Understanding the Likability/Involvement Interaction: the &Quot;Override&Quot; Model

ABSTRACT - Recent research indicates that source likability creates attitude change through a "route" that is more influential in "low (motivational) involvement" situations than in "high involvement" ones. An "override" model of this process is suggested. It argues that high levels of support and counter argumentation enhanced by the motivation, ability and opportunity to so respond to message arguments -- reduce both the production frequency of "affective responses" evoked by likable messages, and their covariance with dependent attitudes, thus leading to the observed interaction. An empirical test of the model is reported.


Rajeev Batra (1985) ,"Understanding the Likability/Involvement Interaction: the &Quot;Override&Quot; Model", in NA - Advances in Consumer Research Volume 12, eds. Elizabeth C. Hirschman and Moris B. Holbrook, Provo, UT : Association for Consumer Research, Pages: 362-367.

Advances in Consumer Research Volume 12, 1985      Pages 362-367


Rajeev Batra, Columbia University

[The data used in the empirical tests reported here are taken from a study sponsored by the Marketing Science Institute; the Institute's support is gratefully acknowledged.]


Recent research indicates that source likability creates attitude change through a "route" that is more influential in "low (motivational) involvement" situations than in "high involvement" ones. An "override" model of this process is suggested. It argues that high levels of support and counter argumentation enhanced by the motivation, ability and opportunity to so respond to message arguments -- reduce both the production frequency of "affective responses" evoked by likable messages, and their covariance with dependent attitudes, thus leading to the observed interaction. An empirical test of the model is reported.


There has been much interest in recent years in the interaction between "two alternative routes" of attitude change and the level of (motivational) "involvement" of the recipient. Various researchers have shown that source (ad execution) likability can also lead to attitude change, in addition to the traditionally-studied "message argumentation" route of attitude change; such source likability, however, appears to be the major "route" of attitude change only when the message recipient is (motivationally) "less involved" in the issue (Chaiken 1980; Gorn 1982).

While this interaction itself is by now commonly accepted, the process mechanism by which it occurs has not received much research attention. This paper suggests a process model of the referenced interaction, called the "override" model, and presents data from a test of the model.


Stated briefly, the "override" model suggested contains the following propositions:

1. Messages (e.g., advertising executions) which are "likable" generate "affective responses" in message recipients, in much the same way that message argumentation evokes the "cognitive responses" commonly studied (Wright 1973; Greenwald 1968).

2. The production frequency of the "cognitive responses" usually studied--in particular, support and counter arguments is the interactive outcome of the message recipient's "motivation," "ability," and "opportunity" to respond to the message in an attribute (i.e., argument) based fashion.

3. When the levels of such "antecedent" motivation, ability and opportunity (to respond to the message in an attribute-based fashion) are at high levels, the number of support and counter arguments produced is high. Such high production of support and counter arguments has two effects on the "affective responses" generated by likable message executions:

a. first, the production frequency of such "affective responses" is itself reduced.

b. second, the effect of the affective responses produced, on dependent attitudes, is reduced these "affective responses" are "overridden" by support and counter arguments in their influence on dependent measures of preference.

4. As a consequence, likable message executions influence brand attitudes far more when the antecedents of cognitive response production (e.g., motivational "involvement") are at low levels.

Because of space restrictions, this paper confines itself to presenting evidence pertinent to the third proposition. The first two propositions are discussed briefly below.

Evidence on the existence and role of "affective responses" has been recently presented by Batra and Ray (1984). Such responses consist of reports of moods and feelings evoked by the ad in the respondent, and are in addition to evaluations of the ad execution itself (such as source derogation or bolstering thoughts). Batra and Ray consider three such "affective response" categories: feelings of upbeat surgency, elation, vigor and activation (acronym "SEVA"); feelings of touching,heartwarming, tender, "social affection;" and feelings of quiet, relaxed, pleasant "deactivation." They show that such "affective responses" are produced by elements of the ad execution which make it more 'likable" and "affective." They also show that such responses have a statistically significant effect on brand attitudes and purchase intentions, especially in "low involvement" situations. In doing so, they extend the "heuristic" route of attitude change (Chaiken 1980) from the likability of the message execution to the (affective) "mediating responses" evoked by such executions.

The second proposition, on the antecedent factors for tie production frequency of support and counter arguments, is supported by evidence from various sources. Evidence on the linkage between "motivational involvement" and the production of these cognitive responses has been provided by Chaiken 1980, Wright 1273, and others. That on the "ability" antecedent is reviewed by Roberts and Maccoby 1973, who show that increased respondent knowledge about the issue leads to increased cognitive response production on message exposure. The effects of response "opportunity" are available in the distraction literature (e.g., Petty, Wells and Brock 1976) and in the effects of the number of arguments in the message (Calder, Insko and Yandell 1974; Chaiken 1980). (The "opportunity" effects of the number of message arguments are assumed to be linear in the range studied here, though they could conceivably be curvilinear, in that too few or too many arguments may inhibit attribute argumentation and instead promote reliance on "heuristic" execution cues.) For a review, see Wright 1981.

We turn now to the third proposition. In this proposition, we state that high production frequencies of support and counter argumentation--caused by high levels of antecedent motivation, ability and opportunity--lead to a suppression both of the production levels of affective responses and, in additions, of their effects on dependent attitudes.

It should be pointed out that Petty and Cacioppo 1979 have already shown that a source credibility x motivational involvement interaction similar to Chaiken's likability interaction is in fact, mediated by the differential production of cognitive responses under the two involvement conditions. They show (their Study 2) that the interaction occurs because subjects under high involvement generated more favorable thoughts and fewer counterarguments to strong than to the weak arguments; under low involvement, neither favorable thoughts nor counterarguments were affected by the argument quality manipulation (1979 p. 1922). They do not, however, examine the role of likability, "affective responses," or of respondent ability and message opportunity.

This paper thus suggests the extension of the results of Chaiken 1980 in the following ways. First, it argues that the effects of the "motivational involvement" antecedent of cognitive response production also apply to the "ability" and "opportunity" antecedents. Second, and more importantly, it suggests that such enhanced cognitive response production serves to (a) reduce the production frequency of "affective responses" evoked by likable message executions and (b) reduce the effects of ("overrides") such affective responses in terms of their influence on (their covariance with) dependent attitudes.

It should be noted that a similar model has recently been suggested by Greenwald and Leavitt 1984, called the "principle of higher level dominance." We turn now to the results of a study attempting to test the suggested model.


The objective of the study summarized below was to vary the level of support and counter argument production by a tactical manipulation of the antecedent "motivation," "ability," and "opportunity" to Produce such responses, and then to investigate the effects of such antecedent levels on the relationships between cognitive responses, affective responses, and dependent attitudes Two hypotheses were tested: one, that high levels of these antecedents would increase the production of support and counter argumentation but decrease that of the affective responses, and two, that these affective responses would have a reduced effect on (decreased covariance with) attitudes when cognitive response production was high.

Stimuli and Design

Forty thirty-second TV ads were selected to create five replications of eight factorial combinations: the motivational incentive to generate support and counter arguments, through the product category featured (high, low); differential knowledge-based ability to generate such responses, through the usage share of the particular brand featured (high, low); and differential opportunities for attribute-based viewer response, through the amount of attribute argumentation in the specific execution (high, low). The stimuli thus covered ten product categories, two brands each, two executions per brand.

The eight factorial combination were used to create two four-ad "blocks" using a half-factorial design. Since there were five replications of each of the eight factorial combinations, five replications were created of each of the two four-ad blocks. None of these within-subject "blocks" repeated a product category (hence also a brand or an execution). Through this design, then, the maximum tactical variance was created within each block on the three antecedent factors (motivation, ability, and opportunity) believed to influence cognitive response production: two of the four ads in each block were "high product category motivation," two were "high usage share" (ability) and two were "high attribute intensity" (opportunity).

It should be noted that the measures actually used in analysis for these antecedent factors were individual-specific and that the design was used for tactical purposes only (see Measures). This was done because while the classification of ads into the factor levels used the researchers' best judgments, the levels of these variables actually operative for individual respondents could conceivably have been very different from those assumed (e.g. levels of motivational involvement). It should also be noted that by varying the attribute-intensity of the particular ad execution, opportunity was also created for differential production of the "affective responses" of interest, since half the ads used were "rational" (attribute intensive) while the other half were "affective" (likable, emotional).


A total of 120 respondents were used, 12 per replicated block. These were housewives and working women, age 20 through 60, from the Palo Alto and nearby areas. While probably more educated and affluent than the average, their assignment to blocks was randomized. Each block was shown to the 12 respondents in four sessions of three respondents each.

Data collection for this study occurred in four phases. First, the ads were rated by judges on various dimensions (see Measures). Second, the women were contacted by telephone approximately one week before their experimental session and were asked embedded pre-exposure questions. The third phase was the experimental session itself. Here, after a first screening of the ad to more nearly equalize prior ad familiarity, the test ads were shown again (in a randomized and rotated fashion). Retrospective verbal protocols were administered to collect cognitive and affective responses (following the Batra and Ray 1984 procedure) and measures were taken of the motivational cognitive response antecedent ("product category involvement"), the ability (knowledge) antecedent, various dependent measures (brand familiarity, attitudes, intentions), and prior brand usage. In the fourth phase of data collection, delayed measures of brand attitudes were taken by telephone a week later. Only some of the data collected were used in the analyses reported in this paper. The measures of interest are detailed below.


The individual-specific "motivational" antecedent of cognitive response production was assessed by asking each respondent how important it was to her, when buying any brand in each of the ten (test and filler) product categories listed, that she bought exactly the brand she did. The response was coded on a 7-point "most important" to "least important scale. Multiple measures were used for the "ability and "opportunity" antecedent constructs. For the "ability" construct, the first was a measure of the number of brands the respondent was aware of (unaided) in the product category, collected during the pre-exposure call. The second was an individual-specific measure on category knowledgeability, which asked the respondent for a self-rating of her knowledge about which features one might look at, in choosing among different brands, with such knowledge coming not only from usage but also from magazines, ads, and friends (5-point "very knowledgeable" to "very unknowledgeable" scale). The third was a measure of the total number of brands ever used by the respondent in that category from the many listed.

The individual-specific measure for the "opportunity" construct was the respondent's rating of the ad being informative or not (7-point scale, "had no information" to "had a lot"). In addition, three judges rated each ad on the number of attributes mentioned in the ad (second measure of "opportunity") and assigned a score on how "rational" the ad execution was (third "opportunity" measure). This last "R-score" was based on a checklist of items such as the demonstration of attributes, the use of a high expertise spokesperson, a "comparative" visual execution, statements regarding component attributes, etc. Two-judge correlations among pairs of the three judges ranged between 70-90% for these scores, and mean ratings across the three judges were therefore used.

While the analysis using antecedent levels (above and below the mean) reported below uses all of these multiple measures, further analysis was conducted to assess convergent validity and measure reliability for these antecedent constructs. This is not reported here, for reasons of space, and may be obtained from the author.

The production frequency of cognitive and affective responses was based, for every individual observation, on the thoughts and feelings reported in the retrospective verbal protocols. The reported thoughts and feelings were classified into support arguments, counter arguments, execution discounting/derogation, execution bolstering, S.E.V.A. (surgency, elation, vigor, activation) feelings, "social affection" (heartwarming, tender) feelings, "deactivation" (relaxing, soothing) feelings, and distraction/irrelevant thoughts. Descriptions and reliability estimates for these categories are available in Batra and Ray 1984, from which this coding scheme was adopted.

Dependent (immediate) brand attitudes were assessed through semantic differential items of "useful-useless," "important-unimportant," "pleasant-unpleasant," and "nice-awful." The mean of these four items was used in analysis.


The analysis reported below examines two issues. First, we look at the relationship between the antecedents of cognitive response (support and counter argument) production--the motivation, ability and opportunity factors--and the production frequency of those cognitive responses and of the "affective responses" studied by Batra and Ray 1984. Here, it is hypothesized that h_-. levels of these antecedents will tend to increase the production frequency of support and counter arguments and suppress that of the three affective responses studied. We then turn to examining the reasons for the relative influence of affective responses on brand attitudes in situations where the cognitive response antecedent factors are at high and low levels. Here, we hypothesize that their reduced effect in "high antecedent" levels is their reduced covariance with brand attitudes in situations where support and counter argument production is high.

The effects of the antecedent "motivation," "ability" and "opportunity" variables on the production frequency of cognitive and affective responses are apparent through the bivariate correlations presented in Table 1.

It can be seen from the table, first, that the motivation measure increases the production of support arguments (and of distractor thoughts), leaving affective response production unchanged. When significant, some of the ability measures increase the production of counter arguments, while decreasing the production of SEVA and social affection feelings. The opportunity measures increase support and counter argument production (with one exception) and decrease the frequency of SEVA, deactivation, and social affection feelings-the one exception is the number of attribute arguments in the ad, which increases counter argument production but suppresses support argument production.

It should be noted that while the different antecedent measures all increase the production of either support or counter arguments, they do differ in which of these two response categories they influence. No theoretical explanation is available for such differential effects on support and counter arguments, and the inconsistency in these effects clearly weakens the support for the relationships hypothesized. In general, however, these bivariate correlations offer some support for the hypothesis that increased levels of the antecedent measures manifest themselves in increased support and counter argument production and in decreased affective response production. (Supplementary multivariate analysis, not reported here because of space limitations, yielded the same conclusions.)

As mentioned, Batra and Ray 1984 have shown that the three affective responses studied here are evoked by likable ad executions,and that these responses have the strength to influence brand attitudes and purchase intentions. We now have data suggesting that in those situations when the motivation, ability and opportunity to produce support and counter argumentation are high, the number of such attribute-based responses rises selectively but significantly, while the production frequency of these affective responses is (usually) suppressed.

The "override" model now hypothesizes that the effect of such affective responses, on brand attitudes, is reduced when the (enhanced) number of support and counter arguments is high: these cognitive responses then "override" the influence of the affective responses generated.

This hypothesis was tested, first, through OLS multiple regression analyses summarized in Table 2, which reports the beta coefficients of the net valence of four groups of mediating responses in predicting post-exposure attitudes. Each regression run is conducted within two subsamples of each antecedent operationalization: cases with antecedent values less than or equal to the mean ("low") and those above the mean ("high"). Use of these "post-hoc" samples obviously makes impossible the causality inferences allowed by use of the experimental cells, but was felt to be appropriate given the individual-specific levels of the antecedent variables, as mentioned earlier in the 'Design' section.

The mediating responses are clustered, yielding net positive valence per cluster, in the four groups of (a) support arguments minus counter arguments (b) source bolstering minus source discounting (c)SEVA plus deactivation plus social affection responses (all positively valenced) and (d) distractor thoughts.

While particular runs differ, it can be seen that for most motivation and ability operationalizations (major exceptions: the opportunity variables) the net valence of support and counter argumentation is much more frequently a significant attitudinal predictor (at p < .05) in the "high" sub-samples than in the "low." Conversely, the sum of the three affective responses (all positively valenced) are significant predictors far more often in the "low" sub-samples than in the "high." (Similar, but slightly weaker, results hold for a purchase intentions criterion measure.)





Note that no statement is being made here about the statistical significance of the difference in coefficient values; we are noting merely the obvious difference in the number of times a particular cluster is significant or not. Note also that the difference in results for the "opportunity" measures is theoretically unexplainable, and qualifies the empirical support for the hypothesis

Given that the "affective responses" appear to significantly influence brand attitudes less frequently in high "antecedent" conditions, it now becomes crucial to determine whether this occurs because affective responses simply have reduced variance in the "high" sub-samples, or because (despite adequately high variance) the covariance between them and attitudes goes down in such "high" antecedent sub-samples.

Table 3 reveals that, for these motivation and ability operationalizations, the diminished effects of the three affective responses on attitudes in the high sub-samples are not, in most cases, the consequence of reduced variance, though mean levels are sometimes lower than in the "low" sub-samples (cf. Table 1 results earlier). F-tests for equality of sample variances show that in only two of seven cases (R-score of at, and ratings of ad informativeness) are the variances significantly unequal at a .05 level of significance. Clearly, therefore, while variances usually remain about the same, the covariances between these affective responses and dependent brand attitudes or purchase intentions reduce dramatically in the high sub-samples, compared to the low. In other words, when the total production of support and counter arguments goes up, because of high levels of the motivation and ability antecedents, the covariances between the sum total of three affective responses, and dependent attitudes, goes down. (Since this analysis did not use experimental randomization, however, rival hypotheses of self-selection biases cannot be entirely ruled out.)



Hence qualified support is found for a process model of the likability/"involvement" interaction which states that high production of support and counter arguments "override," through their net valence, the effect of affective responses on brand attitudes. (The support is qualified by the inconsistency of effects between support and counter arguments in Table 1, and by the "deviant" results 'or the "opportunity" operationalizations in Table 2.) Such affective responses reflect, and are induced by, ad execution likability (Batra and Ray 1984); the high production of support and counter argumentation arises because the antecedent conditions have high levels (Tables 1 and 3). This "override" model thus helps us understand the results reported earlier, where high levels of the motivational antecedent nullify the attitudinal effects of likable sources (ad executions) observed in low motivational involvement conditions (Chaiken 1980; Gorn 1982), and extend them to the ability antecedent as well.


The results of the tests reported are hardly unequivocal, but they do suggest a mechanism for an important contingency in the recent work on "two alternative routes" of attitude change. While Petty and Cacioppo 1979 showed that the moderating effects of "motivational involvement" occur through the differential production of support and counter arguments, we now see that the effect is more general, in that such differential production can also arise from other antecedents. More importantly, we see that the mechanism at work extends to the production of the affective responses studied by Batra and Ray 1984, and there is some evidence that enhanced cognitive response production may "override" the generation and use. by the message recipient, of the affective responses to message likability. This suggests that message recipients may use such "likability cues" (Chaiken 1980) to form attitudes only when attribute response does not or cannot occur.

Importantly, however, the results presented do not support this "override" model uniformly, in that they differ across the "ability" and "opportunity" antecedent constructs and for different measures within them. Further work is clearly needed to develop measures of these constructs which are more reliable and more valid than those used here, and to then provide a test of this model that is less subject to biases from measurement error. Future research in this area might also examine the possibly non-linear relationship of the opportunity antecedent to cognitive response production, as well as the differences between support and counter arguments in such relationships.


Batra, R. and Ray, M.L. (1984), "Affective Responses Mediating Acceptance of Advertising," Working Paper, Graduate School of Business, Columbia University.

Calder, B.J., C.A. Insko and B. Yandell (1974), "The Relation of Cognitive and Memorial Processes to Persuasion in a Simulated Jury Trial," Journal of Applied Social Psychology, 4, 62-93.

Chaiken, S. (1980), "Heuristic versus Systematic Information Processing and the Use of Source versus Message Cues in Persuasion," Journal of Personality and Social Psychology, 39. 752-766.

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. (1968), "Cognitive Learning, Cognitive Response to Persuasion, and Attitude Change," in A.G. Greenwald, T.C. Brock, and T.C. Ostrom (Eds.), Psychological Foundations of Attitudes, New York: Academic Press.

Greenwald, A.G. and C. Leavitt, (1984), "Audience Involvement in Advertising: Four Levels," Journal of Consumer Research, 11, 581-592.

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

Petty, R.E. and J.T. Cacioppo (1979), "Issue Involvement Can Increase or Decrease Persuasion by Enhancing Message-Relevant Cognitive Responses, Journal of Personality and Social Psychology, 37, 1915-1926.

Petty, R.E., J.T. Cacioppo and R. Goldman (1981), "Personal Involvement as a Determinant of Argument-based Persuasion," Journal of Personality and Social Psychology, 41, 847-855.

Roberts, D.F. and Maccoby, N. (1973), "Information Processing and Persuasion: Counterarguing Behavior," in P. Clarke (Ed.), New Models for Communication Research, Beverly Hills, CA: Sage Publications, 269-307.

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

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



Rajeev Batra, Columbia University


NA - Advances in Consumer Research Volume 12 | 1985

Share Proceeding

Featured papers

See More


Teaching Old Dog New Tricks… and Old Bottles New Jeans. The Role of Implicit Theories in the Evaluation of Recycled Products

Alessandro Biraglia, University of Leeds
J. Josko Brakus, University of Leeds
Lucia Mannetti, Sapienza University of Rome
Ambra Brizi, Sapienza University of Rome

Read More


Consumer Response to Innovations: The Differential Effects of Focused and Defocused Attention on Perceived Novelty, Usefulness and Symbolism

Katarina Hellén, Univeristy of Vaasa
Maria Sääksjärvi, Delft University of Technology, The Netherlands

Read More


Let's Get Together and Make a Difference: Experiencing a Community in Donation-Based Crowdfunding

Danit Ein-Gar, Tel Aviv University, Israel

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.