Headline-Visual Consistency in Print Advertisements: Effects on Processing and Evaluation

Edwin R. Stafford, Utah State University
Beth A. Walker, Arizona State University
Vincent J. Blasko, Arizona State University
[ to cite ]:
Edwin R. Stafford, Beth A. Walker, and Vincent J. Blasko (1996) ,"Headline-Visual Consistency in Print Advertisements: Effects on Processing and Evaluation", in NA - Advances in Consumer Research Volume 23, eds. Kim P. Corfman and John G. Lynch Jr., Provo, UT : Association for Consumer Research, Pages: 56-62.

Advances in Consumer Research Volume 23, 1996      Pages 56-62

HEADLINE-VISUAL CONSISTENCY IN PRINT ADVERTISEMENTS: EFFECTS ON PROCESSING AND EVALUATION

Edwin R. Stafford, Utah State University

Beth A. Walker, Arizona State University

Vincent J. Blasko, Arizona State University

[The authors wish to thank Peter Reingen and Michael Hutt for their helpful comments and advice.]

INTRODUCTION

Increased use of unusual visual-verbal combinations in print ads (cf. McQuarrie and Mick 1992) reflects a growing recognition that the level of consistency between the picture and the words influences an ad's effects. An inconsistency exists when the ad's headline and visual deviate from the reader's expectations. While advertisements may feature unrelated, irrelevant, or extremely inconsistent verbal-visual information (Heckler and Childers 1992), the verbal-visual meanings in an ad often reflect a moderate level of inconsistency (Blasko and Mokwa 1986). Unlike an extreme inconsistency which cannot be resolved (Meyers-Levy and Tybout 1989), a moderate inconsistency encourages the reader to uncover the underlying connection between the picture and the words that makes the advertisement understandable. Rhetorical figures (e.g., resonance, irony), for example, are used by advertisers to present the consumer with a moderate inconsistency (McQuarrie and Mick 1992). Consider a dairy association ad picturing healthy adults drinking milk accompanied by the headline, "Everybody Knows that Milk's for Babies." This ad employs irony by using "babies" to refer to adults (literal opposites) (Corbett 1990). Resolution of the inconsistency between "babies" and the picture invokes the interpretation that adults, like babies, need milk.

Research on ad headline-visual consistency has been limited and has produced conflicting results. Researchers have typically examined two extreme forms of verbal-visual combinations C a complete match or complete mismatch of verbal-visual information. Isolating memory for brand information as the dependent variable, inconsistent advertisements have been found to be both more effective (Houston, Childers and Heckler 1987) and less effective (Edell and Staelin 1983) than consistent verbal-visual combinations. Moderate verbal-visual inconsistencies have not been explicitly examined.

This study examines the effect of the degree of consistency between the headline and visual in print ads on processing and evaluation. Particular interest centers on the relative effects of moderately inconsistent headline-visual combinations that employ irony by combining a headline and visual that convey opposite meanings (Corbett 1990). "Opposites" represent a moderate form of inconsistency because they have strong implicit associations and can be readily related to one another (cf. Colombo and Williams 1990). Drawing on the information processing (Mandler 1984) and psychology of aesthetics (Berlyne 1971) literatures, we posit that the process of decoding and interpreting a text is intrinsically rewarding. Further, successful interpretation should create liking for the ad. We propose that the process of resolving a moderate inconsistency within a print ad is most rewarding and may result in more positive ad and, consequently, brand evaluations and purchase intentions when compared to consistent and extremely inconsistent headline-visual combinations. Extending previous work, this study uses ad-related processing, ad and brand evaluation, and purchase intention as dependent measures.

RESEARCH ON VERBAL-VISUAL CONSISTENCY: AN OVERVIEW

Communications and psychology research suggests that consistent verbal and visual information produces better memory and is more easily understood than inconsistent information. Incongruent visual and verbal information is less easily interpretable and reduces memory and learning for text material as compared to consistent visual and verbal material (Bock and Milz 1977; Peeck 1974; Stein, et al. 1987). Although the finding that consumers process visual and verbal information differently has been well established in marketing (cf. Childers and Houston 1984; Lutz and Lutz 1977), research on verbal-visual consistency in print advertisements is limited and the findings are conflicting. Comparing consistent (framed), inconsistent (unframed), and verbal-only print advertisements, Edell and Staelin (1983) found that consistent ads outperformed inconsistent and verbal-only messages in terms of brand-item recall. Evoking a "distraction" explanation, the authors explained that the inconsistent visual impeded comprehension, memory, and evaluation processes by distracting the consumer from thinking about the advertised brand. In contrast, Houston, Childers and Heckler's (1987) study of advertisements containing congruent and incongruent verbal and visual information revealed an advantage for inconsistent verbal-visual combinations. Suggesting that deviations from expectations stimulated elaborative processing which improves memory, the authors found that inconsistent ad information was more memorable than consistent information.

The conflicting findings may be resolved by considering the degree of verbal-visual inconsistency used across investigations. Most studies of verbal and visual information have used an extreme operationalization of inconsistency C verbal information that conflicted with the visual depiction. Here, "inconsistent" information was less effective than consistent verbal-visual information. The inconsistency in the ads used by Houston, Childers and Heckler (1987), however, was more moderate. The authors noted that the verbal information was different from, but not in "sharp contrast" to (p. 68), the visual information. These ads were more effective than their consistent ads. While research has implied either a positive or negative linear relationship between verbal-visual consistency and ad effectiveness, a nonmonotonic relationship may exist. Ads featuring moderate levels of verbal-visual inconsistency may be more effective than either consistent or extremely inconsistent ads.

Evidence suggesting that individuals appreciate moderate levels of discrepancy stems from the psychology (Berlyne 1971) and advertising (McQuarrie and Mick 1992) literatures. Blasko and Mokwa (1986) proposed that ads with headlines and visuals conveying opposite meanings may be particularly effective. Since "opposites" have strong implicit associations and can be readily related to one another (cf. Colombo and Williams 1990), these ads may be easily understood. Further, because the moderate inconsistency between the headline and visual may be perceived as clever, the ad evaluation may be enhanced.

Examining "resonance" in advertising, McQuarrie and Mick (1992) also provide evidence for the success of moderately inconsistent headline-visual combinations. In addition to conveying multiple meanings, by definition, resonant ads contain moderately inconsistent information that deviates from expectations and normal usage, but are nonetheless understandable because the headline and visual are relevant to each other. An ad picturing men's ties arranged to form a bouquet that was described by the caption, "Forget-Me-Knots," illustrates the resonance concept. The researchers found that, compared to consistent verbal-visual combinations, resonant ads generated more positive ad and brand attitudes as well as better unaided recall of ad headlines. Finally, moderate inconsistencies have been studied in the context of understanding the evaluations of new products. Meyers-Levy and Tybout (1989) found that new products that were moderately inconsistent with consumers' existing product category schema received more positive brand evaluations than new brands that were either schema consistent or extremely inconsistent with expectations.

A SCHEMA CONSISTENCY APPROACH TO UNDERSTANDING AD EFFECTS

Mandler (1984) provides a framework for understanding headline-visual consistency effects. Mandler posits that it is the extent to which an object deviates from expectations and the ability of an individual to resolve the inconsistency that influences processing and evaluation. Mandler theorizes that, under conditions of schema consistency, expectations are met, and processing is automatic, not effortful (Sujan 1985). Consequently, a consumer produces fewer thoughts as compared to processing an inconsistency. In addition, the absence of effortful processing reduces the extremity of the evaluation (Meyers-Levy and Tybout 1989). For example, when a print ad for an arthritis foundation features a young girl in a wheel chair accompanied by the headline, "This Little Girl has Arthritis," the reader will extend minimal effort since the headline and visual convey equivalent messages.

More central to Mandler's framework is the relative consequences of moderate and extreme inconsistencies. Inconsistencies stimulate elaborative processing as individuals attempt to integrate the headline and the visual. However, the extent of the inconsistency and the ability of the individual to resolve the inconsistency influence the nature of processing and the valence of evaluation. According to Mandler (1984), moderate inconsistencies can be resolved. For example, an arthritis foundation ad depicting a young girl in a wheel chair followed by the headline "This Little Old Lady has Arthritis," the headline (old lady) and visual (young girl) may be considered moderately inconsistent. While the particular headline-visual combination is unexpected, the inconsistency can be resolved without requiring a fundamental change in the consumer's cognitive structure. Berlyne (1971) suggests that the very process of successfully resolving an inconsistency is rewarding, which may bolster a consumer's evaluation of the advertisement and brand. This is in line with McQuarrie and Mick (1992) who suggest that consumers perceive moderate inconsistencies in ads as clever and enjoy deciphering their meaning.

Extremely inconsistent headline-visual combinations are difficult to reconcile. An arthritis foundation ad featuring a young girl in a wheel chair accompanied by the headline "Tennis Elbow can be an Early Sign of Arthritis" may be considered extremely inconsistent. The connection between the girl and tennis elbow is simply impossible to find. While extreme inconsistencies may heighten attention and stimulate processing, the inability to resolve the inconsistency generates confusion and frustration, which may result in more negative evaluations of the advertisement. This relationship between inconsistency, processing, and evaluation is supported in the aesthetics literature (Berlyne 1971).

In sum, Mandler and Berlyne predict an inverted U-shaped nonmonotonic relationship between schema inconsistency and evaluation. Objects that are moderately inconsistent with expectations are evaluated more favorably than those that are consistent or extremely inconsistent with expectations. Further, the process of inconsistency resolution is proposed to underlie the effects of consistency on evaluation. In advertisements, the picture establishes an expectation for the verbal material that follows (Houston, Childers and Heckler 1987). The level of consistency, then, is determined by the degree of discrepancy between the schema elicited by the picture and the meaning conveyed in the headline.

Though the level of inconsistency is central to understanding an ad's effects, operationalization of moderate and extreme inconsistency is less apparent (Meyers-Levy and Tybout 1989), and this has led to considerable debate in the verbal-visual processing literature. Mandler (1984) suggests that levels of consistency are determined by the ease with which the inconsistency is resolved. Drawing from Heckler and Childers' (1992) framework, we identify relevancy and expectancy as determinants of the degree of inconsistency. Relevancy refers to the extent to which the headline and visual contribute to (or detract from) communicating a unified message. Expectancy is defined as the degree to which the headline fits the pattern of structure evoked by the visual portion of the advertisement. We posit that a consistent headline-visual combination is both expected and relevant. A moderately inconsistent message is relevant, but not expected. Because the headline and visual are relevant to each other, a moderate inconsistency can be resolved with relative ease (Meyers-Levy and Tybout 1989). An extremely inconsistent headline-visual combination is neither expected or relevant. In contrast to moderate inconsistencies, extreme inconsistencies cannot be resolved because they diverge dramatically from a consumer's established patterns of thought.

HYPOTHESES

Based on Mandler's (1984) framework, we advance the following hypotheses:

H1: Moderately inconsistent advertisements will lead to more favorable ad evaluations than consistent or extremely inconsistent ads.

Hypothesis 1 reflects Mandler's central premise that objects that are moderately incongruent will be more favorably evaluated than objects that are congruent or extremely incongruent with expectations. Specifically, the experience of successfully resolving the inconsistency should create liking for the ad. Mandler further proposes that the process of resolving the inconsistency underlies the effect of schema inconsistency on evaluation. Not only should this process affect the number and type of thoughts generated, but it may also affect ad-related beliefs formed as the result of the headline-visual inconsistency (e.g., this ad is confusing, irritating, interesting). Because the effect of inconsistency on evaluation is influenced by the extent of elaboration and the ability of the consumer to resolve the inconsistency, we propose the following four hypotheses:

H2: As compared to consistent advertisements, moderate and extremely inconsistent advertisements will result in more (a) references to the headline-visual combination and (b) ad-related thoughts.

H3: As compared to extremely inconsistent and consistent advertisements, moderately inconsistent ads will (a) generate more positive references to the headline-visual combination, (b) generate more positive ad-related thoughts, and (c) be perceived as more interesting.

H4: As compared to moderately inconsistent and consistent advertisements, extremely inconsistent advertisements will (a) generate more negative references to the headline-visual combination, (b) generate more confusion ad-related thoughts, (c) be perceived as the most irritating, and (d) be perceived as the least comprehensible.

H5: Cognitive responses that reflect the resolution process will mediate the relationship between ad type and ad evaluation.

Finally, consistent with recent work on the relationship between ad evaluations (Aad), brand evaluations (Abrand), and purchase intentions (PI) (cf. Miniard, Bhatla and Rose 1990; Mittal 1990), the following hypothesis is suggested:

H6: Through their impact on ad evaluations, moderately inconsistent advertisements will lead to (a) more positive brand evaluations and (b) greater purchase intentions than consistent or extremely inconsistent advertisements.

METHOD

A 3 (headline-visual consistency) by 2 (product class) between-subjects experiment was designed to examine how verbal-visual consistency affects processing, ad and brand evaluation, and purchase intention. Product was manipulated to assess the generalizability of the findings.

Stimuli

Six full-page black and white advertisements were developed by combining ad type (consistent; moderately inconsistent; extremely inconsistent) with product class (mountain bike and a zoo). These product classes were both relevant and familiar to the subject population. The constant portion of each ad contained a brand name (Klein Mountain Bike; St. Louis Zoo) and logo in the lower right corner and a visual of either a cyclist covered with mud racing a bike on a mountain trail or of a growling lion. Each visual was accompanied by a headline that either conveyed a consistent ("Taking the Dirty Off-Roads"; "Come See the Lions"), moderately inconsistent ("Good Clean Fun"; "Come See the Little Kitties"), or extremely inconsistent ("Tire Repair Made Easy"; "Come See the East African Elephants") message for the bike and zoo ads, respectively. Following studies of advertising rhetoric (e.g., McQuarrie and Mick 1992), we employed irony to manipulate the moderate level of inconsistency. Irony employs contradiction by using a word to convey a meaning opposite to the literal meaning of another word (Corbett 1990). We assumed that this type of contradiction, while unexpected, is resolvable (cf. Williams and Lilly 1985).

Pretests

A pretest was conducted using 115 subjects to determine if the six headlines by themselves produced differences on the dependent measures. Subjects reviewed one of six headlines for ten seconds and then completed the measures of cognitive response, ad-related beliefs, and ad and brand evaluations that were used in the main experiment. Pair-wise comparisons of the subjects' responses who were randomly assigned to one of the six headline conditions revealed no differences between headlines on most of the measures. Three exceptions included ad interestingness (F[2,113]=5.71, p<.004) (where extremely inconsistent headlines were perceived as less interesting than consistent and moderately inconsistent headlines), level of irritation (F[2,113]=3.57, p<.031) (where consistent headlines were perceived as more irritating than moderate and extremely inconsistent headlines), and positive ad-related thoughts (F[2,113]=4.03, p<.020) (where moderately inconsistent headlines generated more positive thoughts than extremely inconsistent but not consistent headlines). However, in each case the pattern of means in the headline-only conditions differed from the results in the main experiment, suggesting that the headline-visual combination and not the headline alone, produced the effects in the main experiment.

A second pretest (n=105) using seven-point scales measured the relevancy (extremely irrelevant-extremely relevant; extremely unrelated-very closely related; correlation=.75) and expectancy (extremely unexpected-extremely expected; extremely surprising-extremely predictable; correlation=.86) for each headline-visual combination. Subjects were randomly assigned to one of six headline-visual combinations. We expected that consistent headline-visual combinations would be both expected and relevant, moderately inconsistent ads would be unexpected and relevant, and extremely inconsistent headline-visual combinations would be both unexpected and irrelevant. As predicted, the main effect of headline-visual consistency on relevancy was significant (F[2,101]=36.83, p<.001). Extremely inconsistent headline-visual combinations were significantly less relevant to each other ( =3.18, where 1 is extremely irrelevant) than the consistent and moderately inconsistent conditions ( =5.67 and =5.53, respectively). The impact of consistency on expectancy was also significant (F[2,101]=21.40, p<.001). As expected, consistent headline-visual combinations were significantly more expected ( =4.70, where 1 is extremely unexpected) than the moderate ( =3.69) and extremely inconsistent ( =2.59) conditions.

Experimental Procedure

Undergraduate students (n=109) were randomly assigned to one of the six treatment conditions. Subjects viewed the ad for 10 seconds and at their own pace completed a thought-listing task, 7-point bipolar scales measuring manipulation effectiveness, and the dependent variables.

Measures

Cognitive Responses. Following exposure to a test ad, written protocols were collected by asking subjects to report any and all thoughts, no matter how simple, complex, relevant, or irrelevant. Consistent with Buckholz and Smith (1991), three judges blind to the experimental conditions each coded two-thirds of the thoughts in terms of their content (product-related; ad-related; combination-related; other) and valence (positive; negative; neutral; confusion). Combination thoughts included references to the relationship between the headline and the visual of the ad (i.e., "The caption did not really fit the picture," "It said 'Come see the Elephants' at the zoo but it showed a lion"). Confusion-related thoughts included statements that indicated difficulty in understanding the advertisement or product (i.e., "It is hard to understand," "I had to read it a couple of times to get the meaning"). All thoughts were coded by two judges. Interjudge reliability was 82.5% for content of thought and 84.8% for valence of thought. Disagreements were resolved by discussion.

Ad-related Beliefs. Beliefs about the ad that reflected subjects' ability to resolve inconsistencies were also assessed. By averaging subjects' responses to 7-point bipolar scales, ad interestingness (interesting-uninteresting; provocative-not provocative; intrigued-bored; alpha=.83), ease of comprehension (easy to understand-difficult to understand; not at all confusing-very confusing; correlation=.79), and the level of irritation generated by the ad (irritating-not irritating; annoying-not annoying; correlation=.78) were measured. Factor analysis confirmed the constructs as three distinct factors.

Evaluations. Subjects' responses to two sets of three 7-point bipolar scales (good-bad; favorable-not favorable; positive-negative) were averaged to provide measures of ad (Aad) and brand (Abrand) evaluation. The coefficient alpha scores for Aad and Abrand were .90 and .94, respectively. The average of two 7-point bipolar scales (likely-unlikely; probable-not probable) served as the measure for the purchase intention (PI) (correlation=.96).

Manipulation Checks. Subjects were asked to rate the extent to which the visual and headline conveyed the same or similar meanings (same-different; similar-dissimilar; correlation=.72) (cf. Houston, Childers and Heckler 1987) and the extent to which the headline and visual conveyed opposite meanings (opposite-not opposite; contradictory-not contradictory; correlation=.88).

RESULTS

A 3 X 2 between-subjects ANOVA revealed no differences across product class on the dependent measures. Collapsing across product class, quadratic trend analysis was used to test the nonmonotonic relationships proposed in hypotheses 1, 3, and 6. The remaining hypotheses were tested using one-way between-subjects ANOVA.

Manipulation Checks

The main effect of headline-visual consistency on similarity was significant (F[2,105]=22.90, p<.001). Planned contrasts revealed significant differences between all treatment conditions. As expected, the headline and the visual in the consistent advertisements were rated as significantly more similar in meaning ( =2.65, where 1 is similar) than the moderate ( =4.42) and extremely inconsistent advertisements ( =5.44). The main effect of headline-visual consistency on the opposite meaning conveyed in the headline and visual was also significant (F[2,105]=30.82, p<.001). As expected, the headline and the visual in the moderately inconsistent ads were perceived as conveying more opposite meanings ( =2.03, where 1 is opposite) than the extremely inconsistent ( =3.60) or consistent advertisements ( =5.29). Together with the pretest results, these findings provide strong evidence that each of the ads conveyed the intended level of headline-visual inconsistency.

Tests of Hypotheses

Table 1 summarizes the analysis of variance findings. Hypothesis 1 proposed that ad evaluations would be more positive for moderately inconsistent advertisements than for consistent and extremely inconsistent ads. The results of the quadratic trend analysis revealed the expected inverted U-shaped main effect of headline-visual consistency on ad evaluation (F[1,103]=20.33, p<.001). Planned contrasts indicated that moderately inconsistent ads were evaluated significantly more favorably than the consistent or extremely inconsistent advertisements. The evaluations of consistent and extremely inconsistent ads were not significantly different. Hypothesis 1 was supported.

Since inconsistent conditions stimulate more elaborative processing than consistent conditions, hypothesis 2 posited that moderate and extremely inconsistent ads would generate more combination thoughts and more ad-related thoughts. ANOVA using combination thoughts as the dependent measure revealed a main effect of headline-visual consistency (F[2,105]=11.60, p<.001). As predicted, planned contrasts revealed that significantly more combination thoughts were generated for the moderate and extremely inconsistent than for the consistent ads. ANOVA also revealed a main effect of ad type (F[2,105]=3.85, p<.024) on the number of ad-related thoughts. As expected, planned contrasts indicated that significantly more ad-related thoughts were generated for the moderate and extremely inconsistent than for the consistent ads. Hypothesis 2 was supported.

Hypothesis 3 proposed that if moderate but not extreme inconsistencies can be resolved and the resolution process itself produces positive affect, as compared to consistent and extremely inconsistent advertisements, moderately inconsistent ads will generate more positive combination thoughts, more positive ad-related thoughts, and be perceived as more interesting. ANOVA results showed main effects of ad type on the number of positive combination thoughts [F(2,105)=7.00, p<.001], on the number of positive ad-related thoughts [F(2,105)=4.87, p<.010], and on perceived interestingness (F[2,105]=3.58, p<.031). As hypothesized, planned contrasts indicated that moderately inconsistent ads generated significantly more positive combination thoughts than either the consistent or extremely inconsistent conditions. Also, planned contrasts indicated that moderately inconsistent ads produced significantly more positive ad-related thoughts than the consistent ads but, contrary to expectations, not the extremely inconsistent ads. In addition, moderately inconsistent ads were perceived as more interesting than consistent but not extremely inconsistent advertisements. Hypothesis 3 was partially supported.

Because extreme inconsistencies cannot be resolved, hypothesis 4 suggested that, as compared to consistent and moderately inconsistent ads, extremely inconsistent ads will generate more confusion-related combination thoughts, more confusion-related thoughts overall, be perceived as most irritating and the least comprehensible. ANOVA results supported the main effects of ad type on confusion-related combination thoughts [F(2,105)=4.80, p<.010], on confusion related thoughts (F[2,105]=17.91, p<.001), perceived irritation (F[2,105]=5.87, p<.004), and ease of comprehension (F[2,105]=39.27, p<.001). As expected, extremely inconsistent ads generated more confusion-related combination thoughts and more confusion ad-related thoughts overall than either the consistent or moderately inconsistent ads. In addition, extremely inconsistent ads were perceived as more irritating than moderately inconsistent but not consistent advertisements. Finally, extremely inconsistent ads were perceived as less understandable than consistent and moderately inconsistent ads. Interestingly, planned contrasts also revealed that moderately inconsistent ads were perceived to be significantly easier to understand than consistent advertisements. Hypothesis 4 received partial support.

Hypothesis 5 was proposed to more directly test the mediation effects of the underlying cognitive processes on ad evaluation. Because references to the particular headline-visual combination may be the best measure of the underlying processes, an index of cognitive responses was formed (positive combination-related thoughts minus confusion-related combination thoughts) to capture both the positive and negative indicators of the inconsistency resolution process. Using the index as a covariate, the results of a quadratic trend analysis suggested that the cognitive response index partially mediated the effect of ad type on ad evaluation. Although the index had a significant effect on ad evaluation [F(1,102)=4.09, p<.045], the nonmonotonic impact of ad type on ad evaluation, while reduced, remained significant [F(1,102)=6.63, p<.002]. Hypothesis 5 was partially supported.

Hypothesis 6 addressed the relationship between level of consistency, brand evaluation (Abrand), and purchase intention (PI). As expected, quadratic trend analysis revealed a significant nonmonotonic relationship between ad type, Abrand (F[1,103]=19.49, p<.001) and PI (F[1,104]=6.68, p<.011). As shown in table 1, moderately inconsistent advertisements resulted in (1) more positive brand evaluations than consistent and extremely inconsistent advertisements, and (2) greater purchase intentions than extremely inconsistent but not consistent advertisements.

TABLE

ANOVA SUMMARY OF DEPENDENT MEASURES

Using Aad as a covariate, quadratic trend tests were run to assess the mediating role of Aad on the relationship between ad type on Abrand and PI. The results of the analyses indicated that (1) Aad partially mediated the effect of ad type on Abrand (F[1,102]=24.51, p<.001), but the nonmonotonic relationship of ad type on Abrand, while reduced, remained significant (F[1,102]=6.06, p<.015), and (2), together, Aad (F[1,100]=16.97, p<.001) and Abrand (F[1,100]=11.47, p<.001) completely mediated the nonmonotonic relationship between ad type and PI (F[1,100]=.28, n.s.). Hypothesis 6 was partially supported.

DISCUSSION AND IMPLICATIONS

Consistent with Mandler's model, a quadratic trend test revealed a nonmonotonic relationship between the level of headline-visual consistency, ad and brand evaluation, and purchase intention. Moderate levels of headline-visual inconsistency produced more positive ad and brand evaluations than ads with consistent and extremely inconsistent headline-visual combinations. Moreover, the moderately inconsistent headline-visual combination generated greater purchase intentions than the extremely inconsistent headline-visual combination.

Analyzing consumers' cognitive responses revealed that inconsistencies stimulate elaborative processing, as compared to consistent headline-visual combinations, moderate and extremely inconsistent ads generated more combination thoughts and a greater number of ad-related thoughts. In a study of product-related processing, Sujan (1985) reported a similar result. In line with our results, consistent information was processed automatically and without extensive elaboration. In addition, Mandler asserts that the absence of elaboration reduces the extremity of affect. In support, we found that ad evaluations for consistent advertisements were less extreme than the evaluations in both inconsistent conditions.

While inconsistencies may stimulate elaborative processing, the consumer's ability to resolve the inconsistency influences the nature of the processing and subsequent valence of evaluation. Because moderate headline-visual inconsistencies can be resolved and the resolution process itself is rewarding, subjects produced more positive combination thoughts and more positive ad-related thoughts than consistent or extremely inconsistent headline-visual combinations. Further, moderately inconsistent ads were perceived as more interesting than consistent ads, providing additional evidence that consumers enjoy deciphering moderately inconsistent messages. Because the meanings of the headline and visual in the extremely inconsistent advertisements cannot be resolved, subjects in the inconsistent ad conditions generated thoughts reflecting confusion and frustration.

In line with related research, we found that extremely inconsistent verbal-visual combinations were less comprehensible than more consistent combinations (e.g., Heckler and Childers 1992). However, we also found that moderately inconsistent headline-visual combinations were more easily comprehended than consistent combinations. This finding may conflict with Heckler and Childers' finding that expected-relevant information (our consistent condition) was easier to understand than unexpected-relevant information (our moderately inconsistent condition). However, literary theory suggests that the use of opposites (i.e., irony) enhances learning and comprehension (Williams and Lilly 1985). Further, opposites are more easily recalled and comprehended than synonyms (Colombo and Williams 1990), and research on inference formation (Kardes 1988) asserts that a consumer's active participation in interpreting an ad with incomplete information results in better recall than when complete information is provided. Recall is often used as an indicator of comprehension (cf. Houston, Childers and Heckler 1987). Moderate inconsistencies invite the consumer to uncover the underlying headline-visual connection that makes the ad comprehensible. Consistent combinations provide the connections for the reader. Finally, the results of the covariance analysis provided partial support that the resolution process mediated the effect of ad type on ad evaluation. Further, the same nonmonotonic pattern of results of ad type on ad evaluation was found for brand evaluation and purchase intentions.

Inconsistency results from the structural relationship among elements (i.e., the headline, visual, and copy) within an ad. The manipulation of inconsistency, however, is not possible without also potentially changing the content of the ad. By using extensive pretesting and manipulation and confound checks, we attempted to control for this possibility. Alternative explanations due to content changes, however, can never be ruled out entirely (McQuarrie and Mick 1992). Moreover, to enhance internal validity, ad copy was omitted. The inclusion of ad copy may alter the processing of headline-visual inconsistencies. However, since only a fifth of those who note an ad in a magazine actually read the copy (Pollay and Mainprize 1984), ad copy may have limited impact on results. Our respondents' cognitive responses revealed that none of our test ads were perceived as unrealistic. Finally, the laboratory setting differs dramatically from the natural viewing context. Future investigations should extend this research by combining multiple methodologies (e.g., McQuarrie and Mick 1992).

REFERENCES

Berlyne, Daniel E. (1971), Aesthetics and Psychobiology, New York: Appleton-Century-Crofts.

Blasko, Vincent J. and Michael P. Mokwa (1986), "Creativity in Advertising: A Janusian Perspective," Journal of Advertising, 15 (December), 43-50, 72.

Bock, Michael and Bernhard Milz (1977), "Pictorial Context and the Recall of Pronoun Sentences," Psychological Research, 39 (1), 203-220.

Buckholz, Laura M. and Robert E. Smith (1991), "The Role of Consumer Involvement in Determining Cognitive Response to Broadcast Advertising," Journal of Advertising, 20 (1), 4-17.

Childers, Terry L. and Michael J. Houston (1984), "Conditions for a Picture Superiority Effect on Consumer Memory," Journal of Consumer Research, 11 (September), 531-563.

Colombo, Lucia and John Williams (1990), "Effects of Work and Sentence-Level Contexts Upon Word Recognition," Memory & Cognition 18(2), 153-163.

Corbett, Edward P. J. (1990), Classical Rhetoric for the Modern Student, New York: Oxford University Press.

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

Heckler, Susan E. and Terry L. Childers (1992), "The Role of Expectancy and Relevancy for Verbal and Visual Information: What is Incongruency?" Journal of Consumer Research, 18 (March), 475-492.

Houston, Michael J., Terry L. Childers, and Susan E. Heckler (1987), "Picture-Word Consistency and the Elaborative Processing of Advertisements," Journal of Marketing Research, 26 (November), 359-69.

Kardes, Frank R. (1988), "Spontaneous Inference Processes in Advertising: The Effects of Conclusion Omission and Involvement on Persuasion," Journal of Consumer Research, 15 (September), 225-33.

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

Mandler, George (1984), Mind and Body, New York: W. W. Norton & Company.

McQuarrie, Edward F. and David Glen Mick (1992), "On Resonance: A Critical Pluralistic Inquiry into Advertising Rhetoric," Journal of Consumer Research, 19 (September), 180-197.

Meyers-Levy, Joan and Alice M. Tybout (1989), "Schema Congruity as a Basis for Product Evaluation," Journal of Consumer Research, 16 (June), 39-54.

Miniard, Paul W., Sunil Bhatla, and Randall L. Rose (1990), "On the Formation and Relationship of Ad and Brand Attitudes: An Experimental and Causal Analysis," Journal of Marketing Research, 27 (August), 290-303.

Mittal, Banwari (1990), "The Relative Roles of Brand Beliefs and Attitude Toward the Ad as Mediators of Brand Attitude: A Second Look," Journal of Marketing Research, 27 (May), 209-219.

Peeck, J. (1974), "Retention of Pictorial and Verbal Content of A Text with Illustrations," Journal of Education Psychology, 66 (6), 880-888.

Pollay, Richard W. and Steve Mainprize (1984), "Headlining of Visuals in Print Advertising: A Typology of Tactical Techniques," in Proceedings: American Academy of Advertising, Donald R. Glover, ed. Denver: American Academy of Advertising, 24-28.

Stein, Barry S., Karla F. Brock, Donny R. Ballard, and Nancy J. Vye (1987), "Constraints on Effective Pictorial and Verbal Elaboration," Memory & Cognition, 15 (4), 281-290.

Sujan, Mita (1985), "Consumer Knowledge: Effects of Evaluation Strategies Mediating Consumer Judgements," Journal of Consumer Research, 12 (June), 31-46.

Williams, Richard N. and John Paul Lilly (1985), "The Effects of Oppositional Meaning in Incidental Learning: An Empirical Demonstration in the Dialectic," Journal of Mind and Behavior, 6 (Summer), 419-434.

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