Effect of Archetypal Embeds on Feelings: An Indirect Route to Affecting Attitudes?

Advertising practitioners have long been the target of charges of unethical conduct stemming from their supposed use of subliminal techniques, though very few studies have shown any value of subliminal embeds in an advertising setting. One way advertising professionals can contend with such

charges is to understand as much as possible about subliminal persuasion, including the situations, if any, in which it occurs. Many studies of subliminal persuasion have investigated the effect of subliminal stimuli on cognitive constructs. The authors add to that work by focusing on affect, specifically three types of feelings identified as important to advertising effectiveness: warm, upbeat, and negative feelings. They found empirical support for the hypothesis that unconscious processing of subliminal embeds has significant effects on the upbeat and negative feelings that subjects report in response to ads. Those feelings in turn had a significant influence on ad and brand attitudes, which suggests that subliminal emb eds may have a small, but indirect effect on attitudes.

Since the 1957 publication of The Hidden Persuaders by Vance Packard, people have been captivated by the concept of subliminal advertising. Subsequent publications, particularly ones by Wilson Bryan Key (e.g., Subliminal Seduction 1973), have been very popular, despite a paucity of empirical evidence that subliminal techniques are used by the advertising industry (Rogers and Seiler 1994), effective in persuading consumers (Moore 1982, 1988; Saegert 1987), or related directly to choice (Trappey 1996). Hence, it is not surprising that advertising scholars are interested in exploring the phenomenon of subliminal persuasion, not necessarily to find ways to exploit it, but to discover its boundary conditions and, indeed, to determine if it works at all.

Moore (1982) suggests that if subliminal messages have any effect, it is likely to be in terms of affective reactions to the message. Key (1973), one of the most vocal accusers of the advertising industry, contends (without providing any methodologically sound support) that subliminal messages are received by an individual's subconscious and stimulate an emotional response. We therefore need to investigate the possible effects of subliminal embeds on feelings rather than on cognitions. Our study extends knowledge about subliminal persuasion by testing its effects on three affective constructs found by Edell and Burke (1987) to be important in the formation of attitude toward the ad ([A.sub.ad]) and attitude toward the brand ([A.sub.b]): upbeat, warm, and negative feelings.

Some evidence suggests that subliminal messages may influence affective reactions to marketing stimuli (e.g., Bagley and Dunlap 1980), psychological stimuli (Zajonc 1980), and person perceptions (Bornstein, Leone, and Galley 1987). Edwards (1990), in fact, reports that subliminal messages may influence only affective reactions, such as moods and feelings. Further, Caccavale, Wanty, and Edell (1982) postulate that an ad containing a subliminal embed might work through feelings rather than cognitions, perhaps because affect may be processed in a system entirely different from the conscious system used to process cognitive information (Zajonc 1980).

Background

"Subliminal" is defined as the processing of stimuli presented below the threshold of conscious awareness (Pratkanis and Greenwald 1988). Subliminal advertising is "the use of words, pictures, and shapes that are purposely inserted into advertising materials so that the viewers of the material cannot perceive the imagery at a conscious level, but rather at a subconscious level" (Rogers and Seiler 1994, p. 38). Pratkanis and Greenwald (1988) identify four types of subliminal stimuli: (1) subthreshold stimuli, which are presented at energy levels that are too weak to be detected by the audience (e.g., flashing the words "Eat Popcorn" onto a screen so quickly that the audience is not aware of them), (2) masked stimuli, which are hidden from the audience by the presentation of some other, overriding stimuli (e.g., briefly presenting the stimulus immediately followed by a bright flash of light), (3) unattended stimuli, which are presented in such a way that the embedded figure is unlikely to be segregated from its figural context (e.g., hiding the figure of a naked body in the curves and lines of a picture of an ice cube), and (4) figurally transfo rmed stimuli, which are words or pictures blurred or distorted to the point that they are unrecognizable (e.g., commands recorded backward and inserted into popular music). Interestingly, in the fourth category, as opposed to the third, the stimulus remains unidentifiable even when focal attention is directed toward it.

One common misunderstanding is confusing "subliminal" with "subtle." Subliminal messages involve specific words or images deliberately hidden within some larger context with the intent of subverting perceptual defenses and affecting attitudes subconsciously. Advertisers often use specific colors to enhance the presentation of their product or subtly influence the mood of the audience. That practice may be subtly manipulative, but it does not involve specific images or deliberate deception. Dramatic presentations, including advertising, often use metaphorical representations to tell a story that goes beyond literal interpretation. For example, in the famous 1984 ad by Apple, the advertiser was represented by the running women, the only image presented in color, whereas the competition (IBM) was represented by the Big Brother character. Though some members of the audience may not "get" such references, consciously or subconsciously, they are not subliminal embeds, but rather dramatic portrayals.

Subliminal techniques are alleged to influence consumers by affecting attitudes subconsciously (Moore 1982). Theories about why that might happen, however, are scarce. Key (and others) claim that subliminal messages work because they incorporate archetypal imagery. Archetypes are "universal symbols which sustain a constant meaning and efficiency in their applications" (Key 1973, p. 56). According to Key, those symbols are constant across time and culture and evoke a strong and consistent response at the unconscious level. One particularly good example of archetypal imagery is sexual imagery. Many of the supposedly subliminal stimuli in the media today seem to consist of sexual images. Such imagery is allegedly able to affect our attitudes and behaviors without our conscious awareness.

Empirical evidence supporting that reasoning is conspicuous in its absence. In fact, Moore (1982, p. 38) concludes, "The notion that subliminal directives can influence motives or actions is ... incompatible with theoretical conceptions of perception and motivation." Our study was not intended to explain how or why subliminal persuasion works, but rather to investigate empirically its effects on feelings evoked in response to ads.

Feelings and Advertising Effectiveness

Affective reactions to marketing stimuli are important to our understanding of consumers' experiences and may occur without conscious awareness (e.g., Zajonc 1980). Batra and Ray (1986) found that affective responses have a direct effect on attitude toward the ad. In their study affective responses involved conscious reactions to advertisements. The authors suggest, however, that emotional effects are equally likely to occur under less conscious and more "involuntary" conditions. Edell and Burke (1987) and Burke and Edell (1989) extended those findings when they determined that feelings, in addition to influencing [A.sub.ad], have a direct impact on [A.sub.b].

Research indicates that subjects often feel both positive and negative emotions in reaction to a particular advertisement (Edell and Burke 1987). Therefore, we need to investigate not only whether subliminal embeds affect feelings, but also the specific nature of the feelings (i.e., upbeat, warm, and negative, Edell and Burke 1987).

We examined the relationship between incidental exposure to sexually suggestive, subliminally embedded messages and the feelings viewers report in response to the ads. In addition, we explored whether such cues influence men and women differently in terms of feelings. Evidence suggests that men react differently to sexual stimuli than do women. For example, Schmidt, Sigusch, and Schafer (1973) found that both men and women had positive arousal emotions upon exposure to erotic stories, but that women reported more negative emotions such as disgust, irritation and repulsion. Linguistic studies reveal that the use of terms for sexual intercourse (a typical subliminal ploy) may be arousing to both genders, but especially men, and may also stimulate anger and disgust (negative feelings) in many people (Rieber, Wiedemann, and D'Amato 1979).

Hypotheses

Research in the psychology domain has shown that supraliminally presented sexually charged words and stories stimulated "emotionally activated" feelings in both men and women, though more so for men (Schmidt, Sigusch, and Schafer 1973). Examination of the scales used to measure that construct shows it to be very similar to the upbeat feelings described by Edell and Burke (1987; e.g., "excited," "cheerful"). Some evidence indicates that subliminal cues might act similarly. For instance, Bagley and Dunlap (1980) reported that sexually oriented subliminal embeds led to subjects reporting a feeling of being "turned on," which might represent more upbeat feelings in general. If subliminally presented sexually oriented cues do act similarly to supraliminally presented ones, we would expect persons exposed to such cues to report a higher level of upbeat feelings than those not exposed, and that men exposed to such cues would report a higher level of upbeat feelings than women.

H1a: Subjects exposed to sexually suggestive subliminal messages embedded in a print advertisement report higher levels of upbeat feelings in response to the ad than subjects not exposed to such messages.

H1b: Men exposed to sexually suggestive subliminal messages embedded in a print advertisement will report higher levels of upbeat feelings in response to the ad than women exposed to such messages.

Similarly, exposure to supraliminally presented sexually suggestive messages has been shown to arouse negative feelings in both men and women, though more so for women (e.g., "disgusted," irritated;" Schmidt, Sigusch, and Schafer 1973). Again, if subliminal cues act similarly to supraliminal ones, we would expect the following effects.

H2a: Subjects exposed to sexually suggestive subliminal messages embedded in a print advertisement report higher levels of negative feelings in response to the ad than subjects not exposed to such messages.

H2b: Women exposed to sexually suggestive subliminal messages embedded in a print advertisement report higher levels of negative feelings in response to the ad than men exposed to such messages.

Finally, Schmidt, Sigusch, and Schafer (1973) report that sexually explicit messages affect feelings similar to Edell and Burke's (1987) "warm" construct (e.g., "calm," "emotional") in both men and women, but report no differences in the effect due to gender.

H3: Subjects exposed to sexually suggestive subliminal messages embedded in a print advertisement report higher levels of warm feelings in response to the ad than subjects not exposed to such messages.

Research in the cognitive tradition generally finds that subliminal embeds do not influence measures of cognitive processing (Pratkanis and Greenwald 1988). Hence, we did not expect to find any effect of embeds on judgments of an ad's characteristics, as such judgments require cognitive resources beyond unconscious levels. To examine exhaustively the influence of subliminal embeds, however, we empirically investigated those effects. Following Edell and Burke (1987), we investigated the effect of embeds on three dimensions of ad judgments: evaluation (e.g., believability, informativeness), activity (e.g., enthusiasm, excitement), and gentleness (e.g., serenity, tenderness).

One final question addressed by the research was whether feelings engendered by subliminal embeds add explanatory power in predicting higher level measures of ad effectiveness. Exposure to subliminal ad elements has been shown not to affect [A.sub.ad] and [A.sub.b] directly (e.g., Caccavale, Wanty, and Edell 1982). Yet, if exposure affects feelings, subliminal embeds may influence [A.sub.ad] and [A.sub.b] indirectly, as feelings are known to influence those outcome measures. Hence, we investigated the indirect effect of subliminal embeds on [A.sub.ad] and [A.sub.b].

H4a: Exposure to sexually suggestive subliminal messages embedded in a print advertisement has an indirect effect on [A.sub.ad] through its effect on feelings.

H4b: Exposure to sexually suggestive subliminal messages embedded in a print advertisement has an indirect effect on [A.sub.b] through its effect on feelings.

Method

Experimental Design

The hypotheses were examined in a repeated-measures experiment that tested how exposure to sexually suggestive embeds contained in print advertisements affects the feelings consumers have in reaction to the ads, how consumers judge the ads, ad attitudes, and brand attitudes. Subjects for the experiment were recruited from two undergraduate marketing classes at a major southeastern university.

The experimental design is reported in Table 1: each subject was exposed to four advertisements; one each from four product categories. Two advertisements were used for each category, one with an embed and one without. Other than the embed, the ads contained identical copy and pictures. We used the stimuli developed by Caccavale, Wanty, and Edell (1982). Table 1 describes the embeds. The ads were arranged so that each subject saw two ads with embeds and two ads without embeds.

Independent Variables

Treatment. To determine the effects of exposure to the subliminal embeds on the dependent measures, we needed two sets of advertisements: one with and one without embeds. Rather than reviewing actual ads for the arguable presence or absence of embeds, we used ads created to control explicitly for that factor. Use of ads created for imaginary brands also controlled for non-manipulated differences, as well as for the effects of prior brand familiarity and preference. Two ads with identical copy and pictures (one with an embed and the other without) for each type of product were needed. In the no-embed condition the ad's picture was not altered from the original photograph. In the embed condition the ad's picture was altered to include but "hide" a sexually suggestive word (i.e., the "unattended stimuli" in the typology developed by Pratkanis and Greenwald 1988). According to persons who believe advertisers use subliminal techniques, such hidden words represent archetypes to which our unconscious responds.

The four product categories used in the experiment were pretzels, cheese, popcorn, and cheese curls. Those categories were selected for two reasons. First, all of the products represent types of snack foods and are very relevant to the student population. Second, the experiment dictated that pictures of the products be physically altered to include an embedded message. Because Caccavale, Wanty, and Edell (1982) failed to find subliminal effects on cognitive measures when using such stimuli, using the same treatment ads as they did provided a conservative test of our hypotheses.

Dependent Measures

Feelings. Twenty-five items used by both Edell and Burke (1987) and Shaver et al. (1987) were used to measure feelings. After viewing an advertisement, subjects received instructions asking them to indicate the extent to which they felt each of the inventory items on a scale of 1 (did not experience the feeling at all) to 5 (experienced the feeling very strongly). Exploratory factor analysis with varimax rotation confirmed the three factors found by Edell and Burke (1987): upbeat, warm and negative. The feeling scales and alpha coefficients are reported in Table 2. Note that the alpha coefficient for negative feelings is .58, a value lower than the rest but considered acceptable for much of the research in marketing (Churchill and Peter 1984).

Ad Judgments. Judgments (i.e., subjects' descriptions of an ad's characteristics such as whether it was convincing, energetic, or soothing) were measured through an inventory consisting of 18 items reported by Edell and Burke (1987). After completing the feelings measures, subjects were asked to report how well each item in the inventory described the ad on a scale of 1 (the item did not describe the ad well at all) to 5 (the item described the ad extremely well). Exploratory factor analysis with varimax rotation confirmed the three-factor structure described by Edell and Burke (1987) as evaluation, activity, and gentleness. The judgment scales and their alpha coefficients are reported in Table 2.

Attitudes. Attitudes toward the ads ([A.sub.ad]) were measured on three 7-point semantic differential scales anchored by "likable" and "dislikable," "favorable" and "unfavorable," and "good" and "bad." Attitudes toward the advertised brands ([A.sub.b]) were measured with the same scales applied to the brand. Scale scores were fanned by averaging the values of these items for each ad. The alpha coefficient was .95 for [A.sub.ad] and .90 for [A.sub.b].

Control Variables

Several control variables captured differences that might obscure the effects related to the dependent measures. Prior attitude toward each product category (prior [A.sub.c]) was measured on the three attitudinal scales ([alpha]=.90). A product category dummy variable was added to account for other possible differences between the ads. A product category by gender interaction variable controlled for possible differences between male and female attitudes toward the products advertised. Finally, product category by embed interactions were included to control for the fact that one product class used the word "fuck" and the others "sex."

Subjects and Procedure

A total of 56 students participated in the study, each of whom provided measures for two embed ads and two no embed ads. The students were divided into two groups according to whether they attended a morning or afternoon section of the marketing class. Each group was randomly assigned a set of ads. The morning group saw ads with embeds for pretzels and popcorn, and ads without embeds for cheese and cheese curls. The treatment categories were reversed for the afternoon group. The order of ad exposure was constant across the two groups. In sum, 56 subjects provided four observations each, for a total of 224 observations.

The experiment was conducted in a classroom setting. Prior to viewing the ads, subjects were told that "advertisers have long been interested in understanding both consumers' thoughts and feelings they had in response to an ad" and that "while much of the focus of ad research has investigated television commercials, this study concerns the types of thoughts and feelings consumers generate in response to print advertisements." They were then given a questionnaire and asked to indicate their attitude toward products from each category. Next, subjects were exposed to a slide of the first advertisement. Though the ad was presented for two minutes, subjects were asked to fill out the scales as soon as they had formed an impression of the ad, so they did not view the ads for a full two minutes before completing the dependent measures (cf. Goodstein 1993). Immediately upon forming an impression of the ad, respondents indicated their [A.sub.Ad], feelings experienced while reviewing the ad, and judgments of the ad's characteristics, respectively.

After all subjects had finished answering these questions, the second ad was presented. Again subjects were asked to fill out the scales as soon as they had formed an impression of the ad. That process was repeated until subjects had completed the scales for all four ads. Finally, subjects indicated their [A.sub.B] for each of the brands advertised. Questionnaires then were collected and subjects were debriefed.

Debriefing

It was important to assure that the subjects did not consciously see the embeds. Indeed, our stimuli were chosen partially because they had been used successfully in that regard in the past (Caccavale, Wanty, and Edell 1982). To be sure that our subjects had not seen the embeds, the experimenter asked the subjects three important questions during debriefing. The first was an open-ended question about the respondent's opinions of the purpose of the study. The second question was more directed, focusing on things that might have been unusual about the ads. Finally, subjects were asked about the ads' pictures. In no case did any of the subjects mention having seen or having suspected that the ads included embeds. The experimenter then circled the parts of the ads containing the embed and asked subjects to examine them carefully, whereupon subjects reported seeing the embedded messages.

Results

Feelings

Three separate ANCOVA models were estimated to test the effects of subliminal embeds on upbeat, negative, and warm feelings (H1a and b, H2a and b, and H3, respectively). Each ANCOVA model used embedded message, gender, product category, and all relevant interactions as exogenous factors. Prior product category attitude also was included as a continuous measure in the model. Table 3 reports the cell means.

Upbeat Feelings. The ANCOVA model with upbeat feelings as a dependent measure shows no main effect of the embed on upbeat feelings ([mean.sub.no-embed]=16.4l, [mean.sub.embed]=l6.2l; F[less than]1, n.s.). Hence, H1a is not supported. The ANCOVA model does, however, reveal a significant gender by message interaction ([F.sub.(1,163)]=4.30 p[less than].04). The level of upbeat feelings generated by men in the message embed condition (mean=17.41) is significantly higher than the level evoked by women (mean=15.09; [F.sub.(1,83)]=4.60 p[less than].03). Interestingly, the level of upbeat feelings generated by women in the message embed condition (mean= 15.09) is directionally lower than the level in the no-embed condition, though the difference is not significant (mean=16.39; [F.sub.(1,83)]=1.54, p[less than].20). Therefore, a sexually suggestive subliminal embed appears to enhance the level of upbeat feelings for men, but not for women, and H1b is supported.

Negative Feelings. The ANCOVA model for negative feelings reveals a significant main effect of the subliminal embed ([mean.sub.no-embed]=9.97, [mean.sub.embed]=11.07; [F.sub.(1,163)]=5.90, p[less than].02), in support of H2a. The gender by embed interaction is not significant in the model ([F.sub.(1,162)][less than]1, n.s.), so H2b is not supported.

Warm Feelings. In terms of warm feelings, no effects due to either embeds or their interactions with gender are significant (all Fs[less than]1, n.s.) and H3 is not supported.

Judgments

We also empirically investigated the effects of exposure to subliminal embeds on cognitive judgments of the ads' characteristics. Three separate ANCOVAs were performed with evaluation, activity, and gentleness judgments as dependent variables. The independent variables were the same as those used in the analysis of the feeling scales. Table 3 reports the cell means. As expected, all main effects of the embeds and their interactions with gender are nonsignificant (All Fs[less than or equal to]1, n.s.). Hence, our findings add further evidence that subliminal embeds do not affect cognitively based judgments.

Attitudes

Before investigating the indirect effects of subliminal embeds on [A.sub.ad] and [A.sub.b], we examined their direct effects by using ANCOVA. As expected, we found no direct effects of the embeds on [A.sub.ad] or [A.sub.b].

To test the possibility that sexually suggestive subliminal embeds have an indirect effect on those measures through their effects on feelings, we performed a systems regression analysis whereby the exogenous variables from our ANCOVAs were used to predict feelings, and those feelings then were used to predict [A.sub.ad] and [A.sub.b]. The results support the hypothesized indirect influence as both upbeat and negative feelings, but not warm feelings, had a significant effect on ad and brand attitudes.

In the first model, [A.sub.ad] was the dependent variable and upbeat, negative, and warm feelings were the independent variables. Consistent with previous findings (Edell and Burke 1987), upbeat feelings were a significant predictor of [A.sub.ad] ([beta]=.41, p[less than].01). Likewise, negative feelings also significantly predicted [A.sub.ad] ([beta]=-.43, p[less than].0l). The effect of warm feelings on [A.sub.ad] did not reach the traditional level of significance ([beta]=.03, p[greater than].10). Analogous results were found for [A.sub.b]. Upbeat feelings ([beta]=.16, p[less than].01) and negative feelings ([beta]=.29, p[less than]0.0l) were significant predictors, while warm feelings ([beta]=.08, p[greater than].10) were not. Hence, preliminary evidence suggests that subliminal embeds may affect those attitudinal measures through their effects on upbeat and negative feelings, as posited by H4a and b.

Discussion

In sum, we found that sexually suggestive subliminal embeds (1) significantly increased upbeat feelings engendered by the ad for men, but not for women, (2) significantly increased negative feelings engendered for both men and women, (3) had no effect on warm feelings, (4) had no direct effects on cognitive measures of ad judgments, and (5) may have affected ad and brand attitudes indirectly through their effects on upbeat and negative feelings.

Previous research dismissing the effects of subliminal advertising on ad reactions did not consider the relationship between the embeds and the feelings that subjects reported in response to the ads. Clearly, subliminal embeds have some influence on the audience, but it appears to be very subtle and to consist almost entirely of effects on feelings rather than cognitive measures. Indeed, the effects on the feelings scales are in opposing directions and may cancel each other out. Perhaps previous subliminal studies showed "no effects" because they measured only attitudes without investigating the nature of the process underlying attitude formation. Negative feelings are detrimental to attitudes toward the ad and brand, whereas upbeat feelings enhance attitude towards the ad and brand. Therefore, the net effect of embeds on attitudes may not be significant, though the underlying mechanisms (feelings) are affected by the embeds.

Our research extends understanding of how feelings engendered by an ad may be formed and how they may be influenced by unconscious cues. Traditional ad measures such as recall, and [A.sub.ad], and [A.sub.b] have not been sensitive to the possible effects of subliminal advertising. However, feelings proved to be sensitive in picking up unconsciously processed cues. Edell and Burke (1987) suggest that feelings might be affected by many unconsciously processed characteristics of an ad. Our findings also suggest that feelings might be determined in part by unconsciously processed factors. Cognitive measures of ad effectiveness (i.e., judgments and attitudes) were not affected by subliminal embeds. Hence, our results replicate findings that subliminal embeds have little effect on cognitive measures of ad effectiveness, while providing evidence that the embeds do influence feelings.

Despite the overall lack of empirical evidence showing that subliminal persuasion may be effective, a large percentage of consumers believe firms use subliminal advertising to enhance sales. Consumers believe the subliminal approach may be effective in influencing behaviors, though such effects are not supported by a meta-analyses of choice data (Trappey 1996). Indeed, subliminally-based self-help products such as smoking cessation and weight reduction aids still permeate the market (e.g., Rogers and Smith 1993). Our results suggest that the influence of subliminal perception is based on feelings evoked by ads rather than on cognitive appraisals. One caveat, however, is that even our measures of feelings required cognitive appraisal of the scale words themselves (cf. Batra and Ray 1986; Edell and Burke 1987; Shaver et al. 1987).

Our results do not suggest that marketers should now begin to use subliminal tactics in their communications. Indeed, our findings show that there would be little, if any, benefit to such an approach. Such tactics could have detrimental consequences for ad effectiveness, as we found that negative feelings engendered by the ad increased for both men and women exposed to a subliminal embed. Though more upbeat feelings were generated by men in response to an embed (which may have also produced a more positive [A.sub.b]), such feelings were directionally decreased in women shown an ad containing an embed. Finally, advertisers know that there are stronger ways to manipulate both feelings and general affect related to ads, such as through music (e.g., Moriarty 1986), congruity (e.g., Goodstein 1993), appeal type (e.g., Burke and Edell 1989), role portrayals (e.g., Richins 1991), and ad images (e.g., Scott 1994).

Despite the evidence that subliminally embedded messages may affect feelings engendered by an ad, practitioners should keep in mind the opposing effects discussed above. Moreover, there is no convincing empirical evidence, and our findings are no exception, showing that subliminal messages affect behavior (Moore 1982; Trappey 1996), which is, of course, the ultimate goal of advertising. Before employing an army of airbrush artists and product stylists to embed sexual images and/or words in their ads, practitioners need evidence that such embeds influence ultimate purchase decisions. In fact, previous research has shown that embeds have little effect on complex behaviors such as purchase intentions and product choice (Caccavale, Wanty, and Edell 1982; Moore 1982, 1988; Trappey 1996). Finally, but perhaps most important, practitioners would be unwise to test ethical boundaries of advertising when consciously processed elements of ads can easily be used to accomplish the same goals more effectively.

One key limitation of our study that should be considered before the results or implications are generalized is that the use of a classroom study dictates certain tradeoffs between experimental control and external validity. For instance, the experimental setting probably stimulated higher overall involvement than might occur in the natural environment, though subjects did not report "seeing" the embedded messages. Holender (1986) comments that evidence for the effectiveness of a subliminal embed is critically dependent on the fact that subjects remain unaware of the subliminal presentation. One may question whether the debriefing was enough to determine unawareness. However, as Dixon (1986) states, it is impossible to be certain that subjects are unaware of an experimental manipulation. The best researchers can hope for is a probabilistic judgment based on answers to questions such as, "which is more likely, that the subjects were honest or that they were lying." Given that criterion, we doubt that all 56 s ubjects would report not seeing the embeds if, in fact, they actually had seen them.

A second limitation of our study is that the covariate for product category was significant in the upbeat and activity models. Including that variable as a covariate controlled for category differences before we estimated our predicted effects, yet, we are left questioning whether subliminal embeds might have differential effects across specific product classes. For instance, subliminal embeds may have a greater influence on upbeat feelings when they are relevant to the product being advertised (e.g., perfume). The effect of relevance between external cues and advertised products has been noted in other advertising domains (e.g., Sengupta, Goodstein, and Boninger 1997).

Recent consumer research has begun to examine unconscious processing effects more closely to understand their relationship with ad reactions (e.g., Grunert 1996). The findings of that research stream have broadened our viewpoint on perceptual processes. The question is no longer "Do unconscious processes occur?" but rather "How might we differentiate the effects of unconscious and conscious processing?" Our study suggests that unconscious perception in general, and subliminal perception in particular, may influence feelings as opposed to cognitive reactions to ads, and is meant to encourage further discussion of that idea.

Andrew B. Aylesworth (Ph.D., Indiana University) is Associate Professor of Marketing, Bentley College.

Ronald C. Goodstein (Ph.D., Duke University) is Associate Professor of Marketing, McDonough School of Business, Georgetown University.

Ajay Kalra (Ph.D., Duke University) is Assistant Professor of Marketing, Graduate School of Industrial Administration, Carnegie Mellon University.

References

Bagley, Gerald S. and B.J. Dunlap (1980), "Subliminally Embedded Ads: A "Turn-On'?" Southern Marketing Proceedings, 269-298.

Batra, Rajeev and Michael L. Ray (1986), "Affective Responses Mediating Acceptance of Advertising," Journal of Consumer Research, 13 (September), 234-249.

Bornstein, Robert F., Dean R. Leone and Donna J. Galley (1987), "The Generalizability of Subliminal Mere Exposure Effects: Influence of Stimuli Perceived Without Awareness on Social Behavior," Journal of Personality & Social Psychology, 53(6), 1070-1079.

Burke, Marian Chapman and Julie A. Edell (1989), "The Impact of Feelings on Ad-Based Affect and Cognition," Journal of Marketing Research, 26 (February), 69-83.

Caccavale, John G., Thomas C. Wanty and Julie A Edell (1982), "Subliminal Implants in Advertisements: An Experiment," in Advances in Consumer Research, Vol. 9, Andrew A. Mitchell, ed., Ann Arbor, MI: Association for Consumer Research, 418-423.

Churchill, Gilbert A. and J. Paul Peter (1984), "Research Design Effects on the Reliability of Rating Scales: A Meta-Analysis," Journal of Marketing Research, 21 (November), 360-375.

Dixon, Norman F. (1986), "On Private Events and Brain Events," Behavioral and Brain Sciences, 9 (1), 29-30.

Edell, Julie A. and Marian Chapman Burke (1987), "The Power of Feelings in Understanding Advertising Effects," Journal of Consumer Research, 14 (December), 421-433.

Edwards, Kari (1990), "The Interplay of Affect and Cognition in Attitude Formation and Change," Journal of Personality & Social Psychology, 59 (2), 202-216.

Goodstein, Ronald C. (1993), "Category-Based Applications and Extensions in Advertising: Motivating More Extensive Ad Processing," Journal of Consumer Research, 20 (June), 87-99.

Grunert, Klaus G. (1996), "Automatic and Strategic Processes in Advertising Effects," Journal of Marketing, 60 (October), 88-101.

Holender, Daniel (1986), "Semantic Activation Without Conscious Identification in Dichotic Listening, Parafoveal Vision, and Visual Masking: A Survey and Appraisal," Behavioral and Brain Sciences, 9(1), 1-23.

Key, Wilson Bryan (1973), Subliminal Seduction: Ad Media's Manipulation of a Not So Innocent America, New York: Signet.

Moore, Timothy E. (1982), "Subliminal Advertising: What You See Is What You Get," Journal of Marketing, 46 (Spring), 38-47. ___ (1988), "The Case Against Subliminal Manipulation," Psychology & Marketing, 5 (Winter), 297-316.

Moriarty, Sandra E. (1986), Creative Advertising: Theory and Practice, Englewood Cliffs, NJ: Prentice Hall.

Packard, Vance (1957), The Hidden Persuaders, New York: David McKay.

Pratkanis, Anthony R. and Anthony G. Greenwald (1988), "Recent Perspectives on Unconscious Processing: Still No Marketing Applications," Psychology & Marketing, 5 (Winter), 337-353.

Richins, Marsha L. (1991), "Social Comparisons and the Idealized Images of Advertising," Journal of Consumer Research, 18 (June), 71-83.

Rieber, Robert W., Carl Wiedemann and Jeanette D'Amato (1979), "Obscenity: Its Frequency and Context of Usage as Compared in Males, Nonfeminist Females, and Feminist Females," Journal of Psycholinguistic Research, 8 (3), 201-233.

Rogers, Martha and Christine A. Seiler (1994), "The Answer Is No: A National Survey of Advertising Industry Practitioners and Their Clients about Whether They Use Subliminal Advertising," Journal of Advertising Research, 34 (March/April), 36-45. ___ and Kirk H. Smith (1993), "Public Perceptions of Subliminal Advertising: Why Practitioners Shouldn't Ignore This Issue," Journal of Advertising Research, 33 (March/April), 10-18.

Saegert, Joel (1987), "Why Marketing Should Quit Giving Subliminal Advertising the Benefit of the Doubt," Psychology & Marketing, 4 (Summer), 107-120.

Schmidt, Gunter, Volkamr Sigusch and Siegrid Schafer (1973), "Responses to Reading Erotic Stories: Male-Female Differences," Archives of Sexual Behavior, 2 (3), 181-199.

Scott, Linda M. (1994), "Images in Advertising: The Need for a Theory of Visual Rhetoric," Journal of Consumer Research, 21 (September), 252-273.

Sengupta, Jaideep, Ronald C. Goodstein and David Boninger (1997), "All Cues Are Not Created Equal: Obtaining Low Involvement Attitude Persistence," Journal of Consumer Research, 23 (March), 351-361.

Shaver, Phillip, Judith Schwartz, Donald Kirson and Cary O'Connor (1987), "Emotion Knowledge: Further Exploration of a Prototype Approach," Journal of Personality & Social Psychology, 52 (June), 1061-1086.

Trappoy, Charles (1996), "A Meta-Analysis of Consumer Choice and Subliminal Advertising," Psychology & Marketing, 13 (August), 517-530.

Zajonc, Robert B. (1980), "Feeling and Thinking: Preferences Need No Inferences," American Psychologist, 35 (February), 151-175.

                            Experimental Design
          Product Category
  Group       Pretzels     Cheese Popcorn Cheese Curls
 Morning       Embed       Plain   Embed     Plain
  N=30
Afternoon      Plain       Embed   Plain     Embed
  N=26
Embed descriptions (from Caccavale, Wanty, and Edell 1982, p. 419)
Pretzels: "Sex" scratched in the side of one pretzel.
Cheese: "Sex" written on two crackers and a phallic symbol etched on the
cheese. Popcorn: "Sex" written on two kernels.
Cheese Curls: "Fuck" constructed out of the cheese curls themselves.
                                Scale Items
           Feeling Scales
Upbeat     Warm           Negative
Amused     Affectionate   Bored
Aroused    Hopeful        Defiant
Attentive  Kind           Fearful
Caring     Moved          Regretful
Cheerful   Peaceful       Sad
Confident  Sentimental    Suspicious
Happy      Warmhearted
Inspired
Interested
Proud
Satisfied
Stimulated
Alpha Coefficients: Upbeat=.87, Warm=.87, Negative=.58.
Ranges: Upbeat 12 - 60, Warm 7 - 35, Negative 6 - 30.
            Judgment Scales
Evaluation  Activity          Gentleness
Bad         Exciting          Gentle
Believable  Humorous          Serene
Informative Imaginative       Soothing
Irritating  Ingenious
Phony       Likable
Ridiculous  Playful
            Unique
            Vigorous
            Worth Remembering
Alpha Coefficients: Evaluation=.77, Activity=.82, Gentleness=.78.
Ranges: Evaluation 6 - 30, Activity 9 - 45, Gentleness 3 - 15.
                 Cell Means of Experimental Conditions [a]
        Negative        Upbeat         Warm
          Male   Female  Male  Female  Male  Female
No       10.67    9.18  16.44  16.39   8.44   7.76
Message  (3.11)  (3.00) (5.96) (4.67) (3.39) (1.97)
Embed     n=39    n=40   n=39   n=41   n=39   n=41
Message  11.85   10.33  17.41  15.09   8.80   7.70
Embed    (3.28)  (3.05) (5.80) (3.91) (2.84) (2.10)
          n=41    n=43   n=41   n=44   n=41   n=44
        Evaluation        Activity        Gentleness
           Male    Female   Male   Female    Male    Female
No        17.13    15.05   12.28   11.25     4.08     3.85
Message   (4.23)   (3.74)  (4.77)  (2.47)   (1.80)   (1.57)
Embed      n=39     n=40    n=39    n=40     n=39     n=39
Message   16.80    15.70   12.98   11.57     4.37     3.55
Embed     (4.15)   (4.41)  (4.27)  (3.82)   (2.03)   (1.32)
           n=41     n=44    n=41    n=44     n=41     n=44
(a.)Standard Deviation in parentheses.