was a novel creativity task that required a group to cooperate and coordinate
1 Two groups wer$ eliminated from all analyses because of the use of incorrect order of measures in one case and a participant being acquainted with an experimenter in another case. Additional variations in degrees of freedom from analysis to analysis are due to missing data.
248 CRAIG AND KELLY
Table 1 Mean Ratings for Objects on Overall Difficulty, Ease of Contribution, Creativity, Ease of Coordination, and Complexity
Object
Fountain Boat Mansion Bridge
n
Al 43 43 41
Difficulty
3.96 3.60 4.30 3.54
Ease of contribution
5.45 5.65 5.84 6.07
Dimension rated
Creativity
5.17 5.44 4.84 5.54
Ease of coordination
2.85 2.91 2.86 2.46
Complexity
4.13 4.47 4.02 4.46
Note. All ratings were made on a 7-point scale ranging from 1 (low) to 7 (high).
efforts to produce a single creative product. Group members were told that the task was primarily concerned with creativity and that group members would jointly draw a single picture of an assigned object on overhead projectors. Three overheads were set up on a table such that the projection from each was focused on a single screen. Each group member was assigned to a single overhead projector and overhead pen (each a different color). The group was then given an object to draw, with instructions that the group product should be one picture projected onto the common screen.
The objects were selected on the basis of pretest scores. Four unitary objects were in- cluded in the pretest: a fountain, a boat, a mansion, and a bridge. Group members from 22 three-person groups drew one, two, three, or all four of the objects during a 1-hr task session. They were asked to rate each object that they drew on several dimensions using a 7-point scale. Objects were rated on (a) how difficult the object was to draw, (b) how easy it was to contribute to the drawing, (c) how easy it was to be creative with the object, (d) how easy it was to coordinate efforts, and (e) how complex the object was to draw. A one-way analysis of variance (ANOVA) followed by Student New- man-Keuls tests to compare individual cell means revealed no significant differences be- tween objects on any of the tested dimensions (all^s > .05; see Table 1). Therefore, mansion and bridge were randomly chosen to be included in the main study.
Procedure
Participants arrived at the lab area in groups of 3 or 4. Each person was given a questionnaire that asked whether he or she knew any of the
other participants. A modified version of the Creative Behavior Inventory (CBI; Hocevar, 1979) also was administered to the participants. If all participants were unacquainted, 3 of the 4 participants were randomly assigned to the group condition. The additional participant, or 1 of the acquainted participants, was assigned to an alternative task. Each group of strangers was given the task- and interpersonal-cohesiveness manipulations for the conditions to which they had been randomly assigned (see description to follow). At the conclusion of the manipulation session, the group was escorted to the task room.
After the group had arrived in the task room, a second experimenter, who was unaware of the manipulated task-cohesiveness conditions, intro- duced the task rules.2 A description of the task rules was distributed to each member. The rules also were read aloud by the experimenter. Rules included such things as drawing on only the assigned overhead, erasing on only the assigned overhead, not using any materials other than those provided for the task, and using only the assigned pen. After the group members indi- cated that they understood the rules, a drawing was held to assign each group member to an overhead projector. The group was shown a pretest figure that consisted of three interlocking geometric shapes, each of a different color. The group was asked to reproduce the pretest figure as accurately as it could. The group was told that this was only a pretest to allow the group to become familiar with the task. When the group indicated to the experimenter that it was finished drawing the practice object, a second question- naire that assessed interpersonal cohesiveness
2 Because of the procedures involved in the interpersonal- cohesiveness manipulation, the experimenters were not naive to the group's assignment to this condition.
GROUP COHESIVENESS AND CREATIVITY 249
and task cohesiveness was administered. Details about this questionnaire are included in the Measures section. The group members were then asked to approach their projectors again to begin the main creativity task.
The group was told,
Your task today is to draw a single picture as a group on the overhead projectors. Each of you has already been assigned to a projector by random draw. Once again, the overhead pens on each projector are each a different color and can only be used on the projector to which they are assigned. All of the rules still apply to this portion of the task. As this study is focusing on creativity, please be as creative as possible in your endeavors here. There is no time constraint. When you as a group decide you have been as creative as you possibly can in your drawing let me know that you are finished. Today your group will start by drawing a
The group was informed that it could draw any type of that it wished. The group was then randomly assigned the object it was to draw, either a mansion or a bridge. The group was not given a time limit for completion of the task. When the group completed the drawing, the participants were debriefed and dismissed.
Cohesiveness Manipulations
Task cohesiveness. Groups were randomly assigned to either an HTC or an LTC condition at the beginning of the experiment, following the initial introduction. Using a procedure similar to that used by Zaccaro and McCoy (1988), the importance of the study was emphasized in a verbally presented cover story relating real-world groups to the experiment. Participants in the HTC condition were given a written copy of a rationale and were asked to silently read along with the experimenter. The rationale emphasized the growing importance of the group in the American workplace and how future employment is dependent on the ability to work well in a group. The rationale also mentioned that job interviews often ask about experience and ability to work in a group setting and that the needs of the country can be best met when people have experience working in a group. The groups in this condition also were informed that the group that was judged to give the "best effort" would be awarded a $30 prize at $10 per group member. They then wrote down their names and addresses so that, should they win, the award could be mailed to them. Those
participants in the LTC condition were not given any information .about a rationale or a reward until after the debriefing.
Interpersonal cohesiveness. Each group also was randomly assigned to either an HIC or an LIC condition. This was done immediately following the initial introduction to the experi- ment. Again, the procedure was adapted from Zaccaro and McCoy (1988). Groups in the HIC condition were asked to sit in three chairs that were prearranged in a triangular configuration so that they were all facing one another. Group members were told the following:
It is important that you feel comfortable working together as you will be working as a group throughout this experiment. So I would like for you to spend a few minutes now getting to know each other. Don't be concerned about how long you spend. I will let you know when it is time to move on to the next part of this experiment.
Each group was allowed approximately 15 min to get to know one another. The group also was asked to come up with a team name. Each member was given a name tag and told,
Your first group decision will be to decide on a team or group name that will represent your group for the remainder of the experiment. Once you have all agreed on a group or team name then write the team name in large letters in the center of the name tag.
Groups in the LIC condition were not given any time to get to know one another and were taken directly to the task session portion of the experiment.
Measures
Interpersonal cohesiveness. Interpersonal cohesiveness was assessed after the pretest drawing. The manipulation check questionnaire items, assessed on 7-point scales, included questions about how much the group member liked the other group members now, expecta- tions about liking in the future, how important it was that members get along, how similar the group member felt to other members, and how important socializing was during the session. This five-item scale demonstrated an acceptable internal reliability (Cronbach's a = .72).
Task cohesiveness. Task cohesiveness was assessed after the pretest drawing. This scale asked group members to report on 7-point scales if they felt their group was engaged in the task and enjoying the task, how important it was to
250 CRAIG AND KELLY
do well on the task, how meaningful the task was, and whether they expected there would be benefits from their group's performance. This five-item scale was found to have good internal reliability (Cronbach's a = .87).
Individual creativity. Each participant com- pleted a modified version of the CBI (Hocevar, 1979). The CBI is a checklist of 89 specific creativity-related achievements, accomplish- ments, and behaviors that a person may have experienced. Hocevar cited several studies in which inventories of creative accomplishments and achievements were the best predictors of later creative behavior (Holland & Nichols, 1964; MacKinnon, 1962; Mischel, 1968). The length of the original questionnaire presented two primary concerns. One was that the time it would take to complete the questionnaire could cause the participants to become fatigued and decrease their interest in the later task. Second, there was concern that completing such a lengthy questionnaire, especially if the majority of the items could not be checked off, would lead to a decrease in mood and possibly creative self-esteem. To avoid these two concerns, the inventory was modified to decrease the amount of time to complete it and the opportunity for a decline in mood or self-esteem. The modified version created for this study organized the majority of these specific experiences into seven categories, such as "musician," "artist," and "writer." Under each category label were all the experiences pertaining to accomplishments in that category. With the categories organized in this way, a musician would proceed to answer questions about musical talent, whereas nonmu- sicians could skip to the next category of experiences more relevant to their accomplish- ments. Each cluster contained between 6 and 16 possible items to check, for a total of 67 items. A total score was determined by summing across all checked items. In addition, the modified version left out some of the items that were suspected to be infrequent in an undergraduate population (e.g., "Have you ever had a math- ematics paper published?"). To determine if the modified version was comparable with Hocev- ar's CBI, the two versions were administered to 23 participants as part of a separate study. There was a strong positive correlation between the two versions, r(23) = .76, p < .000.
Coding of Photographed Drawings
Each group's final picture was photographed, and these photographs were then rated on several dimensions by undergraduate research assistants who served as independent judges unaware of experimental condition. The pictures of the group drawing were rated for creativity and technical quality. The coders that judged the photographs for creativity used a definition of creativity that emphasized uniqueness, differ- ence from others, original ideas, and different views. Two dimensions assessed technical quality: complexity and coherence. Complexity was defined to coders as "the degree to which detail and realism were appropriately added to the drawing." Coherence was defined for our independent raters as "how well the drawing appears to be a single work and not the product of different individuals."
Ratings were made through the use of the sort-resort technique developed by Hackman, Jones, and McGrath (1967). Pictures were first sorted for a particular dimension into three piles (high, medium, and low). Each pile was then resorted. The high pile was sorted into two piles (high-high and high-low), the medium pile into three piles (medium-high, medium, and medium- -low), and the low pile into two piles (low-high and low-low). Finally, judges went through each pile to make sure of item placement. This resulted in seven piles that corresponded to a 7-point scale ranging from 1 {low) to 7 {high) on the rated dimension.
Three coders were first given a practice set of 60 photographs of drawings taken from pretest sessions to code. The reliability for this practice set, using intraclass correlations (Orwin, 1994), was established after two practice trials for creativity at rx = .73, coherence at rx = .81, and complexity at fx = .76. All coders were then asked to code the entire experimental set of photographs. Interrater reliability for the draw- ings was again determined by intraclass correla- tions. The intraclass correlations were as follows for the experimental set of photographs: creativ- ity. ri = -59; coherence, rx = .74; and complexity, rY = .69. Orwin indicated that an intraclass correlation is considered good when it is between .60 and .74 and fair when it is between .40 and .60. Thus, all reliabilities were satisfactory.
GROUP COHESIVENESS AND CREATIVITY 251
Results
Results are presented in three sections. First, the manipulation check results are examined, followed by the results obtained from the analysis of the picture coding. The third section contains the results of the individual-level creativity covariance analysis.
We used 2 X 2 X 2 ANOVAs with two levels of task cohesiveness (high and low), two levels of interpersonal cohesiveness (high and low), and two objects (mansion and bridge) to analyze group-level variables such as creativity of the drawing. We used simple effects tests to decompose higher order interactions. Individual- level data, such as questionnaire responses, were analyzed using a hierarchically nested design with participants nested in groups and groups nested in condition and object, as suggested by Anderson and Ager (1978).
Manipulation Checks
We conducted 2 X 2 ANOVAs on both task- and interpersonal-cohesiveness pretask manipu- lation check scales. A main effect of task- cohesiveness condition was found for the task-cohesiveness scale, F ( l , 57) = 5.12, p < .03. This main effect was in the expected direction, such that those in the HTC condition reported higher levels of task cohesiveness (M = 5.09) than did those in the LTC condition (M = 4.66). No other significant main effects or interactions occurred.
The interpersonal-cohesiveness manipulation check scale also showed the expected main effect of level of interpersonal cohesiveness, F ( l , 57) = 4.13,p < .05. Group members in the HIC condition reported higher interpersonal cohesiveness (M = 5.65) than did those in the LIC condition (M = 5.40). No other significant main effects or interactions occurred. Thus, the manipulations of task and interpersonal cohesive- ness were successful.
Cohesiveness Effects on Drawing Performance
Creativity. Creativity ratings were used to assess performance on the first stage (genera- tion) of the task. Mean creativity ratings are shown in Table 2. We predicted that the interaction of HIC and HTC would lead to the
Table 2 Creativity RatingsJ>y Cohesiveness Conditions and Objects Drawn
Condition
Low task cohesiveness M SD n
High task cohesiveness M SD n
Low task cohesiveness M SD n
High task cohesiveness M SD n
Interpersonal cohesiveness
Low
Mansion
4.7 U 1.63 7
4.38a,,, 1.71 8
Bridge
4.25a 1.75 8
5.79b 1.38 7
High
3.93a 1.17 7
5.63b 1.25 8
4.69a,b 0.92 8
3.75a 1.36 8
Note. Ratings were made on a 7-point scale, with 7 being highly creative. Means with different subscripts (assigned separately for each object) indicate significant differences at p < .05.
most creative performance. ANOVA procedures revealed a three-way Object X Task Cohesive- ness X Interpersonal Cohesiveness interaction, F ( l , 53) = 9.53, p < .01. A simple effects test, used to separately examine the predicted two- way interaction between interpersonal and task cohesiveness for each object, was significant for the object mansion, F ( l , 53) = 3.98,/? < .05. As predicted, groups in the HIC condition per- formed well when also in the HTC condition (M = 5.63), but when task cohesiveness was low, performance was worse (M = 3.93). Groups in the LTC and LIC condition (M = 4.71) and those in the HTC and LIC condition (M = 4.38) produced drawings that were not significantly different from either of the other two possible cohesiveness condition combinations.
Simple effects tests also showed a significant Interpersonal Cohesiveness X Task Cohesive- ness interaction for the object bridge, F( 1,53) = 5.57, p < .02. However, this interaction was not in the predicted'direction. Groups in the HTC and HIC condition (M = 3.75) performed worse than those in the HTC and LIC condition (M = 5.79). Those in the LTC and LIC condi-
252 CRAIG AND KELLY
tion (M = 4.25) also performed worse than those in the HTC and LIC condition. The mean creative performance score for groups drawing a bridge in the LTC and HIC condition (M = 4.69) was not significantly different from any of the scores in the other three cohesiveness condition combinations.
To examine our specific hypothesis that the groups in the HTC and HIC condition would have more creative drawings than the groups in the remaining three conditions, we conducted a means comparison using a one-way ANOVA. The resulting one-way ANOVA for groups drawing a mansion revealed an effect of condition on creativity of the drawing, F ( l , 28) = 4.71, p < .04. This effect was such that those in the HTC/HIC condition had higher creativity ratings (M = 5.63) than the average of those in the other three conditions (M = 4.34). A similar one-way ANOVA was conducted for groups drawing a bridge. There was a marginal effect of condition, F ( l , 30) = 3.53, p < .07, such that those in the HTC/HIC condition had lower creativity ratings (M = 3.75) than the average of those in the remaining three conditions (M = 4.86).
Thus, our prediction for the first stage (generation) of the creativity task was supported when groups were asked to draw a mansion. The HTC/HIC groups were more creative than the groups in the other three conditions. Although a marginal interaction was found for groups drawing a bridge, this interaction was in a direction opposite to that predicted, such that groups in the HTC and HIC condition were lower in creativity than groups in the other three conditions.
Technical quality. Performance on the sec- ond stage of the task, execution, was assessed by ratings of the technical quality of the drawings. Two dimensions measured technical quality: coherence and complexity. We predicted that groups in the HTC condition would score higher on both of these dimensions than would those in the LTC condition. Interpersonal cohesiveness was not predicted to have an effect on the technical quality of the drawings.
Coherence was defined as how well the drawing "hung together" or appeared to be the work of a single person rather than a conglomera- tion of several individuals' work. There was a predicted main effect of task cohesiveness, F ( l , 53) = 8.00, p < .01, such that the drawings of
groups in the HTC condition were judged to be more coherent (M = 5.71) than those of groups in the LTC condition (M = 4.70). No effect was found for interpersonal cohesiveness.
Complexity was defined as the degree of detail in the drawing. There was an interaction between object and task cohesiveness, F ( l , 53) = 5.27, p < .03. Simple effects tests showed that there was an effect of task cohesiveness on complexity for drawings of mansions, F ( l , 53) = 4.19, p < .05. Groups drawing mansions in the HTC condition produced more complex draw- ings (M = 4.66) than groups drawing mansions in the LTC condition (M = 3.71). However, task cohesiveness did not have an effect on the complexity of drawings of bridges, F ( l , 53) = 1.26, p < .27. In addition, no effects were found for interpersonal cohesiveness.
The hypothesis that task cohesiveness would positively affect the technical quality of the group's performance was supported for the coherence dimension. However, the complexity dimension results indicate that task cohesive- ness enhanced the complexity of mansion drawings but not of bridge drawings.
Creative Behavior Inventory
A group creativity score was calculated by averaging the group members' individual CBI scores. The resulting average group CBI score was then correlated with the three outcome variables. Only creativity scores for the draw- ings, r(59) = .49, p < .001, were significantly correlated with group CBI scores. Group CBI scores were not significantly correlated with coherence or complexity. Thus, the generation stage of the task seemed most closely connected to creativity.
Given the strong relationship between CBI scores and creativity ratings, we were interested in the extent to which task cohesiveness and interpersonal cohesiveness could affect creative performance over and above what could be predicted from individual-level group creativity scores. To explore this relationship, we con- ducted a 2 (HIC vs. LIC) X 2 (HTC vs. LTC) X 2 (mansion vs. bridge) analysis of covariance with the group ^average of the individual CBI scores entered as a covariate. As reflected by our correlational results, group CBI scores were strongly related to creativity ratings, F ( l , 52) = 9.54, p < .003. Specifically, higher CBI scores
GROUP COHESIVENESS AND CREATIVITY 253
led to higher group creativity ratings. More important, however, the three-way interaction of Task Cohesiveness X Interpersonal Cohesive- ness X Object remained significant, after statistically controlling for differences in creativ- ity, F ( l , 52) = 6.71, p < .01. Simple effects tests were conducted to separately examine the two-way Task Cohesiveness X Interpersonal Cohesiveness interaction on creativity ratings for each object. Analyses for the object mansion showed no significant effect of group CBI scores, F ( l , 29) = 0.05, p < .83. However, the Task Cohesiveness X Interpersonal Cohesive- ness interaction remained, although it was only marginally significant, F ( l , 29) = 6.71, p < .07. For the object bridge, however, group CBI scores were significantly related to creativity, F ( l , 30) = 18.16, p < .001. However, the two-way Task Cohesiveness X Interpersonal Cohesiveness interaction was no longer signifi- cant, F ( l , 30) = 2.60, p < .12. Thus, the influence of task and interpersonal cohesiveness on creativity ratings was maintained for groups drawing mansions over and above the effects of individual-level creativity scores. However, the effect of cohesiveness on creativity ratings for groups drawing bridges was eliminated when CBI scores were taken into account.
Discussion
The purpose of this study was twofold. First, we wanted to extend the group creativity literature by introducing a creativity task that was more complex than the traditional idea- generation task. Second, we wanted to explore the effects of task and interpersonal cohesive- ness on group creativity. The results reflecting on each of these purposes are discussed in turn.
Creativity Task
The majority of experimental research deal- ing with group creativity has used only a single type of task—an idea-generation or brainstorm- ing task. Although brainstorming certainly is consistent with some definitions of creativity (e.g., novel associations), it provides a some- what limited context in which to explore the creative potential of groups. The simple additive nature of brainstorming tasks does not allow for the potential benefits of synergy and coordina- tion of action that can occur in groups.
The group creativity task introduced here, although fanciful, does capture many of the elements of creativity tasks that real-world groups are asked to perform. Specifically, for groups to perform well on this task, they not only must generate novel associations or cre- ative ideas but also must execute these ideas. In addition, the execution stage of the task involves careful physical and cognitive coordination of actions to produce a high-quality performance. This generation-execution sequence can be found in a number of real-world groups, including advertising creativity teams, imagi- neers, and research and development teams. Although researchers have some preliminary knowledge of the variables that may affect the generation stage of creativity (e.g., production blocking, social inhibition), much more can be learned by using a task that allows examination of the group processes that influence perfor- mance throughout the creative process. A task that is both creative and more complex than idea-generation creativity tasks can allow re- searchers to gain knowledge about group creativity in situations that mirror real-world tasks that require more than the exchange of ideas.
In addition, most group creativity research has focused on comparisons of group and individual output. Because the finding that individuals outperform groups on brainstorming tasks is very robust (McGrath, 1984), most research has examined the negative group variables that inhibit creative performance. The limited task context has not previously allowed group researchers to explore positive group variables and their potential beneficial effects on group creativity.̂ We hope that by introducing a novel and more complex creativity task more research on group creativity will be stimulated.
Effects of Task and Interpersonal Cohesiveness on Group Creativity
The variables that we chose to concentrate on in our initial investigation of group creativity were task and interpersonal cohesiveness. A great deal of research has investigated the effects of cohesiveness on other aspects of group performance primarily because of the facilitat- ing effects that are presumed to emerge. Interesting recent research, however, has sug-
254 CRAIG AND KELLY
gested that different types of cohesiveness may have different effects on performance depending on the type of task under investigation (Zaccaro & Lowe, 1986; Zaccaro & McCoy, 1988). Therefore, we felt that task cohesiveness and interpersonal cohesiveness would have different effects on different elements of our creativity task. Specifically, following the results of Zaccaro and McCoy, we thought that the interaction of high levels of both task and interpersonal cohesiveness would facilitate per- formance on the creativity or generation stage of our creativity task because this stage seemed to be most closely characterized as a disjunctive task. However, we thought that task cohesive- ness alone would facilitate performance on the performance or execution stage of our creativity task because this stage seemed to be most closely characterized as an additive task.
Our first hypothesis was supported for one of the two objects used for this task. Groups that drew mansions performed as predicted, such that those in the HTC/HIC condition received higher creativity scores than those in the remaining three conditions. A combination of HTC and HIC facilitated creative performance, presumably by increasing task-directed effort and decreasing inhibited communication. How- ever, the results for groups that drew bridges did not support our hypothesis. In fact, groups in the HTC/HIC condition actually performed worse on creativity than groups in the remaining three conditions.
The most plausible explanation for these results deals with possible differences between the two objects presented to groups to draw. Pretest data showed that the two objects were equivalent in terms of difficulty, complexity, ease of contributing to the drawing's creativity, and coordinating efforts. However, an examina- tion of the actual drawings of groups in the main study showed a startling difference between the two objects. Groups who were assigned to draw mansions drew a wide variety of different types of mansions. However, groups who drew bridges seemed to primarily use two salient bridge exemplars during the generation stage of the task. In particular, more than two thirds (23 out of 30) of the groups chose to draw the Golden Gate Bridge. An additional 3 groups drew a well-known local bridge. Therefore, it appears that the presence of salient exemplars eliminated the generation stage of the task for
these groups. Our hypothesis concerning creativ- ity dealt only withJhe effects of cohesiveness on the generation stage, so it may have been relevant only for groups who drew mansions.
It is interesting that groups drawing bridges actually performed worse in the HTC/HIC condition. It may be that interpersonal cohesive- ness is important only during the generation stage of the task. When that generation stage is bypassed, however, interpersonal cohesiveness may actually detract from performance. For example, Lott and Lott (1965) suggested that interpersonal cohesiveness may have negative effects on group performance because of the distracting nature of the interaction that often takes place in interpersonally cohesive groups.
We also predicted that task cohesiveness would be important for improving the technical quality of the groups' drawings during the execution stage of the task. Following the results of Zaccaro and Lowe (1986), we also predicted that interpersonal cohesiveness would have no effect on this stage of the task. The results largely supported our hypotheses for both objects and on both measures of technical performance quality. For both groups drawing mansions and groups drawing bridges, HTC led to higher coherence scores, whereas interper- sonal cohesiveness had no effect on perfor- mance. Also, for groups drawing mansions, HTC led to higher complexity scores, whereas again interpersonal cohesiveness had no effect on performance. The results on complexity scores for groups drawing bridges were not statistically significant. Therefore, performance during the execution stage of the task seemed to be directly related to increased effort and may have been relatively unaffected by the uninhib- ited nature of the communication process.
Implications and Future Directions
Examination of cohesiveness and its effects on group performance on an interactive, interde- pendent creativity task is not common. Although the use of idea-generation tasks has done much to further understanding of a group's creative process, more interactive tasks also should be informative about group processes during inter- dependent creative endeavors. Obviously, more research must be conducted to confirm our preliminary findings concerning the effects of task cohesiveness and interpersonal cohesive-
GROUP COHESIVENESS AND CREATIVITY 255
ness on creative performance. However, these findings are provocative in that they give some insight into how real-world creative perfor- mance could be improved.
The fact that interpersonal cohesiveness might lead to more uninhibited communication and hence improve some aspects of creativity is not especially startling, despite the fact that there is little empirical validation of this notion. Task cohesiveness, in contrast, should enhance performance by increasing individual effort or commitment to the group task. However, our manipulation of task cohesiveness involved both a rationale (to increase task importance) and a possible reward (extrinsic motivator). It is possible that because of the multicomponent nature of our manipulations, future research should try a different approach for increasing task cohesiveness. For example, the notion that commitment to the task in the form of task cohesiveness is also an integral component of group creativity is more novel and might be a more important component of task cohesiveness than the aspects that our study emphasized. Overall, task cohesiveness, through its positive effect of increased effort such as we found with other task types, seems to be especially impor- tant to the execution of the creative idea.
Our results seem to imply that real-world groups could improve performance by taking into account both the type of cohesiveness that would facilitate performance and the type of task that the group is being asked to perform. For example, groups that are called on to replicate or reproduce accurate products might best be served by becoming task cohesive and perhaps not encouraged to become too interper- sonally cohesive. However, groups that are required to create, imagine, or generate novel ideas or products might benefit from being encouraged to develop both task cohesiveness and interpersonal cohesiveness to improve group performance.
Other group variables that have been found to facilitate or debilitate group task performance in general also should be investigated with respect to creativity tasks. For example, it might be interesting to know how groups perform on creativity tasks with varying types of group structure. Group composition effects, in particu- lar, may be especially important to group creativity. For example, it could be that diversity of opinion or heterogeneity of ideas could
improve the generation of creative ideas, at least under certain conditions of task and interper- sonal cohesiveness. However, heterogeneity of ideas also may detract from other aspects of performance, especially during the execution or implementation of ideas.
Other variables also might moderate creative performance. As discussed before, adoption of productivity norms is likely to moderate the effect of cohesiveness on performance (Sea- shore, 1954). Coordination issues also might be explored, especially on more complex creativity tasks. For example, Cannon-Bowers, Salas, Blickensderfer, and Bowers (1998) found that groups that were cross-trained on a task tended to operate more efficiently than groups that were not cross-trained. It would be interesting to examine the role of each type of cohesiveness in these groups. Finally, comparing the creative performance of groups with individuals on this task also might provide a different answer to the question of whether groups are really less creative than individuals.
Conclusion
There is much to be learned about group creativity. Empirical efforts directed toward understanding group creativity should use what researchers have learned from simple tasks such as idea generation and attempt to apply this knowledge to more complex interdependent tasks that involve group interaction and coordi- nation. It is these types of interactive tasks that more closely mirror real-world creativity tasks. This study demonstrates the importance of such investigations.
References
Anderson, L. R., & Ager, J. W. (1978). Analysis of variance in small group research. Personality and Social Psychology Bulletin, 4, 341-345.
Bollen, K. A., & Hoyle, R. H. (1990). Perceived cohesion: A conceptual and empirical examination. Social Forces, 69, 479-504.
Cannon-Bowers, J. A., Salas, E., Blickensderfer, E., & Bowers, C. (1998). The impact of cross-training and workload on team functioning: A replication and extension <j>f initial findings. Human Factors, 40,92-101. '•
Carron, A. (1982). Cohesiveness in sport groups: Interpretations and considerations. Journal of Sport Psychology, 4, 123-138.
256 CRAIG AND KELLY
Cohen, D., Whitmyre, J. W., & Funk, W. H. (1960). Effect of group cohesiveness and training upon creative thinking. Journal of Applied Psychology, 44, 319-322.
Diehl, M., & Stroebe, W. (1987). Productivity loss in brainstorming groups: Toward the solution of a riddle. Journal of Personality and Social Psychol- ogy, 53, 497-509.
Evans, C. R., & Dion, K. L. (1991). Group cohesion and performance: A meta-analysis. Small Group Research, 22, 175-186.
Evans, C. R., & Jarvis, P. A. (1980). Group cohesion: A review and re-evaluation. Small Group Behavior, 11, 359-370.
Festinger, L. (1950). Informal social communication. Psychological Review, 57, 411^-16.
Firestein, R. L. (1990). Effects of creative problem solving training on communication behaviors in small groups. Small Group Research, 21, 507-521.
Goodman, P. S., Ravlin, E., & Schminke, M. (1987). Understanding groups in organizations. Research in Organizational Behavior, 9, 121-173.
Gross, N., & Martin, W. E. (1952). On group cohesiveness. American Journal of Sociology, 57, 546-564.
Gully, S. M., Devine, D. J., & Whitney, D. J. (1995). A meta-analysis of cohesion and performance: Effects of level of analysis and task interdepen- dence. Small Group Research, 26, 497-520.
Hackman, J. R. (1976). Group influence on individu- als. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology (pp. 1455-1525). Chicago: Rand-McNally.
Hackman, J. R., Jones, L. E., & McGrath, J. E. (1967). A set of dimensions for describing the general properties of group-generated written passages. Psychological Bulletin, 67, 379-390.
Hocevar, D. (1979). The development of the Creative Behavior Inventory (CBI). Los Angeles: University of Southern California. (ERIC Document Reproduc- tion Service No. ED 170 350).
Hogg, M. A. (1992). The social psychology of group cohesiveness: From attraction to social identity. New York: New York University Press.
Holland, J. L., & Nichols, R. (1964). Prediction of academic and extracurricular achievement in col- lege. Journal of Educational Psychology, 55, 55-65.
Kelly, L., & Duran, R. L. (1985). Interactions and performance in small groups: A descriptive report. International Journal of Small Group Research, 1, 182-192.
Lamm, H., & Trommsdorff, G. (1973). Group vs. individual performance on tasks requiring ide- ational proficiency (brainstorming): A review. European Journal of Social Psychology, 3, 361— 388.
Lott, A., & Lott, B. (1965). Group cohesiveness as interpersonal attraction: A review of relationships with antecedent and consequent variables. Psycho- logical Bulletin, 64, 259-309.
MacKinnon, D. (1962). The nature and nurture of creative talent. American Psychologist, 17, 484— 495.
McGrath, J. E. (1984). Groups: Interaction and performance. Englewood Cliffs, NJ: Prentice Hall.
Mischel, W. (1968). Personality and assessment. New York: Wiley.
Mullen, B., & Copper, C. (1994). The relation between group cohesiveness and performance: An integration. Psychological Bulletin, 115, 210-227.
Nystrom, H. (1979). Creativity and innovation. London: Wiley.
Orwin, R. G. (1994). Evaluating coding decisions. In H. Cooper & L. V. Hedges (Eds.), The handbook of research synthesis (pp. 139-162). New York: Russell Sage Foundation.
Osborn, A. F. (1957). Applied imagination. New York: Scribner.
Paulus, P. B., & Dzindolet, M. T. (1993). Social influence processes in group brainstorming. Jour- nal of Personality and Social Psychology, 64, 575-586.
Paulus, P. B., Larey, T. S., & Ortega, A. H. (1995). Performance and perceptions of brainstormers in an organizational setting. Basic and Applied Social Psychology, 17, 249-265.
Seashore, S. E. (1954). Group cohesiveness in the industrial work group. Ann Arbor, MI: Institute for Social Research.
Shaw, M. E. (1976). Group dynamics: The psychol- ogy of small group behavior (2nd ed.). New York: McGraw-Hill.
Steiner, I. (1972). Group processes and productivity. San Diego, CA: Academic Press.
Stogdill, R. (1972). Group productivity, drive, and cohesiveness. Organizational Behavior and Hu- man Performance, 8, 26-43.
Tziner, A. (1982). Differential effects of group cohesiveness types: A clarifying overview. Social Behavior and Personality, 10, 227-239.
Woodman, R. W , Sawyer, J. E., & Griffin, R. W. (1993). Toward a theory of organizational creativ- ity. Academy of Management Review, 18, 293-321.
Zaccaro, S., & Lowe, C. (1986). Cohesiveness and performance on an additive task: Evidence for multidimensionality. Journal of Social Psychology, 128, 547-558.
Zaccaro, S., & McCoy, M. C. (1988). The effects of task and interpersonal cohesiveness on perfor- mance of a disjunctive group task. Journal of Applied Social Psychology, 18, 837-851.