Why Do Focus Groups Work: a Review and Integration of Small Group Process Theories

Edward F. Fern, Virginia Polytechnic Institute and State University
ABSTRACT - Several small group process theories were reviewed and integrated into a more general theory of social impact. The resulting model suggests alternative explanations and hypotheses for the focus group phenomenon.
[ to cite ]:
Edward F. Fern (1982) ,"Why Do Focus Groups Work: a Review and Integration of Small Group Process Theories", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 444-451.

Advances in Consumer Research Volume 9, 1982      Pages 444-451

WHY DO FOCUS GROUPS WORK: A REVIEW AND INTEGRATION OF SMALL GROUP PROCESS THEORIES

Edward F. Fern, Virginia Polytechnic Institute and State University

ABSTRACT -

Several small group process theories were reviewed and integrated into a more general theory of social impact. The resulting model suggests alternative explanations and hypotheses for the focus group phenomenon.

INTRODUCTION

Focus groups have been in use for approximately twenty five years, yet most of what is known about this popular research tool comes from experiential reports by group moderators. Almost no empirical research has been done on the use of groups and no theoretical foundation has been offered to support the evidence provided by moderators. The purpose of the paper is to (1) explore the group moderators' explanations for why focus groups work, (2) review several small group process theories which may account for the focus group phenomenon, (3) integrate several theories of small group processes into a more general theory of social impact, (4) assess the implications of group process theory for focus group research, and (5) suggest theory based hypotheses for the future study of focus groups.

Several "naive theories" or hypotheses have been advanced by moderators and others as explanations for why focus groups seem to work. One hypothesis depicts the group member as hiding in "the midst of the lonely crowd" [Hess, 1968]. The group is seen as providing security for its members in the same way a crowded city provides a sanctuary for fugitives. Focus group participants, according to this explanation, have a feeling of lowered identifiability or greater anonymity than participants in individual interviews. As a consequence group members feel less inhibited and are more willing to volunteer information, even when it is highly personal.

Another explanation of the focus group phenomenon is that a norm is created which makes it okay to speak out in front of others. In the course of group discussion, the less inhibited individuals "ventilate" their experiences, which encourages others to ventilate theirs. If the "self-starters" are rewarded or receive approval, norms will be established for the more inhibited members. Once the process is well under way it tends to be "self-maintaining and self-reinforcing." As other group members see the atmosphere becoming more permissive, they become less reluctant to discuss personal experiences and habits.

Another explanation suggests that increased excitement and enthusiasm, resulting from a warm-up period, leads participants to want to get their ideas out and expose their feelings [Hess, 1968]. Goldman [1962] sees the group as providing social strength and support which members can draw upon when expressing anxiety-provoking or socially unpopular ideas.

Although these explanations make intuitive senseS one might have more faith in focus groups if small group process theories, which have been tested empirically, were found to support them. To this end a review of group process theory was undertaken. This review looks at studies done on (1) deindividuation theory, (2) social facilitation theory, (3) diffusion of responsibility theory, and (4) social impact-theory.

Deindividuation

According to Zimbardo (1969), the deindividuated internal state is characterized by diminished self-awareness and self-evaluation, and less concern for the evaluation of others. Anonymity, the presence of other people, decreased responsibility, and arousal will all lead to a deindividuated state of mind which in turn results in less restrained or uninhibited behavior. Several studies have supported the prediction that high anonymity and low identifiability result in uninhibited behavior (Dipboye, 1977). The effect of group size or the presence of an audience on deindividuated behavior has been reported less consistentlY.

The effects of group presence, anonymity, and arousal on aggression and deindividuation were studied by Diener (1976). Eighty male introductory psychology students participated in a factorial design study with two levels of group presence (group vs. alone), two levels of anonymity (anonymous vs. nonanonymous), and two levels of arousal (arousal vs. no arousal). Arousal was induced by having subjects perform the rather extreme behavior of throwing bottles against a wall. The bottle throwing activity was eliminated for the no arousal condition. Aggression was defined as "any potentially pain-producing stimuli" delivered to a role player who acted as a target for the aggressive weapons (e.g., foam swords and newspaper balls). Subjects upon entering the experimental room were told they could do "various things" to the role player with the weapons but specific behaviors were not suggested.

Observers recorded aggressive behavior but deindividuation was measured using a post experimental questionnaire. As might be expected, the aroused subjects displayed significantly more aggressive acts than the unaroused subjects. However, individuals in the three person group condition were less aggressive than those left alone with the role player. Also anonymity (whether or not other members and the role player knew the subjects names) did not result in greater aggressive behavior as expected. The latter two findings were contrary to the predictions from deindividuation theory. Responses to the questionnaire items showed that anonymous subjects did feel less concern for social evaluation, but this did not result in a significant difference in aggressive behavior. Perhaps anonymity is not a sufficient condition for causing uninhibited behaviors. A second unexpected finding was that groups were less inclined to behave aggressively than individuals. Moreover, neither groups nor individuals showed evidence of the deindividuated state as measured by the questionnaire. [Although several deindividuation measures were used in this study, social evaluation (i.e., concern for what others would think of them) was the only one on which anonymous subjects showed a significant difference.] In this particular study groups did not cause a release of inhibition.

Another extreme manipulation was used by Gergen, Gergen, and Barton (1973) to test deindividuation theory. Approximately fifty 18-25 year old college students were divided into groups of eight members, four males and four females. In the high anonymity condition, each group was put into a 10 x 12 foot room for an hour with only a pinpoint of red light. The subjects were told "There are no rules ... as to what you should do together." Subjects in this condition were compared with subjects in a lighted room to determine how a situation which is free from normative pressures affects behavior. Compared to the lights-on condition, conversation in the dark became "muted, disjointed, and faltering." Subjects in the lighted room found seats where they remained throughout the session. However, in the dark, subjects moved about continuously and frequently touched each other. Fifty percent of the dark room subjects reported hugging another person with 80 percent reporting sexual excitement. Some subjects reported feeling free, more serious, less anxious to be known by others and less anxious to know others.

In a follow-up study, Gergen and his associates used an additional 22 students similar to those in the first experiment but told them they would meet after the session. Compared to the extreme anonymity situation, these subjects were more likely to be bored, less likely to hug, and in general had less intense relationships. The Gergen report concludes "state of anonymity seems to encourage whatever potentials are most prominent at the moment - whether for good or for ill." This statement is no doubt an attempt to reconcile the reports that in some situations anonymity leads to aggressive behavior and in other situations it leads to sexual behavior. It could be alternatively argued that the red light in the Gergen study and the bottle throwing warm-up in the Diener study provided non-verbal cues as to what behavior was expected of the subjects.

Dipboye (1977), in an extensive review of the literature, attempts to integrate two different theoretical and empirical research approaches to the deindividuation phenomenon. The above studies are examples of the first approach in which lowered personal identifiability reduces moral constraints and results in antinormative behavior such as bottle throwing. For the deindividuated person, lowered identifiability is a positive or desirable experience. The second perspective views man as actively seeking an unique identity. A loss of personal identity is not seen as a pleasurable experience - rather it is a negative experience to be avoided. In this case the individual renews his/her search, presumably through anti-normative behavior, for identity. Contrary to the positive experience associated with the first perspective, being lost in the crowd or being just another face in the crowd is an undesirable experience.

In attempting to integrate these differing perspectives of deindividuation theory, Dipboye hypothesizes three moderating variables (l) structure of the social system, (2) the person's self-evaluation, and (3) prior self-awareness. He notes a common element among the studies demonstrating a release of inhibition: the studies involved groups of strangers in novel situations. In unorganized groups, facing the uncertain reactions of others in the group, deindividuation is likely to reduce inhibiting self-consciousness and arouse positive affect. In an organized permanent group where an individual occupies a role central to his or her self-concept, deindividuation is likely to serve as a threat and therefore arouses negative affect and a search for identity.

Given the unorganized social situations, the reaction of individuals to deindividuation may depend upon their self-esteem. People with low self-esteem should have positive feelings when self-awareness or self-consciousness is reduced due to the effects of (l) anonymity, (2) being part of a large group, or (3) responsibility being diffused. On the other hand, a person with high self-esteem facing the unorganized group may react by seeking out experiences that will heighten self-awareness and react to deindividuation with negative feelings and identity-seeking behavior.

The final moderating variable is prior self-awareness. Reactions to deindividuation may depend on the prior level of self-awareness but Dipboye is not at all clear on how the two concepts are related. Therefore, just what predictions would follow from this variable are uncertain.

Deindividuation theory was thought to provide a plausible explanation of what might occur in interacting discussion groups. A review of the recent studies in this area makes this notion less tenable. Where uninhibited behavior was noted, the studies used extreme manipulations which do not parallel conditions normally found in discussion groups. Dipboye's discussion provides additional insight into what variables and conditions may be necessary to account for the focus group phenomenon. This idea will be pursued later in the paper. First, another theory in which group presence plays an important role will be explored.

Social Facilitation

Unlike deindividuation theory, social facilitation theory posits the mere presence of others is a sufficient condition for explaining behavior in groups. How the presence of other people affects individual performance has been studied using either: (l) the audience paradigm, or (2) the coaction paradigm. Audience refers to the presence of passive spectators during the performance of a task. The presence of other people doing the same task but independently of the subject is termed coaction. Studies employing these paradigms have resulted in some divergent findings. Some studies have shown that the mere presence of others facilitates individual performance while other studies have shown the mere presence of others leads to decrements in performance.

Using the audience paradigm Travis [Reported in R. B. Zajonc, "Social Facilitation," Science, Vol. 149, No. 7 (July 1965), pp. 269-274.], and Dashiell [Ibid.] have shown that the mere presence of others facilitates task performance. Travis found improvement in the performance of a pursuit-rotor task when college students performed in the presence of four to eight people. Dashiell likewise found improvement in simple multiplication or word association tasks when subjects performed in front of others. Not all studies have supported these findings, however.

In a study reviewed by Cottrell [1972], Pessin asked college students to learn nonsense syllables either with others present or alone. When in front of an audience, subjects required more trials in learning the list than when alone, plus the audience condition produced significantly more errors. Other researchers using other tasks have supported the Pessin finding. Zajonc [1965] has attempted to resolve these differences by noting the Travis and Dashiell studies involved well learned tasks while the Pessin studies involved learning or acquiring a set of new responses in front of an audience.

In a review of many past studies involving humans, insects and birds, as well as a wide range of behaviors, Zajonc found support for the idea that coaction and the presence of an audience facilitates learned responses and leads to decrements in the performance of not well learned responses. Zajonc reasons that this effect is due to the audience enhancing the dominant response. The dominant response is the most probable response whether it is the right or wrong one. If the task is well learned, the dominant response is the right response. If the task has not been learned, the dominant response is most likely to be the wrong response or error. Thus, Zajonc offers a simple generalization to cover both situations, "audience enhances the emission of dominant responses."

Zajonc also provides evidence that drive, arousal, and activation increase the chances of emitting the dominant response. The mere presence of others is hypothesized to increase the individual's general drive level. If the right or sought responses are dominant, the then increased drive will improve performance by increasing the probability of correct responses. If, on the other hand, the correct or appropriate responses are subordinate to stronger incorrect responses, the increased drive state will impair performance by increasing the probability of incorrect responses.

In a modification of the Zajonc hypothesis, Cottrell (1972) suggests that individuals learn to anticipate positive and negative outcomes when others are present through classical conditioning. This anticipation and not the mere presence of others is hypothesized to increase the individual's drive level. The anticipation of negative outcomes should lead to decrements in performance of a not well learned task, while anticipated positive outcomes for a well learned task should facilitate performance. The presence of other people may also lead to decrements in performance if, as a result of their presence, each individual feels less responsible for completion of the task at hand.

Diffusion of Responsibility

Two different areas of study on group behavior have led to the explication of a diffusion of responsibility hypothesis. Wallach, Kogan, and Bem (1962) have used a diffusion of responsibility hypothesis to explain "risky shifts" in group choice situations. Diffusion of responsibility to others in a crowd has been offered by Latane and Darley (1970) as an explanation for the lack of bystander intervention in emergency situations. Both the unresponsive bystander and the member of a decision making group face essentially the same dilemma. Standing alone, each individual carries the burden of responsibility for the consequences of their decision to act or to remain passive. Failure to intervene in an emergency results in the individual receiving all the blame when others are not present. Likewise, the responsibility for a risky choice falls on the individual if others are not present. That is, a wrong decision falls on the shoulders of the lone individual. If others are present, however, the individual is less likely to receive all the blame the responsibility is diffused to all of those present.

Latane and Darley [1970] tested the diffusion hypothesis using male and female introductory psychology students in an epileptic seizure study. Subjects were led to believe they would be part of a group discussion on problems associated with college life. Actually all participants except the subject were tape recorded voices. Each subject was put in a room and was told the discussion would be held over the intercom to preserve anonymity. Each participant was to present his or her problems to the group with a free discussion to follow. The epileptic seizure victim spoke first during the discussion pointing out that he was prone to seizures, particularly when he was studying hard or taking an exam. The naive subject spoke last, after the last prerecorded voice was played. When it was the victim's next turn he feigned an epileptic seizure. The major independent variable was the number of other people in the discussion group. The lapsed time from the start of the seizure until the subject left the room to seek help was the dependent variable. The speed of response along with the percent responding is presented in Table 1. Analysis of variance on the reciprocal of response time indicated the effect of group size was highly significant. All comparisons, except between two and three person groups, were significant. According to Latane and Darley the subjects found themselves in a state of conflict and indecision about what to do. On the one hand, subjects worried about the guilt and shame they would feel if they did not help the person in distress. On the other hand, they were concerned about making fools of themselves by overreacting, about ruining the ongoing experiment by leaving their intercoms, and about destroying the anonymous nature of the situation which the experimenter had earlier stressed as important.

TABLE 1

EFFECTS OF GROUP SIZE ON LIKELIHOOD AND SPEED OF RESPONSE

Although diffusion of responsibility is a widely reported phenomenon, other types of social inhibition have been reported, namely social influence and audience inhibition (Latane, Nida, and Wilson, 1978). More recently Petty et al., reported group members put forth less cognitive effort on an evaluative task than did individuals. Seventy-five introductory psychology student i of both sexes were asked to evaluate an editorial and a poem either alone or as one of a group of four or sixteen people. After reading the communication, the students answered four questions that constituted a general evaluation of the communication and a set of three questions that comprised an effort index. As predicted, the individuals felt they put more effort into the task than did the subjects in groups. The researchers claim this finding supports their hypothesis that responsibility diffusion should lead to reduced cognitive effort.

Wallach et al., [1968] provide support for the diffusion of responsibility notion in a study on group risk taking. A rather elaborate procedure and set of instructions was developed to isolate two distinct components of the decision making process - group decision and group responsibility. The participants in this study were 336 male and female college undergraduates enrolled in the summer session. They were given booklets of questions from past College Board Examinations to answer. Each question was coded as to level of difficulty, ranging from questions which 10 percent of previous test takers failed to questions which 90 percent failed. A correct answer could win from 17 cents to $1.50 depending on the level of difficulty of the question. Each group or individual first had to decide on the difficulty level of the question to be tried and then they had to answer it.

The procedure involved experimental conditions representing combinations of group decision and group responsibility. The personal responsibility condition involved a group decision on difficulty but each individual answered each question (therefore no responsibility to the group should be felt). The group responsibility-individual decision condition had each individual deciding on his own level of difficulty. One individual was then randomly selected to answer the question with each member winning or losing based on that individual's performance. It was expected that this condition would make conservative choices in question difficulty. The group responsibility group decision conditions involved the group deciding on the difficulty of the question with an individual, randomly selected in one condition and selected by the group in the other, answering the question. Since, in the random condition, each member faced an equal chance of being selected, felt responsibility to the group should have resulted in a conservative choice in question difficulty. That is, each individual should have chosen a less difficult question than if he or she were only responsible to himself or herself. Presumably if subjects felt responsible to the group they would want to maximize the chances of the group being rewarded for a correct answer.

The results show that in the personal responsibility-group decision condition the shift in question difficulty [The shift index was the difference between mean question difficulty of a practice booklet involving no payoffs and the real booklet containing payoffs.]indicated increased risk taking compared to a control group. In the group responsibility-individual decision condition the shift was as expected in the conservative direction. When group responsibility and group decision were both present there was a strong shift toward greater risk taking whether the responsible person was randomly selected or chosen by the group. The researchers saw these findings as supportive of the diffusion of responsibility hypothesis. Since the individual was answering the questions at a difficulty level determined by the group, he may have felt that the group would have absolved him of the blame for failure. Wallach et al., suggested the group decision brought about diffusion of responsibility in two ways. First, it appeared in making the decision by pushing decisions in the risky direction. Second, it reduces the felt responsibility of any group member designated to act as the group's representative. In the eyes of the responsible group member, the group shares his responsibility since the decision is a group product. Higher risk levels with greater probability of failure apparently could be tolerated more in this situation than when the responsible individual must carry out his responsibility without benefit of communication with the group.

Social Impact Theory

Latane and Nida [1980] have proposed a more general theory of social impact which should accommodate the above theories. Latane defines social impact as "any effect of the presence or actions of other people on an individual." The effect may be "changes in physiological states, motives, cognitions, emotions, beliefs, values, and behavior that occur in an individual as a function of the real or imagined presence or actions of other people." Three basic principles of the theory will be discussed briefly.

The first principle states that the amount of impact experienced by an individual should be a multiplicative function of the strength (S), immediacy (I), and number (N) of social forces or I = f (SIN).

The second principle specifies a marginally decreasing impact on an individual as another person is added to a social situation. That is, the addition of a fourth person should have a greater impact than the addition of the eighth if individual difference variables are held constant. Specifically, Latane proposed social impact (I) should equal some root (t) of the number of sources (N) times a constant (s) or I - sNt, where t < 1.

To support the second principle, Latane and Harkins (1976) sought to determine if a power law applies to the multiplication of impact prediction. Ten of 26 introductory psychology students were used in a pretest with the remaining sixteen used in the main experiment. The main experiment involved four cross-modality matching tasks. In two of the tasks, subjects matched brightness of light with sound levels and loudness with levels of luminance. In the other two, subjects were asked to imagine reciting a poem in front of audiences of size one, two, four, eight, or sixteen, whose faces were presented on colored slides. Subjects were to make the light as bright or the sound as loud as they would be anxious, nervous, or tense in performing in front of an audience whose faces they had just been shown. The audience was composed of males or females in their early teens or late 30's. As predicted from social impact theory, tension increased at a decreasing rate as audience size increased. The exponent for the effect of audience size was estimated to be .52.

Where the second principle deals with an individual as < target of social forces, the third principle deals with others with the individual. Increasing strength, immediacy, and number of other people in this situation should lead to a division of impact. Each person in a group should feel less social force than if he/she were the sole target of the social forces. The effect of a social force from outside the group should be some power of people which is negative but less than one. Therefore, the prediction in this situation would be I - sNt, where 0<t<1. This principle was derived from the diffusion of responsibility hypothesis discussed previously.

Several studies support the division of impact idea. A recent study by Petty et al., [1977] (reviewed in the Diffusion of Responsibility section) examined the effects of group size on cognitive effort and evaluation. If cognitive effort, such as required in critically evaluating a poem or editorial, is viewed as costly to the individual and other people are available to share the cost or load, each individual may be tempted to reduce his/her own share. A trend analysis on the logarithm transforms of group size and effort index scores was significant, accounting for 91 percent of the variance. The exponent implied by the linear trend was -.10, supporting the hypothesis that perceived effort is an inverse power function of group size with an exponent less than one.

FIGURE 1

INTEGRATIVE MODEL OF SOCIAL IMPACT

An Integrative Model of Social Impact

Although most of Latane's studies have dealt with social impact in terms of social inhibition, in certain instances the effect attributed to social impact could be more specifically attributed to 'either deindividuation, identity ' seeking, or the facilitating effects associated with the presence of others. In this section the previously discussed theories will be integrated to help explain the predictions from social impact theory. The resulting model depicted in Figure l is offered to bring the relevant theories into sharper focus.

Social facilitation theory provides one avenue for explaining the multiplication of impact findings. If the presence of other people is facilitating then the contradictions between the experiential evidence offered by focus group researchers and empirical evidence offered by Latane may be explained by social facilitation theory. If the task is spontaneous thought generation, the dominant response may be well rehearsed thoughts or often repeated tales of little information value. The fear of being negatively evaluated by other group members and little promise of reward for participation in group discussion, could inhibit spontaneous thought generation (i.e., in this case the dominant response would be the wrong response). With repetition of previously stated ideas and irrelevant stories occupying much of the group's discussion time, it would seem likely that the number of different ideas generated would increase as group size increased but at a decreasing rate.

Another explanation of the multiplication of impact principle follows from the identity seeking side of deindividuation studies. Although some participants may seek solace by hiding in the crowd, others--most likely those with high self-esteemCmay have a disruptive effect with their attempts to reestablish their lost personal identity. That is, they may become "pests" in order to bring attention to themselves. The focus group moderator whose purpose is to maintain control, may thwart attempts to hide and at the same time serve as a roadblock to identity seeking. Insofar as hiding and seeking activities continue, they may be disfunctional to the group's task. By performing the very activities deemed necessary as control techniques, moderators may be unwittingly attenuating the groups effectiveness.

Contrary to the above lines of reasoning, it would be expected from the synergism hypothesis that the number of thoughts would be a monotonically increasing function of group size. This effect might be noted if the discussants were involved in a well learned task where the dominant response was more likely to be the right or acceptable response than wrong (e.g., discussing personal product experiences).

Two plausible explanations of the synergistic effect seem to be offered by deindividuation theory. The presence of other people may afford each individual some degree of perceived anonymity which may result in less self-awareness and less concern for the evaluations of others. Decreased self-awareness may in turn free the individual from personal inhibition which could result in greater freedom in thought generation. In the second explanation, generalized arousal resulting from the pre-discussion warm-up period may cause reduced personal inhibition. Getting caught up in the carefree excitement of the occasion may tend to reduce self-awareness leading to a deindividuated state.

Social facilitation theory and deindividuation theory are depicted in Figure l as affecting the nature of the individuals contribution to the group discussion (i.e., the quantity and quality or nature of the thoughts presented). These two theories can be viewed as competing explanations of the focus group Phenomenon. Future research can provide tests to determine which theory provides greater understanding of this process.

An illustration may help clarify the predictions derived from these theories. Figure 2 summarizes the effects, on the number of ideas generated as a function of group inhibition, facilitation, and no group effect. If focus groups provide a synergistic effect through interaction, an exponent greater than one would be expected. In this case the number of ideas generated by the group would be greater than the sum of the ideas generated by its members working independently of each other. If group size is inhibiting, an exponent less than one would be expected. The number of ideas generated by the group would be less than the sum of the contributions of the individual members. An exponent of one suggests that group size has no effect or the number of ideas generated by the group is equal to the number of ideas generated by members independent of each other.

FIGURE 2

NUMBER OF IDEAS AS A FUNCTION OF GROUP INHIBITION, FACILITATION, AND NO GROUP SIZE EFFECT

The division or diffusion of impact principle would suggest that, as group size increases, any one individual would feel less compelled to attend to the comments of others or would be less likely to participate in the discussion. Therefore, the number of ideas contributed by each individual should decrease as group size increases. This effect would be modeled by an exponential function but in this case the exponent would be negative as depicted in Figure 3. The diffusion effect may be caused by audience inhibition or diffusion of responsibility for participation to others.

FIGURE 3

DIFFUSION OF IMPACT AS A FUNCTION OF GROUP SIZE

Participation in the group discussion may be withheld due to evaluation apprehension. Potential negative outcomes such as looking foolish could cause some individuals to withhold participation. Also, since there are no explicit goals to strive for in focus groups there is no way for the individual to share in the rewards associated with meeting the group's goals. Additionally, an ambiguous situation calling for unrehearsed behavior, such as in focus groups, may cause participants to look to others for support. If the others respond in a safe or neutral way, the individual may find comfort in doing likewise. If others on the other hand respond brilliantly, the individual may feel his/her contribution would appear foolish by comparison. The net result might be a diffusion of responsibility or an avoidance of responsibility for contributing to the group's output.

Many of the above theories may not be directly applicable to focus groups. Most of these studies had naive subjects performing a task in front of an audience or group of passive spectators. In these studies it is fairly clear as to which direction the cause/effect relationship flows. The experiments were designed to put the subject in a single role--either the source or the target of social forces. However in interacting discussion groups the individual is both the source and target. In effect, the individual switches roles from talker to listener. When talking the individual is a target of social forces emanating from the group and a source of forces directed at the group. When listening the individual is both a target of social forces emanating from the speaker and a source directed at the speaker. As the size of the group increases the magnitude of social forces aimed at the speaker should increase. Also, as the size of the group increases, diffusion of forces from any one speaker should increase. The presence of both the multiplication of impact and division of impact in interacting discussion groups makes researching this phenomenon difficult.

Theoretical Implications For Focus Group Research

The theories reviewed in this paper are only representative of the many that could contribute to understanding why and how focus groups operate. This section will explore the implications of these theories (i.e., why and how they may operate) for the conduct of focus group research. Primary emphasis will be placed on how these theories might be used to develop a more effective focus group technique.

Up to this point no distinction has been made between the different types of focus groups. However, it now becomes useful to differentiate various types of focus group research. In an important step towards developing a theory of qualitative research, Calder [1977] has delineated three distinct focus group approaches: (l) the exploratory approach, (2) the clinical approach, and (3) the phenomenological approach. The exploratory approach defines those situations in which the researcher is looking for ideas or hypotheses that lead to scientific investigation and theory. The clinical approach reflects the perspective of clinical psychology. Its purpose is to uncover the underlying or subconscious causes of behavior. The phenomenological approach seeks everyday or experiential knowledge. approach has been most often used to discover how consumers think, feel, and behave in product purchase situations. As an aid to planning marketing strategies, the phenomenological approach seeks to uncover consumers' product experiences.

The implications which can be drawn from theories about small groups depend upon the specific approach being considered. A few theoretical constructs will be presented and their differential effect depending upon the focus group approach will be explored. First, from studies on deindividuation it was learned that anonymity, decreased responsibility, and arousal may lead to uninhibited behaviors. If group moderators want to vary the degree of inhibition felt by group members they can try to control the amount of anonymity and arousal that participants are subjected to. Turning the lights off to achieve anonymity may be rather extreme but other available means are:

(1) insuring that strangers are recruited, (2) eliminating respondent interaction before the group session, (3) eliminating personal references during the warm-up, and (4) using numerical or alphabetical labels rather than names during the sessions. On the other hand, arousal can be achieved through preplanned warm-up sessions designed for the specific focus group task at hand.

When clinical groups are conducted greater spontaneity and candor may be called for. Openness may be particularly critical when attempting to uncover preconscious motives. Therefore, enhancing respondent anonymity may be effective in helping the moderator achieve greater depth of analysis. Anonymity may also be crucial when the phenomenological approach is used to uncover product experiences -- particularly when consumers are sensitive about discussing the products (e.g., personal hygiene products). In the exploratory approach anonymity is probably a less important factor.

In situations calling for actual product usage experiences (i.e., well-learned responses), the presence of others and induced arousal may facilitate members' participation in the discussion. In the phenomenological approach a few minutes spent raising the level of arousal (i.e., getting the consumers in role) may increase each individual's drive state thereby increasing the respondents' accuracy in providing relevant product use experiences. However, heightened arousal and drive may be inappropriate and even detrimental to the goals of the clinical approach. For exploratory research arousal may or may not be appropriate depending on the nature of the specific task.

Diffusion of responsibility for participation in the group discussion may be directed horizontally toward other group S members or vertically toward the moderator. The effectiveness of clinical groups may not be impaired due to a reduced sense of responsibility because the group's goals are not so obvious (i.e., there is not much responsibility I to be diffused). On the other hand responsibility diffusion may be a problem as group structure becomes more z forma] and/or more rigid. In exploratory research the group participants may be more aware of where the group is going (i.e., more moderator-participant interaction may provide cues) and therefore foster feelings of reduced responsibility as groups size increases. In situations where the moderator's role is more central to the process (e.g., exploratory research), effort may be reduced simply because the moderator is perceived to have greater responsibility for the outcome than the respondents.

Social impact theory provides the potential for a relatively simple and straight forward way to structure focus groups to fit specific tasks. If the task is exploratory in nature (i.e., to generate lots of ideas, thoughts, or hypotheses) than the researcher may be concerned about the size of the group. In a study which compared the number and quality of ideas generated in individual interviews with focus groups of four and eight members, Fern [In press] found evidence to support the diminishing returns hypothesis. The exponent of the power function (I=N ) fitted to the data was .7 with 86% of the variance in number of unique ideas explained by the group size factor. However, if the task is to learn about consumers' experiences in using products then size may not be a critical factor. To date there is very little empirical evidence to suggest what types of focus group tasks are affected by the size of the group.

The relative strength of social forces (e.g., perceived social status) may be a critical variable when using the phenomenological approach. In this approach the role of the moderator should be unobtrusive. Therefore the moderator should attempt to neutralize any apparent differences in social status. Moreover, groups should probably be comprised of respondents with similar social backgrounds. These concerns are Probably unfounded when the task is exploratory in nature. In this approach moderator obtrusiveness is not a problem.

The immediacy (i.e., physical proximity) of social forces is also well within the control of focus group moderators. Moderators may not consciously manipulate immediacy but they seem to have individual preferences for seating arrangements. Some prefer informal seating arrangements around the perimeter of the room (low immediacy) while others prefer a more closely knit aggregation around a small table. Still others may prefer a small table with the respondents facing the moderator (high immediacy). Greater distance among group members may be called for in clinical sessions where the moderator is less involved in the group interaction. However, for exploratory research more control may be required and therefore tighter seating arrangements may be in order. With the one exception noted, none of these relationships have been studied. These notions along with others taken from the literature review provide ideas for needed research on the focus group techniques.

Hypotheses for Future Research

Several hypotheses were derived to demonstrate potential avenues for research on focus groups. These hypotheses do not represent an exhaustive list. However, they do suggest opportunities for future research in this area. The first set of hypotheses test many of the notions from deindividuation theory. If arousal and/or anonymity from being part of a group results in a release of inhibition, the productivity of focus groups may be greater and qualitatively better than the productivity of an equivalent number of individual interviews

H1.1: The overall quantity and quality of information obtained from focus groups will be greater than that obtained from an equivalent number of participants in individual interviews.

H1.2: As group size increases the more willing group members will be to participate in group discussion.

H1.3: As group size increases the more willing group members will be to disclose personal information

H1.4: As group size increases the greater the relative participation of members with high personal self-esteem will be.

H1.5: Focus groups using warm-up sessions will produce more uninhibited responses than groups not using warm-up sessions.

On the other hand, some slightly different hypotheses might be formulated from the Zajonc perspective. The presence of other group members may serve to reduce the amount of creative or unrestrained thought of group participants.

H2.1 The more familiar the group members are with the discussion topic, the more spontaneous and-uninhibited will be their participation in the presence of others.

H2.2 The less familiar the group members are with the discussion topicS the more they will resort to story telling and other irrelevant conversation.

Still other hypotheses may be formulated from the diffusion studies. Reluctance to participate in a discussion which poses the risk of looking foolish and which offers little chance for reward may attenuate the group's performance.

H3.1: As group size increases, the number of voluntary contributions per group member will decrease.

H3.2: As group size increases, the incremental cognitive effort exerted by each individual member will decrease.

H3.3: As group size increases, the amount of responsibility each member feels for the group's task decreases.

Finally, hypotheses can be derived from social impact theory. According to the Latane definition of impact (I), beliefs, attitudes and behavioral intentions can all be dependent variables in the social impact function I - f (SIN) . Additionally, immediacy (I) and strength (S) of social forces can be treated as independent variables. Although no theoretical support has been presented in this review, hypotheses could take the form of:

H4.1 As the other group member's status relative to the respondent's status increases, the more reluctant the respondent will be to participate.

H4.2: As the physical distance among group members decreases (e.g., decreasing the size of the table) responsibility diffusion decreases.

Several theories have been reviewed which may add to the understanding of the focus group phenomenon. An attempt was made to integrate these theories into a more general theory of social impact. The theories reviewed in this paper are not the only viable candidates for explaining this phenomenon. Other theories need to be reviewed and tested if a theory based focus group methodology is to be developed. Currently experiential information is the focus group researcher's only guide. It is hoped that this paper will provide a step in the direction of augmenting experience with theory based research. The result could be more confidence in-using this popular research technique.

REFERENCES

Calder, Bobby J. (August 1977), "Focus Groups and the Nature of Qualitative Marketing Research," Journal of Marketing Research, Vol. 14 pp. 353-364.

Cottrell, Nickolas B. (1972), "Social Facilitation," in C. G. McClintock (ed. ), Experimental Social Psychology (New York: Holt, Rinehart. and Winston).

Diener, Edward (January 1976), "Effects of Prior Destructive Behavior Anonymity and Group Presence on Deindividuation and Aggression," Journal of Personality and Social Psychology, Vol. 33, No. 1, pp. 497-507.

Dipboye, Robert L. (November 1977), "Alternative Approaches to Deindividuation," Psychological Bulletin, Vol. 84, No. 6, pp. 1057-1075.

Fern, Edward F. (In press), "The Use of Focus Groups For Idea Generation: The Effects of Group Size, Group Types Acquaintanceships and The Moderator on Response Quantity and Quality," Journal of Marketing Research.

Gergen, Kenneth, Gergen, Marg M. and Barton, William R. (Oct. 1973), "Deviance in the Dark," Psychology Today, Vol. 7, No. 5, pp. 129-130.

Goldman, Alfred E. (July 1962), "The Group Depth Interview," Journal of Marketing, Vol. 26, No. pp. 61-68.

Hess, John M. (1968), "Group Interviewing," in Robert L. Ring (ed.), 1968 ACR Fall Conference Proceedings (Chicago, Illinois: American Marketing Association), pp. 193-196.

Latane, Bibb and Darley, John M. (1970), The Unresponsive Bystander: Why doesn't he help? (New York: Appleton-Century Crofts).

Latane, Bibb and Nida, Steve (1980), "Social Impact Theory and Group Influence: A Social Engineering Perspective," in P. B. Paulus (ed.), Psychology of Group Influence (Hindsdale, New Jersey: Lawrence Erlbaum Associations).

Latane, Bibb, Nida, Steve and Wilson, Dave (1978), "Group Size and Helping: Review and Critique," Manuscript in preparation, The Ohio State University.

Petty, Richard, Harkins, Stephen, Williams, Ripling and Latane (Fall 1977), "The Effects of Group Size on Cognitive Effort and Evaluation," Personality and Social Psychology Bulletin, Vol. 3, pp. 579-582.

Wallach, Michael A., Kogan, Nathan and Bem, Daryl J. (September 1968), "Diffusion of Responsibility and Level of Risk Taking in Groups," Journal of Abnormal and Social Psychology, Vol. 68, No. 3, pp. 263-274.

Zajonc, Robert B. (July 1965), "Social Facilitation," Science, Vol. 149, No. 3681, pp. 269-274.

Zimbardo, P. G. (1969), The Human Choice: Individuation, Reason, and Order Versus Deindividuation, Impulse, and Chaos," in W. J. Arnold and D. Levine (eds.), Nebraska Symposium on Motivation (Lincoln: University of Nebraska Press, pp. 237-307.

----------------------------------------