Effects of Attribute Salience on the Consistency of Product Evaluations and Purchase Predictions

Sharon Shavitt, University of Illinois
Russell H. Fazio, Indiana University
ABSTRACT - Previous research (e.g., Wilson & Dunn, 1986; Millar & Tesser, 1986) has demonstrated that the context for judging a product can make either affective or cognitive considerations more salient and therefore more important in evaluating that product, and in turn can affect the consistency between product judgments made in different contexts. The present research extends these findings to the specific attribute dimensions relevant to evaluating a product. Our studies examined the consistency between subjects' attitude expressions and their predictions of their purchase behaviors toward a product. The findings demonstrated that, when the attributes that were made salient at attitude expression and at purchase prediction corresponded, the consistency between those judgments was much greater than when the salient attributes did not match. The studies also identified contextual factors as well as product characteristics that heightened the salience of certain attributes when making product judgments. Finally, the results demonstrated that the effect of context on attribute salience was moderated by individuals' predispositions to be influenced by situational factors (i.e., their level of self-monitoring; Snyder, 1974). Implications of these findings for the nature of attitudes are discussed.
[ to cite ]:
Sharon Shavitt and Russell H. Fazio (1990) ,"Effects of Attribute Salience on the Consistency of Product Evaluations and Purchase Predictions", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 91-97.

Advances in Consumer Research Volume 17, 1990      Pages 91-97

EFFECTS OF ATTRIBUTE SALIENCE ON THE CONSISTENCY OF PRODUCT EVALUATIONS AND PURCHASE PREDICTIONS

Sharon Shavitt, University of Illinois

Russell H. Fazio, Indiana University

[This research was supported by Grant MH38832 and by Research Scientist Development Award MH00452 from the National Institute of Mental Health awarded to Russell Fazio. Sharon Shavitt was supported by NIMH National Research Service Award Postdoctoral Fellowship MH17146-04 at Indiana University during portions of this research.]

ABSTRACT -

Previous research (e.g., Wilson & Dunn, 1986; Millar & Tesser, 1986) has demonstrated that the context for judging a product can make either affective or cognitive considerations more salient and therefore more important in evaluating that product, and in turn can affect the consistency between product judgments made in different contexts. The present research extends these findings to the specific attribute dimensions relevant to evaluating a product. Our studies examined the consistency between subjects' attitude expressions and their predictions of their purchase behaviors toward a product. The findings demonstrated that, when the attributes that were made salient at attitude expression and at purchase prediction corresponded, the consistency between those judgments was much greater than when the salient attributes did not match. The studies also identified contextual factors as well as product characteristics that heightened the salience of certain attributes when making product judgments. Finally, the results demonstrated that the effect of context on attribute salience was moderated by individuals' predispositions to be influenced by situational factors (i.e., their level of self-monitoring; Snyder, 1974). Implications of these findings for the nature of attitudes are discussed.

INTRODUCTION

Recent research by Wilson and his colleagues (e.g., Wilson & Dunn, 1986; Wilson, Dunn, Bybee, Hyman & Rotondo, 1984; Wilson, Kraft & Dunn, 1989) and by Millar and Tesser (1986, 1989), presented in this session, has demonstrated that variations in the context for judging a product can make either affective or cognitive considerations more salient and therefore more important in evaluating the product. When the evaluations implied by these cognitive and affective aspects differ, one may report differing attitudes toward the product. This in turn can affect the consistency between the product judgments one makes in different contexts.

The present research extends these findings beyond the fundamental distinction between affective and cognitive considerations. It is our contention that these results have general implications for any attribute dimensions that underlie attitudes. The specific attributes relevant to evaluating a product (e.g., its taste, cost, sophistication) may also vary in salience as a function of the judgment context, and these variations may influence the favorability of one's attitudes. For example, one's self-reported attitude toward caviar may be somewhat unfavorable when attributes relevant to taste or cost are salient, but relatively favorable when the attribute of social sophistication is salient. - This suggests that the salience of attribute dimensions should influence the consistency between people's expressed attitudes and their subsequent judgments in a manner that parallels the previous findings. Specifically, if the attribute that is salient at the time of an initial attitude report is not the same as the one that is salient at a subsequent judgment, the correspondence between these evaluative judgments may be relatively low. However, when the salient attributes correspond, consistency should increase. The present studies tested this hypothesis. [It should be noted that mean differences in the attitudes reported in different judgment contexts were not expected in this research (and were not consistently obtained). One would only expect such differences to the extent that subjects consensually evaluated the target product more favorably in terms of one of the salient attributes than the other. If, for example, people generally agreed that the social image associated with caviar was more desirable than its taste, then the heightened salience of its social image versus its taste attribute should produce unidirectional changes in people's general evaluations of caviar. But people may differ in their relative evaluations of caviar along these dimensions, in which case mean differences in attitudes reported as a function of judgment context might not be obtained.]

These studies also investigated some of the factors that influence which attributes are salient when evaluating a product. In addition to the contextual factors explored in the studies by Wilson, et al. and by Millar and Tesser, other factors were also thought to play a role. For example, the target of the evaluation -- the attitude object or product -could certainly influence the attributes that are salient when evaluating it. For some items, a given attribute may spontaneously be salient when evaluating the item. When evaluating a candy bar, for example, its taste should normally be a salient attribute. Although it can certainly be evaluated along other dimensions, taste may be regarded as its most relevant attribute, and should spontaneously drive on-e's general evaluation of the candy bar. On the other hand, when evaluating a product that is perceived to create a particular social impression (e.g., trendy clothing, a gourmet food item), attributes related to that social impression may spontaneously be salient and drive one's general evaluation.

In the absence of any contextual influences, then, evaluations of a product may spontaneously be driven by a particular attribute or set of attributes. For different products, different attributes should be salient and drive one's evaluations. Thus, products themselves can be employed as manipulations of attribute salience. (See Shavitt, 1989, in press, for a similar point regarding the attitude functions typically associated with certain attitude objects.)

The present research employed both product and context manipulations to vary independently the attributes that were salient when two different evaluative judgments were reported. These studies examined the influence of attribute salience on the relation between the attitudes that subjects initially reported toward a product and their subsequently-predicted likelihood of using the product.

STUDY 1

In the first study, 73 paid subjects (recruited through newspaper ads) reported their attitudes and predicted their behaviors toward one of two beverages.

Manipulations The attribute that was salient at the time that subjects reported their attitudes was manipulated contextually, via an initial questionnaire intended to heighten the salience of one of two attribute dimensions, taste or social impressiveness. On the taste questionnaire, subjects rated 20 different food items (e.g., marshmallows, lima beans, chocolate milk-shake) in terms of "how good they taste to you." Each item was rated on a 5-point scale, with 5 being "very good tasting" and 1 being "not at all good tasting." On the social impressiveness questionnaire, subjects rated 20 different actions (e.g., winning a tennis game, quitting your job, driving a BMW car) for "the extent to which your doing each of these things would make a good impression on others." These ratings were also made on a 5-point scale, with 5 being "will make a very good impression" and 1 being "won't make a good impression at all."

The attribute that was salient when subjects later predicted their behaviors was manipulated by varying the target product. The two beverage products used in this study, Perrier mineral water and 7-Up, were expected to differ in terms of the attributes that would spontaneously be salient when evaluating them (after the impact of the questionnaire manipulation had dissipated). 7-Up was expected to be evaluated typically on the basis of its taste. In contrast, it was thought that Perrier, because of its association with trend-conscious, upwardly-mobile consumers, was likely to be evaluated in terms of its social impressiveness.

It was predicted that the correlation between subjects' reported attitudes and their behavior predictions would be high when there was a match between the attribute whose salience was manipulated at attitude assessment and the attribute expected spontaneously to be salient for the target beverage (i.e., Taste/7-Up or Social Impressiveness/Perrier). On the other hand, when these attributes did not correspond (Social Impressiveness/7-Up or Taste/Perrier), the correlation was expected to be low.

Procedure: Immediately after completing one of the two questionnaires (taste ratings or social impressiveness ratings), subjects reported their general attitude toward one of the two beverages on a 9-point scale (very favorable/very unfavorable). After a delay of several minutes, during which subjects filled out some other measures, they predicted their likelihood of buying or drinking the target beverage in four different situations, such as the following:

"You are having lunch at a sandwich shop that is offering a free beverage with the purchase of a sandwich. How likely would you be to order a Perrier [7-Up] as your free beverage?"

Subjects made their predictions on a 5-point scale, with 5 being "I would definitely choose it" and 1 being "I would definitely not choose it." A similar 5-point scale was used to predict one's behaviors in three other scenarios, which involved taking the beverage to the beach in a cooler, redeeming a coupon, and buying it at the supermarket.

Finally, subjects responded to checks on the types of attributes that were assumed spontaneously to underlie their evaluations of the target beverage.

Results Subjects' responses on the attribute checks supported our assumptions about the attributes that typically underlie attitudes toward these beverages. Furthermore, as expected, when the attribute that was salient at attitude assessment (the attribute in the questionnaire) corresponded with the attribute that was salient at purchase prediction (the attribute associated with the target beverage), the correlation between subjects' attitude reports and their purchase predictions was substantially larger than when those attributes did not match (see Table 1). The interaction of questionnaire type and target beverage, representing this effect, was significant at pc.05 (see Steiner & Darroch, 1969, for the test of an interaction effect on correlation coefficients).

Conclusions These findings suggest that one's evaluative judgments of a product can vary -even within a short time frame -- as a function of what attribute is salient when the judgment is made. The results provided evidence that the salience of product attributes influenced the consistency between one's product judgments. Paralleling the studies by Wilson, et al. (1984, 1986, 1989) and by Millar and Tesser (1986, 1989) on the salience of affective versus cognitive factors, a match in the salience of specific attributes was associated with a greater correlation between attitude reports and behavioral judgments, as compared to conditions in which the salient attributes did not correspond.

TABLE 1

STUDY 1: CORRELATIONS BETWEEN ATTITUDE AND MEAN PURCHASE PREDICTION

The results also suggested that both contextual factors and product factors can influence the salience of product attributes. For some products, a given attribute may spontaneously be salient when a product evaluation is made. Which attribute(s) tend to be salient, and to drive one's judgments, can differ from product to product. A number of measures suggested that, as expected, such differences existed for the two beverage products employed in this study. The results also indicated that contextual factors could heighten the salience of particular attributes and temporarily increase their impact on reported attitudes toward these products.

However, one might expect the impact of such contextual factors to be moderated by other important factors. For example, certain personality characteristics, such as a high level of self-monitoring (Snyder, 1974), may predispose individuals to be affected by the contextual salience of dimensions. High self-monitors, who strive to fit into different social situations, are sensitive to situational cues, and their attitudinal and behavioral expressions are typically responsive to such cues (see Snyder and DeBono, 1989). Also, Kardes, Sanbonmatsu, Voss, and Fazio (1986) have shown that the attitudes of-high self-monitors are less accessible from memory than those of low self-monitors. All of this suggests that high self-monitors may lack strong, pre-existing attitudes to guide their responses. As a result, their responses may be more influenced than those of low self-monitors by whatever attributes are salient in a situation. Thus, the correspondence of those salient attributes may have a bigger impact on the relation between their product judgments.

Furthermore, high self-monitors may tend to be more concerned with certain attributes than with others. Recent research by Snyder and DeBono (1985, 1989) suggests that the product attitudes of high self-monitors reflect a greater concern with social image than the product attitudes of low self-monitors. This suggests that a manipulation that increases the salience of a product's social impressiveness may have stronger effects for high than for low self-monitors.

In the next study, we sought to replicate the present findings on the role of attribute salience in the consistency of product judgments, and to explore the potential moderating effects of self-monitoring.

STUDY 2

In this study, 123 introductory psychology students reported their attitudes and made purchase predictions regarding one of two powdered drink mixes .

Manipulations The attribute that was salient at attitude assessment was manipulated via the same initial questionnaires used in the previous study (taste vs. social impressiveness ratings). There was also a control (no attribute rating) condition, in which subjects completed a filler questionnaire.

The attribute that was salient when subjects predicted their behaviors was varied by having subjects respond to one of two new target products, Wyler's Lemonade Mix or Kool-Aid Drink Mix. The attribute of taste was expected spontaneously to drive evaluations of Wyler's Lemonade Mix. In contrast, it was thought that Kool-Aid, because of its strong image as a product for young children, was more likely to be evaluated in terms of the social impressions elicited by drinking it.

Procedure: Immediately after completing one of the three questionnaires (taste vs. social impressiveness vs. control), subjects reported their general attitude toward one of the two drink mixes. After a delay of several minutes, in which subjects filled out some other measures, they predicted their likelihood of buying or drinking the target drink mix on a measure similar to the one in the first study. Finally, they responded to checks on the attributes assumed to underlie their drink mix evaluations, and completed Snyder's (1974) Self-Monitoring Scale.

Results Overall, the correlations between expressed attitude and the average predicted likelihood of buying or drinking the target product replicated previous findings (see Table 2). When the attribute that was salient at attitude assessment (the attribute in the questionnaire) corresponded with the attribute that was salient at purchase prediction (the attribute associated with the target drink mix), the correlation between subjects' attitude reports and their purchase predictions was greater than when those attributes did not match. As in the previous study, the 2 X 2 interaction of questionnaire type and target drink mix (representing this effect) was statistically significant (p<.05).

How did subjects' level of self-monitoring influence this observed relation between attitude reports and purchase predictions? Table 3 shows the correlations for low and high self-monitors, based on a median-split of self-monitoring scores. As expected, it indicates that the pattern of correlations that was observed in Table 2 emerged strongly for high self-monitors (and yielded a significant 2 X 2 interaction). But, it did not emerge for low self-monitors.

Apparently, high self-monitors, who may be predisposed to be influenced by situational circumstances, were more affected by the context manipulation (questionnaire) designed to heighten the salience of product attributes. This was particularly true for the social impressiveness questionnaire, which involved an attribute that is especially relevant to the concerns of the high self-monitor. Subjects' responses to the attribute check measures indicated that this questionnaire (rating things for the impression they make on others) was far more effective in heightening the salience of the social impressiveness attribute for high than for low self-monitors. (However, there was no evidence that high and low self-monitors differed in the extent to which either taste or social impressiveness attributes were associated with the target products.)

Conclusions: The present study replicated previous results in indicating that the product attributes that are salient at attitudinal and at behavioral expression can have a strong impact on the consistency between those judgments. When the salient attributes corresponded, the consistency between attitude reports and purchase predictions was greater than when the salient attributes did not match.

The results also provided evidence that the salience of attributes can be affected not only by contextual factors and by product factors but by individual differences, as well. Individual differences in self-monitoring apparently moderated the effects of context on attribute salience and, in turn, moderated the pattern of relations between attitudes and purchase predictions. The pattern that had been observed as a function of attribute salience emerged for high but not for low self-monitoring subjects.

It is possible that low self-monitors were less affected by the heightened salience of product attributes because they already possessed highly accessible, pre-existing attitudes that guided their responses to these products (see Kardes, et al., 1986). High self-monitors may have lacked such attitudes. Their attitudinal and behavioral judgments were perhaps more likely to be constructed in response to whatever product attributes were salient at the time.

GENERAL DISCUSSION

Attitude researchers have generally viewed attitude reports as reflecting an individual's single, global evaluation of an object. In contrast, the present findings suggest that attitude responses are not necessarily unitary and can often be somewhat unstable. These and the other studies presented in this session indicate that reports of one's attitude toward a product can vary, even within a short amount of time, as a function of whatever happens to be salient in the judgment context. Research by Wilson, et al. (1984, 1986, 1989) and by Millar and Tesser (1986, 1989) has suggested that the salience of affective versus cognitive considerations can affect the favorability of one's reported attitudes. The present studies suggest that the salience of specific attributes can do the same. This was evidenced by a substantially lower correlation between evaluative judgments that were made when different attributes were salient, compared to conditions in which the salient attributes corresponded.

Thus, attitudes may perhaps be more heuristically conceptualized as clusters of different evaluations of an object (see Zanna & Rempel, 1988). Each evaluation may be specific to a representational domain (affect vs. cognition vs. behavior), a particular attribute, a goal, a setting, etc. Differences in the judgment context (e.g., differences in the goals that are salient) may lead one to access different evaluations, or to compute new ones when existing evaluations do not seem to apply. This conceptualization of attitudes is a particularly useful approach for understanding the stability of attitudinal responses over time and across situations. It also has implications for predicting the relation between attitude reports and behavior: For example, to the extent that the considerations (e.g., attributes, goals, etc.) that are salient at attitude assessment and those that are salient in the behavioral context are likely to differ, and to the extent that those different considerations imply different evaluations of the object, attitude statements should not be good predictors of behavior. However, to the extent that the evaluations implied by different salient considerations are consistent, attitude statements should be more predictive of behavior (see Millar & Tesser, 1989). Similarly, if the same considerations are likely to predominate and to drive one's judgments of an object regardless of context, attitude reports should provide relatively good predictions of behavior.

TABLE 2

STUDY 2: CORRELATIONS BETWEEN ATTITUDE AND MEAN PURCHASE PREDICTION

This view of attitudes also fits well with the nature of consumer attitudes, which are commonly formed and expressed in differing contexts and in response to different stimuli. For example, one may form some evaluations of a product based on several attributes made salient by advertisements, word-of-mouth information, past experiences, etc. Which of these evaluations is subsequently expressed (e.g., through purchase) may depend in part on which of the product attributes is salient in the decision context (e.g., in the store environment).

How these multiple evaluations are stored and structured in memory remains to be determined. In that regard, the results of the present studies are open to a number of interpretations. The present findings suggested that variations in the attribute rating task that subjects performed prior to attitude assessment (taste vs. social impressiveness ratings) influenced the favorability of the general attitudes they reported toward the target product. Was this because the heightened salience of particular attributes led subjects to compute new, attribute-specific evaluations of the target product on the spot? Perhaps subjects' pre-existing evaluations of the target product were not readily accessible in memory. Perhaps the salience of a certain attribute led subjects to think of the target product in a new way, and to re-evaluate it (at least temporarily) with that attribute in mind. Or perhaps the attribute rating task created a perceived demand to generate and report an attribute-specific rating of the target product.

TABLE 3

STUDY 2: CORRELATIONS BETWEEN ATTITUDE AND MEAN PURCHASE PREDICTION FOR HIGH AND LOW SELF-MONITORING INDIVIDUALS

These possibilities have relevance to the attitude-nonattitude distinction proposed by Converse (1970) and Hovland (1959), who observed that individuals sometimes provide attitude responses to an item on a survey (nonattitudes), even though they have no pre-existing evaluation of the attitude object. Converse (1970) noted that such nonattitudes were characterized by unreliability in their measurement. Fazio (1989) suggested that the attitude-nonattitude distinction can be viewed as a continuum, with nonattitudes at the low end, evaluations that are relatively inaccessible in memory closer to the middle, and well-learned, highly accessible evaluations at the high end. In the context of the present research, this distinction suggests that it may be toward the lower end of the attitude-nonattitude continuum where the malleability of attitude reports observed here as a function of judgment context is most likely to emerge. This is consistent with our finding that high self-monitors, [ who are less likely than low self-monitors to possess highly accessible attitudes (Kardes, et al., 1986), evidenced the predicted effects of judgment context, while low self-monitors did not.

Alternatively, it is possible that the present results were obtained because multiple evaluations of the target product already existed in memory. These evaluations (or a subset of them) might have tended toward the high end of the attitude-nonattitude continuum, and been highly accessible in memory. When subjects were asked to express their attitude toward the product after performing the rating task, they may have simply retrieved and reported the evaluation that was the most accessible in that context.

Each of these possibilities can account for the present data. Determining when each of these processes is most likely to occur, however, requires a better understanding of how such multiple evaluations may be organized in memory. What factors govern the accessibility of these evaluations? What are the links between them? How will their evaluative consistency affect the way they are organized and retrieved? When faced with new considerations regarding an object, when will a new evaluation be computed and when will an existing one be deemed adequate? What factors determine whether a newly-formed evaluation will be stored separately from the others or will be used to update an existing evaluation?

Further research is needed to investigate the implications of the present findings for the structure of attitudes. Understanding how multiple evaluations of a product are stored and organized in memory, and the conditions under which they are accessed or recomputed, is an important issue for future study.

REFERENCES

Converse, P.E. (1970). Attitudes and non-attitudes: Continuation of a dialogue. In E.R. Tufte (Ed.), The quantitative analysis of social problems Reading, MA: Addison-Wesley. pp. 168- 189.

Fazio, R.H. (1989). On the power and functionality of attitudes: The role of attitude accessibility. In A.R. Pratkanis, S.J. Breckler, and A.G. Greenwald (Eds.), Attitude structure and function Hillsdale, N.J.: Erlbaum. pp. 153-179.

Hovland, C.I. (1959). Reconciling conflicting results derived from experimental and survey studies of attitude change. American Psychologist, 14, 8-17.

Kardes, F.R., Sanbonmatsu, D.M., Voss, R.T., and Fazio, R.H. (1986). Self-monitoring and attitude accessibility. Personality and Social Psychology Bulletin, 12, 468-474.

Millar, M.G. dc Tesser, A. (1986). Effects of affective and cognitive focus on the attitude-behavior relation. Journal of Personality and Social Psychology, 51, 270-276.

Millar, M.G. & Tesser, A. (1989). The effects of affective-cognitive consistency and thought on the attitude-behavior relation. Journal of Experimental Social Psychology, 25, 189-202.

Shavitt, S. (1989). Operationalizing functional theories of attitude. In A.R. Pratkanis, S.J. Breckler, and A.G. Greenwald (Eds.), Attitude structure and function Hillsdale, N.J.: Erlbaum. pp. 311 -337.

Shavitt, S. (in press). The role of attitude objects in attitude functions. Journal of Experimental Social Psychology. Snyder, M. (1974). Self-monitoring of expressive behavior. Journal of Personality and Social Psychology, 30, 526-537.

Snyder, M. and DeBono, K. (1985). Appeals to image and claims about quality: Understanding the psychology of advertising. Journal of Personality and Social Psychology, 49, 586-597.

Snyder, M. and DeBono, K. (1989). Understanding the functions of attitudes: Lessons from personality and social behavior. In A.R. Pratkanis, S.J. Breckler, and A.G. Greenwald (Eds.), Attitude structure and function Hillsdale, N.J.: Erlbaum. pp. 339-359.

Steiner, I.D. and Darroch, R.K. (1969). Relationship between the quality of counter-attitudinal performance and attitude change. Journal of Personality and Social Psychology, 11, 312-320.

Wilson, T.D. and Dunn, D.S. (1986). Effects of introspection on attitude-behavior consistency: Analyzing reasons versus focusing on feelings. Journal of Experimental Social Psychology, 22, 249-263.

Wilson, T.D., Dunn, D.S., Bybee, J.A., Hyman, D.B., & Rotondo, J.A. (1984). Effects of analyzing reasons on attitude-behavior consistency. Journal of Personality and Social Psychology, 47, 5-16.

Wilson, T.D., Kraft, D., & Dunn, D.S. (1989). The disruptive effects of explaining attitudes: The moderating effect of knowledge about the attitude object. Journal of Experimental Social Psychology, 25, 379-400.

Zanna, M.P., & Rempel, J.K. (1988). Attitudes: A new look at an old concept. In D. Bar-Tal & A.W. Kruglanski (Eds.), The social psychology of knowledge New York: Cambridge University Press . pp. 315-334.

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