Indirect persuasion in advertising: how consumers process metaphors presented in pictures and...

In current advertising practice, it is rare to find magazine ads that lead off with a direct verbal claim such as "Tide gets clothes clean." Instead of straightforward claims that a brand possesses some attribute or delivers some benefit, one encounters pictures of dress shirts stacked in the

freezer (Cheer), detergent bottles wearing miner's helmets (Purex), or measuring cups full of blue sky (Tide; see Figure 1). If the picture is straightforward, then the headline may be metaphorical, with an offer of "complete love for your clothes" (Cheer), or clothes "fresh and clean as sunshine" (Gain). This trend away from direct claims has been in place for decades. By the 1990s, Leigh (1994) found that 74% of magazine ads used rhetorical figures in their headlines. Similarly, Phillips and McQuarrie (2003) report that the incidence of figurative pictures in ads approximately doubled during the time period from 1954 to 1999.

[FIGURE 1 OMITTED]

Clearly, advertisers have come to believe that making indirect claims can confer some advantage over making direct claims. However, the nature of this advantage is by no means obvious (Kardes 1993). On the contrary, copywriting texts are replete with maxims to "be direct," and harp on the importance of clarity and simplicity (e.g., Burton and Purvis 1996). The rationale is that the consumer of magazine advertising is busy and distracted, with little capacity to process advertising messages. Under such circumstances, it would appear to be imperative to use simple direct claims such as "Tide gets clothes clean," accompanied by realistic pictures of expected product benefits. Instead, indirect persuasion attempts in words, and especially in pictures, grow ever more predominant. Why?

We shall argue that indirect claims, such as those using metaphor, may be advantageous because they render the consumer more receptive to multiple, distinct, positive inferences about the advertised brand. In addition, an indirect metaphorical claim presented in a picture enjoys a further advantage because such inferences are more likely to be generated spontaneously at the time of ad exposure. Both processes can be explained using the idea of weak implicature drawn from the branch of linguistics known as pragmatics (Sperber and Wilson 1986). Thus, the intended contribution of the present study is (1) to develop a theoretical explanation for why indirect claims may be advantageous, (2) to show how this explanation allows different types of indirect claims to be distinguished in terms of the processes underlying consumer response, and (3) to demonstrate that pictorial indirect claims may be particularly advantageous. The effect of this demonstration is to provide empirical confirmation for the concerns often expressed about the misleading potential of indirect claims in general and pictures in particular. Hence, a final contribution of this paper is to present the public policy implications of the findings.

INDIRECT CLAIMS IN ADVERTISING

Indirect claims elicit beliefs for which no explicit statements have been made (Smith 1991); instead, consumers are invited to construct multilayered meanings that are not actually given in a text (Dick, Chakravarti, and Biehal 1990; Stern 1992). Indirect persuasion attempts thus rely on consumer inference, in the sense of going beyond what is explicitly stated in an ad (Johar 1995). Although researchers have argued that the way a claim is presented can heavily influence "bottom-line persuasion results," such as brand beliefs (MacKenzie and Lutz 1989, p. 63), the process by which persuasion occurs for indirect claims has not been well articulated. In addition, few attempts have been made to differentiate the category of indirect persuasion attempts, or to specify the process whereby different kinds of indirect claims achieve their effect (Kardes 1993).

METAPHOR AS AN INDIRECT CLAIM

Metaphors represent a type of indirect claim because they make claims in a figurative way rather than in a literal way--the advertising message is not stated outright but only implied (Mothersbaugh, Huhmann, and Franke 2002). Rhetorical figures are artful deviations--incongruities that both require resolution and point the way to resolution (McQuarrie and Mick 1996; Toncar and Munch 2001). For example, when consumers view the Tide ad in Figure 1, they perceive a deviation from expectation: a measuring cup filled with blue sky. Because consumers encounter rhetorical figures all the time (Gibbs 1994; Pollio, Smith, and Pollio 1990), they do not treat the Tide picture as an incomprehensible error. Instead, they realize that the advertiser has invited them to elaborate on how the incongruity can be resolved (Phillips 1997).

The rhetorical figure of metaphor, specifically, invites a comparison of two objects by suggesting: that one object is like another, even though they come from different domains (Stern 1990; Ward and Gaidis 1990). To resolve an advertising metaphor, consumers must draw inferences that find similarities between the two objects. In the case of the Tide ad in Figure 1, consumers have to infer how liquid laundry detergent is like the sky. The content of these inferences comes from the "fuzzy set" of attributes that the two objects share in the consumer's mind (MacCormac 1985). Consumers will first search for a simple inference that associates the two objects; if no simple inference can be found, consumers will entertain multiple alternatives (Sawyer and Howard 1991). For example, in response to the Tide ad, consumers might think that Tide makes clothes as fresh as the breeze, as white as the clouds, or as bright as the sky. It is this openness to multiple alternative interpretations, or weak implicatures (Sperber and Wilson 1986), that may confer a persuasive advantage on indirect claims presented through metaphor.

PICTURES AS AN INDIRECT CLAIM

The use of indirect persuasion attempts in ad pictures has increased even more rapidly than their use in ad words (Phillips and McQuarrie 2003). Moreover, Pollay (1985) documents that during the course of the 20th century, pictures came to occupy an ever-increasing portion of magazine ads, even as the number of words steadily decreased. Thus, any attempt to explain advertisers' increased reliance on indirect persuasion needs to address the delivery of indirect claims through pictures as well as through words. In fact, the ubiquity of pictorial claims suggests that visually presented indirect claims may enjoy some kind of extra or distinct advantage based on a difference in how they are processed by consumers.

Unfortunately, to date, consumer research focused on visual elements in advertising has been the exception rather than the rule. Moreover, the research that has been published has not focused on the inference process for pictures as compared with words. It is well established that pictorial elements in ads, such as color and layout, can influence a variety of advertising outcomes (e.g., Finn 1988; Rossiter and Percy 1983). For instance, it has been shown that a picture can be used to convey a specific belief (Miniard et al. 1991; Mitchell and Olson 1981), and that small alterations to a picture can have a material impact on the favorability of consumer beliefs (e.g., Meyers-Levy and Peracchio 1992; Peracchio and Meyers-Levy 1994). It is even the case that, in some situations, pictures in ads can be more memorable than words (e.g., Childers and Houston 1984). However, none of these studies has examined how pictures compare with words in terms of consumers' receptivity to, or spontaneous generation of, multiple inferences.

In fact, one of the seminal studies of advertising pictures used an extensive set of pretests to eliminate any candidate pictures that conveyed multiple messages (Edell and Staelin 1983, p. 49) and to eliminate any candidate picture-word pairs that differed in believability (p. 50), thus forestalling a test of the central ideas underlying this paper. Notably, given these constraints, Edell and Staelin found unframed pictures to be less persuasive than words or words combined with pictures--a kind of picture inferiority effect. Thus, the present study attempts to close a gap in the literature by comparing specific indirect claims presented alternatively in words and pictures, and demonstrating that the pictorial presentation of such indirect claims may be particularly advantageous.

VERBAL AND PICTORIAL METAPHOR

The requirement to explain advertisers' use of indirect claims in both words and pictures was a driving factor in the selection of metaphor as the specific kind of indirect claim to be examined in this study. A benefit of focusing on metaphorical claims is that both visual and verbal metaphors have been acknowledged and extensively discussed in the literature (Phillips 2003). Use of the same underlying metaphor affords the opportunity to compare pictorial versus verbal means for delivering an indirect claim. Moreover, there is a lively controversy about whether any substantive difference exists between visual or verbal metaphors. Thus, some researchers have hypothesized that metaphor does not occur at the surface level of representation (i.e., pictures versus words), but rather at the level of cognitive thought (Forceville 1996; Hitchon 1997). If the modality of presentation of a metaphor is irrelevant, then there should be no difference in how consumers process an indirect metaphorical claim presented in pictures versus words (see Figure 2 for an example). In support of this line of reasoning, Smith (1991) found that consumers drew similar inferences from (nonmetaphorical) visual and verbal claims.

[FIGURE 2 OMITTED]

In contrast, other researchers, especially semioticians, suggest that ad messages presented in pictures are more "open" to multiple interpretations than similar messages presented in words because the visual message is entirely implicit (Eco 1976; Marchand 1985). Several researchers contend that the openness or ambiguity of pictures in advertising is what makes them so persuasive. Messaris (1997) states that associating two unrelated objects through pictures, such as healthy young adults and cigarettes, or automobiles with mountain scenery, causes viewers to take the association for granted without question. In addition, Tanaka (1994) suggests that advertisers turn to pictures when they don't want to take responsibility for the covert content of their advertising messages; in fact, several researchers have speculated that advertisers purposefully use images to imply messages that cannot be legally verbalized (Pollay 1991; Rossiter and Percy 1983; White 1981).

In light of these perspectives, it appears that indirect claims presented through pictures may differ in important ways from those presented through words, especially in their ability to mislead consumers. Misleading advertising can be defined as a "discrepancy between the factual performance of a product and the consumer's beliefs generated by the advertisement" (Gaeth and Heath 1987, p. 43). If pictures are, indeed, more open to spontaneous positive brand inferences, then the processing differences between indirect claims presented in pictures and words have implications beyond academic theory; they touch on public policy and regulation. As noted above, however, as yet there exists no empirical evidence for the spontaneous generation of multiple inferences from pictures. The experiment described subsequently was designed to address this issue.

CONSUMER PROCESSING OF INDIRECT CLAIMS

In Kardes's (1993) formulation, indirect persuasion attempts can be advantageous because consumers must self-generate the implicitly stated claim. Such self-generated claims are known to be more accessible and less subject to counterarguing (Lee and Olshavsky 1995). The difficulty with this formulation is the restriction that indirect persuasion attempts only enjoy this advantage when consumers are more rather than less involved with the advertising message (Kardes 1988), because only involved consumers will bother to self-generate inferences. If this precise formulation is correct, then the growing prevalence of indirect claims in magazine ads would imply that ordinary consumers have grown more and more involved with ad messages over the years--a dubious proposition.

We test a different explanation of how metaphorical indirect claims are processed by consumers. According to the linguists Sperber and Wilson (1986), there are two basic kinds of inferences that can be drawn in response to a message: (1) strong implicatures, and (2) weak implicatures. For some messages, one inference is likely to be chosen as most relevant by most members of the language community most of the time. In such cases, we speak of strong implicature--the implicature chosen as most relevant tends to vary little across listeners ("strong" in this usage is analogous to "strong signal"). In the case of other messages, a wider and more varied range of inferences is possible. These sorts of utterance place fewer constraints on the inference process, and it is this relaxation of constraints that invests these utterances with a multiplicity of weak implicatures.

Consider the ad in Figure 2 (part B) for Blix cleaner, which presents an indirect claim in the form of a verbal metaphor: "Blix: Stain Grenade." These types of metaphors are among those cited by Sperber and Wilson as prototypical messages that are likely to give rise to weak implicatures. The presence of weak implicatures does not mean that a statement is incurably ambiguous (i.e., lacks a strong implicature altogether); it seems reasonable to expect that many North American consumers would choose "especially powerful stain remover" as a relevant implicature of"Stain Grenade" presented in the context of a household cleaner. The point of Sperber and Wilson's analysis is that alternative inferences are nonetheless more likely than if the claim of extra power had been stated directly. Moreover, these alternatives will be couched differently by different listeners. For one consumer, it will be that Blix has explosive cleaning power, for another, that it will allow her to attack stains. Some of the inferences generated could be considered misleading if stated without substantiation: "Blix is a more powerful cleaner than other brands." Moreover, the consumer need not settle on any of these inferences as the exclusive interpretation of the advertiser's intent; all may be maintained as possibilities. Consequently, an indirect claim open to weak-implicature processing may be said to give rise to a vector of inferences.

Weak implicatures are best thought of as inferences generated as part of an attempt to comprehend advertiser intent. This description of how implicatures operate has several consequences. First, although fallible, a weak implicature is always an attempt to guess what the advertiser is trying to say. As such, weak implicatures generated by a consumer familiar with the genre of advertising will typically take the form of favorable claims about the brand, or more exactly, hypotheses about the specific favorable claim the advertiser intended (Tanaka 1996). Second, because weak implicatures are good-faith attempts to guess advertiser intent, they are less likely to function as distracting thoughts (cf. Edell and Staelin 1983), relative to inferences generally. Third, because a vector of different weak implicatures is generated, persuasiveness of the ad may be enhanced due to diminished counterarguing, inasmuch as a consumer's limited capacity for counterarguing must now be spread across multiple inferred claims, increasing the odds that at least one claim will escape scrutiny. Finally, consumers who are exposed to related but distinct attribute claims (e.g., air bags, antilock brakes, traction control) for a product are more likely to rate a general claim (e.g., automobile safety) as true than consumers who are exposed to one attribute claim repeatedly (Hawkins, Hoch, and Meyers-Levy 2001). In sum, once indirect claims are conceptualized as; a means of increasing the probability that a consumer will be receptive to, or even spontaneously generate, a vector of weak implicatures, it becomes evident why an advertiser might wish to bear the risks of making an indirect rather than a direct claim.

Considered as a theory, the explanation of indirect-claims processing in terms of weak implicature has two additional advantages. First, it offers a genuine alternative to Kardes's initial formulation of how indirect claims; in advertising confer an advantage. Kardes showed that a single inference (the conclusion to a syllogism) could be self-generated and would gain persuasive power thereby; however, this effect was limited to consumers who were more involved with an advertisement. By contrast, Sperber and Wilson describe how some kinds of indirect claims can lead to the generation of multiple inferences via a relatively undemanding process akin to that used to comprehend messages in everyday encounters. Thus, the weak-implicature formulation does not specify a high level of message involvement as a prerequisite. Of course, the two formulations are not mutually exclusive. They may be reconciled by acknowledging that "indirect persuasion" is not a homogeneous category, so that different kinds of indirect persuasion can be explained by invoking different processes.

The second advantage associated with a weak-implicature formulation stems from the fact that, as a theoretical category, "indirect persuasion" is intolerably vague. There are clearly many different alternatives to making a direct claim in advertising, so that the catchall category of "indirect" claims is in need of differentiation. An explanation couched in terms of weak implicature provides one avenue toward obtaining such differentiation. Since it is the lack of constraints on inference that produces weak implicature, the indirectness of a claim can be said to vary directly with the degree to which such constraints are absent. In turn, this makes pictures unanchored by words an important kind of indirect claim to examine, on the semiotic assumption that in the absence of the many interpretive aids built into the structure of verbal language, the interpretation of pictures will be particularly unconstrained. Pictorial claims thus allow a test of the effects of very weak implicature on consumer processing. It seems reasonable to suppose that quite different processes may govern consumer response once constraints on interpretation drop below some threshold.

RESEARCH QUESTIONS

The two propositions to be examined empirically are: (1) that consumers are more receptive to multiple positive inferences concerning the advertised brand after exposure to an indirect claim, whether pictorial or verbal, as compared with a direct claim; and (2) that consumers are more likely to generate multiple positive inferences spontaneously, at the time of exposure, in the case of a pictorial indirect claim as compared with a verbal indirect claim. The first proposition could be tested using a conventional measurement approach. Following ad exposure, consumers would be asked to agree or disagree with a set of possible inferences about the brand, with the expectation that agreement proportions would be higher in the specified treatment condition. However, such a paradigm cannot determine whether consumers spontaneously generated such inferences at the time of ad exposure or only formed these inferences when prompted by the researcher's questions (Johar 1995; Kardes 1988). Consequently, we use a measure of response latency to provide an indication of whether consumers spontaneously generated inferences at the time of ad exposure.

Pioneered by Kardes (1988, 1993) in consumer research and building on work by Fazio and others in social psychology (Bargh and Chartrand 2000; Fazio 1990), response latencies to statements presented following ad exposure have been used to test hypotheses concerning how an ad was processed by consumers. The basic principle underlying such use of response latencies is that the results of a judgment already made can be recalled more quickly than a novel judgment can be constructed; in other words, retrieval is quicker than computation. Thus, in response latency studies, short response times indicate statements that correspond to inferences made at the time of ad exposure that are retrieved at the time of questioning, whereas longer response times indicate inferences that had to be constructed at the time of questioning. This property of response latencies has made them a central variable in contemporary cognitive psychology (Van Zandt 2002), leading to an extensive literature on the best approaches to collecting and analyzing latency data (e.g., Fazio 1990; Ratcliff 1993; Ulrich and Miller 1994). Although as yet little used in advertising and consumer research (see, e.g., Jewell and Unnava 2003; Johar 1995; Kardes 1988), we believe response latencies hold great promise with respect to clarifying the cognitive processes that underlie consumer response to advertising.

METHOD

Stimulus Development

To construct the experimental stimuli, we examined ads in recent issues of popular magazines to locate visual metaphors suitable for experimental manipulation. Working with an initial set of 12 ads, a professional artist extracted the image carrying the visual metaphor and discarded everything else in the source ads. Next, fictitious brand names, with neutral or vaguely positive connotations, were devised for each of the 12 products and combined with a product category identifier (e.g., "Plus Dishwasher Detergent"). To construct the direct verbal claim version (henceforth called "verbal literal"), a large photo of the product package was placed in the center of the ad above the brand name, and then a direct claim, designed to capture the claim implicit in the root visual metaphor, was placed at the top to serve as a headline (e.g., "Clears away tough stains"). The verbal metaphor version was identical except for a change to the wording to create the metaphor (e.g., "Bulldozes tough stains"). For the visual metaphor version, the brand name was again placed at the bottom of the ad and the image carrying the visual metaphor (in this example, a photo of earth-moving equipment clearing a dish) was placed to fill the center of the page; no other text appeared in the visual metaphor version (see Figure 3).

[FIGURE 3 OMITTED]

These 12 initial sets of ad stimuli were pretested to ensure similarity of meaning across the three ad versions within a set and to screen out dead or inapt metaphors. Undergraduate students (n = 99) first rated the extent to which the visual and verbal metaphor ad versions and the verbal literal headlines expressed the same basic idea; this shared meaning was formulated into a strong-implicature statement (e.g., for the dishwasher ad, "Cleans difficult stains from dishes"). Next, for a rhetorical figure to be present, an expression must exhibit artful deviation (McQuarrie and Mick 1996). Dead metaphors--metaphors that have entered conventional usage--are not deviations and fail this test. Likewise, metaphors that are judged inapt--dumb or wrong--fail to meet the artfulness test. To identify dead metaphors, participants rated each verbal headline on a five-point scale as to whether it was "artful, clever" or "plain, matter-of-fact" (McQuarrie and Mick 1996). For the visual metaphor ads, inapt metaphors were identified by asking participants to rate the picture on a five-point scale anchored by "creative, imaginative" or "silly, stupid." Six sets of ad versions survived this screening.

To identify specific weak implicatures associated with the metaphor ads, students (n = 17) examined the six visual metaphor versions and were asked to list the attributes claimed and messages delivered for the brand. A different group of students (n = 12) examined the six verbal metaphor versions and answered the same question. Meanwhile, the authors independently conducted a semiotic analysis of each ad to identify potential meanings. This procedure produced seven to eight candidate statements intended to represent weak implicatures for each of the six ad sets. Wherever possible, we crafted these statements such that they would be misleading in the absence of substantiation. These included claims that probably could not be substantiated (for the dishwasher detergent, "Protects dishes from scratching"), or that were far-fetched (for the fabric softener, "Causes sharp fibers to curl back on themselves").

Next, a substantial pretest was conducted to refine the weak-implicature statements. It is not possible a priori to give an exhaustive specification of the weak implicatures associated with a given statement--weak implicatures arise empirically from the confrontation of a specific statement and a specific language community. In addition, the elicitation procedure used to identify weak implicatures for the metaphors could also have produced "generic" product category inferences such as might be produced in response to virtually any advertisement for the category, whether metaphorical or not. As a case in point, "cleans the dirtiest pots and pans" was an initial candidate for weak implicature for the dishwasher detergent ads. A priori, one might assume that the inclusion of "dirt" would result in a preferential association of this statement with the metaphor versions, given their references to earth-moving equipment. However, this claim appears to have become generic within the dishwasher detergent category, inasmuch as respondents exposed to the verbal literal version were equally likely to accept it as an inference.

Undergraduates at a large North American university (n = 147) saw two instances of each type of ad version (verbal literal, verbal metaphor, visual metaphor) for a total of six ads, one from each of the six version sets. Different subgroups (n = 3) saw different ads, so that all 18 stimuli were rated. Following each ad, participants rated the strong- and weak-implicature statements on a five-point scale anchored by "Almost everyone/no reasonable person would draw this conclusion from the ad." Analysis of the pretest data showed that four ad sets met the criteria (1) that there be a strong implicature shared across all versions, and (2) that there be multiple weak-implicature statements preferentially associated with the metaphor versions.

In a final pretest, we examined whether consumers valued the attributes claimed in the surviving weak implicatures. Recall that for indirect claims to be useful to advertisers, they must lead to the generation of multiple positive inferences. For this purpose, undergraduate students (n = 46) rated each weak implicature on a five-point scale anchored by "a great deal of added value/no added value." Specifically, they were given the product category and asked to assume that a product in that category could deliver the attribute claimed in the strong implicature (e.g., a dishwasher detergent that "cleans difficult stains from dishes"). They then rated the incremental value if that product also delivered the attribute claimed in a weak-implicature statement (e.g., if it was also "an industrial-strength cleaner"). Based on this pretest, we selected three weak implicatures for each of four ad sets that were judged as providing incremental value. We take these to be a sample of the set of all possible weak implicatures that might arise in connection with the metaphorical claims incorporated into the ads.

Participants and Procedure

For the main study, participants were undergraduate students (n = 183) at a different North American university who participated in a subject pool to fulfill a course requirement. A debriefing questionnaire after the study led to the elimination of six participants who either had come close to guessing the hypotheses (3), or failed to follow instructions (3), leaving 177 participants for analysis.

The study was conducted in a small computer lab with four workstations running the Empirisoft DirectRT data collection program. Participants were seated at a computer and read initial instructions on the screen as the experimenter also read them aloud. The first ad stimulus was presented on-screen when the participant pressed a key signaling a readiness to begin; after that point, participants moved at their own pace through the ads and statements, and then finally completed the debriefing questionnaire off-line. Workstations were positioned such that participants could not see the stimuli shown to others.

In the initial instructions, participants were told that advertisers strive to combine pictures and words in ads to create a variety of positive meanings. They were told that advertisers also strive to control the meanings suggested by their ads, because unintended meanings might not be consistent with the advertiser's plan for that brand. Next, they were told that the only way to be sure that an ad conveyed multiple positive meanings while not conveying undesired and unintended meanings was to pretest ads with consumers. Therefore, their task today would be to look at rough versions of ads and to judge whether specific meanings had been intended by the advertiser or not. The purpose of these instructions was to cue participants to examine both the pictures and words in each ad, to expect an ad to have multiple meanings, and to legitimize "no" as well as "yes" responses in the case of specific statements.

Participants saw a total of seven ads and responded to multiple statements for each ad. The first ad shown served as a practice session and responses to it were not analyzed. Next, participants saw a filler ad followed by several statements presented one at a time, the four test ads each followed by statements, and then another filler ad followed by statements. Half the participants saw one filler ad first, half the other. The filler ads and associated statements enabled us to compute, separately from the test stimuli, a covariate measuring how fast or slow each individual participant tended to be in responding based on their latencies in response to these filler--ad statements. The two filler ads were similar to the verbal literal ads (picture plus direct claim), whereas the practice session ad was similar to the visual metaphor ads (metaphorical picture, no headline).

Measures

For each test ad, participants responded to a mix of filler and test statements. Test statements consisted of three weak implicatures and one strong implicature. As an additional protection against yea-saying bias, filler statements included statements that pretesting had shown were very likely to be rejected as inferences from the ads. Order of statement presentation was randomized by the program, with the exception that the first statement presented was always a filler. This procedure was intended to protect against the possibility that respondents might be able to use a visual representation of the ad in short-term working memory to answer the initial statement, a distinct process that might unduly favor visual stimuli. The phrase "The ad was designed to make you think:" appeared at the top of each screen that contained a statement. One statement appeared at a time, and each statement began with the brand name advertised. The program recorded whether the participant answered "yes" or "no," and the latency of this response in milliseconds.

Design and Analysis

The test ad stimuli consisted of four sets of four ad versions--the three versions pretested during stimulus development (unanchored visual metaphor, a verbal metaphor, and verbal literal), and a fourth version, termed an anchored visual metaphor (discussed below). These four ad versions were created for each of four product categories (see descriptions in Table 1). As discussed under stimulus development, each of the first three ad versions within a set had been pretested to ensure that it conveyed a particular implicature ("cleans difficult stains from dishes" in the dishwasher detergent set). The verbal literal ad in each set was designed to convey few if any other weak implicatures beyond the shared implicature (individual participants, of course, may or may not infer any number of other implicatures). This was achieved by including only a package picture and a direct statement of the shared implicature in the verbal literal ad (e.g., "clears away tough stains"). The verbal metaphor ad was identical to the verbal literal ad except that the headline was altered to create a verbal metaphor (e.g., "bulldozes tough stains"). Pretesting showed that verbal metaphor ads were likely to convey additional implicatures beyond the shared implicature. In the unanchored visual metaphor ad, there was no headline, and the package picture was replaced by an image containing a visual metaphor (see Figure 3). Pretesting showed that the visual metaphor versions also conveyed additional implicatures beyond the shared implicature.

The fourth ad version within each ad set, the anchored visual metaphor, was created by combining the picture from the visual metaphor version and the headline from the direct literal version (e.g., for the dishwasher detergent, the picture in Figure 3 showing construction equipment clearing a dish was combined with the headline "Clears away tough stains"). This condition is called the anchored visual metaphor condition because the direct headline anchors the meaning of the picture. Previous research has shown that consumers who are presented with a tentative conclusion for an ad message may stop processing the ad and cease drawing inferences (Santamaria, Garcia-Madruga, and Carretero 1996). Hence, inclusion of a headline making a direct claim may act to reduce the openness of the accompanying visual metaphor (Phillips 2000) and cause fewer weak implicatures to be generated. Inclusion of the anchored visual metaphor condition thus allows a test of the basic contention that it is the lack: of constraints on interpretation that fuels the vector of weak implicatures generated in response to an indirect pictorial claim. If so, the anchored visual metaphor will not cause respondents to produce as many spontaneous weak implicatures as the unanchored visual metaphor.

Order of presentation of the four test ads seen by each participant was randomized by the Empirisoft program. Ad versions were assigned to products in accordance with a latin square. Two latin squares were used, resulting in eight distinct combinations of products and ad versions across the sample as a whole (see Table 2). The basic design analyzed is a 2 (visual metaphor present or absent) x 2 (literal headline present or absent) factorial design. Loglinear analysis was used to apply this design to the analysis of "yes" and "no" responses, and the Linear Mixed Models (LMM) procedure in SPSS 11.5 was used to apply this design to the response latencies. The LMM procedure is a recently developed extension of the more familiar GLM (generalized linear models), MANOVA (multivariate analysis of variance), and ANACOVA (analysis of covariance) procedures (McCulloch and Searles 2001; Wallace and Green 2002) that can accommodate the multilevel covariates used in the analysis of the response latencies, as described below.

Data Transformation and Adjustment for Covariates

Response latencies are notorious for their departures from normality, with a long right tail typically seen in distributions. In the present study, test statement latencies ranged from 537 milliseconds (ms) to 12,632 ms, with a mean of 2,224 ms and a standard deviation of 1,277 ms. Because latency data have become so central to process investigations in cognitive psychology (Van Zandt 2002), a variety of procedures for dealing with distribution problems have been explored. Among the notable conclusions of recent work is that the use of cutoff values to eliminate outliers is difficult to justify in most circumstances (Ulrich and Miller 1994). Nonetheless, Van Zandt (2002, pp. 469-470) offers a compelling demonstration that the presence of even a single outlier can severely bias the mean and the variance for an experimental condition. Current thinking indicates that the best solution to this dilemma is to retain all the data, but to invert the values (Ratcliff 1993; Van Zandt 2002), and that is the procedure we used.

Analyses of responses to filler ads indicated the presence of a marked practice effect, with responses to the last filler ad almost 50% faster than responses to the first. Hence, test ad sequence was entered as a covariate. Similarly, we found a significant correlation between statement sequence and speed of response, so statement sequence was also entered as a covariate. Although both statement sequence and ad sequence were randomized by the program so that they would be unlikely to systematically bias comparisons of the test stimuli, the procedure did not produce exactly equal cell sizes. In any case, it is desirable to explicitly model the effect of these variables in an analysis of covariance inasmuch as it tends to promote a smaller error term and greater statistical power (Kirk 1995). As an additional control, the amount of time spent viewing each ad was entered as a covariate. Finally, participants varied dramatically in their overall slowness or speed of response, so we entered their latency of response to the eight filler-ad statements as a covariate. These four covariates operate variously at the level of the participant, the ad, and the individual statement, and the LMM procedure was developed in part to handle multilevel data of this sort. With the LMM procedure, we were also able to include product as a blocking variable in the analysis. Because not all possible latin squares were used, it is desirable to block on product as a means of extracting variance extraneous to the hypotheses (Kirk 1995).

Manipulation Check: Strong Implicature

If the metaphor ads cannot convey the shared strong implicature, then overall persuasiveness may be compromised, no matter how many weak implicatures are elicited. A test of the strong implicatures across ad conditions using the loglinear procedure confirmed that there were no significant differences in the degree to which the four ad versions conveyed the strong-implicature statements (all [chi square] < 1); in other words, each appeared to be equally effective in conveying the ad's main message. This allows us to interpret the results for the weak implicatures in terms of the conveyance of additional inferences, and not just the conveyance of different inferences.

Treatment Effects: Agreement Proportions for Weak Implicatures

A loglinear analysis of the proportion of "yes" responses to the weak-implicature statements showed a significant positive impact for the presence of a visual metaphor ([chi square] = 42.26, p < .001) and a significant positive impact for the absence of a literal headline ([chi square] = 16.62,p < .001). The interaction between the two factors is not significant ([chi square] < 1). As can be seen from Table 3, the likelihood that a participant will affirm that an ad was designed to convey a weak implicature increases steadily as one moves from the literal headline-only condition, to the verbal metaphor, to the visual metaphor anchored with a literal headline, to the unanchored visual metaphor (all pairwise comparisons were significant at p < .05). The pattern of results indicates that an indirect metaphorical claim, regardless of the modality in which it is presented, is susceptible to a greater number of potential interpretations. Expressing the metaphor in a picture accentuates this effect. However, when a visual metaphor is present but also anchored by a literal statement, participants appear to be somewhat less susceptible to constructing multiple inferences, presumably because they can rely, in part, on the strong implicature of the literal headline to surmise advertiser intent. This is consistent with the idea that it is the relative lack of constraints to the interpretation of pictures that makes possible a larger vector of inferences; once such a constraint is added, in the form of a literal headline, the pictorial effect is reduced.

Treatment Effects: Response Latencies for Weak Implicatures

An LMM was tested with product category as a blocking variable and four covariates: participant's average speed of response, ad position, ad viewing time, and statement position. In addition, a complete factorial design with three factors was specified: whether the response was "yes" or "no," along with the visual metaphor and the literal headline factors. The latter two factors were specified as repeated measures in the LMM, along with an index variable distinguishing the three weak-implicature statements per product. A significant interaction involving the three factors in the design was found, F(1, 1942) = 6.53, p [less than or equal to] .01. As can be seen in Figure 4 and Table 4, the unanchored visual metaphor is the only stimulus type for which "yes" responses are significantly faster than "no" responses. This is the key comparison from the standpoint of the research tradition on response latencies. Differences in "yes" latencies or "no" latencies across stimulus types are not readily interpretable because they confound differences in the processing of the stimuli themselves with differences in the response process (Fazio 1990).

[FIGURE 4 OMITTED]

The significantly faster "yes" latency for unanchored visual metaphors is consistent with the idea that consumers spontaneously generate a vector of weak implicatures at the time of exposure to a visual metaphor. When subsequently asked whether the advertiser intended a specific weak implicature, many participants will answer "yes" because they are able to retrieve from memory a corresponding inference generated earlier. Because retrieval is a relatively fast process, response latencies for such affirmative replies are small. Conversely, participants answering "no" cannot do so until they have finished retrieving the vector of inferences formed at exposure and find no matches. Put another way, because the meaning of a visual metaphor is more open, a larger number of weak implicatures tends to be generated at the time of exposure, which facilitates subsequent "yes" responses in the case of any specific implicature; in turn, the need to complete a search of this same vector entails extra processing time before a subsequent "no" response can be given.

This interpretation is reinforced if we examine both latencies and agreement proportions for the verbal metaphor and anchored visual metaphor conditions. Relative to the literal verbal baseline, verbal metaphors produced higher levels of agreement that the advertiser did intend to convey a specific weak implicature (Table 3). However, response latencies for these affirmative responses were not quicker than those for negative responses; in fact, they are significantly slower. This finding supports the proposition that while verbal metaphors may invite elaborative processing (McQuarrie and Mick 1996), they do not spontaneously elicit weak implicatures at ad exposure. When the experimental protocol subsequently demands that participants in the verbal metaphor condition decide whether a specific weak implicature was intended, some proportion of respondents appear to be receptive, as demonstrated by the high agreement percentages relative to the literal headline condition. Yet this decision would seem to be the result of a computation performed on demand rather than a retrieval of an earlier judgment, as evidenced by the fact that for verbal metaphor, "yes" latencies are significantly slower than "no" latencies.

Relative to the literal verbal baseline, anchored visual metaphors also produced higher levels of agreement that the advertiser did intend to convey a specific weak implicature, as reported in Table 3. However, response latencies for these affirmative responses were not quicker than those for negative responses in this condition. This is consistent with the proposition that consumers tend not to generate spontaneously a vector of weak implicatures at the time of exposure to an anchored visual metaphor. There is no need to do so--an adequate inference about the advertiser's intended message is readily available in the strong implicature provided in the direct headline.

In summary, verbal metaphors and anchored visual metaphors allow multiple interpretations if a subsequent situational cue stimulates a need to interpret the ad text; but only unanchored visual metaphors appear able to provoke such interpretation spontaneously at the time of ad exposure.

GENERAL DISCUSSION

The experiment reported here offers one explanation for why advertisers have increasingly chosen to use indirect persuasion in magazine ads over time. Specifically, the findings show that the use of metaphorical claims in ads appears to make consumers receptive to multiple, distinct, positive inferences about the advertised brand (i.e., weak implicatures), while still conveying the main message of the ad (i.e., the strong implicature). Furthermore, metaphors presented in pictorial form are able to elicit these multiple inferences spontaneously at the time of ad exposure. Participants viewing other forms of metaphor--either verbal metaphors or anchored visual metaphors--appear to generate additional inferences only on reflection when prompted. This pattern of results is consistent with the idea that it is the openness of indirect metaphorical claims--the lack of constraints on their interpretation--that is the source of their persuasive advantage. The fewest constraints, and hence the greatest openness, obtains in the case of indirect metaphorical claims expressed in pictorial form. If advertisers wish consumers to generate multiple positive inferences, to some degree spontaneously, then it is obvious why they might rely more and more on indirect persuasion, especially in pictures.

A further implication of this study is that verbal rhetorical figures in ads may be less effective than visual figures. This corresponds to results recently obtained by McQuarrie and Mick (2003), who found that verbal rhetorical figures had a weaker impact on attitude toward the ad and ad recall than did visual figures. There remains a considerable body of evidence that verbal rhetorical figures in ads can be advantageous, consistent with the argument advanced in McQuarrie and Mick (1996; cf. Mothersbaugh, Huhmann, and Franke 2002; Tom and Eves 1999), to the effect that verbal rhetorical figures invite consumers to elaborate multiple meanings in a way that is pleasurable and memorable. Based on the current study, however, it now appears that visual rhetorical figures may be even more advantageous. An invitation to elaborate may be taken up or not. By contrast, visual figures appear to evoke this elaboration at the time of ad exposure, rendering the positive impacts of such elaboration more probable.

Future research might extend the experimental protocol used here to examine other types of indirect advertising claims. For example, it is possible that other rhetorical figures, such as pun, may also encourage the generation of weak implicature, either spontaneously (when presented through pictures) or as prompted (when presented through words). It would also be interesting to examine a very different form of indirect claim, not resting on a rhetorical figure per se, such as the pictures commonly found in fashion ads. Do these types of indirect claims also confer some advantage and does the underlying process involve the same openness to multiple alternative inferences, or something different? If it is true that direct claims are becoming the exception rather than the rule in magazine advertising, then the kinds of indirect claims available to advertisers are probably numerous, diverse in character, and evolving over time. Future research is needed to differentiate between the various sorts of indirect stratagems available and to link them to specific consumer-processing outcomes. Although the present study focused on multiplicity of positive inferences as a key outcome of the use of indirect metaphorical claims, it is possible that other types of indirect claims may be suitable for achieving other kinds of advantageous outcomes.

Finally, the weak-implicature model of indirect persuasion, unlike the Kardes (1993) model, does not presuppose a high degree of consumer involvement, instead relying on more spontaneous generation of implicatures. Conversely, the logical inferences used by Kardes may have imposed heavier demands on consumer processing than the more pragmatic inferences used here and in Gaeth and Heath (1987). It may be the case that it is only those weak implicatures that take the form of pragmatic inferences that do not require high levels of involvement. In any case, we cannot claim to have reproduced conditions of low task involvement in our experimental protocol because of its reliance on forced exposure and the requirement that participants respond to the ads. It is interesting to note that McQuarrie and Mick (2003) did construct an incidental exposure context and still achieved positive results for rhetorical figures, especially visual figures. Furthermore, in their reanalysis of Starch data (which presumably involved ads that were initially processed under naturalistic conditions), Mothersbaugh, Huhmann, and Franke (2002) also found positive results for rhetorical figures. Nonetheless, an important task for future research is to study inference formation by consumers occurring outside of a laboratory context, under a variety of involvement conditions, and using a purposeful selection of indirect persuasion tools that differ in the demands they impose on consumer processing.

IMPLICATIONS FOR PUBLIC POLICY

Perhaps the most important implications of this study concern its identification of ways and means whereby advertisers can effectively mislead consumers. We put the matter in these stark terms to highlight the darker implications of the results. It has been argued since the 1920s that presenting a claim in pictorial form is particularly useful when the advertiser does not wish the consumer to scrutinize a claim overly closely: "A picture ... can say things that no advertiser could say in words and retain his self respect" (Calkins 1928, p. 141). According to historians of that era, advertisers came to hold the view that "People lack the ability to argue with pictures.... Arguments invited counterarguments, and assertions might provoke skepticism. But pictures deflected skepticism; they inspired belief" (Marchand 1985, pp. 154, 236). More recently, scholars have gone so far as to claim that "nowadays, examples of visual claims that would be unacceptable in verbal form can be found in most kinds of advertising" (Messaris 1997, p. 225). In short, it has been suspected for many years that pictures in ads may be unusually potent persuaders, and that this potency could readily be turned to deceptive purposes.

Pictures are not the only advertising element to raise these concerns. Stern (1992) noted that in light of legal precedent, the crafty advertiser might well decide to avoid direct claims altogether and instead focus on figurative statements and the like. Hence, metaphors and indirect claims more generally might recommend themselves to advertisers who wish consumers to draw certain conclusions, but who dare not make such claims flat out (cf. McQuarrie and Mick 1999; Phillips and McQuarrie 2003).

The literature on deception in advertising thus suggests an additional explanation for the trend away from direct claims in magazine advertising. That is, a tightening regulatory environment over time may have driven advertisers toward indirect persuasion attempts to escape the critical scrutiny of both regulators and consumers. The problem with this explanation is that prior to the present research, there was no empirical evidence that indirect persuasion attempts, outside a high-involvement context, could achieve advertisers' persuasion goals--however much legal immunity these might confer. Indeed, suspicion of the misleading potential of pictures and figurative statements rested on no more than suspicion.

The present study indicates that pictorial metaphors can cause consumers to spontaneously infer multiple positive inferences about the advertised brand, including misleading claims that could not legally be stated outright in the absence of substantiation. To the extent that this finding holds in the case of ads where pictorial metaphor is but one component and with respect to buying behavior in the marketplace and not just judgments in the laboratory, then the suspicion historically directed onto pictures is given an empirical foundation.

If it is true that consumers are peculiarly vulnerable to pictorial metaphors, then legal protections may need to evolve beyond a focus on whether a claim made in words is true or false toward an emphasis on whether a population of typical consumers does or does not spontaneously draw certain inferences from the ad in its entirety, including its pictorial components. The standard could no longer be that of what a reasonable man or woman would do, because it is not reasonable to infer from a picture of a bottle made of berries that the cleaning product contained therein is made of all-natural ingredients. Nonetheless, based on the evidence of this study, it is an empirical fact that such inferences do spontaneously occur when consumers are exposed to pictorial metaphors.

Whether the legal system can evolve to address the possibility of misleading pictorial claims remains to be seen. At the least, this study argues for efforts to educate consumers to attend more closely to pictorial claims and to scrutinize them more critically (as Gaeth and Heath [1987] argue for verbal claims). Such efforts, in turn, require a conviction on the part of scholars and educators that pictorial metaphor is sufficiently potent to be feared. We hope to have contributed to that perception by means of the experimental data on response latencies reported in this study.

TABLE 1
Description of Ad Stimuli and Implicature Statements

                                     Product

                      Plus dishwasher                 Onezip
                         detergent                  slider bags

Visual metaphor   Miniature construction     A chain door lock
stimulus          equipment cleaning off a   attached to a plastic bag
                  dirty dish                 filled with grapes

Verbal metaphor   "Bulldozes tough stains"   "Puts a padlock on
stimulus                                     freshness"

Verbal literal    "Clears away tough         "Seals in the freshness"
stimulus          stains"

Strong-           Cleans difficult stains    Has a tight closure to
implicature       from dishes                keep food fresh
statements

Weak-             Is an industrial-          Gives a tighter seal than
implicature       strength cleaner           other slider bags
statements
                  Protects dishes from       Holds more food than
                  being scratched            other bags the same size

                  Leaves no detergent        Stays shut even when
                  residue                    dropped

                                       Product

                    Able fabric softener       Mills window cleaner

Visual metaphor   Two pots of cacti           A spray bottle composed
stimulus          juxtaposed with two         of purple berries
                  similarly shaped socks

Verbal metaphor   "Removes the cactus feel    "Bring home a fresh
stimulus          from your clothing"         fruit orchard"

Verbal literal    "Removes the scratchy       "Bring home a fresh
stimulus          feel from your clothing"    fruit scent"

Strong-           Makes clothing feel more    Has a fruity smell
implicature       comfortable
statements

Weak-             Works on the most           Uses all-natural
implicature       rugged outdoor clothing     ingredients
statements
                  Causes sharp fibers to      Is environmentally safe
                  curl back on themselves

                  Makes clothing fit better   Is easier to use than
                                              other window cleaners

Note. For each product, the anchored visual metaphor stimulus was
created by combining the picture described in the visual metaphor
row with the headline given in the verbal literal row.

TABLE 2
Conceptual Representation of the Latin Square Designs
Used in the Experiment

Participants in      Unanchored visual      Anchored visual
Group 1 might see:   metaphor for           metaphor for slider bag
                     dishwasher detergent

Participants in      Unanchored visual      Anchored visual metaphor
Group 2 might see:   metaphor for slider    for window cleaner
                     bag

Participants in      Unanchored visual      Anchored visual metaphor
Group 3 might see:   metaphor for window    for fabric softener
                     cleaner

Participants in      Unanchored visual      Anchored visual metaphor
Group 4 might see:   metaphor for fabric    for dishwasher detergent
                     softener

Participants in      Verbal metaphor for    Verbal literal for
Group 1 might see:   window cleaner         fabric softener

Participants in      Verbal metaphor for    Verbal literal for
Group 2 might see:   fabric softener        dishwasher detergent

Participants in      Verbal metaphor for    Verbal literal for
Group 3 might see:   dishwasher detergent   slider bag

Participants in      Verbal metaphor for    Verbal literal for
Group 4 might see:   slider bag             window cleaner

Note: Order of presentation of the four test ads was randomized for
each participant. Only 8 of 24 possible pairings of the four
treatments with the four products were administered.

TABLE 3
Agreement Proportions for Weak Implicatures
for Each Treatment Condition

                               Literal headline

Visual metaphor          Present (%)              Absent (%)

Absent                   21.1 (a),(b)                28.2
                       (Verbal literal)        (Verbal metaphor)

Present                      34.5                    40.9
                  (Anchored visual metaphor)   (Visual metaphor)

(a) Numbers shown represent the proportion of occasions (n = 531,
representing 177 participants responding to 3 weak-implicature
statements) where a participant, when exposed to the indicated
stimulus condition and confronted with a weak-implicature statement,
responded "yes" to the prompt "The ad was designed to make you
think that."

(b) All adjacent pairwise comparisons (verbal literal versus verbal
metaphor, verbal metaphor versus anchored visual metaphor, anchored
visual metaphor versus unanchored visual metaphor) are significant
at p < .05.

TABLE 4
Response Latencies (Milliseconds) (a) for Weak Implicatures

                                Literal headline

                      Present                 Absent

Visual metaphor      Yes       No        Yes           No

Absent            2,130 (c)   2,065   2,161 (b)   2,009 (b),(d)
                  (Verbal literal)       (Verbal metaphor)

Present           2,296 (c)   2,153   2,073 (b)   2,232 (b),(d)
                  (Anchored visual       (Visual metaphor)
                      metaphor)

(a) Raw latencies were inverted prior to analysis by linear mixed
models. Means output by the LMM (Linear Mixed Models) analysis
were then reinverted for purposes of display. The LSD (Least
Significant Difference) adjustment was applied in conducting
pairwise comparisons of cell means.

(b) Indicates that latencies for "yes" and "no" responses
within this stimulus type were significantly different
(p [less than or equal to] .05).

(c) Indicates that latencies for "yes" responses across these
two stimulus types were significantly different (p < .01).

(d) Indicates that latencies for "no" responses across these
two stimulus types were significantly different (p < .001).

REFERENCES

Bargh, John A., and Tanya L. Chartrand (2000), "The Mind in the Middle: A Practical Guide to Priming and Automaticity Research," in Handbook of Research Methods in Social Psychology, Henry Reis and Charles Judd, eds., New York: Cambridge University Press, 253-285.

Burton, Phillip, and Scott Purvis (1996), Which Ad Pulled Best? 8th ed., Chicago: NTC Publishing Group.

Calkins, Earnest (1928), Business the Civiliser, Boston: Little, Brown.

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

Dick, Alan, Dipankar Chakravarti, and Gabriel Biehal (1990), "Memory-Based Inferences During Consumer Choice," Journal of Consumer Research, 17 (June), 82-93.

Eco, Umberto (1976), A Theory of Semiotics, Bloomington: Indiana University Press.

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

Fazio, Russell (1990), "A Practical Guide to the Use of Response Latency in Social Psychological Research," in Research Methods in Personality and Social Psychology. Review of Personality and Social Psychology, vol. 11, Clyde Hendrick and Margaret S. Clark, eds., Newbury Park, CA: Sage, 74-97.

Finn, Adam (1988), "Print Ad Recognition Readership Scores: An Information Processing Perspective," Journal of Marketing Research, 25 (May), 168-177.

Forceville, Charles (1996), Pictorial Metaphor in Advertising, New York: Routledge.

Gaeth, Gary J., and Timothy B. Heath (1987), "The Cognitive Processing of Misleading Advertising in Young and Old Adults: Assessment and Training," Journal of Consumer Research, 14 (1), 43-54.

Gibbs, Raymond W., Jr. (1994), The Poetics of Mind, Cambridge: Cambridge University Press.

Hawkins, Scott A., Stephen J. Hoch, and Joan Meyers-Levy (2001), "Low-Involvement Learning: Repetition and Coherence in Familiarity and Belief," Journal of Consumer Psychology, 11 (1), 1-11.

Hitchon, Jacqueline C. (1997), "The Locus of Metaphorical Persuasion: An Empirical Test," Journalism and Mass Communication Quarterly, 74 (1), 55-68.

Jewell, Robert D., and H. Rao Unnava (2003), "When Competitive Interference Can Be Beneficial," Journal of Consumer Research, 30 (September), 283-291.

Johar, Gita (1995), "Consumer Involvement and Deception from Implied Advertising Claims," Journal of Marketing Research, 32 (August), 267-279.

Kardes, Frank R. (1988), "Spontaneous Inference Processes in Advertising: The Effects of Conclusion Omission and Involvement on Persuasion," Journal of Consumer Research, 15 (September), 225-233.

--(1993), "Consumer Inference: Determinants, Consequences, and Implications for Advertising," in Advertising Exposure, Memory, and Choice, Andrew A. Mitchell, ed., Hillsdale, NJ: Lawrence Erlbaum, 163-191.

Kirk, Roger E. (1995), Experimental Design: Procedures for Behavioral Sciences, 3rd ed., New York: Wadsworth.

Lee, Dong Hwan, and Richard W. Olshavsky (1995), "Conditions and Consequences of Spontaneous Inference Generation: A Concurrent Protocol Approach," Organizational Behavior and Human Decision Processes, 61 (2), 177-189.

Leigh, James H. (1994), "The Use of Figures of Speech in Print Ad Headlines," Journal of Advertising, 23 (2), 17-33.

MacCormac, Earl R. (1985), A Cognitive Theory of Metaphor, Cambridge: MIT Press.

MacKenzie, Scott B., and Richard L. Lutz (1989), "An Empirical Examination of the Structural Antecedents of Attitude Toward the Ad in an Advertising Pretesting Context," Journal of Marketing, 53 (April), 48-65.

Marchand, Roland (1985), Advertising and the American Dream: Making Way for Modernity, 1920-1940, Berkeley: University of California Press.

McCulloch, Charles E., and Shayle R. Searles (2001), Generalized, Linear, and Mixed Models, New York: John Wiley.

McQuarrie, Edward F., and David G. Mick (1996), "Figures of Rhetoric in Advertising Language," Journal of Consumer Research, 22 (March), 424-438.

--, and -- (1999), "Visual Rhetoric in Advertising: Text-Interpretive, Experimental, and Reader-Response Analyses," Journal of Consumer Research, 26 (June), 37-54.

---, and -- (2003), "Visual and Verbal Rhetorical Figures Under Directed Processing Versus Incidental Exposure to Advertising," Journal of Consumer Research, 29 (March), 579-587.

Messaris, Paul (1997), Visual Persuasion: The Role of Images in Advertising, Thousand Oaks, CA: Sage.

Meyers-Levy, Joan, and Laura A. Peracchio (1992), "Getting an Angle in Advertising: The Effect of Camera Angle on Product Evaluations," Journal of Marketing Research, 29 (November), 454-461.

Miniard, Paul W., Sunil Bhatla, Kenneth R. Lord, Peter R. Dickson, and H. Rao Unnava (1991), "Picture-Based Persuasion Processes and the Moderating Role of Involvement," Journal of Consumer Research, 18 (June), 92-107.

Mitchell, Andrew A., and Jerry C. Olson (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects of Brand Attitude?" Journal of Marketing Research, 18 (August), 318-332.

Mothersbaugh, David L., Bruce A. Huhmann, and George R. Franke (2002), "Combinatory and Separative Effects of Rhetorical Figures on Consumers' Efforts and Focus in Ad Processing," Journal of Consumer Research, 28 (March), 589-602.

Peracchio, Laura A., and Joan Meyers-Levy (1994), "How Ambiguous Cropped Objects in Ad Photos Affect Product Evaluations," Journal of Consumer Research, 21 (June), 190-204.

Phillips, Barbara J. (1997), "Thinking Into It: Consumer Interpretation of Complex Advertising Images," Journal of Advertising, 26 (2), 77-87.

--(2000), "The Impact of Verbal Anchoring on Consumer Response to Image Ads," Journal of Advertising, 29 (1), 15-24.

--(2003), "Understanding Visual Metaphor in Advertising," in Persuasive Imagery: A Consumer Response Perspective, Linda M. Scott and Rajeev Batra, eds., Mahwah, NJ: Lawrence Erlbaum Associates, 297-310.

--, and Edward F. McQuarrie (2003), "The Development, Change, and Transformation of Rhetorical Style in Magazine Advertisements, 1954-1999," Journal of Advertising, 31 (4), 1-13.

Pollay, Richard W. (1985), "The Subsidizing Sizzle: A Descriptive History of Print Advertising, 1900-1980," Journal of Marketing, 48 (Summer), 24-37.

--(1991), "Signs and Symbols in American Cigarette Advertising," in Marketing and Semiotics: Selected Papers from the Copenhagen Symposium, Hanne Hartvig Larsen, David Glen Mick, and Christian Alsted, eds., Copenhagen: Handelshojskolens Forlag.

Pollio, Howard R., Michael K. Smith, and Marilyn R. Pollio (1990), "Figurative Language and Cognitive Psychology," Language and Cognitive Processes, 5 (2), 141-167.

Ratcliff, Roger (1993), "Methods for Dealing: with Reaction Time Outliers," Psychological Bulletin, 114 (3), 510-532.

Rossiter, John R., and Larry Percy (1983), "Visual Communication in Advertising," in Information Processing Research in Advertising, Richard Jackson Harris, ed., Hillsdale, NJ: Lawrence Erlbaum Associates, 83-125.

Santamaria, Carlos, Juan A. Garcia-Madruga, and Mario Carretero (1996), "Beyond Belief Bias: Reasoning from Conceptual Structures by Mental Models Manipulation," Memory and Cognition, 24 (2), 250-261.

Sawyer, Alan G., and Daniel J. Howard (1991), "Effects of Omitting Conclusions in Advertisements to Involved and Uninvolved Audiences," Journal of Marketing Research, 28 (November), 467-474.

Smith, Ruth Ann (1991), "The Effects of Visual and Verbal Advertising Information on Consumers' Inferences," Journal of Advertising, 20 (4), 13-24.

Sperber, Dan, and Diedre Wilson (1986), Relevance: Communication and Cognition, Oxford: Basil Blackwell.

Stern, Barbara (1990), "Beauty and Joy in Metaphorical Advertising: The Poetic Dimension," in Advances in Consumer Research, Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, eds., Provo, UT: Association for Consumer Research, 17, 71-77.

--(1992), "'Crafty Advertisers': Literal Versus Literary Deceptiveness," Journal of Public Polio; and Marketing, 11 (1), 72-81.

Tanaka, Keiko (1994), Advertising Language: A Pragmatic Approach to Advertisements in Britain and Japan, London: Routledge.

Tom, Gail, and Annmarie Eves (1999), "The Use of Rhetorical Devices in Advertising," Journal of Advertising Research, 39 (4), 39-43.

Toncar, Mark, and James Munch (2001), "Consumer Response to Tropes in Print Advertising," Journal of Advertising, 30 (1), 55-65.

Ulrich, Roll, and Jeff Miller (1994), "Effects of Truncation on Reaction Time Analysis," Journal of Experimental Psychology: General, 123 (1), 34-80.

Van Zandt, Trisha (2002), "Analysis of Response Time Distributions," in Steven's Handbook of Experimental Psychology, 3rd ed., Hal Pashler and John Wixted, eds., New York: John Wiley.

Wallace, Dennis, and Samuel B. Green (2002), "Analysis of Repeated Measures Designs with Linear Mixed Models," in Modeling Intra-individual Variability with Repeated Measures: Methods and Applications, D. S. Moskowitz and Scott L. Hershberger, eds., Mahwah, NJ: Lawrence Erlbaum Associates, 103-134.

Ward, James, and William Gaidis (1990), "Metaphor in Promotional Communication: A Review of Research on Metaphor Comprehension and Quality," in Advances in Consumer Research, Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, eds., Provo, UT: Association for ,Consumer Research, 17, 636-642.

White, Hooper (1981), How to Produce an Effective TV Commercial, Chicago: Crain Books.

Edward F. McQuarrie (Ph.D., University of Cincinnati) is a professor of marketing in the Leavey School of Business, Santa Clara University.

Barbara J. Phillips (Ph.D., University of Texas at Austin) is a professor of marketing in the Department of Management and Marketing, University of Saskatchewan.