The definition of cooperation allows many procedural variations, as revealed by examination of the cooperation literature which includes procedures differing so greatly that it is surprising that all of them are considered under the same topic. This paper attempts to provide a framework for organizing that literature by (1) indicating some procedural dimensions along which cooperation procedures can be classified, (2) classifying and reviewing briefly the research at the extremes of these procedural dimensions, and (3) indicating the behavioral effects that are necessary to demonstrate control by the various cooperation procedures.

The initial objective was to determine whether an increase in cooperative responses (minimal cooperation) was also accompanied by an increase in the degree of correspondence in the number of reinforcers of the two subjects (maximal cooperation). Correct matching-tosample responses of seven pairs of male adolescents were reinforced with money. On each trial, a subject could (1) give the matching-to-sample problem to his coactor (give or cooperative responses), or (2) take the problem for himself (take responses). The first member of the pair to respond made the choice. Correspondence did increase under this procedure as compared to a baseline where problems were distributed randomly. However, the increased cotrespondence usually resulted from take responses rather than cooperative give responses. This equitable method of problem distribution, designated as sharing, was characterized by the subjects alternately taking problems. The spacing of daily sessions may have been partly responsible for the high degree of correspondence, because correspondence did not increase within the usual number of sessions when the sessions were massed, i.e., all in one day. Daily sessions require cooperative responses, i.e., each subject has to show up each day for the other to earn money, and this dependency upon the coactor's behavior may facilitate some sharing or cooperation to ensure the coactor's attendance.A cooperation procedure may be defined as one in which the reinforcers of each member of a pair are at least partly dependent upon the responses of the other member (Hake and Vukelich, 1972). The cooperation effect is then defined by increases in these responses by both members of the pair. Hake and Vukelich (1972) suggested a second cooperation effect. While increases in cooperative responses are indicative of control by the reinforcer resulting from the cooperation procedure, control by the reciprocal nature of the cooperation procedure also requires equality or at least an increase in the degree of correspondence between the numbers of reinforcers or cooperative responses of the members of the pair. If cooperation is to be considered a social behav-

Human subjects, mostly between 11 and 16 yr old, matched to sample for points that were exchangeable for money. An audit response was defined as a response maintained by allowing a subject access to an existing score on his own (self audit) or a coactor's (coactor audit) performance. In Experiment I, changes from non-social procedures (no coactor) to social procedures (coactor present) increased self and coactor audits. Since both types of audits occurred at about the same rates during cooperation and parallel work procedures, the increases did not depend on the subjects' response interactions. Although Experiment I did not demonstrate that subjects were comparing scores, the frequent occurrence of each kind of audit within a brief time period (interpersonal audit) did indicate that it was reinforcing to have both scores at the same time. These interpersonal audits suggested that the coactor's score increased self audits during social procedures. Experiment II supported this notion: relative to a non-social procedure, self audits increased more during a parallel work procedure when the coactor's score was accessible than when it was not accessible. Thus, increases in other behaviors that occur in the presence of a coactor, i.e., social facilitation, may also result from or be increased by providing a coactor's score.

The study examined the effects of the availability of a non-cooperative response on cooperative responding when cooperation did not have to result in an equal distribution of work or reinforcers. Also, an attempt was made to determine if the cooperative responding was under the control of the cooperation procedure. Pairs of institutionalized retardates were tested in full view of each other. For each subject, reinforcers (money) were contingent upon responses on each of two panels: (1) a matching panel for working matching-tosample problems, and (2) a sample panel for producing the sample stimulus. The matching panels of the two subjects were 6 m apart, but a subject's sample panel could be placed at different distances from his matching panel. For each subject, either his own or his partner's sample panel could be nearest his matching panel such that less walking was required to reach one sample panel than the other. Subjects could work either individually, by producing their own sample stimulus, or cooperatively, by producing the sample stimulus for their partner. Subjects selected whichever solution involved the least amount of walking. The importance of testing for control by the cooperation procedure was indicated by the findings that cooperative-like responses were not always under the control of the cooperation procedure.The essential aspects of a cooperation procedure are (1) that the reinforcers of both individuals are at least in part dependent upon the responses of the other individual and (2) that the procedure allows such responses, designated as cooperative responses, to result in an equitable division of responses and reinforcers (Hake and Vukelich, 1972). The cooperation literature illustrates the range of variations possible on these basic procedural requirements.A recent review of the cooperation literature classified cooperation procedures at the extremes of four procedural dimensions (Hake and Vukelich, 1972). The procedures at the two extremes of three of the dimensions may

An audit response allows access to an existing score from a subject's own performance (self audit) or from his coactor's performance (coactor audit), A previous study found that social stimuli (coactor present) increased audits relative to a non-social (no coactor) condition. The increase, designated a social-stimulus effect, was found to be due more to the coactor's score than to his mere presence. This finding suggested that the difference between self and coactor scores might affect the size of the social-stimulus effect. In the present study, six pairs of human subjects matched-to-sample for points that were exchangeable for money. During a session, matching-to-sample problems were distributed so that a subject's score was ahead, behind, or about even with his coactor's score. The even condition produced the largest social-stimulus effects, i.e., the most audits that could not be attributed to nonsocial variables such as time or number of problems. The even condition may have produced the largest social-stimulus effects because it was the only condition where the major social reinforcer (being ahead) could be both present or absent and, consequently, the even condition was the only one where audits had a discriminative function with respect to the presence of the major social reinforcer.