Social Psychology Quarterly 2019, Vol. 82(2) 158–181 The Formation of Group Ó American Sociological Association 2019 DOI: 10.1177/0190272518813562 Ties in Open Interaction journals.sagepub.com/home/spq Groups

Shane Thye1, Edward J. Lawler2, and Jeongkoo Yoon3

Abstract We examine how task jointness and group incentive structures bear on the nature and strength of the affective and cognitive ties that people forge to a group. The argument is that affective group ties have stronger effects on social order than cognitive group ties. There are two general hypotheses. First, joint tasks generate stronger cognitive and affective ties to groups, whereas group incentives generate cognitive but not necessarily affective ties to the group. Second, affective ties more effectively solve two fundamental problems of social order in groups: (1) sustaining membership (also known as stay behavior) and (2) generating the joint gains of further collaboration (cooperation). The theoretical logic is that joint tasks pro- mote a sense of shared responsibility, and this leads members to attribute their individual emotions to the group as an object, whereas alignment of individual and group incentives does not produce such effects. The theory and hypotheses are tested experimentally in four- person open interaction groups, manipulating task jointness (high, low, none) and incentives (individual based vs. group based). The results generally support the hypotheses underlying the theoretical logic. Affective ties to groups are based primarily on levels of task jointness, and such tasks have stronger effects than incentives on the capacity of groups to retain mem- bership and induce cooperation in social dilemmas.

Keywords affective ties, cooperation, emotion, group ties, social order

A classic issue in sociology is whether promote cooperation on behalf of group or to what degree person-to-group ties goals. In this article, we examine how are instrumental or affective (Durkheim and when fundamental structures and [1933] 1964; Parsons 1951; Simmel processes promote different forms of 1964; Weber [1918] 1968). Group ties group ties that in turn have different are instrumental if they are based pri- marily on individual rewards or payoffs; 1 group ties are affective if the group is an University of South Carolina, Columbia, SC, USA 2Cornell University, Ithaca, NY, USA object of intrinsic worth or value, namely, 3Ewah Womans University, Seoul, Korea an end in itself. Such ties are important Corresponding Author: because how people are tied to a group Shane Thye, University of South Carolina, 309 bears on its stability, for example, the Sloan College, Columbia, 29208, USA. capacity to sustain membership and Email: [email protected] Formation of Group Ties 159 consequences for group stability and company’s marketing and legislative social order. efforts to assisting with their philan- We theorize the general idea that thropic activities. HP reportedly mobi- affective ties to a group generate stronger lized thousands of former employees in person-to-group ties and therefore invoke such activities, which is itself noteworthy more resilient social orders than instru- because Silicon Valley is reputed to be mental ties (Berger and Luckmann a place where company loyalty is as rare 1966; Collins 2004; Parsons 1951). Affec- as a pinstripe suit. As documented in tive ties involve feelings about a group, a New York Times article, many retired whereas instrumental ties are cognitive employees said they joined such efforts and involve a perceived connection to because of their ‘‘affection for the com- a group. Actors with only instrumental pany,’’ and they often noted that these vol- group ties more readily leave groups in unteer involvements make them ‘‘feel response to changes in incentive or good.’’ One volunteer said: ‘‘I feel like I reward structures, whereas actors with have two marriages: a wonderful marriage affective bonds are more inclined to at home for 36 years and a marriage at remain in the group and contribute to HP.’’ Such affective ties to the company cooperative endeavors even if they by retirees are a source of significant bene- involve risk. Affective ties entail a nonin- fits to HP (Lawler, Thye, and Yoon 2009). strumental component in addition to the This news anecdote dramatizes the capac- instrumental component. Affective and ity of affective person-to-group ties to mobi- instrumental ties are certainly inter- lize behavioral commitments to the group. twined in any given context, but the ana- In this study, we test a theory of social lytic distinction is important and has commitments (Lawler et al. 2009, 2015) a long tradition in sociology. We aim to that posits the structural conditions and pull apart and test the distinct effects of processes that determine whether and affective-based and instrumental-based how social interactions generate affective group ties, recognizing that they are con- group ties and more resilient social orders nected. Similarly, in psychology, cogni- in groups or organizations. This theory tion refers to perception and affect to feel- builds on the affect theory of social ings. These are known to be interwoven, exchange (Lawler 2001; Lawler et al. but they are distinguished for many pur- 2008), but it is distinct in three general poses. In our research, a cognitive tie to ways. First, it extends principles of the a group refers to a perceived connection theory beyond social exchange and net- between self and group, whereas an affec- work contexts by expanding the theory’s scope to encompass task groups where tive tie refers to a feeling about the members work on a cooperative task and group. Instrumental ties are cognitive interact openly to accomplish it. A task but not necessarily affective. group is defined here as three or more The importance of affective group ties actors who interact with one another to is anecdotally illustrated by a program produce a collective product or result. that Hewlett-Packard (HP) developed to Second, we elaborate and explicitly test involve retired employees who wished to the crucial role of shared responsibility volunteer for the company.1 Such volun- as the mechanism for affective group teer work ranged from helping with the ties (see Lawler et al. 2009, 2014). 1This example is based on a report in the New This mechanism has not been directly York Times, ‘‘Going to the Company Elders for tested in prior work. Third, we compare Help,’’ (Richtel 2008). how and when two distinct structural 160 Social Psychology Quarterly 82(2)

(exogenous) conditions—the jointness of 1999; Colvin and Boswell 2007; Gottschalg the task and the individual/group incen- and Zollo 2007; Nyberg et al. 2010). tives for accomplishing it—generate Distinct traditions across the social sci- stronger or weaker ties of individuals to ences converge on the notion that to moti- the group. The theory of social commit- vate behaviors in the collective interest, ments highlights the role of person-to- the key is to align individual and group group ties and the kinds of tasks that peo- interests. Nevertheless, a substantial ple work on together. The theory argues body of theory and research indicates that joint group tasks promote stronger that (1) repeated exchanges generate affective ties to the group, in which the individual emotions and (2) under some group becomes an end in itself, whereas conditions, these feelings transform group incentives tend to promote instru- exchange relations into affective objects mental ties to the group, in which the of attachment or commitment (Lawler groupisameanstoindividualgain.In et al. 2000; Lawler and Yoon 1996; Thye this way, we question the prevailing et al. 2014; Thye, Yoon, and Lawler assumption in human resources and orga- 2002). What makes exchange instrumen- nizational behavior that aligning the inter- tal is the repeated flow of benefits and ests of individuals and a group or organiza- reciprocity between the same actors in tion is most crucial to employee retention the context of an exchange network and cooperative, group-oriented behavior (Emerson 1972a; Emerson 1972b; Molm (Nyberg et al. 2010). 2003; Willer 1999). What makes exchange The ties people have to groups are cen- affective is that repeated exchange gener- tral to the classic problem of social order. ates an emotional buzz in the form of At the societal level, Hobbes’s ([1651] everyday feelings of pleasure, satisfac- 1985) framing and solution to the order tion, or excitement. Research now demon- problem involved a contractual (i.e., strates that repeated exchanges are likely instrumental) tie between the members to transform instrumental ties into rela- of a society and the central authority of tional ties due to the cognitions such a group. That tie was based on the capac- as uncertainty reduction or trust and ity of the central authority (i.e., the state) affective states such as pleasure or satis- to serve individual interests, in particu- faction (e.g., Collins 2004; Cook and lar, the basic wants and needs of citizens Emerson 1984; Kollock 1994; Lawler for security and stability. This is an and Yoon 1996; Lawler et al. 2008). In instrumental solution that essentially our theorizing, a cognitive tie is defined aligns individual and group interests. as a perceived connection to the group Contemporary adaptations of this instru- (e.g., Tajfel and Turner 1986), and we mental solution are found, for example, in propose that it likely emanates from game-theoretic accounts of iterated social instrumental incentives, whereas an dilemmas and rational choice analyses of affective tie refers to feelings about the solidarity (e.g., Axelrod 1984; Hechter group and tends to emanate from the 1987; Kollock 1994). This principle more- task activity.2 over is integral to many ‘‘best practices’’ frameworks used by management profes- 2There is growing empirical evidence for the sionals. Here, interest alignment is a impact of task jointness on individual-to-group term often used to capture the market ties in social exchange (Kuwabara 2011; Lawler, Thye, and Yoon 2008) but also in employee-to- logic that shapes compensation schemes organization ties in work organizations such as that essentially define contemporary nursing homes and schools (see Price and Collett employee-employer relations (Cappelli 2012; Taylor and Pillemer 2009). Formation of Group Ties 161

Conceptually, ties may be of multiple For instance, social identity theory sug- forms—to other persons (person-to- gests that people readily define them- person ties) or to the group itself (per- selves as members of a cognitive unit, son-to-group ties). This distinction was even minimal units, without any social integral to Parsons’s (1951) approach to interaction or emotional content (e.g., social order, Mead’s (1934) classic distinc- Tajfel and Turner 1986). tion between specific and generalized Our theoretical argument boils down others, and Simmel’s (1964) comparison to the following general claims: Affective of dyads and triads. Research in social ties to groups emerge and are sustained psychology demonstrates that person-to- if (1) members interact with each other, group ties are distinct, theoretically and (2) that interaction centers on a joint empirically, from person-to-person ties, task, (3) their interactions generate and implicitly they have different a sense of positive emotion and shared implications for social order (Brewer and responsibility, and (4) shared responsibil- Gardner 1996; Prentice, Miller, and ity leads them to attribute their feelings Lightdale 1994; Tajfel and Turner 1986). to the group. This process is predicted to For example, turnover in groups that yield stronger person-to-group ties that are founded on the ties members have to are manifest in (1) intentions to stay in each other should be more disruptive the group and (2) cooperation with other than turnover in groups founded on the members in the context of risk. Prior ties individuals have to the group itself research finds evidence of these effects (for evidence, see Prentice et al. 1994). in standard exchange settings and net- The theory of social commitments (Lawler works (e.g., Lawler and Yoon 1993, et al. 2009) focuses on how person-to- 1996, 1998; Lawler et al. 2008; Thye, group ties develop or strengthen, tracing Lawler, and Yoon 2011), but some impli- these to the social interactions that cations of the theory, especially the pre- unfold while individuals address a com- cise role of shared responsibility, have mon problem or shared goal. never been tested. This research tests For the most part, person-to-group ties the role of shared responsibility and are neglected in contemporary sociologi- extends the argument to an open interac- cal theorizing, much of which is informed tion setting in a way that allows for by imagery or cultural a novel and more complete test of the the- frames. With network imagery, sets of ory in a single study. At issue is whether interconnected person-to-person ties con- or not the hypothesized shared responsi- stitute and sustain larger units, and the bility mechanism occurs in small open larger units as such are not meaningful interaction task groups, as exampled by groups to the actors in those networks a university committee or organizational (Faris and Felmlee 2014; Gorman and team. Marsden 2001; Granovetter 1985; Watts Few scholars in microsociology exam- 2003). For theorists adopting a cultural ine social processes in the context of frame, actors access and use cultural open interaction (cf. McLeer et al. 2011; resources to interact with others in the Smith-Lovin, Skvoretz, and Hudson context of taken-for-granted or assumed 1986). Early sociologists began with ties to a larger social unit (Benford and open interaction task groups (e.g., Bales 2000; Swidler 1986). Neither 1950; Durkheim 1915; Homans 1961), approach treats person-to-group ties as but the study of such groups gradually problematic. Yet, group ties clearly gave way to more controlled experimen- emerge even under minimal conditions. tal settings in which interaction is 162 Social Psychology Quarterly 82(2) constrained to isolate specific phenomena a common condition of work organiza- or processes (Berger et al. 1977; Lawler tions in which employees receive a portion and Yoon 1993; Thye 2000; Willer 1999). of their total compensation from their indi- A noteworthy feature of the current vidual performance and a portion from the research is the focus on task groups collective or company performance. with open interaction. Our research is The theory of social commitments part of an effort, reflected in some work (Lawler et al. 2009, 2014) suggests that of others as well (Allen, Sargent, and high task jointness generates stronger Bradley 2003; Keyton 2016; McLeer cognitive and affective person-to-group et al. 2011), to bring the task group back ties, whereas joint interests generate cog- into contemporary sociological focus. We nitive but not affective ties to the do so by examining small groups who group. Cognitive ties refer to one’s per- work on an engaging problem (i.e., have ceptions of a connection to a group (e.g., high experimental realism) that resem- Tajfel and Turner 1986); such perceptions bles common everyday situations where may or may not have affective elements. people interact with one another to pro- Affective ties involve an attribution of duce a common product (i.e., have high individual emotions to the group.3 The mundane realism). theory of social commitments argues that the latter occurs when the task itself SOCIAL COMMITMENTS THEORY generates a sense of shared responsibil- ity. Shared responsibility is a key mecha- The theory of social commitments (Lawler nism that determines whether individual et al. 2009) is essentially a theory of emotions are associated with or attrib- person-to-group ties, and it provides an uted to the group entity. More specifi- explanation for how affective group ties cally, the shared responsibility induced develop among individuals working on by a joint task leads actors to attribute a common task. In the theory, the task positive feelings or emotions from the and interests are primary ‘‘social connec- task interactions to the group. These are tors.’’ The task structures social interac- termed social unit attributions of emo- tion around a common issue and shapes tion. The idea is that individually felt the activity directed at the group goal. A emotions are attributed to something joint task encourages social interaction actors share or have in common and the and engagement to succeed at the task. group affiliation is a salient object or tar- In other parlance, joint tasks enhance or get for these individual emotions. When strengthen the collective orientation of social unit attributions occur, the result the group. Interests are a complementary is a stronger affective tie to the group in but distinct type of social connector. the case of positive emotions or a weaker Shared interests motivate cooperation as a pathway to individual benefit or gain. 3Cognitive and affective ties are analytically The task is about the process, whereas distinct but empirically intertwined in any given the interests are about the individual- context. Affective ties have cognitive components, and cognitive ties can have some emotional load- level payoffs or outcomes. Applied to ing or charge. Nevertheless, the difference is a task group or organizational context, important. A cognitive tie itself is not sufficient the joint interests of members can be por- to make the group an object of intrinsic value or trayed as the proportion of individual end in itself, whereas an affective tie can produce such a result. We aim to create experimental con- rewards members derive from their indi- ditions that can instantiate the theoretical differ- vidual efforts versus the group efforts. ence, recognizing that there is overlap between This concept of joint interests reflects the affective and cognitive components. Formation of Group Ties 163 tie to the group in the case of negative of shared responsibility is a moderating emotions (Lawler et al. 2008, 2009, 2014). condition, determining when people Task jointness and shared responsibil- make social unit attributions; in other ity are, respectively, the objective and words, it is the interaction of shared subjective conditions for person-to-group responsibility and positive emotions that ties with an affective component. Social drives the predicted theoretical process. unit attribution is the mechanism This is the first study to explicitly test through which individual emotions lead this interaction effect. Figure 1 portrays to affective ties or attachments to the the theoretical principles of social com- group. If task activity generates a sense mitments theory and identifies the key of shared responsibility (Lawler et al. links that lead to the hypotheses to be 2008, 2009), social unit attributions tested here. transform individual emotions into collec- The theoretical model exposes the tive feelings. In comparison, group incen- mechanisms whereby joint tasks and tives that align individual and group group incentives bear on two important interests should not foster social unit dimensions of stability and order in attributions of emotion. The incentive groups—the propensity of members to structure may create a sense of being in stay in the group and also collaborate a group with shared goals and may even when they face a decision that pits their create a mild sense of shared responsibil- own individual interests against those of ity; however, these should be weaker the group (i.e., a social dilemma choice). than those produced by task jointness, In addition, the model has interesting and they should not generate social unit implications for the role of group incen- attributions of emotion.4 Joint tasks are tives. It suggests that group incentives especially important to the sense of alone are not sufficient to produce affec- shared responsibility because they make tive group ties because the social unit individual contributions more indistin- attributions of emotion are not present. guishable, that is, difficult to trace. A minimal group identity may emerge, The theory asserts that if task struc- but without much, if any, emotional con- tures seamlessly interweave individual . Thus, whereas instrumental condi- contributions and generate a group-level tions may be the initial impetus for inter- sense of shared responsibility among action, group ties take on an affective members, they are likely to perceive their component most prominently under con- individual feelings as jointly produced as ditions of task jointness. well (Lawler et al. 2009, 2014). The sense

4 Under some conditions, group incentives may HYPOTHESES create a weak sense of shared responsibility. In fact, the data in Table 1 support this notion, Following are six hypotheses derived F(1, 120) = 3.44, p \ .07. At issue is that group from the theoretical model in Figure 1. incentives create a weaker sense of shared The focus is on how task jointness and responsibility relative to task jointness. Our the- ory assumes that shared responsibility is incentive structure (Hypotheses 1a and grounded in and emerges from the task activity, 1b) generate affective and cognitive ties, that is, the process of doing the task. The the mechanisms through which these individual-group incentives orient individuals effects occur (Hypotheses 2a and 2b), toward or away from the group, but how they and the impact on the strength of the accomplish the task or work together (or apart) to reach the shared goal is not shaped by these group tie manifest in intentions to stay incentives per se. in the group and cooperative behavior 164 Social Psychology Quarterly 82(2)

Figure 1. Model for the Theory of Social Commitments. when members confront a social dilemma Hypothesis 3b. Affective group ties have (Hypotheses 3a and 3b).5 a stronger effect on cooperation in a social dilemma choice than cognitive Hypothesis 1a. Groups with joint tasks group ties. generate stronger cognitive and affec- tive ties to the group than groups METHODS without joint tasks. Overview Hypothesis 1b. Group-based incentives generate stronger cognitive ties to To test the theoretical predictions, an the group than individual-based experiment created a scenario in which incentives but not stronger affective four people undertook first an individual ties to the group. task and then a group task. The experi- Hypothesis 2a. The effects of joint tasks ment manipulated the incentive struc- on cognitive and affective ties are ture (i.e., how much profit or pay for the indirect and mediated by the interac- experiment was dependent on the indi- tion of shared responsibility and posi- tive emotions. vidual task vs. the group task) and the Hypothesis 2b. The effects of group-based jointness of the group task (no jointness, incentives on cognitive ties are direct low jointness, high jointness). The exper- and not mediated by the interaction iment was structured to accomplish three of shared responsibility and positive goals: first, to examine the impact of emotions. task jointness and incentive structures Hypothesis 3a. Affective group ties have on cognitive and affective ties to the a stronger effect on intent to stay group. Cognitive group ties capture the than cognitive group ties. extent to which the individual perceives self and group as interconnected, whereas 5Stay behavior is a standard measure of com- affective group ties reflect the emotional mitment, and a social dilemma taps the propen- attachment to the group (Lawler et al. sity to cooperate in the context of risk. Together, 2008). The second goal was to examine these reflect the capacity of a group to sustain its whether task jointness promotes a sense membership and capture the gains from cooperation. of shared responsibility and positive Formation of Group Ties 165 emotions from the task activity, in partic- the incentive manipulation. There were ular, whether shared responsibility and three conditions in the task jointness positive emotions interact. This mecha- manipulation. In the no jointness condi- nism has yet to be tested. The third goal tion, subjects completed the task together was to test whether affective and/or cog- in the same room but did so individually nitive group ties mediate the effects of and without any communication with task jointness and group incentives on the other subjects. This condition serves two behaviors that reflect the strength of as a control for the other conditions. the person to group tie: stay behavior and These groups do not satisfy our definition cooperation in a social dilemma. Intent to of a task group. However, they do consti- stay is an inclinationtocontinueworking tute a minimal group in the sense that with the same others in the future. Cooper- they occupy the same space, are working ation refers to contributions in a social on the same task, and some portion dilemma, specifically, the degree that of their payoffs are interdependent. actors allocate individual resources to the Research indicates that it does not take group or keep them for themselves. much interdependence to produce group- oriented behavior (Kramer and Brewer 1984). Their inclusion in the experiment Experimental Design and Subjects is to control for such effects and serve as The experiment consists of a 2 3 3 facto- a point of comparison for the other exper- rial design crossing incentive structure imental conditions. In the low-jointness (whether incentives are individually condition, subjects could discuss the lost based vs. group based) and levels of task at sea task for up to 10 minutes but jointness (none, low, high). When incen- then completed it individually and with- tive structures were individually based, out any further communication with the subjects are told that 80 percent of their others. In the high jointness condition, pay will be based on their performance subjects freely interacted for 10 minutes, on the individual task (i.e., the things and then they produced a single collec- that kill us), whereas 20 percent of their tively agreed on ranking. pay will be based on the group task (i.e., A total of 480 undergraduate students lost at sea).6 When incentive structures at a northeastern and southeastern univer- were group based, we reversed these per- sity participated in the experiment for pay- centages (20 percent based on the individ- ment. In all, 120 same-sex tetrads were ual task, 80 percent on the group task). A randomly assigned to one of the six experi- manipulation check on a prequestionnaire mental conditions (20 tetrads per cell; 12 confirmed that all subjects understood female and 8 male each). Because data were collected at two universities, gender 6These tasks were chosen for a variety of rea- and university affiliation were counterbal- sons. First, both tasks are intellectually engaging anced within each experimental condition.7 and novel and subjects are unlikely to have differ- ential knowledge or expertise regarding these subjects. Second, survival tasks are commonly Experimental Procedures used in the examination of group processes (John- son and Johnson 2009). Third, the tasks essen- The experiment created a situation where tially simulate conditions where people come four individuals first worked separately together (as in a work team or organizational committee) to collectively solve a problem with 7In the analyses, we statistically control for an unknown solution. They were chosen specifi- the gender of the subject and location of the cally to create both high experimental and mun- data collection. Neither race nor ethnicity were dane realism. coded or controlled in this study. 166 Social Psychology Quarterly 82(2) on an individual task in private cubicles to measure several theoretically impor- and then worked together in the same tant variables (e.g., perceptions of shared room on a second task. Upon their arrival, responsibility, positive emotions, cogni- subjects were escorted to separate cubi- tive ties); (2) were presented with a social cles, where they gave informed consent dilemma choice, which provides a behav- and then began work on the Phase I (indi- ioral measure of cooperation; and (3) com- vidual) task. The Phase I task was titled pleted a final (post)questionnaire that The Things That Kill Us. The instruc- contained the intent to stay measure of tions for this task explained that people group ties. Subjects were not provided face potentially harmful objects or engage with information regarding the outcome of in dangerous activities every day and the social dilemma until the experiment listed 15 dangers. Subjects were asked was over so this information could not affect to rank order how dangerous a thing is their responses on the postquestionnaire. based on the number of yearly deaths caused by that item. Subjects had 10 Measures of Intervening Variables minutes to complete the task.8 After they completed the Phase I task, Several questionnaire items were admin- subjects were escorted to a room and istered to capture the intervening factors seated around a semi-circular table. specified by the theory and test the Here they completed another ranking predicted interaction effects of positive task similar in form but not content to emotions and perceptions of shared the Phase I task. The second, Phase II responsibility. All were measured on task was titled Lost at Sea. The instruc- a midquestionnaire administered after tions told subjects they had chartered the lost at sea task but before the behav- a yacht for a holiday trip across the Atlan- ioral measures of group tie were adminis- tic Ocean when a fierce fire breaks out on tered. First, we measured positive emo- the ship. Much of the yacht is destroyed tions in the form of pleasure/satisfaction, and is now slowly sinking. However, they which were derived from prior research and the others have salvaged a four- (Lawler et al. 2008, 2009; Lawler and person rubber life raft and managed to Yoon 1996). Subjects were asked to report save 15 items. Their task is to rank the how they are feeling about their experi- importance of those items for their sur- ence right now using a series of bipolar vival.9 Again, subjects were told they had adjectives with a nine-point scale. The 10 minutes to complete the task.10 Sub- items measuring pleasure/satisfaction jects completed the Phase II task with were pleased-displeased, happy-unhappy, some degree of joint interaction (high or satisfied-dissatisfied, and contented- low) or no joint task at all (control). discontented (Cronbach’s a = .92). This Following the Phase II task, subjects measure is virtually identical to that used were escorted back to separate cubicles in previous research and displays consis- where, in the following sequence, they tently good measurement properties (Law- (1) completed a questionnaire designed ler et al. 2000; Lawler and Yoon 1993, 1996). 8See Supplemental Documents in the online Next, we measured perceptions of version of the article for details. shared responsibility on the same mid- 9 See Supplemental Documents in the online questionnaire. For this measure, subjects version of the article. 10The experimenter monitored the group were asked how they would describe through a one-way mirror to ensure no communi- themselves and the others during the cation occurred. task. There were two nine-point Likert Formation of Group Ties 167

Figure 2. Modified Inclusion of Other in Self from Aron, Aron, and Smollan (1992). scale items in the measure; one was Smollan (1992) found that the IOS scale anchored by (1) accountable to the self displayed high levels of alternate forms and accountable to the group and the reliability (a = .95; test/retest r = .85) other by (2) sharing little responsibility and is a good predictor of whether or not and sharing much responsibility (Cron- dating couples would still be dating bach’s a = .93). 3 months later. Aron and Fraley Cognitive ties to the group were mea- (1999:142) offered, ‘‘Over a variety of sured next using a modified version of samples and relationship types, it has the Inclusion of Other in Self scale (IOS; high test-retest and alternate form reli- see Figure 2). The IOS scale asks the sub- ability and indications of convergent and ject to select one of seven diagrams of discriminant validity equaling or sur- overlapping circles, representing self passing a variety of much more elaborate and other. Each choice consists of a circle verbal self-report measures of intimacy or labeled self and another labeled other; we closeness.’’ We view this measure as a par- substituted group for other to generate simonious way to assess the degree that a measure of person-to-group ties. The individuals perceive a tie to the group, first of seven images in the measure por- recognizing that this tie may have both trays the self-circle as touching but not cognitive and emotional elements. Theo- overlapping with the group circle. Then, retically, the IOS measure does not say each consecutive diagram shows slightly anything about people’s feelings or emo- greater overlap of the self and group cir- tions toward the group. The emotional cle. In the final choice, the two circles elements are better captured by our affec- are nearly but not quite superimposed. tive tie measure, and statistically, we Specifically, subjects were asked, ‘‘Which model these as distinct but interrelated of the following diagrams best represents constructs. how you perceive your relationship to the Affective ties to the group were mea- group of participants in the study?’’ This sured using a four-item index adminis- measure is simply coded 1 to 7 to indicate tered on the postquestionnaire. The how much the individual perceived self to measure included a series of bipolar be a part of the group. Aron, Aron, and adjectives on nine-point scales. Subjects 168 Social Psychology Quarterly 82(2) were asked, ‘‘How do you feel about your and give 8 to the group, or allocate all relationship to the group?’’ on the follow- 16 to the group. Tickets allocated to the ing items: bad-good, detached-attached, group would be increased by 50 percent negative-positive, and disloyal-loyal and then distributed equally among the (Cronbach’s a = .94). A variant of this participants. Thus, if all subjects kept measure has been used in past research all of their tickets, each would leave the where it also displayed desirable mea- experiment with 16 tickets. However, if surement properties (Lawler et al. one subject donated all of his or her tick- 2008). Confirmatory factor analysis indi- ets to the group and the other three sub- cates that the four intervening variables jects kept all of their tickets, the donating are distinct.11 subject would receive 6 tickets total, and the others would receive 22 total tickets (16 donated tickets turn into 24 tickets, Measures of Dependent Variables which are divided four ways). In the event Strength of group tie was measured in that all subjects donated all 16 tickets to two ways. First, based on standardized the group, those 16 would turn into 24 measures of group commitment in the total tickets for each subject. Subjects organizational literature (e.g., Price and made their decisions to keep or give the Mueller 1986), we measured the subject’s tickets in private and had no information intent to stay in the group on the post- on the contributions of others. It was only questionnaire. This measure consists of when the experiment was finished that a two-item index appearing on the post- subjects learned the outcome of their deci- questionnaire. The first item asks the sion and were given the tickets. The sub- subject to choose how many (if any) of jects were correctly told that their tickets the participants they would choose to would be placed in a random drawing to work with again. Subjects could respond take place after all study sessions were from zero to all three of the others. The completed, with the winner receiving second item consists of a 9-point Likert a $150 cash prize. This procedure has scale that asks, ‘‘If you could, would you been used by Batson et al. (1995) and like to work with the same participants others to create a basic social dilemma again?’’ Here the scale was anchored by and measure prosocial behavior. For our no, not at all and yes, definitely. The purposes, the number of tickets each sub- responses from these items were stan- ject donated to the group is a measure of group tie strength under conditions of dardized (i.e., converted into Z-scores) 12 before being combined into a single index. risk or possible defection. This index ranged from –1 to 11 (Cron- bach’s a = .69). Second, following the Phase II task, each subject was presented with a social 12 Of course, other factors contribute to cooper- dilemma choice: to keep or allocate 16 lot- ation levels in a social dilemma. Research indi- tery tickets for a $150 cash prize. Subjects cates that forces such as trust, fear, greed, and could allocate these in blocks of 8 to them- the shadow of the future all impact cooperation selves or to the group. In other words, the rates (Simpson 2003). While these are not manip- ulated or controlled in the current study, there is subject could keep all 16, keep 8 tickets no reason to believe these factors would vary across experimental conditions in the manner that affective ties do, especially given the large 11See Supplemental Documents in the online number of subjects and random assignment of version of the article. subjects to conditions. Formation of Group Ties 169

Table 1. Means and Analyses of Variance for Key Measures across Conditions

Experimental Conditions High Joint Task Low Joint Task No Joint Task GBI IBI GBI IBI GBI IBI

Affective tie to the group 6.56a 6.32a 6.30a 6.32a 4.99b 4.92b Jointness F(2, 120) = 52.1*** (.96) (.65) (.64) (.77) (.58) (.63) Incentive F(1, 120) = .52 Cognitive tie to the group 4.78a 4.65a 4.17a,b 3.68b 1.86c 1.65c Jointness F(2, 120) = 163.7*** (.60) (.94) (.96) (1.00) (.39) (.32) Incentive F(1, 120) = 4.07* Shared responsibility 7.12a 6.90a 6.30b 5.70b 3.17c 2.96c Jointness F(2, 120) = 163.1*** (.66) (.93) (1.10) (1.19) (.91) (1.17) Incentive F(1, 120) = 3.441 Positive emotion 6.34a 6.35a 6.33a 6.22a 6.22a 5.91a Jointness F(2, 120) = 1.69 (.95) (.83) (.78) (.65) (.69) (.83) Incentive F(1, 120) = .55 Number of cases 20 20 20 20 20 20

Note: GBI and IBI refer to group- and individual-based incentives, respectively. Standard deviations are in parentheses. Cells in a given row with different superscripted letters (a, b, c) are significantly different using Tukey’s Honest Significant Difference multiple range test for pairwise comparisons, controlling for the overall Type I error rate at a = .05. That is, within each row, values superscripted with an a are significantly different from values superscripted with b or c. 1p \ .07. *p \ .05. **p \.01. ***p \ .001 (two-tailed test).

RESULTS Task Jointness and Incentives: Hypotheses 1a and 1b The results are divided into three sec- tions: (1) analyses of variance to assess Table 1 shows means, standard devia- the hypotheses for the impact of joint tions, and analyses of variance (ANOVA) task and incentives on cognitive and results that test Hypotheses 1a and 1b. affective ties to the group, (2) multilevel Hypothesis 1a predicts that groups with structural equation modeling using joint tasks generate stronger cognitive Mplus version 7.11 (Muthe´n and Muthe´n and affective ties to the group than 2012) to test the mediating mechanisms groups without joint tasks; Hypothesis and effects on the nature and strength 1b predicts an impact of incentive struc- of group ties, and (3) supplementary anal- ture on cognitive but not affective ties to yses to examine other relevant measures. the group. Table 1 contains the means, Our primary purpose is to test the theo- ANOVA results, and post hoc Tukey Hon- retical hypotheses and theorized associa- est Significant Difference (HSD) tests tions between constructs of the Figure 1 (controlling the overall Type I error rate theory of social commitments. Support at a \ .05). for that theory would come in the form The ANOVA results support Hypothe- of statistically significant associations ses 1a and 1b. There was no interaction between those constructs, and we give between task jointness and incentive less emphasis to the size of the effect. structure, so we proceed directly to the While the focus of our experiment and main effects. There are significant main subsequent data analysis is on theory effects of task jointness on both affective testing, we also address issues related to ties to the group (F = 52.1, p \ .001) overall model fit. and cognitive ties to the group (F = 170 Social Psychology Quarterly 82(2)

163.7, p \ .001). Note that the two joint promotes perceptions of shared responsi- conditions (high and low) are significantly bility and positive emotion. The results different than the no joint condition. from an ANOVA on shared responsibility There is also a significant impact of support the theory (F = 163.1, p \ .001). group- versus individual-based incentives High jointness produces more sense of on cognitive ties to the group (F = 4.07, shared responsibility than low jointness, p \ .05) and, as expected, no effect on and low jointness produces more sense affective ties (F \ 1.0). More group-based of shared responsibility than the no joint- incentives produced stronger perceptual ness condition (Ms = 7.01, 6.00, and 3.06, overlap between self and group (i.e., cog- respectively, for high, low, and no joint nitive ties) compared to individual-based conditions). However, there are no signif- incentives. In sum, task jointness and icant effects of high jointness on positive group incentive structure have the emotion (F = 1.69, ns). The means for pos- expected effects. Group-based incentives itive emotion (Ms = 6.35, 6.27, 6.06, promote cognitive but not affective group respectively) are on the positive side of ties, whereas task jointness promotes the Likert scale, but none of these differ- both. ences are statistically significant. Post hoc tests elaborate the patterns Overall, the results for shared respon- across the six experimental conditions sibility affirm a central idea from the (see superscripts in Table 1). Here are theory—namely, that joint tasks generate the most important results. First, the a stronger sense of shared responsibility. two joint task conditions (low and high) No effects occur on positive emotion, but produce stronger affective ties to the the mean levels are on the positive side group than the no joint condition; of the bipolar dimension.13 Even though although the mean for high jointness is task jointness did not have the predicted higher than that for low jointness, the dif- effect on positive emotions, greater ference is not statistically significant shared responsibility may still lead to (combined Ms = 6.24 vs. 6.09). Task joint- more social unit attributions of the emo- ness (M = 4.32) also generates stronger tion that is present. We test for this possi- cognitive ties to the group compared to bility next. The most central theoretical the no joint conditions (M = 1.75), and prediction is an interaction effect between high jointness (M = 4.17) is significantly shared responsibility and positive emo- greater than the low joint condition tion, indicating that shared responsibility (M = 3.93). Second, the strength of cogni- leads people to attribute their positive tive ties is greater with group-based incen- emotion to the group. tives (M = 3.60) than individual-based incentives (M = 3.32). The post hoc tests Mechanisms: Hypotheses 2a and 2b affirm that the group versus individual incentives condition does not have a signifi- In our data, individuals are nested within cant effect on affective group ties. Although groups. Thus, to examine the causal the overall pattern of means based on post mechanisms linking task jointness and hoc tests is generally consistent with incentives to the strength of group ties, Hypotheses 1a and 1b, it is important to we used two-level random coefficients acknowledge that the differences between 13 the high and low joint conditions are not Task jointness effects on positive emotions may require repeated joint tasks, as has been statistically significant in all cases. the case in work finding positive effects of A key prediction from the theory of repeated exchange on positive emotions (Lawler social commitments is that task jointness et al. 2008; Lawler and Yoon 1996). Formation of Group Ties 171

Table 2. Multilevel Structural Equation Analysis for the Theoretical Model (Number of Groups = 120; Number of Individuals = 480)

Dependent Variables Social Intent Shared Positive Cognitive Affective Dilemma to Responsibility Emotion Tie Tie Contribution Stay

Group level Low joint task 2.48*** .05 —— — — (.29) (.19) High joint task 3.62*** .301 —— — — (.25) (.17) Group-based incentive — — .23 ——— (.24) Individual level Positive emotion (PE) — — — — — — Shared responsibility (SR) — — — — — — Interaction (PE 3 SR) — — .08*** .07*** —— (.02) (.02) Cognitive tie — — — — .01 .10* (.03) (.05) Affective tie — — — — .19*** .62*** (.03) (.05) Intercept –3.20*** –.36 –.38 .00 .00 .00 (.56) (.31) (.58) (.12) (.03) (.08) Variance estimate Group intercept residual .78** .37 1.31*** ——— Individual residual — — 1.81*** 1.37*** .56*** 1.32*** Akaike Information Criterion 9,589.81 Chi-square 499.69 (df = 30)***

Note: Low joint task and high joint task are dummy variables, and the omitted category is no joint task; group-based incentive is also a dummy variable, and its omitted category is individual-based incentive; the interaction between positive emotion and shared responsibility is created by mean centering the two variables; entries that correspond to predicting variables are unstandardized estimates; numbers in parentheses are standard errors; the effects of gender and research site are controlled; and the equation errors between affective tie and cognitive tie are correlated for the estimation. 1p \ .07. ***p \ .001 (two-tailed test). modeling to control for the nonindepen- conflating individual-level and unit-level dence of observations. We combined this effects (Zhang, Zyphur, and Preacher, approach with structural equation model- 2009). ing with full maximum likelihood estima- The first step in our analysis is to test tion (ML-SEM). ML-SEM decomposes the the theoretical model in Figure 1. The variance of each variable into two second step is to create a refined model components—individual-level variability that incorporates changes suggested by and a group-level variability. This ena- the modification indices from the first bles causal modeling of the relationships model. Table 2 shows the results of the among variables at each level. As such, initial model, strictly testing the theoreti- ML-SEM provides less biased estimates cal formulation in Figure 1, and Table 3 of coefficients by avoiding the problem of shows the results for the refined model. 172 Social Psychology Quarterly 82(2)

Table 3. Multilevel Structural Equation Analysis for the Refined Model (Number of Groups = 120; Number of Individuals = 480)

Dependent Variables Shared Positive Cognitive Affective Social Dilemma Intent to Responsibility Emotion Tie Tie Contribution Stay

Group level Low joint task 2.31*** –.32 1.98*** ——— (.28) (.16) (.15) High joint task 3.45*** –.05 2.74*** ——— (.25) (.17) (.15)

Group-based incentive — — .24* ——— (.12) Individual level Positive emotion (PE) — — — .47*** —— (.06) Shared responsibility (SR) — — — .13*** —— (.04) Interaction (PE 3 SR) — — .06** .07*** —— (.02) (.01) Cognitive tie — — — — — .16*** (.04) Affective tie — — — — .20*** .60*** (.03) (.05) Intercept –2.90*** .26 –1.82*** .00 .00 .00 (.54) (.28) (.31) (.12) (.03) (.08) Variance estimate Group intercept residual .73** .30 .04 — — — Individual residual — — 1.85*** .91*** .57*** 1.30*** Akaike Information Criterion 9,290.52 Chi-square 318.75 (df = 27) ***

Note: Added pathways are shown in bold. Low joint task and high joint task are dummy variables, and the omitted category is no joint task; group-based incentive is also a dummy variable, and its omitted category is individual-based incentive; the interaction between positive emotion and shared responsibility is created by mean centering the two variables; entries corresponding to predicting variables are unstandardized estimates; numbers in parentheses are standard errors; the effects of gender and research site are controlled; and the equation errors between affective tie and cognitive tie are correlated for the estimation. *p \ .05. **p \.01. ***p \ .001 (one-tailed test).

Most noteworthy, the results of both mod- the cognitive tie. These additional paths els support the central role of shared are shown in bold in Table 3. We also responsibility as a moderator of social dropped the from cognitive tie to unit attributions of the emotion. social dilemma contribution, as indicated The modification indices suggested the by the refined model. The results show addition of four paths and the dropping of that the refined model significantly one path. The four additional paths are as improves the goodness of fit (x2 = 499, follows: (1) low task jointness to the cogni- df = 30 in Table 2 vs. x2 = 318, df =27 tive tie, (2) high task jointness to the cog- in Table 3; x2 improvement = 181 saved nitive tie, (3) shared responsibility to the for 3 degrees of freedom, p \ .001; also affective tie, and (4) positive emotion to see the difference in Akaike Information Formation of Group Ties 173

Figure 3. Test of the Refined Theoretical Model.

Criterion). Tests of the hypothesized the- g = .06, p \ .01, respectively, for Tables oretical mechanisms (Hypotheses 2a and 2 and 3). As expected, the interaction 2b) and the consequences for the strength between shared responsibility and posi- of group ties (Hypotheses 3a and 3b) are tive emotion appears to facilitate the mainly based on this refined model. social unit attributions that generate We present these models together affective and cognitive group ties. Turn- because they are quite consistent and ing to the antecedents of positive emotion the modifications do not pose any serious and shared responsibility, task jointness problems to the original theoretical model significantly increases shared responsi- in Figure 1 (see also Figure 3). Tables 2 bility (g = 2.48 and 3.62, p \ .001, respec- and 3 contain coefficient estimates (g) tively, for low joint and high joint condi- for each test. The first two columns of tions in Table 2; g = 2.31 and 3.45, p \ the tables regress the effects of shared .001 in Table 3). However, the effects of responsibility and positive emotion on jointness on positive emotion differ the dummy variables for the all experi- between the high and low joint condi- mental conditions, and the next two col- tions. In the original theoretical model, umns add the interaction term. The the effect of high jointness on positive most significant support for Hypothesis emotion is positive and marginally signif- 2a is shown in Columns 3 and 4 of the icantly (g = .30, p \ .07, see Table 2), but two tables. The results reveal a significant this pattern is not replicated in the interaction of shared responsibility and refined theoretical model. positive emotion on affective group ties The aforementioned results and the (g = .07, p \ .001 and g = .07, p \ .001, Table 1 ANOVAs do not reveal major dif- respectively, for Tables 2 and 3) and cog- ferences between the low and high joint nitive group ties (g = .08, p \ .001 and conditions. In light of this, we conducted 174 Social Psychology Quarterly 82(2) a supplementary analysis combining the a positive direct effect on affective ties data from both the low and high joint con- (g = .13, p \ .001, see Table 3) but no ditions.14 These results are very similar to effects on cognitive ties. Positive emotions those previously described. Task joint- and perceptions of shared responsibility ness has effects on affective and cognitive play a role in generating affective ties ties through the interaction of shared beyond the social unit attribution process responsibility and positive emotions, and captured by the interaction effect of these group incentive has effects only on cogni- variables. Thus, the core idea about tive ties. The general implication is that it social unit attributions in the theory is does not take much task jointness to gen- affirmed, but the evidence suggests erate the predicted effects leading to a somewhat broader role for shared affective ties. Moreover, the impact of responsibility and positive emotions in the high/low joint conditions combined the promotion of affective group ties. In compared to the no joint condition reveals addition, it is noteworthy that the results the importance of the social interaction consistently indicate that group incen- component of joint tasks. Joint tasks are tives have no effect on perceptions of likely to have the strongest effect when shared responsibility even though mem- people engage in interaction to accom- bers perceive more connection to the plish them. The social interaction could group with stronger group incentives. help explain the one new finding in this reanalysis, specifically, that task joint- Impact on Strength of Group Tie: ness generated positive emotions (g = 1.81, p \ .001). This also may reflect the Hypotheses 3a and 3b boost in sample size and statistical power Hypotheses 3a and 3b predict that affec- that occurs when the two joint conditions tive ties will have stronger effects than are combined. cognitive ties on the capacity of groups Hypothesis 2b predicts that the effects to sustain membership and generate col- of group-based incentives on cognitive lective goods. This is tested by examining ties are direct and not mediated by the the effects of cognitive and affective ties interaction of shared responsibility and on inclinations to leave the group and positive emotions. The results in Table 3 cooperation in a social dilemma, control- support this hypothesis. Group incentives ling for all antecedent variables. The increase cognitive ties (g = .24, p \ .05), last two columns of Table 3 contain the and the results of the refined model indi- main results. As predicted, stronger cate that task jointness also generates affective ties increase both cooperation greater cognitive ties. in a social dilemma and intent to stay in The refined model reveals some effects the group (g = .20, p \ .001; g = .60, p \ for shared responsibility and positive .001), whereas cognitive ties increase emotion not predicted by our theory. First only the intent to stay in the group (g = of all, positive emotions from the task .16, p \ .001). The impact of cognitive activity directly strengthen affective ties ties on intent to stay (g = .16) is signifi- to the group (g = .47, p \ .001, see Table cantly smaller than the impact of 3), but such emotions have no impact affective ties (g = .60), and a test for on cognitive group ties. Second, percep- thedifferenceinthesizeofthetwo tions of shared responsibility also have effects is significant at p \ .01. This sup- ports Hypothesis 3a. Hypothesis 3b is 14The specific results are not reported here but also supported; the impact of affective can be provided on request. ties on social dilemma contribution is Formation of Group Ties 175 significant, but that of cognitive ties is secondary issue is whether or how well not significant.15 the model fits the data. Although no stan- To conclude, the ANOVA and ML-SEM dardized measure of goodness of fit exists results provide support for the broad for multilevel structural equation model- principles of the theory of social commit- ing (SEM), there are some ways to access ments. Greater task jointness generates model fit. To check this, we aggregated stronger cognitive and affective group the data from each group, treating the ties, a greater sense of shared responsibil- group mean on all measures as a single ity, and thus social unit attributions of unit of observation (N = 120), and esti- emotion. Importantly, the experimental mated a single-level SEM. In this model, condition with the highest task jointness the standardized root mean square resid- produces the highest scores on virtually ual (SRMR) = .083, and the Comparative all relevant measures, and the no task Fit Index (CFI) = .80. These results indi- jointness condition always generates the cate a marginal fit as fitness is typically lowest levels. The high and low joint condi- indicated by SMSR \ .05 and CFI . .90. tions do not differ in all cases, which may To further understand this, we used LIS- suggest that it does not take much task REL 7 to estimate each moment in the jointness to produce the sort of effects pre- theoretical model to check the associated dicted by our theory; social interaction itself R2 values. Those R2 values are as follows: may be sufficient to generate a degree of positive emotion (.022), shared responsi- jointness. Overall, comparing the role of bility (.73), affective tie to the group task jointness to group-based incentives, (.66), cognitive tie to the group (.75), task jointness plays a larger role as pre- social dilemma choice (.085), and intent dicted by our theorizing. The more limited to stay (.46). Overall, the data suggest role for group incentive structures stands that the source of the marginal model fit in stark contrast to the common, prevailing emanates from positive emotion and notion in organizational-related fields that social dilemma choice. To strengthen group incentive schemes are the most model fit, future investigators may seek important condition for promoting group alternative ways to model these con- ties and commitment to work organizations. structs and other ways to theorize the process. Assessing Group-Level Goodness DISCUSSION of Fit Overall, the data from the experiment Questions about the ties of individuals to support the hypotheses produced by the social units have been central to sociology theory of social commitments. A from its inception (Durkheim [1933] 1964; Parsons 1951; Simmel 1964; Weber 15An additional possibility for why our mea- [1918] 1968). For the classical scholars, sure of affective ties had stronger, more robust the links between micro and macro phe- effects is because that measure consisted of nomena boil down to whether these are a four-item index while the measure of cognitive based primarily on the interests and ties consisted of a single item. The additional measurement error in the measure of cognitive rationality of individual actors or the sen- ties could be sufficient to attenuate its relation- timents and norms of communities (e.g., ship with any other variable. As a robustness Durkheim [1933] 1964). The former check, we estimated the model predicting cogni- assumes individuals cognitively assess tive ties while including affective ties as a covari- ate. The results do not differ and are available situations and make choices that are from the author on request. best for them. Group ties are important, 176 Social Psychology Quarterly 82(2) but these social units are purely instru- (see Lawler et al. 2008). In the following, ments of individual gain. The latter we elaborate the contributions by high- assumes individuals enact and repeat lighting several implications of the established scripts, patterns, or recipes research. for producing a socially legitimate result. The first implication concerns the role Our theory and research revisits this of affective and cognitive group ties. The larger, more fundamental issue about prevailing theoretical principle in much the ties that people forge to social units sociological work is that group ties are and uses a small groups experiment to fundamentally instrumental and self- examine these broader issues empirically. interest based. Groups form, endure, There are three general contributions and command commitment so long as of this research. First, the research they serve the interests of individuals reveals fundamental ways that people within them. On a practical level, the develop ties to a group and the behavioral implied advice is that if you are forming consequences of such ties, namely, stay- or leading a group, first and foremost, it ing in the group and cooperating when is critical to provide individuals rewards presented with a social dilemma. The and benefits that they value and align focus on and analysis of person-to-group their individual interests with the group’s ties distinguishes this research from collective interests or goals. Our research most other work on social or relational qualifies this conventional wisdom about ties (e.g., Hechter 1987; Lawler and interest alignment by showing that affec- Yoon 1993, 1996). Second, the research tive ties generate more resilient group expands the scope of the theory of social ties and these ties form through distinct commitments by demonstrating the mechanisms. impact of affective and cognitive ties in The second implication concerns the a four-person task group where people role of tasks that people accomplish in interact freely to accomplish a joint task. interaction with others. Tasks link social It is important to note that unlike structures to interactions and interac- research in exchange settings, actors tions to structures. This paper empha- reveal the effects of emotion on group sizes a particular dimension of tasks— ties in the absence of (prior to) informa- their jointness or propensity to generate tion on their success at the task.16 Third, a sense of shared responsibility among this research demonstrates the role of individuals. This research shows that shared responsibility as a crucial condi- joint tasks have important effects, inde- tion for the development of affective pendent of structural incentives, on the group ties. This is the first explicit test propensity of people to form affective of the notion that shared responsibility group ties. Perceptions of shared respon- promotes social unit attributions of indi- sibility play a critical role. Interest align- vidual emotions and feelings from a joint ment promotes a cognitive bond and task, namely, its moderating effect. Ear- reduces propensities to leave the group, lier evidence is only inferential or indirect but it does not necessarily foster a greater sense of shared responsibility or a procliv- 16In related social exchange research (see ity to cooperate when social dilemmas Lawler et al. 2008; Lawler and Yoon 1996), suc- arise. cess at the exchange task (i.e., repeated The third implication, building on the exchanges) cannot be separated from the affective process leading to commitment. In this current second, concerns the conditions and research, the emotional process occurs and is mechanisms under which joint tasks and measured prior to information on task success. group incentives generate person-to- Formation of Group Ties 177 group ties. The theoretical guidance here affective ties to the group. In the case of comes from a theory of social commit- incentives, group-based incentives pro- ments (Lawler et al. 2009). The theory duced closer connections to the group predicts that task interactions produce than individual-based incentives but stronger affective group ties under two have no effects on affective group ties. conditions. First, task interactions gener- In addition, the effects of task jointness ate positive individual emotions. These and incentives on the strength of the feelings can arise from the task activity group tie—intent to stay and cooperation itself and are not necessarily contingent in a social dilemma—are indirect. They on the success of the task activity or operate differentially through the affec- actors’ knowledge of that success, as dem- tive or cognitive ties. Affective ties reduce onstrated here. Second, actors perceive intentions to leave the group and increase a shared responsibility for the task. This inclinations to cooperate in a social sense of shared responsibility may stem dilemma with other group members, from the structure of the task itself, an whereas cognitive ties produce effects on already given institutional definition of the intent to leave but not cooperation that task, or actors’ interpretation of the in a social dilemma. These effects are con- task as they undertake it. Shared respon- sistent with our theoretical predictions sibility transforms individual emotions or and the more general idea that affective feelings into sentiments about the ties provide a stronger person-to-group group. Importantly, this study explicitly bond. tests this moderating role for the sense Our research also raises questions for of shared responsibility and thus provides future investigators. For instance, the evidence for the core mechanism of the degree of task jointness necessary for theory. Moreover, it does so in a context the development of affective group ties that approximates the interactions that is an unresolved issue in the current occur in naturally occurring, everyday study because there are few significant groups, committees, or teams. These two differences between high and low joint- findings suggest that tasks that generate ness conditions. It is quite possible that a sense of shared responsibility are an face-to-face interaction itself is sufficient emergent force for social order in a broad to generate the levels of jointness and array of groups (Lawler et al. 2015). sense of shared responsibility that To assess the independent effects of unleash the process theorized here (see joint tasks, the research orthogonally Ma¨s and Dijkstra 2014). We did find manipulated both the incentive structure that perceptions of shared responsibility (individual or group based) and jointness were higher in the high joint condition of the task. Theoretically, the most telling than in the low joint condition, but this comparison, given the hypotheses, is did not produce differences in affective between a context in which actors work ties or the behaviors that reflect stronger side by side and complete the task inde- group ties. The issue of how face-to-face pendently (no task jointness) and a situa- interaction, net of other factors, alters tion where actors interact freely and pro- the process of developing group ties duce a single, joint task response (high awaits further investigation. task jointness). The comparison between The evidence indicates that group these two conditions strongly supports incentives play a more prominent role in our theoretical predictions. Actors with producing cognitive rather than affective a joint task perceived closer connections ties. However, it is important to acknowl- to the group and developed stronger edge that affective and cognitive ties are 178 Social Psychology Quarterly 82(2) not completely distinct or separable. Any emotions from microlevel experiences to affective tie involves cognitive compo- meso- or macrolevel entities (Turner nents, just as any cognitive tie is likely 2007). Of note, none of these processes to have some underlying emotional ele- necessarily require a social dilemma as ments. On an analytic level, affective the problem of order is usually framed and cognitive ties are more distinct than (see also Lawler et al. 2015). Problems they likely are on an empirical level. of social order can and do occur in the Our study measured and statistically absence of a social dilemma. Social dilem- analyzed cognitive and affective path- mas are indeed important contexts for ways without necessarily eliminating examining the order problem, but they their interrelatedness. While we believe are not the quintessential representation IOS (Inclusion of Other in Self) is a good of the Hobbesian problem as often measure of a perceived tie to the group assumed. Problems of social order that on conceptual grounds, future research occur in the absence of social dilemmas should develop measures that are more warrant more systematic attention by empirically distinct, consulting the field scholars, and as our research suggests, of psychometrics. understanding the role of emotion and Returning to the theme of social order, affect is likely to be crucial. our theory and research, along with recent work by others (e.g., see Burke ACKNOWLEDGMENTS and Stets 2015; Heise, MacKinnon, and Scholl 2015; Turner 2015), points to an The order of authorship is random and does not reflect differential contributions. This material important overarching conclusion: one is based on work supported by the National Sci- solution to the problem of social order is ence Foundation under Collaborative Grant affectively laden and emotionally rich Numbers SBR-9817706 and SBR-9816259 to the person-to-group ties. Recall our experi- University of South Carolina and Cornell Univer- mental groups engage in face-to-face sity. We thank Ashley Harrell, Will Kalkhoff, Jennifer McLeer, and Mostafa Mobli for com- interaction to solve a problem. People in ments and suggestions on an earlier draft. such groups deal with uncertainty along many dimensions, for example, about an SUPPLEMENTAL MATERIAL unfamiliar task, about each other’s skills or competencies given they have no his- Supplemental material for this article is available tory, or how to combine their competen- online. cies to generate a successful result. In such a context, social order must be REFERENCES socially constructed and engineered along Allen, Bellinda C., Leisa D. Sargent, and Lisa the way. In the small groups literature, M. Bradley. 2003. ‘‘Differential Effects of this process of construction is addressed Task and Reward on Interdependence on Perceived Helping Behavior, Effort, and implicitly in Bales’s classic work on emer- Group Performance.’’ Small Group gent status structures and in expectation Research 34:716–40. states theory (Berger, Cohen, and Zel- Aron, Arthur, Elaine N. Aron, and Danny ditch 1972; Berger et al. 1977, 1992; Smollan. 1992. ‘‘Inclusion of Other in the Ridgeway 2011). Moreover, recent work Self Scale and the Structure of Interper- sonal Closeness.’’ Journal of Personality argues that social order is generated or and Social Psychology 63(4):596–612. sustained by the subtle coordination of Aron, Arthur, and Barbra Fraley. 1999. ‘‘Rela- meanings (Rawls 2010), through mutual tionship Closeness as Including Other in verification of identities (Burke and Stets the Self: Cognitive Underpinnings and 2009), or by the spread of positive Measures.’’ Social Cognition 17(2):140–60. Formation of Group Ties 179

Axelrod, Robert. 1984. The Evolution of Coop- Cook, Karen S., and Richard M. Emerson. eration. New York: Free Press. 1984. ‘‘Exchange Networks and the Analy- Bales, Robert F. 1950. Interaction Process sis of Complex Organizations.’’ Research Analysis: A Method for the Study of Small in the Sociology of Organizations 3(4):1–30. Groups. Boston: Addition-Wesley. Durkheim, Emile. 1915. The Elementary Batson, C. Daniel, Tricia R. Klein, Lori High- Forms of the Religious Life. Translated by berger, and Laura L. Shaw 1995. ‘‘Immoral- J. W. Swain. Mineola, NY: Dover. ity from Empathy-Induced Altruism: When Durkheim, Emile. [1933] 1964. The Division Compassion and Justice Conflict.’’ Journal of Labor in Society. New York: Free of Personality and Social Psychology Press. 68:1042–54. Emerson, Richard. M. 1972a. ‘‘Exchange The- Benford, Robert D., and David A. Snow. 2000. ory. Part I: A Psychological Basis for Social ‘‘Framing Processes and Social Movements: Exchange.’’ Pp. 38–57 in Sociological Theo- An Overview and Assessment.’’ Annual ries in Progress. Vol. 2, edited by J. Berger, Review of Sociology 26:611–39. M. Zelditch, Jr., and B. Anderson. Boston: Berger, Joseph, Bernard P. Cohen, and Morris Houghton Mifflin. Zelditch, Jr. 1972. ‘‘Status Characteristics Emerson, Richard M. 1972b. ‘‘Exchange The- and Social Interaction.’’ American Sociolog- ory. Part II: Exchange Rules and Net- ical Review 37:241–55. works.’’ Pp. 58–87 in Sociological Theories Berger, Joseph, M., Hamit Fisek, Robert Z. in Progress. Vol. 2, edited by J. Berger, M. Norman, and Morris Zelditch, Jr. 1977. Zelditch, Jr., and B. Anderson. Boston: Status Characteristics and Social Interac- Houghton Mifflin. tion. New York: Elsevier Faris, Robert, and Diane Felmlee. 2014. Berger, Joseph, Robert Z. Norman, James W. ‘‘Casualties of Social Combat: School Net- Balkwell, and Randall F. Smith. 1992. ‘‘Sta- works of Peer Victimization and Their Con- tus Inconsistency in Task Situations: A sequences.’’ American Sociological Review Test of Four Status Processing Principles.’’ 79:228–57. American Sociological Review 57:843–55. Gorman, Elizabeth H., and Peter V. Marsden. Berger, Peter L., and Thomas Luckmann. 2001. ‘‘Social Networks, Job Changes, and 1966. The Social Construction of Reality. Recruitment.’’ Pp. 467–502 in Sourcebook Garden City, NY: Anchor. of Labor Markets: Evolving Structures and Brewer, Marilynn B., and Wendi Gardner. Processes, edited by I. Berg and A. L. Kalle- 1996. ‘‘Who Is This ‘We’? Levels of Collec- berg. New York: Kluwer Academic/Plenum. tive Identity and Self-Representations.’’ Gottschalg, Oliver, and Maurizio Zollo. 2007. Journal of Personality and Social Psychol- ‘‘Interest Alignment and Competitive Advan- ogy 71(1):83–93. tage.’’ Academy of Management Review Burke, Peter J., and Jan E Stets. 2009. Iden- 32:418–37. tity Theory. Oxford: Oxford University Granovetter, Mark. 1985. ‘‘Economic Action Press. and Social Structure: The Problem of Burke, Peter J., and Jan E. Stets. 2015. ‘‘Iden- Embeddedness.’’ American Journal of Soci- tity Verification and the Social order.’’ Pp. ology 91:481–510. 15–64 in Order on the Edge of Chaos: Social Hechter, Michael. 1987. Principles of Group Psychology and the Problem of Social Solidarity. Berkeley: University of Califor- Order, edited by E. J. Lawler, S. R. Thye, nia Press. and J. Yoon. Cambridge: Cambridge Uni- Heise, David R., Neil J. MacKinnon, and Wolf- versity Press. gang Scholl. 2015. ‘‘Identities, Roles, and Cappelli, Peter. 1999. The New Deal at Work: Social Institutions.’’ Pp. 165–88 in Order Managing the Market-Driven Workforce. on the Edge of Chaos, edited by E. J. Law- Boston: Harvard Business School Press. ler, S. R. Thye, and J. Yoon. Cambridge: Collins, Randall. 2004. Interaction Ritual Cambridge University Press. Chains. Princeton, NJ: Princeton Univer- Hobbes, Thomas. [1651] 1985. Leviathan. New sity Press. York: Penguin Books. Colvin, Alexander, and Wendi R. Boswell. Homans, George C. 1961. Social Behavior: Its 2007. ‘‘The Problem of Action and Interest Elementary Forms. New York: Harcourt Alignment: Beyond Job Requirements and Brace and Jovanovich. Incentive Compensation.’’ Human Resource Johnson, David W., and Frank P. Johnson. Management Review 17:38–51. 2009. Joining Together: Group Theory and 180 Social Psychology Quarterly 82(2)

Group Skills. 10th ed. Upper Saddle River, Behavior in Negotiated Exchange.’’ Ameri- NJ: Pearson. can Sociological Review 58:465–81. Keyton, Joanne. 2016. ‘‘The Future of Small Lawler, Edward J., and Jeongkoo Yoon. 1996. Group Research.’’ Small Group Research ‘‘Commitment in Exchange Relations: Test 47:134–54. of a Theory of Relational Cohesion.’’ Ameri- Kollock, Peter. 1994. ‘‘The Emergence of can Sociological Review 61:89–108. Exchange Structures: An Experimental Lawler, Edward J., and Jeongkoo Yoon. 1998. Study of Uncertainty, Commitment, and ‘‘Network Structure and Emotion in Trust.’’ American Journal of Sociology Exchange Relations.’’ American Sociologi- 100:313–45. cal Review 63:871–94. Komorita, Samuel S., and Craig D. Parks. Ma¨s, Michael, and Jacob Dijkstra. 2014. ‘Do 1994. Social Dilemmas. Boulder, CO: West- Intergroup Conflicts Necessarily Result view Press. from Outgroup Hate?’’ PLoS ONE 9(6): Kramer, Roderick M., and Marilynn B. e97848. Brewer. 1984. ‘‘Effects of Group Identity McLeer, Jennifer, Jake Frederick, Barry Mar- on Resource Use in a Simulated Social kovsky, and Christopher Barnum. 2011. Dilemma.’’ Journal of Personality and ‘‘Standardizing Open Interaction Coding Social Psychology 46(5):1044–57. for Status Processes.’’ Advances in Group Kuwabara, Ko. 2011. ‘‘Cohesion, Cooperation, Processes 28:33–58. and the Value of ‘Doing Things Together’: Mead, George Herbert. 1934. Mind, Self and How Economic Exchange Creates Rela- Society.Chicago:UniversityofChicagoPress. tional Bonds in Exchange Relations.’’ Molm, Linda D. 2003. ‘‘Theoretical Compari- American Sociological Review 74(4):560– sons of Forms of Exchange.’’ Sociological 80. Theory 21:1–17. Lawler, Edward J. 2001. ‘‘An Affect Theory of Muthe´n, Linda K., and Bengt O. Muthe´n. Social Exchange.’’ American Journal of 2012. Mplus User’s Guide (version 7). Los Sociology 107:321–52. Angeles, CA: Muthen and Muthen. Lawler, Edward J., Shane R. Thye, and Jeong- Nyberg, Anthony J., Ingrid Smithey Fulmer, Yoon. 2000. ‘‘Emotion and Group Cohe- Barry Gerhart, and Mason A. Carpenter. sion in Productive Exchange.’’ American 2010. ‘‘Agency Theory Revisited: CEO Journal of Sociology 106:616–57. Return and Shareholder Interest Align- Lawler, Edward J., Shane R. Thye, and Jeong- ment.’’ Academy of Management Journal koo Yoon. 2006. ‘‘Commitment in Structur- 53:1029–49. ally Enabled and Induced Exchange Parsons, Talcott. 1951. The Social System. Relations.’’ Social Psychology Quarterly New York: Free Press. 69:183–200. Prentice, Deborah A., Dale T. Miller, and Jeni- Lawler, Edward J., Shane R. Thye, and Jeong- fer R. Lightdale. 1994. ‘‘Asymmetries in koo Yoon. 2008. ‘‘Social Exchange and Attachments to Groups and to Their Mem- Micro Social Order.’’ American Sociological bers: Distinguishing between Common- Review 73:519–42. Identity and Common-Bond Groups.’’ Per- Lawler, Edward J., Shane R. Thye, and Jeong- sonality and Social Psychology Bulletin koo Yoon. 2009. Social Commitments in 20:484–93. a Depersonalized World. New York: Russell Price, Heather E., and Jessica Collett. 2012. Sage Foundation. ‘‘The Role of Exchange and Emotion on Lawler, Edward J., Shane R. Thye, and Jeong- Commitment: A Study of Teachers.’’ Social koo Yoon. 2014. ‘‘The Emergence of Collec- Science Research 41(6):1469–79. tive Emotions in Social Exchange.’’ Price, James L., and Charles W. Mueller. Pp. 189–203 in Collective Emotions, edited 1986. Absenteeism and Turnover of Hospi- by C. V. Scheve and M. Salmela. Oxford: tal Employees. Greenwich, CT: JAI Press. Oxford University Press. Rawls, Anne Warfield. 2010. ‘‘Social Order as Lawler, Edward J., Shane R. Thye, and Jeong- Moral Order.’’ Pp. 95–121 in Handbook of koo Yoon eds. 2015. Order on the Edge of the Sociology of Morality, edited by S. Chaos: Social Psychology and the Problem Hitlin and S. Vaisey. New York: Springer. of Social Order. Cambridge: Cambridge Ridgeway, Cecilia L. 2011. Framed by Gender: University Press. How Gender Inequality Persists in the Mod- Lawler, Edward J., and Jeongkoo Yoon. 1993. ern World. Oxford: Oxford University ‘‘Power and the Emergence of Commitment Press. Formation of Group Ties 181

Simmel, Georg. 1964. The Sociology of Georg Willer, David. 1999. Network Exchange The- Simmel. Edited by K. H. W. Wolff. Glencoe: ory. Westport, CT: Praeger. The Free Press. Zhang, Zhen, Michael J. Zyphur, and Kristo- Simpson, Brent. 2003. ‘‘Sex, Fear, and Greed: pher J. Preacher. 2009. ‘‘Testing Multilevel A Social Dilemma Analysis of Gender and Mediation Using Hierarchical Linear Mod- Cooperation.’’ Social Forces 82(1):35–52. els: Problems and Solutions.’’ Organiza- Smith-Lovin, Lynn, John V. Skvoretz, and tional Research Methods 12:695–719. Charlotte G. Hudson. 1986. ‘‘Status and Participation in Six-Person Groups: A Test BIOS of Skvoretz’s Comparative Status Model.’’ Social Forces 64:992–1005. Shane Thye is a professor of sociology at Swidler, Ann. 1986. ‘‘Culture in Action: Sym- bols and Strategies.’’ American Sociological the University of South Carolina and the Review 51:273–86. series coeditor (with Professor Edward J. Tajfel, Henry, and John C. Turner. 1986. ‘‘The Lawler) of Advances in Group Processes. Social Identity Theory of Intergroup Rela- His work has appeared in outlets such tions.’’ Pp. 7–24 in Psychology of Intergroup as the American Sociological Review, Relations, edited by S. Worchel and L. W. Austin. Chicago: Nelson-Hall. American Journal of Sociology, and Taylor, Catherine J., and Karl Pillemer. 2009. Social Forces. Professor Thye is currently ‘‘Using Affect to Understand Employee conducting research on status processes Turnover: A Context Specific Application in bargaining relations and the social psy- of a Theory of Social Exchange.’’ Sociologi- chological and neural bases of micro cal Perspectives 52(4):481–504. Thye, Shane R. 2000. ‘‘A Status Value Theory social order. of Power in Exchange Networks.’’ American Sociological Review 65:407–32. Edward J. Lawler is the Martin P. Thye, Shane R., Edward J. Lawler, and Jeong- Catherwood Professor Emeritus of Indus- koo Yoon. 2011. ‘‘The Emergence of Embed- trial Relations at Cornell University, ded Relations and Group Formation in Net- works of Competition.’’ Social Psychology coauthor (with Thye and Yoon) of Social Quarterly 74:387–413. Commitments in a Depersonalized World Thye, Shane R., Aaron Vincent, Edward J. (2009), and coeditor (also with Thye and Lawler, and Jeongkoo Yoon. 2014. ‘‘Rela- Yoon) of Order on the Edge of Chaos: tional Cohesion, Social Commitments, and Social Psychology and the Problem of Person-to-Group Ties: Twenty-Five Years of a Theoretical Research Program.’’ Social Order (2015). His theory and Advances in Group Processes 31:99–138. research deals with emotional founda- Thye, Shane R., Jeongkoo Yoon, and Edward tions of order and stability in exchange J. Lawler. 2002. ‘‘The Theory of Relational relations and groups. Cohesion: Review of a Research Program.’’ Advances in Group Processes 19:89–102. Turner, Jonathan H. 2007. Human Emo- Jeongkoo Yoon is a professor in the tions: A Sociological Theory. New York: School of Business Administration at Routledge. Ewha Womans University in Seoul, Turner, Jonathan H. 2015. ‘‘The Evolution of South Korea. His research focuses on the Social Mind: The Limitations of Evolu- power and exchange theories, with partic- tionary Psychology.’’ Pp. 177–93 in Hand- book on Evolution and Society, edited by ular application to leadership and organi- A. Maryanski, R. Machalek, and J. H. zational change. With Edward J. Lawler Turner. New York: Routledge. and Shane R. Thye, he has developed Watts, Duncan J. 2003. Six Degrees: The Sci- the research program of relational cohe- ence of a Connected Age. New York: Norton. sion theory. He also published such books Weber, Max. [1918] 1968. ‘‘Conceptual Exposi- tion.’’ Pp. 956–1005 in Economy and Soci- as Authentic Leadership, Management by ety, edited by G. Roth and C. Wittich. Ber- Purpose (2015), and Change Management keley: University of California Press. for Sustainable Companies (2010).