Visual Imaging Ability As a Mediator of Advertising Response

John R. Rossiter, University of Pennsylvania
Larry Percy, Gardner Advertising Company
ABSTRACT - This study examines the role of visual imaging ability as a mediating variable in consumer response to advertising. A model is proposed in which visual imaging ability is hypothesized to mediate the effects of both visual and verbal stimulus content in advertising through a process which we term visual reinforcement. Data are provided in support of the model and in support of the visual reinforcement hypothesis. Visual imaging ability appears to be a powerful and hitherto overlooked mediating variable in consumer information processing.
[ to cite ]:
John R. Rossiter and Larry Percy (1978) ,"Visual Imaging Ability As a Mediator of Advertising Response", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 621-629.

Advances in Consumer Research Volume 5, 1978      Pages 621-629

VISUAL IMAGING ABILITY AS A MEDIATOR OF ADVERTISING RESPONSE

John R. Rossiter, University of Pennsylvania

Larry Percy, Gardner Advertising Company

[The authors are indebted to the creative staff at Gardner Advertising Company for preparing the stimuli used in the study; to the research staff at Gardner for collecting the data; and to Frank Hayson of the Wharton School for computer assistance. Professors James Bettman, Kathy Lutz and Richard Lutz of UCLA provided valuable comments on the manuscript.]

ABSTRACT -

This study examines the role of visual imaging ability as a mediating variable in consumer response to advertising. A model is proposed in which visual imaging ability is hypothesized to mediate the effects of both visual and verbal stimulus content in advertising through a process which we term visual reinforcement. Data are provided in support of the model and in support of the visual reinforcement hypothesis. Visual imaging ability appears to be a powerful and hitherto overlooked mediating variable in consumer information processing.

INTRODUCTION

Advertising, in all media except radio, relies heavily on visual as well as verbal information to present the advertised product. Research on consumer response to advertising, however, has focused almost exclusively on verbal information processing: witness the plethora of advertising response studies in which verbalized beliefs, attitudes, recalled message points and so forth represent the main informational constructs. Most importantly, visual information processing has not been represented in any models of advertising response with the exception of one model developed for children by Rossiter (1976). The present study proposes a more elaborate model which examines the nature of visual information processing as it applies to advertising.

Researchers have not been blind to the importance of visual information per se. Studies of consumer spatial behavior in retail location research (for a review see Reynolds and Wells, 1977), of packaging, and of the effects of illustration size in print advertising (for a review see Hendon, 1973) are at least partially concerned with consumer response to visual information. However, these studies have employed S-R designs of an input-output type in which the nature of visual information processing has not been assessed.

The concept of information processing implies that something theoretically necessary if not directly accessible is going on inside the organism which operates on the incoming information and consequently affects the output response. Variables which tap these organismic processes constitute valid 0-variables in the basic S-O-R paradigm assumed by most information processing theorists and are to be distinguished from individual difference variables such as those of the demographic, psychographic or past product usage variety which may be correlated with output responses but which have not yet been shown to affect information processing in any causative or even specifiable manner. 0-variables which mediate verbal information include intelligence, anxiety, and sex (Hovland and Janis, 1959; McGuire, 1969). 0-variables which mediate visual information are beginning to be identified in the cognitive psychology literature. An extensive review can be found in Richardson (1969).

Most of the research on individual differences in visual information processing has been concerned with the mediating effects of these 0-variables on verbal learning (item recall and recognition) and motor learning (performance improvement after visualized practice). However, Richardson speculates that visually oriented 0-variables may also mediate the learning of affective (attitudinal) responses. But no one has yet demonstrated visual mediation of affective learning empirically nor developed a theoretical model that would predict it. This joint theoretical and empirical endeavor is highly relevant to advertising and provides the basis for present study.

The 0-variable examined in this study is visual imaging ability. Visual imaging ability may be colloquially defined as the individual tendency to "think in pictures'' (Richardson, 1969). Although everyone experiences pictorial imagery at some time or other, individuals vary widely in the extent to which they employ visual imagery as a habitual mode of thinking and simultaneously in the extent to which visual imagery occurs easily and spontaneously as a covert response to external stimuli (Paivio, 1969; Richardson, 1969).

THE MODEL

The theoretical model on which the present study is based is shown in Figure 1. The model holds a crucial implication for individual differences in visual imaging ability. Its predicts that visual imaging ability will mediate the processing of both visual and verbal stimuli. The hypothesized processes are explained below for verbal stimuli and then visual stimuli.

Verbal Stimuli

The initial and primary reaction to a verbal stimulus consists of a verbal comprehension response; that is, the verbal stimulus (word, sentence, or larger textual unit) is decoded and understood.

However, it has also been shown that verbal stimuli may simultaneously evoke visual imagery responses. Indeed, the imagery value of a verbal stimulus has consistently been shown to have a more powerful influence on verbal learning than verbal comprehension factors such as the meaningfulness of the stimulus (Paivio, 1969; 1971; Hulse, Deese and Egeth, 1975). Verbal stimuli having the highest imagery value are generally those of a more concrete nature. Concrete verbal stimuli are those which have clear, "real world" referents as opposed to vague, abstract referents. It is easier for most people, for example, to experience visual imagery in response to the word BEER than to the word BEAUTY.

The initial covert information processing response to verbal stimuli therefore has two hypothesized components as shown in Figure 1: a verbal comprehension response and, frequently though not always, a visual imagery response.

FIGURE 1

ADVERTISING RESPONSE MODEL FOR THE PRESENT STUDY

It is the subsequent responses, however, that are of most theoretical interest in the processing of persuasive information. The present model posits that if the verbally comprehended stimulus is recognized as a favorable one (e.g., the word GOOD vs. the word BAD) verbal reinforcement will occur; and that, likewise, if the visual imagery response to the verbal stimulus is a favorable one (e.g., a pleasant, perhaps personally involving mental picture) visual reinforcement will occur.

The verbal reinforcement phenomenon has solid empirical support. A long history of evidence beginning with a classic experiment by Staats and Staats (1957) has demonstrated that favorably evaluated words can function as verbal reinforcers. By associating objects or other words (e.g., the name of a product) with these verbal reinforcers attitudinal responses toward the object or word can be increased. Advertisers, of course, have always known this implicitly, hence their predilection for highly favorable words such as NEW or FREE in copy.

The visual reinforcement phenomenon in the processing of verbal stimuli has yet to be demonstrated. It has been shown by Paivio (1971) that the verbal learning superiority of high imagery value words occurs regardless of their connotative value; that is, "good" words are learned just as well as "bad" words providing their visual imagery values are equivalent. This at first would seem to indicate that verbal reinforcement is not involved in verbal learning but that some type of visual reinforcement, mediated through the word's evoked image, may be. However, "learning" in verbal learning and imagery studies has invariably been defined in terms of recognition or recall. This can be contrasted with the type of affective learning involved in the Staats-type experiments (and in the present study) where the dependent variable is an attitudinal or evaluative response. No one has yet investigated whether the visual imagery value of verbal stimuli relates to affective learning; that is, Paivio's finding that imagery independently increases recognition and recall can not be taken as evidence that imagery could also increase attitude because the latter type of dependent variable has not been employed in imagery studies.

Demonstration of visual reinforcement of attitudes through the use of verbal stimuli would be extremely significant for advertisers. In contrast with verbal learning researchers who try to select connotatively neutral verbal stimuli, advertisers deliberately select affectively loaded verbal stimuli to describe products. If these affectively loaded stimuli not only produce verbal reinforcement (of the GOOD type) but also visual reinforcement (of the pleasant, personally involving mental picture type) then such stimuli may constitute extra reinforcement for affectively learned responses such as attitude.

Visual Stimuli

If verbal stimuli can elicit visual imagery responses then visual stimuli should almost certainly do so. The model in Figure 1 assumes that the initial and primary reaction to a visual stimulus is some type of visual or iconic (picture-like) encoding which we have called a visual imagery response. However, as shown by the smaller arrow in the figure, a verbal reaction may also occur. Let us examine the visual reaction first.

A well known experiment by Shepard (1967) suggests that visual imagery always occurs in response to visual stimuli. Shepard's experiment is interesting because, although it was conducted in the context of experimental psychology, the stimuli employed happened to be magazine advertisements. Subjects in the experiment studied 612 illustrated magazine advertisements at their leisure. They were given a recognition test on 68 pairs of ads in which the pairs consisted of one ad from the originally studied set plus a new ad of the same type as the rest. The task was to identify which of each pair had been in the original set. On an immediate recognition test, subjects were 98.5% accurate. On a delayed test one week later subjects were still 90% accurate. Performance did not fall to the chance level (50%) until four months afterward. This normal but amazing recognition feat implies that people store visual images of every visual stimulus they pay attention to. The alternative, non-imagery explanation, that the subjects in Shepard's experiment generated verbal responses to the 612 advertisements which later aided them in the recognition test, is hardly convincing because this would imply that the subjects could accurately recall a list of 612 words in one trial, which simply cannot be done. Visual imagery therefore seems to be an inevitable response to attended-to visual stimuli.

People may also generate verbal responses to visual stimuli just as they may generate visual responses to verbal stimuli. However, our model suggests, and the evidence on cognition indicates, that intramodal responses (i.e., an initial imagery response to visual stimuli and an initial verbal comprehension response to verbal stimuli) are the typical and dominant ones even though cross-modal responses may also occur. Dominant intra-modal responses are illustrated by the larger arrows in Figure 1.

With visual stimuli, as with verbal stimuli, it is the subsequent responses that are in dispute. One school of thought acknowledges that the initial response to a visual stimulus may be a visual imagery response but that the subsequent responses are verbal. That is, the person mentally "describes" or interprets the visual stimulus in verbal terms. The verbal response then provides the basis for further effects such as recall or recognition. The best known theory in this verbal coding school is the "verbal loop hypothesis" (Glanzer and Clark, 1969). We have already argued that verbal labeling of this type would have difficulty accounting for visual recognition findings such as those of Shepard (1967). However, verbal labeling would appear to be a possibility with fewer visual stimuli than in Shepard's experiment. The verbal labeling viewpoint remains popular (Pylyshin, 1973) and stands as an alternative to the subsequent response interpretation in the present model.

The present model proposes that subsequent responses to visual stimuli may not pass through any "verbal loop" at all. It is hypothesized that the initial imagery response can serve as an internal stimulus which triggers further visual imagery. If this imagery is favorable, visual reinforcement will occur. [This view should be carefully distinguished from the imagery coding theory advanced by Bugelski (1970; 1977). Bugelski's imagery coding theory is the diametric opposite of the verbal coding theories of Glanzer and Clark and Pylyshin in that it holds that all coding is imaginal. However, Bugelski is not referring to a picture-like visual image when he uses the term "image" whereas we, in contrast, are. In this sense our view of imagery follows that of Richardson (1969) who defines visual imagery as a consciously experienced, quasi-perceptual event. We differ from Richardson, however, in that our model hypothesizes that visual imagery may also have reinforcing properties. In particular, subsequent visual imagery responses made to the initial iconic image are capable of assuming reinforcing qualities, hence our concept of visual reinforcement.]

The Nature of Attitudes

Several comments are necessary concerning the nature of the ultimate response in affective learning. In the present case this is an attitude (see Figure 1). Up until this ultimate response our model is quite clearly a modality specific "dual trace" model of the type advocated by Paivio (1971). Our model hypothesizes that both visual and verbal stimuli can produce verbal and visual "traces" in the form of covert stimuli and responses during information processing. However, we do not maintain that the ultimate effect, i.e., an attitude, is modality specific. Following Anderson and Bower (1974) and to some extent Pylyshin (1973) we regard the ultimate retained effect as being abstract and conceptual in nature. In retaining a response such as an attitude there is no requirement that the person also retain the original verbal or visual learning experience in toto. Just as it is possible for people to undergo an attitude change without recalling the original verbal content of the persuasive message (Greenwald, 1968; Calder, Insko and Yandell, 1974) it is similarly possible to undergo a visually mediated attitude change without recalling the original visual imagery that produced it.

Although attitudes themselves are not modality specific, attitudes learned under verbal and visual reinforcement conditions should be retrievable via appropriate cues in either modality: external stimuli similar to those under which the original learning occurred. Consequently, the product attitude response may be elicited by a visual cue, as when the product is seen again in another advertisement or in a store, or by a verbal cue, as when the product name is heard or the name read while filling out an attitude questionnaire.

Finally, the conceptualization of attitude as an automatic, involuntary, elicited response implies that visual (and verbal) reinforcement in our model operates via a classical conditioning process. All that is necessary is the contiguous association of a brand or product stimulus with a visual stimulus. The occurrence of subjectively pleasant visual imagery within the individual beyond the initial iconic encoding response then becomes a sufficient condition for visual reinforcement to occur via classical conditioning. The classical conditioning possibilities of visual imagery were first recognized by Leuba (1940) who referred to images as "conditioned sensations." This view was later elaborated by Mowrer (1960). Since 1960, with attention turning from behaviorism to the newer cognitive theories, psychologists have not addressed the potential classical (UCS) reinforcing value of visual imagery. Instead, behavioristic research on imagery has tended to follow the paradigm developed by Wolpe (1958) which focuses on the operant reinforcing value of visual imagery of a voluntary type (Cautela, 1977). Our concern, however, is with "involuntary" visual imagery and thus the classical conditioning paradigm is appropriate.

Visual Imaging Ability and Visual Reinforcement

Our model requires us to demonstrate that a visually mediated type of reinforcement occurs independently of the "standard" verbally mediated reinforcement. Verbal coding theorists could argue that all reinforcement is verbal, due in the case of visual stimuli to verbal labeling (see dashed line in Figure 1).

Examination of individual differences in visual imaging ability provides a test of the competing hypotheses. Individuals who are high in terms of visual imaging ability should be more likely to generate visual imagery to any stimulus with visual or verbal content and therefore to show greater affective learning in general because of the dual reinforcement idea. But isolation of visual and verbal content provides a more crucial test. If visualizers (this convenient shorthand will be used hereafter) show greater affective learning only in response to verbal stimuli it could be argued that they are simply generating more initial visual imagery responses to words which are then labeled and translated into verbal reinforcement. However, if it could be demonstrated that visualizers also show greater affective learning in response to visual stimuli then a clear-cut case for the validity of visual reinforcement would emerge. This is because the initial imagery response to visual stimuli is common to all individuals (cf. the Shepard experiment). Increased reinforcing effects among high visualizers would therefore make it almost impossible to avoid the conclusion that these individuals are deriving extra visual reinforcement (further spontaneously generated pleasant visual imagery) from the same initial image because any verbal reinforcement from this image would be constant. An extreme counter-argument, that this further imagery is nonreinforcing per se but is verbally labeled and then reinforcing, seems unlikely. In any event, such an argument would not vitiate the reinforcing potentiality of visual information processing in its own right.

THE STUDY

The test of the visual reinforcement hypothesis (and of the advertising response model in general) was devised by exposing consumers to two new brands of beer via separate print advertisements. The brand names were fictional and were equated beforehand to ensure that names per se had no effect on attitude. One brand was presented via visually oriented (strong visual) advertising: a large picture of the product with small verbal copy underneath. The other brand was presented via verbally oriented (weak visual) advertising: large verbal copy with a small picture of the product underneath. A "no picture" control condition was rejected as being atypical of beer advertisements.

Experimental conditions were further arranged so that half the consumers, in a rotated 2x2 design, received advertisements in which the verbal copy was highly concrete in nature (superlative and explicit product claims) and half received advertisements in which the verbal copy was highly abstract in nature (superlative but vague product claims).

Consumers were asked to rate their attitude toward the brand after each exposure as a measure of affective learning produced by the advertisements. Beer was selected as the product category because previous research (e.g., Allison and Uhl, 1964; Woodside, 1972) has shown that brand attitude engendered by advertising is a very good precursor of purchase behavior. Too often a haphazard selection of dependent variables such as beliefs, attitudes, or recall is made. As argued by Rossiter and Percy (in preparation) the dependent "advertising effect" variable should always be dictated by an a priori advertising response model applicable to the particular product category, consumer segment, and brand choice situation. The latter consideration (see also Wright, 1976) makes recall, for example, an unsuitable dependent variable for the beer category except, perhaps, in restaurant ordering situations where no verbal or visual cues are present. For most applications concerning beer advertising, attitude is the appropriate dependent variable.

The central issue in this study was not just whether affective learning would differ by stimulus condition but whether it would also differ predictably across individuals. A measure was therefore taken of visual imaging ability as a predictor variable. For comparison purposes, a nonvisual predictor variable, intelligence, was also included and sex differences were assessed. The main theoretical predictions, however, were: (1) visualizers would show greater affective learning than other consumers because of the greater likelihood of spontaneous occurrence of visual reinforcement from both the verbal and the visual stimulus content in advertising; (2) that this individual difference should be accentuated when the visual stimulus content of advertising favors visual imagery formation, i.e., contains strong visual emphasis; and similarly (3) when the verbal stimulus content of advertising also favors visual imagery formation, i.e., contains concrete copy. Each of these effects would provide evidence in support of the visual reinforcement component of the model.

METHOD

Subjects. Participants in this study were 88 adults (44 men and 44 women) recruited from a typical Midwestern suburban shopping center. All respondents were beer drinkers and an additional attempt was made to ensure a range of ages in the sample in case favored beer product attributes should vary as a function of age.

Materials and design. Four print advertisements representing the possible 2x2 combinations of visual versus verbal emphasis and concrete versus abstract copy were constructed professionally by the staff of a major national advertising agency. Visually oriented advertisements consisted of (a) a brand name at the top of the page, (b) a large picture of a mug of beer bearing the brand label and (c) concrete or abstract copy beneath. Verbally oriented advertisements consisted of (a) a brand name at the top of the page, (b) concrete or abstract copy as the main body of the advertisement and (c) a small picture of a mug of beer bearing the brand label beneath. Brand names (to serve later as the product cue) were in very large type of identical size for the visually and verbally oriented advertisements. Pictures and copy were also identical between versions except that the visually oriented advertisements used a 3:1 picture to copy size ratio whereas the verbally oriented advertisements reversed this ratio. Print size was adjusted slightly between concrete and abstract versions so that the total space allocated to copy was constant within visually and verbally oriented advertisements respectively.

The design of the study required each respondent to be exposed to both the visually oriented and verbally oriented advertising stimuli and to the concrete and abstract copy in order that visual imaging ability could be applied as a predictor on a within-subject basis. This meant that each respondent could see a visually oriented advertisement and a verbally oriented advertisement but, obviously, the brand names and copy content would have to differ.

The copy manipulation was taken care of by the concrete-abstract manipulation: each respondent simply saw the visually oriented advertisement with either concrete of abstract copy then the verbally oriented version with the complementary type of copy; this, and also the order of exposure for visually versus verbally oriented advertisements was rotated across respondents in a balanced design.

The brand name factor required the selection of two different brand names so that two different advertisement types could be rated by each respondent. To make the learning task as realistic as possible it was decided to use realistic brand names rather than abstract symbolic designations such as X or Y. The two brands of beer were introduced as new Bavarian imports (see copy below). Accordingly it was thought that Germanic sounding words would be appropriate as brand names without having immediately meaningful denotation to most consumers. The purpose would then be to select two Germanic sounding words whose connotative or affective associations were as equal and as neutral as possible so that the names alone would not influence product attitude. Nine two-syllable nouns were selected from a German dictionary. These were arranged in two randomly ordered lists. A separate pretest sample of 60 consumers (30 men and 30 women) drawn from the same population as those in the main study were then asked to rate each word on a scale of 1 to 7 in terms of how they "like [7] or dislike [1] each name as a brand name for beer." This task was explained by a short scenario which claimed that a Bavarian beer company was considering several brand names for a new import entry to the U.S. market. Two names, BAUER and LAUFER, were selected on the basis of this pretest as being most equal and closest to the neutral affective rating position. Mean affective ratings for each of the two names were identical (3.8 for men and 3.7 for women) as were their standard deviations (2.0 and 2.1 respectively). By a coin toss, the name BAUER was used in the visually oriented advertisements and LAUFER in the verbally oriented advertisements.

The visual stimuli selected for inclusion in the advertisements consisted of identical pictures of a "foaming mug of beer." The stimuli varied only in size, location, and brand name, as described earlier. The purpose of this selection was to ensure a fairly unitary stimulus (as far as intuitive judgment is concerned since, unlike verbal stimuli, there is no standard taxonomy for visual stimuli) so that the attentional focus, which is presumably the basis of the initial imagery response, would be singular or wholistic for all respondents. Theoretical complications would arise if the visual stimulus were multidimensional. A picture of people drinking beer or a beer bottle in an environmental setting might provoke differential attention to various substimuli within the visual display. This possibility was virtually removed with the "mug shot" which at the same time represented a visual stimulus that would not be atypical of a beer advertisement. The verbal stimuli in the advertisements consisted of either concrete or abstract copy. Three beer product attributes were selected for inclusion in the copy: overall quality, taste, and price. These three attributes have been found in proprietary research to be the main determinants of beer brand attitudes. The limitation to three attributes ensured that verbal processing to generate an attitude could easily take place. The use of five or more attributes might exceed short-term memory capacity (7+2) and result in differential attribute focus across respondents.

The concrete copy was designed to be superlative and explicit, of the "factual documentation" type. The text for the concrete copy was:

BAVARIA'S NUMBER 1 SELLING BEER FOR THE LAST 10 YEARS

WINNER OF 5 OUT OF 5 TASTE TESTS IN THE U.S. AGAINST ALL MAJOR AMERICAN BEERS AND LEADING IMPORTS

AFFORDABLY PRICED AT $1.79 PER SIX-PACK OF 12 OZ. BOTTLES

The $1-79 price was determined from the pretest by asking respondents to estimate what "affordably priced" would mean for the new Bavarian import and then computing the mean estimate.

The abstract copy was designed to present the same copy points in superlative but vague form of the more "emotional" type. The text for the abstract copy was:

BAVARIA'S FINEST BEER

GREAT TASTE

AFFORDABLY PRICED

The concrete copy is longer than the abstract copy. This is necessitated by the inclusion of sufficient information to provide an explicit referent for each attribute claim.

Procedure. Consumers participating in the study first answered some preliminary screening questions and then answered the 15-item Visualizer-Verbalizer Questionnaire (VVQ) developed by Richardson (1977) as a measure of visual imaging ability. The VVQ has been shown to have very high convergent and discriminant validity with other measures of visual imagery, including both other self-report and also physiologically based measures of right- and left-brain processing. Test-retest reliabilities to the order of .90 have been obtained for the VVQ although its internal consistency reliability has not been established (see White, Sheehan and Ashton, 1977). The range for this test is 0 to 15. The VVQ is scored so that high scores designate individuals whose characteristic mode of thinking is exclusively visual; low scores designate individuals whose characteristic mode of thinking is exclusively verbal; with dual visual and verbal types tending toward the middle of the score distribution. This scoring method was ideal for present purposes in that visual imagery as a predictor is unconfounded by dual types of processing.

Each respondent was then shown the first advertisement which was either visually or verbally oriented and contained either concrete or abstract copy, depending on the predetermined experimental condition. After the respondent had finished looking at the advertisement, he or she was asked, "As best you can judge, based on this advertisement, how would you rate this brand?" Respondents were handed a card on which the name BAUER BEER or LAUFER BEER, respectively, was printed followed by four, 7-point bipolar rating scales. The rating scales--good-bad, inferior-superior, unpleasant-pleasant, and interesting-boring, with the end-adjectives in this order--were selected from Osgood, Suci and Tannenbaum (1957) and Fishbein and Ajzen (1975) for their high loadings on the semantic evaluation factor. The scales were scored on a 7-point -3 to +3 system with the sum (range: -12 to +12) constituting the overall affective measure of brand attitude, the dependent variable in the study. This type of multi-item index has been shown elsewhere (Fishbein & Ajzen, 1975) to be a highly valid and reliable measure of attitude. The process was then repeated with the second advertisement assigned to each respondent's experimental condition.

Finally, each respondent completed a 3-minute, timed, Number Series test. The Number Series test (Lumsden, 1959) is a highly reliable and unidimensional measure of general intelligence which correlates highly with standard I.Q. measures. The score range is 0 to 22 in terms of number correct. Its focus on numerical rather than verbal or visual stimuli makes it eminently suitable as an alternative predictor variable in the present study.

RESULTS

The four advertisements employed in the study revealed strong stimulus effects on brand attitude ratings (Figure 2). This conclusion is warranted when it is realized that , in the pretest, the brand names alone elicited neutral attitudes. Visually oriented advertisements were more effective in inducing positive brand attitudes than verbally oriented advertisements (p < .01) and concrete copy was similarly more effective than abstract copy (p < .05). The presence of significant interaction is also evident from Figure 2, with visual emphasis and concrete copy combining to elicit the strongest attitudes. This result would be predicted by the visual reinforcement hypothesis and these overall results lend strong support to our dual reinforcement advertising model.

However, our main concern in this paper is to assess the validity of the visual reinforcement component of the model. This requires an examination of individual differences in affective learning. Accordingly, brand attitude scores were regressed against the three individual difference predictor variables, visual imaging ability (WQ), intelligence (Number Series) and sex, plus copy type (concrete versus abstract). The correlations between the predictor variables did not differ significantly from zero (p > .05) which eliminates the problem of multicollinearity. The results are shown in Table 1. There are three dependent variables to be considered. The first is total attitude, computed as the sum of visually and verbally induced attitudes, which provides a test of the hypothesis that visualizers will reveal greater affective learning from advertising in general. Copy type was not employed as a predictor variable for total attitude as this variable is automatically summed into the total attitude score. The second two dependent variables represent brand attitudes induced by visually and verbally oriented advertisements respectively, recalling that each respondent rated one brand under each condition.

FIGURE 2

BRAND ATTITUDE RATINGS BY STIMULUS CONDITION

Looking first at the total attitude results it is seen that the only significant predictor is visual imaging ability (F = 5.32, p < .05; shown to be stable in randomly split samples). Neither sex nor intelligence revealed significant mediational effects. The regression coefficient for visual imaging ability indicated that visualizers showed the greatest degree of affective learning. This finding supports the first hypothesis of the study. The bivariate correlation between visual imaging ability and total attitude was r = .26 (p < .05) which indicates that individual differences in visual imaging ability accounted for about 7% of the variance in attitude scores on an overall basis.

Looking next at the predictors of brand attitude induced by visually oriented advertising it is seen that copy type is a significant predictor (F = 11.83, p < .01; stable in randomly split samples) as is visual imaging ability (F = 8.58, p < .01; stable in randomly split samples). The sign of regression coefficient for copy type indicated that the former result was due to the previously mentioned interaction favoring visually oriented advertising with concrete copy. The sign of the regression coefficient for visual imaging ability indicated that, as predicted, visualizers showed significantly greater affective learning under visually oriented presentation conditions than other individuals. The bivariate correlation in this case is r = .49 (p < .01) which indicates that individual differences in visual imaging ability accounted for about 25% of the variance in visually induced attitude scores. This finding supports the second hypothesis of the study.

TABLE 1

REGRESSIONS OF BRAND ATTITUDE RATINGS WITH PREDICTOR VARIABLES

However, in view of the apparent interaction between visually oriented advertising and the concrete versus abstract copy dimension a further analysis was conducted by running separate regressions for the 44 respondents who received the visually oriented advertising with concrete copy and the other 44 respondents who received the visually oriented advertising with abstract copy. These results are shown in Table 2. Interestingly, the relationship between visual imaging ability and brand attitudes induced by visually oriented advertising appears to hold only for the combination of visually oriented advertising with abstract copy (F = 12.49, p < .01; stable in randomly split samples). The only significant predictor of brand attitude induced by visually oriented advertising with concrete copy is sex (F = 7.31, p < .05). However, this F value was not stable in a randomly split sample test and should be regarded with caution. Nevertheless, the regression coefficient for the sex variable indicated that women showed significantly greater affective learning from the visual-concrete combination than did men. Implications of these results for the visual reinforcement hypothesis are addressed in the discussion.

Returning to Table 1 and looking finally at the predictors of brand attitude induced by verbally oriented advertising it is seen that none of the predictors had significant univariate F-values. The absence of a significant effect for copy type indicates that visual reinforcement is not differentially triggered in visualizers by concrete copy alone. This finding counts against the third hypothesis of the study.

TABLE 2

REGRESSIONS OF VISUAL BRAND ATTITUDE RATINGS FOR CONCRETE VERSUS ABSTRACT - COPY

DISCUSSION

The overall results produced impressive evidence in favor of the dual reinforcement model of advertising effects. That is, across all individuals, the use of a visually oriented layout and concrete verbal copy--the two factors hypothesized to maximize visual reinforcement--generated the strongest affective learning both separately and, more powerfully, in combination.

However, the main purpose of this study was not to test the model so much as to test the validity of the visual reinforcement component of the model. The validity tests required that individuals who possess high visual imaging ability show differentially greater affective learning (1) under all stimulus conditions and especially (2) under conditions of visually oriented advertising and (3) concrete copy content. The results supported the first two of these tests but not the third. Exclusive visualizers were superior to other individuals in terms of total affective learning and in terms of affective learning induced by visually oriented advertising. But visualizers did not show any superiority to other individuals in terms of affective learning induced by concrete advertising copy.

One post hoc interpretation of this latter result is that visualizers become so preoccupied with the visual elements of advertising that they pay less attention to the copy. This may be accentuated when the copy is longer--as it was in the concrete case. The finding that visualizers showed greater affective learning with the shorter, abstract copy under the strong visual stimulus condition (Table 2) would support this "distraction" hypothesis.

The emergence of a possible sex difference in Table 2 also merits discussion. [The sex difference is described as "possible" because, as noted, the associated F-value was not stable in a split sample test.] The sex difference occurred only in response to the strong visual stimulus-concrete verbal copy combination. Visual imagery is thought to be an activity which occurs, like other spatial processing, in the right brain hemisphere. Verbal comprehension, and other verbal reasoning activities, are thought to occur in the left brain hemisphere. Physiologists have found that women tend to be bilateral information processors whereas men tend to be unilateral processors (see White, Sheehan and Ashton, 1977). If so, there may be a multiplicative effect of visually and verbally induced imagery among women which leads to a stronger apparent preference for the visual-concrete combination. Women may be less "distracted" from longer advertising copy by visual stimuli in advertising because they are more able to process both types of information. This possibility has interesting implications for advertising directed to female audiences.

Also of interest was the failure of differences in intelligence to play a mediational role in the processing of persuasive information. Intelligence differences should be inversely related to persuasion when reception is constant (McGuire, 1969). However, this result has most commonly been obtained with more complex verbal messages than those employed here.

Methodological Considerations

One methodological consideration that is central to the conclusions drawn from this study concerns the possibility that the relationship between visual imaging ability and affective learning (positive attitude inducement) may be due to a "yea-saying" response bias. That is, individuals who checked the "yes" responses on the VVQ test may also have favored the positive extremes of the attitude scales used to measure affective learning.

The "yea-saying" possibility can be confidently ruled out. As noted in the Method section, the WQ test is scored such that "yea-sayers" would fall into the middle of the total score distribution. A person answering "yes" to most or all items would emerge as a mixed imager, not a visualizer. Moreover, the WQ test permits only dichotomous (yes-no) responses so that a tendency toward extreme responses is also precluded. The relationship between visual imaging ability and affective learning therefore remains unaffected by the measurement methods employed in the study.

A second consideration concerns the definition of "concrete" versus "abstract" verbal copy. The copy used in the concrete condition was certainly more specific and factually oriented than the copy used in the abstract condition. The idea here was to provide a singular and explicit referent for each attribute claim in the concrete copy. This is in line with the definition of concrete words as having single and explicit referents rather than multiple and often vague referents characteristic of abstract words (e.g., BEER versus BEAUTY).

However, the concrete copy required more words--regard-less of the individual words' concrete versus abstract status--than the abstract copy. Our attempt to develop concrete versus abstract text was thus confounded by text length.

Nevertheless, the differential effects of the two types of text, not so much for visualizers but for the sample as a whole, remain valid. The fact that visualizers were not affected as predicted by the imagery-producing concrete text suggests that the methodology may have been in error but not necessarily the theory. In practical terms, advertisers might still see the results as generally favoring longer, factually oriented copy over shorter but vaguer superlatives.

Further Applications

The identification of individual differences in visual imaging ability may itself have limited application to advertising unless such individuals can be easily located. To be of practical use, visual imaging ability would have to be shown to be correlated with standard media demographic variables or to be prevalent in the target segment for a given brand.

Individual differences in visual imaging ability have, however, made it possible to validate the existence of the process which we term visual reinforcement--a far more general phenomenon.

The concept of visual reinforcement has provocative implications for advertising. Visual reinforcement resulting from visual imagery may explain the superior learning produced by television commercials even when television commercials and print advertisements are apparently "equated" for visual and verbal content (Grass and Wallace, 1974). The multiple visual stimuli provided in a television commercial video should produce multiple initial imagery responses and thus increase the amount of visual reinforcement experienced by the viewer.

The present study focused on positive visual reinforcement. However, our model would predict that visual imagery can have negatively reinforcing and also punishing consequences for the effective learning of brand attitudes. Advertisements for pain relievers, for example, frequently employ the negative reinforcement principle. Negative verbal reinforcement from the copy or audio may be enhanced by negative visual reinforcement from the visual elements of the picture or video. Similarly, public service announcements entailing health warnings often employ the punishment principle. These too may be enhanced by visual imagery.

Visual reinforcement therefore has many unrecognized applications in advertising and probably in other areas of marketing communication as well. Indeed, any visual or verbal stimulus which reaches the consumer may produce visual reinforcement. The present model, although tested in an advertising context, should encourage examination of visual reinforcement in other marketing communication situations.

Finally, and not to be overlooked, is that visual imaging ability represents the first visually oriented individual difference variable to be identified in consumer research. Perhaps this will speed the development of better theories of visual information processing as an obviously relevant but surprisingly neglected aspect of consumer behavior.

REFERENCES

Ralph I. Allison and Kenneth P. Uhl, "Brand Identification and Perception," Journal of Marketing Research, 1 (August 1964), 80-85.

John R. Anderson and Gordon H. Bower, Human Associative Memory (New York: Wiley, 1974).

B. R. Bugelski, "Words and Things and Images," American Psychologist, 25 (November 1970), 1001-1012.

B. R. Bugelski, "Imagery and Verbal Behavior," Journal of Mental Imagery, 1 (Spring 1977), 39-52.

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

Joseph R. Cautela, "Covert Conditioning: Assumptions and Procedures," Journal of Mental Imagery, 1 (Spring 1977), 53-64.

Martin Fishbein and Icek Ajzen, Belief, Attitude, Intention, and Behavior, (Reading, Massachusetts: Addison-Wesley, 1975).

M. Glanzer and W. H. Clark, "Accuracy of Perceptual Recall: An Analysis of Organization," Journal of Verbal Learning and Verbal Behavior, 5 (January 1963), 289-299.

Robert C. Grass and Wallace H. Wallace, "Advertising Communication: Print Vs. TV," Journal of Advertising Research, 14 (October 1974), 19-23.

Anthony G. Greenwald, "On Defining Attitude and Theory," in Anthony G. Greenwald, Timothy C. Brock and Thomas M. Ostrom (Eds.), Psychological Foundations of Attitudes, (New York: Academic Press, 1968), 361-388.

Donald Wayne Hendon, "How Mechanical Factors Affect Ad Perception," Journal of Advertising Research, 13 (August 1973), 39-45.

Carl I. Hovland and Irving L. Janis, Personality and Persuasibility, (New Haven, Connecticut: Yale University Press, 1959).

Stewart H. Hulse, James Deese and Howard Egeth, The Psychology of Learning (Fourth edition, New York: McGraw-Hill, 1975).

Walter Kintsch, Learning, Memory, and Conceptual Processes, (New York: Wiley, 1970).

C. Leuba, "Images as Conditioned Sensations," Journal of Experimental Psychology, 26 (March 1940), 345-351.

William J. McGuire, "The Nature of Attitudes and Attitude Change," in Gardner Lindzey and Elliot Aronson (Eds.), The Handbook of Social Psychology: Volume 3, (Reading, Massachusetts: Addison-Wesley, 2nd edition, 1969), 136-314.

O. Hobart Mowrer, Learning Theory and The Symbolic Processes, New York: Wiley, 1960).

Charles E. Osgood, George Suci and Percy H. Tannenbaum, The Measurement of Meaning, (Urbana, Illinois: University of Illinois Press, 1957).

Allan Paivio, "Mental Imagery in Associative Learning and Memory," Psychological Review, 76 (May 1969), 241-263.

Allan Paivio, Imagery and Verbal Processes,(New York: Holt, Rinehart & Winston, 1971).

Z. W. Pylyshin, "What the Mind's Eye Tells the Mind's Brain: A Critique of Mental Imagery," Psychological Bulletin, 80 (July 1973), 1-24.

Fred D. Reynolds and William D. Wells, Consumer Behavior (New York: McGraw-Hill, 1977).

Alan Richardson, Mental Imagery, (New York: Springer, 1969).

Alan Richardson, "Verbalizer-Visualizer: A Cognitive Style Dimension," Journal of Mental Imagery, 1 (Spring 1977), 96-110.

John R. Rossiter, "Visual and Verbal Memory in Children's Product Information Utilization," in Beverlee B. Anderson (Ed.), Advances in Consumer Research: Volume 4, (Cincinnati, Ohio: Association for Consumer Research, 1976), 523-527.

John R. Rossiter and Larry Percy, Advertising Theory and Strategy (in preparation).

Carolyn K. Staats and Arthur W. Staats. "Meaning Established by Classical Conditioning," Journal of Experimental Psychology, 54 (July 1957), 74-80.

Roger N. Shepard, "Recognition Memory for Words, Sentences and Pictures," Journal of Verbal Learning and Verbal Behavior, 6 (February 1967), 156-163.

K. White, P. W. Sheehan and R. Ashton, "Imagery Assessment: A Survey of Self-Report Measures," Journal of Mental Imagery, 1 (Spring 1977), 145-170.

Arch G. Woodside, "A Shopping List Experiment of Beer Brand Images," Journal of Applied Psychology, 56 (November-December 1972), 512-513.

Peter L. Wright, "Conditional Consumer Choice Processes and Advertising Strategy," Paper presented at the American Marketing Association's Seventh Annual Attitude Research Conference, Hilton Head, South Carolina, (February 1976).

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