Some Indirect Effects of Case Vs. Base-Rate Data on Information Processing Strategies

ABSTRACT - Recent work provides evidence that the availability of well-defined internal knowledge (i.e., the availability of private knowledge or memory for prior beliefs that is easily interpreted) determines the process employed in forming attitudinal judgments. When well-defined internal knowledge is available, judgments are formed by simply aggregating the salient cues. When "internal cues are weak, ambigUOUS, or uninterpretable (Bem, 1972, p. 2)", self-perception processes, in which attitudes are inferred from past behavior and surrounding circumstances, are invoked to guide judgments (See Chaiken and Baldwin, 1981; Tybout and Scott, in press). In view of these findings, it becomes important to specify the conditions under which well-defined internal knowledge will be available.

Citation:

Carol A. Scott and Alice M. Tybout (1983) ,"Some Indirect Effects of Case Vs. Base-Rate Data on Information Processing Strategies", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 109-111.

Advances in Consumer Research Volume 10, 1983      Pages 109-111

SOME INDIRECT EFFECTS OF CASE VS. BASE-RATE DATA ON INFORMATION PROCESSING STRATEGIES

Carol A. Scott, University of California, Los Angeles

Alice M. Tybout, Northwestern University

ABSTRACT -

Recent work provides evidence that the availability of well-defined internal knowledge (i.e., the availability of private knowledge or memory for prior beliefs that is easily interpreted) determines the process employed in forming attitudinal judgments. When well-defined internal knowledge is available, judgments are formed by simply aggregating the salient cues. When "internal cues are weak, ambigUOUS, or uninterpretable (Bem, 1972, p. 2)", self-perception processes, in which attitudes are inferred from past behavior and surrounding circumstances, are invoked to guide judgments (See Chaiken and Baldwin, 1981; Tybout and Scott, in press). In view of these findings, it becomes important to specify the conditions under which well-defined internal knowledge will be available.

Tybout and Scott (in press) provide a starting point for addressing this issue. They reason that when immediate sensory data are available, well-defined internal knowledge will exist because such data are readily Interpreted. In contrast, when only nonsensory or indirectly acquired data are available, well-defined internal knowledge may be unavailable because such data are often difficult to evaluate and employ in Judgment formation. Consistent with this hypothesis. they found that when subjects had immediate sensory data available by virtue of having tasted a product, their internal knowledge was well-defined and product judgments conformed to information aggregation predictions. When subjects had only nonsensory consensual data regarding others' product evaluations available, their internal knowledge was not well-defined and judgments were consistent with a self-perception process.

The Tybout-Scott findings provide initial evidence on the conditions necessary for well-defined internal knowledge; such knowledge is expected when immediate sensory data are available. However, because this effect is thought to be the result of the engaging and interpretable nature of immediate sensory data, it follows that nonsensory data possessing similar characteristics may also lead to well-defined internal knowledge. In the Tybout-Scott study, the nonsensory data were consensual data presented in base rate form. The pallid abstract nature of such data make them difficult to interpret and unlikely to stimulate associations in memory. Thus, it is not surprising that this data failed to produce well-defined internal knowledge. In contrast, consensual data presented in a vivid, concrete case form might he expected to operate much like sensory data. Such data are likely to be easily interpreted and should stimulate cognitive elaboration (c.f. Kisielius and Sternthal, in review; Nisbett and Ross, 1980).

The current experiment was designed to extend our previous work by varying the nature of nonsensory, consensual data and observing the impact on internal knowledge. Consensual data were delivered in either base rate or case form. When base rate data were employed it was expected that the findings of our previous study would be replicated; the availability of such data should fail to provide well-defined internal knowledge and self-perception processing should occur. However, when case data were available, well-defined internal knowledge was anticipated and information aggregation was expected.

Our predictions were tested under the guise of conducting market research for a new product. [The procedure employed is a direct extension of the one reported in Tybout and Scott (in press).] Participants, who came to a test kitchen, read descriptions of two new soft drinks as a basis for choosing one of these drinks to try. In addition, they were offered a 504 coupon for a popular fast food restaurant if they chose one of the brands. All subJects chose the brand for which the coupon was offered. Next, all subjects received both consensual data and taste data about the soft drink they chose. These two types of data were constrained to be opposite in favorableness (i.e., one was always favorable and one was always unfavorable). The experimental treatments were created by varying three aspects of the data presented: the order of type of information (consensual data first vs. taste data first, the kind of consensual data (base rate vs. case) and the order of data favorableness (unfavorable data first vs. favorable data first). Thus, eight treatments groups were created (l and 2) favorable consensual (base-rate or case) data followed by unfavorable taste data; (3 and 4) unfavorable consensual (base-rate or case) data followed by favorable taste data; (5 and 6) favorable taste data followed by unfavorable consensual (base-rate or case) data; and (7 and 8) unfavorable taste data followed by favorable consensual (base-rate or case) data.

The specific operationalizations of the treatments deserve mention. Taste data entailed drinking a sample of the chosen brand that was either good tasting (favorable) or bad tasting (unfavorable). When consensual data were presented In base-rate form, subjects overheard a research assistant say that more than 80 percent of the women in another comparison taste test preferred the brand the subjects had selected (favorable), or that 80 percent preferred another brand over the one subjects had selected (unfavorable). When consensual data were presented in case form, subjects overheard one of the other women (actually a confederate) say how much she liked the brand selected (favorable) or the alternative brand not selected (unfavorable). The manipulation of the order in which taste and consensual data were presented was designed to vary the availability of well-defined internal knowledge. When either taste data or case consensual data was presented first, well-defined internal knowledge was expected to be available when processing was initiated, whereas when base-rate consensual data were first, individuals were not expected to have access to such knowledge at the time a processing strategy was selected.

After receipt of these taste and consensual data, the dependent measures were administered. These included a cognitive response thought listing task, a set of affective semantic differential scales that formed an affected index, and a set of scales that formed an index of desire for more information before adopting the product, recommending it to friends, etc.

The predicted interaction between the order of consensual and taste data and the kind of consensual data is depicted in Figure 1.

FIGURE 1

THE PREDICTED INTERACTION OF THE ORDER OF TYPE OF INFORMATION AND THE KIND OF CONSENSUAL DATA

Because initially receiving either taste data or case consensual data was expected to result in well-defined internal knowledge being available, an information aggregation process was anticipated in these conditions. In accord with such a process, the incentive for choosing the brand should be interpreted as an additional positive cue (beyond the favorable taste or to (sensual data given) and should enhance evaluation and reduce the desire for further information. However, because initially receiving base-rate consensual data was not expected to make well-defined internal knowledge available, a self-perception process was predicted in this condition. When self-perception occurs, an external cause for behavior, such as an incentive, undermines internal attribution. As a result, subjects in this condition were expected to evaluate the product less favorably and be less certain (i.e., desire further information) than those in the other three conditions. No systematic effect of the order of data favorableness on processing strategy was anticipated.

A significant order of type of information by kind of consensual information interaction was obtained on the affective and information indices. Although this interaction is evidence for the differential effects of base-rate vs. case information on the attitude formation process, the form of the interaction only partially supports our predictions (see Figure 2). When consensual information was presented in base-rate, statistical fashion, the findings were consistent with our hypothesis and replicated in Tybout and Scott (in press). Subjects who had consensual data available rated the product less favorably than those who had taste data available.

FIGURE 2

THE OBTAINED INTERACTION OF THE ORDER OF TYPE OF INFORMATION AND THE KIND OF CONSENSUAL DATA

Further, those subjects who received the consensual data first expressed a greater desire for more information than those who received the taste data first. Finally, the affective index and the information index were significantly correlated when taste data was received first, but not when consensual data was received first. Correlation between these measures may be indicative of more well-defined internal knowledge (cf. Chaiken and Baldwin, 1981).

On the other hand, when consensual information was presented in a concrete, case format, the hypothesis of no differences between consensual first and taste first conditions was not supported. Rather, affective ratings were more positive when consensual data came first than when taste came first. Subjects in the consensual data first condition expressed a desire for less information than those in the taste first condition. Further, the affective and information indices were significantly correlated in the consensual data first condition, but not in the taste first condition.

A final indicator of the differences in the processing of cues resulting from base-rate versus case consensual data is the number of cognitive responses listed. Subjects in the case conditions listed more thoughts than those in the base-rate conditions. This main effect is qualified, however, by an order of data favorableness by order of information type interaction that shows that the number of thoughts differ between base-rate and case conditions only when the first piece of information is unfavorable.

The findings of this experiment raise several questions about the effects of base-rate versus case consensual data on information processing strategies. The results concerning the base-rate data were as expected on the basis of prior theorizing and prior empirical work. The effects of case consensual information, however, were not as expected and are somewhat more difficult to explain.

The prediction in this study was that either case consensual information or taste information would provide a basis for strong internal cues regarding one's attitude toward the product. Thus, the order in with they were presented should have no effect on judgments. The results of this experiment showed that order of presentation did have an effect, with the consensual data first treatment resulting in more favorable evaluations than the taste data first condition. A closer look then at these findings is in order.

If case data are equally as strong or vivid as the taste data, then one possibility is that subjects' judgments would simply show a recency effect. A recency effect is responsible for the effects of the case data, but it is a highly asymmetrical one. That is, subjects showed a recency effect only when the product was good. When the taste data were negative, order of presentation of data type had no significant effect. It may be that the comparative nature of the consensual message is responsible for this asymmetry. Subjects in bad taste conditions would have heard someone say only that they liked this bad tasting drink better than the other one they tasted. Because subjects had no knowledge of this other product, this information may not have been very informative. Subjects in the good taste condition, though, heard that the person preferred another product to the one subjects tried, which may have shed more evaluative light on the product. [A weakness in this explanation lies in the fact that parallel results were not obtained for base-rate data, which were also comparative.] Future experiments could be conducted to test whether a more straightforward message would have a different effect.

An alternative explanation for the effect of order of taste and case consensual may lie in subjects' relative commitment to case and taste of data. Presumably subjects have a greater commitment to inferences drawn on the basis of their own experience than to ones founded on others' experiences. Thus, the inconsistency between consensual and taste data may be less disturbing when consensual data are received before taste data produce a personal commitment to a particular view than when the order is reversed. Although the effects on the desire for information measure are consistent with the speculation of greater uncertainty for the caste-consensual data sequence, a more direct measure of the underlying process and, ultimately, manipulation of the mediating variable would be desirable.

Finally, the finding that the case condition elicited a greater number of cognitive responses than the base-rate condition is encouraging evidence for the greater cognitive elaboration presumed to accompany more vivid communications. But, this difference occurred only when unfavorable data were presented first. It may be that the unfavorable information was somewhat unexpected and thus provided more motivation for processing subsequent cues. The motivation could be heightened by subsequent conflicting information of equal strength. Future experience should investigate whether some additional motivational stimulus is needed in order to produce the illusory vividness effect (cf. Nisbett and Ross, 1980).

Unfortunately, the explanations offered for the case consensus data results are post hoc. Future research projects that include more measures of the underlying process and a greater variety of independent factors should provide more insight into both direct and indirect effects of case and base-race consensual data.

REFERENCES

Bem, D.J., "Self-Perception Theory," in L. Berkowitz (ed.), Advances in Experimental Social Psychology, New York: Academic Press, 1972).

Chaiken, S. and Baldwin, Mark W., Affective-Cognitive Consistency and the Effect of Salient Behavioral Information on the Self-Perception of Attitudes, Journal of Personality and Social Psychology, 41 (July 1981), 1-12.

Kisielius, J. and Sternthal, B., "Detecting and Explaining Vividness Effects on Social Judgments" in review.

Nisbett, R.E. and Ross, L. Human Inference: Strategies and Shortcomings of Social Judgment (Englewood Cliffs, N.J.: Prentice-Hall, 1980).

Tybout, A.M. and Scott, C.A., "Availability of Well-Defined Internal Knowledge and the Attitude Formation Process: Information Aggregation versus Self-Perception," Journal of Personality and Social Psychology, in press.

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