Creating Cultural Synergy

finishit

GROUPSINEQUALITYANDSYNERGY.pdf

Home >Business & Finance homework help >Management homework help >Creating Cultural Synergy

Groups, Inequality, and Synergy Bianca Manago, Jane Sell, Carla Goar

Social Forces, Volume 97, Number 3, March 2019, pp. 1365-1388 (Article)

Published by Oxford University Press

For additional information about this article

Access provided at 19 Nov 2019 20:32 GMT

https://muse.jhu.edu/article/720301

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Groups, Inequality, and Synergy

Bianca Manago, Vanderbilt University Jane Sell, Texas A&M University Carla Goar, Kent State University

W e combine insights from the sociological and psychological traditions of social psychology to examine how status inequality may affect group per- formance. Specifically, we examine how synergy, that is, performance

gains produced through group interaction, is realized in ethnically diverse groups. Research suggests that diverse work groups struggle to capitalize on the strengths of all group members, and in turn, achieve synergy less frequently than homoge- neous groups. We examine how an intervention in the status generalizing process af- fects group members’ influence—and how this affects synergistic gains (n = 50). We study these processes over a three-week period of time. In groups that received an intervention in the status generalizing process (experimental condition), racial hierar- chy is decreased in weeks 1 and 3—but inequality remained in baseline groups. Additionally, in week 3, experimental groups achieved synergy more than baseline groups and did not perform worse than baseline groups on any task.

This project combines psychological and sociological social psychology to analyze the effect of task definition on group performance in ethnically diverse groups. Our project synthesizes the theory of status characteristics and expectation states (from sociology) with theories of group productivity (from psychology) to develop predictions about the effect of status hierarchy on group performance gains, that is, synergy. Using a multi-week laboratory experiment, we test whether varying the definition of the task can decrease racial/ethnic inequality in group in- teractions. Finally, we test our predictions about the relationship between group inequality and group performance.

Developed in the tradition of psychological social psychology, synergy is a group-level phenomenon that describes how successful a group is at answering or completing tasks (Hackman 1987; Larson 2010). Specifically, the concept of

This research was supported in part through NSF Grant 0961940 awarded to the last two authors. The authors would like to thank the departments of sociology at Kent State University and Texas A&M University for their support. The authors are especially appreciative of Luis Romero, Calixto Melero, and Whitney Hoft for helping conduct the experiments. We would also like to thank Charles Samuelson for his feedback during the formative stages of this paper. Finally, the authors are grateful to the reviewers whose feedback has improved the quality of their work. Direct correspondence to Bianca Manago, Department of Sociology, Vanderbilt University, PMB 351811, Nashville, TN 37235-1811, USA; e-mail: [email protected]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Social Forces 97(3) 1365–1388, March 2019 doi: 10.1093/sf/soy063

Advance Access publication on 11 July 2018

Groups, Inequality, and Synergy 1365

mailto: [email protected]

synergy is used to determine whether group interaction increases group perfor- mance by examining whether a group outperforms its average and/or best mem- ber. Rooted in sociological social psychology, group process research has been less interested in such judgment problems, but rather, the focus is largely upon group interactions per se (for discussion, see Berger [2014]). By bridging litera- tures from sociology and psychology, our project makes predictions about the effect of racial/ethnic inequality on group processes and group performance.

Experimental research has been justifiably criticized for the lack of serious atten- tion to race and ethnicity (Goar and Sell 2009; Goar et al. 2013; Hunt et al. 2013). Using experimental methods, this project examines how race/ethnicity affects both group interaction and group performance. We address groups composed of women who self-identify as either White or Mexican American. We examine: (1) whether the defined complexity of a task reduces inequality in a group; and (2) whether a decrease in group inequality increases synergy. We posit that decreasing inequality will help increase synergy within groups. This should be the case because decreasing hierarchy enables open consideration of more people’s ideas.

In our study, decreased hierarchy is achieved not from the imposition of a spe- cific decision-making strategy (e.g., Curşeu, Jansen, and Chappin 2013), or from the addition of further information about group members, but from definitions of the task itself. Specifically, we compare the processes and performance of groups that received the inconsistent complexity intervention, which emphasizes the complexity of the task and diversity of opinions needed to solve that task, to groups that did not receive such an intervention. We find that for the last task completed, groups who received the inconsistent complexity intervention have less hierarchy and also a higher probability of achieving synergy.

Group Performance and Synergy Under some conditions, groups can perform better than individuals alone. Famously, Condorcet’s jury theorem states that a jury decision is more likely to be correct than the decision of a single person as long as the individual jurors are correct more than half the time. The social processes by which groups would do better than individuals reside on some well-examined processes such as error checking (Laughlin 2011; Laughlin and Ellis 1986) and contributing multiple perspectives (Kerr and Tindale 2012). However, the way researchers measure “improved group performance” has been debated.

The concept of synergy provides a way to delineate whether and to what extent groups might perform better than individuals. Synergy is “a gain in per- formance that is attributable in some way to group interaction” (Larson 2010, 4). There are two kinds of synergy: weak synergy and strong synergy. Weak syn- ergy occurs when collective performance is better than the average performance of group members working individually; and strong synergy occurs when the group performs better than the highest-performing individual (Larson 2010). So, to determine if a group achieved weak and strong synergy, one must measure the collective group task performance in comparison to the average and high performance of individual group members, respectively.

1366 Social Forces 97(3)

For decision-making tasks, synergy is determined by the group’s ability to identify and offer the best collaborative group response. However, the decision- making process can be difficult because there are issues of coordination, idea assessment, and discussion. Thus, for decisions in which there are multiple inter- dependent outcomes, group superiority over individual efforts is certainly not guaranteed (Banerjee 1992).

Indeed, groups frequently demonstrate process losses; that is, they perform lower than their most capable member (Kerr and Tindale 2004). There is a long history of research on this phenomenon and exceptions to it (Kerr and Tindale 2004; Stasser and Titus 2003).1 One reason for process loss is articulated through extensions of the original Social Decision Scheme (SDS) Model. For example, the Social Judgment Scheme demonstrates that group members tend to talk more about the information that all group members know, rather than seek- ing out unique information by different group members (Davis et al. 1989; Kameda et al. 2012). This is thought to occur because of the “herding” phenom- ena or conformity to “popular” or common information.

The majority of SDS theories do not consider who offers the suggestions, but rather how many group members offer the solution. In particular, SDS theory and its associated models rarely examine how an individual’s status may impact the likelihood that his/her answer will become the majority answer. In this proj- ect, we intentionally focus on group members’ statuses and examine how group processes affect hierarchy and performance.

Similarity, Diversity, and Inequality in Groups Broadly defined, diversity refers to differences that exist between and among group members. In the group performance literature from psychology, diversity is conceptualized in two ways: deep level and surface level. Deep-level diversity is defined as the breadth of difference that exists among those characteristics that pertain to specific talents, abilities, and knowledge that are not easily per- ceptible without extended interaction. On the other hand, surface-level diversity (sometimes labeled as demographic diversity) is that which differs on easily observable categorical characteristics such as age, gender, and race/ethnicity (Harrison et al. 2002; Harrison, Price, and Bell 1998; Larson 2010; Pelled 1996; Pelled, Eisenhardt, and Xin 1999; Phillips 2014; Tsui, Egan, and O’Reilly 1992).

Past research suggests that deep-level diversity in groups can be advantageous because group members with different skills, abilities, experiences, and perspec- tives can provide a broad range of available member resources for groups to pull from (Horwitz and Horwitz 2007; Jehn, Northcraft, and Neale 1999; Joshi and Roh 2009). However, the impact of surface-level diversity remains unclear. While some research finds that diverse workgroups outperform homogeneous workgroups (Phillips 2014; Phillips, Northcraft, and Neale 2006; Sommers 2006), other research finds that surface-level diversity has either a negative or neutral impact on group performance (Larson 2010; Levine and Moreland 1990; van Dijk, van Engen, and van Knippenberg 2012; Watson, Kumar, and

Groups, Inequality, and Synergy 1367

Michaelsen 1993). Most researchers posit that the negative consequences of surface-level diversity occur because these groups may have more difficulty achieving social integration—that is, quality interpersonal relationships and pos- itive affect toward group members—and this may negatively impact communi- cation and cooperation (Jehn, Northcraft, and Neale 1999; Pelled, Eisenhardt, and Xin 1999). However, more research is needed to understand the reason for the lower performance of groups with surface-level diversity.

In contradistinction to these findings, Herring finds that gender and racial diversity in business organizations is positively related to business outcomes such as sales revenue, number of customers, market share, and relative profits (Herring 2009, 2017). Further, a meta-analysis of diverse work groups suggests that at least some of the negative outcomes associated with racially diverse groups may be due to observers’ biased perceptions of group interactions (e.g., perceived performance or conflict) rather than objective measures of such interaction (e.g., number of ideas generated) (van Dijk, van Engen, and van Knippenberg 2012; Lount et al. 2015). This difference in research findings underscores the need to examine the behavioral group processes involved in the interactions. We argue that by examin- ing status processes, we can improve the study and understanding of diverse work groups. Specifically, we look to the theory of status characteristics and expectation states (SCES) to explain how the status characteristics of the group members may affect group processes and, in turn, group performance.

Status Characteristics and Expectation States Theory The vast literature on the theory of status characteristics and expectation states supports the principle that, other things being equal, group interaction is dic- tated by group members’ differentiating characteristics. There are two kinds of status characteristics, specific and diffuse, which align closely with psychological concepts of deep- and surface-level diversity, respectively (Harrison et al. 2002; Harrison, Price, and Bell 1998; Larson 2010; Pelled 1996). Specific status char- acteristics are characteristics that involve specific task abilities and so create spe- cific performance expectations for the individuals possessing them. For example, if group members received different scores on a math test, and these scores were later made public, this status distinction would structure group interaction, espe- cially if the group was interacting on a math-related task (Berger and Conner 1969; Berger and Fişek 1974; Berger and Webster 2006; Freese 1974).

Other types of status characteristics, diffuse status characteristics, are also activated in groups when they differentiate group members. While specific status characteristics prompt performance expectations that pertain directly to one’s task ability (e.g., math skills), diffuse status characteristics prompt both specific and wide-ranging performance expectations. Diffuse status characteristics are defined within the culture and carry societal stereotypes as well as overall esti- mations of “moral worth.” For example, in most societies, diffuse status charac- teristics include characteristics such as race, gender, and age.

Similar to specific status characteristics, diffuse status characteristics are used in group interaction and decision-making. Individuals’ influence in group

1368 Social Forces 97(3)

decision-making will be shaped by diffuse status information: this results in a “self-fulfilling prophecy,” by which individuals with positive stereotypes are ad- vantaged in the interaction: for example, men have more influence than women, older people have more influence than younger people, and people of color have less influence than White individuals. This reproduction of stereotyping is known as the burden-of-proof process, because it proceeds unless an event or other information interrupts the process (Berger, Cohen, and Zelditch 1972; Berger and Webster 2006).

Interventions in Status Generalization The theory of status characteristics and expectation states (SCES) predicts that people’s influence within the group will differ by the status they have relative to other group members. Influence is a component of the task group’s power and prestige structure, which is created by status and expectations. Further, SCES of- fers an alternative explanation for lack of success: the correct opinions of mem- bers with low-status characteristics (such as racial and ethnic minorities and/or women) are not being considered with the same weight as the incorrect opinions of members with high-status characteristics (i.e., they have less influence). Thus, groups are effectively dismissing the potentially valuable opinions and sugges- tions of low-status members.

Because these status processes can be insidious and negatively affect opportu- nities for those in power-disadvantaged positions, many researchers have con- sidered methods to overcome these status processes. The majority of this research focuses on adding information about other status characteristics pos- sessed by group members. So, for example, some studies have examined how evidence of superior task performance (a specific status characteristic) might decrease the negative effects of the low state of the diffuse status characteristic (see Cohen and Roper 1972; Freese and Cohen 1973; Markovsky, Smith, and Berger 1984; Pugh and Wahrman 1983; Wagner, Ford, and Ford 1986; Webster and Driskell 1978).

Other ways that researchers have interfered with status processes include questioning the influence hierarchy (Ridgeway and Correll 2006) or increasing participation rates of low-status group members (Walker, Doerer, and Webster 2014). One technique that is particularly important for our proposed research is what Cohen called a “Multiability” setting. In these settings, the emphasis is upon discussing and demonstrating how different settings and tasks require many different abilities (Cohen 1982; Cohen and Lotan 1997).

Building on the vast work of Cohen and colleagues, Fişek (1991) developed a theoretical elaboration of status characteristics theory, which Goar and Sell (2005) further elaborated on to focus on how information about the task itself might decrease inequality in interaction. Their formulation did not involve any interventions related to the characteristics of the participants, nor did they actu- ally change the task (as did Cohen). They simply defined the task as requiring multiple abilities that are unrelated, and therefore unlikely all to be possessed by one person. When the task requires multiple skills that have been defined as

Groups, Inequality, and Synergy 1369

uncorrelated, generalization leads to expectations for those multiple skills, and the expectations are uncorrelated. Those high and low expectations may not completely cancel out, but the aggregate will be less differentiating than would be the case if generalization occurred only to one skill.

The definition (labeled “Inconsistent Complexity,” or IC) encourages consult- ing others in the group since the task contains many different aspects associated with a variety of skills. By emphasizing the diverse nature of skills needed to suc- ceed at the task, we predict the IC intervention will reduce group inequality. By decreasing hierarchies that may inhibit low-status group member contributions, we predict that group performance will also be increased.

Scope Conditions and Hypotheses This project bridges two literatures to examine: (1) if the inconsistent complexity intervention reduces inequality; and (2) if (in)equality has an effect on group per- formance and synergy. To combine different theoretical frameworks, the same scope conditions must apply to both theories (Foschi 1997; Webster and Sell 2007). We have mentioned that the synergy literature considers task-oriented groups, which are interdependent and collectively oriented. The theory of status characteristics and expectation states (SCES) shares these scope conditions.

To meet these scope conditions, we consider the task, group history, and group composition. The task itself is particularly important for testing ideas about synergy. In the groups we consider, we control on the history of group members’ interaction by only using newly formed groups of strangers. To con- trol for surface-level diversity or diffuse status characteristics, we consider groups composed of women who are all undergraduate students of about the same age. The ethnic composition of the groups is all the same: two women who identify as White and one woman who identifies as Mexican American.2 We use this particular group composition of two majority group members and one minority group member to maintain consistency and comparability with past research, particularly Goar and Sell (2005).

We anticipate that ethnicity—when accurately assessed by group members—will function as a diffuse status characteristic, with White group members possessing the high state of the characteristic and Mexican American group members posses- sing the low state. Because both Mexican American and White group members act on this status relation, we expect that without an intervention in the status pro- cesses, White group members will dominate decision-making. However, we seek to intervene in this process by changing the definition of the situation—that is, by de- ploying incompatible complexity—to decrease inequality within the group. Because none of the group members have any prior training or expertise, a willingness to consider the opinions of others (that is, the willingness to be influenced) should also create synergy. Therefore, under the specified scope conditions, we predict:

1. Relative to no explicit definition, defining tasks as involving incompatible complexity (IC) increases interactional equality (as indicated by influence or its complement, deference);

1370 Social Forces 97(3)

2. Increased equality does not lead to decreased success on tasks; and 3. Increased equality increases synergy.

Experimental Design Overview An experimental design allows us to test our predictions. The primary advantage of experiments is the ability to test causal relationships by controlling for poten- tial confounding variables. Consequently, experiments do not attempt to directly generalize to existing empirical populations. Instead, they are designed to test theoretical principles that might eventually be applied to empirical settings. For this reason, they do not attempt to represent naturally occurring events.3

In this study, we controlled the composition of groups and only manipulate group members’ beliefs about the task. Specifically, all groups were composed of two self-identified White group members and one self-identified Mexican American group member. We controlled for possible gender, age, and education effects by recruiting only women at the same university of around the same age. Groups were then randomly assigned to one of two conditions to test our hypotheses.

In one condition, participants were told that “some individuals are better at these tasks than others” (baseline condition), and in another condition partici- pants were told that “the complexity of the task makes it unlikely that one indi- vidual would be good at every aspect of the task” (experimental condition). In total, we ran 50 groups, each composed of one Mexican American woman and two White women, for a total of 100 White women and 50 Mexican American women as participants.

Procedure Undergraduate students who were enrolled at a university in the southwest United States were recruited for paid research studies. Interested students com- pleted a one-page questionnaire containing contact information and basic demo- graphic information, including gender, class rank, age, race, and availability. We contacted students who self-identified as female and White or Mexican American and scheduled them for three weekly one-hour sessions.4

Upon arriving for the first session, participants individually read and signed the informed consent forms. They then were asked to sit at a round table where researchers placed completed name tags to ensure that the Mexican American woman was always sitting in the same location. At this time, participants were given a copy of the completed recruitment sheet for every group member5 and were asked to find their own sheet to affirm its correctness. The race, name, and gender of every participant was highlighted to ensure that all participants were aware of the racial/ethnic self-identity of all.

Next, the participants watched an instructional video that described the tasks. During this video, a university professor defined the task as a group problem-solving

Groups, Inequality, and Synergy 1371

task, and invested participants in the success of the group by offering an additional cash bonus at the end of Session 3 based on group performance. For both conditions, the senior researcher also emphasized that groups working together often performed better than those working alone. The description of the task in the instructional video included the intervention that differentiated the baseline and experimental groups (see below).

At the end of the video, participants were instructed to work on a task. The tasks given throughout the duration of the experiment were similar in nature. Each task provided a vignette that described a hypothetical situation in which the group became stranded in a particular environment: moon, ocean, or desert.6

The group was then provided a list of 12 to 15 salvaged items that might aid in survival. Group members were asked to work as a group to rank the items from most important to least important and to provide a reason for why each item was ranked as it was. Since participants were instructed to develop their reason- ing for different rankings, a high degree of group interaction was required, al- lowing us to examine group processes and synergistic gains.

To complete the task, participants first went to separate cubicles and worked individually for seven minutes. At the end of the seven minutes, group members returned to the center table and began to work on the same task collectively for 20 minutes. For the collaborative portion of the task, participants were told they could reference their individual sheets but not change their answers. The interac- tive group portion of the session was videotaped. Upon completing the task, the researcher administered a questionnaire that required participants to recall the instructions they received, as well as rate the performance of individual members and the group as a whole. This allowed us to determine if all the scope condi- tions were met and if the manipulations were clear to participants. When all group members completed the questionnaire, participants were paid $20 for their participation and allowed to leave. They were not told the results of their individual or group effort. The entire session took approximately one hour to complete.

Task 2 (completed in week 2) and Task 3 (completed in week 3) proceeded similarly to Task 1, except that participants did not complete a consent form or view the instructional video again. Since this was the first study of its kind, we thought the tasks themselves might have an effect. Therefore, we controlled par- ticipants’ experiences during the three time periods by showing all groups the tasks in the same order. In this way, we controlled time spent together and any task-specific effects, thereby eliminating familiarity as a factor that might con- tribute to differential task performance between conditions.

At the end of the third session, participants were given an additional question- naire to complete. This questionnaire asked for a more in-depth account of par- ticipants’ feelings toward the members of their group and the group as a whole. Once the questionnaires were complete, participants were told how they per- formed on the group task for each of the three sessions. They were paid $20 for their participation and given an additional $15 bonus, which they were told was the highest possible bonus that a group could earn for their performance on the

1372 Social Forces 97(3)

tasks. For participating in all three sessions, each participant earned a total of $75.

Incompatible Complexity Intervention (Experimental Manipulation) The independent variable was the definition of the task, as it was described in the instructional video during Session 1. We predicted that manipulating the task definition between the baseline and experimental conditions would modify participants’ expectations by providing the Incompatible Complexity interven- tion. In the video, groups in both conditions were given the same descriptive information about the task they would be completing. Groups in both condi- tions were told that individuals working together performed better than those working alone for all tasks they would encounter.

However, we varied instructions when providing direction about the team- work necessary to complete the task. In particular, in the baseline condition we provided instructions that mirrored real-life experience for most tests; namely, baseline groups were told that some individuals perform better than others:

This task today and the other tasks you will work on during the second and third group meeting are similar. We do know that some people do better at these tasks than other people. Even though some people do bet- ter while others do worse, we are trying to find out what makes some groups more successful than other groups at tasks like these.

Later in the video these instructions were reinforced, with baseline groups being told:

After the 20-minute allotted time for the group task has passed, we will give you a short questionnaire to fill out. This questionnaire will ask you to assess quality of the answers your group provided and more specifi- cally the quality of answers offered by individual group members.

On the other hand, groups in the experimental condition received the Incompatible Complexity intervention, meant to modify participants’ expecta- tion states by differing from everyday experience; namely, experimental groups were told that a variety of knowledge and skills were necessary for success:

The task today and the other tasks you will work on during the second and third group meeting involve using many, many different skills and abilities. We know that, generally speaking, some group members will have some special skills and abilities and others will have other skills and abilities. So, while everybody will have some ability to contribute to the task, it would be extremely unusual or even impossible for a single individual to be good at every single aspect necessary for these tasks.

These instructions were similarly reinforced later in the video, with experimental groups being told that they would fill out a questionnaire asking them to assess the kinds of abilities associated with success at the task and then also to assess the abilities of each group member.

Groups, Inequality, and Synergy 1373

Manipulation Checks As mentioned, after the videotaped instructions, group members answered a questionnaire, which asked about pertinent scope conditions. In both condi- tions, participants responded that: people do better when they work together (100 percent of participants responded in the affirmative to this), and all answers were not equally correct (90 percent of all participants responded in the affirma- tive, with no difference between conditions). If any group members did not cor- rectly identify the race/ethnicity of their other group members on the questionnaires, their group was not included in the analysis.7

To ensure that the manipulation was salient, group members were given a questionnaire that asked a series of true/false questions, including: (1) some peo- ple just seem to do better, overall, than others at the task; and (2) there are many different abilities and skills important for the task. Those in the Baseline- Contact condition were much more likely to agree than those in the Inconsistent Complexity condition (χ2 = 51.4, p < 0.001) that “some people just seem to do better, overall, than others at the task,” and those in the Inconsistent Complexity condition were more likely to agree that “there are many different abilities and skills important for the task” (χ2 = 28.7, p < 0.001).

Dependent Measures To compare the two conditions, we examined the dependent variables of influ- ence and group performance. We operationalized influence by measuring the amount of deference in discussion groups; that is, how much individuals change their answers in response to someone else’s disagreement (Berger et al. 1992; Berger, Cohen, and Zelditch 1972; Berger, Rosenholtz, and Zelditch 1980). To examine group performance, we look at both: (1) correctness: the proximity of the group ranking to the correct answers; and (2) synergy: the ability of a group to perform better than its members.

Task deference is calculated by first summing the absolute difference between the individual score and group score for each item on each task, and then divid- ing that total by the number of items in the task.8 The difference between the individual score and the group score was calculated for each item, summed together, and divided by the total number of items. When carried out for all of the items, this provides an average of the amount of deference given per each item across each task. Compared to those with low deference scores, individuals with high deference scores changed their answers more between the individual and group portion of the task. Deference is not a zero-sum measure, and we can- not determine to whom the deference is given. However, we can tell if indivi- duals change their answers based on the group interactions and processes, which is important for achieving synergy.

Correctness was measured by comparing group answers to those of experts. To measure correctness, we summed the difference between the expert and indi- vidual/group ranking for each item and divided this number by the total number of items in the task. This allowed us to measure how far, on average, the group

1374 Social Forces 97(3)

and/or individual was from the expert rankings. Thus, the lower the total score, the better, as it indicates that the individual was closer to the experts’ values. We calculated both group and individual correctness.

Synergy was measured, as suggested by Larson (2010), by comparing the group correctness to individual correctness (see table 1). To determine if groups achieved strong or weak synergy, we compared the lowest (best) and mean (average) individual group member score to the total group score. If the group score was lower (better) than the lowest (best) individual score, the group was said to have achieved strong synergy. Similarly, if the group score was lower than the average individual score in that group, the group was said to have achieved weak synergy. For example, let’s say that the three group members had scores of 2, 3, and 4 and the group score was 2.5. The average participant score for the group would be 3, and in this instance, the group would have achieved weak synergy (2.5–3 < 0), but not strong synergy (2.5–2 > 0). If the group score was 1.5, they would have achieved both weak and strong synergy.9 We created two separate dichotomous variables: that is, they either did (1) or did not (0) achieve weak synergy, and they either did (1) or did not (0) achieve strong synergy.

Results Analytic Strategy10

To assess the effect of our experimental manipulation on deference, we use a series of hierarchical linear regression models. By including a level in the model for both participant group and task number, hierarchical linear regression mod- els allow us to control for the nested nature of the data, that is, individuals within groups and groups within Tasks 1, 2, and 3.11 First, we examine the def- erence across all three tasks, comparing both experimental and baseline groups and White and Mexican American group members. We do so by conducting

Table 1. Measurement

Measure Calculation

Average per item deference Average deference = Σ (|Rgroup(i)−Rparticipant(i,1)|)/# of items Correctness Participant score (PS) = Σ (|Rcorrect−Rparticipant|)/# of items

Group score (GS) = Σ (|Rcorrect−Rgroup|)/# of items Average participant score (APS) = [(PSi,1 + PSi,2 + PSi,3)/3]

Strong synergy1 Strong synergy = GSi−Best individual scorei < 0 Not strong synergy = GSi−Best individual scorei ≥ 0

Weak synergy2 Weak synergy = GSi−APSi < 0 No synergy = GSi−APSi ≥ 0

Note: (1) Group’s score is better than its best individual member’s score. (2) Group’s score is better than the average of its individual scores.

Groups, Inequality, and Synergy 1375

three separate hierarchical linear models of average deference on experimental condition (our categorical indicator variable): one model for all participants (N = 150), one for White participants only (N = 100), and one for Mexican American participants only (N = 50). Finally, we use post-estimation Wald tests to test for predicted differences in overall deference between the baseline and experimental condition. All contrasts are evaluated at the p < 0.05 level, and all tests are one-tailed, according to our predictions.

Next, we examine the effect of experimental condition on overall deference for each task individually (see table 3). To measure the effect of our experimen- tal manipulation on deference, we conduct separate hierarchical linear regres- sion analyses for each task with participant as the unit of analysis and group as the level in our hierarchical model. Again, the experimental condition is included as a categorical indicator variable, and we use post-estimation Wald tests to test for the predicted difference between conditions.

In addition to the deference between experimental and baseline conditions, we examine within-group stratification. Our prediction is that the differentiation between Mexican Americans and Whites will be less in the experimental than in the baseline condition. To test this, we consider how the positions of most, mid- dle, and least deferential differ. By chance alone, if there was no status differenti- ation, we would expect that the Mexican American group member would occupy each position at an equal rate. However, due to status-based discrimina- tion, this is improbable without an intervention. Therefore, we expect that in the experimental group, Mexican Americans will occupy each position at a rate that is closer to what would be expected by chance alone. However, in the baseline group we expect that Mexican Americans will be more likely to occupy the bot- tom (least influential) position and less likely to control the top position than would be expected by chance alone. Using chi-square tests, we compare the probability that—across all 25 groups within the condition—the Mexican American group member will occupy each position at a rate that is comparable to what we would expect by chance alone. We do this separately for the baseline and experimental condition for each task.

We then examine the effect of the manipulation on within-group inequality in two ways. First, we conduct a hierarchical linear regression of the deference on the interaction of ethnicity and experimental condition for each task individu- ally. In these task-specific models, group is a level in the hierarchical model. Again, we use post-estimation Wald tests to test for predicted differences: first, comparing the difference between Anglo and Mexican American group mem- bers within condition, and then, comparing the difference between White and Mexican American group members from the baseline condition to that of the experimental condition (see difference column, table 4).

We now turn to our examinations of group performance. First, we compare baseline and experimental group performance simply with a measure of correct- ness, that is, average per item difference from experts. To do so, we estimate three separate linear regression models—one for each task—of correctness on experimental condition.12 We then use post-estimation Wald tests to compare correctness between the baseline and experimental conditions. Of note, there

1376 Social Forces 97(3)

were no significant differences between the members of the baseline or experi- mental conditions’ individual scores on any of the tasks (all contrasts, p = n.s.). Thus, all differences between the experimental and baseline groups that are observed in group scores are likely due to interaction. (That is, we have no other measurements that can account for the differences.)

To measure probability of synergy attainment (i.e., our dichotomous measure of whether a group achieved either strong or weak synergy), we conduct binary logistic regression models separately at each time point, with experimental con- dition as the indicator variable and either weak or strong synergy attainment as the dependent variable. We test for differences in synergy attainment between conditions using post-estimation tests of predicted probabilities by experimental condition (Long and Freese 2014).

Finally, we examine how the probability of synergy attainment is affected by experimental condition and group member deference (for the third task only). To do so, we conduct a binary logistic regression analysis, and include experi- mental condition as an indicator variable. Additionally, we include measures of the average deference given by White participants in each group, and the defer- ence given by the Mexican American group member.

Findings Deference or influence Table 2 illustrates the pattern of deference by experimental condition across all studies. As predicted, all participants are willing to accept more influence (defer more) in the experimental condition than in the baseline condition (p < 0.01). This is especially true for White participants in the experimental condition, who defer about 2.15 rankings on each item, compared to White participants in the baseline condition, who defer about 1.88 rankings per item (p < 0.01). Mexican American participants in the baseline and experimental groups defer to about the same amount (p = n.s.).

Table 3 shows the amount of deference given by members of the experimental and baseline groups for each task. We find that compared to the baseline condi- tion, individuals in the experimental condition deferred more on Tasks 1 and 3 (all contrasts, p < 0.05), but not significantly on Task 2 (p = n.s.).

Additionally, we can examine the stratification within different groups (figure 1). By determining who deferred the most, second, and least within each group, we

Table 2. Means and (SE) for per Item Deference by Condition—All Studies

Baseline Experimental Difference

All participants (N = 150) 2.02 (0.79) 2.25 (0.86) 0.23**

White participants (N = 100) 1.88 (0.74) 2.15 (0.87) 0.27**

Mexican American participants (N = 50) 2.30 (0.82) 2.44 (0.81) 0.14

Note: (1) * p < 0.05 ** p < 0.01 *** p <0.001; (2) Means estimated using hierarchical linear models, post-estimation comparisons made using Wald tests.

Groups, Inequality, and Synergy 1377

can examine the overall pattern by condition compared to what we would expect by chance alone. If there was no effect of race/ethnicity, we would expect that 33 percent of the time, the Mexican American member would be in the first, second, or third position.

Out of the 25 experimental groups, for Tasks 1, 2 and 3, we do not see signifi- cant differences between chance alone and the distribution for the Mexican American group members’ most deferential position. However, in the baseline con- dition, Mexican Americans are significantly more likely to occupy the most deferen- tial position. Therefore, as predicted, the experimental conditions were more equitable in terms of within-group deference than were the baseline conditions.

Group Performance and Synergy We hypothesized that groups with less inequality in deference would perform better than groups with more inequality. As noted, the experimental condition had overall more deference and less within-group racial/ethnic inequality than the baseline condition. Now we examine how this increased equality affects group performance and the probability of attaining synergistic performance gains.

Table 3. Means and (SE) for per Item Deference by Condition (N = 150)

Task Baseline Experimental Difference

1 1.94 (0.63) 2.19 (0.80) −0.26*

2 2.38 (0.89) 2.59 (0.84) −0.20 3 1.73 (0.69) 1.96 (0.83) −0.23*

Note: (1) * p < 0.05 ** p < 0.01 *** p < 0.001; (2) Means estimated using hierarchical linear models, post-estimation comparisons made using Wald tests.

Figure 1. Mexican Americans’ group position by deference given

Task 1 Task 2 Task 3 Task 1 Task 2 Task 3

Control Experimental

N um

be r

of P

ar ti

ci pa

nt s

in P

os it

io n

Top (Least Deferential) Middle Bottom (Most Deferential)

* * *

Note: *p < 0.05 **p < 0.01 ***p < 0.001; significance indicate results from chi-square tests comparing findings to what is expected by chance alone.

1378 Social Forces 97(3)

As illustrated in table 4, baseline and experimental groups performed simi- larly on Tasks 1 and 2. However, on Task 3, the experimental group performed better than the baseline group. This supports our second prediction that the experimental groups would not perform worse than the baseline groups. While these measures illustrate the general differences in overall performance, they do not speak to the performance gains, that is, synergy.

We predicted that groups that had less inequality would be more likely to cap- italize on group member contributions and achieve synergy. As indicated in table 5, experimental groups were overall more likely to achieve both weak and strong synergy than baseline groups (both, p < 0.05). This overall difference is largely driven by Task 3.

For Task 1, out of the 50 total groups, 15 experimental and 16 baseline groups achieved strong synergy; 8 experimental and 8 baseline groups achieved weak synergy; and 2 experimental compared to 1 baseline group achieved no synergy (all contrasts, p = n.s.). Similarly, for the second task, 10 experimental and 8 baseline groups achieved both strong and weak synergy; 11 experimental and 9 baseline achieved weak synergy; and 6 of both experimental and baseline groups achieved no synergy (all contrasts, p = n.s.). Thus, in the first two tasks, it appears that the Incompatible Complexity intervention has little impact on synergy.

Table 4. Means and (SE) for Difference between Correct Score and Group Score (N = 50)

Task Baseline Experimental Difference

1 2.20 (0.47) 2.40 (0.58) −0.20 2 4.74 (0.56) 4.75 (0.65) −0.01 3 4.48 (0.39) 4.09 (0.82) 0.39*

Note: (1) * p < 0.05 ** p < 0.01 *** p < 0.001; (2) Means estimated using hierarchical linear regression model, post-estimation comparisons made using Wald tests; (3) Correctness is measured by difference from experts. The lower the score, the better. Higher numbers indicate larger difference between expert and individual or group scores.

Table 5. Groups’ Predicted Probabilities of Achieving Synergy by Condition (N = 50)

Weak synergy Strong synergy

Task Baseline Experimental Difference Baseline Experimental Difference

Overall −0.70 −0.83 −0.13* −0.33 −0.42 −0.09*

1 0.96 0.92 0.04 0.64 0.60 0.04

2 0.76 0.76 0.00 0.32 0.40 −0.08 3 0.24 0.64 −0.40*** 0.08 0.28 −0.20*

Note: * p < 0.05 ** p < 0.01 *** p < 0.001; (2) Means estimated using hierarchical logistic regression model, post-estimation comparisons made using Wald tests.

Groups, Inequality, and Synergy 1379

However, the third task shows significant differences between the baseline and experimental groups. On the third task, 7 experimental and 2 baseline groups achieved both strong and weak synergy; 9 experimental and 4 baseline groups achieved weak synergy; and 9 experimental versus 19 baseline groups achieved no synergy at all (p < 0.05). Thus, it appears that diverse task groups are not negatively affected by the Inconsistent Complexity manipulation, and in fact, at least for the third task, score better than individuals.

Task 3—Synergy Figures 2a and 2b represent how deference affects the predicted probability that groups will achieve weak and strong synergy, respectively. In Task 3, groups in the experimental condition (solid lines) achieved synergy significantly more fre- quently than groups in the baseline condition (dashed lines). This is represented by the fact that the solid lines are consistently higher than the dashed lines. Additionally, as White group members (blue line) defer more, the predicted probability of achieving both weak (figure 2a) and strong synergy (figure 2b)

Figure 2a. Predicted probability of achieving weak synergy, Task 3

0 .2

.4 .6

.8 1

P re

di ct

ed P

ro ba

bi li

.2 .6 1 1.4 1.8 2.2 2.6 3 3.4 3.8 4.2 4.6

Average of Group Members' Deference

Mexican American Control Mexican American Experimental White Control White Experimental

Note: Dotted line indicates an out-of-sample prediction.

Figure 2b. Predicted probability of achieving strong synergy, Task 3

0 .2

.4 .6

.8 1

P re

di ct

ed P

ro ba

bi li

.2 .6 1 1.4 1.8 2.2 2.6 3 3.4 3.8 4.2 4.6

Average of Group Members' Deference

Mexican American Control Mexican American Experimental White Control White Experimental

Note: Dotted line indicates an out-of-sample prediction.

1380 Social Forces 97(3)

increases. However, this pattern is not true for Mexican American group mem- bers. That is, as Mexican American (black line) group members defer more, the predicted probability of achieving synergy decreases.

Summary of Findings As predicted, groups that received the inconsistent complexity intervention were more likely to have more equitable hierarchies in terms of deference. This was apparent in Task 1 and Task 3 (and for Task 2 in terms of position analysis). For Task 3, the experimental groups were more 40 percent more likely to achieve weak and 20 percent more likely to achieve strong synergy than the baseline groups. Furthermore, we see that this pattern is related to overall defer- ence. Otherwise said, as Mexican American group members defer less and White members defer more, groups are more likely to achieve weak synergy. This pattern also holds for the probability of achieving strong synergy on Task 3. For none of the tasks do the experimental groups do more poorly than the baseline groups, and on the third task, experimental groups do better overall and achieve synergy at much higher levels.13

Discussion and Conclusion This paper combines insights from research on group performance, specifically synergy, with status characteristics and expectation states research on group processes. Drawing on status characteristics and expectation states research, we argue that who offers suggestions or gives and receives agreements can help explain group processes and performance. Specifically, we predict that the status of group members may create interactional inequality and inhibit the open exchange and consideration of ideas that will affect group performance. We examine whether an Inconsistent Complexity intervention that involves only defining the tasks as involving multiple and complex abilities, but not changing the tasks at all, might change the interaction for diverse groups faced with differ- ent tasks over time.

We find that the intervention seemed to substantively change interaction on two of the three tasks, such that hierarchy was lessened in terms of deference. This lessening never had negative effects on performance; and on the last task, the defi- nition created better-performing groups that had higher levels of synergy; that is, they performed better than the individuals themselves. These findings demonstrate that not only does greater equality in task-oriented groups not hurt task perfor- mance, but also it may, in fact, improve performance—at least for the kinds of tasks we consider—where no one person has more expertise than another.

But the research also points to the importance of considering the tasks them- selves in group interaction. With a few exceptions (e.g., Shelly and Shelly 2009, 2017), status characteristics and expectation states theories rarely consider the variation in tasks on which group members interact. Our results point out that we should, indeed, be considering the tasks themselves. On the second task (Lost at Sea), all groups did poorly and groups in the experimental conditions were

Groups, Inequality, and Synergy 1381

more hierarchically structured than for the other tasks. This is concerning and is an important issue to investigate further.

A likely scenario is that situations characterized by high degrees of ambiguity create the conditions most likely to lead to stereotyping and subsequent inequal- ity (Dovidio and Gaertner 2000; Hodson, Dovidio, and Gaertner 2002; Melamed and Savage 2013). Lost at Sea seemed to be viewed as “more difficult” and more ambiguous by the group participants. Indeed, in terms of correctness, participants did the worst on this task. Fourteen out of the 50 groups made spe- cific comments about the difficulty of this task. Transcripts show no indication that subjects struggled as much with Tasks 1 or 3.14

The differences in group processes across the tasks point to the importance of conceptualizing group tasks. It is important to note, however, that all three tasks had the same general structure; that is, they did not vary the type of task (for example, decisions were not about guilt or innocence, or solving mysteries as many of the studies in organizational psychology). It seems unlikely that a sec- ond interaction, that is, timing alone, could explain the poor performance and increased inequality, because the outcomes are exactly reversed in the third task administered at the third point in time. So, it would seem that the original defini- tion of these tasks and the focusing of individual group members on differing abilities did have longer effects. This is an important finding and supplements the results of Goar and Sell (2005).

While the inconsistency manipulation seemed to lessen inequality, the effects were not as dramatic as those for Goar and Sell (2005) and indicate that some reassessment of the manipulation is necessary. Our study did consider a different ethnic group (Mexican Americans) and different tasks (to enable the measure- ment of influence or deference) from those in Goar and Sell (2005). It is possible that these differences made the intervention less powerful. So, for example, to measure deference, participants had to develop an initial opinion, and this may develop a personal commitment to that initial opinion that would be stronger for higher-status (White) participants. Future work should test different kinds of diffuse status characteristics (e.g., gender or different ethnicities) to ensure that the hypothesized theoretical scope is empirically supported.15 Additionally, fur- ther research may manipulate the proportion of high- and low-status group members to examine if this affects group inequality.

Even with these caveats, our findings have important implications for class- rooms, research groups, and teams. The “decay” of the manipulation at the sec- ond point in time is contradicted by the third point of time, when results are particularly strong. So, it suggests that emphasizing the inconsistent complexity of tasks can have long-lasting effects on decreasing inequality and allow groups to weather through backsliding.

Our test was an extremely conservative test of the manipulation. As far as we know, it is the only such laboratory, experimental test that is done over this length of time. It was only administered once in the beginning of a three-week period. That manipulation lasted only a few minutes, yet it had effects three weeks later. The effects on hierarchy were relatively strong, especially for the first and third task. Further, the effects are stronger than the effects of contact

1382 Social Forces 97(3)

and diversity alone, since contact and diversity were present in both the experi- mental and baseline conditions. Such a definition is extremely easy to use; it takes little time and requires few resources. Further, if the manipulation had an effect in such a “small dose,” it is likely to have a more substantial effect when the definition is re-emphasized over time. Our results also suggest a productive integration of synergy research with research on status characteristics and expec- tation states. Such integration would serve both literatures and allow them to expand.

Notes 1. There is a venerable tradition in group process research demonstrating that groups

underperform relative to individuals; this work was applied to the study of panic, for example (Mintz 1951). Additionally, superordinate goals have been long investi- gated as means to decrease conflict within and across groups (McClendon and Eitzen 1975; Sherif 1966). However, in our study we expose all our groups to a superordi- nate goal. It is the definition of the superordinate goal that differs between the condi- tions we examine.

2. We use the racial classifications of “Mexican American” and “White,” as they reflect the self-identifications of our study participants.

3. There are many discussions on the logic of experiments and generalization. Recent delineations of the role of experiments in social science include Lucas (2003); Sell (2018); Webster (2016); and Willer and Walker (2007).

4. Two researchers—one White and one Mexican American—jointly administered each session and alternately instructed the group. Having experimenters alternately give instructions averaged any ethnic-based effects of the experimenters, who held posi- tions of power relative to the group members.

5. Researchers redacted personal information on the group copies such as class rank, phone number, and e-mail address to control the information participants had about each other. Redacting this information also protected the privacy of participants.

6. Tasks came from two sources. Survival on the moon can be found on the NASA website (https://www.nasa.gov/pdf/166504main_Survival.pdf) or in Hall and Watson (1970). Survival in the desert and on a wrecked ship came from Johnson and Johnson (2009).

7. Eight groups were excluded on this basis. To compensate for the loss of these groups, eight other groups were run in the correct conditions.

8. When only two people interact and disagree about a response, one can infer influence by whether a group member stays with their original opinion or changes their opin- ion. A person either stays with an initial response (rejects influence) or changes an initial response (accepts influence or demonstrates deference). In our study, we have a similar measure. We measure how much the original decision of a participant dif- fers from the group response. In this way, we measure how much the group has affected each group member.

9. Groups that achieved strong synergy always achieved weak synergy, but those that achieved weak synergy did not always achieve strong synergy. That is, if groups out- performed their best member, they must have outperformed their average member. But if groups performed better than their average member, they did not necessarily perform better than their best member.

Groups, Inequality, and Synergy 1383

https://www.nasa.gov/pdf/166504main_Survival.pdf

10. To access script and data files, please visit the first author’s website at: biancama- nago.com.

11. We use likelihood ratio tests to test the null hypothesis that the random effects in the hierarchical model are statistically insignificant (i.e., that the group has no bearing on the regression estimation). Chi-square suggests that we can reject the null hypoth- esis, meaning that group is an important hierarchical variable (p < 0.001 for all).

12. We also estimated this relationship using one model with task as a variable, and there were no substantively meaningful differences in magnitude or statistical significance.

13. This finding supplies a caveat to the Goar et al. (2013) article, which discusses task performance. While the experimental groups in this study never did worse, they did better on one task of the three.

14. In Task 3, experimental groups performed better than baseline groups in terms of synergy. However, individuals did less well in terms of correctness when compared with the group’s level of correctness. Examining the transcripts, and individual/group data, we found that the variance on a single item seemed to exert undue influence on performance scores on Task 3.

15. Different ethnic groups may generate different degrees of inequality. For example, Thye and Harrell (2017) found very strong status effects for race when they con- sidered African Americans versus Whites, much stronger than for gender. Research by Skvoretz and Bailey (2016) also suggests that diversity complicates interaction, as performance expectations are filtered through ethnic-based notions of familiarity and suitability.

About the Authors Bianca Manago is currently an assistant professor of sociology at Vanderbilt University. Her research questions fall at the intersection of medical sociology, social psychology, and inequality.

Jane Sell is a professor at Texas A&M University. Her areas of research inter- ests include the development and decay of cooperation in social dilemmas and how different factors affect racial and ethnic equality within small groups. Along with Murray Webster Jr. she edited Laboratory Experiments in the Social Sciences (Elsevier, 2014).

Carla Goar is an associate professor in the Department of Sociology at Kent State University. Her research areas include group dynamics, race and inequal- ity, and adoption.

Supplementary Material Supplementary material is available at Social Forces online.

References Banerjee, Abhijit V. 1992. “A Simple Model of Herd Behavior.” Quarterly Journal of Economics 107(3):

797–817.

1384 Social Forces 97(3)

Berger, Joseph. 2014. “The Standardized Experimental Situation in Expectation States Research: Notes on History, Uses, and Special Features.” In Laboratory Experiments in the Social Sciences, edited by M. Webster and J. Sell, 269–93. Oxford: Elsevier.

Berger, Joseph, Bernard P. Cohen, and Morris Zelditch Jr. 1972. “Status Characteristics and Social Interaction.” American Sociological Review 37(3):241–55.

Berger, Joseph, and Thomas L. Conner. 1969. “Performance Expectations and Behavior in Small Groups.” Acta Sociologica 12(4):186–98.

Berger, Joseph, and M. H. Fişek. 1974. “A Model for the Evolution of Status Structures in Task-Oriented Discussion Groups.” In Expectation States Theory: A Theoretical Research Program, edited by J. Berger, T. Conner, and M. Fisek, 53–84. Cambridge, MA: Winthrop Publishers.

Berger, Joseph, Robert Z. Norman, James W. Balkwell, and Roy F. Smith. 1992. “Status Inconsistency in Task Situations: A Test of Four Status Processing Principles.” American Sociological Review 57(6):843–55.

Berger, Joseph, Susan J. Rosenholtz, and Morris Zelditch Jr. 1980. “Status Organizing Processes.” Annual Review of Sociology 6(1980):479–508.

Berger, Joseph, and Murray Webster. 2006. “Expectations, Status, and Behavior.” In Contemporary Social Psychological Theories, edited by P. J. Burke, 268–300. Stanford, CA: Stanford University Press.

Cohen, Elizabeth G. 1982. “Expectation States and Interracial Interaction in School Settings.” Annual Review of Sociology 8(1982):209–35.

Cohen, Elizabeth, and Rachel Lotan. 1997. Working for Equity in Heterogeneous Classrooms: Sociological Theory in Practice. Sociology of Education Series. New York: Teachers College Press.

Cohen, Elizabeth G., and Susan S. Roper. 1972. “Modification of Interracial Interaction Disability: An Application of Status Characteristic Theory.” American Sociological Review 37(6):643–57.

Curşeu, Petru Lucian, Rob J. G. Jansen, and Maryse M. H. Chappin. 2013. “Decision Rules and Group Rationality: Cognitive Gain or Standstill?” PLoS One 8(2):e56454.

Davis, James H., Tatsuya Kameda, Craig Parks, Mark Stasson, and Suzi Zimmerman. 1989. “Some Social Mechanics of Group Decision Making: The Distribution of Opinion, Polling Sequence, and Implications for Consensus.” Journal of Personality and Social Psychology 57(6):1000–12.

Dovidio, John F., and Samuel L. Gaertner. 2000. “Aversive Racism and Selection Decisions: 1989 and 1999.” Psychological Science 11(4):315–9.

Fişek, M. Hamit. 1991. “Complex Task Structures and Power and Prestige Orders.” Advances in Group Processes 8:115–34.

Foschi, Martha. 1997. “On Scope Conditions.” Small Group Research 28(4):535–55. Freese, Lee. 1974. “Conditions for Status Equality in Informal Task Groups.” Sociometry 37(2):174–88. Freese, Lee, and Bernard P. Cohen. 1973. “Eliminating Status Generalization.” Sociometry 36(2):177–93. Goar, Carla and Jane, Sell. 2005. “Using Task Definition to Modify Racial Inequality within Task Groups.”

The Sociological Quarterly 46:525–43. ———. 2009. “The Use of FAD Funds: Expanding Experimental Investigations of Race/Ethnicity in

Sociology.” ASA Footnotes 37(3):7. Goar, Carla, Jane Sell, Bianca Manago, Calixto Melero, and Bobbi Reidinger. 2013. “Race and Ethnic

Composition of Groups: Experimental Investigations.” Advances in Group Processes 30:47–75. Hackman, J. Richard. 1987. “The Design of Work Teams.” In The Handbook of Organizational Behavior,

edited by J. W. Lorsch, 315–42. Englewood Cliffs, NJ: Prentice-Hall. Hall, Jay, and Wilfred Harvey Watson. 1970. “The Effects of a Normative Intervention on Group Decision-

Making Performance.” Human Relations 23(4):299–317. Harrison, David A., Kenneth H. Price, and Myrtle P. Bell. 1998. “Beyond Relational Demography: Time and

the Effects of Surface- and Deep-Level Diversity on Work Group Cohesion.” Academy of Management Journal 41(1):96–107.

Groups, Inequality, and Synergy 1385

Harrison, David A., Kenneth H. Price, Joanne H. Gavin, and Anna T. Florey. 2002. “Time, Teams, and Task Performance: Changing Effects of Surface-and Deep-Level Diversity on Group Functioning.” Academy of Management Journal 45(5):1029–45.

Herring, Cedric. 2009. “Does Diversity Pay? Race, Gender, and the Business Case for Diversity.” American Sociological Review 74:208–24.

———. 2017. “Is Diversity Still a Good Thing?” American Sociological Review 82(4):868–77. Hodson, Gordon, John F. Dovidio, and Samuel L. Gaertner. 2002. “Processes in Racial Discrimination:

Differential Weighting of Conflicting Information.” Personality and Social Psychology Bulletin 28(4): 460–71.

Horwitz, Sujin K., and Irwin B. Horwitz. 2007. “The Effects of Team Diversity on Team Outcomes: A Meta-Analytic Review of Team Demography.” Journal of Management 33(6):987–1015.

Hunt, Matthew O., Pamela Braboy Jackson, Samuel H. Kye, Brian Powell, and Lala Carr Steelman. 2013. “Still Color-Blind? The Treatment of Race, Ethnicity, Intersectionality, and Sexuality in Sociological Social Psychology.” Advances in Group Processes 30:21–45.

Jehn, Karen A., Gregory B. Northcraft, and Margaret A. Neale. 1999. “Why Differences Make a Difference: A Field Study of Diversity, Conflict, and Performance in Workgroups.” Administrative Science Quarterly 44(4):741–63.

Johnson, David, and Frank Johnson. 2009. Joining Together: Group Theory and Group Skills, 10th ed. Upper Saddle River, NJ: Pearson.

Joshi, Aparna, and Hyuntak Roh. 2009. “The Role of Context in Work Team Diversity Research: A Meta- Analytic Review.” Academy of Management Journal 52(3):599–627.

Kameda, Tatsuya, Thomas Wisdom, Wataru Toyokawa, and Keigo Inukai. 2012. “Is Consensus-Seeking Unique to Humans? A Selective Review of Animal Group Decision-Making and Its Implications for (Human) Social Psychology.” Group Processes & Intergroup Relations 15(5):673–89.

Kerr, Nobert L., and R. Scott Tindale. 2004. “Group Performance and Decision Making.” Annual Review of Psychology 55:623–55.

———. 2012. “Scientific Inquiry on How Groups Decide: The Davisonian Approach.” Group Processes & Intergroup Relations 15(5):577–84.

Larson, James R. 2010. In Search of Synergy in Small Group Performance. New York and London: Psychology Press.

Laughlin, Patrick R. 2011. “Social Choice Theory, Social Decision Scheme Theory, and Group Decision- Making.” Group Processes & Intergroup Relations 14(1):63–79.

Laughlin, Patrick R., and Alan L. Ellis. 1986. “Demonstrability and Social Combination Processes on Mathematical Intellective Tasks.” Journal of Experimental Social Psychology 22(3):177–89.

Levine, John M., and Richard L. Moreland. 1990. “Progress in Small Group Research.” Annual Review of Psychology 41:585–64.

Long, J. Scott, and Jeremy Freese. 2014. Regression Models for Categorical Dependent Variables Using Stata, 3rd ed. College Station, TX: Stata Press.

Lount, Robert B., Jr., Oliver J. Sheldon, Floor, Rink, Katherine W. Phillips 2015. “Biased Perceptions of Racially Diverse Teams and Their Consequences for Resource Support.” Organization Science 26(5): 1351–64.

Lucas, Jeffrey W. 2003. “Status Processes and the Institutionalization of Women as Leaders.” American Sociological Review 68(3):464–80.

Markovsky, Barry, Le Roy F. Smith, and Joseph Berger. 1984. “Do Status Interventions Persist?” American Sociological Review 49(3):373–82.

McClendon, McKee J., and D. Stanley Eitzen. 1975. “Interracial Contact on Collegiate Basketball Teams —Test of Sherif’s Theory of Superordinate Goals.” Social Science Quarterly 55(4):926–38.

1386 Social Forces 97(3)

Melamed, David and Scott V. Savage. 2013. “Status, Numbers and Influence.” Social Forces 91(3): 1085–104.

Mintz, Alexander. 1951. “Non-Adaptive Group Behavior.” Journal of Abnormal and Social Psychology 46 (2):150–9.

Pelled, Lisa Hope. 1996. “Relational Demography and Perceptions of Group Conflict and Performance: A Field Investigation.” International Journal of Conflict Management 7(3):230–46.

Pelled, Linda Hope, Kathleen M. Eisenhardt, and Katherine R. Xin. 1999. “Exploring the Black Box: An Analysis of Work Group Diversity, Conflict and Performance.” Administrative Science Quarterly 44(1): 1–28.

Phillips, Katherine W. 2014. “How Diversity Works.” Scientific American 311(4):43–7. Phillips, Katherine W., Gregory B. Northcraft, and Margaret A. Neale. 2006. “Surface-Level Diversity and

Decision-Making in Groups: When Does Deep-Level Similarity Help?” Group Processes & Intergroup Relations 9(4):467–82.

Pugh, M. D., and Ralph Wahrman. 1983. “Neutralizing Sexism in Mixed-Sex Groups: Do Women Have to Be Better Than Men?” American Journal of Sociology 88(4):746.

Ridgeway, Cecilia, and Shelley J. Correll. 2006. “Consensus and the Creation of Status Beliefs.” Social Forces 85:431–54.

Sell, Jane. 2018. “Definitions and the Development of Theory in Social Psychology.” Social Psychology Quarterly 81(1):8–22.

Shelly, Robert K., and Ann C. Shelly. 2009. “Speech Content and the Emergence of Inequality in Task Groups.” Journal of Social Issues 65(2):307–33.

Shelly, Robert K., and Ann Shelly. 2017. “Impact of Variations in Task for Group Path-to-Task Research.” Paper presented at the Annual Southern Sociological Meetings, Greenville, SC.

Sherif, Muzafer. 1966. Common Predicament: Social Psychology of Intergroup Conflict and Cooperation. Boston, MA: Houghton Mifflin.

Skvoretz, John, and Jasmón L. Bailey. 2016. “‘Red, White, Yellow, Blue, All Out but You’: Status Effects on Team Formation, an Expectation States Theory.” Social Psychology Quarterly 79(2):136–55.

Sommers, Samuel R. 2006. “On Racial Diversity and Group Decision Making: Identifying Multiple Effects of Racial Composition on Jury Deliberations.” Journal of Personality and Social Psychology 90(4): 597–612.

Stasser, Garold, and William Titus. 2003. “Hidden Profiles: A Brief History.” Psychological Inquiry 14(3): 304–13.

Thye, Shane R., and Ashley Harrell. 2017. “The Status Value Theory of Power and Mechanisms of Micro Stratification: Theory and New Experimental Evidence.” Social Science Research 63:54–66.

Tsui, Anne S., Terri D. Egan, and Charles A. O’Reilly. 1992. “Being Different: Relational Demography and Organizational Attachment.” Administrative Science Quarterly 37(4):549–79.

van Dijk, Hans, Marloes L. van Engen, and Daan van Knippenberg. 2012. “Defying Conventional Wisdom: A Meta-Analytical Examination of the Differences between Demographic and Job-Related Diversity Relationships with Performance.” Organizational Behavior and Human Decision Processes 119(1): 38–53.

Wagner, David G., Rebecca S. Ford, and Thomas W. Ford. 1986. “Can Gender Inequalities Be Reduced?” American Sociological Review 51(1):47.

Walker, Lisa Slattery, Sharon C. Doerer, and Murray Webster Jr. 2014. “Status, Participation, and Influence in Task Groups.” Sociological Perspectives 57(3):364–81.

Watson, Warren E., Kamalesh Kumar, and Larry K. Michaelsen. 1993. “Cultural Diversity’s Impact on Interaction Process and Performance: Comparing Homogeneous and Diverse Task Groups.” Academy of Management Journal 36(3):590–602.

Webster, Murray. 2016. “Research Opportunities.” Social Psychology Quarterly 79(1):5–21.

Groups, Inequality, and Synergy 1387

Webster, Murray, and James E. Driskell. 1978. “Status Generalization: A Review and Some New Data.” American Sociological Review 43:220–336.

Webster, Murray Jr., and Jane Sell. 2007. “Theory and Experimentation in the Social Sciences.” In The SAGE Handbook of Social Science Methodology, edited by W. Outhwaite and S. Turner, 190–207. Amsterdam, Netherlands: Elsevier.

Willer, David, and Henry A. Walker. 2007. Building Experiments: Testing Social Theory. Palo Alto, CA: Stanford University Press.

1388 Social Forces 97(3)