Attention, Memory, Attitude, and Conation: a Test of the Advertising Hierarchy


Esther Thorson, Annie Chi, and Clark Leavitt (1992) ,"Attention, Memory, Attitude, and Conation: a Test of the Advertising Hierarchy", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 366-379.

Advances in Consumer Research Volume 19, 1992      Pages 366-379


Esther Thorson, University of Wisconsin-Madison

Annie Chi, University of Wisconsin-Madison

Clark Leavitt, Ohio State University

One of the earliest theories of how advertising works was the hierarchy of effects model (Lewis, 1898). This model posits that the processes of attending to a commercial, learning and remembering its content, developing attitudes, and generating conative responses occur in a sequential causal chain. Lavidge and Steiner (1961) articulated a relatively recent version of the hierarchy model. They suggested that the steps in responding to advertising included: awareness of the ad, knowing what the brand has to offer, liking the brand, preferring the brand to competitors, conviction about the purchase, and purchase. Although advertising texts (e.g., Bovee & Arens, 1982; Wells, Burnett, & Moriarty, 1989) do not always specify a particular theory when they talk about how advertising works, they too suggest that first the ad must catch and hold attention. Then people must comprehend and remember its contents. Next, there must be attitudinal impact, and finally, conation.

Thus much of the scientific and heuristic literature of advertising theory adheres to the notion that responses to commercials fall into a single causal flow of effects. That is, no matter how complicated or simple, hierarchy models all derive from the central assumption that to be effective, advertising must "move people through a series of processing steps." In addition to positing a series of processing stages, hierarchy models also assume (Preston and Thorson, 1983) that to claim that a commercial has been effective, each step in the hierarchy must occur and it must occur in the specified order.

But is it true that an ad must propel people through these steps in order to reach conation? Graphic representation of this idea is shown in Figure 1. This model basically suggests that first there is attention to the ad. As a result, a memory for it is created. Based on the content of that memory, attitudinal responses develop, and as a result of these attitudes, conation eventually occurs.

To evaluate the model of Figure 1, we must examine the links between contiguous stages in the hierarchy. There is significant support in the literature that attitudinal and conative responses are correlated with each other. Indeed, a recent highly publicized finding sponsored by Advertising Research Foundation has shown that the best predictor of purchase is liking for ads. (Haley, 1990).

The link between memory and attitudinal responses is problematic. Gibson (1983) criticized the use of recall measures to evaluate ads because of the "common finding" that recall has little or no correlation with consumer choices in the marketplace. However, Stewart (1986) has presented compelling evidence that commercials that are well-recalled and comprehended and which present a message that clearly differentiates the brand, do score higher on persuasion. His findings suggest, in fact, that perhaps a number of variables that increase the impact of processing may lead to commercial memory being correlated with persuasion.

The relationship of attention to the other steps of the hierarchy is also unclear. MacKenzie (1986) demonstrated that attention to a brand attribute in an ad affected perceived importance of that attribute. His study did not look at the subsequent impact on attitude. Olney, Holbrook, & Batra (1991) showed a close link between attention and attitudinal responses when attention was measured as the amount of viewing time people give the ad when they have the opportunity to zip the ad instead of watching it. This study did not, however, look at memory. Thorson, Zhao, & Friestad (1988) showed strong links between attention and memory when attention was measured as the percent of time people actually watched the TV screen. There were not, however, significant links between attitude toward the ad or brand and the eyes-on-screen measure of attention. The authors suggested that attention affects memory because without watching, input is drastically reduced and there is nothing to remember. However, once people watched a commercial, their attitudes were influenced by the attributes of the ad, eliminating the impact of simply attending to the ad.

Attributes of Ads

Perhaps most importantly in the hierarchical schema, however, are attributes of the ad itself and how well each of these attributes predict the various responses to the ad: attention, memory, attitude and conation. Probably the most studied link in recent studies has been between ad attributes and attitudinal responses. Many studies have looked at the emotional attributes of ads and attitudinal responses (e.g., Stayman and Aaker, 1987; Edell and Burke, 1987; Thorson & Friestad, 1989). Holbrook and Batra (1987) showed that ad content measured as emotional, threatening, mundane, sexy, cerebral, and personal predicted both pleasure and arousal responses to ads and attitude toward the ad and brand. Other attributes of ads examined for their links with attitudinal impact have included informational and transformational appeals (Edell & Burke, 1987), and presence and type of illustrations (Mitchell, 1986).

There have also been studies of ad attributes and memory. One of the earliest attempts to predict recall of TV commercials was by Haller (l972). He used experts' judgments of four commercial dimensions: (1) the extent to which the visual aided the audio in getting the message across; (2) the stopping power of the message, (3) message clarity, and (4) the extent to which the message spoke personally to the viewer. Thorson and Snyder (l984) and Thorson (l983) used predictions from a psycholinguistic model of comprehension (Kintsch & van Dijk, 1979) to predict industry measures of recall for commercials. Measures such as the number and arrangement of idea units in the scripts predicted about 33 percent of the variance in recall scores.



In experimental studies of consumer responses to television commercials, Thorson & Friestad (1989) and Thorson & Page (1990) have shown that commercials that create emotion in viewers are better recalled than commercials neutral in emotional impact.

Other studies have examined the impact of ad attributes on both memory and attitude. Stewart and Furse (l985, l986) defined 153 variables and used them to predict Research System Corporation's three (ARS) measures of ad effectiveness: recall, key message comprehension, and persuasion. After removing product category effects, Stewart and Furse found variables that accounted for 13 to 26 percent of the variance in recall, 8 percent of the variance in key message comprehension, and 9 to 11 percent of the variance in persuasion. This aspect of the impact of ad attributes is important. McGuire (1969), another of the hierarchy theorists, pointed out that the different stages in the hierarchy are influenced by different aspects of ad structure. For example, attention may be increased by such ad variables as color, brightness contrast, camera movement, and gender of speakers. But memory may be increased by such variables as repetitiveness in the script, simplicity, and novelty. Although each stage in the hierarchy of effects operates contingently on the stage prior to it, it also is influenced by different ad attributes than the stages preceding and following it.

Indeed, the notion that different attributes of commercials differentially affect success in transiting different stages in the hierarchy has driven many attempts to build lists of attributes of commercials that will lead to "successful" consumer responses at various levels of the hierarchy--memory, persuasion, and liking for commercials. The search for a complete specification of all ad attributes relevant to eventual impact is termed "copy testing," and of course this area of advertising research also has a long history.



In the copytesting tradition, but emphasizing the hierarchical nature of responses to ads, Leavitt (1970) presented people watching ads with 52 adjectives that he found most used by consumers in focus groups. Leavitt expected that the dimensions that would emerge from factor analysis of the adjectives would reflect the four stages of the basic hierarchy model: attention, learning or memory, attitude change, and conative impact. Four dimensions did emerge. They included: Stimulation(amusing, energetic, novel, slow, and worn out); Relevance (convincing, credible, relatistic, irritating, and confusing); Gratification (agreeable, attractive, tender, warm), and Familiarity (well-known, new, saw before). In validation studies with the four dimensions, Leavitt found that Stimulation was related to attention. Relevance was related to attitude and conation. However, none of the four dimensions was related to memory.

The four dimensions from Leavitt's study have face validity; that is, they appear to be reasonable responses to commercials. They also find much support in the copy-testing literature (e.g., see Ju, Stout, & Leckenby's review (1990) of copy research). The items also seemed to relate to the critical stages of the hierarchy model. Thus it was the 52 items that made up Leavitt's original system that were employed in this study.

Models and Research Questions

The main thrust of the study reported here was to test the predictions of the quintessential hierarchy model of advertising effects. This model posits that ad characteristics will drive attention to the ad, which will in turn affect memory for the ad, and then attitude toward the ad and its brand, and finally, conative responses to the ad. Based on the uncertainties about the mediating relationships of these variables reflected the literature reviewed above, a second form of hierarchy was also deemed important to test. Specifically, the impressive literature demonstrating that memory and attitudes are not correlated suggests the model depicted in Figure 2. In this model, there are two routes of responses to ads: (1) the comprehension/learning route and (2) the evaluative route. In the learning route, ad characteristics drive attention, which in turn determines comprehension and memory. In the evaluative route, ad characteristics drive attention, which in turn determines attitude toward the ad, attitude toward the brand, and intention to purchase. Under this model, ads that are remembered are not necessarily those that create desire and intention to purchase.

A second thrust of the study was to ask whether the same form of the model was applicable both to ads that created emotion and those that did not. When an ad creates emotion in the viewer, the memory engram for the experience is enhanced over non-emotional conditions (Squires, 1986; Mishkin, Malamut, & Bachevalier, 1984). The engram is stronger in that recall likelihood is higher (Thorson & Friestad, 1989), and the engram is richer in the sense that viewers can report more information about the ad. The engram also contains residue from the experience of the emotion itself and thus is affectively charged. This means that the emotion itself can serve as a recall cue (Clark & Teasdale, 1985; Page, Thorson, & Heide, 1990). Given the enhanced salience of the engram created by emotional commercials, it seems likely that the engram would be more likely to affect attitudinal responses. Actually, Stewart's (1986) finding that ads leaving stronger memory traces and having more brand differentiating impact are more likely to produce greater persuasion is consistent with this view of the impact of emotion stemming from a commercial. We therefore expect that there is more likely to be a direct link between memory and attitudes for the emotional commercials, that is, they will show a pattern more like the classical hierarchy. This result is not expected for nonemotional commercials.

Concepts and Measurements

The concepts employed here included all the steps hypothesized in the hierarchy models. A brief theoretical and methodological explication for each concept follows.


There are many ways to conceptualize people's attention to television and television commercials. Attention can be thought of as how often people look at the screen under natural and free viewing conditions (e.g., Anderson & Lorch, 1983). Attention can be thought of as how much time people voluntarily watch a commercial when they are enabled to avoid it if they prefer (e.g., Olney, Holbrook, & Batra, 1991). Or, in a forced viewing situation, attention can be conceptualized as the intensity with which people experience their viewing (e.g., Reeves & Thorson, 1986). All three of these conceptualizations seem legitimate and important. Methodologically, however, the simplest of these definitions to operationalize is the third one, and for that reason, the self-reported intensity of watching was employed here.


Memory can also be defined in a number of different ways. Perhaps the most difficult and challenging memory task for viewers of commercials is to free recall ads. In this task, the subject must provide his/her own cues for remembering. This means the recall task is also the most free of the biasing impact of cues such as product category, brand name, information about the commercial execution itself, and so on. The recall method was for that reason adopted here.

Attitudes and Conation

Attitude toward the ad and brand are the most common measures of people's attitudes that develop at least partially in response to viewing ads. Although there may be a number of dimensions of liking for ads, we were here most interested in a global response of liking or not liking rather than to the intricacies of dimensions of liking (Olney, Holbrook, & Batra, 1991) or aspects of the antecedents to the liking. The same rather global conceptualization of attitude toward the brand was also preferred here.

Conation is another concept with a number of reasonable possible definitions. Conation can be the actual selection or purchase activity. More often, however, in studies that attempt to tie conation to specific ads, self-reported intention to purchase is used as the critical measure of behavior toward the brand. That definition is used here.

A Methodological Caveat

To test the accuracy and generalizability of the two alternative models, and to compare their applicability to emotional and nonemotional commercials, one would need to sample extensively from emotional and nonemotional commercials. Given this is so, how many ads compose a large sample, and how does one know when the sample is representative? Without a taxonomy of all possible ads, these are difficult questions. Thus it was necessary to limit the scope of the study reported here. In the study, a sample of on-air ads was collected from prime time programming in 1989. These ads were then tested for emotional impact on viewers. Six commercials producing high positive emotional impact on viewers and six commercials producing little or no positive impact were chosen for the present study.

Analytic Approach

The basic idea of flow through successive stages of the hierarchy model and the alternative model is tested by the procedure suggested by Holbrook and Batra (1987) and Olney, Holbrook, and Batra (1991, p. 441). This procedure:

"tests for mediating effects via the rule that Y mediates the effect of X on Z if and only if (1) X is related to Z, (2) Y is related to Z, and (3) X is related to Y, and (4) when Z is regressed on X and Y is controlled for, the significance of X in explaining Z decreases (partial mediation) or disappears entirely (complete mediation)."


Stimulus materials

Twelve commercials were selected from a large pool of those sampled from prime time programming. Extensive pretesting of the commercials in original pool had shown that some ads generated relatively high positive emotional responses, while others showed very little or no emotional response at all. Three seven-point scales were used to index the emotional-neutral dimension. The items included: warm feeling-neutral feeling; emotional-not emotional, and pleasant-unpleasant. The mean emotion rating for the six emotional commercials was 5.8 and for the six neutral commercials, 2.0. The difference between the two groups was significant (F(1,60) = 14.8, p < .01). Three random orders of the twelve commercials were dubbed onto 3/4 inch videotape.




Fifty-nine students in an elementary marketing course in a large midwestern university were tested in the study. Subjects were given credits toward a classroom exam as the reward for participating in the study.

Dependent Measures

Commercial attributes were measured using the 52 items developed by Leavitt (1970). These items are shown in Table 1.

Attention to each ad was measured with three seven-point items describing "how closely did you attend to this commercial:" no concentration-full concentration, watched intensely-watched vaguely, and paid full attention-paid no attention.

Attitude toward the commercials (Aad) was measured with three seven-point items: good-bad, unpleasant-pleasant, and like very much-dislike very much. Attitude toward the brand advertised (Abr) was measured with the same three items.

Purchase Intent served as the index of conation, and was measured with three seven-point items: likely to-unlikely to, plan to-don't plan to, and intend to-don't intend to.

Memory for the commercials was measured at the end of the viewing period with an open-ended questionnaire that asked subjects to "list the brand and/or the basic idea of each commercial you remember seeing."

Procedure and viewing conditions

Subjects were tested in a quiet classroom where they could comfortably view a 19-inch television. The students were randomly assigned to one of the three orders of the 12 stimulus ads. After the interviewer indicated interest in people's responses to commercials and consent forms had been signed, subjects viewed each ad once, then responded to the 52-item adjective checklist, then responded to the attention and attitudinal items. After all 12 of the commercials were completed,a five-minute delay task was introduced. Then subjects were asked to free recall any of the 12 commercials they had seen. After the recall procedure, subjects were thanked and excused.




Ad Characteristics. A factor analysis was conducted on the 52 items used to measure attributes of the commercials. Items loading greater than .5 on each factor were used to form five factors. (See Table 1). The five factors accounted for 75.2% of the total variance. Four of the factors, Stimulation, Liking, Personal Relevance and Familiarity were virtually identical to the four factors reported by Leavitt (1970). The five factors were used as the ad message variables.

Discriminant Validity. To assure that the resulting factors capture different dimensions of ad characteristics, the validity of the measures was assessed by examining the correlations between the factors and dependent measures. The criterion against which discriminant validity is determined is that the factors should relate differentially to at least some of the dependent measures (Carmines & Zeller, 1979).

The correlations are displayed in Table 2. Among the five factors, only personal relevance is related to attention (r = .34). Liking and personal relevance are related to memory but in differing directions (.20 vs -.17). Personal relevance is the only variable not related to any of the attitude variables. The relations of the other four variables to attitudes differ in strength. Although Stimulation is correlated with attitude toward the ad, it is not related to brand attitude. Finally, all factors are related to purchase intention. The relations differ in strength. This pattern of correlations supports the position that the adjective measures are valid in representing different dimensions of ad characteristics.

Reliabilities of Attention, Aad, Abr, and Purchase Intention. The three attention items were combined to measure the intensity with which subjects attended to the commercial. Reliability assessments were performed for each of the twelve test commercials across subjects. Alpha coefficients ranged from .74 to .96, with an overall reliability of .93.

For attitude toward the ad, the three items regarding subjects' evaluation of the ad were combined. Alpha coefficients for the Aad indices ranged from .88 to .95. The overall scale reliability was .93. Similarly, the Abr index measure was constructed by combining three attitudinal items toward the advertised brand. The reliability coefficients ranged from .89 to .90, with overall scale reliability of .95. The purchase intention indices showed relatively high reliability across all twelve commercials, from .94 to .99. The overall scale reliability was .99.

Recall. Memory was indexed by the free recall test. Subjects who could correctly recall any information from a test commercial received a score "1" for that commercial. Failure to recall and incorrect recall was scored "0". Although free recall can capture whether subjects remember anything from the commercial, the dichotomous measure reduces memory variation between subjects. This limitation should be kept in mind when evaluating the results reported below. Descriptive statistics for each of the dependent variables are available from the authors.

Tests of the Research Questions

The purpose of the study was to test two competing hierarchic models, the classic hierarchic (one-way flow) model and the two-route hierarchic model (Figure 1). Hence, the analyses involved the following decisions: (1) first, to determine the relation between ad characteristics and purchase intention; (2) second, to find the appropriate model explained by the present data; and (3) to test the mediating effects of the proposed variables on the relationship between ad characteristics and purchase intention.

Several hierarchical regression analyses were performed in the effort to reach the decisions discussed above. Tables 3-6 display the beta coefficients and R2s that were produced when various mediation steps in the hierarchy were considered.

Predictive Impact of the Ad Characteristics. In Table 3, the key regression results for purchase intention as the dependent variable and the various independent variables are presented. Column 1 in the table shows that when ad characteristics alone were entered in the regression equation, they accounted for significant variance in purchase intention (R2=49.7%, F=8.55, p < .01). Among them, Credibility, Liking, and Personal Relevance had significant positive influences on purchase intention.



Once the relationship between ad characteristics and purchase intention is established, the next task is to test the two models proposed here. First, we examine the classic hierarchy model.

Mediation as Predicted by the Classic Hierarchy Model The classic hierarchic model showed in Figure 1a presents a one-way flow of advertising effectiveness from ad characteristics to purchase intention with various interventions of attention, memory, attitude toward ad, and attitude toward brand. Because the mediation involves four intervening variables, several hierarchical regression equations were built to test each intervening variable step by step. The regression results are, then, subject to meeting the required conditions for mediating effects discussed above.

Columns 1 and 7-10 in Table 3 show that some links are missing from the successive stages of the classic hierarchic model. The missing links may occur in attention and memory. When attention and memory were added to the regression equations (Column 7 and 8), the total variance was not improved to a significant level (R2=51.6% to 54.03%, hierarchical F tests= 2.04, 2.06, p > .05). In addition, the two variables did not perform well as predictors of purchase intention (Columns 2 and 3). Thus, the results do not meet the condition of mediating effects in that attention and memory should be related to purchase intention. The beta coefficients of the ad characteristics did not decrease or disappear when attention and memory were controlled for. Credibility, Liking and Personal Relevance still affected purchase intention strongly and positively.

However, Columns 1, 6, and 10 reveal that the mediation between ad characteristics and purchase intention exists. When all the mediators were controlled for, the effects of ad characteristics decreased substantially but were still significant, with the exception of Personal Relevance. The source of mediation may come from attitude toward ad and/or attitude toward brand (Aad, Abr). When Aad and Abr were entered, respectively, in the regressions, the effects of Credibility and Liking dropped (Columns 9, 10). Aad and Abr by themselves are also strong predictors of purchase intention (Columns 4, 5).

In conclusion, the classic hierarchic model is not supported by the present data. Ad characteristics affect purchase intention significantly. The effects are both direct and partially mediated. However, the mediation does not flow through attention, memory, to attitudinal responses, and finally to conative responses. Rather, the effects of some ad characteristics, Credibility and Liking, operate via Aad and Abr on purchase intention. Personal relevance, however, has a strong direct effect on purchase intention, which overwhelms the mediation of Aad and Abr.

Mediation as Predicted by the Two-Route Hierarchic Model. Figure 1b presents the two-route hierarchic model of advertising effectiveness. The model proposes a distinction between learning from ads and evaluative responses to ads. Basically, it suggests that the evaluative route of responses to ads is not affected directly by learning. Here, we first examine the distinction by looking at the relationship between memory and Aad, which separates comprehension/learning from evaluation of ads. Equations 5-6 in Table 4 provide the first evidence to support the hypothetical distinction. Attention and memory do not affect Aad (beta=.18, .03, p > .05).

We now first examine the evaluative route, which leads successive responses to purchase intention in reaction to an ad. Given the evidence that attention affects neither purchase intention nor Aad, the hypothesis that the evaluative route to conative responses is driven by attention is not supported. Thus, further tests of the evaluative route will exclude attention from the regression analyses. Moreover, to test the causal order of mediations in the evaluative route, (1) the links between ad characteristics, Aad, and Abr were tested, then (2) the links between ad characteristics, Aad, Abr, and Purchase Intention were tested.

Results from Equations 7-10 in Table 4 show support for the first link, that is, that Aad mediates the effects of ad characteristics. Aad and Abr were highly correlated. This means that in further analyses, the two variables may suppress each other's effects and thus interpretive caution should be kept in mind. Nonetheless, when Aad was controlled for, the effects of ad characteristics on Abr decrease and even disappear for Credibility (beta= .32 to .01). For Credibility, Aad manifests complete mediation. In contrast, for Liking and Familiarity, Aad only partially mediates the effects on Abr. Moreover, Equations 11-13 support the second causal link. After both Aad and Abr are controlled for, the effects of Credibility on Purchase Intention decrease from .40 and .29 to .28. The effects of Liking show the same descending pattern. Similarly, Personal Relevance has a strong direct influence on purchase intention, and it is not mediated by attitude toward the ad.

Equations 1-4 in Table 4 provide slight support for the comprehension/learning route. Both Personal Relevance and Familiarity have strong direct effects on attention. However, ad characteristics and attention do not show direct influence on memory. Yet, the effect of attention is improved and approached significance level when ad characteristics were controlled for (beta= .27, p = .07). The possible explanation for this nonsignificant effect is the memory measurement used in the study. Memory was measured by whether or not subjects recalled the ad in the study but not how much they could remember about the ad. Again, then, the problem with the dichotomous nature of the memory measure must be kept in mind.

In conclusion, the two-route hierarchic model is supported but with some qualifications. First, the model proposes that the comprehension/learning route and the evaluative route are both driven by attention. However, the results obtained suggest that attention only influence the comprehension/learning route but without having impact on attitude toward ad. Thus, the effects of ad characteristics on purchase intention are operated via attitude toward ad and attitude toward brand without the intervention of attention. Second, the relationship between attention and memory is hypothesized with a caution over the methodological concern. The resultant model is displayed in Figure 1c.

Mediation as Predicted for Effects of Emotional Commercials . Table 5 displays the key regression results for testing the hierarchic model of advertising effectiveness for emotional commercials. Equations 7-15 provide support for the evaluative route to the final conative responses to ads. With Aad controlled for, the effect of Credibility on Aad disappears. The effects of Liking and Familiarity decrease substantially. Aad shows a complete mediation for credibility and partial mediation for liking and familiarity. The effect of Liking on purchase intention is completely mediated via Aad and Abr, but the effect of Credibility is partially mediated. Personal Relevance turns out to have overwhelming direct effect on purchase intention.

Tests of the comprehension/learning route reveal more interesting applications to the two-route hierarchic model. First, the link between ad characteristics and attention disappear for emotional commercials. Equations 1 and 2 show no significant effects of ad characteristics on attention and memory (R2= 10.7%, 9.7%; F= 1.03, .93, p > .05). However, influence of learning on the evaluative process of advertising is still found. Results from Equations 4-6 indicate that both attention and memory have direct effects on Aad. Moreover, the effect of attention is partially mediated by memory. The advertising effectiveness model for emotional commercials is illustrated in Figure 2a. One implication from this model is that for emotional commercials the learning route and the evaluative route converge to influence attitude toward brand and purchase intention.

Mediation as Predicted for Effects of Non-emotional Commercials. Similar tests were performed for non-emotional commercials. The results are shown in Table 6. The empirical findings, again, validate the evaluative route of responses to ads (see Equations 6-14). Credibility and liking have effects on attitude toward ad. The effects of Liking and Familiarity on attitude toward the brand are mediated through attitude toward the ad. Credibility and Liking influence purchase intention via attitude toward ad and attitude toward brand. Personal relevance has a strong direct effect on purchase intention without intervention.

The comprehension/learning process of advertising is partially supported in that a direct effect is found between ad characteristics and attention.







However, both ad characteristics and attention indicate a weak connection with memory (Equations 2 and 3). It seems that both personal relevance and familiarity have a positive impact in drawing attention. Nontheless, the attention paid to the ad does not necessarily produce better memory for it.


Given that the present study involved forced viewing of commercials in a lab setting, the consistency of its results with studies using a variety of other methodologies is striking. First, in the overall sample of commercials, the classic hierarchy of effects is not supported. There is a memory branch and an attitudinal/conative branch and the two do not correlate with each other. This is consistent with the large number of studies that have argued that across the great unwashed world of commercials, recall is not a good index of persuasion. Also of considerable interest is the fact that in the two-branched hierarchy, attention influences processing in the memory branch, but not in the attitudinal branch. This is the same result as found by Thorson, Zhao, & Friestad (1988), even though their study was carried out with non-forced viewing, and with the commercials embedded in ordinary TV programming environment. These results provide strong evidence that watching a commercial makes viewers remember it. But once watching has occurred, it has no intrinsic influence on whether viewers like the ad or brand. That impact is presumably carried by the quality and persuasiveness of the commercial itself.

A third, and perhaps most interesting way in which the present results are consistent with previous findings, is that when commercials are subdivided into those that create emotion in viewers and those that do not, the emotional commercials generate a clear linkage between memory and attitudes. Like in Stewart's large scale study of commercials, those ads that are remembered well and have special impact (in his case, successfully differentiated the brand) are also more likely to show high persuasion scores. Again, it is notable that the present results coincide with Stewart's findings when his study involved so many more commercials and when the persuasion score was significantly different from the one used here. Stewart used the ARS measure of brand change to index persuasion, while here the measure was simply of liking the ad and brand and intention to purchase.

Somewhat puzzling was the result that for emotional commercials, ad characteristics did not predict attention. An obvious explanation was that the emotional ads were all watched with high intensity and the resultant lack of variance in attention scores prevented their prediction by the ad characteristics. Unfortunately the mean attention scores for the emotional commercials simply were not high enough or even consistent enough to provide much support for this explanation. In the overall sample, Personal Relevance and Familiarity predicted attention. Another explanation for the lack of relationship between ad characteristics and attention may have been that the impact of Relevance and Familiarity were eliminated by the impact of the emotional response to the ads. This idea seems plausible when we also consider that Personal Relevance and Familiarity had a very high impact on processing the nonemotional commercials. When emotion is not around to carry the day for a commercial, then it makes sense that how relevant people think the commercial is and how well they already know the brand would affect how closely they watch.

It is also interesting to consider that even when the attention to memory to attitude to conation link is quite strong, as it is for the emotional commercials, the ad characteristics that predict each step by itself differ. None of the five ad characteristics drive attention or memory, Stimulation and Credibility drive Aad, Credibility, Liking, and Familiarity drive Abr, and Credibility and Personal Relevance drive Purchase Intention. This finding helps to elucidate how McGuire's matrix of persuasion model may operate. Ad characteristics do indeed influence different stages of the hierarchy of responses, just as McGuire posited. However, the stages themselves also influence each other. For example, for the emotional commercials, strong recall is associated with more positive Aad, more positive Aad with more positive Abr, and a greater intention to purchase.

Finally, it is interesting to note that the five dimensions of ad characteristics probably reflect the differences in emotional and nonemotional commercials, but the impact of emotional commercials is not captured by these dimensions. That this is so is confirmed by the finding that the structure of the hierarchical flow of effects is itself changed by emotional impact in commercials. This finding is again consistent with a frequent finding in advertising research. Edell & Burke (1987), Stayman & Aaker (1988), and others have found that even when other antecedents of attitude are taken into account, the impact of individual emotional responses to ads often have independent direct effects on attitude.


The consistency of the present study with others using different commercials, different viewing conditions, different dependent measures, and different subjects argues for its generalizability. But as with all studies, the extent to which its conclusions apply must come under close scrutiny and some skepticism. The use of Leavitt's scale of ad attributes rather than one of the many other copytesting systems available may have influenced the pattern of the linkage between stimulus characteristics and flow of processing. The self-reported measure of intensity of attention paid may have significantly changed the pattern of influence flow from ad structure to the other stages of the hierarchy. A different and larger sample of commercials and the method of categorization into emotional and nonemotional might also be expected to change the observed pattern of the processing. Measuring memory with a dichotomous free recall index may have reduced the variation in memory sufficiently to lessen the opportunity to observe its true relations with the other stages of the hierarchy. Testing of adults or the elderly or any other population group might yield different results from those observed in college students. And certainly intention to purchase is a different measure from actual purchase behavior. As always, the narrower-than-desirable scope of this study motivates a call for further research on the question.

Subject to these limitations, however, the results reported here are encouraging. In spite of those who have long decried the value of looking at memory for commercials, there are under certain conditions, clear relationships between memory, attitudes, and conation. In spite of the age of the hierarchy model and the many attacks on its validity (e..g, Preston & Thorson, 1983), with some modifications, the model remains strongly predictive of viewer behavior. And finally, the study suggests that there is a role for both memory and attention to play. In an advertising literature dominated by attitude theory and measurement, these other two processes are often given short shrift. But in a media world burdened with high levels of clutter, attending and remembering some ads rather than others may soon come to be seen as critically important.


Anderson, Daniel K., & Lorch, Elizabeth P. (1983). Looking at television: Action or reaction? In J. Bryant and D.R. Anderson (Eds.), Children's Understanding of Television. New York: Academic Press.

Bovee, Courtland L, & Arens, William F. (1982) Contemporary Advertising. Homewood, IL: Richard D. Irwin.

Burke, Marian C., & Edell, Julie A. (1989). The impact of feelings on ad-based affect and cognition. Journal of Marketing Research, 26, 69-83.

Carmines, Edward G., & Zeller, Richard A. (1979). Reliability and Validity Assessment. Newbury Park, CA: Sage.

Clark, D.M., & Teasdale, J.D. (1985). Contraints on the effects of mood on memory. Journal of Personality and Social Psychology, 48, 1595-1608.

Edell, Julie A., & Burke, Marian D. (1987). The power of feelings in understanding advertising effects. Journal of Consumer Research, 14, 421-433.

Gibson, Lawrence D. (1983). Not recall. Journal of Advertising Research, 23(1), 39-46.

Haller, T. B. (1972). Predicting recall of TV commercials. Journal of Advertising Research, 12, 43-46.

Holbrook, Morris, & Batra, Rajeev (1987). Assessing the role of emotions as mediators of consumer responses to advertising. Journal of Consumer Research, 14, 404-420.

Ju, Kuen-Hee, Stout, Patricia A., & Leckenby, John D. (1990). An annotated bibliography of copy research: 1983-1987. Working paper, College of Communication, University of Texas at Austin.

Kintsch, Walter, & van Dijk, T.A. (1978). Toward a model of text comprehension and production. Psychological Review, 85, 363-394.

Lavidge, R.J., & Steiner, Gary A. (1961). A model for predictive measurement of advertising effectiveness. Journal of Marketing, 25, 59-62.

Leavitt, Clark (1970). A multidimensional set of rating scales for television commercials. Journal of Applied Psychology, 54, 427-429.

Lewis, E. St. Elmo (1898). As cited in Edward K. Strong, Jr. (1925) The Psychology of Selling and Advertising. New York: McGraw-Hill.

MacKenzie, Scott B. (1986). The role of attention in mediating the effects of advertising on attribute importance. Journal of Consumer Research, 13, 174-195.

McGuire, William J. (1969). An information-processing model of advertising effectiveness. In H.L. Davis & A.J. Silk (Eds.), Behavioral and Management Sciences in Marketing. New York: Ronald.

Mishkin, M., Balamut, B., & Bachevaleir (1984). In G. Lynch, J.L. McGaugh, and N.M. Weinberger (Eds.), Neurobiology of Learning and Memory. New York: Guilford, 65-77.

Mitchell, A.A. (1983). The effects of visual and emotional advertising: An information processing approach. In Larry Percy & Arch G. Woodside (Eds.), Advertising and Consumer Psychology. Lexington, MA; Lexington Books.

Olney, Thomas J., Holbrook, Morris, B., & Batra, Rajeev (1991). Consumer responses to advertising: The effects of ad content, emotions, and attitude toward the ad on viewing time. Journal of Consumer Research, 17, 440-453.

Page, Thomas J., Thorson, Esther, & Heide, Maria Papas (1990). The memory impact of commercials varying in emotional appeal and product involvement. In Stuart Agres, Julie A. Edell, and Tony M. Dubitsky (eds.), Emotion in Advertising: Theoretical and Proactical Explorations. Westport, CT: Quorum Books.

Preston, Ivan (1982). The Association Model of the advertising communication process. Journal of Advertising, 11, 3-15.

Preston, Ivan, & Thorson, Esther (1983). The expanded Association Model: Keeping the hierarchy concept alive. Journal of Advertising Research, 24, 59-66.

Reeves, Byron, and Thorson, Esther (1986). Watching television: Experiments on the viewing process. Communication Research, 13, 343-361.

Squires, Larry R. (1986). Mechanisms of memory. Science, 232, 1612-1619.

Stayman, Douglas M., & Aaker, David A. (1987). Repetition and affective response: Differences in specific feeling responses and the mediating role of attitude toward the ad. Working paper, University of Texas at Austin.

Stewart, David W. (1986). The moderating role of recall, comprehension, and brand differentiation on the persuasiveness of television advertising. Journal of Advertising Research, (April/May), 43-47.

Stewart, David W., & Furse, David H. (1985). Analysis of the impact of executional factors on advertising performance. Journal of Advertising Research, 24(6), 23-26.

Stewart, David W., & Furse, David H. (1986). Effective Television Advertising: A Study of 1000 Commercials. Lexington, MA: Lexington Books.

Thorson, Esther (1990). Television commercials as mass media messages. In James J. Bradac (Ed.), Messages in Communication Science: Contemporary Approaches to the Study of Effects. Beverly Hills, CA: Sage.

Thorson, Esther, & Friestad, Marian (1989). The effects of emotion on episodic memory for TV commercials. In Pat Cafferata & Alice Tybout (Eds.), Advertising and Consumer Psychology. New York: Lexington Press.

Thorson, Esther, & Page, Thomas J. (1990) On the ubiquity of Aad effects. Paper presented at the American Academy of Advertising, Orlando, April.

Thorson, Esther, Zhao, Xinshu, & Friestad, Marian (1988). Attention over time: Behavior in a natural viewing environment. Paper presented at the American Academy of Advertising, Chicago.

Wells, William, Burnett, John, & Moriarty, Sandra (1989). Advertising: Principles and Practice. Englewood Cliffs, NJ: Prentice Hall.



Esther Thorson, University of Wisconsin-Madison
Annie Chi, University of Wisconsin-Madison
Clark Leavitt, Ohio State University


NA - Advances in Consumer Research Volume 19 | 1992

Share Proceeding

Featured papers

See More


Decisional Conflict Predicts Myopia

Paul Edgar Stillman, Ohio State University, USA
Melissa Ferguson, Cornell University, USA

Read More


Foods for Sharing: The Social Value of Handmade Foods

Xin Wang, Nanjing University
Chunqu Xiao, Nanjing University
Xingyu Duan, Nanjing University
Hong Zhu, Nanjing University

Read More


Doing Worse but Feeling Better: Consequences of Collective Choice

Nuno Jose Lopes, University of Navarra
Elena Reutskaja, IESE Business School

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.