Discussion Paper on Message Variables and Persuasion

Ruby Roy Dholakia, University of Rhode Island
Steve Hoch, Northwestern University
ABSTRACT - The five papers in this session focus upon selected message variables and their effect on audience responses. Klebba and Unger cake a look at a well known source for Chrysler cars and relate perception of his credibility and its effect to the audience's possession of positive and negative information. The focus of the Percy and Rossiter paper is on print advertising; they examine the effects of picture size and color on brand attitude response. Belch relates one- and two-sided messages to message modality while Zirkhan and Martin attempt to predict advertising response from cognitive complexity and individual perception of advertising structure. The fifth paper by Sullivan and O'Connor examine the relation between behavioral intention and audience response to creative aspects of televised messages for public services.
[ to cite ]:
Ruby Roy Dholakia and Steve Hoch (1983) ,"Discussion Paper on Message Variables and Persuasion", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 36-39.

Advances in Consumer Research Volume 10, 1983      Pages 36-39


Ruby Roy Dholakia, University of Rhode Island

Steve Hoch, Northwestern University

[The discussion of papers 1, 4 and 5 is by the first author and papers 2 and 3 by the second author.]


The five papers in this session focus upon selected message variables and their effect on audience responses. Klebba and Unger cake a look at a well known source for Chrysler cars and relate perception of his credibility and its effect to the audience's possession of positive and negative information. The focus of the Percy and Rossiter paper is on print advertising; they examine the effects of picture size and color on brand attitude response. Belch relates one- and two-sided messages to message modality while Zirkhan and Martin attempt to predict advertising response from cognitive complexity and individual perception of advertising structure. The fifth paper by Sullivan and O'Connor examine the relation between behavioral intention and audience response to creative aspects of televised messages for public services.

The Impact of Negative and Positive Information on Source Credibility in a Field Setting

The paper deals with a topic of some practical importance. There appears to be a belief among advertisers that sources with negative attributes will not be as persuasive as those without these disadvantages. This can be seen in the dropping of Billy Jean King and Anita Bryant as product endorsers. However, as well known people are used as endorsers and spokespersons, it may be increasingly difficult to select and continue to use someone with only positive attributes. It is interesting, therefore, to examine whether negative information about a source leads to lower acceptance of a message.

By using Lee Iacoca, the authors focus on a source currently being used in Chrysler commercials. It increases the external validity of the research design; however, it does create problems for internal validity of the study. The student subjects are categorized into 4 groups based on self reports about their knowledge of Iacoca's earlier association with: only the Mustang (positive), only the Pinto (negative), both cars (positive and negative), and neither car (no information). While it is heartening to see that there is some consistency between the information state and source perception, it is not clear whether the rating of the source is a function of the information about the source or vice versa. This creates a specific problem in relating perceptions of the source to the perceptions of Chrysler cars.

The authors find that the affective dimension of source credibility to be positively and significantly related to perceptions of product quality and safety while the cognitive dimension is positively and significantly related only to perceived product quality. Since the affective component of the credibility dimension differed significantly only between the Positive and Both groups, it may lead one to conclude that it is better not to use a source with some negative attributes. However, the stud found no significant differences on the affective dimension of source credibility between the Positive and Negative groups even though they supposedly had diametrically opposite information. It is likely, therefore, that an individual with only positive information about Iacoca rated Chrysler cars similarly to an individual with only negative information about the source. The data seems to imply that it is better to provide only positive, only negative or even no information about a source rather than mixed (both positive and negative) information.

Some of the data analysis obscures the relationships between possession of information and perception of source credibility. For example, testing of hypothesis 1 leads the authors to conclude that possession of negative information does not lead to differences in perception of the source on the-affective dimension but does create differences on the cognitive dimension. In fact. they see it leading to a higher rating of Iacoca on the cognitive dimension. This is both counter-intuitive and contrary to the data. The collapsing of the four groups into two (know of association with Pinto and don't know) leads to the no information cell (n = 130) being averaged with the positive only cell (n = 47) in the don't know group with the no information group dominating the mean rating because of its larger size.

The paper raises some interesting issues. However, the field setting and self selection of groups clouds the analysis. Perhaps an experimental design can more fruitfully examine the issues. Such a design may allow experimental manipulation of source perceptions through differential information conditions and examination of their influence on persuasion. Such a study could also address the issue of information integration (Anderson 1971) impression formation (Richey, et. al. 1975) and their relationship to negative and positive information. Furthermore, such an approach could not only look at source perceptions and attitudes but also behavioral constructs since there is some evidence l:o believe that a less credible source may be more effective than a higher credibility source under certain situations (e.g., Dholakia and Sternthal 1976; Powell 1965).

Effects of Picture Size and Color on Brand Attitude Responses in Print Advertising

Percy and Rossiter provide some interesting findings on the effects of visual content upon persuasion. Although the results are clearly exploratory and not completely straightforward, it is interesting that such seemingly minor manipulations, picture size and color, produced any differences in attitude at all. The differences in the stimuli were clearly noticeable, but why they should have influenced attitudes is not clear. There are two issues that need examination: 1) how did picture size and color influence subjects' perception of Esprit in the current study? and 2) what are the possible cognitive processes underlying the influence of visual content upon persuasion?

The picture size effect is quite mysterious. Only one cognitive belief (naturalness) is affected at the individual belief level. When an additive scale is formed. differences in overall belief are still not significant. It is only when the beliefs are weighted by subject supplied importance ratings that the effect of picture size becomes somewhat significant. What can explain these results? It appears that there is some sort of an "interaction" between a subject's belief and a subject's importance rating. In other words, a subject that sees the Esprit brand as exceedingly natural also sees the natural attribute as very important. Two things could be going on here. The belief by importance "interaction" could be nothing more than a statistical quirk in the scaling process. More interestingly, however, the picture size manipulation could somehow be affecting the importance weights that subJects assign to the different product attributes. Maybe a larger image of the product, where it is easy to discriminate the frosty bottle and glass and the mineral spring effervescence, encourages subjects to attach greater weight to the natural attribute. It might be worthwhile to reanalyze the data using attribute importance ratings as dependent variables. There is no reason to suppose that importance weights are fixed, unalterable by the visual stimulus

Percy and Rossiter see the color manipulation as influencing emotional response to the-product without necessarily altering belief structure. Another possibility is that the study was not really measuring response to the product; instead, subjects were responding to the "rough" artwork used as the stimuli. Virtually ali color ads are "better looking" than black and white versions, but whether the color effect reported here is anything more than a momentary color "high" is not clear. However, the tact that color has a mild effect upon intention argues that the emotional response might have carried over from the ad to the product itself.

Why might two seemingly minor visual elements like picture size and color have persuasive impact? To answer that question requires an understanding of the underlying cognitive processes. Percy and Rossiter mention three related explanations: increased attention, "image carryover", and "consumption imagery". All of these explanations imply that differences in visual content encourage people to allocate processing resources in particular ways. Certain thoughts are aroused while others are suppressed. A larger picture might focus attention upon the physical attributes of the brand (refreshing, cold, tingly, erotic). At the same time the large image might suppress other thoughts not directly related to the product (the noveau-riche product name Esprit, the annoyance at being harassed by a market researcher in a mall). There are a couple of ways to test whether this is what is realLy going on. First is the cognitive response approach where direct measures of mediating thoughts are collected. Analysis or the cognitive responses might reveal whether manipulations in visual content alter the content of subjects' thoughts. Second it visual content is operating at the level or cognitive resource reallocation (directing attention to the product rather than other thoughts that the subject might have), then it should be easy to find a case where the persuasive effect of visual content is negative rather than positive. A ready example might be something like cod liver oil or Pepto-Bismol, where increased attention to the product itself would probably arouse quite a few negative associations.

One final comment concerns the authors' statement that picture size seems to influence "cognitive-perceptual processing" while color influences "affective processing". Given the equivocal nature of the results, this conclusion seems to overstep the bounds of the data. First, Percy and Rossiter note, it is not clear why picture size did not influence the affect attitude measure since it had done so in a previous study. Moreover what is at purely affective response? Is it a warm feeling? If so, then measurement of psychophysiological activity would seem appropriate. The idea that the effects of certain visual elements in advertisements night not be under direct cognitive control is intriguing, but needs to be studied in a more direct manner. While there can be no argument with taking multiple dependent measures, one can question the measurement of purely affective responses with attitude scales. No matter how you ask it, the subject cannot avoid some form of cognitive activity when deciding where to mark the "affective" attitude scale.

The Effects of Message Modality on One- and Two-Sided Advertising Messages

At an emotional level, we want to see two-sided appeals work: if an advertiser is "Man enough to admit it", then he/she cannot be all bad. While there are several very rational arguments to support the efficacy of two sided appeals, convincing empirical demonstrations of their superiority have been rare. Instead of discussing why Belch found no differences between one- and two- sided appeals, this discussion will address two related questions: 1) why don't two-sided appeals work more often, and 9) is there a set of guidelines that will increase the probability of success of two-sided, "man enough to admit it" appeals?

What are the necessary conditions for successful two-sided appeals? Belch's introductory review suggests several factors. First is education. The audience has to be educated enough to "get it". One might expect that two-sided appeals would not work as well with children who would see the negative information it face value, rather than as the conciliatory gesture it was meant to be. Second, two-sided appeals work best when the audience is negatively predisposed to the position advocated by the appeal. The Haas and Linder (1979) strategy of mentioning negative information upfront could reduce the counterargumentation tendencies of the hostile audience. One might regard two-sided appeals as the strategy of last resort when trying to persuade an audience that dislikes the advertised product. Third, the audience must be aware of counterarguments to the appeal. If the audience is not aware of counter arguments, then a two-sided appeal could simply provide a new set of defenses to fend off the persuasive attempt. Therefore, using a two-sided appeal for a new or unfamiliar product might be ill-advised because it is less likely that counterarguments would be available to the audience. Fourth, the audience must have adequate time to process the two-sided appeal. If time is limited, then there would be less opportunity for counterargumentation which in turn would reduce the relative advantage of the two-sided appeal. Also, processing limitations or lack of attention might preclude subjects making the necessary inferences that increase the credibility of the two-sided source.

The fifth condition is that the two-sided appeal must contain some kind of refutation, explicit or implicit, of the negative information included in the appeal. It is clear what constitutes an explicit refutation--a direct frontal assault on the opposition's arguments. The explicit refutation seems best suited to two-sided appeals within the context or direct comparative advertising. An implicit refutation is a somewhat more elusive animal. The basic idea is that while the advertiser admits inferiority on some attributes, the audience makes one of two inferences: 1) the negative attributes really don't matter, or 2) the so-called negative attributes are actually positive attributes. In two of the studies mentioned by Belch (Etgar and Goodwin, 1982; Swinyard, 1981), the negative attributes focused upon the trade-off between price and quality. For instance, the lack of full-service grocery shopping reinforced the positive attribute that the advertiser offered the lowest prices in town, while a high priced beer reinforced the notion that the advertised product was indeed the highest quality premium beer. This tactic has long been recognized in personal selling and is called the "Boomerang" technique--turning a negative point or sales objection into a product advantage. (Prospect: "I'm too busy to see you." Salesman: "Sir, I've come to see you because you're busy. If you weren't, then you wouldn't need our product.") The "boomerang" refutation seems the most promising. Consider the case where the negative attributes are not very important to the audience but are still a real product liability (rather than a hidden asset). In this case the advertiser must hope that the increased source credibility and decreased counterargumentation afforded by a two-sided appeal will outweigh the negative impact from the overt recognition of true product deficits. However, the successful implicit refutation is tricky because if it is too subtle the audience will miss it and be left with only the negative information.

So why don't two-sided appeals work more often? The above discussion suggests that problems occur because a complex set of pre-conditions must be satisfied in order for the two-sided appeal to be more persuasive than its one-sided counterpart. If the audience is intelligent, negatively predisposed to the appeal, is aware of counterarguments to the appeal, and has adequate processing time, then the two-sided appeal might be more effective. These conditions are necessary but probably not sufficient. The appeal also must provide an implicit refutation of the negative arguments or at the very least admit inferiority only on attributes that the audience feels are unimportant. When will all these conditions be met? The answer is not very often. The prevalence of heterogeneous audiences could mean that the two-sided appeal is self-defeating. While the two-sided appeal might increase purchase probability for those people who don't currently like or buy the product (those negatively predisposed), it might also decrease purchase probability for those people who currently like and buy the product by providing a new arsenal of counterarguments to justify not continuing to buy. Moreover the successful use of implicit refutation seems more a creative art than an easily applied change in copy. In summary, these comments constitute a warning about the potential pitfalls in using a persuasion technique that has rational underpinnings but requires a delicate balance between standing conditions.

Message Characteristics and Audience Characteristics: Predictors of Advertising Response

A recent episode on the 'Hi and Lois' cartoon series shows the housewife's selection of detergent brands based entirely on her liking for the advertisements themselves. The study reported in this paper attempts to predict such a response. The emphasis is on aesthetic response which is defined as 'how people feel after seeing an ad!

The two independent variables used in the study are measures of cognitive complexity and message complexity. Both these variables are operationalized at the individual level specifically for the product chosen - calculators. Cognitive complexity is made a function of the number of attributes listed by a respondent and their categorization into similarity groups. The procedure, therefore, controls for response bias to a degree by imposing both criteria to represent cognitive complexity by the subject's R score.

The second variable deals with message complexity. This is measured by the cloze test which looks at 'an individual's ability to provide deleted words in a passage of writing.' The authors imply a positive relationship between cognitive complexity and the perceived pleasantness of an ad and state a non-linear (inverted U) relationship b.tween perceived message complexity and the perceived pleasantness of an ad. Furthermore, they feel an ad must have a minimum level of complexity to be employed (i.e., it cannot be too simple) and this level will depend on an individual's level of cognitive complexity. Therefore, an interactive relationship is posited between message complexity and cognitive complexity influencing the pleasantness rating of an ad.

The analysis, however, focuses only on the relationship between the two independent variables individually and ad response. Discriminant analysis finds the predictors to be statistically significant. The discriminant function is then used to predict ad responses of a holdout sample and is found to perform slightly above the chance level (59.2% correctly classified).

The statistical analysis fails to consider the important interactive relationship between the two predictor variables. This relationship is of greater theoretical and practical relevance for the design of marketing communications. Mischel (1973) postulated that cognitive and behavioral construction competencies of an individual are affected by individual variables including cognitive complexity. One would expect, therefore, perceived ad complexity as measured by the R-score to be related to the individual level of. cognitive complexity. In terms of practical implications, implicit in the research strategy is a belief that-message complexity can be manipulated to achieve optimal pleasantness, and only the level of audience's cognitive complexity must be known to execute this message variable. Since the analysis does not examine the relationship between cognitive complexity and perceived message complexity, this particular study-cannot shed further insight into the question. It remains, however, an issue for further research.

The specific statistical tool--discriminant analysis-appears also to be limited in usefulness for the analysis undertaken in this research. Given the availability of a scaled response, perhaps regression analysis would be more appropriate to help explain degrees of liking for an ad. The explanatory variable would be, of course, the interaction between cognitive complexity and message complexity.

Search for a Relationship Between Viewer Responses to Creative Aspects of Televised Messages and Behavioral Intention

This paper's focus is on two aspects of communication response: judgments about creative aspects of an ad and intention to perform specific behaviors emanating from the ad message. Hopefully, such research can identify features of ads that affect audience response, including behavioral intention.

The authors use public service announcements (PSA) as message stimuli which are rated on an adjectival list developed by Leavitt (1975). Factor analyses of the responses reveal a factor structure similar to earlier investigations and the authors comment favorably on the reliability and validity of the adjectival response list. However, when the factors are used to predict behavioral intentions no significant relationships are found between the variables. This is despite the two exposures to each message in the study.

The research effort raises more questions than it answers and the authors list some of them for further investigations. A problem that arises at the design of any research study is one of expected relationships between variables; in this instance, the relationship between responses to creative aspects and behavioral intent. It is likely that other variables such as attitude towards the issue intervene between the attitude towards the ad and behavioral intentions. The strength of relationships between these variables are also likely to he affected by individual and situational variables such as product familiarity, initial attitudes, etc., that determine cognitive processes (Sternthal and Craig, 1989). This becomes especially important for the types of behaviors examined in this study--wearing of seat belts, observing of 55 m.p.h. limit--which have been the topic of both persuasive communication and political commentary in recent years. It is highly unlikely that a message (even with two exposures) can directly affect behavioral intentions through responses to its creative aspects alone. We need to hypothesize and measure other intervening variables, including attitude towards the issue and towards the behavioral actions.

The study raises an interesting issue concerning the conceptualization of behavior and behavioral intent. In most marketing studies, the behavioral intent refers to a single act such as trial or a single purchase. Behaviors elicited by PSAs are usually quite different. In this study, it refers to a series of acts; not just putting on the seat belt once or observing the speed limit once but through repetitive behavior over time. To audience members the behavioral intent scale measures a different degree of commitment. It is not surprising, therefore, that a lack of relationship between responses to creative aspects of an ad and behavioral intent is found. The stage is set for more effective conceptualization and measurement of relevant variables that link immediate exposure effects and future behavior.


The five papers discussed in this section examined disparate elements of the message variable. It is difficult to draw connections between them and no such effort will be undertaken. Most of them pose research problems that require further investigations before the specific studies can meaningfully explain a relationship between the selected message variable and persuasion.


Anderson, N. H. (19713, "Integration Theory and Attitude Change," Psychological Review, 78, 3, 171-206.

Dholakia, R. R. and Sternthal, 8. (1976), "Highly Credible Sources: Persuasive Facilitators or Persuasive Liabilities?", Journal of Consumer Research, 3, 923-32.

Etgar, M. and Goodwin, S. A. (1982), "One-Sided Versus Two-Sided Comparative Message Appeals for New Brand Introductions," Journal of Consumer Research, 8, 460-464.

Haas, R. G. and Linder, D. (1979), "Counterargument Availability and the Effects of Message Structure on Persuasion," Journal of Personality and Social Psychology, 23, 219-233.

Leavitt, C. (1975), "Theory as a Bridge Between Description and Evaluation of Persuasion," in Advances in Consumer Research, Vol. 2, ed. M. J. Schlinger, Chicago: Association for Consumer Research. 607-13.

Mischel, W. (1973), "Toward A Cognitive Social Learning Reconceptualization of Personality," Psychological Review, 80, 4, 959-83.

Powell, F. (1965), "Source Credibility and Behavioral Compliance as Determinants of Attitude Change," Journal of Personality and Social Psychology, 9, 669-76.

Richey, Marjorie H., Koenigs, Robert J., Richey, Harold W. and Fortin, Richard (1975), "Negative Salience in Impressions of Character: Effects of Unequal Proportions of Positive and Negative Information," Journal of Social Psychology, 97, 933-941.

Sternthal B. and Craig, C. S. (1982), Consumer Behavior: An Information Processing Perspective, Englewood Cliffs, NJ: Prentice-Hall.