Consumer Research Validity: the Effect of Social Settings on Cognitive Responding to Television Commercials

ABSTRACT - The social setting in which exposure to television commercials takes place is a critical factor that has been neglected in consumer research. Although psychological theories suggest this factor will have important effects, much copytesting research continues to be carried out without explicit attention to social settings. The effect of the presence of others on cognitive responding to and acceptance of television advertisements was examined to assess the validity of cognitive response procedures in advertising copytesting. Results of a within-subjects study showed that varying social conditions (alone, interacting group, non-interacting group) had no effect on cognitive responding or persuasion.

Citation:

Timothy C. Brock and Sharon Shavitt (1984) ,"Consumer Research Validity: the Effect of Social Settings on Cognitive Responding to Television Commercials", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 18-23.

Advances in Consumer Research Volume 11, 1984      Pages 18-23

CONSUMER RESEARCH VALIDITY: THE EFFECT OF SOCIAL SETTINGS ON COGNITIVE RESPONDING TO TELEVISION COMMERCIALS

Timothy C. Brock, The Ohio State University

Sharon Shavitt, The Ohio State University

[The authors are grateful to Stephen G. Harkins & Richard E. Petty for their helpful suggestions.]

ABSTRACT -

The social setting in which exposure to television commercials takes place is a critical factor that has been neglected in consumer research. Although psychological theories suggest this factor will have important effects, much copytesting research continues to be carried out without explicit attention to social settings. The effect of the presence of others on cognitive responding to and acceptance of television advertisements was examined to assess the validity of cognitive response procedures in advertising copytesting. Results of a within-subjects study showed that varying social conditions (alone, interacting group, non-interacting group) had no effect on cognitive responding or persuasion.

INTRODUCTION

With increasing frequency, consumer researchers have raised concerns about the external validity and construct validity of their investigations (Sawyer, Worthing, & Sendak, 1979; Calder, Phillips, & Tybout, 1982). It has been argued, for instance, that consumer research that fails to consider explicitly "background" factors may yield results that are not generalizable to the target populations (Lynch, 1982). The unexplored and undetected effects of these neglected factors can affect our key manipulations in serious ways, often making treatment main effects difficult to interpret. The choice of factors to include deliberately in our studies must be guided by theoretical models and by concerns for realism.

One critical factor which has been virtually ignored in research on advertisements is the effect of social settings on cognitive performance and persuasion. This factor is vital for both the external validity of consumer research experiments and for the theories derived from them. In real life, for example, market researchers sometimes obtain subjects from community organizations, e.g., churches, and these subjects are often acquainted. At other times, total strangers are used, but they view commercials together with other strangers. The effect of the presence of others on the cognitive performance of individuals has always received substantial research attention from social psychologists (Allport, 1920). They have found that performance is sometimes augmented by the presence of others, whether those others are merely alongside (Zajonc, 1965) or whether those others are members of a more salient and evaluative group (Cottrell, 1972). Improvement of performance tue to others' presence is referred to as social facilitation. However, others' presence is not always facilitating. Individuals reduce their output when they perceive themselves to be part of a group that is collectiveLy responsible for 1 task (Latane, Williams, & Harkins, 1979). Reduction of performance due to others' presence is referred to as social loafing.

Social loafing has been demonstrated over a variety of situations and tasks, including cognitive performance. Petty, Harkins, & Williams (1980), using a cognitive response approach, asked subjects to evaluate the quality of stimuli. Those who thought they alone were responsible for the evaluations generated more positive thoughts and gave more positive evaluations to a high-quality stimulus than those who believed they were part of a group. However, with a low-quality stimulus, individuals gave more negative reactions than group members did. This interaction demonstrates that cognitive effort was diffused among group evaluators. Without fully understanding these social effects on cognitive processing and persuasion, we run the risk of interpreting some consumer research results in a misleading way. When unanticipated factors (such as social effects) interact with the variables we manipulate, the construct validity of our theories and the external validity of our research is jeopardized (Lynch, 1982).

Unfortunately, existing basic research is not so definitive that we can apply it readily -- without further data collection -- to prediction of the impact of social conditions on cognitive responses to advertisements. Research on cognitive performance suffers from the failure to establish different social conditions and to confirm that the established conditions were so perceived by the subjects. Often, subjects have merely been told about the presence of other recipients of a message, without their seeing these alleged other recipients or without independently confirming that they perceived they did or did not share responsibility with others for task performance (Petty, Harkins, and Williams, 1980). Thus, existing studies have lacked naturalism. A second deficiency has been the failure to employ a within-subjects experimental design, in which effects attributable to individual differences (in personality, mental capacity, liking for commercials) could be statistically excluded. All existing research lacks the sensitivity that is conferred by the within-subjects approach (Petty, Cacioppo, and Harkins, 1982). A third defect is that existing studies of social influence effects on cognitive performance have not included all the dependent measures in which we are interested, particularly total number of cognitive responses. Total number of responses generated by subjects while viewing an ad would provide an index of how extensively the ad is being processed. In a cognitive response-listing paradigm where subjects are asked to list their thoughts after receiving a persuasive message, the length of the thought-list may reflect cognitive effort.

Is cognitive processing of TV commercials affected by real-life viewing conditions? Often viewers are alone, sometimes others are nearby who are strangers, and, very often, acquaintances, friends, and family members share the same TV set. In real life there is no possible control over the viewing environment. Hence, copytesting research, to increase external validity, should attempt to be systematically representative of frequently-occurring real-life viewing circumstances. It is common, for instance, for copytesters to bring together unacquainted participants in settings (such as focus groups and location groups) in which a slight 'group feeling' might occur (Greenberg, Goldstucker, & Bellenger, 1977). Moreover, studies of persuasion conducted by social psychologists and consumer researchers typically entail running subjects in groups. Does this group method lead to social loafing and, hence, reduced cognitive performance? Strong evidence (Fern, 1982) exists that individuals generate more ideas than do members of focus groups. This suggests that allowing interaction among subjects in a cognitive response-listing paradigm may have an effect on the data. Strictly speaking, social loafing is limited to situations where felt responsibility is explicitly induced, and does not occur when the persons do not share responsibilities (Williams, Harkins, & Latane, 1981). However, subjects may assume shared responsibility even when it is not explicitly manipulated. Thus, when subjects are asked to process a persuasive message while other subjects are present, they may infer a sharing of responsibility. Having formed this perception of the situation, subjects may spontaneously diffuse responsibility, and, thus, cognitive effort.

Other elements of the cognitive response-listing situation may bias the data when others are present. For example, a "production norm" could evolve among subjects when cognitive responses are collected in groups. A subject, witnessing that others are continuing to list thoughts long after he has finished, may generate a few more thoughts for the sake of conforming. Reduction of thought output may also occur as a function of such conformity.

What will be the effects of social viewing conditions on cognitive responses and persuasion? This is the important question addressed in the present studies. Three experimental conditions were established, each designed to represent both a common real-life viewing setting as well as a condition of interest to psychological theories (e.g., social facilitation and social loafing). In the Alone condition, subjects viewed and responded to TV commercials by themselves -- only the experimenter was present. The Aggregate condition included other non-interacting viewers; this condition can be thought of as a "mere presence" condition (cf. Zajonc, 1965). In the Group condition, the viewers interacted very briefly and became somewhat acquainted before viewing the communications. These three conditions may be thought of as reflecting a continuum of decreasing group salience, from moderate to zero: that is, the presence of other viewers was made explicit in the Group condition, it was incidental in the Aggregate condition, and it didn't exist at all in the Alone condition.

METHOD

Overview

An overview of the experimental design is shown in Table 1. The study was conducted twice, in June and in August, 1982. There we-e nine commercials, seven for nationally advertised products and two political advertisements for an Ohio gubernatorial candidate. Each viewer reported to three sessions on the same day; and every viewer reacted to the same order of commercials. One condition was run in each session. The order of the three social conditions was counterbalanced in a Latin-square design. For example, on June 2, and August 18, viewers first participated in the Group condition, then in an Alone condition, and finally, in an Aggregate condition. In each of these sessions, subjects viewed three commercials. One of the commercials, Detergent 1, appeared twice, as the first and ninth (last) commercial viewed.

In sum, the independent variables or classifications were: Social Conditions (Alone, Aggregate, Group); Order of Social Conditions (e.g., Group first, Alone first, or Aggregate first, etc.); Session (First, Second, and Third); and Replication (June or August). The between-subjects factors were Order of Social Conditions and Replication; the within-subjects factors were Social Condition and Session.

Recruitment of Subjects

June. Twenty-four paid participants were recruited by means of notices in the campus newspaper soliciting staff members (non-students) who generally to their own cooking and laundry. Within the assigned day, subjects were scheduled for three one-half hour sessions at times convenient for them. In spite of these efforts, not all arrived at the scheduled times and this resulted in usable data from 17, instead of 24, subjects.

TABLE 1

OVERVIEW OF THE EXPERIMENTAL DESIGN

August. Because of the no-show rate in June, over-scheduling was used in August with similar recruitment notices; the overbooking resulted in usable data from 23 subjects.

Materials

Eight different videotaped television commercials were shown to subjects on color screens, SiX for nationally advertised products and two for an Ohio gubernatorial candidate. The order in which they were shown is as follows: Detergent 1, Decaffeinated Coffee 1, Candidate 1, Cooking Oil 1, Detergent 2, Candidate 2, Decaffeinated Coffee 2, Cooking Oil 2, Detergent 1 (repeated). Three commercials were shown in each of three sessions (see Table 1). The order of presentation of the ads was constant, with different types of products advertised within each session.

Dependence measures. Four kinds of measures (cf. Petty et al.,1981) were used: 1) Subjects listed their cognitive responses to each at on a standard thought-list form. They wrote thoughts in boxes and rated their thoughts for favorability toward the message. The sum of the positively-rated thoughts was divided by total number of thoughts listed to yield one of four dependent variables derived from the thought lists: proportion of positive thoughts. The proportion of negative thoughts (the second dependent variable) was derived in the same way. Counterarguments, which were coded by judges as more elaborated arguments against major portions of the ad, and were also scored as proportions of the total thoughts, comprised the third dependent variable. Finally, total thoughts was the critical fourth dependent variable, reflecting the amount of cognitive processing in each of the social conditions. 2) On four seven-point self-rating scales, subjects indicated their agreement with the message, how convinced they felt, whether the message was enjoyable, and to what extent they were distracted from paying attention to it. 3) A postexperimental questionnaire measured the effectiveness of the social treatments with rating scales about perceptions of others and feelings of acquaintanceship (see Appendix, "Checks on Treatment Effectiveness"). The first item, a count of other participants, was used only in the August study. 4) The final questionnaire included a brand checklist (asking subjects which brands of the advertised products they normally use) and questions about age and income.

Procedure

Site. The experiment was run in a language laboratory at a major university. The room was equipped with 32 individual carrels and two color TV screens at the front of the room. Subjects remained in their carrels for the duration of each session, which included viewing the commercials and filling out the paper-and-pencil measures described above

Social conditions. In the Alone condition, a subject participated by herself, with only an experimenter present. In the Aggregate condition, a subject was with 5 to 7 confederates throughout the session, but no interaction took place. The confederates were trained to avoid any interaction with the true subject and, in all other ways, to resemble true subjects. The true subject was assigned to a center seat so that she could be aware of other "subjects" around her.

In the Group condition, all of the day's true subjects (4 to 8) participated together, as well as one confederate. The role of this confederate was to stimulate acquaintanceship by being the first to respond to some "let's get acquainted" questions put to the group by the experimenter. The experimenter began by saying that, before viewing the upcoming commercials, she was interested in getting to know a little bit about the participants. Subjects were asked to stand up, so that they could see each other better, and respond to questions such as "where are you from?", "where do you work?" and "do you have any plans for how to spend the money you earn for participating in this project?' The experimenter addressed these questions first to the confederate, and then, in turn, to each of the true subjects. The confederate replied in an open and enthusiastic fashion that was emulated by most subjects. This stimulated acquaintanceship procedure took about six minutes.

Administration of dependent measures. At their first session, subjects were thanked for their participation in what was referred to as a "communications project." They were told that the experimenter was interested in their reactions to some television commercials. Then, the session began: the Alone and Aggregate sessions started with the viewing of the first commercial for that session. Each Group session began with the question-and-answer procedure described above, followed by the viewing of the first commercial. After subjects were shown the first TV ad, they were asked to "take about four minutes" to fill out the thought-listing form and use plus('), minus(-), or zero(0) to indicate favorableness-unfavorableness. After four minutes had expired, the experimenter collected the thought-box form and gave subjects the four attitude scales (agreement, convincingness, etc.). Following the thought-listing and agreement ratings for the third and last commercial of a session, the treatment effectiveness checks (see Appendix) were administered. At the end of the third and final session of the day, the subject was also given the brand checklist and the background questionnaire. The subject was then paid, thanked for her participation, and dismissed.

RESULTS

Treatment Effectiveness Checks

Recall that a principal deficiency of previous research has been the failure to set up different social conditions in a naturalistic fashion and to confirm that the different conditions were so perceived by the subjects. In the present Aggregate condition, trained confederates simulated actual subjects so that the true subject would perceive that others were present, although there was no social interaction. In the Group condition, social interaction and acquaintanceship were stimulated by a confederate who answered the experimenter's questions in an open and enthusiastic manner and by subjects standing up and observing each other answering the same questions. Checks on the effectiveness of these treatments were administered at the end of each session (see Appendix: 'Checks on Treatment Effectiveness').

Count of other participants. The first item, a count of other participants by the subjects, was used in the August replication. All subjects in the Alone condition circled "O" whereas subjects in the Aggregate and Group conditions circled "5" to "8", corresponding closely to the actual numbers of other persons who were present.

Sum of three acquaintanceship ratings. Scores for the three treatment effectiveness checks in the Appendix were summed and the means are recorded in Table 2. (Some subjects did not complete these measures in all social conditions and this accounts for the reduced N.)

Analyses of variance indicated that feeling of being acquainted showed a highly significant difference between conditions (2<.0002). Post-hoc pairwise comparisons (Least Significant Difference tests) indicated that subjects in the Group condition (Mean = 10.6) felt more acquainted than subjects in either the Aggregate (Mean = 5.8) or Alone (Mean = 6.4) conditions (p's<.05). There was no reliable difference between Aggregate and Alone conditions on the summed treatment effectiveness measure. Four subjects in Alone conditions appeared to misunderstand the instructions for they used the high end of the scale, "7", to indicate their perceptions. This accounted for the high mean scores in the Alone condition on June 4 and August 18.

Rating of "feeling part of a group". Analyses of variance on each of the treatment effectiveness checks (questions, two, three, and four in Appendix) all showed significant differences between conditions p's<.05). Post-hoc Least Significant Difference (LSD) tests on the second and fourth questions yielded significant differences between subjects in the Group condition and subjects in either the Aggregate or Alone conditions (all p's<.05). Perhaps the key single item is the third question about feeling oneself to be part of a group. The mean scores shown in Table 3 suggest that feeling of groupness increased from Alone condition to Aggregate condition to Group condition. These differences reached conventional levels of significance between the Group and Alone conditions (LSD test: p<.05).

In sum, it appeared that subjects not only correctly perceived how many others were present during the viewing and thought listing, but they also had feelings of being acquainted and of being part of a group to a greater extent in the Group than in the Aggregate or Alone conditions. Thus, different social conditions for viewing TV commercials were established in a naturalistic fashion and these differences were clearly felt and expressed by the participants.

Effects of Social Viewing Conditions on Cognitive Responses

A multivariate analysis of variance was performed with Replication (June vs. August) and Order (see rows of Table 1) as between-subjects factors and Social Condition (Group vs. Alone vs. Aggregate) as the within-subjects factor. The analysis incorporated eight dependent variables: the four cognitive response measures [Indices were formed for the cognitive response dependent measures by averaging the cognitive response data across each of the three ads in a session. Thus, the total number of thoughts index was the mean of the thoughts listed for each of the three ads. The three remaining cognitive response measures were represented as proportions (of positive thoughts, negative thoughts, or counterarguments). These proportions were subjected to arc sin transformations and the analysis was conducted on the means of these transformed proportions for each of the three ads.] and the four attitudinal measures (agreement, etc.). For each of the above factors the multivariate F was nonsignificant, p>.05 (see Table 4 for mean scores in each social condition).

TABLE 2

SUM OF THREE ACQUAINTANCESHIP RATINGS IN GROUP, AGGREGATE, AND ALONE CONDITIONS: MEAN SCORES

TABLE 3

MEAN RATING OF 'FEELING PART OF A GROUP IN GROUP, ALONE, AND AGGREGATE CONDITIONS

Session (first vs. second vs. third) was represented as a Social Condition X Order of Conditions interaction; this interaction yielded a significant multivariate F, p<.05. Examination of this effect for separate cognitive response dependent variables showed two significant univariate Fs: total thoughts (p<.001) and proportion of counterarguments (p<.05). (Untransformed means are shown in Table 5). Apparently, subjects listed more thoughts and a greater proportion of counterarguments during the second than during the first or third sessions. This "second session peak" effect suggests that thought-listing and counterarguing is facilitated with practice, but is diminished after several trials. There were no other significant univariate Fs for the cognitive resPonse data.

TABLE 4

MEAN SCORES ON EACH DEPENDENT MEASURE IN GROUP, ALONE, AND AGGREGATE CONDITIONS

TABLE 5

TOTAL THOUGHTS AND PROPORTION OF COUNTERARGUMENTS IN EACH SESSION (ORDER X CONDITION INTERACTION)

Effects of Repeating Detergent 1

Subjects saw Detergent 1 a second time, as the ninth or last commercial. Sign tests, adjusted for large samples, were performed for the five dependent measures listed above. The only significant effect, p<.01, was for total thoughts: Detergent 1 elicited more thoughts the second time around.

CONCLUSIONS

The present studies succeeded in remedying the principal shortcomings of applying previous formulations to social influence on cognitive responses to advertisements- Alone, aggregate and group conditions were realistically created. Error due to individual differences was eliminated by using a sensitive within-subjects experimental design. The cognitive response measures of interest in evaluating TV advertisements were employed.

The overall absence of effects of different social viewing conditions, and the robustness of this finding across two studies, permits three important conclusions:

1. Present group data collection methods can be continued. Neither theoretically predicted source of bias, social loafing or social inhibition, was detected.

2. Present group data collection methods, in which unacquainted persons are "aggregated", may yield cognitive responses that are no different in magnitude and favorability from those that would be elicited from the same persons when they view TV alone or when they view TV together with friends. The results have attested to the external validity of the present method. When faced with a task for which distribution of responsibility is unspecified, subjects do not appear to spontaneously diffuse responsibility for message processing. Also, cognitive response data do not appear to be threatened by conformity to thought-production norms.

3. Cognitive responses are robust. Thoughts about a TV commercial are not readily modifiable by social changes in the viewing or thinking environment.

We began with the concern that a "background factor" of social settings may pose a serious threat to the validity of consumer research. In order to definitively rule out any effects of such a factor, a sensitive experimental design which employs high-powered statistical tests was required,,(cf. Lynch, 1982). The highly sensitive "within-subjects design was used, allowing definitive conclusions to be drawn from its powerful statistical analyses. The adoption of this rarely-used approach is strongly advocated as a standard design in establishing the validity of consumer research.

APPENDIX

CHECKS ON TREATMENT EFFECTIVENESS

REFERENCES

Allport, Floyd H. (1920), "The Influence of the Group Upon Association and Thought," Journal of Experimental Psychology, 3, 159-182.

Calder, Bobby, J., Lynn W. Phillips, and Alice M. Tybout (1982), "The Concept of External Validity," Journal of Consumer Research, 9, 240-244.

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

Fern, Edward F. (1982), "The Use of Focus Groups for Idea Generation: The Effects of Group Size, Acquaintanceship, and Moderator on Response Quantity and Quality," Journal of Marketing Research, 19, 1-13.

Greenberg, B. A., J. L. Goldstucker, and D. N. Bellenger (1977), "What Techniques are Used by Marketing Researchers in Business, " Journal of Marketing, 41, 61-68.

Latane, Bibb, Kipling D. Williams and Stephen G. Harkins (1975), "Many Hands Make Light the Work: The Causes and Consequences of Social Loafing," Journal of Personality and Social Psychology, 37, 822-832.

Lynch, John G., Jr. (1982), "On the External Validity of Experiments in Consumer Research," Journal of Consumer Research, 9, 225-239.

Petty, Richard, E., John T. Cacioppo, and Stephen G. Harkins (1982), "Group-Size Effects on Cognitive Effort and Attitude Change" in Hare, A. P., Blumberg, H. H., Kent, V., and Davies, M. (Eds.), Small Groups: Social-Psychological Processes. Social Action, and Living Together, London: John Wiley and Sons.

Petty, Richard E., Stephen G. Harkins, and Kipling D. Williams (1980), "The Effects of Group Diffusion of Cognitive Effort on Attitudes: An Information-Processing View," Journal of Personality and Social Psychology, 38, 81-92.

Petty, Richard E., Thomas M. Ostrom, and Timothy C. Brock (1981), Cognitive Responses in Persuasion, Hillsdale, N.J.: Lawrence Erlbaum Associates.

Sawyer, Alan G., Parker M. Worthing, Paul E. Sendak (1979), "The Role of Laboratory Experiments to Test Marketing Strategies," Journal of Marketing, 43, 60-67.

Williams, Kipling, D., Stephen G. Harkins, and Bibb Latane (1981), "Identifiability as a Deterrent to Social Loafing: Two Cheering Experiments," Journal of Personality and Social Psychology, 40, 303-311.

Zajonc, Robert B. (1965), "The Requirements and Design of a Standard Group Task," Journal of Experimental Social Psychology, 1, 71-88.

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