2

EXPERIMENTAL DESIGNS FOR
RESEARCH ON SMALL GROUPS

The Five Ps

Patrick R. Laughlin

UNIVERSITY OF ILLINOIS AT URBANA–CHAMPAIGN

As richly attested in the chapters of this book, small groups are studied in many disciplines, including anthropology, behavioral accounting, behavioral economics, communications, management, media studies, organizational behavior, political science, social psychology, social work, and sociology, and by many methods, including laboratory experiments, surveys, participant and external observation, discourse analysis, archival research, and increasingly in our technologically integrated world, by the vast resources of the World Wide Web and the internet.

This chapter considers illustrative experimental designs for laboratory experiments on small groups. Laboratory experiments entail the full power and logic of the scientific method: (a) test of hypotheses, whether point predictions mathematically derived from quantitatively formulated theory, predictions deductively derived from well-formulated theory, less formal predictions suggested by previous results, plausible intuitions, or pure exploratory curiosity; (b) random assignment of participants to experimental conditions; (c) accepted methods of analysis of the results; and (d) accepted criteria of inference from the analyses.

Experimental designs in small group research are embedded within the “Five Ps:” Problems, Procedures, Processes, Performances, and Principles. Problems are questions about small group phenomena. For example, do groups perform better than individuals on logical reasoning? How do groups combine or aggregate the different beliefs or preferences of the group members, such as correct or incorrect on a geometric proof, or guilty or not guilty in a jury trial, in a collective group response? What is the effect of group experience on subsequent group member learning, beliefs, or preferences?

Procedures include the research setting, the group task and objective, the structure of the group such as roles of formal leader or group recorder, demographic and personality variables such as gender and extroversion, and the norms of expected behavior such as careful consideration of the views of other members.

Processes are how the group members interact with each other and influence each other in achieving the objective of the group task. Processes may be assessed as in classical rhetorical theory and current discourse analyses by who says what to whom under what circumstances with what effect. Processes may also be assessed by proposing formal models such as unanimity or majority by which groups combine the beliefs or preferences of the group members in a collective response and competitively testing the predictions of the models against the obtained data. Baron and Kerr (2002) call these two approaches to understanding group processes the social influence approach and the social combination approach.

Performances are what the groups do, such as propose a solution, make a judgment, decide guilt or innocence, or write a report. The correspondence between what the groups do and the objective of the group task defines the degree of success or failure.

Principles state the generalized relationships between independent and dependent variables. Sets of principles may then be incorporated into theories that provide understanding of the phenomena of interest. Principles are also called postulates or propositions.

We now consider illustrative problems that have motivated experimental research on small groups and present corresponding experimental designs.

Problem One: Do Groups Perform Differently from Individuals?

Symbolizing a group condition as G and an individual condition as I, we may schematize the design of an experiment to investigate Problem One as Design 1.

Each participant is randomly assigned to either the group or individual condition.

Design 1 has been widely used in research on group versus individual performance on memory, learning, problem solving, decision making, evaluative judgments (attitudes), allocation of resources, and other tasks. For example, extensive research on group versus individual problem solving has supported the robust generalization that groups perform better than the average individual on a wide range of tasks (for illustrative reviews, see Hastie, 1986; Hill, 1982; Kerr & Tindale, 2003; Levine & Moreland, 1998; Stasser & Dietz-Uhler, 2001).

Problem Two: What are the Social Combination Processes by
which Groups Map a Distribution of Group Member Beliefs or
Preferences to a Collective Group Response?

Many groups combine or aggregate the preferences of their members for different alternatives by some process such as unanimity, majority, or truth wins (a single correct member suffices for a correct group response) to formulate a collective response. What are these social combination processes?

Assume that a number of individuals first respond to some measure of memory, problem solving, decision making, attitudes, allocation of resources, and so forth. Some of these individuals are then randomly assigned to respond to the same measure as a cooperative group, and some to respond again as individuals. We may schematize this as Design 2.

Since the individual responses at Administration One are known, the composition of the groups (e.g., in a six-person group, five members are correct and one member is incorrect) at Administration Two are known, Design 2 therefore can test different social combination models such as majority wins or truth wins. For example, both Laughlin and Ellis (1986) and Stasson, Kameda, Parks, Zimmerman, and Davis (1991) found that five-person groups followed a truth-wins process on elementary algebra, geometry, and probability problems, where a single correct individual at Administration One sufficed for a correct group response on Administration Two. In general, tests of social combination models have found that the proportion of group members that is necessary and sufficient for a collective group response (problem solution, decision, judgment, choice, etc.) is inversely proportional to the demonstrability of the proposed group response (Laughlin & Ellis, 1986).

Competitive tests of the social combination processes may also be conducted from the control individuals on Administration Two; see Davis (1973) for the underlying logic and analytic methods.

Problem Three: What is the Effect of Group Experience on
Subsequent Group Member Responses?

We now expand Design 2 to a third individual administration measure of memory, problem solving, decision making, attitudes, allocation of resources, and so forth.

Design 3 addresses group-to-individual transfer.Transfer may be specific, where the same task is repeated on the third administration, or general, where the third task is a new problem, decision, or judgment, etc. of the same general class. For example, Laughlin and Ellis (1986) demonstrated specific transfer with Design 3 when the same ten algebra, geometry, and probability problems were administered three times, and Stasson et al. (1991) demonstrated general transfer with Design 3 by using the same five algebra, geometry, and probability problems for the first two administrations, and then new problems that could be solved by the same general equation or approach on the third administration.

Problem Four: What is the Effect of Repeated Group
Experience on Group-to-Individual Transfer?

The previous illustrative studies of Laughlin and Ellis (1986) and Stasson et al. (1991) on group-to-individual transfer used a single training session and a single transfer session, assessing two issues: (a) group versus individual training performance (Problem Two); and (b) group-to-individual transfer (Problem Three). Laughlin, Carey, and Kerr (2008) used Design 4 to address two further issues: (c) sufficiency and (d) completeness. Sufficiency is the issue whether one training session is sufficient for group-to-individual transfer. Completeness is the issue whether group-to-individual transfer is complete (i.e., whether individual performance on the transfer problems is at the level of performance by comparably experienced groups on the same problems).

The group or individual task was to solve letters-to-numbers problems as presented on an interactive computer terminal. The ten letters A, B, C, D, E, F, G, H, I, J were randomly coded without replacement to the ten numbers 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 (e.g., A = 3, B = 5, etc.). Instructions explained that the objective was to identify the complete coding in as few trials as possible. Each trial for the groups consisted of four stages: (a) the group members discussed and proposed an expression in any number of letters and the operators of addition and subtraction (e.g., A + B =?, AB + G =?, D + E + J – C =?); (b) the computer gave feedback on the answer in letters (e.g., A + B = G); (c) the group discussed and proposed a number for one or more letters (e.g., A = 6, J = 4); (d) the computer indicated whether proposed numbers were correct or incorrect. The full correct coding solved the problem, whereas anything less than the full correct coding required another trial.

Sets of three persons solved four problems as cooperative groups or individuals on separate terminals. For example, in Condition 2 they solved three problems as a group and then separated to solve the fourth problem as individuals. The groups performed better than the individuals on each problem. Group-to-individual transfer occurred on each of Problems Two, Three, and Four. One group experience was sufficient for transfer to occur. Transfer was complete on Problems Two and Three, but not on Problem Four, due to exceptional performance by the groups in Condition 1.

Problem Five: Do Groups Perform Better than the Best
of an Equivalent Number of Individuals and Does
This Depend upon Group Size?

Recall the well-established finding that groups perform better than individuals on a wide range of problems. Virtually all of this research has compared an equal number of groups and individuals (say, 40 three-person groups and 40 individuals). With random assignment to group and individual conditions, this compares groups and the average individual. A more stringent test of group versus individual performance would compare groups of a given size with the best of an equivalent number of individuals (say, 40 three-person groups and the best 40 of (40 × 3) = 120 individuals).

Laughlin, Hatch, Silver, and Boh (2006) used Design 5 to address Problem Five by comparing 40 two-person groups with the best 40 of 80 individuals, 40 three-person groups with the best 40 of 120 individuals, 40 four-person groups with the best 40 of 160 individuals, and 40 five-person groups with the best 40 of 200 individuals.

The 40 five-person groups had significantly fewer trials to solution than the 40 best of the 200 individuals. Four of the five individuals were randomly selected from each of the 40 replications, and the four-person groups had significantly fewer trials to solution than the 40 best individuals. Similarly, three of the five individuals were randomly selected from each of the 40 replications, and the three-person groups had significantly fewer trials to solution than the 40 best of 120 individuals; two of the five individuals were randomly selected from each of the 40 replications, and the two-person groups performed at the level of the best 40 of the 80 individuals and better than the second-best individuals. In summary, groups of size three, four, and five performed better than the best of an equivalent number of individuals, but groups of size two performed at the level of the best of two individuals.

Design 5 also addresses the effect of increasing group size. The groups of size three, four, and five performed better than groups of size two but did not differ from each other. In conclusion, groups of size three were necessary and sufficient to perform better than the best of an equivalent number of individuals on the highly intellective letters-to-numbers problems.

Problem Six: What is the Relative Importance of Multiple
Hypotheses and Multiple Evidence in Collective Induction?

Collective induction is the cooperative search for generalizations, rules, and principles. Groups such as scientific research teams or auditing teams observe patterns, regularities, and relationships, propose hypotheses to explain them, and conduct experiments to evaluate the predictions from the hypotheses. If the observations and experiments support the predictions, the hypotheses become more plausible; if the results fail to support the predictions, the hypotheses are revised or rejected.

In an experiment by Laughlin and Bonner (1999) four-person groups induced a rule that partitioned ordinary playing cards of four suits (clubs, diamonds, hearts, spades) and 13 cards per suit (ace, deuce,… king) into examples and nonexamples of the rule. Aces were assigned the numerical value 1, deuces 2, treys 3, …, tens 10, jacks 11, queens 12, kings 13. The rule could be based on any characteristics of the cards, such as suit (e.g., diamonds; spades), number (e.g., eights; jacks; multiples of three), or any numerical and/or logical operations on suit and number of any degree of complexity (e.g., diamond or spade jacks; diamonds eight and above, or spades seven and below; two clubs alternate with two hearts; two even clubs alternate with two odd hearts).

The instructions to the groups explained that the objective was to induce a rule that partitions ordinary playing cards into examples and nonexamples of the rule in as few trials as possible. The problem began with a known positive example of the rule, such as the eight of diamonds for the rule “two diamonds alternate with two clubs,” placed on a table. A trial consisted of the following four stages. First, each group member wrote his or her hypothesis (proposed correct rule) on a member hypothesis sheet. Second, the group members discussed until they reached a consensus on a group hypothesis, which a randomly selected group recorder wrote on a group hypothesis sheet. Third, the group members discussed to consensus on the play of any of the 52 cards on each of four arrays. Fourth, the experimenter classified the card on each array as either an example or nonexample of the correct rule. Examples were placed to the right of the known example on the table, and nonexamples were placed below the known example. As this cycle continued on successive trials, four progressive arrays of examples and non-examples developed in the order of play. There were ten trials of hypotheses and card plays, with no feedback on either member or group hypothesis until after the final trial.

The groups solved rule-induction problems in one of the nine conditions of a 3 × 3 factorial design, inducing the rule from one, two, or four arrays (sets of card plays) and proposing one, two, or four hypotheses per trial, as in Design 6

Performance improved with increasing arrays of evidence (card plays) but not with increasing hypotheses. This suggests that multiple evidence is relatively more important than multiple hypotheses in collective induction. Groups may be able to propose sufficient hypotheses to induce the rule, but they need evidence to evaluate the hypotheses.

Principles

Consideration of the aggregate body of findings from experimental research in an area may lead to the formulation of Principles (Postulates, Propositions). For example, Laughlin and Hollingshead (1995) proposed a theory of collective induction in the form of eight postulates and Laughlin (1999) later formulated four further postulates integrating experimental research.Table 2.1 presents these 12 postulates. Postulates 1–6 set collective induction in a general theory of group decision making. Postulates 7–8 formalize the social combination processes in collective induction. Postulates 9–12 generalize research on collective versus individual induction (Postulate 9), the relative importance of multiple hypotheses and multiple evidence (Postulate 10), influence in simultaneous collective and individual induction (Postulate 11), and the relative effectiveness of positive and negative hypothesis tests (Postulate 12). Similarly, Hollingshead (1998) aggregated her research on transactive memory in nine propositions.

Table 2.1 Collective induction: 12 postulates (Laughlin, 1999)

Thus principles (postulates, propositions) are integrated in a theory of the phenomena of interest. Theory may be defined as “a set of interrelated constructs, concepts, definitions, and propositions that present a systematic view of phenomena by specifying relations among variables, with the purpose of explaining and predicting the phenomena (Kerlinger, 1988, p. 9).”

Summary

These illustrative experimental designs for research on groups may be considered within the framework of the five Ps: problems, procedures, processes, performances, and principles. Although the illustrations are largely from research in the broad substantive area of group problem-solving, the designs may be used for research on a wide variety of issues in small groups, including evaluative judgments (attitudes), jury decisions, decision under risk or uncertainty, allocation of resources, social dilemmas, and the other issues represented in the other chapters of this book. For example, the coding analyses of influence processes in small groups of Meyers and Seibold in this volume could be conducted for the groups within one or more of the illustrative designs.

Recommendations

Recommendation one: use experimental designs that address
a number of problems in the same study

The illustrative designs address a number of problems in the same study. Similarly, we recommend that researchers design their experiments to address a number of different issues in their domain of interest. It is often informative to add an initial measure of knowledge, attitudes, personality traits, or preferences in order to conduct social combination analyses of the group process, possible covariance analyses, and regression analyses. Similarly, it is often informative to add a subsequent individual measure to determine the effect of the group interaction on the group members, such as group-to-individual transfer, or mere acquiescence in a group response that does not affect member change.

Recommendation two: use designs with more than two levels of the
independent variables

Designs with more than two levels of independent variables are more likely than designs with only two levels to: (a) determine the form of the function that relates dependent variables to independent variables; (b) determine the relative importance of two or more independent variables; (c) discover relationships that may be integrated in principles; and (d) test competing theoretical predictions. For example, Kerr and Kaufman-Gilliland (1994) used eight equally spaced levels of share size in an endowment game to test the competing social identity and commitment theories of the effects of communication on cooperation.

Recommendation three: anticipate the data analysis
when designing the experiment

Both data analytic methods such as analysis of variance or regression and accepted criteria of inference from the results are based on assumptions about the nature of the data. For example, it is frequently possible to design continuous response measures which allow more powerful data analytic techniques than dichotomous response measures. It is often possible to use randomized block designs and covariance designs to reduce the systematic variance from demographic, personality, or secular variables. Data periodontists may be helpful later, but we should brush and floss our experimental designs carefully in the planning stage.

Recommendation four: cooperate with other researchers

A major difficulty in conducting experimental research on small groups is obtaining sufficient participants for requisite statistical power when the group rather than the individual is the unit of analysis. Assuming limited allocations from departmental participant pools, group studies may be run in tandem with individual studies, both within and across laboratories and classes. Participant resources may be combined and groups or individuals run according to show-up rates. Studies which require periods of intervening activity or filler tasks, such as control individuals during the period when groups are interacting or delayed memory studies, may be coordinated and run cooperatively with other researchers: one researcher's filler task may be another researcher's task of interest. As with the previous recommendation, cooperation may be anticipated during the design stage and the appropriate data analyses may be incorporated in the research.

Coda

Our English word “problem” comes from the Greek προβαλλειν, to throw forward, literally a spear and by extension an ongoing situation. Our English word “decision” comes from the Latin decidere, to cut off, literally an arm with a sword and by extension our current more abstract meaning. Decisions in the design, conduct, analysis, and communication of experimental research on small groups are necessary cutoff points; the ongoing problems continue to motivate further research.

References

Baron, R. S., & Kerr, N. L. (2002). Group process, group decision, group action (2nd ed.). Philadelphia, PA: Open University Press .

Davis, J. H. (1973). Group decision and social interaction: A theory of social decision schemes . Psychological Review, 80, 97–125.

Hastie, R. (1986). Review essay: Experimental evidence on group accuracy. In G. Owen & B. Grofman (Eds.), Information pooling and group accuracy (pp. 129–157). Westport, CT: JAI.

Hill, G. W. (1982). Group versus individual performance: Are N + 1 heads better than one? Psychological Bulletin, 91, 517–539.

Hollingshead, A. B. (1998). Distributed knowledge and transactive processes in decision-making groups. In M. A. Neale, E. M. Mannix, & D. H Gruenfeld (Eds.), Research on managing groups and teams (Vol 1, pp. 103–123). Greenwich, CT: JAI.

Kerlinger, F. N. (1988). Foundations of behavioral research. New York : Holt, Rinehart, and Winston.

Kerr, N. L., & Kaufman-Gilliland, C. M. (1994). Communication, commitment, and cooperation in social dilemmas. Journal of Personality and Social Psychology, 66, 513–529.

Kerr, N. L., & Tindale, R. S. (2003). Group performance and decision making. Annual Review of Psychology, 56, 623–655.

Laughlin, P. R. (1999). Collective induction: Twelve postulates. Organizational Behavior and Human Decision Processes, 80, 50–69.

Laughlin, P. R., & Bonner, B. L (1999). Collective induction: Effects of multiple hypotheses and multiple evidence in two problem domains. Journal of Personality and Social Psychology, 77, 1163–1172.

Laughlin, P. R., Carey, H. R., & Kerr, N. L. (2008). Group-to-individual problem-solving transfer. Group Processes and Intergroup Relations, 11, 319–330.

Laughlin, P. R., & Ellis, A. L. (1986). Demonstrability and social combination processes on mathematical intellective tasks. Journal of Experimental Social Psychology, 22, 177–189.

Laughlin, P. R., Hatch, E. C., Silver, J. C., & Boh, L. (2006). Groups perform better than the best individuals on letters-to-numbers problems: Effects of group size. Journal of Personality and Social Psychology, 90, 644–651.

Laughlin, P. R., & Hollingshead, A. B. (1995). A theory of collective induction. Organizational Behavior and Human Decision Processes, 61, 94–107.

Levine, J., & Moreland, R. L. (1998). Small groups. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (Vol. 2, pp. 415–146). New York : McGraw-Hill.

Stasser, G., & Dietz-Uhler, B. (2001). Collective choice, judgment, and problem solving. In M. A. Hogg & R. S. Tindale (Eds.) Blackwell handbook of social psychology: Group processes (Vol. 3, pp. 31–55). Oxford : Blackwell Publishers.

Stasson, M. F., Kameda, T., Parks, C. D., Zimmerman, S. K., & Davis, J. H. (1991). Effects of assigned group consensus requirements on group problem solving and group members’ learning. Social Psychology Quarterly, 54, 25–35.