Operationalizing Involvement As Depth and Quality of Cognitive Response

ABSTRACT - This paper conceptualizes message response involvement as situational states characterized by the depth and quality of the cognitive responses evoked by the message. Data is presented to argue, however, that the inherent multidimensionality of such cognitive responses makes the operationalization of such a construct necessarily dependent on the nature of the theory and/or application in which such a construct is used. One theoretical framework is presented, and a potential method of researching such operationalizations is discussed.


Rajeev Batra and Michael L. Ray (1983) ,"Operationalizing Involvement As Depth and Quality of Cognitive Response", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 309-313.

Advances in Consumer Research Volume 10, 1983      Pages 309-313


Rajeev Batra, Stanford University

Michael L. Ray, Stanford University

Lauranne Buchanan (of Stanford) collaborated equally in planning and conducting the pre-study reported in this paper. Financial support for this project was provided in part by the Marketing Management Program of the Graduate School or Business, Stanford University and in part by the Marketing Science Institute.


This paper conceptualizes message response involvement as situational states characterized by the depth and quality of the cognitive responses evoked by the message. Data is presented to argue, however, that the inherent multidimensionality of such cognitive responses makes the operationalization of such a construct necessarily dependent on the nature of the theory and/or application in which such a construct is used. One theoretical framework is presented, and a potential method of researching such operationalizations is discussed.


For all the interest it has generated amongst researchers, the construct of involvement has proved to be extremely difficult to define and, consequently, operationalize. In this paper we hope to show that, in the communication and advertising context, involvement may usefully be conceptualized and operationalized as the depth and quality of message-evoked cognitive responses. We argue, however, that the measurement of involvement in terms of such cognitive responses will necessarily be arbitrary, and that such measurement must therefore be closely tied to theoretical frameworks which make use or the construct as a mediator in persuasion communication processes in various applied settings.

We first present our characterization of message response involvement, with a brief discussion of its antecedents and consequences. Next, we indicate why any operationalization of such involvement will necessarily be incomplete; this point is supported with the results of a pre-study which demonstrates the multidimensionality or message induced cognitive response. We then present a theoretical framework which could be used to shape the measurement of such involvement, but urge that further exploratory research to guide such operationalization be pursued, before final answers are sought.


Two Kinds of Involvement

We believe that a major reason why there continues to be a lack of consensus about the definition and measurement of involvement is simply that the term "involvement" is used interchangeably to describe two qualitatively different phenomena: involvement with a product class and involvement with a message.

"Product class involvement" usually refers to an individual's predisposition to, for example, make a brand choice (in that product category) with care and deliberation, perhaps due to high levels of perceived risk and the like. Such involvement should therefore endure across time, though there could clearly be temporal differences in the intensity of such involvement (Houston and Rothschild 1977; Rothschild 1979). It seems appropriate to ascribe a motivational character to such involvement. (To characterize such product class involvement in this fashion is not, of course, equivalent to adducing evidence that it is a useful construct, or even that it exists. See Ray 1979).

"Message response involvement", on the other hand, can only exist as a very situational state, being specific to the processing of a particular message by a particular individual at a particular point of time. It is a term used to characterize the way in which that specific message gets processed; this manner of processing varies across product classes, brands within a produce class, messages for a given brand, message reception situations, and the individuals who receive that message.

Message response involvement, therefore, exists not as an enduring predisposition, but as an interactive outcome of many situational factors.

For that reason, such message response involvement is not merely motivational in origin; situational variations in such involvement could be due to differences in the situational opportunity to "get involved" (due to media mode effects) as well as the message recipient's ability to get involved (due to the existence or otherwise of prior knowledge structures, in the recipient, dealing with the content of the message). Note, importantly, that in defining message response involvement in this fashion we are drawing a distinction between the antecedent factors of involvement (the motivation, ability, and opportunity to respond) and the state that is a degree of "involvement.

This supra-motivational conceptualization or message response involvement has its origins in work by Wright (1973, 1975, 19-81), and is, in fact, what Krugman, a pioneer in this area, originally intended (1966-67, ?. ;84). Further, the distinctions made above between the two basic kinds of involvement are in substantial agreement with Houston and Rothschild (1977) and Mitchell (1979), though the terminology may, or course, be different.

Message Response Involvement As Cognitive Response

In this paper, we are concerned with the operationalization of such message response involvement alone. We propose that if it seems reasonable to conceptualize such message response involvement as a situational state, it seems reasonable also to measure or calibrate such states by the kinds or cognitive responses that the message actually receives in that state. For semantic clarity, let us clarify that by "cognitive responses" we also include those responses which are commonsensically called "affective."

Characterizing message response involvement through such state-descriptive cognitive responses, however, raises the question: involved in what? Are we talking about the recipient's involvement in the message as a whole? In certain aspects of the message? Clearly, to be "involved" implies a heightened expenditure of mental effort on some aspects of the stimulus. for involvement of any kind should lead to some separation of one's stimulus field into some figure and ground, if only the separation of what one is involved in from what one is not (or less) involved in (see, for example, the characterization of involvement as intensity and direction, by Mitchell (1979)).

Depending on how one answers this question, one could develop various alternative ways of measuring and classifying message evoked cognitive responses.

Cognitive Response Measures: Conventional Alternatives

One could, for instance, measure only the intensity of such message-evoked cognitive response to calibrate this involvement state. To do so, one could use -- as one option -- Krugman's coding of cognitive responses into "connections" (1965, 1966-67). These measure the extent to which the ad content is related (by the respondent) to specific incidents or events in that person's life.

There are at least two problems with this approach. The first is that Krugman's construct and coding scheme have never been validated and remain somewhat idiosyncratic. The second is that in such coding the affective valence, as well as the attribute-based-or-not dimension, are both totally ignored. As a consequence, while one can use a "connections" count to generate and test hypotheses relating to the such "depth of processing", as was done by Webb and Ray (1979), such coding does not permit the study of processes in which the "direction" or "basis" of response is also important.

This drawback is avoided by the use of the other popular coding scheme, where those cognitive responses deemed relevant are coded into support-arguments,counter-arguments, source-discounting thoughts, and curiosity thoughts (e.g. Wright 1973). This scheme gets at response valence (or direction) but not intensity, unless some or these code categories are postulated to be more or less "deep" or effortful than others. (Wright, 1973, for example, found that respondents in a condition given less motivation and opportunity to respond produced far fewer support arguments and source derogations than those in the high involvement manipulation; such responses were thus presumably more effortful.)

However, this coding scheme too suffers from problems, in that it selects out certain responses as being irrelevant, and (with the exception of source derogation) focuses on support and counter argumentation which deal with attribute-based message content. Wright explicitly excludes emotional reactions not accompanied by attribute argumentation, or simple statements of like or dislike for the product (see Appendix A, p. 62, in Wright 1973). Further, the category of source derogation is used only for statements of advertisement/advertiser distrust or derogation, and for dislike for the overall means used by the advertiser in this presentation. Other, less extreme, reactions to the ad execution are presumably ignored.

Were we to use such a scheme, therefore, we would implicitly be defining the phenomenon of interest as the degree of responses, both pro and counter, to the attribute assertions in the ad, as well as source derogations. The problem with this coding scheme, therefore, is that it "a priori" excludes from analysis other kinds of mediating cognitive responses which could potentially help us better understand message response involvement and its antecedents and consequences.

Cognitive Responses: Recent Perspectives

For instance, Wright's coding scheme does not provide all the data needed to investigate the sorts of hypothesized processes that could underlie brand preference formation in low involvement conditions, as recently suggested by some authors (see, for instance, Greenwald, Leavitt and Obermiller 1980; Batra and Ray 1982). Such hypotheses have suggested that, in low involvement message response situations, advertising-induced performances for brands are based not only on the respondent's multi-attribute evaluation of the brand, but are substantially impacted also by the liking for the ad execution itself, as well as "mere exposure" effects caused by advertising frequency.

To study such hypotheses, and to explore the nature or message response involvement, one would therefore need data not only on how message recipients respond to the attribute assertions made in the ad, but also on their liking or disliking of the ad itself and the feelings and emotions evoked by the ad.

It seems necessary, therefore, to go beyond just consumer reactions to the attribute assertions in the ad, in order to fully operationalize involvement. This suspicion is reinforced by an examination of the results of Wright's 1973 study which shows (var. A2) a drop from 43% to 27% in explained variance when moving to low involvement conditions and using just attribute based coding categories for cognitive response.

There is yet another unrelated reason why the inclusion of apparently "irrelevant" responses may aid analysis: if a message tends to evoke a higher than usual amount of "distracting" (from the selling point) processing, one would expect such high levels of "irrelevant responses" to detract from the time given to the rehearsal and elaboration of message assertions, and thence to the "involvement with" such messages. Support for such hypotheses would seem to be readily forthcoming from the literature on distraction effects (Festinger and Maccoby 1964; Osterhouse and Brock 1970; Petty, Wells and Brock 1976).

In sum, then, message response involvement may be conceptualized as a situational state, to be described by the message-evoked cognitive responses that comprise that state; however, analytic coding of such cognitive responses should use far more coding categories than are used in current schemes.


We have begun exploratory work on the development of a data collection and analysis methodology which uses written retrospective verbal protocols to develop a data-based measure of message response involvement and to study low involvement message processing. Our preliminary data seem to indicate that there may be various types of message response involvement, all described through some cognitive response descriptors, and that it may be impossible to avoid a theory-driven (but still arbitrary) selection from these types, to serve whatever research purpose one has at hand.

This Pre-Study

The methodological pre-study to be discussed here was concerned with (1) the identification of a data collection methodology which encourages the reporting of the "liking" and "irrelevant" responses which current schemes seem to neglect, provided such responses are natural and "valid", and (2) the development of a coding scheme which uses logical, mutually exclusive, and collectively exhaustive categories which go beyond current coding schemes in their use of feeling-type, ad evaluation, and "irrelevant thought" data (see also Wright 1980; Wright and Rip 1981).


The "best data collection methodology" objectives were addressed through a sequence of mini-tests, using experimental groups with about 8-10 adult female respondents per group. The objective in trying out these variations in data collection methodology was to identify that variant which most encourages the reporting of non-attribute based responses and "irrelevant" responses while at the same time discouraging the reporting of ad evaluations offered in a "respondent as judge of ad effectiveness" mode.

In each of these groups, respondents answered preliminary questions (on awareness, attitude, usage, prior familiarity, etc.) and were then shown eight test commercials. The screening of each commercial was followed by the collection of message-induced cognitive response retrospective verbalization data, with the key question being, "What thoughts and feelings came to mind while you watched that commercial?"

It is in the instructions that the treatments differed across groups: the instructions were differentially directive ("heavy" = highly directive, "light" = less directive), either with or without an example of a "real" verbal protocol, either with or without a "practice" screening and protocol, and with different ways of data collection ("standard", "cued", "structured").

The different data collection methods are not of interest here and will not be described further. Suffice it to say that in terms of quantity, face validity, and sensitivity of results, the best retrospective verbal protocol method seemed to involve: fairly non-directive opening instructions, an example of previously obtained cognitive responses to a commercial, a practice trial in recording cognitive responses, a standard open-ended cognitive response question (as opposed to a structured one suggesting certain responses or one allowing respondents to "cue" themselves to responses during commercial exposure) followed, for each commercial, by "structured" semantic differential "affective" questions, designed both to jog memories and to ensure that all dimensions of cognitive responses were reviewed by the respondent.

This structured supplemental "affective" questionnaire was developed on the basis of a separate group study as well as industry research (Hells, Leavitt and McConville 1971; Schlinger 1979) and psychological research on the components of emotions (Pribram 19803. This was done because of our observation that while it was easy (in the standard, blank page format) to code the frequency of support arguments, counterarguments, and the like, it was nearly impossible to gauge the intensity and nuances of more affective responses from free-form written verbalizations.


We are concerned in this paper not with differences across the different data collection groups, but in an exploratory Look at the different types of cognitive responses that the eight different commercial evoked, as coded

The coding scheme itself was developed somewhat iteratively ; logical preconceptions regarding the possible different types or responses were modified when actual coding was done, and when checks were made to ascertain the reliability (across coders, and for the same coder across time) of such coding. A coding manual was developed using actual responses from the first few groups. It was checked in a separate group study which asked respondents to sort forty actual responses from previous groups to see if those separate respondents categories were different from our codes. They were not.

See Table 1 for the resulting general coding scheme, which used 12 categories.




Our first observation is that on an aggregate basis "affective" and "irrelevant" responses to commercials are at least as numerous as the responses to attribute assertions that most current coding schemes highlight. Thus, for our sample of 552 protocols (69 respondents for 8 commercials each), the sum total of all positive feelings/liking/evaluation was the largest category in frequency of response (32%), followed by a negative feeling/derogation/evaluation (27%), followed by the total of all support arguments, both connection based and otherwise (25%), and then the total of all counterarguments (16%)

Since there was no dependent variable in this pre-study, it is impossible to assert with confidence that these non-attribute responses explain a part of the variance in such dependent measures that attribute-based responses fail to explain. Further, space limitations preclude a full presentation of our data. Nevertheless, in the illustrative cases discussed below, it should be apparent that these additional variables offer a far richer view of ad-evoked cognitive responses than current codes allow.

Consider the differences between commercials "G" and "D" in our study. There were no statistically significant differences between them in terms of total support arguments, total counter arguments or support arguments minus counter arguments. Nor was there a statistically significant difference between them in terms of brand-related connections. One could easily conclude, based on the conventionally used coding schemes, that these two commercials were equivalent in terms of "involvement" levels.

Even a gross "total positive feelings/liking/evaluation" count, however, shows up one significant difference: commercial "D" evoked significantly (a = .06) more positive "affective" responses than commercial "G", while being equivalent on other categories. One might therefore call "D" more involving.

At a more disaggregated level, however, other subtle differences emerged. The "D" commercial was "liked" more; it evoked more positive "emotional feelings"; yet it was less positively "evaluated" and evoked more source derogations (a < .05). Our inferences on net involvement must therefore be guarded (even if one argues that ad "evaluations" are irrelevant to considerations of "involvement"), because a source derogation is, by its definition in our coding scheme, a fairly intense reaction.

While commercials "D" and "G" were different in execution styles, commercials "K" and "B" were both very "affective" executions. A gross look at the total number of counter arguments or total positive affective reactions leaves them indistinguishable. But "K" evoked more positive "emotional" feelings, while "B" was better "liked" (G < .05). Since "emotional feelings" are more intense in our coding scheme than mere "liking" of an ad execution, "K" could be considered more "involving".

However, "K" has fewer brand-relevant connections and more irrelevant thoughts (indicating, perhaps, that emotional reactions are less easy to "contain" within "relevant bounds"). Should we still consider it the more "involving" commercial?

The point of these examples is not only to show the richness of the analysis that these "extra" code categories make possible, but also that commercials scoring high on different dimensions can make competing (and equally plausible) claims to being more "involving". These examples also show that restricting analyses to attribute assertion responses is clearly inadequate. Our semantic differential data, not presented here because of space limitations, shows how commercials frequently differ in the nuances of the affective and motivational responses they evoke.

We therefore believe that, because of the multidimensionality of the way in which people can get "involved" with commercials --in number of responses to the ad's attribute assertions, in the depth of the emotionality or evoked feelings, in the strength of the desire-to-buy that gets evoked, in the number of brand related "conscious bridging experiences" that occur, and others--there should be no hasty attempt to define one best measure of cognitive response for message response involvement.

The work reported briefly here is very preliminary and limited; in sample size, in dimensions tried, in variations in data collection approaches, in formal tests for reliability and validity, in statistical analysis. This was, after all, an exploratory pre-test. We believe, however, that our basic conclusion still holds good.

A Possible Theoretical Guide

Given the necessity of a theoretical scheme to guide the choice of a cognitive response measure of involvement, what should one choose? We have suggested elsewhere (Batra and Ray 1982) the equating of the degree of involvement with the degree of responding to the attribute assertions in the ad. This is consistent with the "brand" versus "non-brand" distinction mentioned by Mitchell, Russo and Gardner (1981); bears some resemblance to the "phonemic" versus "semantic" levels in "depth of processing" studies; is consistent with research on the reduction in the number of attribute dimensions used in decisions by people under high time pressure and distraction ("low opportunity"?) conditions (e.g., Wright 1974); and relies heavily on research results that seem to indicate crucial differences between two routes to the creation and modification of attitudes: one based on brand attributes, the other on ad execution and ad frequency.

While studies like Mitchell and Olson (1981), Gorn (1982), and others, lend credence to the hypothesis that advertising-induced brand preference formation in general is mediated through these two mechanisms, we have argued (Batra and Ray 1982) that the crucial difference between these two routes is that one consists of a conscious, voluntary attribute-based evaluation of the brand, while the other uses a less effortful, involuntary liking for the brand (which, in turn-, consists of the conditioned affect from a conscious liking for the ad execution, as well as a pre-conscious "mere exposure" liking effect of ad frequency). We have suggested therefore that "low involvement" brand preference formation has a larger "percentage contribution" of the involuntary, less effortful, liking component, as compared to "high involvement" message reception situations. Implications for advertising strategy, both creative and media, have then been drawn.

Such theoretical reasoning would lead to a useful operationalization of message response involvement that compares such involvement in terms of the degree of attribute-based response, as a percentage of total response. However, this is only one of many possible operationalizations, and our general point, again, is one about the non-finality of such operationalizations.

Next Steps

Two studies are now underway. One involves the further study of the measurement and determinants of message response involvement. The second would test the hypothesis, already supported in some other studies (e.g. Rothschild 1979; Ray et al. 1976; Webb and Ray 1979, and Wright 1973), that advertising effectiveness differs for situations with different quantity and quality of cognitive response.

The study on measurement of involvement will be done by collecting data (through schemes such as ours) on the depth and quality of cognitive response evoked by messages which systematically differ in the hypothesized antecedent factors of such involvement, and by then relating these antecedent factors to the different types of cognitive response evoked.

By varying the product class of the brand in the ad, the brand itself, and the message execution nature of the commercial, for consumers with different prior levels of awareness and usage, there should be enough variation in these "antecedent" variables of message response involvement. Data on the cognitive responses to these messages will then be analyzed.

We suggest an exploratory approach to the analysis of these data: coding the cognitive responses using a coding scheme such as that developed in the pre-study reported above, which uses about 12 micro-categories; and then relating these 12 (or fewer) types of responses to the antecedent variables mentioned, using canonical correlation.

It may thus he possible to find a linear combination of cognitive response variables which seems to match definitions of message response involvement, such as that of attribute-based cognitive response. And it should show which sorts of antecedent variables covary most with such response.

In addition to such an operationalization of message response involvement, and this identification of those antecedent variables that maximally covary with it, one could then use this operationalization as the dependent measure and the ability, motivation, and opportunity to respond as three independent variables (constructs, with multiple measures) in a causal modeling framework, to find the relative contributions of those three to message response involvement, thus helping us to understand it better


Batra, R. and Ray, M.L. (1982), "The Two Components of Affect and their Implications for Advertising Strategy," Paper presented at the Colloquium on Involvement, New York University, June.

Festinger, L. and Maccoby, N. (1964), "On Resistance to Persuasive Communications," Journal of Abnormal and Social Psychology, 68, 359-366.

Gorn, G.J. (1982), "The Effects of Music in Advertising on Choice Behavior: A Classical Conditioning Approach," Journal of Marketing, 46, 94-101.

Greenwald, A.G., Leavitt, C. and Obermiller, C. (1980), "What is Low Consumer Involvement?" In Gorn, G.J. and Goldberg, M.E. (Eds.) Proceedings or the Division 93 Program, 88th Annual Convention, American Psychological Association, pp. 66-74.

Houston, M.J. and Rothschild, M.L. (1977), "A Paradigm for Research on Consumer-Involvement," Unpublished paper, Graduate School of Business, University of Wisconsin, Madison.

Krugman, H.E. (1965), "The Impact of TV Advertising: Learning Without Involvement," Public Opinion Quarterly, 99, 349-356.

Krugman, H.E. (1966-67), "The Measurement of Advertising Involvement," Public Opinion Quarterly, 30, 583-596.

Mitchell, A.A. (1979), "Involvement: A Potentially Important Mediator of Consumer Behavior," In W.G. Wilkie (Ed.), Advances in Consumer Research (Vol. 6), Ann Arbor: Association for Consumer Research.

Mitchell, A.A. and Olson, V.P. (1981), "Are Product Attribute Beliefs the Only Mediator or Advertising Effects on Brand Attitude?" Journal of Marketing Research, 18, 318-339.

Mitchell, A.A., Russo, J.E. and Gardner, M. (1981), "Strategy Induced Low Involvement Processing of Advertising Messages", Working Paper, Graduate School of Industrial Administration, Carnegie-Mellon University, Pittsburgh, PA 15913.

Osterhouse, R.A. and Brock, T.C. (1970), "Distraction Increases Yielding to Propaganda by Inhibiting Counterarguing," Journal of Personality and Social Psychology, 15, 344-358.

Petty, R.E., Wells, G.L., and Brock. T.C. (1976), "Distraction Can Enhance or Reduce Yielding to Propaganda: Thought Disruption Versus Effort Justification," Journal of Personality and Social Psychology, 34, 874-884.

Pribram, K.H. (1980), "The Biology of Emotions and Other Feelings," In Emotion: Theory, Research and Experience (Vol. I.: Theories of Emotion), New York: Academic Press, pp. 245-269.

Ray, M.L. (1979), "Involvement and Other Variables Mediating Communication Effects As Opposed to Explaining All Consumer Behavior," In William L. Wilkie, (Ed.), Advances In Consumer Research, Vol. VI, Ann Arbor: Association for Consumer Research.

Ray, M.L., Ward, S. and Reed, J.B. (1976), "Pretesting or Anti-Drug Abuse and Education and Information Campaigns," In R. Ostman (Ed.), Communication Research and Drug Education, Beverly Hills, CA: Sage Publishing, 193-219.

Rothschild, M.L. (1979), "Involvement Strategies for High and Low Involvement Situations," In Maloney, J.C. and Silverman, B. (Eds.) Attitude Research Plays for High Stakes, Chicago: American Marketing Association.

Schlinger, M.J. (1979), "A Profile of Responses to Commercials," Journal of Advertising Research, 19, 37-46.

Webb, P. and Ray, M.L. (1979), "Effects of IV Clutter," Journal of Advertising Research, 19, 7-19.

Wells, W.D., Leavitt, C., and McConville, M. (1971), "A Reaction Profile for TV Commercials," Journal of Advertising Research, 11, 11-17.

Wright, P. (1973), "The Cognitive Processes Mediating Acceptance of Advertising," Journal of Marketing Research, 10, 53-62.

Wright, P. (1974), "The Harassed Decision Maker: Time Pressures, Distractions, and the Use of Evidence," Journal of Applied Psychology, 59, 555-561.

Wright, P. (1975), "Consumer Choice Strategies: Simplifying vs. Optimizing," Journal of Marketing Research, 19, 60-67.

Wright, P. (1980), "Message-evoked thoughts: Persuasion research Using Thought Verbalizations," Journal of Consumer Research, 7, 151-175.

Wright, P. (1981), "Cognitive Responses to Mass Media Advocacy," In Petty, R.E., Ostrom, T.M., and Brock, T.C. (Eds.) Cognitive Responses to Persuasion, Hillsdale, N.J.: Lawrence Erlbaum.

Wright, P. and Rip, P.D. (1981), "Retrospective Reports On the Causes of Decisions," Journal or Personality and Social Psychology, 40, 601-614.



Rajeev Batra, Stanford University
Michael L. Ray, Stanford University


NA - Advances in Consumer Research Volume 10 | 1983

Share Proceeding

Featured papers

See More


Red Flag! The Consequences of Alerting Consumers to Fake Reviews

Jared Joseph Watson, New York University, USA
Amna Kirmani, University of Maryland, USA

Read More


Donate Today or Give Tomorrow? Adding a Time Delay Increases Donation Amount but not Willingness to Donate

Emily Powell, New York University, USA
Minah Jung, New York University, USA
Joachim Vosgerau, Bocconi University, Italy
Eyal Pe'er, Bar-Ilan University

Read More


Exploring Consumers’ Technology Dreams and Nightmares: A Collage-Elicitation Study

Céline Del Bucchia, Audencia Business School
Cristel Russell, American University, USA
Burlat Claire, Audencia Business School

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.