The Affective and Cognitive Dimensions of Pictures in Advertising: an Extension of Mitchell &Amp; Olson

Arthur E. Heimbach, University of Washington
Richard F. Yalch, University of Washington
ABSTRACT - Research by Mitchell and Olson (1981) was reviewed to develop an experiment examining the effect of pictorial and verbal information on product attitudes. Subjects were exposed to six ads. Four were pictures of animals representing different levels of likableness and softness and two were different verbal claims of softness. Beliefs about and attitudes toward the product and attitudes toward the ad were assessed. The results differ from Mitchell and Olson in that the pictorial ads did not yield more favorable attitudes than the verbal ads and the attitude toward the ad measure was not a consistently useful predictor of brand attitudes.
[ to cite ]:
Arthur E. Heimbach and Richard F. Yalch (1988) ,"The Affective and Cognitive Dimensions of Pictures in Advertising: an Extension of Mitchell &Amp; Olson", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 178-183.

Advances in Consumer Research Volume 15, 1988      Pages 178-183


Arthur E. Heimbach, University of Washington

Richard F. Yalch, University of Washington


Research by Mitchell and Olson (1981) was reviewed to develop an experiment examining the effect of pictorial and verbal information on product attitudes. Subjects were exposed to six ads. Four were pictures of animals representing different levels of likableness and softness and two were different verbal claims of softness. Beliefs about and attitudes toward the product and attitudes toward the ad were assessed. The results differ from Mitchell and Olson in that the pictorial ads did not yield more favorable attitudes than the verbal ads and the attitude toward the ad measure was not a consistently useful predictor of brand attitudes.


Although originally directed at exploring the Fishbein model's specification that attitudes are mediated solely by attribute beliefs, Mitchell and Olson's (1981) article also raised the interesting question of how similar information presented visually or verbally may affect information processing. Specifically, they found that consumers evaluated brands of facial tissue differently depending on whether the brand was associated with one of three pictures or the word "soft." Further, the different evaluations of the brands could not be explained by cognitive beliefs on five attributes including softness that were assessed. Hence, Mitchell and Olson concluded that advertisements could alter attitudes in ways other than by belief changes and that much of these influences could be measured as an attitude toward the advertisement. This important conclusion helped launch the investigation of attitudes toward an advertisement as an important determinant of attitudes toward the advertised product (cf., Gardner, 1985 for a review). It also was instrumental in starting a stream of research examining the relative persuasive powers of pictures and words (cf. Kisielius & Sternthal, 1986 for a review).

The purpose of this paper is to reexamine the attitude toward the advertisement and attitude toward the advertised product relationship by reviewing Mitchell and Olson's study and identifying limitations inherent in their design. On the basis of this review, three major issues are identified which we feel limit generalization of their conclusion regarding Aad and Aproduct Next, a revised procedure is suggested to address these limitations. An experiment incorporating these changes is presented and the results are summarized. Finally, issues for future research on the relationship between attitude toward the ad and attitude toward the advertised product are offered.

Before beginning, we would like to express our admiration and appreciation of Mitchell and Olson's insight in developing a revolutionary view of attitude change and a clever research design. However, all research can be improved and this may contribute to our understanding of the phenomenon investigated (see Mitchell, 1984 for another extension of Mitchell and Olson, 1981). To this end we reviewed and modified Mitchell and Olson's work much as they sought to extend Fishbein's attitude theory (p. 320).


Mitchell & Olson were interested in whether two advertising message factors, amount of repetition (none versus some) and mode of presentation (verbal versus pictorial) might result in different attitudes toward a product without noticeable differences in beliefs toward the advertised product. Using a within subjects research design, each subject was exposed to four ad treatments. These included a photograph of a kitten, a photograph of a sunset, a card with an explicit verbal claim of softness and a picture of an abstract painting. Each message was associated with a different brand of facial tissue, identified by the letters I, J, L and R. Effects of the message were assessed using Ahtola's (1975) vector model. Accordingly, beliefs about the brands on various levels of five attributes (softness, strength, attractiveness, economy and color), and each subject's evaluation of several levels of each attribute on a good-bad scale were determined. Attitudes toward each advertisement and each picture were also assessed.

Mitchell and Olson analyzed their data using a one-way ANOVA, which assumes that the four ads differed in many ways. However, their hypotheses and interpretations suggest that their design could be conceptually construed as a 2x2 factorial, which allows a separation of the cognitive aspects of the advertising stimuli from the affective. As shown in Table 1, the picture of the abstract painting represents a control group in that it should not communicate that the facial tissue is soft and should not stimulate positive emotions. -On the other hand, the advertisement with the word soft on it should communicate that the tissue is soft but also without stimulating positive emotions. The picture of the sunset was expected to stimulate positive emotions but without the association of softness. Finally, the picture of the kitten might be expected to communicate softness as well as to stimulate positive emotions. As discussed later, there are some problems with the experimental manipulations when viewed in this manner but we think that it clarifies their study as well as introduce the extensions discussed in this paper.



The results of their analysis revealed that there were no significant effects attributable to message repetition but there were differences in response across the four advertisements. The brand associated with the picture of the kitten was more positively evaluated than the brand paired with the word soft. Secondly, although the verbal claim of softness created significantly stronger beliefs about the softness of the tissue relative to an abstract painting, the verbal claim ad did not result in more favorable attitudes than the abstract painting. Finally, the verbal message did not result in significantly stronger beliefs about softness than the picture of the sunset and attitudes were less favorable toward the brand associated with the verbal message than for the one associated with the picture of the sunset.

On the basis of these results and other analyses, Mitchell and Olson concluded that an attitude change without belief change had been demonstrated. The most convincing evidence they cited involved the more favorable attitudes toward the product in response to the -picture of the sunset compared to the verbal claim of softness. As shown in table l, this supposedly compares cognitive information without emotional information with emotional information without cognitive information. The attitude change without belief change was further verified using regression analysis in which there was a significant effect of attitude toward the ad on attitudes toward the advertised product beyond the beliefs and evaluation of beliefs component as specified in Fishbein's model (Mitchell and Olson 1981, p. 326).


Demand Effects

All subjects were exposed to all four treatments before belief measures and attitude ratings were obtained. This creates the possibility of demand effects and treatment interactions. That is, subjects may have determined the nature of the experiment, guessed the hypotheses, and altered their behavior in some way. Further, seeing a picture of a kitten after a picture of a sunset may sensitized subjects to the many dimensions on which they differ. To their credit, Mitchell and Olson conducted tests for these effects. A test for demand character had each subject write his/her perception of the purpose of the study and it showed no problems. A similar conclusion was reached after comparing groups varying in the order of presentation of the verbal claim and kitten ads. However, these do not rule out problems that might occur without a subject's awareness or other treatment interactions. A between subjects design virtually eliminates this concern.

Difficulty in Altering Affect without Changing Beliefs

Mitchell and Olson expected that both the picture of the kitten and the verbal claim of softness would alter the subjects' beliefs about the tissue's softness. Therefore, any attitudinal differences between these two conditions could be attributed to affectual differences resulting from liking of the kitten. Further, they expected the picture of the sunset to communicate only emotional information and not alter the subjects' beliefs about the product. These expectations were not entirely met. The results of the analysis of the beliefs revealed that beliefs about softness were significantly higher in the verbal claim condition than in the abstract painting condition. However, they were not significantly higher than in the sunset picture condition and actually significantly lower than in the kitten picture condition. Further, beliefs about the attractiveness of the colors of the tissue were significantly higher in the sunset picture condition compared to the verbal claim condition. Thus, the efforts to alter beliefs without altering emotions and to alter emotions without altering beliefs were not successful.

In their article, Mitchell and Olson acknowledge that the claim of "very soft" as opposed to "soft" would be more equivalent to the softness associated with a kitten and probably create a significantly higher softness belief than that resulting from the sunset picture ad. Further, analyzing their data using a two-way ANOVA would have provided a direct test of the relative influence of cognitive and emotional information.

Belief-Affect Confounded with Verbal-Visual

As mentioned in the introduction, another contribution of Mitchell & Olson was the demonstration that pictorial information might have more impact than an equivalent verbal message. Evidence relevant to this comes from the comparison of the verbal claim of softness for one brand of facial tissue with the association of a picture of a kitten with another brand. The latter resulted in more positive beliefs about the softness of the tissue. However, this conclusion came about because the design involved manipulations of verbal and visual information as well as affect and belief oriented messages. Thus, the critical comparison of the affect only (sunset picture) with the belief only message (verbal claim of softness) actually involves two differences. One is the desired difference in being oriented toward changing beliefs or changing feelings. However, another is that one message is pictorial and the other is verbal. Because of this, observed differences may be attributed to the differences in belief-affect orientation or to the verbal-visual presentation difference. As both differences are of interest, it would be desirable independently to manipulate both.


The discussion of Mitchell and Olson has identified three limitations of the research. One concern was that the belief-oriented ad (verbal claim of softness without positive affect) was not effective in altering subjects' beliefs about the brands' level of softness. To address this problem, verbal messages were developed that were equivalent to the kitten picture. One stated directly that the brand was very soft and the other did this in a more vivid manner by claiming that the brand was as soft as a kitten.

Another concern is that Mitchell and Olson's design mixed picture-word differences wi h affect-belief differences. This was addressed by having four ad stimuli involving pictures varying in their ability to communicate the softness of the tissue and to stimulate liking of the brand. Further, the three picture conditions used by Mitchell & Olson involved vastly different stimuli (kitten, sunset and abstract painting), which appeared to result in a variety of unwanted belief inferences (for example, the attractive color belief resulting from associating a brand with a sunset). Only pictures of animals were used to minimize belief inferences on other dimensions.

Finally, in order to alleviate a lingering concern with within subjects designs involving manipulations that can interact with each other (seeing a verbal claim of softness before seeing a picture of a kitten), the extension used a between subjects design.


Six different conditions were employed in a between subjects experimental extension of Mitchell & Olson's study. Four used pictures of animals associated with a brand of facial tissue and two had verbal messages. The four animals represented the two dimensions of softness and likableness as illustrated in Table 2. The two verbal messages used the words "very soft" (corresponding to Mitchell & Olson's suggestion that this was more equivalent to the picture of a kitten than the claim of soft) or the words, "soft as a kitten." The latter was thought to create a verbal equivalent of pictorial information contained in the kitten picture ad and thereby provide a better test of the communication effectiveness of verbal versus visual information. Eight different classes of students were used as subjects. Each class was split into two groups and assigned to different rooms. Once the experiment started, the subjects were exposed to a slide containing the advertisement (picture or words, a name card with the word "Breeze tissue" on it, and a picture of an unbranded box of tissues). After an exposure period of about ten seconds, the subjects were asked to complete a questionnaire. Upon completion, they were debriefed and excused from the session. To protect against demand effects, all subjects were told that they were participating in a brand name recall experiment. The directions for the questionnaire indicated that the project involved brand name recall and that different subject groups would be exposed to various treatment groups.




The instrument used in this study was composed of nine separate measurement sections. The first section consisted of simple demographic and informational questions not pertinent to the study. The second section gathered product attitudes using four semantic differential questions (good - bad, dislike very much- like very much, pleasant - unpleasant, and poor quality - good quality). Section 3 had a set of nine product characteristic importance questions (e.g., It is important to me that this product be as soft as possible). The fourth section contained six bipolar scales for evaluating the advertisement (good - bad, irritating - not irritating, like - dislike, not interesting - interesting, pleasant unpleasant, awful - nice). 1 he fifth section had fourteen questions concerning beliefs about the brand (e.g., This brand is probably softer than the average brand). Section six was composed of two separate parts; three questions about attitude toward purchasing the brand and one question reflecting an intention to buy the brand. Section seven consisted of a battery of questions on beliefs about the softness and likableness of the various animals used in the different ads and served as a manipulation check. The eighth section gathered responses pertaining to the cover issue of brand name recall and section nine provided open ended questions testing for demand artifacts.


A manipulation check was done using the subjects' evaluation of a list of four animals on two dimensions, softness and likableness, using bipolar adjectives like soft/hard, dislikable/likable, furry/rough, and friendly/unfriendly. Subjects in the verbal conditions evaluated the four animals used in the experiment. Subjects in the pictorial conditions evaluated the three animals used in the other ads and a tiger as a substitute for the animal used in their ad.



As seen in table 3, the perceptions of the animals corresponded to the dimensions that they were supposed to represent. The kitten was perceived as soft and likable, the whale as not soft and somewhat likable, the tarantula as soft but not likable, and the rhino as not soft and not likable. From these results, one might assume that the slides produced the reactions intended. However, we found that the ratings of the verbal concept of an animal did not correspond to perceptions of a brand associated with a picture of the animal. This was particularly true of the whale as will be discussed later.

Table 4 shows the different responses to the 14 belief statements by type of ad stimulus. The relevant individual beliefs were combined into five scales (for example, softness consisted of softer than average, soft enough for a baby, and soft enough to be used frequently for a cold). Only beliefs about softness were significantly affected by the association of the brands with an advertisement. A one-way ANOVA of these responses revealed a significant difference between the two verbal advertisements and all four pictorial advertisements. Thus, as suggested by Mitchell and Olson, the change from the description of the tissue as soft to "very soft" and as suggested by us, to "soft as a kitten," substantially increases perceptions that the tissue is soft. However, contrary to expectations, there were no significant differences in the overall softness belief measure among the four pictorial groups.

Next, the effect of the manipulations on attitudes toward the product were analyzed. This was done in several ways. First, following the design illustrated in table 2, a two-way ANOVA was run using the dimensions of softness and likableness as operationalized by the pictures of four different animals. The results revealed main effects of likeness (F = 28.0, p < .001) and softness (F = 7.1, p < .01) but an insignificant interaction. However, the differences were not as expected. Although the brands associated with the more likable animals (kitten and whale) were evaluated more favorably than those associated with the less likable animals (rhino and tarantula), the brands associated with the softer animals (kitten and tarantula) were evaluated less favorably than those associated with the less soft animals (rhino and whale). These results can be attributed to a highly favorable reaction to the whale ad. Based on Mitchell and Olson's results and the evaluations of different animals (see table 3), it had been expected that the kitten ad would be the most effective.

Second, a comparison was made of attitudes toward the product in response to the visual and verbal ads (see table 4 for the means). This was done by running a one-way ANOVA using the two verbal ads and the four ads with pictures. The results revealed differences only between the two verbal claim ads and the ad with a picture of a tarantula. The same result occurred when attitude toward-the act of purchasing the brand of tissue was the dependent measure. These results differ from Mitchell and Olson's finding that an affect-oriented ad (picture of sunset) was more effective in altering attitudes toward a brand than a belief-oriented verbal claim ad. In this study, even the most liked picture (whale) did not result in more favorable attitudes toward the product than the verbal claim ads.

The final stage in the data analysis was to examine the mediating role of beliefs. According to Mitchell and Olson, a failure of beliefs to mediate the influence of attitudes toward the ad in a regression analysis of attitude toward the brand as a function of beliefs about the brand demonstrates that there is a noncognitive influence of advertising on consumers' attitudes toward products.



The first part of the regression analysis involved a series of regressions with attitudes toward the product as the dependent measures and the subjects' beliefs about the brand of facial tissue on the attributes of softness, color, absorbency, toughness, and economy. Regressions were run with all treatment groups and separately for those exposed to ads with pictures and hose to the verbal only ads. They were also run using the belief scores alone and the belief scores weighted by the importance ratings. For all cases, softness was a significant determinant of attitudes toward the product. Among the other beliefs, only beliefs about color was significant and only for the groups exposed to the pictorial ads (t = 2.00 for the difference in the regression coefficient for the pictorial groups compared to the verbal groups). This finding corroborates Mitchell & Olson's finding that color beliefs may be affected by pictures relative to words.

Additional analyses comparing various formulations of the beliefs and their importance weights did not improve upon the basic finding that softness and color were the only measured beliefs influencing attitudes toward the product Therefore, the remaining analyses used only these two beliefs as the measures of the mediating effect of beliefs on the effect of the advertisements. These analyses were done on a group by group basis and are reported in table 5.

The results generally support Mitchell and Olson's findings that attitudes toward the advertisement do affect attitudes toward the advertised brand in a way not measured by the beliefs about the brand. The only condition where attitudes toward the advertisement did not have a significant effect on attitudes toward the product in a way not mediated by beliefs was in reaction to the picture of the whale.


This study extends Mitchell and Olson's (1981) research in several ways. First, by changing from a within subjects to a between subjects design, attitudinal effects attributable to different ads were smaller. This is likely to be true because the sequential presentation of various ads causes subjects to note many attribute differences because of differences with the other ads which would be ignored in a single presentation. We think that this was most true in reactions to the picture of the kitten. It did not result in particularly high evaluations of the brand of facial tissue as being soft even though relative to four other animals, the kitten concept was rated as softest. Thus, although Greenwald (1976) has argued persuasively for within subjects designs, we still worry that they tend to cause type one errors in a world already oriented toward rejecting the null hypothesis (see Greenwald 1975).

Our second issue was the effect of affect and belief manipulations on attitudes toward the brand. We feel that we were only partially successful here. Although the concepts of the four animals differed substantially on these two dimensions, product beliefs and attitudes resulting from an association of a brand with a picture of a particular animal did not work as expected. Specifically, a very favorable reaction to the brand of tissue associated with the whale resulted in a reversal of the expected softness effect (animals considered to be less soft resulted in more positively evaluated brands). Further, even though a kitten was clearly considered to be the most soft of the animals used, the brand of tissue associated with a picture of a kitten was not rated as more soft than brands associated with the other animals.

The third issue dealt with the effects of verbal and visual information. Here, our results differ from Mitchell dc Olson. First, by using words like very soft and soft as a kitten as opposed to "soft," we were able to cause subjects to rate the brand of tissue as being softer compared to when it was associated with a picture of a soft animal. As a result, with the exception of the whale ad condition, attitudes toward the brand were most positive in the two verbal conditions. Clearly, a "picture superiority effect" was not evident in our research.


What can we say about altitude toward the ad as an influence on attitudes toward the product in a way not mediated by beliefs? Although we did not consistently find an independent effect of attitudes toward the ad, it was a significant predictor in most cases and other research has tended to support the existence and importance of attitude toward the at (Gardner, 1985). The remaining issue is how does it occur? Much has been written about below awareness processing of emotional information and this seems to be a feasible phenomenon. Alternatively, reactions to the animals paired with the brand may become associated with the product through the cognitive association of the two. In this way, feelings toward kittens become associated with the brand especially when they constitute the only information about the brand: That is, if one thinks that kittens are playful, playfulness might then become an attribute of any product associated with kittens.



It is possible that what Mitchell and Olson and other researchers on attitudes toward the ad have discovered is the brand image. Currently popular with advertising agencies using names like brand personality and brand character, these appear to affect consumers' evaluation of products, especially fungible ones. Thus, the facial tissues associated with the pictures of kittens, whales, tarantulas and rhinos are similar to products associated with Green Giant, Charlie the Tuna, the Keebler Elves and the Pillsbury Doughboy. Attitude toward the ad research could be integrated with brand image research. Further, attitude measurement may need to be modified to include the image as well as the physical attributes.

It is interesting to note that the response to the whale ad, which was the most affect-oriented and the least belief-oriented, showed that attitudes toward the advertisement did not have a significant influence on attitudes toward the product when beliefs about the brand's softness were included in the regression analysis. This is a surprising finding in that the whale picture generated the most positive feelings toward the brand and toward the advertisement. Further, the picture of the whale was not expected to influence beliefs about the softness of the tissue. However, it seemed that many beliefs became more positive in response to the whale ad, demonstrating a "halo effect." Under such circumstances, attitude toward the ad becomes a redundant predictor of attitude toward the product.

Although acceptance of altitude toward the ad as an independent determinant of attitude toward the product has been near universal as evidenced by frequent ACR conference sessions on the topic and consumer behavior textbook treatment of the issue of low involvement communication, our research has shown that this relationship is complex and still deserving investigation.


Ahtola, O. T. (1975), "The Vector Model of Preferences: An Alternative to the Fishbein Model," Journal of Marketing Research, 12, 52-54.

Gardner, Meryl Paula (1985), "Does Attitude Toward the Ad Affect Brand Attitude Under a Brand Evaluation Set?" Journal of Marketing Research, 22 (May), 192-198.

Greenwald, Anthony G. (1975), "Consequences of Prejudice Against the Null Hypothesis," Psychological Bulletin, 82, 1-20.

Greenwald, Anthony G. (1976), "Within-subjects Designs: To Use or Not to Use?" Psychological Bulletin 83, 314-320.

Kisielius, Jolita and Brian Sternthal (1986), "Examining the Vividness Controversy: An Availability-Valence Interpretation," Journal of Consumer Research, 12 (March), 418-431.

Mitchell, Andrew A. (1984), 'The Effect of Verbal and Visual Components of Advertisements on Brand Attitudes and Attitude Toward the Advertisement," unpublished manuscript, Graduate School of Industrial Administration, Carnegie-Mellon University, Pittsburgh, PA 15213.

Mitchell, Andrew A. and Jerry C. Olson (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitude?" Journal of Marketing Research,-18 (August), 318-332.