If It Isn't a Duck Then Why Did It Quack? Competing Explanations For an Observed Effect of Illustrations in an Advertisement

Peter R. Dickson, The Ohio State University
Robert E. Burnkrant, The Ohio State University
Paul W. Miniard, The Ohio State University
Hanumantha R. Unnava, The Ohio State University
ABSTRACT - A further examination of the availability-valence hypothesis found that illustrations added to advertising copy enhanced belief formation and attitude toward the product. The results illustrate the value of taking as much care in the choice of dependent measures as is taken in choosing the manipulations to test the hypotheses.
[ to cite ]:
Peter R. Dickson, Robert E. Burnkrant, Paul W. Miniard, and Hanumantha R. Unnava (1986) ,"If It Isn't a Duck Then Why Did It Quack? Competing Explanations For an Observed Effect of Illustrations in an Advertisement", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 153-157.

Advances in Consumer Research Volume 13, 1986      Pages 153-157

IF IT ISN'T A DUCK THEN WHY DID IT QUACK? COMPETING EXPLANATIONS FOR AN OBSERVED EFFECT OF ILLUSTRATIONS IN AN ADVERTISEMENT

Peter R. Dickson, The Ohio State University

Robert E. Burnkrant, The Ohio State University

Paul W. Miniard, The Ohio State University

Hanumantha R. Unnava, The Ohio State University

ABSTRACT -

A further examination of the availability-valence hypothesis found that illustrations added to advertising copy enhanced belief formation and attitude toward the product. The results illustrate the value of taking as much care in the choice of dependent measures as is taken in choosing the manipulations to test the hypotheses.

INTRODUCTION

In a recent study, Kisielius and Sternthal (1984) undertook an examination of what they call the "availability-valence hypothesis." Their findings that instructions to image and the presence of story-board illustrations reduced subJects' judgments about a fictitious brand of shampoo were explained by the following set of propositions:

1) people process new information about an object by relating it to available information in memory,

2) the affective valence of information brought to mind by external stimuli in turn determines the favorableness rating of the external stimuli,

3) the picture analogs (illustrations) used in their research and their instructions to image advertising copy enhances cognitive elaboration that rakes available in memory information that is less favorable to the advocated message position and, consequently,

4) adding illustrations to advertising copy may result in a decrement in favorable brand attitudes.

In the Marketing Teleconference Consortium organized by the Oklahoma State University on November 28, 1984 entitled "Assessing the Black Box: When is a Duck a Duck," scholars from different institutions raised concerns as to whether Kisielius and Sternthal (K&S) had considered and eliminated alternative explanations for the major effects observed in their study. It was suggested that a research approach employing manipulation check measures and measures of moderating beliefs associated with specific claims made in the advertising copy right have helped eliminate competing explanations and led to a better test of theory. m accordance with this philosophy, the research reported here replicates K&S's study, incorporating additional measures which we believed would help further explain the judgmental effects of pictures.

The K&S results were surprising in view of the literature which has shown pictures to frequently have a favorable effect on learning (i.e., recall). Childers and Houston (1984) examined the role of interactive pictorial stimuli in which verbally presented brand names were accompanied by pictures that directly portrayed the brand name. They found that pairing pictures with verbal brand names enhanced brand name recall when people were engaged in sensory but not semantic processing. However, picture conditions yielded better recall than verbal conditions in both sensory and semantic processing groups after a two day delay. Lutz and Lutz (1977) have also found interactive pictures to enhance recall. This was not the case for non-interactive pictures (i.e., pictures that were separated from the brand name) which generated recall levels equivalent to those obtained in the verbal only condition. In a recent partial replication of the Lutz and Lutz study, Biron and McKelvie (1984) found that picture conditions resulted in superior recall to verbal conditions even when the illustration and verbal portions were separated.

K&S's pictorial stimuli were large drawings with the verbal claim appearing at the bottom of the page in small standard one-eighth inch type. Each verbal claim was about a sentence long. In the research reported here, their verbal and pictorial conditions are replicated. A second pictorial condition is added that is designed to draw greater attention to the verbal claim. In this condition, the verbal claim appears in large black lettering at the top of the page. If K&S's results are due to subjects' paying insufficient attention to the verbal claims, we would expect the negative effect of the picture condition to be eliminated in the large print picture condition.

An alternative possibility that could account for their results is that the illustrations employed in the picture conditions may have adversely affected the evaluation of the advertisements, and this negative evaluation may have generalized to the product. The pictures employed by RES were black line drawings or sketches on white paper. They were not at all like finished advertisements. Examination of differences in stimulus evaluations between verbal and pictorial advertisement conditions would address this issue. We, therefore, asked subjects to evaluate each of the 13 story-board pages in the advertisement on good-bad, desirable-undesirable and pleasant-unpleasant scales.

A third possibility is that their illustrations may have been irrelevant to their verbal claims. Pictures appear to facilitate learning when they are relevant to the verbal material, but have a detrimental effect on learning when they are irrelevant (Wollen, Weber and Lowry 1972; Wollen and Lowry 1971). Manipulation check questions included in our study permit an examination of this possibility.

A fourth consideration is that K&S had people rate the shampoo on three attributes after seeing all 13 pages. Yet it is likely that the effects of pictures on acceptance of each page's verbal claim will vary from page to page. Therefore, in addition to asking for summary evaluations of the shampoo, we assess subjects' beliefs about each of the claim made in the advertising stimuli to which they were exposed. We expect that some of these beliefs will be affected by the picture manipulation and that others will not be affected by this manipulation.

Relatively little is known about what differentiates effective and ineffective or counterproductive pictures. Whether or not pictures are interactive with verbal material, whether they are relevant to that material, whether people are engaged in semantic or sensory processing and whether they have sufficient time to fully process the stimuli have been shown to be important moderating variables. It is likely that other variables are also important. For example, message titles and pictures that help the reader better understand the meaning of the verbal material have been found to increase the cognitive effort employed in the comprehension of that material (Britton, Holdredge, Curry and Westbrook 1979) and the recall that results (Bransford and Johnson 1972, Dooling and Lachman 1971). Some of these variables will be explored in this research by examining differences in subjects' ratings between effective and ineffective pictorial ads. While we recognize the post hoc nature of this kind of data investigation, we believe it is a fruitful precurser to future research.

A final concern is methodological in nature and focuses on the particular manner in which K&S assessed subjects' attitude toward the shampoo. Their attitude measure, reproduced in Table 2, consisted of three semantic differential scales. As can be seen, the first two scales appear to tap subjects' beliefs or cognitive attitude rather than their affective attitude. Consequently, it is debatable whether K&S even tested the effect of elaboration valence on affective attitude toward the target object. If this is a valid concern, then the K&S findings should be interpreted using a cognitive based theory rather than an affect transfer theory.

The following study is an attempt to replicate and extend the K&S findings by exploring competing explanations for observed moderating effects of illustrations on advertising copy. The findings should therefore be of interest to three audiences: advertising practitioners, information processing theorists, and those interested in continuing the philosophy of science debate on the most appropriate methods for testing theory and advancing knowledge.

METHOD

Subjects and Procedures

A total of 555 students enrolled in a massive undergraduate marketing course participated in the study. In order to maximize comparability with the K&S study, we employed the same copy, pictures and dependent measures. [We would like to thank Jolita Kisielius and Brian Sternthal for their generous cooperation in supplying us with the experimental materials from their study.] Only the instructions were altered slightly to reflect that the product was being considered for introduction into the Columbus market rather than the Chicago area.

A booklet containing the advertising stimuli was distributed to subjects during class under the supervision of eight experimenters. Subjects were directed to examine each of the 13 pages that contained the phrases comprising the advertising copy (see Table l for the 13 phrases). They were exposed to each page for five seconds. After the directed page by page study of the advertisement, subjects responded to a series of measures described below

TABLE 1

THE ADVERTISING COPY

At the end of the exercise we also verbally instructed subjects to reread the initial instructions and place an "x" on the back of the booklet if they had not conscientiously followed the instructions or if they had answered any or all of the questions flippantly. Fifty nine participants did so and were excluded from the analysis. A further 50 subjects were dropped from the analysis because of missing data or clearly inconsistent responses to the K&S and attitude measures that indicated a lack of care and/or effort.

Independent Variables

The experiment involved a 2x3x2 between-subjects factorial design. The first factor manipulated the presence or absence of the following paragraph in the initial instructions:

"To compensate for the unfinished form of the ad, we would like you to view the ad in a particular way so as to get a good feel for the product. As you are presented with the ad, please try to construct a mental picture or image of the product as it is being described. For example, if an ad about a dishwashing liquid stated that it had high sudsing value, you might imagine a sink full of dishes covered with soap bubbles. Certain characteristics of a product may seem hard to picture in your mind. Nevertheless you should attempt to create some sort of image.

The second manipulation involved three story-board formats. The first (verbal only) presented the reader with a type written phrase at the bottom of each of the 13 pages. The rest of the page was conspicuously blank. In the second condition (picture-words), the K&S illustrations were added. We also explored the framing effect of the words on the picture (rather than vice versa) by creating a word-picture condition. This involved placing the phrase at the top rather than the bottom of each page in very bold printed letters some 4-5 times larger than in the second condition.

The third factor was a blocking on sex. The female students received a pink booklet with female illustrations, the males a blue booklet with male illustrations.

Measures

Upon completion of the advertising message, subjects first responded to the three bipolar scales used by K&S (see Table 25. Our study went further by having subjects respond to (1) five evaluative semantic differential attitude scales (see Table 2), (2) 11 general belief statements about Celebrate, (3) 16 very specific belief judgments using the words from the advertising copy but including three false product claims, and (4) questions that evaluated each of the 13 story-board pages. The last set of measures consisted of again presenting each of the 13 stimulus pages on the left facing page. The right facing page contained measures asking subjects to rate the stimulus page on three evaluative scales, followed by three scales assessing the ease of understanding, ease of relating to, and meaningfulness of the page. The next three questions measured how favorably disposed the subject was toward the entire page, the words alone, and the illustration alone (in the conditions where illustrations were presented). For the illustration conditions, the final four measures captured the extent to which each picture was perceived to be consistent with the accompanying words, similar to the words, related to the words, and whether it was perceived to clearly illustrate the specific point or feature made by the text (e.g., "The picture clearly illustrates that Celebrate has a handle built into the side of the bottle so it won't slip out of the hands"). For the verbal only condition, however, the final four measures assessed subjects' perceptions of the ease in which images could be formed for the words (consisting of two measures) and the words could be related to subjects actual shampoo experiences, as well as their agreement with the statement "Nothing comes to mind when I read these words.

TABLE 2

RATING SCALES EMPLOYED IN THIS RESEARCH

Subjects answered these questions after studying each story-board page for as long as they desired. These measures, although taken some time after the initial paced exposure to the advertisement, enabled us to explore the impact of specific word-picture combinations on related specific and general beliefs as well as the impact of the entire set of 13 pages of text and illustrations on overall attitude toward the Product.

HYPOTHESES

For the purposes of this paper we have chosen to only present a number of preliminary findings that address the following hypotheses:

H1: The imaging instructions will reduce affective evaluation of the advertised product (shampoo).

H2: The presence of illustrations accompanying the advertising copy will reduce affective evaluation of the product.

H3: Framing the pictures with the copy by presenting the words at the top of the page in very large printed letters will moderate the detrimental effect of the illustrations on attitudes.

These three hypotheses are derived from availability-valence hypothesis. -The first two are directly implied from the propositions presented above. The third hypothesis is based on the theory that external contextual information as well as internal available information will shape the interpretation of the distal cues. This implies that the pictures may well influence interpretation of the copy and vice versa. If the words are presented boldly at the top of the page rather than at the bottom of the page, then the subject may be more likely to read the copy first before studying the pictures. This will result in less picture interference with the words and an interpretation of the picture more consistent with the copy. It may also reduce the impact of relevant and (perhaps more importantly) irrelevant information on the interpretation and assimilation process which is likely to occur if the subject is first exposed to the illustrations and then reads the relevant copy (see Edell and Staelin 1983).

We anticipate that the effect of imaging and the effect of the illustrations on attitudes will be accompanied by similar effects on beliefs about product claims made in the copy. However, we expect that the effect of adding each of the 13 different story-board illustrations to each of the phrases will vary and depend on the picture-word consistency.

RESULTS

Findings for K&S Measure

The scores on the three item K&S measure were summed for each individual. It yielded a coefficient alpha of .47 (n = 446) which is lower than the alpha of .60 (n = 43) reported by K&S. The second scale (easy/difficult to hanle) was problematic as reflected by a corrected item-total correlation of only .17. Indeed, alpha rose to .63 if this item was deleted. The K&S scale was then submitted to a 2x3x2 factorial analysis of variance. Contrary to K&S's findings, neither the imaging nor pictorial manipulations had a significant effect. The only significant main or interaction effect was that females rated Celebrate higher (P < .004) than did males.

Findings for Our Attitude Measure

The above analyses were then replicated for our five item attitude measure. This measure received a coefficient alpha of .85. Our attitude measure also responded to the experimental manipulations quite differently than the K&S scale with the exception of the imaging instructions again failing to have a significant effect. Thus, El was not supported.

Contrary to K&S's findings, pictures had a significant (p < .04) positive effect on attitude. The mean score across the five scale items was 4.98 in the verbal only condition, 5.25 in the picture-small print condition, and 5.26 in the picture-large print condition. Differences between the verbal only condition and each of the picture conditions were significant (p c .05). However, H2 was not supported as the pictures increased rather than decreased affect toward the product. H3 was also not supported, as affect was constant across the two picture conditions. Other than the effect stemming from the pictorial manipulation, none of the remaining main or interaction effects were significant (p > .1). Consequently, we collapsed the design into a simple single factor design consisting of two groups: the verbal only subjects versus the picture subjects (including both picture versions). Unless otherwise noted, subsequent analyses are based on a comparison of these two conditions.

The above results raise questions about the tenability of the availability-valence hypothesis as presented and tested by K&S. First, the validity and the consistency of the three item K&S measure can be questioned. Second, we were not able to replicate the previous findings using the K&S measure. Third, and of greatest concern, the illustrations enhanced affective attitudes toward the product. Granted the effect explained very little of the overall variance, but it was still statistically significant in the wrong direction'

To explore whether the illustrations contributed to the formation of stronger beliefs about the product, which in turn influenced attitude, we examined the impact of pictures on specific beliefs. It should be noted that the availability-valence hypothesis does not explicitly propose that illustrations influence learning and acceptance of the verbal claims which in turn influence attitude. Rather, it proposes a direct transfer of valence from internally generated available thoughts to the target object.

Effect of Illustrations on Beliefs

Comparisons of product-related beliefs between picture and verbal only conditions revealed a number of significant (p < .05) differences and several that approached significance (see Table 3). It is reassuring that some, but not all, beliefs were influenced by the message manipulation. This finding eliminates the possible contention that subjects employed an evaluative inferential process in responding to the measures. Only if all (or at least most) beliefs varied in sympathy with the attitude measures across the message conditions would such an argument seem plausible.

For those beliefs affected by the manipulation, pictures enhanced subjects' beliefs about the product. For all but one of these beliefs, higher belief scores reflect greater processing accuracy and/or acceptance of the claim.

TABLE 3

RESULTS FOR SELECTED BELIEFS

This was not the case for the statement "Celebrate comes in four special formulas" as the advertisement states it comes in three special formulas. For this incorrect product claim, subjects in the pictorial conditions had less accurate perceptions although they did on average disagree with the statement. But this result is understandable when one appreciates that the illustration accompanying this claim depicts five people. This inconsistency between the pictorial and semantic stimuli apparently had a detrimental impact on the picture condition subjects' accuracy.

An important question at this juncture is whether message impact on affect was driven by the above differences in cognition. If the message influences affect through beliefs, then one should expect that controlling these beliefs should eliminate the message impact on attitude. Accordingly, various sets of beliefs were covaried to test their ability to mediate the influence of pictures on affect. The first set of covariates were the beliefs associated with the bottle design (beliefs 2-5, 7 and 8 in Table 3). These covariates completely eliminated (p > .3) the impact of pictures on attitude. The second set of covariates consisted of beliefs focusing on hair treatment (beliefs 6, 8, 9). Again the effect of pictures was eliminated (p > .1) although the beliefs pertaining to number of formulas was not a significant (p > .7) covariate. The final covariance analysis used the single remaining belief (belief l in Table 3). Although this belief was a significant (p < .001) covariate, it did not completely mediate the impact of pictures (p < .05). In sum, these analyses suggest that pictures influenced particular product beliefs (especially the beliefs about bottle design) which in turn impacted on product attitude.

An important question that arises at this point is why do some pictures enhance associated beliefs while others do not? To address this concern, we compared those stimulus pages which did (p < .05) versus did not (p > .1) have a significant effect on product beliefs. The pages where the effect approached significance were omitted. [The page presenting the potentially conflicting picture-word information about the number of formulas was also omitted.] These comparisons were performed on various measures assessing favorability toward stimulus elements, ease of understanding, meaningfulness, and consistency. The results are summarized in Table 4.

TABLE 4

RATINGS OF EFFECTIVE AND INEFFECTIVE PICTURES

As can be seen, these two groups differed across nearly all of the measures except for those representing the meaningfulness and ease of understanding and relating to the page. Pages which influenced product beliefs generated more favorable responses than pages failing to influence beliefs. The presence of such differences must be cautiously interpreted, however, as evidence for a variable's role as an antecedent of the picture effect. For example, although illustrations were rated more favorably for pages having an effect on belief, further data analysis suggest that subjects' favorability toward the illustration was unrelated to the pictorial effects observed here. In particular, comparisons between the picture and verbal only conditions for those pages generating stronger beliefs revealed that subjects had a more favorable (p < .05) evaluation of the page in the verbal only condition' If these page evaluations were responsible for the impact of pictures on beliefs, then the verbal only condition should have enhanced beliefs. That the verbal only condition produced more favorable page evaluation but weaker beliefs suggests that subjects' evaluative reactions to the illustration were not responsible for the effect of pictures on beliefs. Indeed, covarying these page evaluations did not eliminate the impact of the picture manipulation on beliefs.

DISCUSSION

K&S summarized their findings by pointing out that "on four occasions, verbal information presented alone was shown to induce more message-consistent judgments than verbal information accompanied by pictorial analogs" (p. 61). They argue that this effect is due to the "pictorial analogs" inducing greater elaboration of information that is less favorable than that elicited by the verbal claims. The results reported here present a dramatic contrast to their findings. Using their stimuli and procedures, verbal information accompanied by pictures produced a more favorable response than did verbal information alone. One could interpret this effect as also being consistent with the availability-valence hypothesis by arguing that in our study the pictures induced elaboration of information that was more favorable than the information available in memory for subjects in the verbal only condition. However, the same theory would then be employed to explain two diametrically opposed effects obtained following the same procedure and using the same stimuli. This points out a serious weakness in the availability-valence hypothesis. In a very real sense it is nonfalsifiable. No matter what effects are obtained on attitude, it can always be argued that they are consistent with the valence of information available in memory. The theory does not permit us to predict on an a priori basis whether pictures or other stimuli will have a favorable or an unfavorable effect on the valence of available information.

This research went beyond the use of a multi-item attitudinal scale as a single dependent variable. Subjects also reported their level of agreement with statements made in the verbal claims. Significant belief effects emerged for six of the thirteen verbal claims made in the K&S stimuli. These belief scores show more agreement with the claim in the pictorial conditions than in the verbal only condition. When these beliefs were employed as covariates in separate analyses of variance, they eliminated the effect of message on attitude.

These findings appear to be inconsistent with implications of the availability-valence hypothesis. K&S state as a general proposition that "pictorial analogs" should enhance elaboration. The resulting effect depends on the valence associated with this elaboration. They argue that "the richness of cues in the picture condition was likely to have enhanced the availability of idiosyncratic information" (p. 56). If this idiosyncratic information were the cause of the obtained effect on attitude, then covarying for the belief scores obtained on ratings of the verbal claims should not eliminate the effect of pictures on attitude. The finding that covarying for these beliefs did eliminate the effect on attitude suggests that the obtained effect of pictures on attitude was not due to elaboration of idiosyncratic information.

If we assume as they do that pictures enhance "elaboration of message relevant information" (p. 61), we should expect to find a consistently greater level of learning in the picture condition across their stimuli. That is, if presenting the claim with a picture increases stimulus relevant elaboration we should obtain significant belief strength effects for each of the verbal claims. The finding that only some of the beliefs were significant suggests that the effect on learning may depend upon the type of picture employed or upon the relationship between the picture and the accompanying verbal claim.

In a general sense, these findings also support the usefulness and importance of carefully examining obtained effects on a series of dependent variables. Without the belief measures discussed above, we would have been unable to address alternative explanations for the obtained effects on attitude. On the other hand, in hindsight, we feel that our dependent measures of attitude toward each page and picture-word inconsistency were weak because of the manner in which this data was collected. It is questionable whether the responses to these measures fully reflected the processing that occurred during the initial paced exposure to the advertisement.

It should be acknowledged that the effects observed in this research did not account for a large proportion of the variance in the data. We did not obtain a significant effect on the imaging manipulation, and the manipulation of the order and visual dominance of the verbal claim over the illustration was not effective. We followed K&S by administering our study in a classroom, and this may have adversely affected precision. While the distractions and loss of control that are common in classroom exercises may well have added error, there is no reason to believe that classroom administration would have a biasing effect on our results. This is particularly true here because Kisielius and Sternthal also employed classroom administration. We recommend that future experimental research: reduces the number of pages and claims in the advertisement, manipulates time of exposure by a greater amount, manipulates the attractiveness and the consistency of specific picture-word combinations within the advertisement, and is administered in small groups or individually. The value of using appropriate measures of affect and beliefs has been demonstrated in the above study.

REFERENCES

Brion, Jose and Stuart J. McKelvie (1984), "Effects of Interactive and Noninteractive Imagery on Recall of Advertisements," Perceptual and Motor Skills, 59, 799-805.

Bransford, John D. and M.K. Johnson (1972), "Contextual Prerequisites for Understanding: Some Investigations of Comprehension and Recall," Journal of Verbal Learning and Verbal Behavior, ii, 717-726.

Britton, B.K., T.S. Holdredge, C. Curry and R.D. Westbrook (1979), "Use of Cognitive Capacity in Reading Identical Texts with Different Amounts of Discourse Level Meaning," Journal of Experimental Psychology: Human Learning and Memory, 5, 262-270.

Childers, Terry L. and Michael J. Houston (1984), "Conditions for a Picture-Superiority Effect on Consumer Memory," Journal of Consumer Research, 11, 643-654.

Dooling, D.J. and R. Lachman (1971), "Effects of Comprehension on Retention of Prose," Journal of Experimental Psychology, 88, 216-222.

Edell, Julie A. and Richard Staelin (1983), "The Information Processing of Pictures in Print Advertisements," Journal of Consumer Research, 10, 45-61.

Kisielius, Jolita and Brian Sternthal (1984), "Detecting and Explaining Vividness Effects in Attitudinal Judgments," Journal of Marketing Research, 21, 54-64.

Lutz, Kathy A. and Richard J. Lutz (1977), "Effects of Interactive Imagery on Learning: Application to Advertising," Journal of Applied Psychology, 62, 493-498.

Wollen, Keith A. and Douglas H. Lowry (1971), "Effects of Imagery on Paired-Associate Learning," Journal of Verbal Learning and Verbal Behavior, 10, 276-284.

Wollen, Keith A., Andrea Weber and Douglas M. Lowry (1972), "Bizarreness Versus Interaction of Mental Images as Determinants of Learning," Cognitive Psychology, 3, 518-523.

----------------------------------------