Assessing the Effects of Visible Consumption on Impression Formation

Russell W. Belk, University of Illinois, Urbana-Champaign
ABSTRACT - This paper reports a series of exploratory experiments measuring the impact of visible consumption of products and services on the impressions which others form about the consumers of these items. A method labeled "the detective study" is illustrated for measuring these impressions, and comparisons of effects are made between different types of consumption items and between male and female perceivers.
[ to cite ]:
Russell W. Belk (1978) ,"Assessing the Effects of Visible Consumption on Impression Formation", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 39-47.

Advances in Consumer Research Volume 5, 1978      Pages 39-47


Russell W. Belk, University of Illinois, Urbana-Champaign


This paper reports a series of exploratory experiments measuring the impact of visible consumption of products and services on the impressions which others form about the consumers of these items. A method labeled "the detective study" is illustrated for measuring these impressions, and comparisons of effects are made between different types of consumption items and between male and female perceivers.

The belief that a person's possession and expenditures reveal something about the person may be one of the strongest cultural universals affecting consumer behavior. In virtually all cultures, visible products and services are the bases for inferences about the status, personality, and disposition of the owner or consumer of these goods. Relevant cues may be noted not only in the number and type of goods consumed, but also in such features are their style, color, uniqueness, condition and brand name. Where other information about a person is known, this information is integrated with visible consumption information in developing an overall impression of another. There are also many instances in which visible consumption cues dominate the overall impression, either because the cues are highly distinctive or because little additional personal information is available due to observing strangers or newly established or casual acquaintances. Nevertheless, while consumer behavior research has examined the messages about self or ideal self concept which consumers seemingly intend to display through their selections of products and services, [E.g., Grubb and Grathwohl (1967); Birdwell (1968); Grubb and Hupp (1968); Dolich (1969); Mason and Mayer (1970); Hamm and Cundiff (1969); Ross (1971); Grubb and Stern (1971); Greeno, Sommers, and Kernan (1973); and Landon (1974).] we have devoted little attention to the messages which consumption selections actually do convey to others. The broad intent of this research, therefore, was to initiate a program examining the influences which configurations of visible consumption items have on impression formation.


There has been little systematic research into the effects of visible consumption on impression formation, but there have been occasional studies examining isolated consumption factors affecting person perception. Perhaps the classic study was Haire's investigation of the image of NescafT Instant Coffee, using subject descriptions of the presumed owner of a grocery shopping list including either instant or regular coffee (Haire, 1950). As indicated by subsequent replications, consumer images based on even these small differences in consumption patterns are significant and reliable as long as the product images remain constant (Westfall, Boyd, and Campbell, 1957), but may change as the images of the products change (Webster and Von Pechmann, 1970). A somewhat more visible item of consumption which has received some attention in studies of person perception is clothing. Several researchers have found through unobtrusive experimental designs that those in higher status clothing (e.g., a suit and tie rather than casual clothes) seem to gain more compliance with their requests (e.g., signing a petition [Suedfeld, Bochner, and Metas (1971); Darley and Cooper (1972).]) and also serve as models whose behavior is more imitated (e.g., crossing a street against a traffic light [Lefkowitz, Blake, and Mouton (1955).]). Holman (1976) has conducted an experimental study using photographs of a woman dressed in several different clothing ensembles, and found a number of differences in the inferred traits of the woman. While somewhat less conspicuous than clothing, cosmetics have also been found to affect person perception. In a study of the 1950's, McKeachie (1952) found that women wearing lipstick were judged by a sample of college men to be more frivolous, introspective, anxious, conscientious, and interested in the opposite sex than women not wearing lipstick. And Calder and Burnkrant (1977) recently found that a woman described as buying Revlon mascara was judged by college females to be more popular and competent than a woman described as buying Walgreens' mascara. Automobile ownership is another highly visible consumption characteristic which has been found to be related to person perception and related behaviors. Doob and Gross (1968) found that a more expensive and newer automobile which faked being stalled at a stoplight, received fewer horn-honking responses than a less expensive and older automobile. And Grubb and Hupp (1968) found that owners of Volkswagens and Pontiac GTOs were able to draw distinct and reasonably accurate profiles of the other group of owners' self concepts, knowing only the automobile which they owned.

In each of these studies, however, the investigated consumption items were examined in isolation from other consumption characteristics which might typically be apparent. It seems reasonable that people normally make judgments of others based on such broader configurations of visible consumption items. For instance, it is possible that a person who is wearing a suit and a tie with white stockings might be judged to be attempting to enact a role with which he is uncomfortable. In order to allow for such inferences, the present study sought a method which could systematically manipulate a composite of visible consumption items.


The method chosen for the present research was to present each subject with a consumption profile representing a single cell from a factorial design. The experiment was presented in the guise of a "detective study" being done to help the New York City Police Department to more effectively locate the owners of unidentified property which was lost or stolen and later recovered by or turned in to the police. There are several advantages to this method. One desirable feature of the cover story is that it allows the presentation of a large number of consumption items without appearing forced. For instance, by including a wallet or purse in a particular group of items, their contents can also be manipulated. This allows the use of business cards, membership cards, credit cards, and specialty merchandise such as pocket calendars and matchbooks, all of which can manipulate the stimulus person's use of various moderately visible services. In addition, the use of this cover story removes any elements of social desirability which might otherwise inhibit the expression of consumption-related stereotypes. Subjects were told that the groups of items they would be shown in the descriptive form of "property tags" were ultimately returned to their owners, and that the characteristics of these owners were known. Thus, the stated purpose of the study was to compare perceptions of owners to the actual characteristics of the owners in order to see how much accuracy was possible from the items alone. This rationale also allowed the presentation of more than one group of unrelated items, amounting to participation in more than one experiment by the same subject.

Treatment Variables

All subjects were part of five different 25 factorial designs. There were ten subjects per cell for a total of 320 subjects participating in these experiments. Half of the subjects in each cell were males and all were undergraduate students at the University of Illinois. The five experiments differed in the visibility of consumption items included, in the sex of the stimulus person, and in whether or not the locational context where the items were reportedly found was manipulated. The two experiments which will be reported here both involved moderately visible consumption items for which locational context was not varied, but one involved a female stimulus person and the other involved a male stimulus person. The treatment levels for these two experiments are shown in Tables 1 and 2.

All treatment levels which a particular subject was to receive were simultaneously presented on 3 inch by 5 inch cards with the first two treatments presented on the same card and the order of the cards randomized. While it is not possible to claim that the two experiments were identical except for the sex of the stimulus person, both had a wallet or purse, a sports event, a restaurant, a personal care item, and either a travel mode or a travel abode. This rough equivalence allows comparisons between males judging either male or female stimulus persons, and females judging either male or female stimulus persons.

It was hypothesized that perceivers of the same sex as the stimulus person would be able to make more complete and consistent judgments about the person based on these visible consumption items (i.e., lower error variance) than could those of the opposite sex. The underlying assumption is that a person of the same sex will be able to bring more self-knowledge about use of the consumption items to bear on inferences about the stimulus person. For two of the consumption items (wallet/ purse and hair-care/lipstick) there may also be a greater familiarity with the nature of the consumption items themselves by a same sex observer. There has been a large amount of research on the effects of observer sex on person perception in studies not involving consumption stimuli [E.g., Taft (1955); Levy and Schlosberg (1960); Warr and Knapper (1968); and Tagiuri (1969).] but the findings are inconsistent. The choice of the college subject pool, if it had an effect, probably biased results away from the hypotheses, since at this prime age for dating, the students may be particularly alert to cues indicating the character of members of the opposite sex.

Descriptor Variables and Covariates

Based on pretests involving open-end responses describing the owners of many of the stimulus item configurations, the demographic and adjectival scales shown in Table 3 were selected as dependent variables ("descriptors''). These same variables, with the exception of the last two items, were used to obtain a self-description of the subject following participation in the five experiments. Because of the college subject pool, residence, family income, occupational class, and years of education were measured referring to subjects' parents.

The subject descriptions will serve as covariates in future analyses, but at present it is of interest to compare the intercorrelations of these covariates, the intercorrelations of the descriptor variables, and the cross correlations between covariates and descriptor variables, in order to consider whether several types of response styles may have affected judgments. This information is summarized in Figure 1 for the moderately visible male consumption items experiment using data from the 160 male subjects. [The absolute magnitudes and patterns of correlations differed somewhat by subject sex and stimulus person sex, but the relative strengths of the covariate intercorrelations versus the "projection diagonal" correlations versus the descriptor intercorrelations retained the same ordering over the four combinations of subject and stimulus person genders.] An examination of the three matrices in Figure 1 allows several types of potential insights into subject responses. If the configurations of the covariate intercorrelations and the descriptor intercorrelations were equivalent, this could be taken as evidence of inherent clusters of traits, and factor or cluster analysis might be used to reduce the dependent measures prior to further analysis. Although there is some similarity between these two matrices, the differences in the number of significant correlations forces rejection of this conclusion and course of action. If on the other hand, the covariate intercorrelations were stronger than the descriptor intercorrelations, then one possibility would be that "halo effects" exist in self ratings and stimulus persons are viewed more discriminatingly. Another possible interpretation of such a pattern would be that while certain traits naturally occur together in the subjects (e.g., educated/successful), the consumption items presented are unnatural combinations which cause normal trait associations to fall apart. Since this pattern did not emerge, neither of these interpretations need concern us. A third possible pattern, and the one which is found in the data in Figure 1, is that the covariate intercorrelations will be weaker than the descriptor intercorrelations. The explanation in this case is clearer. The tendency to judge others using bundles of related attribute judgments has been referred to as "cojudgment" in person perception studies not involving consumption items (Warr and Sims, 1965). This phenomenon appears to be due to a reliance on implicit theories of personality types when judging others. That is, when judging others we tend to rely upon our knowledge or impressions of people in general to go beyond surface traits in deciding what characteristics people are likely to possess. This tendency leads to the use of groups of normally related personality characteristics. An overlapping but less plausible explanation for this pattern is that consumption characteristics reveal patterns of personality characteristic relationships which would not normally be found together. For instance, possession of a Rolls Royce automobile might lead to judgments that the owner is of high income and also ostentatious, even though the two characteristics are not normally correlated. However, if we assume that an observer tries to not only interpret the characteristics of a stimulus person, but also to form a familiar interpretation which will allow further predictions of that person's demeanor and behavior, then the cojudgment interpretation of descriptor intercorrelations is the most plausible.









One further possible basis for judgments of stimulus persons is examined in the "projection diagonal" of Figure 1. These are the correlations between the corresponding covariate and descriptor items. High positive correlations here would indicate that the subjects tend to project their own perceived traits onto the stimulus person. As the weak correlations indicate, this projection does not appear to have occurred. Instead it seems that while there may be some basis for expecting inherent trait clusters, cojudgment occurs in judging others to a greater extent than it does in self-descriptions. From the point of view of data analysis, this indicates that data reduction in the set of dependent measures should not be expected to yield a set of underlying variates which represent general person traits common to stimulus persons as well as subjects of both sexes. Avoiding data reduction also eliminates the potential problem of averaging the differences in relationships which may occur across cells of the experimental design.


Because of the strong likelihood that multivariate analyses of variance would form different linear composites of the dependent variables across the two experiments and two subject groups of interest, MANOVA procedures were rejected for the present comparisons. While it would be possible to include subject sex as a sixth independent variable in univariate analysis of variance, the complicated potential interactions between sex and the treatment variables would prove difficult to interpret. Therefore, the method of analysis chosen was to run separate ANOVAs for each dependent variable in each of the four combinations: (1) male subjects, male stimulus person experiment; (2) male subjects, female stimulus person experiment; (3) female subjects, male stimulus person experiment; and (4) female subjects, female stimulus person experiment. A sample ANOVA for the dependent variable "unattractiveness" with the data from combination 1, is presented in Table 4.

It may be seen from Table 4 that three main effects and three interactions are significant. The way in which these variables affect attractiveness judgments will be described shortly. For the moment, it is more important to note that these effects were able to account for 62 percent of the variance in attractiveness ratings, based on variance components estimates. [See Dwyer (1974).] This is the measure which is to be used for testing the hypothesis that subjects will be able to judge those of the same sex more completely and consistently based on consumption items. Table 5 shows a summary of proportions of variance accounted for in each dependent variable for the four subject/stimulus person combinations.

Looking only at the average proportions of variance accounted for over all dependent measures, it appears that females are clearly better judges of both male and female consumers than are males. Female judgments were more consistent across judges, leading to higher levels of explained variance. Male and female stimulus persons, on the other hand, appear to be equally easy to judge. This is true whether the perceiver is male or female. Examining the proportions of variance accounted for in each separate independent measure, it appears that females are consistently the best judges of income, seriousness, weight, friendliness, and interestingness, while males more readily assess responsibility. Judges of the same sex as the stimulus person seem to do somewhat better at estimating age, life cycle stage, residence, occupation, and attractiveness, while judges of the opposite sex as the stimulus person were able to more consistently estimate generosity, happiness, likeableness, emotionality, successfulness, aggressiveness, and amount of money missing, using consumption characteristics. Generally, however, it must be concluded that these patterns of differences fail to conform to an easily identifiable sensitivity stereotype such as one sex being more able to agree on affective judgments while the other sex focuses on more objective or cognitive characteristics.





Although the hypothesized ability of perceivers to judge those of the same sex more completely based on their consumption patterns was not supported in total variance explained, it is still possible that such effects are reflected in male and female differences in their interpretations of a particular consumption characteristic. In order to assess this possibility, Tables 6 and 7 provide a summary of the direction of main effects found to be significant (~ < .05) in the two experiments, broken down by sex of the perceiver. Although space precludes presenting interactions, a substantial portion of the variance in these two experiments was accounted for by these main effects. In Table 6 and to an even greater degree in Table 7 it may be seen that the implications of these consumption item clues differs between male and female subjects. For the wallet and hand or shoulder bag items it does appear that the subjects presumed to be personally most familiar with these items were able to provide more consistent descriptions of their owners. While this lends support to the hypothesized bases for sex effects in impressions formed from visible consumption, the same pattern does not emerge for the lipstick and hair-care items.






For the two experiments for which results were presented, given the confidence level used in testing, nineteen main effects could be expected to occur by chance over the two subject groups. In fact, more than five times this many main effects were found to be significant. Even moderately visible and sometimes subtle consumption differences were found to produce consistent differences in the impressions formed of the consumers of these products and services. Of course, the particular effects obtained in these experiments can be expected to change with different groups of perceivers and over time. It is also likely that effects may differ somewhat with different presentation contexts. For instance, if consumption items are presented via photographs or videotapes, the cue that these items were lost or stolen disappears, the physical features of the stimulus person are introduced, and selective perception is likely to be more operative. Nevertheless, the detective study methodology presents an efficient way to examine the combined effects of a number of visible consumption items on impression formation.

An examination of the effects summarized in Tables 6 and 7 reveals one disturbing but expected result. For example, consider the effect of the air/bus treatment in Table 6 on females. Not only was the air traveler judged by relatively objective wealth-related criteria as being higher income, higher occupational status, more highly educated, and missing more money than the bus traveler, but he was judged by criteria objectively unrelated to wealth to be more likeable, successful, interesting, generous, responsible, attractive, and aggressive! In part, this halo effect may be due to the business student subjects' identification with the socially distant reference group of executive/air travelers, [See Cocanougher and Bruce (1971).] but it also seems apparent that the more objective judgment of wealth creates a bias toward an array of other positive characteristics which bear no logical relationships to income. Whether estimates of income are in fact focal inferences which bring with them a number of collateral judgments in forming impressions based on visible consumption is a question which must await more detailed examinations of information processing and integration of such cues. Two somewhat parallel processes which may shed some light on this question are investigations of the integration of brand attribute information into an overall brand attitude, and the integration of information about the products and brands carried by a retailer into the overall impressions of the retailer.

We need to know much more about the role of visible consumption in impression formation. As Figure 2 illustrates, impression formation is an interactive process involving images of products and services as well as images of people consuming these items. Inferences about a product user are determined in part by the products he or she is seen to use or consume, provided that some prior image of these products exists for the perceiver. Similarly, inferences about the image of a product are partially determined by those seen to use or consume the product, provided that some prior image of these users exists for the perceiver. In addition to previously received external information about the person or product, inferences may be aided by information about the consumption situation in which a product is being used (e.g., an outfit being worn in a church rather than on a tennis court), and information about the role being enacted by the consumer (e.g., wearing an outfit as a salesperson in a particular store rather than as a customer in this store). The effects of product image on inferences about product user image are likely to be strongest where little prior information about the person exists, but Bem (1967) argues that some such inferences may take place even when the perceiver is the stimulus person. Whether or not such self-perception is normally a part of impression formation via visible consumption, however, it is clear that in order to know more about the role and function of product images, we need to know more about how the use of visible products and services influences our impressions of people.




Darrel J. Bem, "Self Perception: An Alternative Interpretation of Cognitive Dissonance Phenomena," Psychological Review, 74 (1967), 183-200.

A1 E. Birdwell, "A Study of the Influence of Image Congruence on Consumer Choice," Journal of Business, 41 (1968), 76-88.

Bobby J. Calder and Robert E. Burnkrant, "Interpersonal Influence on Consumer Behavior: An Attribution Theory Approach," Journal of Consumer Research, 4 (1974), 29-38.

A. Benton Cocanougher and Grady D. Bruce, "Socially Distant Reference Groups and Consumer Aspirations," Journal of Marketing Research, 8 (1971), 379-381.

John M. Darley and Joel Cooper, "The 'Clean for Gene' Phenomenon: The Effect of Students' Appearance on Political Campaigning," Journal of Applied Social Psychology, 2 (1972), 24-33.

Ira J. Dolich, "Congruence Relationships Between Self Images and Product Brands," Journal of Marketing Research, 6 (1969), 80-84.

Anthony N. Doob and Allen E. Gross, "Status of Frustrator as an Inhibitor of Horn-Honking Responses," Journal of Social Psychology, 76 (1968), 213-218.

James Dwyer, "Analysis of Variance and the Magnitude of Effects: A General Approach," Psychological Bulletin, 81 (1974), 231-237.

Daniel Greeno, Montrose S. Sommers, and Jerome B. Kernan, "Personality and Implicit Behavior Patterns," Journal of Marketing Research, 10 (1973), 63-69.

Edward L. Grub and Harrison L. Grathwohl, "Consumer Self-Concept, Symbolism, and Market Behavior: A Theoretical Approach," Journal of Marketing, 5 (1967), 22-27.

Edward L. Grubb and Gregg Hupp, "Perception of Self, Generalized Stereotypes, and Brand Selection," Journal of Marketing Research, 5 (1968), 58-63.

Edward L. Grubb and Bruce L. Stern, "Self-Concept and Significant Others," Journal of Marketing Research, 8 (1971), 382-85.

Mason Haire, "Projective Techniques in Marketing Research," Journal of Marketing, 14 (1950), 649-656.

Bruce C. Hamm and Edward W. Cundiff, "Self-Actualization and Product Perception," Journal of Marketing Research, 6 (1969), 470-472.

Rebecca C. Holman, "Communicational Properties of Women's Clothing: Isolation of Discriminable Clothing Ensembles and Identification of Attributions Made to One Person Wearing Each Ensemble," unpublished dissertation, University of Texas, 1976.

E. Laird Landon, Jr. "Self-Concept, Ideal Self-Concept, and Consumer Purchase Intentions," Journal of Consumer Research, 1 (1974), 44-51.

Monroe Lefkowitz, Robert R. Blake, and Jane S. Mouton, "Status Factors in Pedestrian Violation of Traffic Signals,'' Journal of Abnormal and Social Psychology, 51 (1955), 704-706.

Nissum Levy and Harold Schlosberg, "Woodworth Scale Values of the Lightfoot Pictures of Facial Expression," Journal of Experimental Psychology, 60 (1960), 121-125.

Joseph B. Mason and Morris L. Mayer, "The Problems of the Self-Concept in Store Image Studies," Journal of Marketing, 34 (1970), 67-69.

William J. McKeachie, "Lipstick as a Determinator of Personality," Journal of Social Psychology, 36 (1952), 241-244.

Ivan Ross, "Self-Concept and Brand Preference," Journal of Business, 44 (1971), 38-50.

Peter Suedfeld, Stephen Bochner, and Carol Metas, "Petitioner's Attire and Petition Signing by Peace Demonstrators" A Field Experiment," Journal of Applied Social Psychology, 1 (1971), 278-283.

Ronald R. Taft, "The Ability to Judge People," Psychological Bulletin, 52 (1955), 1-23.

Renato Tagiuri, "Person Perception," in Gardner Lindzey and Elliot Aronson (eds.), Handbook of Social Psychology, 2nd Edition, Volume 3 (Reading: Addison-Wesley, 1969), 395-449.

Peter B. Wart and Christopher Knapper, The Perception of People and Events (London: Wiley, 1968).

Peter B. Warr and Alan Sims, "A Study of the Cojudgment Process," Journal of Personality, 33 (1965), 598-604.

Frederick E. Webster, Jr. and Frederick von Pechmann, "A Replication of the 'Shopping List' Study," Journal of Marketing, 34 (1970), 61-63.

Robert L. Westfall, Harper W. Boyd, and Donald T. Campbell, "The Use of Structured Techniques in Motivation Research," Journal of Marketing, 22 (1957), 134-139.