Involvement and You: 1000 Great Ideas

Joel B. Cohen, University of Florida
[ to cite ]:
Joel B. Cohen (1983) ,"Involvement and You: 1000 Great Ideas", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 325-328.

Advances in Consumer Research Volume 10, 1983      Pages 325-328


Joel B. Cohen, University of Florida

Anyone who presumes to write about the concept of involvement should select an involving title: otherwise your credibility to write about this topic is at risk. This is the first of the 1000 great ideas . Actually, I only have one or two other ideas to offer, and they're better described as "solid" than as "great." [I prefer to think of the title as "puffery" rather than as false and deceptive. Anyone still recall those Charles Atlas ads in comic books?] Now that you've foregone other alternatives and have taken the trouble to read this far you can either say, "I've been had...again" (which, I should point out, is not entirely a flattering self-revelation) or you can say, "I know I'm going to benefit immeasurably from the rest of this paper." Frankly, if it's all the same to you I'd prefer that you be in the latter and far more receptive frame of mind as you read what I have to say. I will be brief and only occasionally obtuse.


[Many of the points made in this section are taken from Cohen (1982b).]

I do not subscribe to the practical view of steaming full speed ahead on research and implications without first taking the trouble to carefully define the concept of interest and place it within a theoretical framework. There's been a fair amount of that with respect to "involvement." It's not always been easy, therefore, to know if we're actually going anywhere or merely steaming in circles.

Each of the three papers I shall comment on shortly offers a number of interesting ideas and suggests some promising directions for future research. These papers and the involvement literature of recent years more generally acknowledge the plurality of views regarding the meaning of the term "involvement." While the field may elect to continue using the term to refer, on the one hand, to pre-existing motivational aspects of either an individual, an object or a setting, and at other times to an integrated set of information processing operations (i.e., high involvement vs. low involvement models), doing so interferes at a minimum, with the field's ability to communicate about it both internally and externally. The first order of business, it would seem then, is to try to agree on what involvement is.

There is a tradition of work on ego-involvement and self-concept/value-relatedness by a host of psychologists at least as far back as Freud (e.g., object cathexis) and extending through most contemporary work on both motivation and personality. A direct outgrowth of this work are attempts by consumer behavior researchers to characterize objects and issues as either high or low in involvement, in the sense of personal importance or relevance (e.g., Robertson, 1976; Houston and Rothschild, 1977; Lastovicka, 1979). The work by the Sherifs and colleagues such as Cantril and Hovland related ego-involvement notions to the attitude change area through an analysis of anchoring effects on perceptual judgment (see Sherif, Sherif and Nebergall, 1965). That the motivational aspects of involvement cannot be subsumed within the ubiquitous high to low operationalization should be clear from their research on involvement as commitment to a position. Though associated with issue-involvement, predicted responses to messages reflect an often opposing tendency to reject issue-relevant but position-discrepant ideas.

Were a term such as ego-involvement or self-concept involvement or enduring product involvement to be used in consumer behavior, therefore, it is reasonable to think that little confusion would exist. These terms suggest a readiness to respond to a particular set of stimuli. The term "involvement" by itself, however, seems to refer to an actualized interaction with a stimulus rather than a mere potential to do so. So the suggestion offered here is that the single term (i.e., involvement) not be used to refer to inherent properties of an individual, situation or object. The practice of dressing up simple scales designed to measure how important something is or how much interest the person may have in some object by discussing these as "involvement" is, in my opinion, not helpful. These terms are quite satisfactory in their own right as indicants of certain pre-existing potentials.

Another definitional option is succinctly presented in this session's Batra and Ray paper: "involvement may usefully be conceptualized and operationalized as the depth and quality of message-evoked cognitive responses." Now, given our "gut-level" sense (we shall become slightly more cerebral shortly) for what involvement might be, it is entirely reasonable to suppose that one's cognitive responses will differ as a function of his degree of involvement (though exactly how they will differ is yet to be determined). Nevertheless, such responses must occur subsequent to the onset of involvement. We're going to pile a lot of excess baggage on the term if we start equating it with outcomes (e.g., cognitive responses) of succeeding information processing operations. In addition, we shall lose those aspects of involvement which are not reflected in such measures. While it may turn out that certain types of cognitive responses most often occur under high involvement, such observables are indicants of the construct and not the construct itself. Were it otherwise, there would be no need for such a construct, and we should instead substitute the operational equivalent into our involvement-based theory.

So, whether it's a special partitioning of cognitive responses, a count of the number of personal connections, the amount of information correctly recalled (or perhaps distorted in the direction of some guiding schema under "really high involvement") or some other perfectly sensible "window" on information processing activity or outcomes, all such operations tap involvement-mediated consequences and not involvement per se.

What, then, is a good definition of involvement? There is an advantage to having parsimonious constructs, in the sense that "clean" and straightforward building blocks can then be more readily assembled into larger and useful theories and tested. Relationships among overly broad constructs are necessarily imprecise, and this impairs our ability to refute propositions and reformulate theory. Accordingly, it seems preferable to separate interesting antecedent and consequent variables from the construct of involvement and then to build these variables back into a larger involvement-based theory for the purpose of studying relationships among them. We shall return to this theme a little later on. From this perspective, then, involvement might fundamentally be viewed as a state of activation, and since an essential aspect of involvement is its selectivity, the activation is, in addition, directed to some portion of the psychological field. When we speak of someone as being highly involved, say in his or her own thoughts, we are referring to both the level of activation and the fact that attention seems to be focused inward, probably to the relative exclusion of things going on around the individual. Similarly, someone highly involved in solving a problem appears to have a single-minded purpose and is devoting a substantial proportion of his or her attention and information processing capacities to that task.

Proceeding from this conceptualization, then, a particularly appropriate operationalization of involvement is likely to be closely linked to some measure of attention, ideally one which takes account of capacity available at the time for competing tasks. Secondary task procedures have been used to test attentional hypotheses of a similar nature, and these might be readily adapted to involvement as well as information load research (Moore, 1982). For example, Johnston and Uhl (1976) exposed subjects to a series of words ant, as a secondary task, tones. Since both encoding words and detecting tones requires attention, the speed with which tones were detected provides an index of how much attention was being paid to the words (or, put another way, how involving the word encoding task was). Focused activation and attention are very close cousins, but conceptually at least the former is intended to be more inclusive (i.e., incorporating certain drive-like properties which have effects other than on attention). By involvement I think we mean the former

So, in summary, this definition separates: (1) the activation directed toward a stimulus (i.e., involvement) from, (2) the particular goals, beliefs and interests as well as particular stimulus factors (e.g., novelty, intensity, complexity, distraction) that, as antecedent variables, affect the level and degree of focus of involvement and from, (3) subsequent involvement-induced responses (from attention through cognitive responses to possibly overt behavior). Such a definition is sharply in contrast to some. Krugman, for example, quite clearly does not wish to equate involvement with attention, interest, or excitement but rather with a special class of responses to the advertisement, namely personal connections to one's life experiences and goals. By our account his definition seems to contain some "excess baggage" including as it does involvement-mediated responses. Once again, a person having a higher activation level toward a particular advertising message ("high advertising message involvement") is likely to do a number of things including being better able to comprehend and retrieve the information and to engage in greater elaboration of the content of the message. Thus, specific predictions regarding the impact of involvement on the amount of effort, thought, learning and ultimately belief and attitude change can be mate and studied by including involvement as a treatment manipulation (though the further one moves toward persuasion the more likely one is to find involvement interaction effects). Rather than defining involvement so that it will exclusively refer to any one out of the set of possible responses to an advertisement (e.g., number of thoughts generated, number of correct retrievals), it seems conceptually more appealing to separate the construct of involvement from the set of involvement-mediated responses. A useful next step would be to develop a conceptual framework in which hypothesized links between involvement and sets of antecedent and consequent variables would be carefully laid out. This should lead to a systematic program of research.

Mitchell (1979, 1981) has provided a definition of advertising involvement which is much more consistent with that offered here. Re conceives of involvement as an internal state whose motivation/activation properties are evoked by a particular stimulus. Not only is his a more parsimonious view of this construct than most earlier definitions, but he seeks to enrich the concept by embedding it in a larger-model. Even in Mitchell's work, however, one observes the considerable reluctance within the field to deviate from the practice of reserving the term "high involvement" for involvement with the advertising message, and referring to anything else (including heightened attention to other aspects of an ad) as "low involvement." I feel we should discontinue the practice of referring to the consumer as in a state of "low involvement" just because the individual's attention is disproportionately going to something he or she may be interested in (e.g., aspects of advertising execution) rather than what the advertiser is interested in (i.e., the message)-. Within advertising it is natural to consider the message and aspects of its presentation as important separable factors, though this is, of course, an arbitrary division of the stimulus field. Either, both, or neither of these may result in a high level of involvement for an individual, and so the factors producing such differences in both the intensity and direction of involvement should be studied.


The major contribution of each of the session's papers is its particular focus on one or a number of potentially promising links between involvement and certain antecedent and consequent variables.

Proceeding from the very compatible "situational state" view of involvement taken by Batra and Ray and looking at subsequent processing activity (e.g., cognitive responses) as conceptually linked (i.e., rather than as operational equivalents), several interesting hypotheses are developed. The relationship between degree of involvement with an ad and-the degree of attribute-based response as a percentage of total responses to the at is a key element in their conceptualization of the overall process. I would not stress the similarity between this relationship and memory researchers' (e.g., Craik and Tulving, 1975) depth of processing models because such a large proportion of even low-involvement mediated responses to the "stories" told in commercials are apt to be semantic (i.e., at the level of meaning) rather than "shallow" (e.g., visual, auditory). Nevertheless, in focusing on the greater relative attention to advertising content rather than execution (under high message involvement), Batra and Ray are in essence paving the way for a program of research that encompasses the initial encoding and subsequent processing and retrieval of the information as well as alternative routes to attitude formation.

The development of appropriate coding schemes for cognitive responses to such aspects as attribute assertions and ad execution makes a great deal of sense. Since an average of .5 cognitive responses per person per commercial is not an atypical result, however, it should be clear that a great deal more information processing is being carried out than we are recording. The authors also point out that "affective" responses to commercials seemed to be quite important and that "it was nearly impossible to gauge the intensity and nuances of more affective responses from free-form written verbalizations." This probable limitation of cognitive response measures is important to keep in mind since a characteristic feature of many high involvement episodes is their heightened emotionality. Whether supplementary measures such as evaluative semantic differential scales will satisfy the need to register more "purely" affective responses is not clear. This has always proven to be among the most difficult of reactions to tap. Perhaps we should look again at some of the techniques for more continuous recording of liking-disliking previously explored in theater tests. While subject-controlled dials may not be as precise or objective as physiological measures, they may in fact Prove to be more useful for our purposes!

Batra and Ray want to equate predominantly high involvement processing with "a conscious, voluntary attribute-based evaluation of the brand" and predominantly low involvement processing with "a less effortful, involuntary liking for the brand. [Petty and Cacioppo (1981), in a somewhat related fashion, discussed two routes to attitude change; one termed "central" (i.e., issue-relevant) and the other "peripheral."] While the "degree of attribute-based response as a percentage of total response" would appear to be a useful index of message involvement in many cases, this may not be as general as it might first appear. Park and Young's session paper emphasizes the fact that overall liking for a brand does not necessarily have-to follow an attribute-based combinatorial rule. An advertising message, for instance, that succeeds in creating a "unit relationship" (in balance theory terms) between the brand ant, say, some important self concept-related value may produce a positive affective response without a great deal of attribute-based processing. Or, a message that succeeds only in creating enough perceived similarity between the brand and a positively evaluated object (possibly another product) to have the consumer categorize these together may produce positive affect toward the brand. Also, as ads and brands advertised become more familiar, it is likely that consumers will "short circuit" any systematic consideration of attributes and will access instead an already formed overall evaluation, even when the ad is - particularly involving. Thus, ones cognitive responses (to the extent they mirror actual processing operations; see Lynch and Srull, 1982) may not be focused on attributes to the degree Batra and Ray suggest under high involvement.

As already mentioned, the Park and Young paper offers some provocative ideas concerning "affective involvement" as distinct from "cognitive involvement." [I should thank the authors for giving special attention to some of my recent writings on both involvement and particularly non-attribute based evaluation processes.] Whereas Batra and Ray stress the latter (in the sense of cognitive responses focused on attributes) virtually as a definition of high involvement, Park and Young's paper gives credence to two alternative high involvement mechanisms. These may be conveniently summarized as "analytical" (using attribute-based combinatorial rules to derive an overall evaluation) and "nonanalytical" (the evaluation is an outcome of exemplar-based categorization). So, in the second of these, a person in a high involvement state would engage in sufficient "top-down" processing to retrieve from memory at least one relevant product exemplar (i.e., an instance of a comparable liked or disliked product) and would assess its overall similarity to the to-be-categorized brand. [Think about spotting a fairly large dog down the street and trying to decide whether this is a "nice dog" or a "dog to be avoided." One way to do this is to picture an exemplar from each category and make some overall comparison. In scanning our mental pictures, some specific features may then become salient, possibly because they seem to discriminate instances of the two categories, and these features may be used for a more detailed analysis if necessary.]

As someone who has previously argued that non-analytical processes should be given serious attention in our field, I am pleased to see their application to involvement. The conditions under which people may engage in nonanalytical product categorization and evaluation are yet to be determined, including the involvement state of the individual. The study reported by Park and Young may simply have been too ambitious given the lack of prior research on this topic. One difficulty is that the study apparently does not control for prior experience or knowledge about the product. What impact people's previous processing of advertisements and other information for shampoo might then have on their responses to the present commercial and their ability to effectively role-play a person exclusively using one (and only one) processing strategy cannot, therefore, be determined. Since the manipulation check data is not reported to cut down the length of their paper we have very little to go on in empirically evaluating the success of these complex and demanding role-playing assignments. There is, for example, in the "affective involvement" (nonanalytical processing) condition no way of knowing whether any particular exemplar was even used and what categorization response might then have resulted for the advertised brand. In addition, the role-playing manipulation may have guided subjects in their responses to the dependent measures since certain of the cognitive responses and Fishbein scale responses seem most "appropriate" in each condition. In short, trying to identify which mode of processing is being employed (or trying to restrict processing to one mode) is a challenging task, and we may have to proceed in a step by step fashion, initially using less complex stimulus materials and settings.

Park and Young's second hypothesis predicts a lack of relationship between attribute-based responses and attitude toward the commercial on overall brand attitudes. The evidence cited in support of it cannot be used to make the case for a nonanalytical processing (affective involvement) explanation. All factors that are consistent with a null hypothesis must also be considered in such cases (e.g., failure of manipulations, measurement problems). Ignoring other strictly data analysis issues, the correlations themselves can at best be no more than suggestive. In and of themselves they are hardly evidence that a particular type of information processing occurred. In summary, then, Park and Young offer an intriguing account of an alternative high involvement process, and I feel certain they will follow this up with careful empirical work.

Deighton's session paper focuses on the low involvement state and, I'm happy to note, argues for a broadened and more useful examination of the concept in relation to its antecedents and consequences. He draws an analogy between a consumer and a scientist, suggesting that high and low involvement tend to correspond respectively to "good" scientific practice (e.g., formulate and test hypotheses by disconfirmation) and "poor" scientific practice (e.g., the use of schema-consistent, rather than evidence-consistent inferences). This is an interesting parallel. Whether these are merely hypothetical end points along some abstract continuum or actually correspond in important respects to consumers' behavior is, of course, the key empirical question. While we await pertinent evidence on this it may be worth noting that recent research in psychology has not been quite as kind to "strong" versions of schema theory (such as the view presented by Deighton) as the paper suggests. In a review article, Alba and Hasher (1983) first note that there have really been a collection of schema theories rather than one well articulated theory. As such its propositions are sometimes hard to pin down. However, all schema theories adopt the position that during encoding there is a net reduction in the amount of information that will be stored in memory. Certain encoding processes are presumed to be selective, filtering out non-essential information or choosing a portion of the information for further processing; while other processes create a more holistic representation of the information that is carried further, often enriching it with background information and inferences. Research on memory storage is not generally supportive of schema theories: memory for events is typically more detailed and accurate and not so tightly integrated, and people seem to be able to distinguish between the information itself and "enriching" interpretations they may have made. The particular findings in the literature seem to have a good deal to do with the procedures used to measure information retrieval.

This is not to argue that people fail to engage in "top-down" processing or do not organize encoded information into interconnected and perhaps hierarchical memory structures. It does suggest, however, that we should not simply assume that the "adoption of a schema" inevitably has certain consumer behavior outcomes. In fact we should probably regard the "adoption of a schema" as a little too vague a notion until we investigate specifically what this means in more concrete terms. Otherwise we may interpret the filling of any information gap a consumer faces as schema-directed biases rather than, for example, an acknowledged inference made on the basis of seemingly reliable and pertinent information. While the former may qualify as "poor scientific practice" the latter may actually be a prudent and sensible response.

Despite these limitations in our current understanding of how schemas function, Deighton's overall perspective is itself a useful contribution. He offers an insightful account of how advertising may effect how we think about and structure a product decision. ("Effective advertising determines the questions that are posed, even before it seeks to influence the answers that are found.") This goes beyond the mere transmission of information or a direct attempt to create more favorable attitudes.

A basic tenet of Deighton's thesis is that advertising together with "uninvolved thinking" is apt to produce a considerable amount of bias (perhaps schema-driven) in subsequent information processing operations, particularly inference. His concluding section on confirmatory bias, while ambiguously linked to degree of involvement (one might speculate about a U-shaped relationship) is particularly provocative. Though we should bear in mind that predictions based on somewhat similar biases toward cognitive consistency were seldom upheld when such behavior was seen by people to be counter to their best interests, there are at a minimum subtle forces at work in the direction of confirmatory biases (e.g., availability, inadequate consideration of less salient factors). So, possibly because of memory priming effects which direct thinking down advertising-stimulated pathways or an inherent difficulty in reasoning through or operationally carrying out disconfirmation tests (e.g., for ingredient-effectiveness claims), some amount of confirmatory bias may operate in the marketplace. This is an interesting point to be considered by those who have stressed the "self correcting" aspects of competitive market forces and specifically, the unquestioned ability of product labels to offset overzealous and borderline advertising practices. I for one very much look forward to some good empirical research on many of the worthwhile ideas presented at this session.


Alba, J.W. and Hasher, L. (1983), "Is Memory Schematic," Psychological Bulletin, in press.

Cohen, J.B. (1982a), "The Role of Affect in Categorization: Toward A Reconsideration of the Concept of Attitude," in Jerry Olson (ed.), Advances in Consumer Research, 9, 94-100.

Cohen, J.B. (1982b), "Involvement: Separating the State from Its Causes and Effects," A paper presented at the Involvement Colloquium held at New York University, June 3-4.

Craik, F.I.M. and Tulving, E. (1975), "Depth of Processing and the Retention of Words in Episodic Memory," Journal of Experimental Psychology: General, 104, 268-294.

Houston, M.J. and Rothschild, M.L. (1977), "A Paradigm for Research on Consumer Involvement", Unpublished Paper, University of Wisconsin.

Johnston, W.A. and Uhl, C.N. (197 i), "The Contributions of Encoding Effort and Variability to the Spacing Effect on Free Recall," Journal of Experimental Psychology: Human Learning and Memory, 2, 153-160.

Lastovicka, J.S. (1979)j "Questioning the Concept of Involvement Defined Product Classes", In William L. Wilkie (ed.), Advances in Consumer Research, 6, Ann Arbor: Association for Consumer Research, 174-79.

Lynch, J.G. and Srull, T.R. (1982), "Attentional Factors in Consumer Choice: Concepts and Research Methods," Journal of Consumer Research, 9, 18-37.

Mitchell, A.A. (1979), "Involvement: A Potentially Important Mediator of Consumer Behavior", In William L. Wilkie (ed.), Advances in Consumer Research, 6, Ann Arbor: Association for Consumer Research. 191-95.

Mitchell, A.A. (1981), "The Dimensions of Advertising Involvement", In Rent B. Monroe (ed.), Advances in Consumer Research, 8, Ann Arbor: Association for Consumer Research. 25-30.

Moore, D.L. (1982), Behavioral Measurement of Information Load. Working Paper. University of Florida.

Petty, R.E. and Cacioppo, J.T. (1981), Attitudes and Persuasion: Classic and Contemporary Approaches, Dubuque, Iowa: Wm. C. Brown.

Robertson, T.S. (1976), "Low Commitment Consumer Behavior", Journal of Advertising Research, 16, 619-27.

Sherif, C.W., Sherif, M., and Nebergall, R.E. (1965), Attitude and Attitude Change: The Social Judgement Involvement Approach, Philadelphia: Saunders.