Some of the earliest and most basic ideas about groups in sociology concern group size. Cooley (1922) described how people universally are members of primary groups, which are small in size, face-to-face, highly intimate, cooperative, and enduring. Prototypes include families and groups of friends. Although people also participate in secondary groups, which are typically larger, less personal, more formally organized, and more limited in duration and purpose, primary group participation appears necessary for teaching children the requisites for participating in society, including language, basic social skills, values, and identities. Simmel (1950) described the profound effects on interpersonal relations when the smallest group, the dyad (pair), expands to a triad. The dyad is unique because it can be destroyed by the loss of a single member, and this feature often leads to a degree of intimacy and closeness not found in larger groups. Adding another person, a third party, to form a triad dramatically transforms the character of the relations. Simmel noted three different roles served by a third party. One role is as a nonpartisan or mediator, which can serve to draw the members of the pair closer together. For example, the birth of a child may enhance the emotional bond between the parents, or a third person may try to settle a dispute between two friends. Another role is as the tertius gaudens ("the third party who enjoys"). The third party benefits when two members in conflict try to win the third party's support. For example, parents who are in conflict may compete to win the affection of their child by using gifts and favors. A third role is to divide and conquer. Here the third party benefits by actively encouraging conflict in the pair. For example, a parent may gain greater control over two children by having them compete for the parent's favors and affection. Simmel thus drew attention to the fact that the triad (and larger groups) brings into play processes that are impossible in the dyad. A systematic account of the effects of group size, however, awaited the emergence of an experimental research tradition.
In sociology the term group has been used very loosely, referring at one extreme to small aggregations of people whose members are mutually aware of each other and can potentially interact (McGrath 1984) and at the other extreme to enormous aggregations whose members simply share some characteristic, as in the case of an "ethnic group." In studies of group processes, the more restricted definition has been used. The groups studied have generally been created explicitly for the purpose of research, rather than being ones that existed naturally for other purposes.
Group size has been studied in groups facing two very different circumstances. The first concerns groups that are formed to achieve some common purpose or goal. Interest has focused on how people behave in such groups and conditions that enhance or impede their effectiveness in achieving the goal. The second concerns groups facing collective dilemmas, where each member chooses between two actions with contrasting consequences. One action maximizes each member's own interest, whereas the other action maximizes collective interests, that is, the interests of the group as a whole. The dilemma is that each member profits more from the selfish choice, but if all members make that selfish choice each is worse off than if all chose to maximize collective interests (Komorita and Parks 1994).
As Steiner (1972) notes, purposes for which groups are formed can vary on a continuum from task performance at one extreme (e.g., solving a problem, creating some product) to sociability at the other (e.g., simply enjoying one another's company). Groups whose purpose is primarily sociability have been little studied. Assuming that sociability is defined by spontaneous, reciprocal conversation, enjoyment should be highest in groups with no more than five or six members, the largest group size in which this kind of interaction occurs easily. This rule may be qualified in larger groups, however, if the situation is arranged so that people interact in smaller subgroups.
Most experimental research has focused on groups that are more task oriented. Here the most consequential feature is the nature of the task, or task structure. The first major program studying task-oriented groups experimentally was conducted by Bales and associates (Bales 1950). Bales and other early researchers focused on groups attempting to solve decision or judgmental tasks, problems for which there is no correct answer and where differences of opinion are likely regarding the resolution (e.g., a personnel issue facing an administrator in an organization). Groups were told to discuss the assigned problem for a few minutes and then arrive at a group recommendation. Bales's method of interaction process analysis (IPA) coded verbal and nonverbal communication among group members as they addressed two basic ongoing concerns: the task and the relationships among the members. Research by Bales and associates showed the feasibility of measuring ongoing behavior in discussion groups and was instrumental in the emergence of small-group research as a distinctive field in the social sciences.
Group size is a fundamental consideration in discussion groups because the number of possible symmetrical relations between pairs of members increases much more rapidly than the number of members added to the group. The formula is where x=the number of symmetrical relationships and n=number of members (Bossard 1945). In addition to the pair relations, there are also relations between each group member and the group as a whole. If group members interact for a fixed amount of time, adding group members forces changes in the nature of existing relations and limits the number and nature of new relations (e.g., time available with each member, degree to which conversation is reciprocal).
As group size increases, groups tend to adopt a more direct and organized approach in soliciting information, and a task leader is more likely to emerge or be chosen. The top communicator initiates a greater proportion of the group's communications, and the differences in amount of communication by other members diminish (Bales et al. 1951). An increasing proportion of communication is directed toward the group as a whole, and the number of members who participate minimally by simply listening and giving emotional expressions increases (Bales and Borgatta 1955). In addition, the larger the group, the greater the conflict, the less likely members are to reach agreement on controversial issues, the less the conformity to group norms is, and the less the members are satisfied with the group and its activities (Levine and Moreland 1998; Thomas and Fink 1961).
Latané and L'Herrou (1996) suggest that group size is important in determining whether or not subgroups holding a minority opinion are likely to form and persist over time. As the number of members grows larger, the larger the number of members who may hold minority views, and hence the greater the opportunity for them to interact with each other and avoid exposure to majority views. The emergence of minority subgroups in groups is also affected by whether the group has an odd or even number of members. When opinion is divided, odd-numbered groups (e.g. 3, 5, or 7 members) are more likely to break into majorities and minorities, avoiding the possible even splits of even-numbered groups. With this pattern, oddnumbered groups should have fewer deadlocks and should reach decisions more quickly than even-numbered ones (Bales and Borgatta 1955).
The ideal size of discussion groups is often considered to be four to six members, a number large enough to ensure some diversity in member resources but small enough so that everyone can participate. Groups naturally occurring in public are rarely larger that five or six people, and more often contain two or three (Moreland et al. 1996).
Juries are discussion groups entrusted by the state to make decisions of the utmost consequence for the accused. Because legislation now permits juries smaller than twelve members, jury size has emerged as a variable of interest in studies of actual and mock (experimenter-created) juries. Although twelve-person juries should be less likely to reach unanimous verdicts (or verdicts based on a two-thirds majority) than smaller juries (e.g., six or eight members), the predicted difference is small and it has not been found in the studies themselves (Davis 1989).
Group size affects performance for other types of tasks as well. Steiner (1972) describes the effects for tasks having products or solutions whose adequacy can be judged objectively. Tasks are unitary when they cannot be readily divided into subtasks and where mutual assistance is impractical (e.g., working one arithmetic problem, hammering a nail). Tasks are divisible when they can be divided into subtasks, allowing different group members to work on the parts before the final product is assembled (e.g., working ten arithmetic problems, assembly-line production). Steiner identified several types of unitary tasks. A task is disjunctive when any single member can supply the group's product (e.g., the answer to a question). Here potential group productivity depends on the resources of the most competent member. The larger the group, the greater the likelihood that at least one member will have the needed resources. As the group grows larger, however, each additional person adds a smaller increment to the group's potential. Productivity gains are thus a decelerating function of group size. A task is conjunctive when the resources of the least competent member determine group productivity (e.g., mountain climbers trying to reach the summit while roped together). Here the larger the group, the greater the likelihood that at least one member will be low in competence. As the group grows larger, productivity decreases, with each additional person adding a smaller decrement to the group's potential. Productivity losses are thus a decelerating function of group size. A task is additive when individual resources are added to obtain the group product (e.g., stacking firewood, offering ideas to improve a product). Here, the larger the group, the greater the group productivity.
For divisible tasks, variation can occur both in the nature of the subtasks and in the way that subtasks are combined to form the final product. A subtask can be disjunctive, conjunctive, or additive if two or more group members are working on it, and the process of assembling the subtasks can be disjunctive, conjunctive, or additive, thus creating a large number of possibilities. Where the rule for combining subtasks is additive or disjunctive, group productivity should be a positive, decelerating function of group size. The larger the group, the more likely it is that the subgroups can be staffed optimally, but communication, coordination, and matching members to the subtasks most appropriate for their interests and skills can become increasing problems. An increase in group size has the opposite effect, however, where the rule for combining subtasks is conjunctive, For tasks where the labor is divided and it is critical that each member make the correct response (e.g., an airline crew, a team of surgeons), the larger the group, the greater the chance that one of the critical members (or subgroups) will fail to perform adequately, and hence the group will perform poorly.
Steiner notes that actual task productivity often fails to reach potential productivity because groups use faulty or limiting processes to address the task and may also fail to motivate members sufficiently. Both problems typically increase with group size. With divisible tasks, for example, assignments given to members may not provide a good match between member skills and the subtasks they are to perform. In face-to-face groups where each member contributes orally (e.g., giving novel ideas on some topic), larger groups derive decreasing benefits as group size increases simply because only one person can talk at a time. The best-studied example of motivational losses involves additive tasks where individual performances are simultaneous and anonymous. For example, where group members each make some response on cue (e.g., pulling on a rope, shouting), individual contributions are smaller the larger the group, an effect that is termed "social loafing."
In addition to a task structure, most groups also have a reward structure—the arrangement of payoffs or rewards that motivate members to work on the task. The type and nature of the reward structure may be imposed by a third party (e.g., a supervisor, leader), be part of the group's history, or be chosen by the group itself. In most experimental studies of task-oriented groups, the reward structure is cooperative: The payoff or reward is a result of the group's efforts in meeting some criterion and is shared by all members (although not necessarily equally). The payoff may be intangible (e.g., the satisfaction of solving the problem or completing the product) or tangible (e.g., money or prizes). Reward structures can also be competitive, where rewards are distributed unequally to members based on relative individual performance. Cooperative and competitive contingencies are often compared with a third alternative, individual contingency, where a member receives a reward when he or she meets an individual performance criterion; this is the case in most work-for-pay groups in industry.
These reward structures are not equally appropriate for all tasks. In particular, cooperative rewards are effective across a range of tasks (Johnson et al. 1981; Qin-Zhining et al. 1995) and are uniquely appropriate when the task requires collaborative activities such as response coordination, task subdivision, or information sharing. Collaboration is rewarded with cooperative rewards, increasing the likelihood that the criterion for the group reward will be met. By contrast, competitive rewards are effective only where task responses can be made independently by each person. With competitive rewards, blocking another's responses, not collaboration, is likely to lead to winning. When competition is appropriate, though, it is often more cost-effective (more responses made per unit of reward), easier to implement, and capable of producing short-term performance rates that are higher than those of the other conditions (Schmitt 1987). Competitive motivation can also arise outside of the formal reward structure. Under cooperative conditions, members may work harder simply to be the best performer in the group.
A group member's motivation to perform a task can vary greatly depending on reward structure, and group size affects important aspects of that structure. With a cooperative structure, all members are rewarded, but inequities may exist in the size of the rewards received (e.g., some people contributing to a task may get more money than others). In general, people working on tasks expect their rewards to be proportional to their contributions (Homans 1974; Walster et al. 1978). Thus, if person A and person B have similar task skills, A will be upset if A and B contributed equally to the task but A received a reward half the size of B's. Person A will not be upset, however, if A made a contribution half the size of B's to the task. People who are inequitably underpaid relative to their contributions often work less hard on future tasks and may choose to leave the group (Marwell and Schmitt 1975). In the 1990s professional sports revealed a number of cases where athletes earning millions of dollars per year refused to play for their teams (or played less energetically) because comparable performers on their own or other teams earned more. In cases where the total amount of the cooperative reward for completing a task is proportional to the number of group members (e.g., $50 for a five-member group and $100 for a ten-member group), the larger the group, the greater the potential for larger inequities, that is, several of the members receiving a large share of the reward. If large inequities are present, productivity gains are likely to be a decelerating function of group size (and member discontent may provoke a change in distribution).
In arranging for competition, the reward structure is defined by the unequal rewards distributed to winners and losers at the end of the contest. In most contests the distribution is fixed in advance and is known to the competitors. As with inequitable cooperative structures, the variety of competitive structures depends on the number of competitors in the group, assuming that the total contest amount is proportional to the number of group members (i.e., the larger the group, the larger the contest amount). Two properties of the distribution are relevant. One is the proportion of competitors receiving rewards in each contest. At one extreme, only one competitor receives a reward; at the other extreme, all competitors are rewarded, but in varying amounts. With a single winner, the larger the size of the group, the larger the competitive reward. As has been shown with large lottery prizes, the larger the competitive reward, the greater the motivation of group members to compete, at least in a single contest. However, if there is a series of contests and a difference in competitive skills causes some members to lose continually, the lack of earnings will lead to their withdraw from the group, thus lowering group productivity (Schmitt 1998). Maximizing the proportion of competitors rewarded should encourage poorer performers to remain in contests regardless of group size. When more than one competitor is rewarded, variation can occur in a second property—reward differential or spread, that is, the difference between the highest and lowest reward amounts in each contest. Again, the larger the size of the group, the larger the total competitive reward, hence the larger the possible differential. Maximizing reward differential more highly motivates those who have a chance of winning but gives those who lose frequently less incentive to continue over a series of contests. In sum, group size is a factor when competitive reward structures specify that few are rewarded or have reward differentials that are extreme. Although increasing group size (and the reward pool) is likely to increase member motivation and productivity in initial contests, over a series of contests these gains are likely to be a decelerating function of group size, as those who earn little contribute less or quit.
With individual reward structures, the potential effects of group size are similar to those for cooperative structures. Again assuming that the total reward amount for the group is proportional to the number of members, the larger the group, the greater the size of the reward inequities that are possible in arranging the individual rewards. With large inequities, productivity gains are likely to be a decelerating function of group size.
Several types of social dilemmas have been investigated. The best-known type is prisoners' dilemma. Originally conceived as a two-person game, an n-party prisoners' dilemma in which the number playing the game can be varied has been used to study the effects of group size (Komorita and Parks 1994). Each group member has two choices: to cooperate (C), which maximizes payoffs for the group as a whole, or to defect (D), which maximizes the individual's own payoff. The actual payoffs for each member depend both on own and others' choices. Payoffs for each choice increase with the proportion of members who make the C choice, but the D always produces the higher individual payoff. Finally, the payoff if everyone chooses C is greater than the payoff if everyone chooses D, the selfish choice. Each group member chooses repeatedly over a number of trials. This basic dilemma has numerous everyday counterparts, as when commuters each prefer to use a private automobile instead of a bus, but if each does so the resulting chaos leaves everyone dissatisfied.
Other types of social dilemmas involve a pool of sources to which the group members have access (Pruitt 1998). One is the commons dilemma, based on the Tragedy of the Commons (Hardin 1968), in which a village's common land is overgrazed because of the selfish actions of the individual herdsmen. As studied experimentally, subjects take turns removing resources (e.g., money) from a pool that is replenished periodically based on the amount remaining. The pool can be productive indefinitely, provided that the subjects don't destroy it by taking all the resources. Another is the public goods dilemma, in which the resource pool is built up through individual contributions, as when people contribute money to support public television or some charity. Here the temptation is for individuals to "free ride" and let others make the contribution. As studied experimentally, subjects take turns contributing resources to a pool that is later enhanced by the experimenter and then divided equally.
Studies have found that the larger the group in social dilemmas, the less the cooperation and the greater the selfish behavior, although there is little change in groups larger than eight members (Pruitt 1998). Various explanations of this relation have been proposed (Komorita and Parks 1994; Pruitt 1998). One possibility is that any defection breaches the trust required for cooperation, and that if one person defects, others view cooperation as unlikely and follow suit. The possibility of one person defecting increases as the group grows larger. In addition, as the group grows larger, each selfish individual response becomes less identifiable and may be seen as less responsible for the lower group payoffs. By the same logic, a cooperative response may also be seen as contributing less to the group product as the group grows larger. Finally, any opportunities by the group to communicate the group's cooperative interests or to sanction defectors are more difficult to carry out in larger groups.
Any social interaction among group members must take place in some context, for example, people standing, seated at a table, at stations on an assembly line, or in separate offices linked by a computer network. Physical contexts are very important because they affect both the nature of the interaction among members and the ease with which various members can interact. The nature of interaction among group members is determined in part by the number and kind of stimuli presented by those who interact. Stimuli may be verbal in the form of oral or written communication, or they may be nonverbal—information transmitted without using language (e.g., facial expressions, gestures, posture, appearance, voice quality, rate and temporal patterning of speaking).
The context in which a group functions determines whether verbal and nonverbal stimuli are transmitted among group members. If the context includes face-to-face contact, speakers can deliver messages using both verbal and nonverbal stimuli. Nonverbal stimuli add elements that can be both enriching and distracting. For example, a speaker's facial expressions and gestures can reveal the strength with which a position is held, the truthfulness of a message, or the degree of intimacy sought by the speaker. Aspects of the speaker's demeanor or appearance can also distract listeners from paying close attention to the message. People differ, however, in how they use such stimuli and hence the meaning given to a message. This variability is greatly reduced if the messages are written. In addition, the quality of interaction may differ, depending on the presence of the verbal and nonverbal stimuli. For example, people are more likely to harm others (e.g., deliver punishment or bad news) when communication is restricted and nonverbal behavior is absent compared with face-to-face contact (where people are more fully personalized). In general, restricted communication does not lead to the full development of interpersonal relations, and it minimizes cues that reveal differences in status, power, and prestige among group members. Hence, differences in interaction and influence among group members tend to be less extreme than those in face-to-face groups (McGrath, 1984). Restricted communication is more likely as groups become large and more formally organized.
Another aspect of context concerns the timing of information transmitted among group members. When people interact, verbal and nonverbal stimuli may be presented synchronously (i.e., immediately and in real time), or asynchronously (i.e., at the member's own time, place, and pace). Face-to-face interaction, telephone conversations, and video conferences are synchronous. Letters, memos, videotapes, electronic mail, and answering machine messages are asynchronous. Synchronous interaction, whether in person or electronically mediated, constrains interaction because only one person normally talks at a time. With an increase in group size, a few members typically lead or dominate the conversation. Asynchronous interaction, by contrast, typically permits any number of messages to any or all members, and the fact that the messages are necessarily recorded means that they are normally available for comparison and review later.
Asynchronous communication has increased enormously in popularity in recent years as computer-mediated electronic mail has become available in all organizations and many households. Computer-mediated communication has unique facilitating features: the ability to link anyone who has a network connection and the opportunity to send messages instantaneously to any number of people at very low logistical and social cost. For some tasks, the use of asychronous communication changes the effects of group size on productivity, compared with traditional face-to-face synchronous communication. For example, Valacich and associates (1995) investigated groups formed to generate new ideas, that is, to "brainstorm," where different group members have the potential for generating different ideas. In face-to-face interaction, larger groups derived decreasing benefits as group size increased from five to ten members. In computer-mediated groups, where members communicated via typed messages, larger groups derived increasing benefits as group size increased. Computer-mediated groups appear to use the advantage of reviewing the ideas of others to avoid redundant ideas and build new ones. The study of computer-mediated groups with other tasks promises to reveal further distinctions between synchronous and asynchronous interaction (see Kiesler 1997 for discussion of a number of computer- and Internet-related issues).
An intriguing and potentially distinguishing feature of computer-mediated compared with face-to-face groups is that group size itself may become difficult to define and detect (Sproull and Faraj 1997). People in face-to-face groups take up space, and their physical presence and nonverbal reactions can affect others' behavior even if a group member says nothing. By contrast, the readers of computer-mediated messages are typically invisible, and speakers may have little notion of who has seen or read a message.
In both industry and education, computers have been used to create humanlike partners who interact with people in various ways. Kiesler and associates (1996) investigated subjects' responses in a prisoners' dilemma game where the "partner" was known by the subject to be either a person or computer based. In both conditions, the subject and partner discussed options on each trial, and the partner asked the subject for commitments. For both types of partners, discussion and agreements with the subjects facilitated cooperation, although the effect was stronger with the human partner. Thus, nonhuman partners can produce "social" responses in people provided that the partners make humanlike use of social stimuli and responses.
Much of the early research on context concerned constraints on who could communicate with whom in small groups. These constraints determine the group's communication network (for a summary, see Shaw 1981). For example, two contrasting networks in groups of three or more members are the circle, where each member can communicate with two adjacent members, and the wheel, where one member occupies a central position and others can communicate only with that central person (the "hub"). Wheels and circles are examples of highly centralized and decentralized networks, respectively. Groups studied in such networks have typically been small (e.g., three to five members), and communication has usually occurred via written notes. Comparisons of group problem solving using various networks have found that when tasks are relatively simple and require that members collect their information, centralized networks are more efficient than decentralized ones (with fewer messages and errors). Where tasks are more complex and require that members perform additional operations on their information, decentralized networks are more efficient than centralized ones (Shaw 1981). With complex tasks the communication overload (termed "saturation") experienced by the member in the central position slows the attainment of a solution. Saturation is likely to be a problem in all networks as the group increases in size, and it should emerge more rapidly when interaction is synchronous.
Communication networks arrange the contacts among group members in a decisive manner, but even where members are in face-to-face contact, aspects of the setting can make interaction between some members more likely than others, thus producing networklike effects. Seating arrangement is an example. Groups discussing problems are frequently seated at tables. One of the earliest findings was that people tend to communicate with others across the table and facing them instead of with those seated alongside them (Steinzor 1950). As groups grow larger, people are frequently seated at rectangular tables. Studies have found that people at the table's end positions tend to participate more, are seen a having more influence, and are more likely to be chosen as leaders than those on the sides (for a summary, see Shaw 1981). Networklike arrangements can also be created if group members are instructed to follow a particular communication pattern in a discussion (e.g., "Talk only with the leader."). Instructions can also be used to limit style of expression (e.g., "Give ideas for solving the problem but don't criticize others' ideas.").
In conclusion, differences in group size pose both opportunities and problems for the members. Because the nature of the consequences depends on the type of group task, reward structure, context, and the skills of the group members, little can be said about a group if only its size is known. Under certain circumstances, however, group size is highly consequential for the group's performance and stability. It should be noted that for groups that form in everyday life, unlike those created for experimental purposes, problems posed by size do not necessarily condemn a group to subpar performance. Everyday groups usually have both pasts and futures, and are often skillful in identifying problems and in making changes that help mitigate them.
Bales, R. F. 1950 Interaction Process Analysis: A Method forthe Study of Small Groups. Cambridge, Mass.: Addison-Wesley.
Bales, R. F., and E. F. Borgatta 1955 "Size of Group as a Factor in the Interaction Profile." In A. P. Hare, E. F. Borgatta, and R. F. Bales, eds., Small Groups: Studiesin Social Interaction. New York: Knopf.
Bales, R. F., F. L. Strodtbeck, T. M. Mills, and M. E. Rosenborough 1951 "Channels of Communication in Small Groups." American Sociological Review 16:461–468.
Bossard, J. J. S. 1945 "The Law of Family Interaction." American Journal of Sociology 50:292–294.
Cooley, C. H. 1922 Human Nature and the Social Order. New York: Chas. Scribner's Sons.
Davis, J. 1989 "Psychology and the Law: The Last Fifteen Years." Journal of Applied Social Psychology 19:199–230.
Hardin, G. 1968 "The Tragedy of the Commons." Science 162:1243–1248.
Homans, G. 1974 Social Behavior: Its Elementary Forms, 2nd ed. New York: Harcourt Brace Jovanovich.
Johnson, D. W., G. Maruyama, R. Johnson, D. Nelson, and L. Skon 1981 "Effects of Cooperative, Competitive, and Individualistic Goal Structures on Achievement: A Meta-Analysis." Psychological Bulletin 89:47–62.
Kiesler, S. (ed.) 1997 Culture of the Internet. Mahwah, N.J.: Erlbaum.
——, L. Sproull, and K. Waters 1996 "A Prisoner's Dilemma Experiment on Cooperation with People and Human-Like Computers." Journal of Personalityand Social Psychology 70:47–65.
Komorita, S. S., and C. D. Parks 1994 Social Dilemmas. Dubuque, Iowa: Brown and Benchmark.
Latané, B., and T. L'Herrou 1996 "Spatial Clustering in the Conformity Game: Dynamic Social Impact in Electronic Groups." Journal of Personality and SocialPsychology 70:1218–1230.
Levine, J. M., and R. L. Moreland 1998 "Small Groups." In D. T. Gilbert, S. T. Fiske, and G. Lindsley, eds., TheHandbook of Social Psychology, vol. II, 4th Ed. Boston: McGraw-Hill.
Marwell, G., and D. R. Schmitt 1975 Cooperation: AnExperimental Analysis. New York: Academic Press.
McGrath, J. E. 1984 Groups: Interaction and Performance. Englewood Cliff, N.J.: Prentice-Hall.
Moreland, R. L., J. M. Levine, and M. L. Wingert 1996 "Creating the Ideal Group: Composition Effects at Work." In E. H. Witte and J. H. Davis, eds., Understanding Group Behavior, vol. II. Mahwah, N.J.: Erlbaum.
Pruitt, D. G. 1998 "Social Conflict." In D. T. Gilbert, S. T. Fiske, and G. Lindsley, eds., The Handbook of SocialPsychology, vol. II, 4th ed. Boston: Mcgraw-Hill.
Qin-Zhining, D. W. Johnson, and R. T. Johnson 1995 "Cooperative Versus Competitive Efforts and Problem Solving." Review of Educational Research 65:129–143.
Schmitt, D. R. 1987 "Interpersonal Contingencies: Performance Differences and Cost-Effectiveness." Journal of the Experimental Analysis of Behavior 48:221–234.
Schmitt, D. R. 1998 "Effects of Reward Distribution and Performance Feedback on Competitive Responding." Journal of the Experimental Analysis of Behavior 69:263–273.
Shaw, M. E. 1981 Group Dynamics, 3rd ed. New York: McGraw-Hill.
Simmel, G. 1950 The Sociology of George Simmel; Translated, edited, and with an introduction by Kurt H. Wolff. Glencoe, Ill.: Free Press.
Sproull, L., and S. Faraj 1997 "Atheism, Sex, and Databases, The Net as a Social Technology." In S. Kiesler, ed., Culture of the Internet. Mahwah, N.J.: Erlbaum.
Steiner, I. D. 1972 Group Process and Productivity. New York: Academic Press.
Steinzor, B. 1950 "The Spatial Factor in Face-to-Face Discussion Groups." Journal of Abnormal and AppliedPsychology 45:552–555.
Thomas, E. J., and C. F. Fink 1961 "Effects of Group Size." Psychological Bulletin 60:371–385.
Valacich, J. S., B. C. Wheeler, B. E. Mennecke, and R. Wachter 1995 "The Effects of Numerical and Logical Group Size on Computer-Mediated Idea Generation." Organizational Behavior and Human DecisionProcesses 62:318–329.
Walster, E., G. W. Walster, and E. Berscheid 1978 Equity: Theory and Research. Boston: Allyn and Bacon.
David R. Schmitt