Interpersonal and Group Contingencies

Abstract

Responding by individuals in groups has been a prominent interest of the field of psychology. Experimental analyses of human behavior have provided some unique findings of the role that the environment, including both social and nonsocial stimuli, may have on individual responding. Cooperative and competitive contingencies, previously evaluated in animal and human operant research, provide unique insight into applied interventions, in particular group contingencies. The current manuscript attempts to bridge these two literatures to foster the development of more effective technologies and lines of experimental or translational research that may better inform interventions in the applied realm.

The behavior of individuals in groups has been a prominent focus of numerous fields and includes some of the most popular findings in psychology (e.g., Asch, 1956; Fischer et al., 2011; Haney, Banks, & Zimbardo, 1973). Behavior analysis has not devoted similar attention to patterns of responding in the presence of complex social stimuli, despite behavior analytic technologies being applied in settings in which social stimuli are ubiquitous. The unique control that may occur under complex social contingencies may best be observed during group contingency interventions, which are commonly arranged in educational and contingency management programs (Dallery, Meredith, & Glenn, 2008; Meredith, Grabinski, & Dallery, 2011). Despite their frequent use, group contingency interventions may often be insufficiently described, although effective in their application (Maggin, Johnson, Chafouleas, Ruberto, & Berggren, 2012). Because of the ubiquity of complex social stimuli and the need for interventions to be applied in group settings, behavior analysts would do well to pursue lines of research in these areas. Previous experimental research has evaluated numerous variables that may influence responding in pairs or small groups of participants, often seeking to evaluate arrangements in which cooperative or competitive responding occurs or is disrupted. Additional research on applied interventions in social contexts, such as group contingencies, have evaluated similar arrangements, although the connections between these literatures have yet to be drawn. The current article describes experimental literature on social behavior, in particular describing competitive and cooperative contingencies largely conducted from the late 1960s to the early 2000s. The methods and findings of this research are considered for their implications for applied research and interventions in group contingencies. In addition, deficiencies in applied research are highlighted as a call for greater experimental research to be conducted in an effort to bridge these two literatures and extend the functional approach to social behavior.

Social Behavior

Schmitt (1998a) described behavior as social “. . . when its causes or effects include the behavior of others” (p. 471). Skinner (1953) likewise defined social behavior as “. . . the behavior of two or more people with respect to one another or in concert with respect to a common environment” (p. 297). Given the definitions by Schmitt (1998a) and Skinner (1953), social behavior constitutes a reciprocal interchange in which the antecedents or consequences of some response include the behavior of others, consistent with the definition of a social episode by Baum (2017).

When considering the behavior of individuals in a social episode, analyzing a single member’s behavior is consistent with behavior-analytic methodologies (Hineline, 1980; Skinner, 1953; Todorov, 2009). The differential effects of various social and nonsocial stimuli on a member’s behavior in a social episode should be considered, in particular because these stimuli exert control over responding based on individual history. Joint attention may serve as a prominent example. As a notable deficit in the behavior of individuals with autism spectrum disorder, joint attention includes the use of gesture and eye contact to recruit the attention of another individual following some environmental stimulus (Dube, MacDonald, Mansfield, Holcomb, & Ahearn, 2004). In this way, a stimulus that may be nonsocial (e.g., a cup falling from the table and spilling milk on the floor) evokes a response that includes looking at the cup on the floor, looking at the parent, and looking again at the cup. This response may be referred to as initiating joint attention (Dube et al., 2004; Mundy & Crowson, 1997). Although the initiating response is emitted by one actor, a characteristic response by the coactor is to orient to whatever nonsocial stimulus evoked the initiation by the other. As such, the responding by two members in a dyad may be separately evoked by a nonsocial stimulus (e.g., spilled milk) for the actor and a social stimulus (e.g., the actor’s gaze shift) for the cofactor.

The example above may best be described as a leader–follower arrangement. Leading and following procedures serve as one example of a social episode in which the behavior of the leader and the follower may be under control of different stimuli. In particular, the leader’s behavior is under control of nonsocial stimuli (i.e., stimuli that are not produced by others’ behavior) external to the social episode, whereas the follower’s behavior is controlled by the behavior of the leader (Skinner, 1953). Skinner (1962) described one arrangement in which two pigeons were placed in adjacent compartments and trained to respond to a three-button array in which a single button was operative. Each pigeon had a separate array of buttons and both pigeons pecking the operative button simultaneously produced reinforcement for both. The pigeon serving as the leader reportedly explored the various buttons until the operative key was identified, whereas the follower pecked the corresponding key to the leader. Skinner notes that the role of leader or follower changed between the pair as a function of relative food deprivation with the more deprived bird serving as the leader. Further, Skinner noted persistent coordinated responding of the pair outside of this paradigm in which the birds emitted other coordinated responses such as simultaneous drinking and head-bobbing, suggesting that other responses had come under control of social stimuli without additional training.

In a similar arrangement, Hake, Donaldson, and Hyten (1983) placed two rats in adjacent chambers separated by a clear partition. In the first experiment, a trial began with the two response keys being illuminated in the “leader” rat’s chamber. Once the leader selected the correct key, the keys in the other chamber became illuminated and selecting the corresponding correct key by the “follower” rat produced reinforcement. The authors observed the follower rat respond with high accuracy within twenty 200-trial sessions. Further, when the leader and follower roles were changed, the previously leading rat acquired the follower response in a similar number of trials. The authors then sought to evaluate whether the visual stimulus or some other stimulus (e.g., auditory) controlled responding. The authors introduced white noise and the partition was made opaque. Under these conditions, chance levels of responding were observed. Moreover, when the clear partition was re-introduced, but the auditory stimuli continued to be blocked (Experiment 4), three times as many sessions were required to produce criterion-level responding compared to when both auditory and visual stimuli were present. These findings suggests that correct responding in a leader–follower arrangement had come under control of both visual and auditory stimuli, both of which may be considered social.

Schmitt and Marwell (1968) were similarly interested in coordinated responding under the control of social stimuli in humans. The authors used similar paradigms to Skinner (1962), but included college students. The authors noted shortcomings in prior research by Lindsley (1966) in which coordinated responses by participants may have been coincidental as a result of the arrangement employed. In particular, the participants’ responding may not have come under control of their partner’s responding, but instead participants may have emitted such high rates of responding that close temporal proximity of responses was inevitable (which may have also been the case in the methods described by Skinner [1962]). To elucidate control by social stimuli, Schmitt and Marwell (1968) placed two participants in separate rooms with coordinated plunger pulls producing reinforcement. The first plunger-pull by either participant illuminated a light on the other participant’s panel for 3 s. After 3 s, the light was no longer illuminated and pulling the plunger within 0.5 s produced reinforcement. Reinforcement was followed by a signaled 3-s timeout interval. The authors found that coordinated responses were a function of a combination of timeout lights and response lights with moderate to high levels of cooperative responding being observed even when only response light was presented. In contrast, when using the procedures described by Lindsley (1966), participants’ cooperative responding still occurred at moderate to high levels during all conditions, suggesting that responding was not controlled by the responses of the coactor.

The above examples by Skinner (1962), Hake et al. (1983), and Schmitt and Marwell (1968) all include contingencies that establish coordinated responding in pigeons, rats, and humans, respectively. These leader–follower arrangements are a unique type of social episode as the stimuli controlling each members’ responding differed. Nevertheless, the dynamic features of these arrangements suggest that the role of leader and follower may shift within a dyad (Skinner, 1962). In addition, each study included the equitable arrangement of response and reinforcement among group members, which may have sustained responding by individual members. These arrangements are consistent with research on cooperative contingencies.

Cooperation

The definition of cooperation may differ based on theoretical orientation (Marwell & Schmitt, 1975), although prominent definitions contain five elements: goal-directed behavior, rewards for each participant, distribution of responses among the participants, and coordination of responses, which must be social (Marwell & Schmitt, 1971; Marwell & Schmitt, 1975). Of these, goal-directed behavior and rewards for each participant are identified as necessary in defining cooperation with the other elements clarifying specific relations within a cooperative contingency (Marwell & Schmitt, 1975). Thus, cooperative contingencies may be defined as those in which responding by members produces a reinforcer for all members of the group (Hake & Vukelich, 1972; Schmitt, 1998a). Hake and Vukelich (1972) suggest that the frequency of cooperative responses and reinforcers should be equitable over time. That is, although periods of inequity may occur, cooperative contingencies are associated with the possibility of eventual equity. One method adopted by those interested in assessing responding under cooperative contingencies is a forced response-sharing procedure (Hake & Vukelich, 1972), which operates on a single schedule of reinforcement, identical to the leader–follower arrangements described above. These procedures may also be referred to as coordinated response arrangements, as pairs of participants must emit responses that produce reinforcement for both (Schmitt, 1998a). Using this paradigm, variables that may affect cooperation, such as periods of inequity, can be evaluated. Take, for example, one of the earliest studies on cooperation by Azrin and Lindsley (1956) in which pairs of children received a single jelly-bean following coordinated responses. The frequency of cooperative responses increased as a function of reinforcement with the jelly-beans being divided among the pair in different ways. In two pairs, a single member took all jelly-beans until the partner stopped emitting coordinated responses. The authors reported that periods without coordinated responding were followed by a verbal agreement between partners resulting in an increase in cooperative responding and equal division of the reinforcer. The disruptive effects of unequal reinforcement on coordinated responding is easily extended to natural examples; however, natural settings may commonly include alternative methods to obtain the reinforcer, more consistent with concurrent, rather than single, schedules.

Schmitt and Marwell (1972) examined cooperative responding in dyads under concurrent schedules. The authors were interested in whether participant would exhibit “irrational” behavior in response to reinforcer inequity between members. To do so, a greater magnitude of reinforcement was arranged for cooperative responding relative to the alternative independent responding. Once cooperative responding was consistently emitted by the participants, the authors assigned the pairs of participants to one of three inequitable reinforcement conditions: small, moderate, and large reinforcer inequity. In these conditions, cooperative responding by one member produced two, three, and five times the amount of reinforcement relative to the other member. The authors observed a greater proportion of time allocated to individual responding as the size of inequity increased. Nevertheless, even in the large reinforcer inequity condition, cooperative responding was only disrupted in 4 of 10 dyads. This finding suggests that some participants, but not all, allocated responding to individual arrangements, associated with the lowest magnitude of reinforcement, as a result of inequity.

Shimoff and Matthews (1975) used an arrangement similar to Schmitt and Marwell (1972), but included a confederate in each dyad. The authors describe individual responding by the participant across a number of conditions. Similar to Schmitt and Marwell (1972), inequitable reinforcement had different effects on cooperative responding. That is, for one participant, a clear preference for individual responding under reinforcer inequity was observed. Two participants demonstrated a similar preference, but only after the confederate was given the opportunity to share the reinforcer in the inequitable condition, but did not. For another participant, cooperative responding persisted across all conditions. These findings suggest that reinforcer inequity may disrupt cooperative responding, but other variables, such as opportunity to restore equity, are also implicated.

The opportunity to restore equity has been evaluated using two response options: giving (also referred to as sharing) or taking. Opportunities to give or take reinforcers within a dyad produces distinct results (Schmitt & Marwell, 1972; Shimoff & Matthews, 1975). In previously noncooperative dyads, a giving response by the advantaged partner produced durable increases in cooperative responding (Schmitt & Marwell, 1972). Although giving entails the receipt of reinforcement by one member at the cost of the other, taking procedures produce reinforcement for the member emitting the response and may be emitted by the advantaged or disadvantaged participant. Indeed, Schmitt and Marwell (1972) reported 7 of 10 advantaged participants emitted a taking response. Nevertheless, taking responses were emitted by both members until the advantaged participant ceased taking, effectively reducing the inequitable reinforcement resulting from cooperative responding. Although these responses may restore cooperative responding under inequity, taking has been shown to disrupt cooperative responding under equitable conditions (Schmitt & Marwell, 1971a, 1971b). Taking may be particularly detrimental if participants are able to emit the response and quickly switch to the individual responding alternative, effectively blocking a taking response by the coactor. Under these conditions, taking responses resulted in participants refusing to continue to participate in the study (Schmitt & Marwell, 1971a). In a rendition of the taking response, Marwell and Schmitt (1975) included a response that only removed a reinforcer from the coactor. Under equitable conditions, the authors found that only 1 of 12 participants ever emitted this destroy response, which was followed by noncooperative responding in the dyad. The findings of these studies suggest that when inequitable reinforcement is arranged, give and take responses may restore equity over time. In addition, taking responses, when available under equitable conditions, resulted in noncooperation, as did destroy responses. In each of these studies, unique patterns of responding were observed under different levels of reinforcer inequity and responses that could be emitted by members may restore equity. Thus, a critical feature of these arrangements was the presence of feedback stimuli, typically a reinforcer counter, indicating that reinforcers were being delivered unequally.

Performance feedback serves a critical role in cooperative contingencies by signaling when inequitable conditions exist. Feedback regarding each members’ performance was continually available in the above-described studies on reinforcer inequity, which was necessary for members to determine unequal distribution of reinforcers. When stimuli indicating performance or reinforcer delivery are not continuously available, some research has arranged for participants to emit self-audits, which show their own performance, or coactor audits, which show the performance of their partner. Hake, Vukelich, and Kaplan (1973) assessed participants’ self- and coactor-audit responses under alone, parallel work, and cooperative arrangements. The authors observed higher rates of self- and coactor audits when another participant was completing an identical task (parallel work) than when the participant worked alone. Moreover, self-auditing increased only when a coactor’s score was accessible to the participant, suggesting that other members’ performance affects participants’ auditing, not just the presence of another person. Schmitt (1987) found similar increases in response rate under individual contingencies when coactor scores were accessible, suggesting that these stimuli may affect responding even when they do not affect reinforcement. Vukelich and Hake (1974) also found that auditing was more likely to occur when the participant and coactor scores were even, compared to when the participant was behind or ahead. These findings suggest that feedback regarding a coactor’s performance may serve a discriminative function based on its relation to the participant’s own performance (Hake et al., 1973; Vukelich & Hake, 1974), even though the coactor’s performance does not affect the availability of the consequences of the participant’s responding (Hake et al., 1973; Schmitt, 1987; Vukelich & Hake, 1974).

Of the above described studies on cooperative contingencies, the majority included concurrent schedules of reinforcement, such that reinforcers were available for responding on more than one alternative. Maximal reinforcement for each participant would typically be considered interdependent because it necessitated the coordinated responding of both members of the dyad (Hake & Vukelich, 1972). As an alternative, a dependent cooperation procedure involves the behavior of each organism producing reinforcement for the other (Hake & Vukelich, 1972). This has also been termed “exchange responding” (Schmitt, 1984) or “fate control” (Kelley & Thibaut, 1978). In one example, Boren (1966) trained pairs of stumptail macaque monkeys to emit a lever pressing response that produced reinforcement for the other monkey in the pair. The monkeys were initially trained to engage in alternating responses, such that lever pressing for one produced a reinforcer for the other. The contingency alternated so that the work requirement and reinforcement alternated for each monkey. The alternation requirement was then removed and a free-responding procedure was introduced such that either monkey could press their lever, which produced reinforcement for the other. In one pair, a single monkey continued to emit the response, whereas the other did not (but was fed gratuitously). In the second pair, responding decreased in both monkeys over time. As a result, the authors concluded that this arrangement would have eventually led to starvation of both monkeys and terminated the study.

Other studies have demonstrated similar problems with dependent cooperation procedures. For example, Baron and Littman (1961) arranged an operant chamber with a lever and food dispenser at opposite ends. Two rats were placed in the chamber with a press of the lever resulting in food at the other end of the chamber (0.66 m away). The authors found that one “worker” rat would press the lever, whereas the other “dependent” rat would consume the dispensed pellet before the worker could cross the chamber. As a result, a decline in the total response rates for eight of nine pairs of rats was observed. The findings of the studies using dependent arrangements are concerning likely because they do not involve the critical features of cooperative contingencies. Instead, responding by a single member produces inequitable reinforcers for the other, possibly without reciprocation.

The finding of Baron and Littman (1961) are unsurprising as the lever pressing of the “worker” rat likely contacted extinction. Conditions may also be arranged such that responding by each member produces reinforcement independent of the group. For example, in a slight variation to the chamber arranged by Baron and Littman, Grott and Neuringer (1974) arranged a single dispenser and lever at the end of a tunnel, which limited access to a single rat. The authors found that when three rats were placed in the chamber, turn taking was observed across a number of reinforcement schedules. Similar sharing or turn-taking behavior was found in a study by Hake, Vukelich, and Olvera (1975). In their study, pairs of high school students were given the choice to either give or take a match-to-sample problem, which produced money when completed. When a trial began, the problem was distributed based on the taking or giving response by whichever participant engaged in the response first. The authors observed that the majority of participants received reinforcement by taking, not receiving, problems. However, participants were observed to alternate taking responses, resulting in equity among participants.

The paradigms used by Grott and Neuringer (1974) and Hake et al. (1975) produced a unique form of cooperation in the form of turn taking as it was only possible for a single member to emit the response at a time. In this way, brief periods of inequity necessarily are presented until each member is allowed to emit a response. Subsequent descriptions of exchange contingencies have described conditions under which sharing responses occur only after extended periods of reinforcer accumulation for a single member (Matthews, 1977; Matthews & Shimoff, 1979; Schmid & Hake, 1983). Critical to these arrangements is a history of temporary inequity, followed by equity (or near equity; Schmid & Hake, 1983). Matthews and Shimoff (1979) observed high levels of trusting behavior in pairs of undergraduate students. In subsequent conditions, time and exchange-uncertain conditions resulted in the reduction of trusting behavior. Thus, the potential that the partner would be unable to reciprocate, which would result in unequal reinforcers, produced low levels of trusting behavior.

The above-described studies demonstrate various environmental conditions under which cooperative responding may be developed, disrupted, and restored. Critical to all arrangements is the equal distribution, or potential for equal distribution, of reinforcement. This may include interdependent arrangements, when coordinated responding among actors is required, or dependent arrangements, when reinforcement for one actor is presented solely on responding by the other. Disruption to cooperative responding may occur under periods of reinforcer inequity; however, alternative responses such as taking or giving may restore equity. These experimental arrangements may guide greater understanding of interdependent or dependent responding within groups, yet additional arrangements should be considered. In a slight variation on the distribution of reinforcement, competitive contingencies may be arranged through unequal reinforcement or responding for a limited reinforcer.

Competition

Competitive contingencies are said to be arranged when a reinforcer is obtained by fewer than all those who responded (Schmitt, 1976) or when reinforcer distribution is unequal, based on individual performance (Schmitt, 1998a). Competitive arrangements also frequently require independence of responding, rather than the follow-the-leader and coordinated responding arrangements that are commonly included in cooperative contingencies. That is, although participants’ responding is necessarily affected by the responding of the coactor(s), responses can be emitted independent of others. In general, competitive contingencies are associated with the greatest number of responses per reinforcer (Schmitt, 1981), because a single reinforcer may be arranged for responding by an entire group. This advantage may become less pronounced as the number of participants in a group increases, with some research showing triads being less likely to complete the reinforcer sequence compared to dyads (Schmitt, 1976). Nevertheless, when given the choice among individual, cooperative, and competitive contingencies, participants may select competitive arrangements when the magnitude of the reinforcer is greater (Schmitt, 1976). Similar variables to those affecting responding in cooperative arrangements, such as performance feedback and concurrent schedules, are commonly arranged in competitive contingencies; yet the function of these variables differ under these conditions.

Competitive arrangements have typically included stimuli regarding the performance of each member (Schmitt, 1976). Schmitt (1998a) assessed the role of performance feedback during 2-min competitive contests. In the feedback condition, the total number of competitive responses by each participant was displayed and removed in the no-feedback condition. The authors report very small differences between conditions, potentially due to the brief duration of the contests. In a subsequent study, Schmitt (2000) arranged for participants to separately audit their own performance and that of the coactor. High rates of interpersonal auditing, defined as separate self- and coactor-audit responses being emitted within a brief period, were observed. These interpersonal audits were associated with an increase in rates of competitive responding when scores were similar, consistent with previous findings of Vukelich and Hake (1974). As previously described, Vukelich and Hake did not include a competitive arrangement (i.e., reinforcement was independent of responding by the coactor), yet auditing still occurred at high rates when performance feedback was available. These findings suggest that similar performance may be discriminative for reinforcement of increased rates of responding, although larger discrepancies in performance may not serve the same discriminative function, in particular if the feedback is delivered near the end of the experimental session (Schmitt, 1986, 2000). Thus, the effects of performance feedback may differ based on discrepancies in performance, but also the frequency and timing of feedback.

Performance feedback is uniquely implicated in competitive arrangements as discrepancies between participants’ performance may serve an additional discriminative function under lean schedules of reinforcement. That is, responding under competitive contingencies is typically associated with a greater number of responses for the same reinforcer. Yet similar performance, identified through auditing, may serve a similar function to the near-miss effect in gambling research (e.g., MacLin, Dixon, Daugherty, & Small, 2007). That is, “nearly winning” or “approaching winning” may produce more durable responding than conditions under which performance is highly discrepant. Nevertheless, competitive contingencies include intermittent and unequal schedules of reinforcement for responding by participants in a group by definition, such that each individual will contact the contingency relative to their own performance and that of their competitors (Schmitt, 1986, 1998a). As a result, reinforcement may be isolated to certain individuals and may reduce participation in the competitive contingency if the distribution of reinforcement is not equitable (Matthews, 1979). To examine this potential effect, Schmitt (1998a) arranged reinforcement contingencies that were either fixed (i.e., winner gets 100% of reinforcer, loser gets 0% of reinforcer) or proportional (i.e., reinforcer amount matched proportion of total responses) in dyads of college students. The authors observed greater responding in competitive contingencies when reinforcement was proportional to each participant’s responding. Less interpersonal auditing was also observed under this condition (Schmitt, 2000).

Despite the advantages of proportional reinforcement, fixed arrangements may be more commonly used, likely due to the ease with which they are implemented (Schmitt, 1986). In fixed competitive contingencies, the reinforcer and the criterion for reinforcement is set before responding begins. Although there is still more to be learned about the effects of competitive contingencies, if fixed contingencies are to be used, Schmitt recommends including additional contingencies to promote continued responding across participants. Take for example a competitive contingency in which the classroom that sells the most raffle tickets would receive some prize. An additional contingency may be arranged with little additional cost that would allow for a smaller prize to be delivered to the classroom that sold the most raffle tickets each day for the duration of the contest. This additional contingency may produce continued responding in classrooms that are unlikely to win the larger contest.

The findings of experimental research on leader–follower arrangements and cooperative and competitive contingencies have broad implications for behavior analytic interventions in applied settings. The variables found to affect responding under cooperative and competitive arrangements, such as performance feedback, auditing, taking, giving, and destroying responses may likewise occur in group contingency interventions. Applied research and practice may benefit from considering methods used in the experimental literature that may contribute to the effective arrangement of group contingency interventions.

Group Contingencies

Group contingencies, also referred to as “group-oriented contingencies” (Gresham & Gresham, 1982; Litow & Pumroy, 1975) are the application of operant contingencies at the level of an entire group (Litow & Pumroy, 1975). In group contingencies, the behavior of individuals within a group affect the delivery of some consequence to its members. Three types of group contingencies are classified based on (1) which members’ responses will inform the criterion and (2) who receives the consequence once the criterion is met. Each type of group contingency has received extensive attention in educational settings (see Hayes, 1976; Litow & Pumroy, 1975; Little, Akin-Little, & O’Neill, 2015), although they may be programmed in other settings as well. The oft-cited strengths of group contingencies are the economical and practical feasibility (Hayes, 1976; Litow & Pumroy, 1975). For example, when applied in a classroom, a teacher may arrange for a reinforcer to be delivered to the entire group based on responding at some criterion-level by all members of the class. This arrangement is likely more economical than individual contingencies in which separate criteria and reinforcers are available for each student. It may also be more practical as the reinforcer is interdependent on responding of the members of the group, requiring the teacher to deliver a single consequence and measure responding by a single member or a subset of the group.

When reinforcer interdependence is arranged, additional social consequences may be presented by group members to ensure the reinforcement criterion is met. These social consequences may occur as a collateral effect of the programmed contingency, but may exert substantial control over group members’ responding (see review by Hayes, 1976). Skinner (1953) described “group control” as instances when two or more individuals seek to exert control over another to increase the reinforcement or reduce the punishment of the group. This description of group control is consistent with interdependent or dependent cooperation arrangements (Hake & Vukelich, 1972), described above. The group contingency literature has likewise included interdependent and dependent arrangements, although the arrangements differ from procedures used in the experimental literature. The description of the different types of group contingencies below are presented in an attempt to illustrate how the applied interventions may include similar or unique procedures from those previously described in the experimental literature on cooperative and competitive contingencies.

Interdependent Group Contingency

Interdependent group contingencies include the delivery of some consequence to an entire group contingent on all members’ behavior meeting some criteria. As the most common type of group contingency (Little et al., 2015), these arrangements have been applied to a number of problems including recycling (Hamad, Cooper, & Semb, 1977), energy consumption (Slavin, Wodarski, & Blackburn, 1981), and aberrant behavior (Barrish, Saunders, & Wolf, 1969). Interdependent group contingencies are frequently touted as the most ideal group contingency arrangement for their inclusion of cooperative contingencies, the most common example of which is the Good Behavior Game (GBG; Barrish et al., 1969).

The first application of the GBG involved a classroom of fourth-grade students being split into two teams. The class received a description of 11 rules and were told that any instance in which a rule was broken, a point would be given to that student’s team. Whichever team had the fewest points, or if both teams had fewer than five points, all members would receive special privileges. Significant reductions in problem behavior were observed during the intervention with the authors anecdotally reporting students requesting to play the game during periods when it was unavailable (Barrish et al., 1969). Considerable research has since evaluated this and other interdependent group contingencies, allowing for a greater analysis of variables that affect the efficacy of these arrangements or those that produce collateral effects (Flower, McKenna, Bunuan, Muething, & Vega, 2014; Tingstrom, Sterling-Turner, & Wilczynski, 2006).

Interdependent group contingencies and the GBG are consistent with cooperative arrangements that employ response-sharing procedures (Hake & Vukelich, 1972) operating on a single schedule of reinforcement. That is, a particular consequence is only available for responding under a single schedule requirement, through participation in the group contingency intervention. Operating under a single schedule may introduce additional concerns, such as students “sabotaging” the game for the group (e.g., choosing not to play or emitting responses that affect other members’ access to the reinforcer; Harris & Sherman, 1973). This result is consistent with previous findings when “taking” or “destroy” responses were emitted during cooperation procedures (Schmitt & Marwell, 1971a, 1971b, 1972). Nevertheless, procedural differences between interdependent group contingencies and cooperation arrangements differ in important ways. In particular, previous experimental research has included the opportunity for participants to avoid conditions when taking was available (Schmitt & Marwell, 1971a), signaled when a taking response occurred and allowed the participant to avoid the loss, or allowed a participant to emit an avoidance response that cost a portion of the reinforcer (Schmitt & Marwell, 1971b). The findings of these studies largely suggest that when taking is available and the participant cannot block the coactor’s taking response, participants allocate responding away from the cooperative arrangement to the lower-paying alternative. They may likewise do so when participants are able to block the taking response, but in so doing incur a cost. Nevertheless, when taking can be avoided, participants continued to cooperate. It is unclear how these findings may translate to interdependent group contingencies that operate on a single schedule (i.e., when no alternative is available). If we were to speculate, perhaps this would be in the form of a taking response being followed by another participant presenting some aversive consequence, which may further reduce the likelihood that the response criterion is met for that day. Instead, it may be imperative that interdependent group contingencies arrange for alternative methods for participants to access a reinforcer to reduce the disruptive effects of taking responses on cooperative responding.

The distinction between taking and destroying responses by Marwell and Schmitt (1975) is exceptionally pertinent to the single-schedule arrangements in interdependent group contingencies. As described above, the latter does not result in the reinforcer being delivered to the individual emitting the destroying response. In their study, only 1 of 16 participants emitted the destroy response, suggesting that a critical feature of taking may be the receipt of the reinforcer following a taking response. As such, the sabotaging that may occur in interdependent group contingencies may most closely align with taking, rather than destroying responses. If this is correct, it is likely that the response produces some reinforcer beyond spoiling the contingency for the group. Arranging alternatives that may produce access to the same reinforcer may be one method to reduce the probability of taking responses. Nonetheless, some research suggests that persisting in cooperation may reduce the likelihood that taking responses occur over time.

Marwell, Schmitt, and Bøyesen (1973) examined the effect of a confederate in cooperative arrangements who continued to emit cooperative responses even when taking responses were emitted by the coactor, referred to as a pacifist strategy. The authors observed high rates of cooperation, taking followed by cooperation, or taking with noncooperation by participants. These results suggest that the emission of cooperative responses by the confederate, even when taking responses occurred, resulted in high rates of cooperation overall. Notably, in a sample of participants from the United States, 8 of 20 emitted a take response followed by continued noncooperation. Under this response pattern, participants did not contact the pacifist strategy, likely because of the possibility of retribution. It is unclear how these findings would translate to single-schedule interdependent group contingencies, because there are no noncooperation alternatives available. Perhaps taking would be followed by eventual cooperation, although this arrangement remains to be studied.

Interdependent group contingencies and cooperation arrangements also differ in that the reinforcers are typically shared in interdependent group contingencies. That is, a single activity (e.g., five minutes of extra recess) may be available for all members in the group (Barrish et al., 1969; Harris & Sherman, 1973). In experimental studies on cooperative contingencies, each individual accumulates some reinforcer separate from her peer. This leads to additional questions about reinforcer selection in interdependent group contingencies. Of course, if a prescribed consequence does not function as a reinforcer for all participants, then taking, destroying, or other aberrant responses may be emitted by a subset of members. Experimental research might evaluate this further in terms of reinforcer inequity.

The arrangement of the GBG also includes a competitive contingency, such that the team with the fewest disruptive behaviors will receive special privileges unless both teams emit fewer than five disruptive behaviors. These combined competitive and cooperative contingencies have received some attention in the behavioral literature. For example, Okun and Di Vesta (1975) examined the effect of differing intragroup (within-group) contingencies (i.e., equal or unequal consequences) and intergroup (between-group) contingencies (i.e., competitive or noncompetitive) on the completion of a problem-solving task. The authors found that the most efficient problem completion was associated with equal consequences in the intragroup contingency and competition in the intergroup contingency. Thus, a cooperative contingency between group members and a competitive contingency with other groups may produce the greatest levels of responding. Goldman, Stockbauer, and McAuliffe (1977) conducted a similar study and found that when intergroup competition contingencies were present, the effect of intragroup contingencies was small; however, when intergroup cooperation contingencies were present, intragroup contingencies had a greater effect on performance. The inclusion of competitive and cooperative arrangements in the GBG may increase the efficacy of this interdependent group contingency. Nevertheless, problems associated with “taking” or lack of reinforcement for individual responding may still occur under this arrangement, although they may be more pronounced in dependent group contingencies.

Dependent Group Contingency

In dependent group contingencies, a consequence is delivered to the entire group contingent on the behavior of a specific individual, or a subset of the group (Litow & Pumroy, 1975). This arrangement deviates from the research cited above on dependent cooperative procedures in an important way: in dependent group contingencies, the delivery of some consequence to all members is dependent on responding by a single member. Thus, for the target member, her reinforcement is dependent on her own responding. In contrast, the experimental research on dependent cooperative procedures involve access to some consequence for one member being made contingent on responding by another, and vice versa. It is noteworthy that dependent cooperative procedures were associated with problematic patterns of responding in dyads when each members’ access to some consequence is entirely dependent on responding by another (e.g., Baron & Littman, 1961; Boren, 1966). The arrangements utilized in dependent group contingencies may be effective because they do not require complete dependence as reinforcement is arranged for all members (including the target member), if the response requirement is met. Nevertheless, the target member’s responding may be affected by additional variables, consistent with group control.

When the delivery of a consequence for an entire group is dependent on responding by a single member, the group may arrange for additional consequences to be delivered contingent on the target member’s behavior. Skinner (1953) suggested that group control would be exerted when the behavior of the target member is necessarily interlocked with the behavior of the other members in the group. Dependent group contingencies are a unique example of this “interlocking” because the behavior of the nontarget members does not inform the contingency. Dependent group contingencies may be avoided because of the extent of group control, in particular if responding does not meet the criterion and aversive consequences may be delivered by other members (Davis & Blankenship, 1996; Hayes, 1976), yet they may also be used for this very effect. Take, for example, how dependent group contingencies may be applied in multichild families. When a child does not finish eating their vegetables, the caregiver may arrange a dependent group contingency in which all of the children will go to the park once the child finishes the meal. In this arrangement, it may be very likely that the other children attempt to prompt and reinforce the eating behavior of their sibling or present aversive stimuli following periods without eating. This pattern of responding by the child’s siblings may be expected as it is the only response that they may emit to affect the contingency. That is, the target response of eating vegetables is not available to them and attempting to consume their sibling’s vegetables would be treated as subverting the contingency. The effects on collateral responses by the nontarget members are likely the most interesting products of dependent group contingencies; however, they have not been extensively evaluated in applied realms.

Recent applied research has suggested that dependent group contingencies may be associated with positive interactions between peers. Ennis, Blair, and George (2016) found that teachers reported that the dependent group contingency was associated with the greatest levels of peer support compared to the other group contingency interventions. Experimental research may likewise evaluate the behavior of nontarget members when the reinforcement criterion is met inconsistently. There are numerous variables that may affect responding (e.g., level of interdependence, type of task, or prior history). In addition, other collateral effects of this contingency may be observed, even when additional responding does not produce differential consequences. For example, in an attempt to increase attendance rates of youth group employees, Berkovits, Sturmey, and Alvero (2012) programmed a dependent group contingency whereby a group of employees received a monetary bonus based on the attendance of the group leader. If the group’s leader attended all sessions, the entire group received a bonus. Although the attendance of the other group members had no impact on bonus earnings, an increase in attendance for all members was observed during the intervention. Although these collateral effects may be ideal, there are alternative methods for increasing performance by all members in dependent group contingencies, such as arranging unknown components.

Despite the possible advantages of identifying the member whose behavior informs the contingency, this component may be “randomized” or made “unknown” to the group (Cariveau & Kodak, 2017; Hawkins, Musti-Rao, Hughes, Berry, & McGuire, 2009; Heering & Wilder, 2006; Little et al., 2015). Although not necessarily random, it may be ideal for the individual whose behavior is informing the contingency be unknown or change frequently. This variability would be consistent with programming indiscriminable contingencies, which may produce higher rates of responding by all members of the group and greater responding per reinforcer. This may be contrasted with instances in which the target member is known. Under these conditions, nontarget members may emit lower rates of the target response. Instead, they may emit responses more consistent with group control, arranging for additional contingencies to increase the probability that the target member’s responding would meet the criterion.

The parallels that may be drawn from the experimental and applied research on social behavior and dependent group contingencies are equivocal at this time. Translation is particularly difficult due to the procedural variations in dependent cooperative arrangements and dependent group contingencies. In particular, the arrangement of one member’s responding produces reinforcement for all members of the group (including herself) in dependent group contingencies. This arrangement has not been evaluated in the experimental literature on social behavior, which has largely focused on isolated contingencies in which responding by both members is required. Responses such as taking, giving, or destroying may be of similar interest in dependent group contingencies. It is noteworthy that these responses may be emitted by the target or nontarget member under different conditions, although the effects may be similar to those observed under reinforcer inequity. For example, if one member’s responding produces a reinforcer for herself and her partner, despite her partner not being required to emit any responses, taking may be emitted by the target member. Instances of taking under these conditions would be a unique finding from those on reinforcer inequity (e.g., Schmitt & Marwell, 1972; Shimoff & Matthews, 1975) because this would be more consistent with response inequity. Nevertheless, response inequity would likely affect responding in a similar fashion. In addition, giving responses by the nontarget member may also increase in order to ensure that responding by the target member persists. Additional experimental and translational research is needed in arrangements that more closely model dependent group contingencies to evaluate these potential patterns of responding. Furthermore, responding by the target and nontarget members are likely affected by distinct variables, which may require measurement of each members’ responding or use of confederates as done by Shimoff and Matthews (1975) rather than responding by pairs (e.g., Hake et al., 1975; Matthews & Shimoff, 1979; Schmitt & Marwell, 1972).

Additional research should further consider the effects of single or concurrent schedules in dependent group contingencies on the behavior of the nontarget members. Single schedules may be associated with greater punishment when responding by the target member does not meet criterion, whereas concurrent schedules may be associated with defection from the dependent group contingency. Future research might also examine the role of alternative responses by the nontarget members that may contribute to the response requirement, such as helping.

Independent Group Contingencies

In independent group contingencies, a response criterion is set for the entire group with the consequence being delivered based on each members’ performance independent of the group (Litow & Pumroy, 1975). That is, the same contingency is arranged for the entire group, but consequences are delivered based on individual members’ performance. This arrangement is often contrasted with individual contingencies, which may include different response requirements or reinforcer magnitude for individual members. Instead, independent group contingencies are identical for all members and is consistent with independent responding alternatives in studies on cooperative and competitive arrangements (Hake & Vukelich, 1972; Schmitt & Marwell, 1972). In the previously described experimental studies, independent alternatives were arranged so that responding outside of the cooperative or competitive contingency could occur, albeit with a reduced reinforcer. Responding on this alternative was seen as deviation from cooperation or competition. Likewise, independent group contingencies are not arranged in a way that facilitate competitive or cooperative responding as each member’s behavior only affects her access to the consequence. As a result, this group contingency would not result in group control.

In one example of an independent group contingency, Dallery et al. (2008) arranged for money vouchers to be received by eight habitual cigarette smokers if the individual participant’s carbon monoxide levels were at or below the assigned criterion, independent of all other participants’ performance. This arrangement effectively reduced carbon monoxide levels for seven of eight participants. In another example, Allen and Iwata (1980) arranged an independent group contingency to increase the number of exercises completed by ten adults with intellectual disabilities. Participants were given access to games (e.g., ball passing, tag, running relays) only when they had completed all of the assigned exercises. The results showed that the group contingency effectively increased exercise participation for all participants. In both examples, the contingencies were arranged such that the reinforcer was delivered independent of responding by the other members.

Independent group contingencies are present in a number of conditions beyond those that may be explicitly programmed in educational settings. For example, consequences for criminal behavior are delivered to the offending member (you do not pay your neighbor’s speeding ticket). These independent group contingencies may be effective in controlling individual members’ behavior and are arranged such that each member must meet the criterion independent of other members. This lack of interdependence effectively eliminates the potential for group control. Instead, these arrangements may be useful in promoting responding at some criterion for individual members. For example, if each member will be expected to read at some level of proficiency, it will be critical that the response-criterion be met by each member, independent of the reading proficiency of her peers.

When independent group contingencies are appropriate, additional considerations may be made to increase the efficacy of the intervention. Performance feedback is one area that may facilitate performance in independent group contingencies. Although each member’s performance only affects their own reinforcement, the findings of Hake et al. (1973) suggest the potential utility of feedback regarding individual and other members’ performance. In particular, these authors observed higher rates of self, coactor, and interpersonal audits during cooperative arrangements in which the participants could assist the other in completing a task, and during parallel work arrangements in which the participants worked on similar tasks independently. These findings suggest that performance feedback may be solicited when available, even though the performance of others does not influence an individual member’s reinforcement. In addition, previous work has shown that greater responding may occur simply as a result of the presence of another individual completing a similar task, termed “social facilitation” (Hake & Laws, 1967; Zajonc, 1965). Applied studies might evaluate the delivery of feedback regarding individual and other members’ performance in independent group contingencies. In applied work, feedback may only be provided once the independent group contingency is no longer in effect (i.e., at the end of the intervention period). In addition, the criterion for reinforcement may be arranged in an all-or-nothing fashion in which the individual’s performance either met the criterion, and produces reinforcement, or does not meet the criterion, and does not produce reinforcement. The effects of this arrangement are unclear and the experimental literature does not provide guidance as to how best these contingencies should be properly arranged, although variables related to the timing of performance feedback (Hake et al., 1973; Schmitt, 1998a, 2000) and reinforcer inequity (Schmitt & Marwell, 1970, 1972; Shimoff & Matthews, 1975; Matthews, 1979) may provide some guidance.

Similar to the experimental research, it may be most effective to arrange for independent group contingencies to be available as an option when an interdependent or dependent group contingency is also arranged. Arranging for concurrent schedules may be valuable for a number of reasons. Concurrent schedules may reduce the likelihood of group control during instances in which a single member’s responding may result in the loss of a reinforcer for the group. In these instances, members may defect to the independent group contingency to maximize reinforcement. The ability for individual participants to choose to participate in any group contingency has not been described in the literature and allowing an alternative may increase the social validity of these procedures. In research, including an independent group contingency as an alternative to a dependent or interdependent group contingency might also allow for further evaluation of those variables that may affect cooperative responding. In this way, researchers may further evaluate methods to promote sustained cooperation in applied settings and identify the controlling variables responsible for performance therein.

Additional Considerations and Future Research

The types of group contingencies described above often include a number of additional variables that may influence the efficacy of the arrangement. Prior history among participants has also received little attention in studies on group contingencies, despite the potential effects that history may have on responding under certain arrangements. Marwell and Schmitt (1975) describe three studies in which participants received minimal exposure prior to beginning a cooperative arrangement study. Participants either briefly saw the other participant prior to the experiment, were given 10 min to talk before the experiment, or were visible throughout the experiment. None of these conditions resulted in higher rates of cooperation than conditions without prior exposure. In two additional studies, participants included married couples and best friend pairs. The authors observed consistent cooperation in married couples with 7 of the 20 friend pairs engaging in predominantly noncooperation and 11 pairs emitting at least one taking response. The authors speculate that these differences in responding may be a result of shared economies in married couples, which is likely not the case in friend pairs. The history between members in group contingency research has received little description or been inadequately controlled. As a result, patterns of responding may not adhere to those observed in married couples or best friend pairs described by Marwell and Schmitt (1975). Instead, groupmates may have a pervasive history of negative interactions, potentially the reason why a group contingency intervention is being introduced. The extent to which this history affects responding in group contingencies is unknown, but likely an important variable to be considered in concert with other social stimuli.

Verbal behavior and contingency-specifying stimuli may also play a significant role in the efficacy of group contingencies and responding in cooperative arrangements (Hake & Vukelich, 1973). Schmitt (1984) described face-to-face interactions as allowing for members to introduce additional contingencies beyond those programmed in the experimental paradigm. Numerous researchers describe verbal interactions between participants, such as bargaining (Azrin & Lindsley, 1956; Marwell & Schmitt, 1975), requesting point totals (Hake & Vukelich, 1973), or discussing methods for sharing problems (Hake et al., 1975). This communication may have considerable effects on responding in group arrangements; however, researchers have not adequately controlled for these variables, in particular in group contingency research.

Verbal behavior emitted by the experimenter describing the programmed contingencies arranged in a given study are also common in cooperative contingencies (Schmitt, 1998a) and group contingency interventions (Hayes, 1976). The ubiquity of these instructions may lead to uncertainty regarding whether the observed effects in group contingency interventions are a result of the arranged contingency or instructional control. Hake and Vukelich (1973) describe responding by three subjects to be, at least in part, under instructional control. Under certain conditions, human behavior may initially adhere more closely to instructed, rather than programmed, contingencies (Baron & Galizio, 1983; Galizio, 1979). Schmitt (1998a) recommended that instructional procedures should only describe or model the available responses, but not response rates. Moreover, it may be critical that participants are exposed to the programmed contingency for prolonged durations to ensure that steady state responding is observed (Marwell & Schmitt, 1975), instead of control by verbal stimuli (see Schmitt, 1998b). Future research might evaluate the effects of various instructional procedures on responding in group contingencies.

Performance feedback is a final area that should receive greater attention in group contingency interventions. The discriminative or motivative effects of the individual and other members’ performance is critical to better understanding behavior in groups (Schmitt, 2000). Future research might consider various methods of performance feedback that may affect individual responding. Moreover, the effects of discrepant performance by members in groups of more than two individuals under cooperative arrangements may be another area of fruitful research. Past research on performance feedback has predominantly included cooperative contingencies (Hake et al., 1973), competitive contingencies (Schmitt, 1998a, 2000), or independent contingencies (Hake et al., 1973; Vukelich & Hake, 1974) in dyads. It is unclear how performance feedback regarding multiple members, and potentially disparate performance among these members, influences responding. These variables may be of considerable interest as the number of individuals within a group has been shown to affect performance in group arrangements (Schmitt, 1984) and monitoring of group members’ performance may be a critical variable for further consideration.

Conclusion

The behavior of individuals in groups has been suggested as a critical test for the adequacy of a technology of human behavior (Skinner, 1953). Research on group contingency interventions continue to be published in behavior analytic and educational outlets, yet a considerable decline in human operant research on cooperation and competition arrangements has occurred (Schmitt, 1995). The current article describes some areas that may be bridged between these two literatures, although considerable work has yet to be done. We hope that the current article illustrates the utility and need for both experimental and applied research on interpersonal and group contingencies to further our understanding of social behavior and contribute to more effective interventions.