CATHERINE W. M. YEUNG AND ROBERT S. WYER, JR.*

Transcription

1 1 Affect, Pictures and Consumer Judgment: The Impact of Initial Appraisals of a Product on Evaluations and Expectancy-Confirmatory Information Processing CATHERINE W. M. YEUNG AND ROBERT S. WYER, JR.*

2 2 *Catherine W. M. Yeung is a doctoral student in Marketing, and Robert S. Wyer, Jr., is a Visiting Professor of Marketing, at the Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong ( cyeung@ust.hk or mkwyer@ust.hk). The research was performed as part of a Ph.D. dissertation by the first author, and was supported by grants RGC HKUST 6053/01H and RGC HKUST 6022/00H from the Hong Kong government. Appreciation is extended to Agnes Chan for assistance in collecting data for the studies reported.

3 3 When consumers encounter a picture of a product before they receive specific information about its attributes, they form an initial impression of its desirability. They then use this impression as a basis for their later evaluations of the product independently of specific attribute information that they receive later. When the picture of the product elicits affective reactions, the extraneous affect they happen to be experiencing at the time they see the picture influences their product evaluations through its mediating impact on their initial impression rather than information processing at the time of judgment. Three experiments confirmed implications of these effects.

4 4 People often use the affect they are experiencing as information about their feelings about an object they are asked to judge and, therefore, as a basis for evaluating the object (Schwarz and Clore 1996; Wyer, Clore, and Isbell 1999). The recognition of this possibility has stimulated numerous investigations of the informational influence of affect on consumer judgments and decisions (Adaval 2001; Gorn 1982; Pham 1998; Pham, Cohen, Pracejus, and Hughes 2001; Shiv and Fedorikhin 1999). This research has circumscribed the conditions in which affect has an impact on product evaluations, and has examined the cognitive mechanisms that underlie this impact. A common finding is that people often confuse their actual feelings about a product with the feelings they happen to be experiencing for other, totally unrelated reasons. Consequently, these extraneous feelings can also have an impact on the judgments they make (Adaval 2001; Pham 1998). In most of this research, however, the informational influence of affect is assumed to occur at the time judgments are reported. In many situations, people form an initial impression of a stimulus before they receive specific information about it and make a judgment. In these circumstances, affect could have an influence at this earlier, impression-formation stage of processing rather than at the time of judgment. This possibility is particularly important to consider in the consumer domain. Consumers first encounter with a product often occurs when they see it in a store window, or come upon a picture of it in a magazine. Under these conditions, they may make a spontaneous appraisal of the product on the basis of its physical appearance alone (see Lazarus 1982, 1991, for a more general discussion of this appraisal process). Although the appraisal is cognitive (Lazarus, 1982), it can often give rise to positive or negative affective reactions, and these reactions, in turn, can provide the basis for a general impression of the product as desirable or undesirable. Once this initial impression is formed, consumers might use it as a basis for evaluating the product independently of any more specific information about the product that is presented subsequently (for evidence of this possibility in other domains, see Bodenhausen and Wyer 1985; Lingle and Ostrom 1979; Srull and Wyer 1989). The influence of affective reactions on initial impression formation has not been investigated, however. The three experiments to be reported examined the nature of this influence and circumscribed the conditions in which it occurs. Experiment 1 showed that when subjects receive information about a product s attributes without having made an initial appraisal of it, the affect they happen to be experiencing only influences their evaluations of the product when they believe that their affective reactions to the product are a relevant basis for evaluating it. When subjects are exposed to an attractive picture of the product, however, the affect they are experiencing at the time they see the picture influences their later product evaluations independently of other, more specific attribute information they receive subsequently, and regardless of the specific judgmental criteria they might normally use to evaluate the implications of this information. A second study confirmed these conclusions and, in addition, demonstrated that the affect-based initial impressions that subjects form of a product not only influence their judgments but also stimulate attention to additional information that confirms the validity of these impressions. Finally, experiment 3 showed that the impact of affect was restricted to conditions in which people s picture-based appraisals are likely to elicit affective reactions. That is, a picture that showed details of a product s construction without conveying its overall appearance eliminated the informational influence of affect that was evident

5 5 in earlier experiments. The basis for predicting these effects and a discussion of their implications are reviewed in the pages to follow. EFFECTS OF INITIAL APPRAISALS Effects on Judgments People rarely conduct an extensive analysis of all of the information they receive (Taylor and Fiske 1978; Wyer and Srull 1989). Rather, they typically base their judgments on the first relevant criterion that comes to mind. Therefore, if the initial information they receive about a stimulus stimulates them to form an initial impression of it as desirable or undesirable, they may use this impression as a basis for the judgments they report without considering more specific information that becomes available subsequently (Chaiken 1980, 1987). The impact of initial impressions on judgments is well documented. In the person domain, initial impressions can be based on the name of a person that activates an ethnic stereotype (Bodenhausen 1988; Bodenhausen and Wyer 1985), or a cluster of trait adjectives (Srull and Wyer 1989). They can also be influenced by the source of the information about the object to be judged (cf. Chaiken 1980). In all cases, people appear to use the impression they have formed as a basis for judgment without considering the implications of more specific information they receive later. Effects on Information Seeking Although people may base their judgments of a stimulus on their initial impressions of it, they do not completely ignore information they encounter later. Rather, they may use this information to confirm the implications of the impression they formed at the outset. In Bodenhausen and Wyer s (1985) study, for example, subjects based their judgments of a person on a cultural stereotype that was activated by the person s name. Nevertheless, they thought about implications of the information they received later, giving particular attention to aspects that confirmed their stereotype-based expectations (as evidenced by better recall of these aspects than other, expectancy-disconfirming information). This confirmatory information processing could have several determinants. On one hand, people s initial impression of a stimulus could give rise to an implicit hypothesis about what the stimulus is like and individuals might actively seek information that confirms this hypothesis (Snyder 1981). This tendency could be reflected in not only the type of information that people actively seek (Snyder and Swann 1978; Swann and Guliano 1987) but also their selective attention to information they actually receive (Snyder and Cantor 1979; Snyder and Uranowitz 1978).

6 6 In Snyder et al. s studies, the hypotheses that subjects tested were given to them by the experimenter. However, hypotheses could be self-generated as well. When people expect a priori that a stimulus will be of a given type, for example, this expectation could serve as an implicit hypothesis that they attempt to confirm spontaneously. Chernev (2001), for example, found that people who had an initial preference for a product tended to be more influenced by information that confirmed this preference than information that disconfirmed it. This suggests that initial preferences can influence information seeking as well as hypotheses that are provided by someone else. The tendency to engage in confirmatory information seeking is not necessarily a result of conscious deliberation. Wyer and Srull (1989) assume that when people expect a stimulus to belong to a given category and have an opportunity to obtain more information about it, they spontaneously activate a set of trait and behavior concepts that apply to typical members of this category. Once thee concepts become accessible in memory, they concepts not only provide the basis for questions about the target but also guide their attention to information they actually receive. This tendency may be particularly evident when initial impressions are based on affect. Pham et al. (2001) found that affective reactions to a stimulus are likely to elicit thoughts about the stimulus that are evaluatively consistent with them. Concepts that are activated by these thoughts could bias information seeking in much the same way as expectations based on other, non-affective criteria. We elaborate this possibility presently. THE IMPACT OF AFFECT ON APPRAISALS, INITIAL IMPRESSIONS, AND INFORMATION PROCESSING The influence of affective reactions on judgments is well established in both consumer research (Adaval 2001; Pham 1998; Shiv and Fedorikhin 1999) and elsewhere (for reviews, see Clore, Schwarz and Conway 1994; Schwarz and Clore 1996; Wyer et al. 1999). However, although this influence has been attributed to several factors (Bower 1981; Forgas 1995; Gorn 1982; Isen, Shalker, Clark, and Karp 1978), its primary impact results from its use as information. In many cases, the affect can be elicited by the stimulus itself and, therefore, is objectively relevant to its evaluation (Schwarz 2001; for an empirical example of its effects on product evaluations, see Shiv and Fedorikhin 1999). However, people s judgments can also be influenced by the affect they happen to be experiencing for reasons that have nothing to do with the stimuli being judged. This is particularly likely when people cannot distinguish clearly between the affect that is actually elicited by the stimuli they are judging and the affect they happen to be experiencing for other, irrelevant reasons. Under these conditions, they are likely to misattribute a portion of this extraneous affect to their feelings about the stimulus. Consequently, these extraneous feelings can combine with their actual affective reactions to the stimulus to influence their evaluations. Thus, happy persons, relative to unhappy ones, report themselves to be more satisfied with their life as a whole (Schwarz and Clore 1983; Strack, Schwarz and Gschneidinger 1987). They also report being more amused by jokes and cartoons (Strack, Martin, and Stepper 1988) and make more favorable evaluations of themselves, other persons and products (Levine, Wyer and Schwarz 1994; Ottati and Isbell 1996; Pham 1998). As noted earlier, however, the affect that people experienced in most of this research was assumed to exert its influence at the time of judgment. Under these conditions, people are likely to use

7 7 their affect as a basis for judgment only if they consider it relevant to the judgment they are making. Not all evaluative judgments are based on affect, as Zanna and Rempel (1988) note. This is particularly true in the product domain. In some cases, products are typically evaluated on the basis of hedonic criteria (e.g., taste, physical attractiveness, comfortableness), and these judgments are likely to require an assessment of subjective feelings. In other cases, however, product evaluations are based primarily on utilitarian considerations (e.g., quality of workmanship, or the ability to perform specified functions). As Pham (1998) and Adaval (2001) both showed, people s affective reactions only influence judgments that are based on hedonic criteria. When judgments are based on utilitarian considerations, subjects apparently do not consider their feelings to be relevant, and so the extraneous affect they happen to be experiencing has no impact. This contingency, however, may only be evident when people consciously assess the relevance of their affective reactions to their judgments at the time they are asked to report these judgments. Suppose people who see a picture of a product make a spontaneous appraisal of it that elicits affective reactions, and they form an initial impression of the product on the basis of these reactions. Once this affect-based impression is formed, it may determine the judgment they report later regardless of other considerations. If this is so, the affective reactions that provide the basis for their impressions should influence their evaluations regardless of the judgmental criteria they might otherwise apply. SUMMARY AND INITIAL HYPOTHESES The implications of the preceding discussion can be summarized as follows. First, suppose subjects are asked to evaluate a product on the basis of specific attribute descriptions without having the opportunity to make a global appraisal of it. In this case, they are likely to base their evaluations on criteria that are salient to them at the time of judgment. To this extent, the affect they happen to be experiencing should have an impact on judgments that are based on hedonic criteria but should not influence judgments that are based on utilitarian considerations. This should be true regardless of whether the affect they are experiencing is elicited by the product s attributes or is being experienced for other, unrelated reasons. To formalize: H1: If subjects are asked to evaluate a product on the basis of information about its specific attributes without having made an initial appraisal of it, the extraneous affect they are experiencing will have a positive influence on their evaluations of products that are based on hedonic criteria. (That is, they will evaluate products more favorably if they are feeling happy than if they are not.) However, their feelings will have no effect on their evaluations of products that are judged on the basis of utilitarian criteria. However, now suppose subjects are exposed to an attractive picture of the product before they receive information about its attributes. In this case, subjects should make a spontaneous cognitive appraisal of the product on the basis of its physical appearance (Lazarus 1982, 1991). This appraisal is theoretically holistic and does not involve a detailed analysis of individual features (Lazarus 1982). In

8 8 many cases, however, the appraisal can elicit affective reactions, and these reactions can be used as a basis for an initial impression of the product as desirable or undesirable. Once subjects form this impression, they should later base their judgment on it regardless of the attribute information they receive later and independently of the specific criteria they might otherwise apply in making their judgments. To this extent, extraneous affect that subjects happen to be experiencing at the time they appraise the product should become confused with the affect elicited by the picture and, therefore, should influence the impression they form. This affect-based impression should determine the judgments they report later. Moreover, the effects of this impression should be evident regardless of whether the product is usually evaluated on the basis of hedonic criteria or utilitarian ones. On the other hand, extraneous affect should only have an impact on subjects judgments if they are experiencing this affect at the time they appraise the product and form an initial impression of it. Suppose subjects have already formed an initial impression of the product at the time the extraneous affect is induced. Then, this affect should not influence their impressions, and consequently it should have no impact on the judgments they report later. The first two studies to be reported evaluated this contingency. In each experiment, subjects were induced to feel happy or unhappy for reasons that were irrelevant to the product evaluation task they were asked to perform. In some cases, this affect was induced before a picture of the product was presented and, therefore, existed at the time the appraisal of the product was made. In other cases, however, the picture of the product was presented at the outset, and extraneous affect was not induced until afterwards. Under these conditions, this affect should theoretically have no effect. Specifically: H2: When extraneous affect is induced before a picture of the product is presented subjects will evaluate the product more favorably when they are feeling happy than when they are not, and this will be true regardless of whether hedonic or utilitarian criteria are typically used to evaluate it. When extraneous affect is not induced until after the product is presented, however, it will have no impact on product evaluations. EXPERIMENT 1 The first experiment provided an initial test of hypotheses 1 and 2. Subjects were asked to evaluate a product (running shoes) on the basis of a series of attribute descriptions. Some subjects were told to use hedonic criteria for evaluating the product and others were told to use utilitarian criteria. Before receiving the product information, subjects in were induced to feel either happy or unhappy by recalling and writing about a pleasant or unpleasant personal experience. In two conditions (relevant to the test of hypothesis 2), subjects were shown an attractive picture of the product. In these conditions, extraneous affect was either induced at the outset, before the picture was presented, or not until afterwards (and, therefore, not until subjects had presumably made an initial appraisal of the product). In a third condition (relevant to a test of hypothesis 1), no picture of the product was presented.

9 9 Method Overview. Subjects were 104 introductory marketing students at the Hong Kong University of Science and Technology who were paid $80 HKD ($10 US D) for their services. Between eight and 10 subjects were assigned randomly to each cell of a 2 (induced affect: positive vs. negative) x 2 (judgment criterion: hedonic vs. utilitarian) x 3 (picture conditions: no picture vs. affect-picture order vs. picture-affect order) design. Product and Attitude Selection. Sports shoes were selected on the basis of focus group discussions, which indicated that both hedonic and utilitarian criteria were important to their evaluation. To confirm this assumption, 39 undergraduate marketing students who did not participate in the main experiment were asked to indicate the extent to which they were likely to consider each of several different features when evaluating a pair of sports shoes. Two features, quality of construction and how it would feel to wear the shoes, were assumed to exemplify utilitarian and hedonic criteria for judgment, respectively (for confirmation of this assumption, see Adaval 2001). These features, which were embedded in a number of others in the questionnaire, were evaluated along a scale from 0 (would not consider at all) to 10 (would definitely consider). The importance of each criterion was high (8.69 vs for utilitarian vs. hedonic criteria, respectively) and did not differ from one another, F (1,38) = 2.91, p >.10. Thus, the two types of criteria were equally relevant to the evaluations of products of the type we selected. The 39 students who participated in the product-selection pretest were also asked to generate a list of features they would personally consider when buying a pair of sport shoes. The six most frequently mentioned features were selected for use in constructing attribute descriptions. Three of these features were reworded to convey favorable values (specifically, comprehensive air cushioning, made of breathable material, and striped rubber outside for grip ), and three were worded to convey unfavorable values ( becomes dirty easily, is not wide enough, and 30% of the air cushioning will leak after one year ). The favorableness of these features was confirmed on the basis of ratings by a different group of 37 undergraduates who evaluated each attribute along a scale from 5 (very unfavorable) to +5 (very favorable). Ratings of the three favorable attributes ranged from 3.22 to 4.16 (M = 3.64). Ratings of the unfavorable attributes ranged from to (M = -2.94). These attributes were presented to subjects in the main study in a manner to be described.

10 10 Procedure, No-Picture Conditions. The experiment was conducted in Chinese. Subjects were told they would take part in two separate studies that were unrelated to one another. In the first study, subjects were induced to experience either positive or negative affect using a procedure similar to that employed by Schwarz and Clore (1983) and Adaval (2001). Specifically, they were told that the study concerned the construction of a database on the personal experiences of college students. With this preamble, subjects under positive affect induction conditions were told to identify a recent event that was very important to them and that made them feel happy when they thought abut it. In contrast, subjects under negative affect induction conditions were told to identify a recent event that made them feel unhappy whenever they thought about it. In each case, they were told to imagine the experience in as much detail as possible and try to re-experience the feelings they had at the time, and then to write down a description of these feelings as well as the events that elicited them. They were given between 15 and 20 minutes to write their descriptions. After completing the affect-induction task, subjects were introduced to the second study. This study was ostensibly concerned with how people made decisions in actual shopping situations in which people happen to see a product in a store window and then, after noticing it, enter the story in order to learn more about it. Subjects under hedonic-criterion conditions were then told to imagine that they wanted to buy a pair of sport shoes that would feel comfortable, and that it was particularly important to consider how the shoes would feel when wearing them. In contrast, subjects under utilitarian-criterion conditions were told to assume that they wanted to buy a pair of shoes whose of construction was of high quality, and that it was particularly important to consider how well the shoes were made. Subjects were given a list of the 6 attribute descriptions in one of two random orders. After reading the descriptions, they turned over the information sheet and completed a product evaluation questionnaire in which they estimated their liking for the product along a scale from 0 (not at all) to 10 (very much). Procedure, Picture Conditions. The procedure in these conditions was similar to that employed in no-picture conditions except for the introduction of a picture of the product and the point at which the picture was presented. In affect-picture order conditions, affect was induced at the outset, as in no-picture conditions. Then, after being told about the product evaluation task and the criterion they should use as a basis for judgment, subjects were shown an attractive picture of the sports shoes on an overhead projector, being instructed to imagine that they had seen the product in a store window. Pictures were shown on an overhead projector for about 5 sec. Then, subjects were given the list of attribute descriptions and evaluated the product. In addition, they estimated their first impression of the product they had formed at the time they saw the picture of it. This rating was made along a scale from 0 (not at all favorable) to 10 (very favorable). In picture-affect order conditions, subjects were introduced to the product-judgment task at the beginning of the experimental session, told the judgmental criterion they should use, and shown the picture of the product. Then, they were reminded that in many cases, consumers cannot check out the features of a product they want to consider immediately after they have seen it, and that to simulate these conditions, we would like them to perform an unrelated task before they received information about the product they were evaluating. On this pretence, they were administered the affect-induction task under

11 11 instructions similar to those described earlier. Then, after performing this task, they read the list of product attribute descriptions and made ratings similar to those in affect-first conditions. Manipulation Checks. To confirm the effectiveness of the affect-induction procedure, subjects were asked immediately after making their product evaluations to report their feelings at this moment (i.e., while they had been filling out the product evaluation questionnaire). In addition, they were asked at the end of the experiment to report how they had felt at the time they had described their personal experience. In each case, responses were made along four scales pertaining to how happy, good, unhappy and bad they were feeling. These scales ranged from 0 (not at all) to 10 (very). The average of each subject s responses to the second two items was subtracted from the average of his or her responses to the first two items and used as an overall index of the positive affect the subject was experiencing. To confirm our manipulation of the judgmental criteria, subjects were asked upon completion of the experiment to indicate the extent to which they had thought about each of several features of sports shoes while rating them. Two of these features (quality of construction and how long the shoes would last) exemplified utilitarian criteria, and two others (how they would feel while wearing the shoes and how much they would enjoy wearing them) exemplified hedonic criteria. Responses to each pair of items, which were reported along scale from 0 (not at all) to 10 (very much), were averaged to provide a single index of the extent to which subjects reported using each type of criterion. Results Manipulation Check. The effectiveness of the affect-induction procedure was confirmed. Subjects recalled feeling happier at the time they described their life experience if this experience was a happy one (M = 3.69) than if it was an unhappy one (M = -4.18), F (1, 102) = , p <.001. They also reported feeling happier immediately after making their product evaluations in the former condition than in the latter (0-.83 vs. 2.08, respectively), F(1,102) = 22.74, p <.001. In neither case did this difference depend significantly on other experimental manipulations (p >.10). Our manipulation of the judgment criteria was also successful. Subjects who were told to use a hedonic criterion for judgments reported using hedonic criteria to a greater extent than utilitarian ones (7.23 vs. 5.51, respectively). When subjects were told to use a utilitarian criterion, they reported using hedonic criteria relatively less than utilitarian ones (7.21 vs. 7.32). The relative use of the two criteria significantly differed under the two instructional conditions, F (1, 102) = 12.75, p <.01. Product Evaluations. Product evaluations are shown in the left half of table 1 as a function of induced affect, judgment criteria and picture conditions. An overall analysis of these data as a function of induced affect, picture conditions and judgment criterion revealed that judgments were generally more favorable when subjects were experiencing positive affect (M = 4.72) than when they were not (M = 3.91), F (1, 92) = 4.25, p <.05. However, although the three-way interaction of affect, picture conditions and judgment criterion was not reliable (p >.10), planned comparisons provided support for both of the hypotheses we considered Insert table 1 about here

12 12 According to hypothesis 1, the influence of extraneous affect on product evaluations by subjects who were not exposed to pictures should depend on the relevance of affect to the evaluation they are asked to make. That is, affect should influence product evaluations if these evaluations are based on hedonic criteria but not if they are based on utilitarian criteria. To evaluate this hypothesis, we considered only conditions in which no pictures were presented. As table 1 shows, induced affect had a positive impact on judgments when subjects were told to use a hedonic criterion (6.50 vs under positive vs. negative affect conditions, respectively), but had no effect at all on judgments when subjects were told to use a utilitarian criterion (4.00 vs. 4.13). The interaction implied by these results was significant, directional F (1, 92) = 3.46, p <.03, 1 and confirms both hypothesis 1 and results obtained earlier by Pham (1998). According to hypothesis 2, extraneous affect should only have an influence if it is induced before the picture is presented, and this should be true regardless of he criterion that subjects are told to apply. If subjects have already seen a picture of the product at the time extraneous affect is induced, their initial impression of the product should have already been formed, and so this affect should have little impact. To evaluate this hypothesis, we considered only conditions in which pictures were presented. Extraneous affect that was induced before a picture of the product was presented had a positive impact on evaluations of the product (5.25 vs. 3.77, when affect was positive vs. negative, respectively), F (1, 92) = 4.46, p <.05. When subjects had already seen a picture of the product at the time affect was induced, however, this affect had no influence at all on evaluations of the product (3.67 vs. 3.91, respectively), F < 1. This was true regardless of the judgment criterion that subjects were told to use. These conclusions are confirmed by a planned contrast of the influence of extraneous affect under the two order conditions. This contrast, which is equivalent to the interaction of induced affect and picture-affect order, was significant, directional F (1, 96) = 3.28, p <.05, and was not contingent on judgment criterion (F < 1). These data provide support for hypothesis 2. First Impressions. Hypothesis 2 assumes that the picture of a product elicits an initial appraisal of it, and that the affect associated with this appraisal, along with other affect they happen to be experiencing at the time, influences their initial impression. We did not obtain judgments of the product immediately after pictures were presented. 2 However, subjects recall of the impressions they had at the time they saw the pictures are consistent with this assumption. These data are shown in the right half of table 1. When affect was induced at the outset, happy subjects recalled their first impressions to be more favorable than unhappy subjects did (4.88 vs. 3.25, respectively). When pictures were presented first, however, subjects recalled having similar impressions regardless of whether they felt happy or not ( This test, which involves a comparison of the mean of half the cells of the design with the mean of the other half, is equivalent to a one-tailed t -test, where F = t 2 ; thus, the F-ratio reported here is equivalent to t(92) = 1.85, p <.03, one-tailed. For further discussion, see Keppel (1991, pp (Here and subsequently, comparisons are evaluated in relation to a pooled error term computed on the basis of the overall analysis.) 2 It seemed likely that asking subjects to make an initial judgment of the product immediately after seeing the pictures would induce a demand to be consistent and, therefore, could artifactually influence the judgments that were reported later.

13 13 vs. 3.54, respectively). The interaction of induced affect and picture-affect order approached significance, F (1, 68) = 3.64, p <.06. Thus, the interactive effects of picture-affect order and induced af fect on subjects initial impressions were very similar to the effects of these variables on the overall product evaluations that subjects reported after receiving specific attribute information. A further indication of the mediating effects of initial impressions on product judgments was obtained from a reanalysis of these judgments under the two picture conditions using initial impressions as a covariate. Although the covariate had a substantial impact on judgments, t(64) = 6.03, p <.01, the interactive effects of picture-affect order and was reduced to nonsignificance (F < 1). More generally, the proportion of variance accounted for by main and interactive effects of experimentally manipulated variables on judgments was reduced from 12.0% to 3.4% after eliminating their effects on initial impressions. These data are obviously not definitive. For one thing, subjects may not have actually remembered their initial impressions and, therefore, may have reconstructed these impressions on the basis of the final judgments they had made. Nevertheless, these data are consistent with the assumption that subjects final evaluations were based largely on the global appraisals they had formed at the outset, independently of the specific attribute information they received subsequently. EXPERIMENT 2 Experiment 2 provided a conceptual replication of the first experiment and explored further implications of the conceptualization underlying it. Two modifications or the earlier study were made. First, telling subjects explicitly the criterion they should use to evaluate a product is rather artificial. In purchasing situations, the criteria that consumers apply are typically activated spontaneously, depending on the product they are considering. To provide a closer simulation of conditions similar to those that consumers encounter outside the laboratory, we asked subjects in this experiment to consider products to which they were likely to apply different criteria spontaneously, without explicit instructions to do so. Second, we examined further implications of the proposed conceptualization. As noted earlier, subjects who have based their judgments on an initial impression of the object they are judging may selectively attend to information that confirms the implications of this impression (Chernev 2001; see also Snyder 1981). This may be particularly true when the impression is based on a picture. Pham et al. (2001) found evidence that the feelings induced by pictures elicited spontaneous thoughts that were evaluatively consistent with them. As Wyer and Srull (1989) suggest, concepts activated by thoughts about stimulus could influence both the type of information that people actively seek and the information they actually receive to which they attend. Both possibilities were examined in this experiment. Subjects before receiving attribute descriptions of the product they were considering were asked to indicate what additional attributes of the product they would like to know about. Then, after specific attribute information had been presented and subjects had evaluated the product, we asked them to recall the information they had received. To the extent subjects engage in confirmatory information processing, the effects of extraneous affect when pictures were presented might be reflected in both of these indices.

14 14 The effects of extraneous affect on information seeking when no pictures were presented were somewhat less clear. We assumed that subjects in this condition would base their judgments on the affect they were experiencing at the time of judgment without necessarily considering the attribute information they received. If this is so, the attribute information they seek and to which they attend might not be a function of the affect they are experiencing. On the other hand, suppose people s feelings elicit thoughts that are consistent with these feelings (Pham et al. 2000) and these thoughts activate evaluatively consistent concepts. People might then seek and attend to information that is consistent with these activated concepts (Wyer and Srull 1989). If this is so, the concepts could bias the sort of information to which people attend regardless of whether they actually use this information as a basis for judgment. The data collected in this experiment permitted this possibility to be evaluated. Method Subjects were 103 Hong Kong university students who were paid $80 HKD ($10 USD) for their services. They were assigned randomly to 12 cells of a three-factor design involving induced affect (positive vs. negative), product type (hedonic vs. utilitarian) and picture conditions (no picture vs. pictureaffect order vs. affect-picture order). Selection of Product Categories. A focus group discussion involving postgraduate students was held to select examples of (a) products that were typically evaluated on the basis of utilitarian criteria and (b) products that were typically judged on the basis of hedonic criteria. Based on these discussions, salad dressing and backpacks were selected for further consideration. Forty-one undergraduates who did not participate in the main study were then asked to indicate the extent to which they would think about each of four features while purchasing salad dressing: (a) the quality of the product, (b) the various uses of salad dressing, (c) how they would feel when tasting it, and (d) how much they would enjoy using it. Similarly, they indicated the extent to which they would think about four analogous features when considering the purchase of a backpack: (a) the quality of its construction, (b) the various uses of the backpack, (c) how they would feel when carrying it, and (d) how much they would enjoy using it. These estimates were made along a scale from 0 (not at all) to 10 (very). Subjects ratings of the first two (utilitarian) attributes of each product and their ratings of the second two (hedonic) attributes of the product were averaged to provide a single index of the importance of each alternative criterion. As expected, utilitarian criteria were considered more important than hedonic criteria for judging backpacks (8.27 vs. 7.67) but less important than hedonic criteria for judging salad dressing (7.50 vs. 8.45). Selection of Attributes. To select attributes of each product that differed in favorableness, eight attributes of each product were compiled and given to 48 undergraduate students with instructions to estimate both the favorableness of each attribute and the importance of knowing the attribute in order to make a purchasing decision. These ratings, which were made along a scale from 0 (not at all) to 10 (very). Based on these ratings, three favorable and three unfavorable attributes were selected for each product. Favorable attributes of salad dressing included 95% fat free and no preservatives, whereas unfavorable attributes included artificial coloring and only available in large containers. Favorable attributes of backpacks included machine washable and water resistant, and unfavorable attributes included heavy (800g) and becomes dirty easily. Favorable and unfavorable attributes were evaluated

15 15 similarly in each domain (7.70 vs. 2.57, respectively in the case of salad dressing, and 7.57 vs. 1.92, respectively, in the case of backpacks). Moreover, they were similar in importance (6.42 vs. 5.88, for favorable and unfavorable attributes of salad dressing, respectively; 6.47 vs for favorable and unfavorable attributes of backpacks, respectively). Question Selection List. Before receiving product information, subjects in the experiment were given a list of questions and asked to select the ones they would most prefer to ask in order to obtain information about the product. The procedure we used was similar to that employed by Swann and Guliano (1978), who found evidence that people tend to ask questions, confirming answers to which will confirm the validity of the hypothesis they are testing. Specifically, we constructed a list of 12 questions for each product. In each list, five questions were worded so that an affirmative answer would confirm the existence of a favorable attribute ( Is it tasty?, Is it low in fat?, etc., in the case of salad dressing; Does the zipper work smoothly?, Is it water resistant?, etc., in the case of backpacks). Five others were worded so that affirmative answers would confirm the presence of an unfavorable attribute ( Has artificial coloring been added?, Do I need to finish it quickly after opening?, etc., in the case of salad dressing; Would it get dirty easily?, Is it difficult to carry?, etc., in the case of backpacks). The remaining two questions (e.g., Is the product expensive or cheap?, Is it made of nylon or cloth? etc.) were unbiased. To confirm the relevance of the questions to a purchasing decision, a group of 24 undergraduates were asked to indicate the importance of knowing the answer to each of the questions in making a decision to buy the product to which it was relevant. These ratings, along a scale from 0 (not at all) to 10 (very), are also summarized in table 3. All questions were considered important, and their importance did not depend on whether they were positively or negatively biased (for backpacks, 6.23 vs. 6.89, respectively; for salad dressing, 628 vs. 6.40, respectively). Thus, both types of questions were equally diagnostic. Each set of 12 questions was arranged in a list in random order with the constraint that the mean serial position of positively-biased and negatively-biased questions was the same. These lists were presented to subjects under conditions to be described. Procedure. The procedures we employed in this experiment were similar to those in experiment 1 with two exceptions. First, both pictures of the products and attribute descriptions were presented on a computer. Second, subjects before receiving information about the product s specific attributes were given the opportunity to seek information about the product they were considering. The procedure we used resembled that employed by Snyder and Swann (1978). That is, the experimenter told subjects to imagine a situation in which they have encountered the product they were considering and would like to acquire more information about its specific features. Subjects were then given a list of the 12 questions described in table 3 and asked to write down the five questions they would most prefer to ask about the product. 3 Subjects wrote down the questions in the order of their preference. Subjects were then told they would actually be given information about 6 of the product s attributes. They were told that although this information might or might not answer the questions they had selected, it was nevertheless relevant to an evaluation of the product. With this preamble, subjects were 3 Before s electing questions from the list, subjects were given an opportunity to generate five questions of their own they would like to ask. Most of these questions were not codable as either positively biased or negatively biased, and so analyses of these data we re not performed.

16 16 shown the six pieces of attribute information, one at a time, on the computer screen. The order of presenting the attributes was counterbalanced within each experimental condition. Each subject pressed the space bar to receive the first piece of information and then, after reading it, pressed the bar again to receive the second piece, and so on until all six pieces had been read. 4 Judgments and Recall. Subjects after receiving the product information estimated how well they would like the product along a scale similar to that employed in experiment 1. After making this rating and completing the manipulation check questionnaire, they were told that in understanding how people make judgments of a product, it is sometimes useful to know what information about the product they can recall. On this pretence, they were asked to write down all of the attribute descriptions they could remember. They were told to write them down in the order they came to mind, regardless of whether they actually took the attributes into account in making judgments. Results Manipulation Check. Subjects who were asked to report their feelings immediately after evaluating the product described themselves as happier if they had written about a happy life experience (M = 2.58) than if they had written about an unhappy one (M = 0.49), F (1, 91) = 12.36, p <.01. Moreover, they recalled feeling happier at the time they described their experience in the former condition than in the latter (4.29 vs. 2.81, respectively), F(1, 91) = , p <.01. Product Evaluations. An overall analysis of judgment data yielded a main effect of affect, F(1, 91) = 5.83, p <.05, and an interaction of affect and product type, F(1, 91) = 3.80, p <.06. Although the three-way interaction of these variables and picture conditions was not reliable (p >.10), the pattern of data was quite consistent with hypotheses and with the results of experiment 1. Mean product evaluations are summarized in table 2. The two hypotheses were evaluated in a series of planned comparisons similar to those conducted in the first experiment Insert table 2 about here Under no-picture conditions, extraneous affect had a substantial effect on judgments of the hedonic product (7.11 vs. 4.67, when subjects were induced to feel happy vs. unhappy, respectively). If anything, however, it had a slight contrast effect on judgments of the utilitarian product (3.41 vs. 3.78, respectively). The predicted interaction of induced affect and product type implied by these differences was quite significant, directional F(1, 91) = 4.74, p <.02 (see footnote 1). When subjects who had been induced to feel happy or unhappy were exposed to a picture of the product, however, their feelings had a positive impact on their judgments (5.63 vs. 4.01, when induced affect was positive vs. negative, respectively), directional F (1, 91) = 5.82, p <.01, and this was true regardless of whether the criteria used to the evaluate the product were typically hedonic (6.38 vs. 4.38, respectively) or utilitarian (4.88 vs. 3.63, respectively). In contrast, when subjects saw a picture of the 4 The time required to read each piece of information was recorded. However, analyses of reading times yielded no results of interest, and so these data will be ignored in the discussion to follow.

17 17 product at the outset, extraneous affect had minimal effect on their judgments (4.95 vs. 4.39, respectively, averaged over the two product types). A planned contrast of the impact of inducing affect before pictures were presented and the impact of inducing it afterwards (equivalent to the interaction of induced affect and picture conditions) was significant, directional F(1,91) = 2.86, p <.05, and did not depend on the type of product being evaluated (F < 1). Hypothesis 2 was therefore supported. Information Seeking. The affect that subjects were experiencing at the time they saw a picture of the product was expected to influence the favorableness of the attributes about which they sought informat ion. In contrast, affect that was induced after subjects had seen a picture was expected to have little influence on information seeking. In the absence of pictures, the extent to which subjects made an initial appraisal of the product, and thus the influence of extraneous affect on information seeking, was less clear a priori. To evaluate these effects, we restricted our attention to the first three questions that subjects selected, which were likely to be particularly indicative of the type that subjects were disposed to ask spontaneously. These questions could pertain to either favorable attributes, unfavorable attributes, or neutral ones. The questions that referred to favorable and unfavorable attributes were recoded as either evaluatively congruent or evaluatively incongruent with the affect that subjects were experiencing. The proportion of questions selected of each type was then determined for each subject separately, and these proportions were analysed as a function of product type and picture conditions. Data pertaining to these analyses are shown in table 3. Subjects were generally more likely to select questions about attributes that were evaluatively congruent with the affect they were experiencing (M =.46) than questions about attributes that were incongruent with this affect (M =.34), F (1, 91) = 6.44, p <.01. However, this difference was more apparent when pictures were not presented (M diff =.24) than when they were (M diff =.05), F(1, 91) = 3.94, p <.05. Moreover, the bias in seeking affect-congruent versus affect-incongruent questions in no -picture conditions was greater when the product being judged was hedonic (M diff =.30) than when it was utilitarian (M diff =.17), F(1, 91) = 10.08, p <.01. When pictures were presented, ho wever, the bias was negligible both when affect was induced after pictures were presented (M diff =.08) and when it was induced beforehand (M diff =.03), and this was equally true regardless of the type of product being considered Insert table 3 about here Thus, contrary to expectations these data provide no evidence that subjects were likely to seek expectancy-confirming information when this expectancy was based on a picture of the product. In the absence of a picture, however, subjects did appear to seek information that was evaluatively consistent with the affect they were experiencing, and this tendency was greater when the product was typically evaluated on the basis of hedonic criteria rather than utilitarian criteria. Recall. Because only six attributes were described in the information we presented, the overall level of recall was quite high (M = 5.3) and did not differ appreciably over experimental conditions. To obtain an index of the relative accessibility of the information, we therefore restricted consideration to only the first three attributes that subjects listed. The mean proportion of these three attributes that were

18 18 favorable is summarized in table 4 as a function of experimental variables. (The proportion of unfavorable attributes recalled is of course the mirror image of the number of favorable ones recalled.) Insert table 4 about here The effects of experimental variables on recall are very similar to their effects on product evaluations (table 4). An overall analysis of the data yielded a significant interaction of induced affect, product type and presentation conditions, F (2, 91) = 4.87, p <.01. This interaction was broken down into a series of planned contrast similar to those in evaluating judgment data. When subjects did not see a picture of the product, the attributes they called of a hedonic product were more likely to be favorable when they felt happy (M =.85) than when they felt unhappy (M =.44). However, they recalled fewer attributes of a utilitarian product in the former condition than the latter (.38 vs..60, respectively). This conclusion is confirmed by an interaction of induced affect and product type under no -picture conditions alone, directional F (1, 91) = 27.90, p <.01. When pictures were presented, however, the influence of induced affect depended on when the affect was induced. Specifically, happy subjects were more likely to recall favorable attributes under these conditions (M =.55) than unhappy subjects were (M =.43), F(1, 91) = 3.77, p <.06. However, although this difference was pronounced when affect was induced before pictures were presented (.64 vs..44, respectively), F (1, 91) = 7.28, p <.01, it was negligible when pictures had been presented at the outset (.44 vs..42, respectively). Although a planned contrast of the impact of induced affect on the recall of favorable attributes in affect-picture order conditions with its effect under picture -affect order conditions was not reliable, F (1, 91) = 2.21, p >.10, the difference is consistent with implications of hypothesis 2. Correlational Analyses. The parallel influences of affect on judgments and recall in this study raise the question of whether the effect of affect on one variable mediates its influence on the other. For example, subjects affective reactions might influence their selective attention to attribute information (Forgas and Bower 1987), and this differential attention could bias product evaluations that are based on this information. According to the conceptualization we propose, however, affect influences product evaluations through its impact on subjects initial impressions, independently of the attribute information that subjects receive later. That is, its influence on the attention that subjects pay to attribute information (as reflected in the recall of this information) is only a result of their attempt to confirm the implications of the impression they had formed earlier. The results of correlational analyses are more consistent with the second of these possibilities. That is, the proportion of favorable attribute items that subjects recalled was virtually uncorrelated with their product evaluations under no-picture conditions (r =.06, p >.10), and was only marginally related to judgments under conditions in which pictures were presented (r =.22, p <.10). Moreover, the effects of affect on judgments and their contingency on other experimental variables were not appreciably affected when variance due to recall was eliminated. Therefore, these data suggest that the influence of affect on judgments was unlikely to have been mediated by its impact on the attribute information that subjects were able to recall. Rather, its influence on judgments and its influence on recall were independently

19 19 affected by its impact on the initial impressions that subjects formed of the product before the attribute information was presented. EXPERIMENT 3 Two assumptions underlie our interpretation of the results of experiments 1 and 2. First, we assumed that when subjects encounter an attractive picture of a product, their spontaneous appraisal of it elicits affective reactions, and the impression they form on the basis of these reactions influences their later judgments. Not all pictures elicit affect, however. Some, for example, might convey information about a product s utilitarian characteristics (e.g., the type of material or construction). Pictures of this sort might stimulate a spontaneous cognitive appraisal of the product, but this appraisal would not elicit affect. To this extent, extraneous affect would not become confused with appraisal-elicited affect and, therefore, should not have an influence on the initial impressions that subjects form. Consequently, it should have no impact on judgments that are based on this impression. A second possibility, however, is that pictures that do not convey a product s physical appearance do not stimulate a general appraisal of the product and, consequently, do not lead an initial impression to be formed of it. That is, subjects might treat the information in the picture simply as additional information that they consider along with the attributes that are described later. To this extent, extraneous affect should have an impact similar to its effect when no pictures are presented at all. A third experiment was conducted to evaluate these possibilities. A second assumption was also evaluated in this study. That is, subjects judgments of the product when pictures are presented were presumably based on the initial impressions they formed of it independently of the attribute information they received later. However, because equal numbers of favorable and unfavorable attributes were presented in this experiment, the net implications of which were neutral, this assumption could not be directly evaluated. In experiment 3, the favorableness of the attributes was manipulated. If subjects base their evaluations primarily on their initial impressions, the effect of this manipulation should be minimal. Method Design and Subjects. Ninety-three marketing students received $60 HKD (about $8 USD) for participating. They were randomly assigned to cells of a 2 (induced affect: positive vs. negative) x 3 (picture conditions: hedonic vs. utilitarian vs. none) x 2 (attribute information: predominantly favorable vs. predominantly unfavorable) design. Procedure. The procedure was identical to that conducted under affect-first, hedonic-criterion conditions of experiment 1 (i.e., conditions in which extraneous affect was most likely to have an impact). That is, subjects were first induced to feel happy or unhappy by recalling a personal life experience. Then, they were asked to make evaluations of a pair of running shoes, emphasizing that they should consider how the shoes would feel when wearing them. Under hedonic-picture conditions, they were shown a moderately attractive picture of the shoes exterior, as in experiment 1. In utilitarian-picture conditions, the picture showed a cross-section of a running shoe that conveyed its internal structural characteristics but did not indicate what the shoes actually looked like.

20 20 Six attribute descriptions were presented, five of which (two favorable, two unfavorable and one neutral) were identical in all cases. However, the remaining one was either favorable ( made of breathable material) or unfavorable ( made of non-breathable material ). The effectiveness of this manipulation was confirmed on the basis of data from an independent group of 15 subjects who were not exposed to either pictures or the affect-induction manipulation. These subjects evaluated the product significantly more favorably when the majority of the product attributes were favorable (M = 4.71) than when the majority were unfavorable (M = 2.87), F (1, 13) = 4.07, p <.06. Subjects after receiving the information estimated their liking for the product and then completed questions pertaining to the manipulation check on affect (see experiment 1). In addition, subjects who were shown a picture estimated the extent to which they liked the picture and also indicated how informative it was about the quality of the shoes. These latter judgments were made along a scale from 0 (not at all) to 10 (very much). Results Manipulation Checks. Subjects affective reactions were manipulated successfully, as in other studies. Subjects described themselves to be happier at the time they rated the product when they had written about a positive life experience than when they had written about an unpleasant experience (2.94 vs. 0.02), F (1, 91) = 13.60, p <.01, and recalled feeling happier at the time they wrote about their experience in the former case than in the latter (4.01 vs. 2.81), F(1, 91) = 78.62, p <.01. Characteristics of the pictures were also manipulated successfully. That is, subjects liked the hedonic picture more than the utilitarian one (4.33 vs. 2.70), F (1, 91) = 6.96, p <.01, but considered the utilitarian picture to be relatively more informative about the shoes quality (5.60 vs. 3.61), F (1, 91) = 14.71, p <.01. An overall analysis of liking for the product as a function of induced affect, picture conditions and attribute favorableness yielded a significant main effect of affect, F(1, 81) = 7.72, p <.01, and a significant interaction of affect and picture conditions, F (2, 81) = 3.05, p <.05. However, no effects involving attribute favorableness were reliable (p >.10). The implications of these results are discussed in the context of the issues to which they pertain. Effects of Pictures. We speculated that if subjects who saw a utilitarian picture formed an initial impression of the product on the basis of a nonaffect-based appraisal, and if they used this impression as a basis for their judgment, the extraneous affect they were experiencing would have little impact on these judgments. If, on the other hand, subjects did not form an initial impression on the basis of the picture, the effect of extraneous affect should be similar to that observed under no-picture conditions. Results are most consistent with the first possibility. That is, extraneous affect had a positive influence on judgments when no pictures were presented (4.36 vs. 3.38, when subjects were happy vs. unhappy, respectively), F (1, 81) = 5.59, p <.05, and this difference was even greater when they had seen a hedonic picture of the product (4.87 vs. 2.72, respectively), F(1, 81) = 10.99, p <.01. When subjects saw a utilitarian picture, however, they evaluated the product nonsignificantly less favorably when they were feeling happy than when they were not (3.48 vs. 3.66). This difference was significantly different from the effects of affect under the other two conditions combined (4.62 vs. 3.05), F (1, 89) = 5.24, p <.05, as indicated by the aforementioned interaction of affect and picture conditions in the overall analysis.