3

RUNNING LABORATORY
EXPERIMENTS WITH GROUPS

Gwen M. Wittenbaum

MICHIGAN STATE UNIVERSITY

Fledgling scholars who want to understand groups by conducting laboratory experiments will find many sources of information about how to design experiments (e.g., Haslam & McGarty, 2004; Kirk, 1995; Laughlin, this volume; Smith, 2000). Graduate students in social sciences typically learn about experimental design in their research methods courses and benefit from having an advisor or graduate committee assist with their experimental design for a degree-required project. When embarking to conduct the laboratory experiment, however, the fledgling scholar may have few sources of support for how to run it. Graduate students may compare notes with their peers for advice on how to run an experiment or use procedures typically used by the faculty advisor. There appears to be a dearth, however, of published tutorials on how to run a laboratory experiment, let alone one on groups. The purpose of this chapter is to demystify the process of conducting laboratory experiments on groups by highlighting personal research examples and behind-the-scenes procedures that rarely are reported in the method section of a journal article.

I began conducting laboratory experiments on groups as a graduate student in thefield of social psychology 20 years ago. Throughout this time, I have studied both laboratory groups that interact face-to-face and groups that are imagined or anticipated. There are many ways to examine group dynamics in a laboratory experiment. First, I will offer a typology of commonly usedgroups in laboratory experiments. Second, I will provide research examples of these different types of laboratory groups. Third, I will detail the step-by-step procedures that will help to make alaboratory experiment on groups run smoothly the first time. Finally, I finish with conclusions about lessons learned.

Types of Laboratory Groups

Some fields have used laboratory experiments to understand group dynamics morethan others. Social psychologists who study groups heavily rely on the laboratory experiment. In acontent analysis of the top social psychology journals, Moreland, Hogg, and Hains (1994) reported that 76 per cent of published studies on groups utilized laboratory experiments. A similar conten analysis of organizational psychology journals reported that 50 per cent of the group research in this field used the laboratory experiment method (Sanna & Parks, 1997). Likewise, communication scholars have been criticized for relying too heavily on laboratory experiments (Frey, 1994). Other fields, such as management, sociology, kinesiology, and political science may rely less thanpsychologists and communication scholars on laboratory experiments as a way to understand groups. However, nearly every social scientific field that houses group research uses the laboratory experiment. The ubiquity of the method suggests that there are many approaches to studying laboratory groups.

When scholars study group dynamics in a laboratory experiment, they typically examine groups that fall into one of the following four categories: (1) anticipated groups; (2) imagined groups; (3) confederate groups; or (4) interacting groups. In anticipated groups, participants complete tasks on their own with the expectation that they will interact with fellow group members at a later time. Often, the interaction expectation is fabricated on the part of the experimenter who wants to study how people think and behave when they expect to interact in a group, even though no group actually exists. This type of group is similar to imagined groups, which only exist in the mind of the participants. However, participants in imagined laboratory group believe that they are interacting with fellow group members. In actuality, those group members donot exist, yet their behavior is simulated by fabricated messages transmitted typically through amediated channel (e.g., note passing, online chat). In confederate groups, participants believe that they are interacting with other unsuspecting participants. In actuality, one or more othergroup members are trained accomplices (i.e., confederates) of the experimenter, constrained toact in a certain way. To participants, these confederate groups appear to be in interacting groups. However, an interacting group is used here to mean natural, in-person group interaction of several participants working toward a common goal. The interacting group simulates in the laborator the type of groups that we typically think of in natural contexts, such as committees, juries, andteams. Interacting groups can communicate face to face or using a range of media such as video, phone, computer conferencing, email or text messaging, or in virtual worlds or social networking sites. As most of my research has been conducted in face-to-face groups, I will focus on face-to-facegroups (FTF) in this chapter. See Wirth Feldberg, Schouten, van den Hooff, and Williams (this volume) for a discussion of experimental research using computer games and virtual worlds.

For each group type, I will provide a research illustration from the area of group decision-making – a foundational area of study that lends itself to investigation across the different types of laboratory groups. Table 3.1 provides a description of each type of laboratory group along with a summary of the following research examples.

Table 3.1 Summary of types of laboratory groups

Research Examples

Anticipated groups

Group decision-making requires members to integrate information to achieve a consensus. Efforts toward reaching a consensus, however, may begin before the group meets. This wasthe premise behind Wittenbaum, Stasser, and Merry's (1995) research on tacit coordination. I was a graduate student when I worked on this project, and it was an eye-opening introduction to the creative ways of studying groups in the laboratory. We reasoned that group members form expectations about other group members’ likely areas of expertise and assume what kind of information other members likely will attend to before discussion based on their expertise. Group membersmay help to facilitate group task completion before the group's meeting by adjusting the kind of information that they attend to and remember. We thought, however, that the kind of information to which group members would attend would depend on the anticipated group task. When members anticipate a group decision-making task, they would attend to similar information as others in order to facilitate agreement. Whereas, when members anticipate a collective recall task, they would try to remember information that others likely would not, in the interest of maximizing the group's recall output. This idea lent itself perfectly to a laboratory experiment using anticipated groups.

Participants in Study 1 were told that they would interact with others on a group task but would have the opportunity to share surface information (in the form of a college life questionnaire) with each other before the group convened. After completing the college life questionnaire, the experimenter returned with copies of questionnaires presumably completed by the othereventual group members. These bogus questionnaires were designed so that each of the other membersappeared to be expert in one of the domains relevant to the materials used for the group task. Participants perused these bogus questionnaires and then anticipated one of two group tasks (i.e., the experimental manipulation): a group decision regarding the best of three student body president candidates or collective recall of the candidates’ information. After participants read theinformation about the three hypothetical candidates, they unexpectedly were asked to remember, individually, as much of the candidate information as possible within six minutes. From the researcher's perspective, the study was over after the free recall task. Participants, nevertheless, were placed into groups with others from that experimental session to follow through with the guise of the study. Just as expected, those who anticipated a group decision-making task better remembered information in the others’ areas of expertise, whereas those who anticipated a collective recall task better remembered information outside of others’ areas of expertise. This study showed that group member efforts to facilitate a group decision begin before the group meets.

Imagined groups

The research of Wittenbaum et al. (1995) showed that group members who anticipate a group decision-making task focus on knowledge that others probably know. This finding is consistent with research that shows that members of decision-making groups favor discussing sharedinformation that all members know over unshared information that a single member knows(for a review, see Wittenbaum, Hollingshead, & Botero, 2004). Wittenbaum et al. (1999) reasoned that group members may prefer to discuss shared information because that information can be socially validated (as accurate and important), leading members to positively value such informationwhen mentioned. Likewise, group members who communicate much shared information may receive more positive evaluations from others compared to those who communicate much unshared information.

We wanted to begin our investigation in a setting that was more controlled than FTF groups, where members could not see one another and potentially form evaluations based on other member features (e.g., physical appearance, nonverbal behavior). We also needed to control the information communicated by the target group member. Therefore, we generated a paradigm using imagined groups. Participants were seated in a room and expected to exchange written information with adyadic partner in another room within the same social interaction laboratory. Participants read information about two hypothetical academic job candidates and indicated their preferred candidate. “Discussion” occurred via the exchange of written information. Participants listed ten pieces of candidate information to share with their dyadic partner, which was collected by the experimenter and presumably delivered to the dyadic partner across the hall. The experimenter returned with a list of ten pieces of candidate information, presumably handwritten by the participants’ partner, but actually fabricated by the experimenter to have mostly shared information or mostly unshared information (i.e., the experimental manipulation). The lists of information werewritten in shorthand and contained a slight mix of shared and unshared information to enhance their perceived authenticity. After reading the bogus list of information, participants rated the knowledge, task competence, and credibility of their dyadic partner and themselves and indicated theirpreferred candidate. Results from Experiment 2 showed that participants gave not only their dyadicpartner but also themselves higher ratings of task competence and knowledge when the partner communicated mostly shared rather than mostly unshared information – an effect called mutual enhancement. Although participants did not engage in group decision-making, this study showed that group members form evaluations of one another based on the information that they share en route to consensus formation.

Confederate groups

One of the challenges of achieving group consensus is the ability to resolve disagreement among members. Inevitably, some group members have more influence in changing the preferences of others. Social impact theory (Latane, 1981) predicts that a member with higher immediacy (e.g., physical proximity) and strength (e.g., status) will exert greater influence on the group. To isolate the impact of a single group member, Hart et al. (1999) tested the immediacy and strength components of social impact theory using confederate groups.

In their study, Hart et al. (1999) composed three-person groups of two participants and one confederate, where the immediacy and strength of the confederate were experimentally manipulated. A rigged seat assignment ensured that the confederate sat either 4 ft (high immediacy) or 10 ft (low immediacy) away from the two participants. The group task was to reach a consensus on the aptitude ratings of 40 hypothetical graduate school applicants. An ice-breaking exercise before the group started this task permitted the confederate to state his status as being either high (a senior who had applied to graduate programs) or low (a freshman merely doing the experiment for class credit). First, participants individually rated the graduate school applicants. Then, the group discussed the applicants to achieve consensus on the ratings ofeach. The confederate took a minority position in this discussion by arguing that, contrary to most students, he believed that research experience was the most important criterion and GPA was the least. After discussion, group members again rated the applicants so that the researchers could determine whether participants’ postdiscussion ratings shifted toward the confederate's position (when compared to their prediscussion ratings). Results showed that participants’ ratings shifted closer to the confederate's position when the confederate had high ratherthan low status, but only in the low immediacy condition. Contrary to social impact theory, the confederate was not more influential in the high than the low immediacy condition. As this research shows, the confederate group is a useful tool for understanding the influence of a minority opinion holder in decision-making groups.

Interacting groups

Much of the research on group decision-making has used FTF groups (for a review, see Kerr & Tindale, 2004). FTF groups allow researchers to study natural communication and influence processes among members as well as the type and quality of group decision. Although it may be more difficult to isolate or control a certain communication or influence process in FTF groups, the beauty of studying such groups is the ability to examine processes as they naturally occur and relate to one another. An article adapted from my dissertation (Wittenbaum, 1998) provides an illustration. I wondered whether high-status group members, compared to low-status members, were less likely to fall prey to the discussion bias favoring shared information and more likely to influence the group's decision.

I recruited students to participate in four-person, mixed-sex groups of two men and two women. To manipulate member status experimentally, I assigned two (experienced) members to work on a practice task similar to the eventual group's task and two (inexperienced) members to work on an unrelated task. Because I wanted to control forthe effects of member sex, the two members within the same experience condition were always the same sex (so, either the two experienced members were women, or they were both men). Participants drew slips of paper from a bag that indicated the task on which they would work individually. Although this drawing appeared random, it was rigged to permit assignment to member status and to control for member sex. Specifically, the bag contained only slips intended for members in the experienced condition. The experimenter let either the two women choose first or the two men choose first, claiming that the final two slips were obviously assignments to be inexperienced, so there was no need for the final two members to reach into the bag. If they did challenge the experimenter, however, there was a secret compartment at the bottom of the bag where the experimenter could reach inand grab two slips of paper that indicated inexperienced assignments. Members were separated to work on their practice task individually. After doing so, the group was reunited to commence their group decision-making task. First, they read individually information about two hypothetical job candidates, which biased experienced members to prefer a different candidate than the inexperienced members. After indicating their preferred candidate in private, the experimenter removed the materials and audio-recorded the group's discussion and arrival at a decision. After discussion, members indicated their postdiscussion candidate preference.

Results showed that inexperienced members mentioned more shared than unshared information, replicating the typical pattern seen in groups. Experienced members, however, mentionedshared and unshared information equally often. It was more difficult than expected to get group members to prefer the intended candidate. In 35 of the 56 groups, the experienced and inexperienced members preferred the intended job candidate before discussion. Note that this is one of the limitations of FTF groups relative to confederate groups: in the latter, the researcher is able to guarantee the confederate's position and arguments. Ignoring the 21 groups where the preferencedistribution among members did not work out, the data showed that experienced members were no morelikely than inexperienced members to have their preferred candidate adopted as the group's decision. Interestingly, when examining group decisions as a function of information mentioned, inexperienced members mentioned more information than experienced members when the group chose the candidate preferred by inexperienced members. Experienced members did not have to do the same when they won. So, even though both types of members were equally influential, inexperienced members had to work harder to get their way. This experiment highlights the benefit of FTF groups in helpingto understand the communication processes of different types of group members and how such communication relates to group decisions.

Summary and caveat

The preceding research examples highlight how different types of laboratory groups aid the understanding of different aspects of group decision-making dynamics. Although most group decision-making research uses FTF groups, other group types can isolate processes regarding how members prepare for group discussion, form impressions of other group members, or respond to a minority opinion holder. Researchers who want to conduct experiments on group decision-making can select the group type that best enables them to answer their questions of interest.

It would be a mistake, however, to assume that the different types of laboratorygroups must be used to answer different research questions. Research on social ostracism (i.e., being ignored or excluded by an individual or group) shows, to the contrary, that researchers can test the same hypothesis across the different laboratory groups. Replication of a common effect across laboratory group types demonstrates the validity of an effect across different procedures and manipulations. According to Williams (1997, 2001, 2007), being ostracized results in an immediatereduction in psychological needs for self-esteem, belonging, control, and meaningful existence. I will illustrate how this effect has been demonstrated using all four types of laboratory groups.

Nezlek et al. (1997) studied the effects of rejection on nondepressed individuals who anticipated possibly joining a group (i.e., anticipated groups). False feedback from imagined others indicated that either participants were one of the three chosen for the upcoming group task (inclusion condition) or they were not chosen and would work alone (exclusion condition). Participants felt more positive and accepted when they were in the inclusion rather than the exclusion condition. Williams, Cheung, and Choi (2000) developed an imagined-group paradigm, Cyberball, to establish a baseline of when ostracism effects occur. In it, participants believe that they are playing a game of ball-toss over the internet with other players (who, in actuality, donot exist). Zadro et al. (2004) showed that participants who were included in a three-person game of Cyberball (i.e., received the ball one-third of the time) reported higher self-esteem, control, belonging, and meaningful existence compared to those in the exclusion condition (who received theball twice early on but not again during the six-minute game). Williams et al. (2002) showed that ostracism hurts when using a confederate group paradigm. Two trained confederates discussed euthanasia with a single participant, either including or ignoring the participant during the last ten minutes of discussion. Participants who were included reported higher mood, self-esteem, belonging, control, and meaningful existence compared to those who were ostracized. Finally, Wittenbaum, Shulman, and Braz (2010) established a FTF group paradigm for studying social ostracism. In groups of three, one member lacked (by random assignment) task-relevant information relative to the other two members. This uninformed member was less involved (based on observer ratings) in the group discussion while working on the task relative to the two more informed members and, as a result, felt lower self-esteem, belonging, control, and meaningful existence compared to informed members.

Like group decision-making research, social ostracism can be studied using the different types of laboratory groups. Including the latter research examples as a counterpoint to group decision-making research highlights two conclusions. First, the different types of laboratory groups can examine two important aspects of group life: task and social–emotional, with group decision making exemplifying the former and social ostracism the latter. Second, different laboratory approaches can help both to unveil various processes related to a particular group phenomenon and to replicate a common effect across paradigms. In the case of group decision-making, researchers used different types of laboratory groups to understand different aspects of the process, whereas in the case of social ostracism, the same effect was demonstrated across the different group types. Most researchers, however, will select a particular laboratory group type because of its unique ability to enable the investigation in question and its suitability given available resources.

Researchers who want to understand interaction dynamics in natural group discussion, such as nonverbal and verbal communication between members, are best served by using FTF groups. In Wittenbaum et al. (2010), despite wanting to replicate classic ostracism effects, we also wanted a FTF group paradigm that would allow us to examine the dynamic communication between included and excluded group members – something that the more artificial group types did not permit. In this way, FTF groups resemble the kind of group dynamics we encounter in everyday life (i.e., high ecological validity) more than the other laboratory group types. Confederate groups work well for understanding how group members respond to particular behaviors of others that may be difficult to promote naturally (e.g., a minority opinion holder). Moreover, researchers can record participant behavior in response to confederates, like Williams and Sommer (1997), who observed more smiling and laughing among participants who were included rather than excluded in a ball-toss gamewith two confederates. Because confederate groups are resource intensive (as I will argue later), imagined groups often can be used in their place to answer similar questions and offer more experimental control over others’ behavior. This is, in part, why Williams et al. (2000) developed the Cyberball version of the confederate ball-toss game. Finally, anticipated groups lend themselves toward understanding group member thought and behavior before group interaction, such as how members prepare for group task completion or respond to being chosen (or not chosen) for group membership.

Laboratory Set-up and Data Collection

A great research idea and design is no guarantee that the laboratory experiment will run without a hitch. Numerous considerations regarding personnel, experimental materials, participant recruitment, laboratory set-up, and data collection will help to make the laboratory experiment on groups run smoothly. I personally have found it helpful to be a bit anally retentive and obsessive compulsive when setting up and running a laboratory experiment. Others may mock my excessive tendency toward organization, but after trial and error, they typically come to see the benefits of my ways. The following section reveals my tips for how to run a laboratory experiment on groups as well as some of the challenges and frustrations.

Personnel

The best way to run a laboratory experiment on groups is to have a team of capable personnel to assist with various aspects of the work. Research assistants can help to recruitparticipants, collect the data in the laboratory, code and enter the data, and act as confederates. For an experiment using FTF groups, for example, I might have six research assistants to assist with data collection in the laboratory and two to four different assistants to code data from recorded discussions. I typically recruit the best students from my group communication courses and a few who seek me requesting to do an independent study. Qualities that I look for depend on the type of work that needs to be done.

For data collection, I look for socially skilled, graceful, well-spoken, responsible undergraduate students who have an ability to act. These students also make great confederates and, fortunately for me, are easy to find in a communication department. The field of communication tends to attract students who enjoy public speaking and social interaction. All of the group experiments that I have run require the research assistant to deliver verbal instructions to an individual or group. Because some studies using anticipated or imagined groups require giving false feedback to participants or pretending that other group members are in neighboring rooms when they actually are not, theatrical skills are helpful. For example, we were able to run the tacit coordination (Wittenbaum et al., 1995) and mutual enhancement (Wittenbaum et al., 1999) experiments witha single participant if the experimenter opened and closed doors to neighboring rooms and spoke in these rooms as if other participants were in there. In addition, research on FTF groups often involves coding group discussion. For this type of investigation, I seek some assistants who exclusively will code and not collect data (because I do not want coders to know the study's purpose or manipulations). An undergraduate student who is very smart, meticulous, focused, self-motivated, and enjoys working alone is the prototype for an assistant who codes and enters data. Often, the kind of student who would be a great assistant collecting data in the laboratory is not the same kind of student who would make a great coder. As a result, I recommend matching assistant interests and personality to the best-fitting role and taking time to train them properly.

Confederate groups require some research assistants to play the role of confederate. Finding people to play the confederate role convincingly and training them to do so takes time. Confederate paradigms are resource intensive, as one or more confederates must attend each experimental session in addition to the other research assistants who run the session. Williams et al. (2000) developed an imagined-group paradigm for studying social ostracism (i.e., Cyberball), in part, because the confederate procedure was resource intensive. Prior to Cyberball, Williams (1997) used a ball-toss paradigm with two trained confederates who ignored the lone participant in a game of ball-toss. Only one participant could be run per session and required the attendance of two trained confederates. In this way, confederate groups can be a slow and resource-intensive method of studying groups compared to using imagined groups.

Design and organization of materials

Fledgling laboratory experimenters may focus their efforts on developing the experimental manipulations, questionnaires, and measurements to the neglect of support materials that aid research assistants in collecting data. These support materials include such things such asthe experimenter log, verbal instructions, detailed procedure, procedural summary and reminders, and organization aids. The researcher needs to think about every detailed aspect of the experiment. Which questionnaires are written on and which are reused? Where are completed questionnaires stored? How do research assistants coordinate with each other in the laboratory? How do they keep trackof which groups are recorded on which tapes or DVDs? How are recordings stored? Are there times when participants are sitting idle and a filler task would be helpful? If the laboratory is shared, how should materials be stored when other researchers are using the laboratory? The list of questions seems endless. To the extent that these details are worked out in advance, the laboratory sessions will run much smoother.

Office supplies will help to organize the materials. I begin with a three-ring binder with added sleeves to house the experimenter log (e-log). This log includes a table of the experimental design where assistants can place a tally in the box where a participant or group was run. For each session, assistants complete a sheet that summarizes the activities. The procedures are sufficiently complex typically to require a pair of assistants to run each session. So, they write their names, the date, and time of the session. For each group, they indicate the group number, condition, room, and assistant who ran it. There is a comments area where assistants can write anything out of the ordinary (e.g., participant suspicion, lack of compliance, experimenter errors, late arrivals). The e-log houses a detailed document of the procedure and verbal instructions read by the research assistants. The experiment is divided into steps, starting with pre-session preparation (e.g., determining condition to be run) and ending with postsession clean-up (e.g., coding conditions on the data, storing data). For each step, the procedure identifies what the assistant should do and say. The e-log also contains a printed list of the students who signed up for eachlaboratory session. Hanging on the wall in the experimenter's room is a single-page chart depicting a brief summary of the procedure and my infamous top ten list(see Table 3.2).

Table 3.2 A Sample list of topten recommendations for laboratory research assistants

Because the social interaction laboratory that I use is shared with other researchers, I have become adept at making experiments well organized, yet compact. A plastic file box with a handle can house all of the materials. Each hanging folder inside the box is labeled with a step number in the procedure and contains the materials for that step. For example, the folder forStep 1 houses the informed consent forms. Steps that require an aggregation of materials, for example, the bogus questionnaires from the Wittenbaum et al. (1995) tacit coordination study, already have those materials compiled into a labeled folder ready to hand to participants. The box also contains envelopes for completed questionnaires (e.g., group decisions) and scanner sheets on which participants record responses for (typically postdiscussion) questionnaires that are reused. A compact CD case holds the mini-DVDs from the video recording of group discussions, and there is room for storing the e-log. The operation feels a bit like Dr Seuss's “The Cat in the Hat” in that all of research materials come out of the box and fill the laboratory during the session but neatly pack up into the box when it is time to go.

Several other practices can help research assistants to prepare and organize materials. Colors can make the experimental procedure run more smoothly. In FTF group studies where each member receives different instructions or information, we often color-code each member's materials to keep them straight. Our social interaction laboratory contains several small rooms with doors on which we hang an expandable folder that can be used for placing the materials that the group or individual will receive for the next experimental step. When recording group discussion, research assistants write the group's number on a dry erase board and display it at the start of the recording. This will let eventual coders know which group they are watching. Researchassistants are less likely to forget to record groups if instructions to do soare integrated with the verbal instructions and if a pair of assistants who can catch each other's mistakes run each session. All of these organization tools may seem a bit excessive, but they do ease complex procedures for research assistants and help to minimize errors.

Participant issues

Prior to data collection, researchers should obtain approval from their organization's Institutional Review Board (IRB). The IRB will assure that the proposed experimental procedures abide by the standards of ethical treatment of human participants. Researchers will have to specify to the IRB all procedures used with participants, including methods of recruitment, research materials, informed consent, use of deception, and debriefing. If recording group discussions, consent forms may require two signatures from participants – one indicating their consent to participate and the other, their consent to be recorded. For experiments that use deception (e.g., those using anticipated groups, imagined groups, or false feedback), experimenters mustperform a thorough debriefing. In all of the ostracism experiments, for example, the experimenters help to restore participants’ sense of well-being by admitting that inclusion vs. exclusion was randomly assigned. It is assuring to know that the discomfort felt during cyberostracism dissipates by the end of an experimental session, even before debriefing occurs (Zadro, Boland, & Richardson, 2006). All of the experiments provided as examples in this chapter used college students as participants, a common procedure for laboratory experiments on groups. Some academic departments have subject pools where undergraduate students participate in research in exchange for class credit. Many of these pools today use internet-based computer software as the interface whereresearchers post sessions for their experiments and students sign up to participate. These websites make participant recruitment a breeze, as students can make and cancel appointments at their convenience. Researchers without access to a subject pool may need to recruit participants by visiting college classrooms or posting advertisements around campus or in newspapers.

When studying FTF groups, participant recruitment can be frustrating. The collective information sharing described in Wittenbaum (1998) was a part of my dissertation. The project was larger than described in that article, requiring over 500 participants. Fortunately, I had grant funding that allowed me to pay each participant $12 for nearly two hours of research participation. Still, the undergraduate students at Miami University were not terribly motivated to offer two hours of their time for $12. I supplemented participants from the psychology department subject pool with other students from across campus whom I painfully recruited by going door-to-door in residence halls. I would collect data by day, recruit participants in the evening, and call students the night before their scheduled appointment to remind them to come. It was a grueling schedule. Part of the challenge was the requirement to have exactly two women and two men in each group. For each session, I needed to ensure not only enough participants to achieve four-person groups, but also a mixture of men and women to achieve the desired group composition. As in all of my laboratory group experiments, I overbook one or two additional students to guard against a couple failing to show up to their appointment. If the additional students show up, however, there must be an alternative task that they can complete individually for the same amount of compensation or credit as those who work in groups. The procedure for my dissertation tried my patiencebecause I could have four students show up, but if they were three women and one man, I could not obtain the sex composition that I needed. The result, in this case, was two hours of my scheduled time without any group data collection. At other times, I would have three students show up. In desperation, I would run to the building's lobby trying to recruit an additional student on the spot to achieve my desired group size of four. Some sessions would go unfilled and therefore were cancelled. Collecting data on FTF groups can be frustrating to achieve enough participants to form the desired group size and obtain the desired number of groups. Admittedly, working at a largeuniversity is helpful in this regard. The subject pool in my current department can have upwards of 1, 000 students in it within any given semester. Additionally, the subject pool software sends automatic reminders by email the day before a student's appointment to reduce the likelihoodof absence. This has made research on FTF groups more feasible.

Laboratory set-up

Before conducting laboratory experiments on groups, researchers should make sure that they have the necessary space and resources. All of my group experiments were conducted insocial interaction laboratories that contained three or four smaller rooms used by individual participants or small groups and an additional small room for the experimenter and materials. The small rooms enable the study of anticipated and imagined groups. Individuals seated in one room can betold that the other group members are scattered throughout the various other rooms. Some theatricsare required to make this cover story believable. FTF groups usually are placed at a round or rectangular table in a small room. Dividers can be placed on the tables during periods of the session when members complete tasks individually and in private. If studying the interaction dynamics within FTF groups, the small rooms can be equipped with concealed recording devices or allow space fora video camera atop a tripod. It is important to position the video camera so that all members easily can be viewed, although problems still can result. A group member can move to block the view of another member, preventing the observation of the blocked member's nonverbal behavior. Likewise, when two or more members speak at once, inability to see a member in the recording can make it difficult to know who is speaking. Having group members wear lapel microphones using a multichannel recording device can help to distinguish speakers. To help coders determine the speaker when watching the recording, begin discussion with each member introducing themselves and their member number or label each member with a numerical placard affixed to the table. Sound-proofing in the laboratory is helpful. Without it, noise from closing doors and delivering experimental instructions in other rooms may be picked up during the recording of a group's discussion. To make sure that the first experimental session will run smoothly, I recommend performing a dry run with the research assistants. During the dry run, the research assistants walk through the entire laboratory procedure without participants. This practice session often reveals whether the space, materials, and research assistants are ready for the study to begin.

Data collection

Now that the study is ready to go, when should data collection commence? If using a subject pool, timing during the semester is important. Students who participate in research at the beginning of the semester tend to be more engaged and motivated compared to those during the final weeks. If possible, try to collect data during the beginning of the semester. Group research tends to be more demanding of participants compared to other topics of study, so the more engaged participants are in the laboratory procedure, the more likely the experiment will work out as hoped. If recruiting participants from a more advanced class, make sure that the topic of investigation has not yet been covered in class. For this reason, my group communication students can participate in group decision-making research during the first half of the semester, before I cover that topic in class. Naïve students are less likely to guess the purpose of the study. As described earlier, data collection can be painstakingly slow, so the ability to achieve the desired sample size can be a blessing. If data collection is strained, as sometimes occurs during the middleof the semester, I will postpone it for another time. Peeking at the results at this point may reveal, surprisingly, that the effects are stronger than expected and emerging as statistically significant with a smaller-than-anticipated sample size. This occurred when running the social ostracism research (Wittenbaum et al., 2010), where the effects of being ignored were quite strong on the measures of threatened needs. Happily, we ended data collection sooner than expected.

Conclusion

Running laboratory experiments on groups can be both a rewarding and trying experience. The purpose of this chapter was to provide examples of how groups can be studied using laboratory experiments, illustrate best practices in running experiments on groups, and establish realistic expectations about what to expect from the process. I close with four distilled lessons from my experience conducting laboratory experiments on groups.

First, it is important to match the method to the research question. Assuming that the researcher wants to understand causal relations within group dynamics in a controlled environment, the researcher must determine which type of experimental laboratory group best fits the question of interest. Although I have always been interested in understanding group dynamics, sometimes this involves studying individuals who merely anticipate or think about other group members. At other times, it involves creating a small group to complete a shared task. Although the types of laboratory groups mentioned herein are commonly used methods of studying groups experimentally, there are countless other ways.

This leads to the second lesson: experimenting on laboratory groups requires creative thinking. It took time and many brainstorming sessions to generate a paradigm for studying social ostracism using FTF groups or one for how members tacitly coordinate in anticipation of group task completion. With clever thinking, it may be possible to study just about any group dynamic using a laboratory experiment.

Third, successfully pulling off a laboratory experiment on groups requires a lotof planning, organization, and, occasionally, some theatrics. The team of research assistants is critical for helping to make that all happen.

Finally, studying laboratory groups requires patience. It will take longer to complete than laboratory experiments on individual phenomena. However, the richness of data from group interaction and individuals who believe or expect to be a part of such interactions is difficult to exceed.

References

Frey, L. R . (1994). Introduction: Revitalizing the study of small group communication . Communication Studies, 45, 1–6 .

Hart, J. W., Stasson, M. F, & Karau, S. J. (1999). Effects of source expertise and physical distance on minority influence. Group Dynamics: Theory, Research, and Practice, 3, 81–92.

Haslam, S. A., & McGarty, C. (2004). Experimental design and causality in social psychology research. In C. Sansone, C. C. Morf, & A. T. Panter (Eds.), The Sage handbook of methods in social psychology (pp. 237–264). Thousand Oaks, CA: Sage.

Kerr, N. L., & Tindale, R. S. (2004). Group performance and decision making. Annual Review of Psychology, 55, 623–655.

Kirk, R. E. (1995). Experimental design. Pacific Grove, CA: Brooks Cole.

Latane, B. (1981). The psychology of social impact. American Psychologist, 36, 343–356.

Moreland, R. L., Hogg, M. A., & Hains, S. C. (1994). Back to the future: Social psycho logical research on groups. Journal of Experimental Social Psychology, 30, 527–555.

Nezlek, J. B., Kowalski, R. M., Leary, M. R., Blevins, T., & Holgate, S. (1997). Personality moderators of reactions to interpersonal rejection: Depression and trait self-esteem. Personality and Social Psychology Bulletin, 23, 1235–1244.

Sanna, L. J., & Parks, C. D. (1997). Group research trends in social and organizational psychology: Whatever happened to intragroup research? Psychological Science, 8, 261–267.

Smith, E. R. (2000). Research design. In H. T. Reis, & C. M. Judd (Eds.), Handbook of research methods in personality and social psychology (pp. 17–39). Cambridge : Cambridge University Press.

Williams, K. D. (1997). Social ostracism. In R. M. Kowalski (Ed.), Aversive interpersonal behaviors (pp.133–170). New York : Plenum.

Williams, K. D. (2001). Ostracism: The power of silence. New York : Guilford.

Williams, K. D. (2007). Ostracism. Annual Review of Psychology,58, 15. 1–15.28.

Williams, K. D., Cheung, C. K. T., & Choi, W. (2000). CyberOstracism: Effects of being ignored over the Internet. Journal of Personality and Social Psychology, 79, 48–62.

Williams, K. D., Govan, C. L., Croker, V., Tynan, D., Cruickshank, M., & Lam, A. (2002). Investigations into differences between social- and cyberostracism. Group Dynamics, 6, 65–77.

Williams, K. D., & Sommer, K. L. (1997). Social ostracism by one's coworkers: Does rejection lead to loafing or compensation? Personality and Social Psychology Bulletin, 23, 693–706.

Wittenbaum, G. M. (1998). Information sampling in decision-making groups: The impact of members’ task-relevant status. Small Group Research, 29, 57–84.

Wittenbaum, G. M., Hollingshead, A. B., & Botero, I. C. (2004). From cooperative to motivated information sharing in groups: Moving beyond the hidden profileparadigm. Communication Monographs, 71, 286–310.

Wittenbaum, G. M., Hubbell, A. P., & Zuckerman, C. (1999). Mutual enhancement: Toward an understanding of the collective preference for shared information. Journal of Personality and Social Psychology, 77, 967–978.

Wittenbaum, G. M., Shulman, H. C., & Braz, M. (2010). Social ostracism in task groups: The effects of group composition. Small Group Research, 41, 330–353.

Wittenbaum, G. M., Stasser, G., & Merry, C. J. (1996). Tacit coordination in anticipation of small group task completion. Journal of Experimental Social Psychology, 32, 129–152.

Zadro, L., Boland, C., & Richardson, R. (2006). How long does it last? The persistence of the effects of ostracism in the socially anxious. Journal of Experimental Social Psychology, 42, 692–697.

Zadro, L., Williams, K. D., & Richardson, R. (2004). How low can you go? Ostracism by a computer lowers belonging, control, self-esteem, and meaningful existence. Journal of Experimental Social Psychology, 40, 560–567.