Goal Control of Attention to Advertising: The Yarbus Implication

RIK PIETERS MICHEL WEDEL*

An eye-tracking experiment with four processing goals and a free-viewing condition reveals goal control of attention even during a few seconds of self-paced ad exposure. An ad-memorization goal enhanced attention to the body text, pictorial, and brand design objects. A brand-learning goal enhanced attention to the body text but simultaneously inhibited attention to the pictorial design. This supports the thesis that ad informativeness is goal contingent. Differences in pupillary diameter between ad objects but not between processing goals reect the pupils role in maintaining optimal vision. Implications of the ndings for advertising theory and avenues for future research are indicated.

hat do consumers look at in advertisements, and do the goals that they have in mind systematically inuence what they attend to? Is attention predominantly a dumb process guided by the perceptual salience of the ad stimuli, after which the goals of consumers control higherorder cognitive processes? Or, do consumers processing goals already have early and systematic effects on patterns of visual attention to ads? In view of the potential implications of these questions for advertising theory and the importance of advertising to brand and rm performance, it is surprising that almost no research has directly addressed them. Not only in advertising but also in attention research at large, most research effort thus far has been devoted to understanding reexive control by stimulus factors (Itti 2005; Treisman and Gelade 1980), in advertising mainly the size of ads and the objects contained therein (Janiszewski 1998; Lohse 1997; Pieters and Wedel 2004). As a result, the role of goals in attention to ads (and other complex visual stimuli) is still largely unknown, which has been identied as an important gap in our knowledge (Henderson and Hollingworth 1998; Kingstone et al. 2003; Pashler, Johnston, and Ruthruff 2001). The pioneering work of Alfred Yarbus (1967) offers an optimistic initial answer. He recorded the eye movements

*Correspondence can be addressed to Rik Pieters, professor of marketing, Tilburg University, Department of Marketing, P.O. Box 90153, The Netherlands; e-mail [email protected], or Michel Wedel, Pepsico Professor of Consumer Science, Smith School of Business, University of Maryland, College Park, MD 20742-1815; e-mail [email protected]. We thank Verify International for the eye-tracking data. John Deighton served as editor and Donald Lehmann served as associate editor for this article. Electronically published June 7, 2007

of a single participant, who looked at the painting The Unexpected Visitor by Ilya Repin, under different task sets for 3 minutes each. Inspection of the raw eye recordings during each of these specic tasks led to the conclusion that depending on the task in which a person is engaged, that is, depending on the character of the information that he must obtain, the distribution of the points of xation on an object will vary correspondingly, because different items of information are usually located in different parts of an object (Yarbus 1967, 19). In other words, Yarbuss thesis is that the informativeness of objects in a scene is contingent on the specic processing goal that is activated and that eye movements reect the extent to which these objects are informative. Our research pursues the implications for advertising. It investigates the inuence of consumers processing goals on visual attention to ads and four design objects that they comprise, to assess the extent to which advertising informativeness is contingent on consumers processing goals. However, it is not at all evident that there should be systematic effects of processing goals on attention to advertising under natural exposure conditions. This is because the top-down, voluntary control of attention by goals involves slower serial processes than the fast parallel processes underlying bottom-up, reexive control by stimulus factors (Treisman and Gelade 1980; Wolfe 2005). Therefore, processing goal effects may have a lower likelihood of surfacing during the few seconds that consumers typically spend on ads during self-paced exposure than when viewing The Unexpected Visitor for a forced exposure of 3 minutes. Moreover, consumers are usually exposed to multiple ads surrounded by editorial material rather than to a single isolated scene, and the ensuing competitive clutter may favor reexive control and hinder systematic goal control (Itti 2005).
224
2007 by JOURNAL OF CONSUMER RESEARCH, Inc. Vol. 34 August 2007 All rights reserved. 0093-5301/2007/3402-0009$10.00

GOAL CONTROL OF ATTENTION TO ADVERTISING

225

Heterogeneity across consumers (Rosbergen, Pieters, and Wedel 1997) further impedes the expression of goals in identiable attention patterns. In addition, processing goals during ad exposure are typically more abstract than Yarbuss concrete instructions to search for specic cues, such as the ages of the people in a painting, which may further attenuate reliable goal effects to emerge. In fact, in an inuential review, Viviani (1990) questioned more generally whether any higher-order mental process would be systematically reected in eye movements. It is thus not apparent that reliable goal effects on attention to advertising exist at all. Therefore, we rst propose a conceptual framework that describes how visual attention to objects in ads may be contingent on consumers processing goals. We focus on four general design objects, each with a specic function in advertising, that is, the brand, pictorial, headline, and body text, rather than on idiosyncratic, natural objects, which are present in some but not in other ads and are difcult to dene uniquely. This enables predictions about the informativeness of these design objects under particular processing goals, which we test in an experiment with a baseline condition of free viewing and four processing goals, with 220 participants and 17 ads, using eye movements and pupillary diameter as process measures of attention.

ameter has been found to enlarge with rising processing intensity, it is deemed a reporter variable for the latter, although there is no direct causal relationship, and the underlying cortical mechanism is still unknown (Beatty and Lucero-Wagoner 2000). Attention selection and engagement are determined bottom-up by the salience of ad objects and top-down by their informativeness to consumer goals. A large psychological literature addresses salience and informativeness. We next review these two concepts.

Ad Objects and Salience

The brand, pictorial, headline, and body text are design objects with specic functions in ads (Book and Schick 1997). The brand contains all textual and pictorial signs that uniquely identify the brand, such as its name, logo, and trademark. The pictorial conveys one or more aspects of the message in (photo)graphic form and often shows the product in its usage context. The headline contains the leading text, usually in large font size, expressing the key message of the ad and anchoring the meaning of the pictorial. The body text is in smaller font size and provides support for the message claim and details about features and benets. Ad objects capture attention reexively and immediately when they are salient. The salience of an ad object depends on its local contrast with other objects on basic perceptual features, such as surface size, form, color, and luminance (Itti 2005; Wolfe 2005). The surface sizespaceof ad objects is a perceptual feature with a privileged status in visual attention (Yantis 2000, 80). First, a larger surface size of an object facilitates gure-ground segmentation and increases its salience, so that larger objects tend to pop out (Treisman and Gelade 1980). Second, because the eyes move in space, objects that occupy more of the ads space have a higher likelihood of being xated, even if eye xations would be randomly distributed. In past research on visual attention to advertising, surface size effects on attention have been found to be pervasive (Pieters and Wedel 2004; Wedel and Pieters 2000). It is therefore important to control for the surface size of ad objects, in order to make accurate and unbiased inferences about their informativeness, based on eye movements (Rayner et al. 2001).

GOAL CONTROL OF ATTENTION TO ADVERTISING

The sheer volume of visual stimuli that simultaneously call for attention surpasses the processing capacities of any consumer. Attention therefore selects specic stimuli and the objects they contain and further engages in these, which occurs covertly in the visual brain. In most real-world situations, covert attention and overt attention, which are expressed in observable eye movements, are tightly connected such that covert attention can be inferred from patterns of eye movements (Findlay 2005).

Attention Selection and Engagement

Eye movements on ads involve xationsmoments that the eye is relatively stableand saccadesballistic movements between xation locations to redirect the gaze. During a xation an object or location in the ad is projected via the line of sight onto the fovea, the small area of the retina with the highest acuity, for detailed visual processing (Rayner 1998). Attention selection involves bringing an ad object into the focus of attention and is reected in at least one eye movement to xate it. Attention engagement is the process of sustaining attention to an already selected object, and it has duration and intensity components. Duration of attention is reected in longer gaze durations on the object, which is achieved by rexations (Russo and Leclerc 1994). Intensity of attention has been linked to changes in pupillary diameter. The pupil is the eyes aperture through which light passes, and changes in its diameter ensure optimal vision under varying conditions of luminance, distance to the stimulus, and required visual angle. Because the pupillary di-

Processing Goals and Informativeness

Consumers may freely view and explore ads without a particular task in mind (Janiszewski 1998; Kahneman 1973). But they may also pursue a specic goal, such as learning new information about the advertised brand or determining the attractiveness of the ad (Lichtenstein and Srull 1987). Processing goals focus consumers attention on the stimuli that they are exposed to and on the information contained in them. Such goals are of intermediate abstractness between low-level concrete goals, such as gauging the age of people (Yarbus 1967) or searching for a specic target in a scene (Wolfe 2005), and high-level abstract goals, such as the pursuit of freedom or health (Austin and Vancouver 1996).

226

JOURNAL OF CONSUMER RESEARCH

Processing goals vary in target and content. The goal target can be the ad as a whole or the brand that is advertised (Keller 1987). With respect to goal content, we distinguish evaluation goals (Hilton and Darley 1991; Wyer and Srull 1989), which involve judging the value of the target on a specic dimension (such as its usefulness or attractiveness), from learning goals (Dweck and Leggett 1988), which involve acquiring knowledge about the target and storing it in memory for later use. Only a few of these different processing goals have been used in prior attention research (Henderson and Hollingworth 1998) and mostly without explicit comparisons between them. We examine the inuence of four processing goals, as dened by their target and content, spanning a wide range of possible intermediate processing goals, and compare these to a baseline condition of free viewing.1 Objects in ads receive attention voluntarily when they are more informative. That is, processing goals activate representations of objects in long-term visual memory that are perceived to be relevant to the goal (Friedman 1979). These representations are held in working memory and compared with the objects actually appearing in the ad, which happens faster and more accurately to the extent that they match. Activating a processing goal may enhance or inhibit the association with certain objects. For example, when the participant in the study of Yarbus (1967) tried to determine the material circumstances of the family depicted in the painting of The Unexpected Visitor, attention to the furniture in the room was enhanced but attention to the faces and clothing of the people was inhibited. The nal attention priorities of ad objects are the integration of their salience and informativeness, which have similar effects on attention (Spratling and Johnson 2004). Attention selects and engages in the prioritized ad objects through enhancement and inhibition, leading to eye movements to foveate them. Information extracted during eye xations on ad objects updates the attention priorities until the goals are sufciently attained and attention moves on. Surprisingly, previous quantitative research (reviewed in Henderson and Hollingworth 1998) has typically compared attention to different objects in the same scene, or to the same object in different scenes, to determine the informativeness of these objects. This assumes that informativeness is a property of objects per se or of their relationship with the rest of the scene, which is misguided. Instead, and in line with the Yarbus thesis, we propose that the informativeness of an ad object resides in its linkage to a processing goal, which thus can only be revealed by comparing attention to that ad object between different processing goals. That is, because their salience is invariant across goals, the informativeness of ad objects can be separated from their salience by comparing attention to them between different processing goals.
1 Free viewing serves as a base condition that has been prevalent in advertising and scene perception research yet may itself consist of individual specic goals.

PREDICTIONS
The current embryonic stage of knowledge about goal control of attention to advertising and other complex scenes (Kingstone et al. 2003; Pashler et al. 2001) permits a number of broad predictions only, which we present here. We predict that attention to ad objects is highest for learning goals, intermediate for evaluation goals, and lowest for free viewing, based on the following reasoning. During free viewing, object salience primarily determines attention, which happens quickly (Friedman 1979; Janiszewski 1998). Conversely, when a specic processing goal is activated, information from the ads needs to be processed, evaluated, integrated, and stored, which requires attention. Thus, we expect that less attention will be deployed during free viewing than under each of the processing goals. Further, we predict higher levels of visual attention under learning goals than under evaluation goals. This is because focused attention is required during learning tasks to determine the novelty and quality of the information in the ad, to integrate this with existing knowledge, and to store the results in memory for future use, which takes attention time. However, evaluation is generally a fast and immediate process (Ferguson and Bargh 2003), in particular if it concerns feelings of appreciation, so that the divergence in attention to ad objects between learning goals versus evaluation goals and free viewing will be large. H1: Attention to ad objects is largest for learning goals, intermediate for evaluation goals, and least for free viewing. We predict that compared to the brand, pictorial, and body text, attention to the headline will be least prone to processing goal control, because of the headlines perceptual salience and coordinating role in advertising. That is, the headline stands out due to its large font type and central position in the ad, and it summarizes the ads main message and anchors the meanings in the pictorial (Book and Schick 1997) so that it is highly salient and informative across all processing goals (Rayner et al. 2001). In line with this, attention engagement with the headline diminishes little across repeated ad exposures (Pieters, Rosbergen, and Hartog 1996). H2: Compared to the brand, pictorial, and body text, attention to the headline is least prone to processing goal control. More specically, one would clearly expect attention to the brand object to be higher under brand-directed than under ad-directed goals and free viewing (Keller 1987). Under a brand-directed goal, consumers need to retrieve brandrelevant information from memory and match this with the information about the brand in the ad, which takes time. Under an ad-directed goal and free viewing, however, the brand is not as informative, and attention to it should be less and mostly determined by perceptual salience. Second, we predict attention to the pictorial to be highest

GOAL CONTROL OF ATTENTION TO ADVERTISING

227

under an ad-memorization goal, when the aim is to learn the perceptual features of the ad in detail. Memory for pictures is vast and efcient (Childers and Houston 1984; Costley and Brucks 1992), which should stimulate consumers to prioritize picture processing when learning ads for future recollection. Third, we predict attention to the body text to be highest for brand learning. Text processing is an effortful and slow process (Rayner 1998), on which usually less than a second is spent (Wedel and Pieters 2000). However, the body text in print ads typically provides detailed information, which is relevant to learn about the brand. Moreover, there appears to be a cultural predisposition to anticipate such detailed, precise information to reside in the body text rather than the pictorial of ads (Kenney and Scott 2003). Therefore, one would expect enhancement of attention to the body text of print ads under brand learning, even if it would not contain the desired or other relevant information. H3: Attention is highest for the brand under branddirected goals, for the pictorial under ad-learning goals, and for the body text under brand-learning goals. Support for the hypotheses would point out that far from being a mechanic, dumb process driven reexively by the salience of ad stimuli and the objects contained therein, visual attention to advertising adapts rapidly to the requirements of specic processing goals, such that higher-order processes can be systematically inferred from it. It would reveal how the informativeness of ads is contingent on the specic processing goal that is activated.

canned fruit (Del Monte), pasta sauce (Heinz), ice cream (Hertog), sausage (Huls), frozen pizza (Iglo), cookies (Liga), cake (Maitre Paul), bread (Meggle), cheese (Milner), spices (Pataks), healthy snacks (Quaker), frying oil (Remia), soft drink (Sisi), fruit juice (Tropicana), and dairy drink (Yakult). All ads contained brand, pictorial, headline, and body text design objects. The surface sizes of the four ad objects in each of the 17 target ads were determined with specialized software by drawing the appropriate boxes and polygons around them. In case the brand name appeared in the headline, it was counted to the brand object.

Data Collection
Data collection was done in cooperation with Verify International, a company that specializes in eye-tracking research. All instructions and stimuli were presented on NEC 21-inch LCD monitors in full-color bitmaps with a 1,280 # 1,024 pixel resolution. Throughout the experiment, participants continued to the next page by touching the lower right corner of the (touch-sensitive) screen, as when paging. Infrared corneal reection methodology was used for eye tracking (Duchowski 2003). The specic equipment (Verify International) leaves participants free to move their heads in a virtual box of about 30 centimeters. Cameras track the position of the eye and head, allowing continuous correction of position shifts. Measurement precision of the eye-tracking equipment is better than 0.5 degree of visual angle. Upon entering the facilities of the market research company, participants were seated behind the eye trackers. After a brief warm-up eye-tracking task, they read the following instructions on the computer screen: Now a set of advertisements follows, with the subsequent text for each of the ve conditions: baseline free viewing: Explore these freely as you would at home or in a waiting room; ad learning (memorization): Try to memorize the advertisements. Try to remember each advertisement as well as possible; ad evaluation (appreciation): Determine how attractive or unattractive each advertisement is to you. Try to take account of the beautiful and ugly aspects of each advertisement in your judgment; brand learning: Use the advertisements to collect information about the brand products. Try to learn something new about each branded product; brand evaluation: Use the advertisements to determine how good or bad the brand products are. Try to form an evaluation of each branded product using the advertisement. In the remainder of the article we will use the terms ad-memorization and ad-appreciation goals to denote our specic operationalization and to distinguish them from the brandlearning and brand-evaluation goals. Each of the 17 target ads was shown with the accompanying editorial page from the magazine, preceded by the front cover and followed by the back cover of the magazine. Since previous research has shown serial-position effects on attention to print ads (Wedel and Pieters 2000), the order of ads was randomized across individuals. Nine ads were on the right page and eight were on the left page.

EXPERIMENT Method
Two hundred and twenty regular consumers were recruited by a market research agency for the experiment (52% females; 27% between 16 and 25 years of age, 15% between 26 and 35 years, 39% between 36 and 45 years, and 19% between 46 and 55 years). All had normal or corrected-to-normal vision, and none had participated in eye-tracking research before. They received $15 for participation. Participants were randomly assigned to one of ve conditions of a one-factor between-subjects design, that is, the baseline free-viewing condition and, respectively, (1) ad memorization, (2) ad appreciation, (3) brand learning, and (4) brand evaluation. All material in the ve conditions was identical, except for the processing goal instruction described later. All 17 full-page ads from the most recent issue of Allerhande magazine were selected. This is a free, popular magazine of the largest retailer in the Netherlands, containing articles and ads about food and cooking, and none of the participants had seen this issue before (although some may have been familiar with some ads). The ads were for frozen vegetables (brand: Bonduelle), food wraps (Casa Fiesta),

228
TABLE 1

JOURNAL OF CONSUMER RESEARCH

RAW ATTENTION DATA BY PROCESSING GOALS ACROSS TARGET ADS Advertisement object Brand Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD .738 .546 .856 .351 .839 .839 2.951 1.351 .844 .363 .723 .781 3.001 1.409 .877 .329 1.055 .995 2.975 1.285 .828 .378 .759 .731 2.858 1.441 .859 .348 .821 .846 2.975 1.371 .873 .333 .844 .785 2.944 1.233 Pictorial 3.277 .797 .969 .173 1.508 1.293 3.388 .883 .973 .164 1.313 1.190 3.479 .840 .990 .097 1.900 1.417 3.409 .758 .929 .257 1.339 1.052 3.266 1.108 .980 .139 1.557 1.398 3.414 .856 .973 .163 1.443 1.288 3.369 .796 Headline .664 .304 .814 .389 .901 .952 2.817 1.473 .803 .398 .833 .907 2.849 1.515 .839 .368 1.018 1.025 2.866 1.436 .630 .483 .585 .955 2.179 1.758 .793 .405 .903 1.091 2.755 1.538 .863 .344 .976 .920 2.928 1.280 Body text .894 .546 .666 .472 .821 1.550 2.291 1.706 .610 .488 .611 1.167 2.187 1.809 .696 .460 .939 1.733 2.357 1.666 .777 .417 .778 .765 2.693 1.561 .706 .456 1.054 1.955 2.419 1.660 .689 .463 .921 1.659 2.317 1.618

Processing goal and measures Surface size (dm2) Overall (n p 220): Selection (probability) Duration (sec) Pupillary diameter (mm) Free viewing (n p 45): Selection (probability) Duration (sec) Pupillary diameter (mm) Ad memorization (n p 45): Selection (probability) Duration (sec) Pupillary diameter (mm) Ad appreciation (n p 44): Selection (probability) Duration (sec) Pupillary diameter (mm) Brand learning (n p 45): Selection (probability) Duration (sec) Pupillary diameter (mm) Brand evaluation (n p 43): Selection (probability) Duration (sec) Pupillary diameter (mm)

Visual Attention Indicators and Analysis Approach

Attention was indicated by (1) selection, (2) gaze duration, and (3) pupillary diameter. Selection denotes whether the particular ad object is xated (1) or not (0) by a particular participant. Gaze duration is the total time (in seconds) that a participant who selected a particular ad object attends it. Pupillary diameter is measured in millimeters during attention engagement with a particular ad object. Since the distance between a participants eye and the ad may inuence pupillary diameter and possibly other eye-tracking measures (Beatty and Lucero-Wagoner 2000), we use it as a control

variable in the analyses. Table 1 gives a summary of the attention indicators by goal condition and ad element. The characteristics of our data make the application of standard ANOVA techniques problematic. That is, the dependent variables and control are selection, gaze duration, pupillary diameter, and eyead distance measured on brand, pictorial, headline, and body text for 17 ads and for the 220 participants assigned to one of the free-viewing, ad-memorization, ad-appreciation, brand-learning, and brand-evaluation goal conditions. Attention selection is a binary variable (yes-no), and gaze duration, pupillary diameter, and eyead distance are positiveobserved only if an ad object is selected by attentionand correlated. Furthermore, in-

GOAL CONTROL OF ATTENTION TO ADVERTISING

TABLE 2

229

INFLUENCE OF PROCESSING GOALS ON DURATION OF ATTENTION: MODEL-BASED MEANS AND STANDARD DEVIATIONS Processing goals Attention indicator and ad object Overall Gaze duration: Brand Pictorial Headline Body text Mean SD Mean SD Mean SD Mean SD Mean SD (Baseline) free viewing .888a .120 .923a .104 1.133b .202 1.247 .121 .251a .212 Advertisement Memorization 1.238c .122 1.342b .106 1.459c .209 1.395 .126 .759b .215 Appreciation .847a .123 1.000a .107 1.029b .210 1.085 .124 .274a .221 Learning 1.047b .118 1.030a .103 .871a .205 1.319 .124 .967c .213 Brand Evaluation 1.056b .122 1.047a .107 1.074b .211 1.379 .123 .723b .219

NOTE. Boldface font indicates coefcients with a 95% posterior credible interval that does not cover zero. Coefcients with different superscripts row-wise have posterior credible intervals that overlap less than 5%.

dependent of processing goals, the surface sizes of ad objects strongly affect attention, and these effects need to be accounted for. Finally, goals vary between participants, and participants are heterogeneous (Hutchinson, Kamakura, and Lynch 2000). To fully account for these aspects of the data-generating mechanism, we developed a statistical model (see the appendix) formulated and estimated in the Bayesian framework (Rossi, Allenby, and McCulloch 2005). In the model, processing goals inuence attention parameters hierarchically (top-down), consistent with recent views in the visual perception literature (Lee and Mumford 2003). We compute quantities that are directly interpretable as means of the attention measures for each of the ve goal conditions and the standard deviations of differences between them. We do this as an integral part of the Bayesian estimation procedure, which enables appropriate statistical testing based on accurate assumptions about the data-generating mechanism and additionally provides results that can be interpreted as those of ANOVA.

RESULTS
Attention selection ranges between 97% for the pictorial and 67% for the body text (table 1). Average gaze duration is between 1.54 seconds for the pictorial and 0.82 seconds for the body text, and average pupillary diameter is between 3.39 mm for the pictorial and 2.29 mm for the body text. The average time that participants attended to each of the ads is a mere 4.1 seconds.

Overall Differences between Processing Goals

The statistical analysis reveals that there are no differences between processing goals in attention selection and the intensity of attention engagement. Goal effects on selection may be absent because ads contain four perceptually distinct objects that vary on multiple perceptual features, which

makes selecting them easyselection of the average ad object is around 90%and the activated goals do not call for the selection of a single object at the expense of all others, as is the case in target search tasks. Also, the intensity of attention engagement, as reected in pupillary diameter, does not differ signicantly between processing goal conditions, varying from 3.43 to 3.55 millimeters. Since exposure to print ads is self-paced, gaze durations are around 4 seconds only and easily expandable, so that it may be more efcient to adapt the required attention resources for a particular goal by adjusting the duration rather than the intensity of attention per unit of time, that is, by attending longer instead of harder. Processing goals did have the predicted systematic inuences on the duration of attention engagement, as revealed by the model-based adjusted means in table 2. The gains and losses in duration of attention under the four processing goals relative to free viewing are plotted in the left part of gure 1. In support of hypothesis 1, the duration of attention to the average ad object was much higher under the two learning goals, ad memorization (1.238 seconds) and brand learning (1.047 seconds), than under free viewing (0.888 seconds), an increase of 39% and 18%, respectively. Unexpectedly and counter to hypothesis 1, the duration of attention under the ad appreciation goal (0.847 seconds) was as low as under free viewing, rather than being intermediate. Attention engagement under brand evaluation (1.056 seconds) was as high as under brand learning. The average duration of attention to ads as a whole under ad appreciation and free viewing were, respectively, 3.388 and 3.552 seconds only, which indicates the rapidity with which ad-appreciation goals were implemented. Although the surface sizes of ad objects had signicant effects on all attention measures, there were no differences between processing goals in the effects of the surface sizes on attention. This underlines the goal control of attention to the ad objects per se, regardless of their size.

230
FIGURE 1 GOAL CONTROL OF ATTENTION DURATION: GAINS AND LOSSES RELATIVE TO THE FREE-VIEWING CONDITION

JOURNAL OF CONSUMER RESEARCH

Differences in Attention to Specic Ad Objects

We present the signicant effects of goals on the duration of attention engagement with the ad objects in table 2 (none of the other effects were signicant). The predicted inuence of processing goals on attention engagement to specic ad objects is prominent. To facilitate comparisons of these processing goal effects on each of the ad objects, we plot the gains and losses relative to the baseline free-viewing condition in the right part of gure 1. In support of hypothesis 2, there are no signicant differences between processing goals in duration of attention to the headline but large differences for the other three ad objects. Note that across the 17 ads, the average number of words in the headline (4.8) is much smaller than the average number of words in the body text (61.1). Still, the duration of attention to the headline across the ve goal conditions is 2.2 times longer (1.285 seconds) than to the body text (0.595 seconds) and sufcient to fully absorb its meaning (Rayner 1998). This underscores the high informativeness of the headline, irrespective of the activated goal, and reveals the quite different impact that goals have on attention to the headline and body text, although both are text. In support of hypothesis 3, duration of attention to the pictorial was highest under ad memorization (1.459 seconds) and much higher than under free viewing (1.133 seconds; the average gain under ad memorization being 326 milliseconds across the 17 target ads). In further support of hy-

pothesis 3, duration of attention to the body text was highest under brand learning (0.967 seconds) and much higher than under free viewing (0.251 seconds, a gain of 285%). Counter to our predictions however, attention to the brand was not higher under brand-directed than under ad-directed goals. On the contrary, attention to the brand was highest under an ad-memorization goal (1.342 seconds) and equal to or less than that amount under the other three processing goals and free viewing (1.000 seconds, on average). The attention advantage of over 300 milliseconds for ad memorization is large in view of the overall short gaze durations. We return to this unexpected result in the nal section. Figure 1 gives the complete attention pattern for each processing goal relative to the free-viewing situation. It demonstrates how ad memorization clearly raises attention to the brand, pictorial, and body text. The attention pattern under ad appreciation is remarkably similar to that under free viewing, which points to the possibility that consumers partially adopted an ad-appreciation goal during free viewing, which we will discuss later. Recall that in support of hypothesis 3, duration of attention to the body text was largest under brand learning. Figure 1 shows that this attention gain for the body text comes with a simultaneous attention loss for the pictorial under this goal. Duration of attention to the pictorial was 1.133 seconds during free viewing but dropped to 0.871 seconds during brand learning (a 23% loss). This pattern of simultaneous enhancement and inhibition of attention is remarkable and the rst evidence to our knowledge of such effects during rapid exposure to complex scenesbeyond demonstrations for neurons in the visual brain (Spratling and Johnson 2004). Finally, the body text gains attention under the brandevaluation goal (0.723 seconds) but less so than under the brand-learning goal (0.967 seconds), and this gain is not accompanied by a simultaneous loss of attention to the pictorial. As a consequence, the overall duration of attention to the ads under the brand-evaluation and brand-learning goals is the same, although their attention patterns for specic ad objects differ substantially.

DISCUSSION
The four processing goals promoted unique patterns of visual attention during self-paced brief ad exposures. Compared to free viewing, (1) informativeness of the ad objects was similar under an ad-appreciation goal, (2) the body text was more informative under a brand-evaluation goal, (3) the body text was more but the pictorial was less informative under a brand-learning goal, and (4) body text, pictorial, and brand were all more informative under an ad-memorization goal. The different patterns of attention for each of the processing goals were striking and reveal how the informativeness of ads and the objects contained therein, rather than being universal, depends on the specic processing goal that is activated. The processing goal effects on ad attention emerged while controlling for the surface size of the ad objects, the effects of which were not found to be goal contingent, which only strengthens the processing goal results.

GOAL CONTROL OF ATTENTION TO ADVERTISING

231

Differences between goals in duration of attention were largest for the body text and smallest for the headline. This reveals how specic the goal contingency of informativeness is, because even though both ad objects are textual, the headline was equally informative for all goals, whereas the body text was least informative under free viewing and most informative under the brand-learning goal. Thus, the brandlearning goal did not indiscriminately promote processing of all textual information but selectively rendered body text more informative. These results are even more telling given that we used familiar brands, which should attenuate text attention. Surprisingly, the duration of attention to the brand was higher under an ad-directed (memorization) goal than under the predicted brand-directed goals. In hindsight, the nding is consistent with the qualitative results of Yarbus (1967), who found that a memorization goal caused his participant to closely examine all scene objects. It is also consistent with results obtained by Keller (1987), who found that consumers recalled fewer ad claims under a brand-directed (evaluation) goal than under an ad-directed (memorization) goal. He speculated that brand-directed processing might promote increased attention to the headline and body text but decreased attention to the pictorials in ads, yet without more detailed processing data, however, these explanations cannot be further tested (Keller 1987, 326). Our study provides these processing data and shows that, counter to his speculation, an ad-memorization goal enhanced attention to the body text, brand, and pictorial, whereas a brand-evaluation goal only enhanced attention to the body text without inhibiting attention to the pictorial. The nding that the duration of attention to the pictorial was highest under ad-memorization goals is noteworthy in view of the reported picture-superiority effect in memory. Explanations of the picture-superiority effect have emphasized the better organization and retrieval of pictorial relative to textual information, but our results show that ad-memorization goals already prioritize pictorials during the encoding stage. In this way, our ndings stress the value of future study to unravel the effect of processing goals on retrieval and storage processes. We did not observe goal effects on pupillary diameter, which presumably reports intensity of attention engagement. Yet, we did nd differences in pupillary diameter for the ad objectsthe pupil being signicantly wider for the pictorial than for the brand and body text. A likely explanation of this result is therefore that goal effects on intensity of attention were absent because consumers managed their attention resources by looking longer instead of harder, as revealed in systematically longer gaze durations. In part, the absence of goal effects on pupillary diameter may be due to the specic set of ads that we had in the study, which may not have been sufciently complex to evoke differences in attention intensity across goal conditions. But our ndings may also raise some doubts about the utility of pupillary diameter as a measure of attention intensity when exposure to print ads and similar scenes is self-paced. Future research

that systematically varies the mental load of consumers, such as by time pressure or the introduction of secondary tasks, might follow up on these speculations. We nd it of interest that the attention patterns for free viewing and the ad-appreciation goal were brief and remarkably similar. This not only reveals the predicted, fast implementation of the ad-appreciation goal but also hints at the possibility that consumers may have adopted an adappreciation goal during free viewingas the default goal. More generally, it may hold true that ads themselves prime processing goals, for example, when objects in a particular ad are prominent means to an end that they then activate. Stimulus-based goal activation has recently been examined in other domains (Shah and Kruglanski 2002) and appears relevant for advertising as well. Based on our results we speculate, for example, that text-dominant ads activate a brand-learning goal and pictorial-dominant ads activate an ad-appreciation goal. Future research may test these speculations on default and bottom-up primed goals to expand the theory of goal control of advertising. In sum, the present ndings support Yarbuss thesis that the informativeness of objects in scenes is goal contingent and that eye movements reect the human thought processes; so the observers thought may be followed to some extent from records of eye movements (1967, 190), even during the brief moments that consumers choose to attend to ads.

APPENDIX MODEL FORMALIZATION AND ESTIMATION

The model is a mixed-outcome multivariate hierarchical Bayes regression model, with a probit part for attention selection (0/1) and a Tobit part for attention engagement (1 0). There are k p 1, , K processing goal conditions, namely, free viewing (k p 1), ad memorization (k p 2), ad appreciation (k p 3), brand learning (k p 4), and brand evaluation (k p 5); l p 1, , L attention indicators, namely, attention selection (l p 1), gaze duration (l p 2), pupillary diameter (l p 3), and a control variable, eye-stimulus distance, l p 4); assessed for m p 1, , M ad objects, namely, the brand (m p 1), pictorial (m p 2), headline (m p 3), body text (m p 4); for i p 1, , I participants and j p 1, , J ads. The ( J # LM ) matrix Yi p ( yi, j,l,m ) contains the attention measures for participant i, the ( J # M ) matrix X p ( xj,m ) contains the log(surface size) of each ad object as deviations from their means, and, nally, the ( I # K ) design matrix Z p ( z i,k ) indicates the K goal condition for every participant, that is, z i,k p 1 for participant i with goal k. The duration and intensity of attention are only observed if an ad object is selected, and otherwise are unobserved.

232

JOURNAL OF CONSUMER RESEARCH

We let yi, j,1,m denote the observed attention selection indicator. Then yi, j,l,m p 0 iff yi , j,1,m 0, p 1, l p 1 iff yi , j,1,m 1 0 p yi , j,l,m; l ( 1 describes that selection of ad objects occurs for positive attention priorities, with probabilities P( yi, j,1,m p 1) p P( yi , j,1,m 1 0). In equation A1, yi, j,1,m reects covert attention priorities leading to attention selection. Duration of covert attention engagement, yi , j,2,m , is reected in the duration of overt gaze, yi, j,2,m, and its intensity, yi , j,3,m, in the pupillary diameter, yi, j,3,m, only if yi, j,1,m p 1. Similarly, the control variable, eye-stimulus distance, yi, j,4,m , is only observed if an ad object is selected for attention. We then specify Yi p Qi + X Si + Ei . (A2) (A1)

in the goal control parameters in L Q and L Q to be tested, that is, pairwise differences between (a) the K goal conditions for each of the L attention indicators and (b) the K goal conditions for each of the M ad objects and L attention indicators separately. We specify a matrix R containing all Q p L (M + 1)

(K 2)

contrasts, WQ p R # vec(L Q ) and WS p R # vec(L S ). Thus, R is a Q # KLM matrix. We compute WQ and WS at every sweep of the MCMC chain, which enables us to test the signicance of all required goal contrasts by computing their posterior means and standard deviations.

REFERENCES
Austin, James T. and Jeffrey B. Vancouver (1996), Goal Constructs in Psychology: Structure, Process and Content, Psychological Bulletin, 120 (3), 33875. Beatty, James and Brennis Lucero-Wagoner (2000), The Pupillary System, in Handbook of Psychophysiology, ed. John Cacioppo, Louis G. Tassinary, and Gary G. Berntson, Cambridge: Cambridge University Press, 14262. Book, Albert C. and C. Dennis Schick (1997), Fundamentals of Copy and Layout, Lincolnwood, IL: NTC/Contemporary Publishing Group. Childers, Terry L. and Michael J. Houston (1984), Conditions for a Picture-Superiority Effect on Consumer Memory, Journal of Consumer Research, 11 (September), 64354. Costley, Carolyn L. and Merrie Brucks (1992), Selective Recall and Information Use in Consumer Preferences, Journal of Consumer Research, 18 (September), 46474. Duchowski, Andrew T. (2003), Eye Tracking Methodology: Theory and Practice, New York: Springer. Dweck, Carol S. and Ellen L. Leggett (1988), A Social-Cognitive Approach to Motivation and Personality, Psychological Review, 95 (2), 25673. Ferguson, Melissa J. and John A. Bargh (2003), The Constructive Nature of Automatic Evaluation, in The Psychology of Evaluation, ed. Jochen Musch and Karl Kristoph Klauer, Mahwah, NJ: Erlbaum, 16988. Findlay, John M. (2005), Covert Attention and Saccadic Eye Movements, in Neurobiology of Attention, ed. Laurent Itti, Geraint Rees, and John Tsotsos, Amsterdam: Elsevier, 11416. Friedman, Alinda (1979), Framing Pictures: The Role of Knowledge in Automatized Encoding and Memory for Gist, Journal of Experimental Psychology: General, 108 (3), 31655. Henderson, John M. and Andrew Hollingworth (1998), Eye Movements during Scene Viewing: An Overview, in Eye Guidance in Reading and Scene Perception, ed. Geoffrey Underwood, Amsterdam: Elsevier, 26993. Hilton, James L. and John M. Darley (1991), The Effects of Interaction Goals on Person Perception, Advances in Experimental Social Psychology, 24, 23567. Hutchinson, J. Wesley, Wagner A. Kamakura, and John G. Lynch Jr. (2000), Unobserved Heterogeneity as an Alternative Explanation for Reversal Effects in Behavioral Research, Journal of Consumer Research, 27 (December), 32444. Itti, Laurent (2005), Models of Bottom-Up Attention and Sa-

Here Qi is a (1 # LM ) and Si a ( M # LM ) matrix of personspecic parameters, the latter containing the own-effects of the surface sizes of each object on the attention measures for it and consisting of L ( M # M ) concatenated diagonal matrices, Si,l p diag(ji,l,m). We assume that Ei N (0, W), where the full ( LM # LM ) covariance matrix W accommodates the dependencies between the L attention indicators for the M ad objects. We allow goals to inuence attention hierarchically: Qi p Z i L Q + Fi, S i p Z i L S + Gi . Here S i p vec(diag(Si,l ), l p 1, , L ) , L Q is a ( K # LM ), and L Q is a ( K # LM ) matrix of parameters, allowing for separatebut correlatedgoal control of attention due to the ad object itself (L Q ) and due to its surface size in the ad (L S). We are primarily interested in L Q , and the posterior means and standard deviations of these parameters for duration of attention (l p 2) are presented in table 2. We assume Fi N (0, FQ ), Gi N (0, FS ), and Cov (Fi , Gi ) p FQ,S, so that the model accounts for individual differences through a normal distribution of coefcients and allows all effects of ad objects on attention to be correlated. We specify standard proper conjugate diffuse prior distributions for all parameters (N(0, 10 4 ) for regression parameters, and IW( P + 1, IP ) for a generic ( P # P ) covariance matrix). All model parameters are estimated simultaneously by standard MCMC procedures, employing a data augmentation step to draw the latent attention variables, Yi, and postprocessing to obtain identiable parameters of the multivariate probit (Rossi et al. 2005). We use 25,000 draws, with a burn-in of 10,000, thinning every third draw, and monitor convergence through plots of the draws of the hyper-parameters. The chains converge well before 10,000 draws. The formulated hypotheses require specic contrasts (A3)

GOAL CONTROL OF ATTENTION TO ADVERTISING lience, in Neurobiology of Attention, ed. Laurent Itti, Geraint Rees, and John K. Tsotsos, Amsterdam: Elsevier, 57682. Janiszewski, Chris (1998), The Inuence of Display Characteristics on Visual Exploratory Search Behavior, Journal of Consumer Research, 25 (December), 290301. Kahneman, Daniel (1973), Attention and Effort, Englewood Cliffs, NJ: Prentice-Hall. Keller, Kevin Lane (1987), Memory Factors in Advertising: The Effect of Advertising Retrieval Cues on Brand Evaluations, Journal of Consumer Research, 14 (December), 31633. Kenney, Keith and Linda M. Scott (2003), A Review of Visual Rhetoric Literature, in Persuasive Imagery, ed. Linda M. Scott and Rajeev Batra, Mahwah, NJ: Erlbaum, 1756. Kingstone, Alan, Daniel Smilek, Jelena Ristic, Chris Kelland Friesen, and John D. Eastwood (2003), Attention Researchers! It Is Time to Look at the Real World, Current Directions in Psychological Science, 12 (5), 17680. Lee, Tai Sing and David Mumford (2003), Hierarchical Bayesian Inference in the Visual Cortex, Journal of the Optical Society of America A, 20 (7), 143448. Lichtenstein, Meryl and Thomas K. Srull (1987), Processing Objectives as a Determinant of the Relationship between Recall and Judgment, Journal of Experimental Social Psychology, 23, 93118. Lohse, Gerald L. (1997), Consumer Eye Movement Patterns on Yellow Page Advertising, Journal of Advertising, 26, 6173. Pashler, Harold, James C. Johnston, and Eric Ruthruff (2001), Attention and Performance, Annual Review of Psychology, 52, 62951. Pieters, Rik, Edward Rosbergen, and Michel Hartog (1996), Visual Attention to Advertising: The Impact of Motivation and Repetition, Advances in Consumer Research, 23, 24248. Pieters, Rik and Michel Wedel (2004), Attention Capture and Transfer in Advertising: Brand, Pictorial and Text-Size Effects, Journal of Marketing, 68 (April), 3650. Rayner, Keith (1998), Eye Movements in Reading and Information Processing: 20 Years of Research, Psychological Bulletin, 124 (3), 372422. Rayner, Keith, Karen M. Rotello, Andrew J. Stewart, Jessica Keir, and Susan A. Duffy (2001), Integrating Text and Pictorial

233 Information: Eye Movements When Looking at Print Advertisements, Journal of Experimental Psychology: Applied, 7 (1), 21926. Rosbergen, Edward, Rik Pieters, and Michel Wedel (1997), Visual Attention to Advertising: A Segment-Level Analysis, Journal of Consumer Research, 24 (December), 30514. Rossi, Peter E., Greg A. Allenby, and Rob McCulloch (2005), Bayesian Statistics and Marketing, New York: Wiley. Russo, J. Edward and France Leclerc (1994), An Eye-Fixation Analysis of Choice Processes for Consumer Nondurables, Journal of Consumer Research, 21 (September), 27490. Shah, James Y. and Arie W. Kruglanski (2002), When Opportunity Knocks: Bottom-Up Priming of Goals by Means and Its Effects on Self-Regulation, Journal of Personality and Social Psychology, 84 (6), 110922. Spratling, Mike W. and Mark H. Johnson (2004), A Feedback Model of Visual Attention, Journal of Cognitive Neuroscience, 16 (2), 21937. Treisman, Anne and Gordon Gelade (1980), A Feature Integration Theory of Attention, Cognitive Psychology, 12, 97136. Viviani, Paulo (1990), Eye Movements in Visual Search: Cognitive, Perceptual and Motor Control Aspects, in Eye Movements and Their Role in Visual and Cognitive Processes, ed. Elisabeth Kowler, Amsterdam: Elsevier, 35393. Wedel, Michel and Rik Pieters (2000), Eye Fixations on Advertisements and Memory for Brands: A Model and Findings, Marketing Science, 19 (4), 297312 Wolfe, Jeremy M. (2005), Guidance of Visual Search by Preattentive Information, in Neurobiology of Attention, ed. Laurent Itti, Geraint Rees, and John K. Tsotsos, Amsterdam: Elsevier, 1014. Wyer, Robert S. and Thomas K. Srull (1989), Memory and Cognition in Its Social Context, Hillsdale, NJ: Erlbaum. Yantis, Steven (2000), Goal-Directed and Stimulus-Driven Determinants of Attention Control, in Control of Cognitive Processes: Attention and Performance XVIII, ed. Stephen Monsell and Jon Driver, Cambridge, MA: MIT Press, 73103. Yarbus, Alfred L. (1967), Eye Movements and Vision, New York: Plenum.