Content-Type: text/html

Ocular Responses to Web Ads -

Running Head: OCULAR RESPONSES TO WEB ADS

Wait!  Why is it not Moving?
Attractive and Distractive Ocular Responses to Web Ads



Nokon Heo
Doctoral Candidate

S. Shyam Sundar
Assistant Professor

Smita Chaturvedi
Doctoral Student

PENN STATE UNIVERSITY


Contact Information:
S. Shyam Sundar
College of Communications
Penn State University
212, Carnegie Building
University Park, PA 16802-5101

Voice:  (814) 865-2173
Fax:  (814) 863-8044
E-Mail:  <a href="/cgi-bin/wa?LOGON=A3%3Dind0109A%26L%3DAEJMC%26E%3D8bit%26P%3D273554%26B%3D--%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D_-1420465%253D%253D_%26T%3Dtext%252Fplain%3B%2520charset%3Diso-8859-1" >[log in to unmask]</a>






Paper submitted to the Advertising Division for consideration of presentation at the annual conference of the Association for Education in Journalism and Mass Communication, Washington, DC, August, 2001

Wait!  Why is it not Moving?
Attractive and Distractive Ocular Responses to Web Ads


A       B       S       T       R       A       C       T


Participants (N = 46) in a 2x2x2 within-participants factorial experiment manipulating animation (animated, static), position (top, bottom), and product-involvement (high, low) of banner ads were exposed to eight online news pages.  Their ocular responses (i.e., horizontal and vertical eye movements) were measured during their browsing.  Results indicate that static ads and top ads tend to distract from news reading, and it takes a high-involvement product to attract visual attention toward animated and bottom ads.


Wait!  Why is it not Moving?
Attractive and Distractive Ocular Responses to Web Ads

Despite the phenomenal growth of the internet as an important marketing tool, researchers and ad professionals alike are reluctant to provide a concrete answer to the question, "Does Web advertising work?" (Sundar, Narayan, Obregon & Uppal, 1998).  One reason is the absence of standardized measures for accurately assessing the effectiveness of the new media vehicle in a manner comparable to the assessment of reach and efficacy of such traditional media as television and print (Dreze & Zufryden, 1998; Kassaye, 1999).  Until now, Web advertisers have relied on a kind of "traffic report" that provides estimates of click-through rates, reach, frequency, and gross rating points (see Dreze & Zufryden, 1998 for more information about these measures).  These quantitative measures, however, can be misleading because unique technological features of the new medium, such as interactivity and navigability, make the reception situation quite unlike the passive attendance to traditional mass me
dia.  Due to the difficulty in identifying visitor characteristics and monitoring visitor traffic and flow patterns, the accuracy of reach and frequency measures in the Web medium is limited (Dreze & Zyfryden, 1998).  It is known, for instance, that the reliance on click-through rates alone can underestimate the value of Web advertising because banner ads do generate strong impressions in users, as evidenced by increases in advertising awareness and brand perceptions, even when they don't click on those ads (Briggs & Hollis, 1997).  Although this kind of traffic report can be useful information about the number of pages requested, it can hardly be translated into estimates of audience's specific moment-to-moment viewing behavior (Dreze & Zufryden, 1998).
The present study is an attempt to reconcile this problem by providing an alternative technique for measuring effectiveness of Web advertisements.  Toward this end, we monitored Internet users' moment-to-moment responses during Web browsing.  More specifically, we tracked users' eye movement while they read news stories on the Web to see if certain features of banner ads could generate cognitive responses, visual attention in particular, among Web users.  The particular banner ad features examined in this study were animation, position, and product involvement.  Animation and position are structural features that are independent of content, while product involvement is a content-related ad feature.
This paper will first describe eye-tracking techniques and review literature pertinent to the effects of animation, position, and product involvement on users' eye movement.  Based on the literature review, a set of hypotheses will be proposed.  It will then describe the methods and results of an experiment conducted to test these hypotheses.  Finally, it will discuss both practical and theoretical implications of the findings.
Eye-Tracking and Visual Attention
         In recent years, there has been an increased interest in physiological assessment of psychological responses to advertising, as a way of learning more about information processing of persuasive messages (Borse & Lang, 2000; Fox, Krugman, Fletcher, & Fischer, 1998; LaBarbera & Tucciarone, 1995; Alwitt, 1985).  Typical physiological responses that have gained considerable research attention include skin conductance response, heart rate, and brain activity during processing of commercial communications.  A lesser-known response, eye movement, has been used by researchers to assess the effectiveness of print ads since it allows for making inferences about cognitive processing of ad materials (Fox et al., 1998; Krugman, et al., 1994).  For instance, Fox et al. (1998) tracked readers' eye movements during exposure to print ads to provide a physiological assessment of visual attention to ad features.
When using eye movements to determine visual attention to ads, researchers are interested in examining two different elements, both of which are useful for making inferences about cognitive processing of persuasive communications.  First, researchers measure the amount of viewing time or "dwell time," i.e., time spent looking at a feature of the ad.  This is based on the assumption that viewers fixate their eyes on ad features as long as they cognitively process the ad.  Thus, duration of dwell time is assumed to be positively associated with amount of cognitive effort allocated to process ad material.  The second element of eye movement that is also a useful indicator of visual attention is the location within an ad that captures a viewer's attention.  In other words, the position of the viewers' eyes during exposure to an ad can provide valuable information about how and which parts of the ad are processed by viewers.  It is a direct way of pinpointing the source of users' infor
mation acquisition from - and other responses to -  commercial communication messages.  The present study monitored this element of eye movements to see if certain features of banner ads direct users' visual attention to the ads.  Examining the position of the users' eyes can be crucial in assessing the effectiveness of Web advertising considering the fact that most portals and Websites of major media organizations feature numerous banner ads on a single page, occupying all parts of the screen and even leading to concerns about crowding and ad clutter on Websites (Napoli, 1999).
        Tracking eye movements is accomplished by monitoring the activities of six muscles around the eyes which are innervated by the sympathetic nervous system (see Stern, Ray, & Davis, 1980 for more information about physiological basis of eye movement).  The working of these muscles controls the movement of both eyes in horizontal, vertical, and circular directions.  In studying eye movements to describe visual information processing of such materials as printed ads or news articles, the most interesting eye movement is saccadic eye movement, which is fast, voluntary movements of eyes from one fixation point to another (Lehtela, 1981).  Typically present in reading, saccades are characterized by a high initial acceleration and final deceleration (Stern, Ray, & Davis, 1980).  Hence, the presence of saccadic eye movement during reading can be interpreted as an active allocation of cognitive resources to the material.  Conversely, the absence of it can be interpreted as a lack of cognit
ive effort devoted to reading.  This interpretation of saccadic eye movement is significant in the context of Web environment where Web users read written materials from a computer screen, and saccades associated with this pattern of reading may be interrupted by the presence of distracting elements on the screen, such as a flickering banner ad.  From an advertiser's perspective, this distraction can be translated into evidence showing the effectiveness of banner ads.  For example, a user reading a news article on a Website can be distracted by an animated banner ad at the top of the site, as evidenced by an increase in vertical eye movement and a corresponding decrease in saccadic horizontal eye movement.  Therefore, recording users' moment-to-moment eye movements during exposure to Web screens can provide vital information about their visual attention to various features and locations on the screens.
        Recording of eye movement and position can be done by various means (see Young and Sheena, 1975 for a comprehensive summary of eye movement recording methods).  The most popular measurement technique electro-oculography (EOG) records potential differences from electrodes placed around the eyes.  This potential is present as the result of the difference in potential between the front (the cornea) and the back (the retina) of the eyeball (Stern, Ray, & Davis, 1980).  This corneo-retinal potential creates an electrical field in front of the head, and this field changes in orientation as the eyeballs rotate.
Eye-Tracking and Web Advertising
         The use of eye-tracking techniques by researchers to assess effectiveness of Web advertising is limited.  A few studies have monitored subjects' eye movements to examine the attention-getting potential of certain content characteristics of print ads (e.g., Fox et al., 1998; Krugman et al., 1994).  These studies, however, measured dwell time or fixation duration as an indicator of visual attention.  Most recently, Heo & Sundar (2000b) used EOG recording of Web users' vertical and horizontal eye movements while browsing the Web, and found them to be significantly correlated with memory of banner ads.
Animation and Eye Movement.  Animation is one of the unique technological features of banner advertising that carries moving images and graphics to simplify or enhance the presentation of persuasive communication (Ellsworth & Ellsworth, 1995).  Given the increasing use of animation by Web advertisers (Cleland & Carmichael, 1997), researchers have endeavored to assess the effectiveness of animation in the context of Web advertising.  Studies have demonstrated that animation generates strong cognitive and emotional responses in users including orienting responses (Borse & Lang, 2000), emotional arousal (Heo & Sundar, 2000a), ad memory (Li & Bukovac, 1999), and attitude change (Sundar, Otto, Pisciotta, & Schlag, 1997).  Only one study, however, tracked users' eye movements and documented attentional bias in response to animation (Heo & Sundar, 2000b).  In this study, researchers found that the presence of animation was not associated with EOG recording of visual attention.  The autho
rs interpreted the result in terms of the unique sample they used for their study (high school students).  They argued that their participants were very serious about the experiment and therefore paid more attention to the news stories rather than the ads on the experimental Web pages shown to them, leaving very little variance to be explained by the animation manipulation.
However, there is a great deal of empirical evidence suggesting that animation attracts visual attention.  For example, Borse and Lang (2000) found that animated banner ads elicit orienting response (OR), as indicated by decrease in users' heart rate.  The orienting response is an automatic allocation of cognitive resources to sudden and significant visual stimuli (Lang, 1990) and it presupposes that the respondents saw the stimuli.  Animation is also associated with mild physiological arousal characterized by increased skin conductance response (Heo and Sundar (2000a).  Indirect evidence showing the attention-getting effects of animation comes from memory tests of animated banner ads.  Researchers have found that Web users remember animated ads better than static ads (Li & Bukovac, 1999) suggesting that animation elicits greater eye-fixations, something that is to be expected when the visual field is visited by an unusual object (Loftus, 1976), in this case an animated ad.
        Several theoretical mechanisms have been proposed by researchers to explain the attention-getting effect of animated ads.  One of them stems from the so-called "distinctiveness theories," rooted in the work of von Restorff (Wallace, 1965).  According to these theories, a visual stimulus that is either semantically or perceptually isolated from other stimuli in the visual field can "pop up" in the early stages of visual information processing and, therefore, increase rehearsal of that item in memory (Jenkins & Postman, 1948).  The superior memory effects of vivid visual stimuli to pallid and dull stimuli can also be interpreted in terms of "distinctiveness theories" (Taylor & Wood, 1983).  Simply stated, these theories explain the effects of animated ads in terms of the "figure-ground" relationship in visual perception.  This relationship suggests that a "distinctive" visual stimuli, be it object or word, becomes the "figure" in the visual field and receives greater attention (and
 therefore, more processing resources) than does the "ground" that it stands against (Beattie & Mitchell, 1985).  According to these theories, then, an animated banner ad can be the "figure" that stands out of the "ground," such as news articles or other static images on a Web page.  In fact, a number of eye tracking studies of newspaper reading show that readers first look at the pictures and graphics for a brief moment before they read the text (Stone & Glock, 1981; Brandt, 1954).  This is because pictures and graphics "stand out," so to speak.  If anything, animated versions of these visual elements on the screen would stand out even more, thereby representing a greater degree of distinctiveness.  Thus, we might predict that animation contributes to greater visual attention.
Another theory that is also relevant to our discussion of the attention-getting effects of animation is the limited-capacity theory of television viewing (e.g., Lang, 1995; Lang et al., 1993).  According to this theory, certain structural features of media stimuli, such as cuts, zooms, edits, and movement elicit ORs in viewers and entail greater involuntary allocation of cognitive resources to encode the message (Lang, 1990).  The end-results of this process typically includes increased SCR (Heo & Sundar, 2000a) and decreased heart rate (Borse & Lang, 2000).  Based on the evidence from previous work on the attention-getting effects of animation and theoretical mechanisms explaining these effects, we propose the following hypotheses regarding the effect of animated ads (embedded in news Websites) upon users' eye movements:
H1a: Animated ads will generate lesser number of horizontal eye movements than static ads.
H1b: Animated ads will generate greater number of vertical eye movements than static ads.
The first hypothesis is based on our reasoning that users' horizontal eye movements indicate the degree to which users pay attention to the news content, and that animation will have a distractive effect on users' news reading performance, which will be manifested by inhibited saccadic eye movements.  In other words, any reduction in the volume of users' horizontal eye movements during exposure to news articles can provide indirect evidence of animation effects on users' visual attention to the banner ads.  Greater vertical eye movements, on the other hand, mean lesser attention to the news content and greater attention to the ads, particularly when the ads are positioned either at the top or bottom of the visual field.
Position and Eye Movement.  The psychological effects of ad position have been documented in the context of print (Lohse, 1997) as well as Web advertising (Heo & Sundar, 2000a; 2000b).  Both anecdotal and empirical evidence in print media suggest that readers pay greater attention to ads at the top than those at the bottom (Lohse, 1997).  This is understandable when one considers the typical information-acquisition pattern we have developed throughout the history of print culture.  The so-called "spatial information acquisition theory" posits that people usually acquire information serially, beginning in the upper left corner of the text and proceeding from left to right and from top to bottom.  And newspapers and other print media reinforce this pattern by presenting the most important information at the beginning of the page.  In fact, research on the psychology of ad position has shown that readers pay more attention to ads near the beginning of the heading than to ads near the end of the heading (Lohse, 1997; Rhodes, Teferman, Cook, & Schwartz, 1979).  We expect a similar result in the context of Web advertising because of the similarities in presentation technique and message selection between such traditional print media as newspapers and magazines and online news Websites.  The important news stories still occupy the top location, while less important ones are pushed below.  On
line users will more than likely orient themselves toward the information on an English-language Website in a manner similar to their orientation while reading English-language newspapers-serially, from top to bottom.  Therefore, positioning a banner ad at the top of a Web page will draw greater visual attention from users.  In other words:
H2a: Ads positioned near the top of the page will generate lesser number of horizontal eye movements than those near the bottom of the page.
H2b: Ads positioned near the top of the page will generate greater number of vertical eye movements than those near the bottom of the page.
That is to say, top ads occupy a prime position on the screen, thus attracting visual attention and resulting in reduced reading of news content, as evidenced by decreased number of horizontal eye movements and increased vertical movement.
Product Involvement and Eye Movement.  Unlike animation and position of an ad that are independent of the nature of content, product involvement is a key content-related ad feature.  Typically defined by the degree of perceived personal relevancy and importance evoked by a product ad (Zaichkowsky, 1985), product involvement is known to influence various aspects of cognitive processing of persuasive communication messages.  In general, researchers have found that the degree of perceived involvement is positively associated with individuals' cognitive engagement in the ad (Hupfer & Gardner, 1971) as well as the likelihood of engaging in extended problem-solving and decision-making processes during exposure to advertising messages (Petty, Cacioppo, & Schumann, 1983).
Although direct evidence showing the connection between perceived level of involvement and eye movement is rare, some indirect evidence suggests that the two might be linked.  For instance, in the context of Web advertising, Briggs and Hollis (1997) found that high-involving products generated a greater click-through rate than low-involving products.  They also found that high-involving products were remembered better than low-involving products suggesting that visual attention mediated the memory effects.
The OR literature provides a more direct link between involvement level of advertised products and degree of visual attention paid to ads.  The orienting theory suggests that certain media stimuli that are perceived personally relevant and significant elicit involuntary attention responses in viewers (Lang, Newhagen, & Reeves, 1996; Maltzman, 1979).  The physiological assessment of this kind of short-term attention with older visual media such as television strongly suggests that Web users will pay greater amount of visual attention to high involving products than low involving products on a given Website.  As long as users perceive certain products as personally relevant and important, then the Web ads featuring those products should elicit short-term visual attention.  Therefore:
H3a: Ads featuring high involving products will generate lesser number of horizontal eye movements than ads featuring low involving products.
H3b: Ads featuring high involving products will generate greater number of vertical eye movements than ads fearing low involving products.
        That is, when a Web user sees a banner ad for a product with which he/she has high involvement, then the user is likely to interrupt his/her reading to look at the ad, thereby resulting in reduced horizontal saccades and increased vertical eye movement.
Method
To examine the effects of three independent variables-ad animation, ad position, and product involvement-on eye movement, a 2 x 2 x 2 within-participants factorial experiment was designed.  The first independent variable, animation, had two values: animated or static.  The second independent variable, ad position, had two values: top and bottom.  The third independent variable, product involvement, also had two values: high and low.  To measure the primary dependent variable, visual attention, we tracked participants' eye movements via online EOG recordings while they are reading news articles on Websites.
Design Overview
All participants (N = 46) were asked to read 8 different online news stories on a computer screen.  Embedded within each news page was a banner ad positioned either at the top or bottom of the page, either animated or static, and featured either a high-involvement or low-involvement product/service.  To control for order effects, we created four separate presentation orders.  Within each order, all eight news stories were randomized as well to avoid possible confounding effects of news content.  After reading all 8 news stories, participants filled out a paper-and-pencil questionnaire designed to gather basic information about computer use and demographics.
Participants
Forty-six male and female undergraduate students enrolled in communication classes were recruited in exchange for extra course credit.  Participants were told that the study concerned online presentation of media content and signed a consent form before the experiment.  Each participant was randomly assigned to one of four presentation orders.
Stimulus Material
For the experiment, we created eight separate news Web pages.  To increase external validity, we used actual news reports that appeared in major online newspaper sites.  In selecting stimulus news stories, we excluded any news story that was judged to be controversial, time-bound, and emotion-laden in order to minimize content effects.  In addition to news stories, we also obtained a pool of banner ads from existing commercial Websites.  These banner ads featured different product categories, and each product was evaluated by a separate group of individuals for product involvement.  Based on the literature on product involvement, we came up with an inventory of question items that were intended to tap into the perceived level of personal relevancy and importance of a product.  A pretest was conducted applying this inventory to a vast pool of banner ads.  This selection process distinguished four high and four low involving products from the rest in the pool.
The animated banner ads had either text or objects that either moved or were flashing during and after the loading, whereas the static banner ads had the same objects or text but they appeared motionless on the screen.  The static version of the animated ads was created by disassembling the animation into still images and choosing the one that had the greatest amount of information about the product.  That is, the exact same banner ads were shown to participants in either static or animated form.  (Of course, the presentation of these ads to participants was counterbalanced such that a given participant saw either the animated or static version of a given ad, not both; for every participant who saw the animated version of Ad A and the static version of Ad B, there was another participant who was exposed to the static version of Ad A and the animated version of Ad B, and so on).
Experimental Online News Websites
        To mimic actual news Websites, all eight stimulus news Web pages included the masthead for The Chicago Tribune.  The top banner ads were positioned right above the masthead, while the bottom banners were near the bottom of the page following the news stories.  Each news story was modified to fit the screen so that participants were not required to scroll down the page to read the whole story.  Each story appeared on the screen for 25 seconds before the computer automatically moved onto the next page.  In the pretest, we found that 25 seconds was enough time for participants to read the article.
Measurement of Eye Movements
The main dependent variable, visual attention, was operationalized by vertical and horizontal eye movements.  Eye movements were measured by online recording of electrooculography (EOG) while participants were reading news stories.  To record EOG, we placed six AG/AGCL electrodes around the participants' eyes.  Horizontal eye movements were recorded by placing an electrode at the outer edge of each eye.  Vertical eye movements were recorded with electrodes placed right above and below the eye.  The remaining two electrodes served as grounds and were placed around the participants' neck.
Once the electrodes were placed, horizontal and vertical eye movements were recorded via separate channels on a recording computer that was connected to a DC amplifier with lead cables. The DC recording of eye movements generated a series of peaks and crests on both channels.  To precisely determine where the eyes were looking at a given time during reading, it was necessary to perform calibrations of eye position after subjects read the last news story.  For calibrations, participants were instructed to look straight ahead (i.e., center of the screen), up, down, left, and right for five seconds at a time, with each of these points being marked on the EOG record.  Using these marked points, we identified four different types of eye movements: vertical peaks, vertical crests, horizontal peaks, and horizontal crests.  Vertical peaks were any positive deflections above the mid-point between the up and straight-ahead positions, while vertical crests were any negative deflections below
 the mid-point between the straight-ahead and down positions.  Similarly, we identified both horizontal peaks and horizontal crests by determining the mid-point between straight-ahead and right and between straight-ahead and left positions respectively.  Deflections to the right of the midpoint between straight-ahead and right positions were considered horizontal peaks, while deflections to the left of the midpoint between straight-ahead and left were considered horizontal crests.  For the analysis, we combined both vertical peaks and vertical crests additively to compute the total number of vertical eye movements.  Greater vertical movement would indicate greater attention to the ads.  Both horizontal peaks and horizontal crests were also combined into one index to be used as the indicator of horizontal eye movement.  For this index, greater scores would mean greater attention to news content.  In other words, lesser the number of horizontal eye movements, lesser the attention to
 news stories and, by implication, greater the attention to banner ads.
Product Involvement
To evaluate each banner ad for product involvement, we modified Zaichkowsky's (1985) Involvement Inventory and obtained 12 questions, all using a 7-point semantic differential scale.  The items were unimportant/important, irrelevant/relevant, useless/useful, beneficial/not beneficial, matters to me/don't matter, uninterested/interested, boring/interesting, unexciting/exciting, appealing/unappealing, essential/nonessential, wanted/unwanted, and not needed/ needed.
Procedure
        The experiment was administered to participants individually in a laboratory.  Upon arrival, the participant was greeted by the experimenter and told that the experiment in which they were about to participate concerned browsing several online news pages.  The participant filled out a written consent form before the experiment began.  The experimenter then led the participant to a room, which housed the Web browsing computer and EOG recording devices.  After the participant was seated in front of the computer, the experimenter placed electrodes on the participant's face.  Then the participant was given a verbal instruction regarding experimental procedure.  After giving the instruction, the experimenter went into another room to check if the recording computer was producing appropriate readings.  The experimenter then returned to the experimental room and instructed the participant to turn on the computer monitor and follow a series of written instructions on the screen.  For all participants, the first instruction read:
Welcome to the Study!
The next few screens will display snapshots of news pages from the Websites of The Chicago Tribune.  Please read each screen as you would normally read a web page. Each screen will stay on the computer monitor for ONLY A FEW SECONDS. Therefore, PLEASE DO NOT USE YOUR MOUSE unless instructed.
DO NOT CLICK ON THE LINKS. DO NOT SCROLL DOWN.
Click HERE to Start and take your hands off the mouse.

After reading the instruction, the participant was instructed to click on the hyperlinked word "HERE" to begin the experiment.  While the participant was reading the instruction, the experimenter returned to the other room to start the EOG recording.  Once the participant clicked on the hyperlinked word to begin the experiment, another written instruction was displayed on the screen.  For all participants, the second instruction read:
For the next few seconds, we will be checking to see if the equipment functions properly. Please look at the center of the screen and remain absolutely still. The experiment will begin once this screen changes to a new screen. DO NOT USE THE MOUSE OR THE KEYBOARD during the entire procedure.

As soon as the second screen was displayed, the experimenter, who could monitor the subject's browsing activity on a remote screen in the observation room, stated the EOG recording.  After the second written instruction, there were forty seconds of baseline period during which the participant was instructed to remain calm and look at the center of the screen.  After the baseline period, the participant was automatically linked to the first news web page for 25 seconds.  After 25 seconds, the participant was led via the HTML "refresh content" command to the second page.  After another 25 seconds, the participant was led to the next page, and so on, till they had gone through all eight pages.  After the last page, the participant was led to the calibration pages.  This time, the participant was told that he or she would see a series of paired blinking eyes on the computer screen.  The participant was then told that their task was to follow the blinking eyes without moving their head
s.  For all participants, the calibration instruction read:
One the next screen, you will see a pair of blinking eyes.  Please stare steadily at the eyes without moving your head.  Follow the eyes as they move from one part of the screen to another.

The calibration consisted of five Web pages, each one with a pair of blinking eyes.  The first page had a pair of blinking eyes at the center of the screen for five seconds.  After five seconds, the computer automatically linked the participant to the second calibration page for another five seconds.  This time the blinking eyes appeared at the left-hand side of the screen, midway between top and bottom.  The blinking eyes then moved to the top of the screen, midway along the width of the screen.  Then, it moved to the right-hand side.  This calibration process ended when the participant viewed the blinking eyes appearing at the bottom center of the screen.  After this task, the participant was debriefed and thanked for the participation.  During the entire course of the experiment, the participant was not allowed to use the mouse or keyboard.  And all copies of Graphic Image Format (GIF) files and HTML pages were stored in the Web browser's cache to ensure immediate appearance of
 both images and text.
Data Analysis
To test all hypotheses, a three-way within-participants factorial analysis of variance (ANOVA) was conducted to detect main effects of the three independent variables as well as interaction effects, on each of the two eye-movement indices.


Results
Manipulation Check
        In preparation for performing a manipulation check of the product involvement variable, the twelve product involvement items were subjected to a factor analysis.  After rotation, the analysis identified two factors: "Utility" and "Interest."  Two product involvement indices were created by adding the scores for items loading under each factor.  The results of two t-tests with the two indices as the dependent variables showed a statistically significant difference in participants' perceptions of the products, t (44) = 5.31, p < .001 for Utility and t (44) = 12.56, p < .001 for Interest.  As expected, participants rated high involving products significantly higher in involvement scales than low involving products.
Hypothesis Tests
        H1a predicted that animated ads would generate lesser number of horizontal eye movements than static ads indicating that animation has attention-getting potential.  A three-way ANOVA testing on horizontal eye movements showed no significant main effect for animation.  Therefore, the result did not support the hypothesis in that both animated (M = 13.7) and static ads (M = 13.2) generated about the same number of horizontal eye movements.  H1b predicted that animated ads would generate greater number of vertical eye movements than static ads.  The ANOVA test with vertical eye movement as the dependent variable failed to show a significant main effect for animation, F(1, 41) = 1.90, p > .05.  (Due to an equipment failure, data for four participants' vertical eye movements could not be obtained.  The mean number of vertical eye movements for animated ads and static ads were 20 and 22, respectively.
        H2a predicted that ads near the top of the page would generate lesser number of horizontal eye movements than ads positioned near the bottom of the page.  The ANOVA test showed a significant main effect for position, F(1, 45) = 13.37, p < .001.  Consistent with the prediction, top ads (M = 12.9) generated lesser number of horizontal eye movement than bottom ads (M = 14) indicating that participants were paying less attention to the news content when there was an ad at the top (as opposed to the bottom) of the page.  H2b predicted that ads positioned near the top of the page would generate greater number of vertical eye movements than ads positioned at the bottom.  The ANOVA test with vertical eye-movement showed a significant effect for position, F(1, 41) = 7.31, p < .05, but in the opposite direction.  The average number of vertical eye movements for top ads was 19 whereas that for bottom ads was 24.  This is not what we expected in that participants paid greater amount of atten
tion to the ads when they appeared at the bottom of the page than at the top.
        H3a predicted that high involving product ads would generate lesser number of horizontal eye movements than low involving ads.  As expected, the ads that were perceived as high involving generated lesser number of horizontal eye movements (M = 13.1) than the ads perceived as low involving (M = 13.9), F(1,45) = 5.51, p < .05.  Therefore, the hypothesis H3a was supported.  H3b predicted a greater number of vertical eye movements as a function of product involvement.  The ANOVA test with vertical eye-movement failed to support the hypothesis, F(1, 41) = .70, p = .40.  The difference between high (M = 22) and low (M = 21) involving products was statistically non-significant.
Exploratory Findings
        In addition to the observed main effects, we also found several significant interaction effects between independent variables.  These interactions perhaps underlie the lack of significance on some of the hypothesized main effects.
        Although not hypothesized, the ANOVA test discovered a significant interaction effects between animation and involvement on horizontal eye movements, F (1, 45) = 8.33, p < .05, such that for low involving ads, animation did not matter (see Figure 1), but for high involving products, animation made a significant difference with static ads (M = 12.3) generating lesser number of horizontal eye movements than animated ads (M = 13.9).  In general, static ads show a differentiation on horizontal eye-movement based on involvement level of the product advertised whereas animated ads do not.

Figure 1.  Mean Horizon Eye Movements As a Function Of Animation and Product Involvement




        Another significant interaction effect was detected between product involvement and position, F (1, 45) = 4.87, p < .05.  In this interaction effect, the location of an ad was not a significant factor for high involving products.  However, location made a significant difference for the ads that were perceived as low involving (see Figure 2).  More specifically, for high involving products, positioning an ad either at the top or bottom generated the same number of horizontal eye movements.  However, for low involving products, putting an ad at top generated lesser number of horizontal eye movements.  In other words, when readers saw a low involving ad at the top of the page, they paid more attention to the ad and lesser attention to the news story.

Figure 2. Mean Horizontal Eye Movements As a Function of Product Involvement and Position



        With regard to vertical eye movements, we found a near significant interaction effect between product involvement and animation, F (1, 41) = 2.94, p = .09, shown in Figure 3.

Figure 3. Mean Vertical Eye Movements As a Function of Product Involvement and Animation


Animation did not make a statistically significant difference for high involving products.  However, for low involving products, static ads generated greater number of eye movements (see Figure 3).   This finding is consistent with horizontal eye movement data which showed that low involving static ads generated the least number of horizontal eye movements.  Stated another way, this interaction shows that static ads did not differ along involvement level of product whereas animated ads did.
Discussion
        Today's Web users are bombarded with an array of visually stimulating banners.  To make banners visually appealing to the users, Web advertisers use animation and embed the banners in different locations.  The present experiment tested the potential effects of these two features in relation to product involvement.  Toward this end, we monitored participants' eye movements using EOG.  The EOG recordings of both horizontal and vertical eye movements were used to operationalize visual attention.  Among the six proposed hypotheses, two were supported.  This section provides both theoretical and practical implications for these findings.
The results indicate that top position and high involving products capture users' incidental attention, as evidenced by lesser number of horizontal eye movements.  In other words, reading was inhibited in the presence of an ad at the top of the page.  This effect will be more prominent when the banner features a high involving product.  The interaction effect observed in this study between involvement and position further supports this claim.  These findings suggest that although position and product involvement alone can be psychologically meaningful to users, combining them will result in a kind of "double dose" effect.  The data also showed that putting a low involving banner ad near the bottom was least effective.  The position main effect can be interpreted in terms of "spatial pattern of information acquisition theory" discussed previously (Lohse, 1997).  According to this theory, Web users will start looking at the top of the page to search information because the informati
on on the Web moves from the top to bottom as in a newspaper.  This similarity between newspaper and the Web will predispose the users to orient toward the top of the screen.
An alternative explanation for position effects can be provided by "novelty effects."  Even if today's Web advertisers utilize any available space on a Web page, it is unusual to see a banner ad at the bottom of the page.  This tendency not to have an ad near the bottom might have kept participants from looking at the ad.  The finding that the low involving product at the bottom location generated greatest number of horizontal eye movements can also be explained by novelty effects.  In fact, the same argument can be made with regard to the null effects of animation on both horizontal and vertical eye movements.  Today's Web advertising is dominated by the use of animation.  It is hard to encounter static banners any more.  So, observing an animated banner ad would likely not trigger users' visual attention.  In fact, in this study, we found that static ads were more likely to distract from news reading than animated ads even if the difference was not statistically significant.  This finding illustrates the novelty effects of static banners.
One last explanation for null effects of animation can be explained in terms of "floor effects" or "clown's pants' effects."  The idea is that given the nature of the experiment, participants might have completely overlooked the ads.  This explanation is plausible given the research finding by the Poynter Institute that online news readers tended to read text before they looked at photos or other graphics (Walker, 2000).  Or, some of the dazzling images of banner ads were so distracting with no highlighted information that the users might have consciously ignored the ads (Lynch & Horton, 2000).
The superior attention-getting effect of high involving products can be explained in terms of OR.  This theory suggests that banner ads featuring personally relevant and important products capture users' immediate attention or sudden look.  Typically involuntary in nature, this response is characterized by a sudden decrease in heart rate and a mild increase in skin conductance response.  The fact that product involvement level was associated with a systematic pattern in our EOG recordings ratifies the validity of using EOG as an indicator of OR.  The relationship between product involvement and horizontal eye movements can be explained by the "limited capacity information processing model" (Lang, 1990).  This theory assumes that the human beings' cognitive resources or processing capacity is limited.  So, the allocation of these resources during information processing can determine what and how much information is acquired.  According to this theory, the sudden advent of a visuall
y important stimulus like high involving banner ad elicits OR, and this OR automatically allocates the resources to process the new information.  Since the total available resource is limited, there remains lesser amount of resources that can be allocated to other cognitive activities like reading a news story on a Website.  In this experiment, we assumed that this effect would be manifested by a lesser degree of attention given to reading news stories.  The fact that we found a lesser number of horizontal eye movements in response to high involving products lends support to this theory.
In addition to the two main effects of position and product involvement, we found some interesting interaction effects.  Overall, the interaction findings suggest that high-involving bottom ads and high-involving static ads distract from reading more than their low-involving counterparts.  In terms of looking response, static ads do better than animated, counter to intuition (it's as if our participants are getting a pause when a particular stimulus ad does not move, and wondering, "Wait a minute! Why isn't this ad moving?"), but if the animated ad features a high-involvement product, then it generates as much of a look response as the static ad.  Therefore, product involvement appears to be a critical variable in determining the attractive and distractive functions of animation as well as position of banner ads.
Our study demonstrates powerful novelty effects in that not only static ads attract greater attention than animated ones, but also bottom ads do better at getting visual attention than top ads.  However, other research has shown that animated ads generate higher skin-conductance responses, an indication of users' orientation toward the ad.  Given that SCR goes up but vertical movement doesn't, it appears to be a conscious, learned reaction on the part of Web users.  They notice the animated ads but have learned not to physically move their eyes toward them.  This explanation is buttressed by the fact that our participants showed reduced horizontal movement only when the ads were static and high-involvement.  Clearly, more research is needed on the attractive and distractive features of banner ads in order for us to gain a more complete understanding of users' responses and reactions, so that we may be able to devise accurate measures of ad effectiveness in the new medium.
References
Alwitt, L. F.  (1985).  EEG activity reflects the content of commercials. In L. F. Alwitt & A. A. Mitchell (Eds.), Psychological processes and advertising effects:  Theory, research, and application (pp. 129-155).  Hillsdale, NJ:  Lawrence Erlbaum, Associates.
Beattie, A. E., & Mitchell, A. A. (1985).  The relationship between advertising recall and persuasion: An experimental investigation.  In L. F. Alwitt & A. A. Mitchell (Eds), Psychological processes and advertising effects: Theory, research, and application (129-155).  Hillsdale, NJ: Lawrence Erlbaum, Associates.
Brandt, H. (1954).  The Psychology of seeing.  New York : The Philosophical Library.
Briggs, R., & Hollis, N. (1997).  Advertising on the Web:  Is there response before click-through?  Journal of Advertising Research, 37(2), 33-45.
Cleland, K., & Carmichael, M. (1997).  Banners that move make a big impression.  Advertising Age, 26-28.
Dreze, X., & Zufryden, F.  (1998).  Is Internet advertising ready for prime time?  Journal of Advertising Research, 38, 7-18.
Ellsworth, J., & Ellsworth, M. (1995).  Marketing on the internet: Multimedia strategies for the World Wide Web.  New York: John Wiley & Sons, Inc.
Fox, R. J., Krugman, D. M., Fletcher, J. E., & Fischer, P. M. (1998).  Adolescents' attention to beer and cigarette print ads and associated product warnings.  Journal of Advertising, 27(3), 58-68.
Heo, N., & Sundar, S. S.  (2000a).  Emotional responses to Web advertising; The effects of animation, position, and product involvement on physiological arousal. Paper presented at the annual convention of the Association for Education in Journalism and Mass Communication, Phoenix, AZ.
Heo, N., & Sundar, S. S.  (2000b).  Visual orientation and memory for Web advertising: A study of animation and position effects.  Paper presented at the annual convention of the International Communication Association, Acapulco, Mexico.
Hupfer, N. T., & Gardner, D. M.  (1971).  Differential involvement with products and issues: An exploratory study.  In D. M. Gardner (Ed.), Proceedings, Association for consumer research.  College Park, MD: Association for Consumer Research.
Jenkins, W. O., & Postman, L.  (1948).  Isolation and spread effect in serial learning.  American Journal of Psychology, 61, 214-221.
Kassaye, W. W.  (1999).  Sorting out the practical concerns in World Wide Web advertising.  International Journal of Advertising, 18, 339-361.
LaBarbera, P. A., & Tucciarone, J. D.  (1995).  GSR reconsidered: A behavior-based approach to evaluating and improving the sales potency of advertising.  Journal of Advertising Research, 35, 33-44.
Lang, A.  (1990).  Involuntary attention and physiological arousal evoked by structural features and emotional content in TV commercials.  Communication Research, 17(3), 275-299.
Lang, A. (1995).  Defining audio/video redundancy from a limited capacity information processing perspective.  Communication Research, 22, 86-115.
Lang, A., Newhagen, J., & Reeves, B. (1996).  Negative video as structure: Emotion, attention, capacity, and memory.  Journal of Broadcasting and Electronic Media, 40(3), 460-477.
Lehtela, J.  (1981).  Differences in eye movement data recorded by electro-oculography and corneal reflection techniques.  Human Factors, 23(6), 661-665.
Li, H., & Bukovac, J. L. (1999).  Cognitive impact of banner ad characteristics: An experimental study.  Journalism and Mass Communication Quarterly, 76(2), 341-353.
Lohse, G. L. (1997).  Consumer eye movement patterns on Yellow Pages advertising.  Journal of Advertising, 26(1), 61-73.
Lynch, J. & Horton, S.  (2000).  Yale Centre for Advanced Media WWW Style Manual. [On-line].  Available:  http://info.med.yale.edu/caim/manual/pages/editorial_style.html
Maltzman, I.  (1979).  Orienting reflexes and significance and bilateral SCR to potentially phobic pictures.  Journal of Abnormal Psychology, 93, 41-46.
Napoli, L. (1999, June 17).  Banner ads are under the gun and on the move.  CyberTimes [On-line].  Available: http://www.nytimes.com/library/tech/99/06/cyber/articles/17advertising.html.
Petty, R. E., Cacioppo, J. T., &  Schumann, D.  (1983).  Central and peripheral routes to advertising effectiveness: The moderating role of involvement.  Journal of Consumer Research, 10, 135-146.
Rhodes, E. W., Teferman, N. B., Cook, E., & Schwartz, D. (1979).  Yellow Pages advertising: An empirical analysis of attributes contributing to consumer interest, liking, and preference.  Journal of Professional Services Marketing, 6(2), 35-44.
Stern, R. M., Ray, W. J., & Davis, C. M. (1980).  Psychophysiological Recording.  New York: Oxford University Press.
Stone, D. & Glock, M.D. (1981).  How do young adults read directions with and without pictures? Journal of Educational Psychology, 73 (3), 419-426.
Sundar, S. S., Narayan, S., Obregon, R., & Uppal, C.  (1998).  Does Web advertising work?  Memory for print vs. online media.  Journalism & Mass Communication Quarterly, 75 (4), 822-835.
Sundar, S. S., Otto, G., Pisciotta, L., & Schlag, K. (1997).  Animation and priming effects in online advertising.  Paper presented to the Advertising Division at the annual conference of the Association for Education in Journalism and Mass Communication, Chicago, IL.
Taylor, S. E., & Wood, J. V.  (1983).  The vividness effect:  Making a mountain out of molehill?  In R. Bagozzi & A. Tybout (Eds.), Advances in consumer research (vol. 5), Ann Arbor: Association for Consumer Research.
Walker,D. (2000).  Words on the web more powerful than pictures.  [On-line].  Available:  http://it.mycareer.com.au/cgi-bin/http
Wallace, W. P.  (1965).  A review of the historical, empirical, and theoretical status of the von Restorff phenomenon.  Psychological Bulletin, 63, 410-425.
Young, L. R., & Sheena, D.  (1975).  Methods & designes: Survey of eye movement recording methods.  Behavior Research methods & Instrumentation, 7(5), 397-429.
Zaichkowsky, J. L.  (1985).  Measuring the involvement construct.  Journal of Consumer Research, 12(3), 341-352.