The Influence of Level of Involvement on the Feature Matching Process in Consumer Preference Judgments

Alan Strathman, University of Missouri-Columbia
David S. Boninger, University of California-Los Angeles
Sara M. Baker, Ohio State University
ABSTRACT - Previous research has indicated that Tversky's (1977) feature matching model of similarity judgments can be applied to judgments of preference (Houston, Sherman, & Baker, 1989). However, this research has examined preference judgments only under conditions of low involvement. This study investigated the hypothesis that the feature matching process by which the features of the more recently presented product (the subject) drive the comparison of features with the earlier product (the referent) would be attenuated under conditions of high involvement. As predicted, under high involvement, the subject was preferred equally under unique positive and unique negative conditions.
[ to cite ]:
Alan Strathman, David S. Boninger, and Sara M. Baker (1992) ,"The Influence of Level of Involvement on the Feature Matching Process in Consumer Preference Judgments", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 777-781.

Advances in Consumer Research Volume 19, 1992      Pages 777-781

THE INFLUENCE OF LEVEL OF INVOLVEMENT ON THE FEATURE MATCHING PROCESS IN CONSUMER PREFERENCE JUDGMENTS

Alan Strathman, University of Missouri-Columbia

David S. Boninger, University of California-Los Angeles

Sara M. Baker, Ohio State University

ABSTRACT -

Previous research has indicated that Tversky's (1977) feature matching model of similarity judgments can be applied to judgments of preference (Houston, Sherman, & Baker, 1989). However, this research has examined preference judgments only under conditions of low involvement. This study investigated the hypothesis that the feature matching process by which the features of the more recently presented product (the subject) drive the comparison of features with the earlier product (the referent) would be attenuated under conditions of high involvement. As predicted, under high involvement, the subject was preferred equally under unique positive and unique negative conditions.

INTRODUCTION

People are frequently in situations in which they must make preference judgments by comparing alternatives. Individuals contemplating a vacation may be faced with the task of determining which of two or more vacations they prefer. At other times they must make preference decisions between competing automobiles or alternative houses. Recent research indicates that, if one controls for the attractiveness of the alternatives, relative preference can be affected by the order in which the alternatives are considered. Houston, Sherman and Baker (1989) found that when people were asked to compare the desirability of alternatives, they were more sensitive to the unique information provided by the more recently presented alternative. Their results were consistent with predictions derived from a feature matching process (Tversky, 1977) and indicated that Tversky's model of similarity judgments can be applied to judgments of preference. The current research seeks to further our understanding of the role played by feature matching in consumer preference judgments and to determine whether the feature matching process will manifest itself under conditions of heightened involvement.

Feature Matching

Tversky (1977) presented a feature matching process that attempts to explain comparison processes in judgments of similarity. According to Tversky (1977), comparisons of objects involve a subject of comparison and a referent. In a comparison of the form "A is like B" (e.g., Ford is like Chevrolet) "A" is the subject, and "B" is the referent. Perceived similarity differs depending upon whether people are asked whether A is like B or B is like A.

In Tversky's model, objects may be represented as sets of features. These objects are assumed to have shared features (i.e., features belonging to both a and b) and/or unique features (i.e., features belonging to a but not b and features belonging to b but not a). It is argued that, in a similarity judgment task such as judging the similarity of a to b, people naturally focus on the subject. They compare the features of the subject with the same features of the referent. According to Tversky (1977), the features of the subject are weighted more heavily than the features of the referent. Thus, "the features of the subject control the agenda of the comparison" (Houston et al., 1989, p. 122), and judged similarity is affected more by the unique features of the subject than the unique features of the referent. Tversky (1977) reported research with pairs of countries (e.g., U. S. A. - Mexico), pairs of figures, and pairs of letters to support the propositions of the feature matching process.

Houston et al. (1989) examined implications of the feature matching process for preference judgments. For each of several categories (e.g., automobiles, vacation spots) they constructed feature-based descriptions of alternative items or persons. The descriptions were constructed so they had either unique good and shared bad features or unique bad and shared good features. Overall desirability of the features used in each description was held constant making each alternative equally desirable. In Experiment 1, participants were instructed to try to form a general impression of each item or person. They were then shown a set of category descriptions. After people viewed the first item in each pair they were given a second item for each pair and asked to indicate their preference on a 12-point scale varying from strongly favor the first item to strongly favor the second item. It was expected that, by keeping participants unaware of the experimental task until they viewed the second item in each pair, the second item in each pair would become the subject of the comparison (cf., Agostinelli, Sherman, Fazio, & Hearst, 1986). Greater preference was shown for the first item when both the first and second item contained unique bad features than when both the first and second item contained unique good features. These results are consistent with use of a feature matching process in which the second item is the subject and the first item is the referent.

In Experiments 2 and 3, Houston et al. (1989) attempted to make the first item the subject of comparison. Although these attempts to manipulate the subject of comparison were unsuccessful, the feature matching process was replicated such that when both items had unique negative features the first item was preferred more strongly than when both items had unique positive features. To summarize, the Houston et al. (1989) results are consistent with the position that, when subjects are exposed to complete presentations of the items (i.e., all the features at once) they spontaneously form preference judgments by a feature matching process upon their exposure to the second item in a category.

Involvement

An important way to further our understanding of the feature matching process is to prescribe its limits. Research that has examined the role of involvement or personal relevance on persuasion has shown that increasing involvement increases the tendency of people to engage in a careful consideration of message arguments (Petty and Cacioppo, 1986). For example, Petty, Cacioppo and Schumann (1983) manipulated involvement by leading people to believe they would be able to choose a product as a gift to take home. The high involvement group was led to believe they would be able to take home a brand of disposable razor whereas the low involvement group believed they would be able to take home a product in a different product class. Subjects were then exposed to a series of ads including an ad for a disposable razor which contained either strong or weak product arguments. Data indicated that argument strength (i.e., whether the described product features were weak or strong) had a greater impact on attitude toward the product under high involvement than low involvement. This result is consistent with the position that subjects engage in greater elaboration of the message arguments under high involvement than low involvement.

In Houston et al. (1989), level of involvement was not manipulated and subjects were given no expectations that their task was self-relevant or had any important consequences. Thus, there is no evidence that suggests that feature matching, a low effort strategy, is prevalent in important contexts where the alternatives themselves or possible outcomes of the decision heighten the level of involvement.

We hypothesized that conditions of high involvement should motivate people to more carefully consider the information about the first product at the time that it is presented. Moreover, we expect that high involvement subjects will more thoroughly process that information than will low involvement subjects. When subsequently induced to make a preference judgment, they will be more likely than low involvement subjects to utilize the more thoroughly processed information in their decision making. In other words, greater consideration of information from the first product presented has the effect of equalizing the weights of the features assigned to the subject and the referent. Consequently, we expect that increasing involvement will moderate or eliminate the biasing effects of unique features of the second item in preference decisions.

Finally, we expect that these effects will be reflected in recall (Gati & Tversky, 1987). According to the feature matching process, the unique features of the subject drive the comparison process. If this is the case, one would expect greater recall of the unique features of the subject than of the referent. However, we hypothesized that if individuals in the high involvement condition are exerting enough effort to attenuate the feature matching effect, then their recall scores should parallel this. That is, these individuals should recall equivalent amounts of unique information from the first product as from the second product.

The following study was conducted in order to achieve three goals. First, because Houston et al. provide the only demonstration of the use of a feature matching process in preference judgments, our first step was to replicate their findings with vacation spots, a category they utilized. Next, we examined the use of a feature matching process under conditions of manipulated low involvement. If Houston et al. subjects were indeed uninvolved in the decision process, then data from subjects manipulated to be in a low involvement state should mirror the Houston et al. findings and our own replication. Finally, we examined the use of the feature matching process under conditions of high involvement.

METHOD

Subjects

Subjects were 146 students at Ohio State University who participated in partial fulfillment of a course requirement. Subjects participated in groups ranging in size from 10 to 25.

Materials

Subjects were presented with descriptions of two products for each of two product categories: vacation spots, and pens. In addition, subjects were presented with descriptions of two products from the soft drink product class, which was included only as a filler product for the manipulation of involvement. Each product description consisted of a list of six descriptive phrases (features). For example, features for vacation spots included "beautiful scenery", "overcrowded", and "reputation as a tourist trap." Features for pens included "expensive", "ink smudges" and "looks professional."

Pretests were conducted to create products with comparable sets of features. Each feature was rated for positivity and importance using a procedure similar to that of Houston et al. (1989). Pretest subjects were given a large list of possible descriptive phrases for one of the product categories. For each of the descriptive phrases, subjects were asked to rate on 9-point scales how positive they found the feature (l-negative, 9-positive) and how important they found that feature (l-not at all important, 9-very important).

Mean positivity ratings were then computed for each descriptive feature. For each product category, features whose mean positivity ratings were not statistically different and whose importance ratings were similar were chosen. For each product category, three lists of features were constructed. Each list consisted of three positive features and three negative features. The pairing of one set of two lists of features resulted in product descriptions that shared positive features but had unique negative features, while another pairing resulted in products which shared negative features but had unique positive features. Within each pairing, order of list was counter-balanced. This procedure insured that subjects were choosing between products which were evaluatively equal.

Procedure

The procedure in the present experiment was derived from the procedure employed by Houston et al. (1989). Subjects were randomly assigned to be in unique positive or unique negative conditions for vacation spots. These same subjects also participated in an additional condition in which involvement with the pen decision was manipulated.

Subjects were seated for the experiment in a manner which prevented interaction. They were told that the purpose of the experiment was to investigate how individuals formed impressions of products. Therefore, they would be presented with descriptions of three types of products.

At this point the manipulation of involvement occurred. This manipulation is modeled after the procedure employed by Petty, Cacioppo, and Schumann (1983). Subjects were told that in appreciation for their participation, they would receive a small gift. High involvement subjects were told their gift would be a pen and that they could choose their pen from among a variety of pens. In addition, subjects were told that the pen they would see described would soon be marketed in their city.

Low involvement subjects were told that they could choose their gift from among a variety of soft drinks and that the soft drinks they would see described would soon be marketed in their city. Therefore, these subjects were expected to be uninvolved with the pen descriptions (cf., Petty, Cacioppo, & Schumann, 1983).

Subjects were then presented with the first booklet which contained descriptions of three products, one from each of the product categories. Subjects were paced through the descriptions and were given 20 seconds to view each descriptive list.

After finishing with this booklet, subjects were given a second booklet. Subjects were told that the booklet contained descriptions for another brand in each of the product classes contained in the first booklet. Subjects were to view each of the descriptions and then state a preference for either the product they saw first or second for each of the product types. Preferences were expressed on a 12-point scale with endpoints labeled "strongly prefer first product" and "strongly prefer second product."

In addition to this preference scale, subjects responded to two measures designed to assess the effectiveness of the involvement manipulation. Subjects also responded to a series of ancillary measures and were asked to recall as much of the information concerning each of the pen descriptions as they could.

RESULTS

We first attempted to replicate Houston et al.'s findings. In addition, we tested two hypotheses: 1) subjects in a manipulated low involvement state would use the feature matching process in making their preference judgments as in Houston et al., and 2) use of the feature matching process would be attenuated for high involvement subjects.

Subjects expressed their preference for each product category on a 12-point scale (l = strongly prefer first product, 12 = strongly prefer second product). T-tests were conducted on responses to this scale for the vacation product category as a function of feature matching condition (unique positive features vs. unique negative features). This analysis revealed a significant feature matching effect, t(135) = 2.87, p<.01. These data replicate those of Houston et al. (1989) and indicate that individuals who viewed vacation spots with unique positive features preferred the second product (M = 6.5) more than individuals who viewed vacation spots with unique negative features (M = 5.1).

In order to check the effectiveness of the involvement manipulation, subjects were asked to indicate on two self-report measures how hard they were concentrating, and how much attention they were paying to the descriptions of the pens. A mean of these two measures was computed and served as our measure of involvement. A t-test of this measure comparing high involvement subjects to low involvement subjects revealed a significant difference, t(127) = 2.08, p<.05, indicating that high involvement individuals reported paying more attention to, and concentrating harder on the pen descriptions than did low involvement individuals.

Next, we examined preferences of low involvement subjects as a function of feature matching condition. As expected, this analysis clearly indicated that subjects who received products with unique positive features more strongly preferred the second item (M = 6.8) than did subjects who viewed products with unique negative features (M = 4.9), t(135) = 2.21, p<.03. Because these results parallel the Houston et al. findings and our own replication reported above, it seems likely that the Houston et al. study was conducted in a low involvement context.

Finally, we examined preferences of high involvement subjects as a function of feature matching condition. As predicted, t-tests indicated that individuals in the high involvement condition were no more likely to favor the second product when they saw products with unique positive features (M = 5.9) than when they saw products with unique negative features (M = 5.2), t(135) = .90, p>.35. Thus, the feature matching process was eliminated in a high involvement context.

In addition, subjects were asked to recall as many of the features from each of the pen descriptions as they could. We expected that under low involvement, recall of the unique features of the subject of comparison (i.e., the second product presented) would be higher than recall of the unique features of the referent. However, we expected that under high involvement, this recall difference would disappear. The number of unique features recalled from the first product was compared with the number of unique features recalled from the second product for high and low involvement. This analysis indicated that low involvement subjects recalled more unique features from the second product (M = .59) than from the first product (M = .09), t(65) = 1.91, p = .06. In contrast, high involvement subjects did not recall a greater number of unique features from the second item (M = .63) than from the first item (M = .53), t(75) = .37, p>.70.

DISCUSSION

The results of this study further attest to the role of Tversky's feature matching model in preference judgments while also increasing our understanding of the process and the conditions under which it may be attenuated. Under low involvement conditions, the process by which the features of the more recently presented product (the subject) drive the comparison and retrieval of features from the earlier product (the referent) was exhibited for both vacation spots and pens. For each category, two pretested, evaluatively equal brands were presented. In each case, if the subject had unique positive features it was preferred, but if the subject had unique negative features, the referent was preferred.

This study also elucidates conditions under which the feature matching process does not determine preference judgments. When making comparisons using the feature matching process, people may consider the features of the first product only once the features of the second product are presented. This allows the features of the second product to drive the comparison (Houston et al., 1989). If this tendency was eliminated, then the biasing effects of order that are necessary in the feature matching process might be greatly reduced. In accordance with this reasoning, we hypothesized that conditions of high involvement would motivate people to more carefully consider the information presented about the first product at the time that it is presented, thereby increasing the weights assigned to the features of the referent and making them more nearly equal to the weights assigned to the features of the subject. Under such conditions, the feature matching effect should be significantly attenuated.

In the present study, an examination of the effect of level of involvement supported this hypothesis. For low involvement subjects, the feature matching effect was significant, but for high involvement subjects, the effect was eliminated such that the subject was preferred equally under unique positive and unique negative conditions. This effect of involvement on feature matching may be understood in terms of Tversky=s assumption that weights are associated with the features of each item. For low involvement, the weights assigned to the features of the subject are greater than those assigned to the referent. In fact, the weights assigned to the unique features of the referent may fall to zero such that those features are not considered in the comparison process and will not be recalled. However, under high involvement, the weights for the features of the subject and the referent are more likely to be equal and non-zero, thus increasing the overall recall.

Thus, the use of the feature matching process in making preference judgments appears to be limited to low involvement settings such as those in which the difference among the alternatives is considered relatively unimportant or the outcomes of the decision are not considered serious. It may be worth noting that a real-world manipulation of involvement may be stronger than that accomplished in the laboratory. Such a manipulation should result in a pattern of data which is even more dramatic than the present data.

We also found support for the hypothesis that the influence of the process of feature matching on recall of unique features would be affected by level of involvement. Because the information about both products is considered more seriously under high involvement conditions, these individuals recalled an equivalent amount of unique information from the first product as from the second product. However, under low involvement more unique features were recalled from the subject than from the referent. Because this recall effect occurs with unique features only, a simple recency explanation is untenable.

These data are encouraging because they mirror the data concerning the effects of involvement on preference judgments. In both cases, differences under low involvement were significant whereas there were no differences under high involvement. Our hypotheses regarding the effects of involvement on preference judgments and on recall of unique features both stem from the more general prediction that the operation of the feature matching process is influenced by level of involvement. Thus, these parallel findings give us greater confidence that the results concerning the effects of involvement on the feature matching process are reliable.

If in fact high involvement subjects are engaging in feature matching, then the biasing by-product of the feature matching process may be reduced because all information is being considered. Alternatively, high involvement subjects may not be using a feature matching process at all. The recall data may be useful in distinguishing the feature matching process from a memory-based recall/accessibility process as has been suggested by Houston and his colleagues. If the feature matching process is not occurring under high involvement, then a more memory-based process may be functioning. In this case, subjects form evaluations of the products for the comparison task by using features of the products which are accessible in memory (Houston et al., 1989). If this memory-based process is utilized under high but not low involvement, then we would expect higher correlations between recall of unique features and preference under high involvement (Hastie & Park, 1986).

Analyses of the recall data lend tentative support to the above hypotheses. Correlations between recall of unique features and preference were computed for high and low involvement for unique positive and unique negative features. The correlations under high involvement were .16 and .15, while under low involvement, they were .07 and .06. These correlations approach statistical significance (p = .17) for high involvement subjects only. Although these data clearly are not conclusive, they do suggest the possibility that a feature matching process determines preference judgments in low involvement settings while a memory-based accessibility process determines preference judgments in high involvement settings. One goal of future research may be to more clearly resolve this issue.

The present study advances our understanding of the process by which people make preference judgments in a number of important ways. First, it affirms the surprisingly influential role of the feature matching process in individual preference judgments. When comparison alternatives are close on the evaluative scale and people are not especially motivated, the unique features of the subject of comparison will play an inflated and critical role in the preference judgment. This may occur in many real world settings in which information is serially presented such as when viewing successive advertisements from the same product class or when shopping in several stores for a product.

More importantly, our findings suggest an important limiting condition of the feature matching process. Individuals in high involvement settings engage in sufficient initial processing of comparison alternatives to wipe out the feature matching effect. Indeed, there is tentative support to suggest that instead they may be using a more memory-based process. Thus, the biasing process of feature matching may be less prevalent in important contexts where the alternatives themselves or possible outcomes of the decision heighten the level of involvement. This may be especially true in situations where the self is implicated or is an alternative in the comparison process. In such cases, we may expect more thoughtful and elaborative strategies to supersede the feature matching process.

REFERENCES

Agostinelli, Gina, Steven J. Sherman, Russell H. Fazio and Elliot S. Hearst (1986), Detecting and identifying change: Additions versus deletions. Journal of Experimental Psychology: Human Perception and Performance, 12, 445-545.

Gati, Itamar and Amos Tversky (1987), Recall of common and distinctive features of verbal and pictorial stimuli. Memory and Cognition, 15, 97-100.

Hastie, Reid and Bernadette Park (1986), The relationship between memory and judgment depends on whether the judgment task is memory based or on-line. Psychological Review, 93, 258-268.

Houston, David A., Steven J. Sherman, & Sara M. Baker (1989), The influence of unique features and direction of comparison on preferences. Journal of Experimental Social Psychology, 25, 121-141.

Petty, Richard E. and John T. Cacioppo (1986), Communication and persuasion: Central and peripheral routes to attitude change. New York: Springer-Verlag.

Petty, Richard E., John T. Cacioppo and David Schumann (1983), Central and peripheral routes to advertising effectiveness: The moderating role of involvement. Journal of Consumer Research, 10, 135-146.

Tversky, Amos (1977), Features of similarity. Psychological Review, 84, 327-352.

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