The Impact of Salient Advertisements on Reading and Attention
on Web Pages
Jaana Simola
University of Helsinki and Aalto University School of
Economics
Jarmo Kuisma, Anssi Öörni, and Liisa Uusitalo
Aalto University School of Economics
Jukka Hyöna¨
University of Turku
Human vision is sensitive to salient features such as motion. Therefore, animation and onset of
advertisements on Websites may attract visual attention and disrupt reading. We conducted three eye
tracking experiments with authentic Web pages to assess whether (a) ads are efficiently ignored, (b) ads
attract overt visual attention and disrupt reading, or (c) ads are covertly attended with distraction showing
up indirectly in the reading performance. The Web pages contained an ad above a central text and another
ad to the right of the text. In Experiments 1, 2, and 3A the task was to read for comprehension.
Experiment 1 examined whether the degree of animation affects attention toward the ads. The results
showed that ads were overtly attended during reading and that the dwell times on ads were the longest
when the ad above was static and the other ad was animated. In Experiments 2 and 3, the ads appeared
abruptly after a random time interval. The results showed that attention (i.e., the time when the eyes first
entered an ad) was related to the ad onset time. This happened especially for the ad to the right, indicating
that ads appearing close to the text region capture overt attention. In Experiment 3B the participants
browsed the Web pages according to their own interest. The study demonstrated that salient ads attract
overt visual attention and disrupt reading, but during free browsing, ads were viewed more frequently and
for longer time than during reading.
Keywords: reading, eye movements, online advertising, abrupt onset, task-orientation
Previous research on the processing of online advertisements
has mainly focused on how frequently and for how long ads are
viewed. A common finding in these studies has been that users
rarely look at ads directly. Researchers have thus suggested that
Web users learn strategies that help them ignore ads and maintain
their focus on the primary task (e.g., Stenfors, More´n, &
Balkenius, 2003). These findings have led some authors to
question the applicability of visual attention theories to the Web
environment, and to suggest that Web designers should not
worry about the ads because they are too small to interfere with
the users’ primary task (Diaper & Waelend, 2000). A possible
reason for why users have been observed to ignore online
advertising might be that the studies have used small ad graphics,
simplistic Web page layouts (e.g., word arrays or lists on a
white background) and the tasks have been somewhat artificial
(e.g., reaction time or visual cueing tasks) when compared to
everyday online tasks (Burke, Hornof, Nilsen, & Gorman, 2005;
Day, Shyi, & Wang, 2006; Stenfors et al., 2003; Zhang, 2000).
The present study used standard-sized online ads and a Web
page layout adopted from an authentic Web portal to investigate
the allocation of attention toward advertisements during online
reading. Another aim was to correct earlier misconceptions
about the applicability of attention theories to the Web environment
by demonstrating that the theories do generalize to the
visual environment of the Internet. We also discuss practical
implications of the present results for Web page design.
Previous eye tracking studies on advertising have mainly examined
the visual processing of print advertisements (Radach, Lemmer,
Vorstius, Heller, & Radach, 2003; Rayner, Miller, & Rotello,
2008; Rayner, Rotello, Stewart, Keir, & Duffy, 2001; Wedel,
Pieters, & Liechty, 2008). It is important, however, to investigate
the processing of online ads because, in contrast to print media, the
Web environment allows the use of dynamic ad formats, such as
animated and abruptly appearing ads. According to prior research
such design cues capture visual attention efficiently, and thus
processing of online ads is likely to differ from the processing of
print advertisements. Moreover, the ubiquity of the online environment
warrants research into the effects of online advertising on
users’ attention.
This article was published Online First May 23, 2011.
Jaana Simola, Cognitive Science and CBRU, Institute of Behavioral
Sciences, University of Helsinki, Helsinki, Finland, and Department of
Marketing and Management, Aalto University School of Economics, Finland;
Jarmo Kuisma, Anssi Öörni, and Liisa Uusitalo, Department of
Marketing and Management, Aalto University School of Economics; Jukka
Hyöna¨, Department of Psychology, University of Turku, Turku, Finland.
This study was supported by grants from Helsingin Sanomat Foundation.
The authors thank Jari Lipsanen for his advice in statistics, and TeliaSonera
for providing the Web platform and advertisements. Portions of Experiments
1 and 3A were presented in the Conference on Neuroeconomics,
Copenhagen, Denmark, May, 2008.
Correspondence concerning this article should be addressed to Jaana
Simola, Institute of Behavioral Sciences, P.O. Box 9, FI-00014 University
of Helsinki, Finland. E-mail: jaana.simola@helsinki.fi
Journal of Experimental Psychology: Applied © 2011 American Psychological Association
2011, Vol. 17, No. 2, 174–190 1076-898X/11/$12.00 DOI: 10.1037/a0024042
174
Attention to Online Ads
The amount of information in our visual environment far exceeds
what our visual system is able to process and assimilate into
a conscious experience. The function of visual attention is to
enhance the processing of selected stimuli. Models of visual attention
often include the concept of a saliency map that integrates
information across different features (e.g., intensity, orientation,
color, etc.) into a single map encoding the visual saliency of
objects (Itti & Koch, 2000). The maximum of the saliency map
corresponds to the most salient location, which is thought to attract
visual attention. On the other hand, inhibiting the most salient
location and directing attention away from it requires voluntary
effort.
Salient changes in the visual field may capture attention involuntarily
in a stimulus-driven, bottom-up manner. Previous research
has shown that during parallel search for a singleton target, participants
cannot override the stimulus driven activation by salient
stimulus attributes even when they are told to do so (Theeuwes,
1994). Voluntary attention, also termed as top-down or goaldirected
attention, refers to attentional allocation based on individuals’
current goals. Top-down attention allows users to actively
maintain attention on the primary task or to shift attention voluntarily
from one display region to another. Previous studies have
demonstrated that the ability to selectively ignore distractors depends
on the presence of an attentional set for target and distractor
properties, suggesting that top-down control of attention is not
possible without knowledge of the target and distractor features
(Theeuwes & Burger, 1998). Moreover, Theuuwes and Burger
observed that when the salient element was unpredictable and
changed from trial to trial, it was not ignored and therefore
interfered with visual search.
Previous studies suggest that Web users mainly rely on topdown
strategies that help them ignore ads present in their visual
field (Dre`ze & Hussherr, 2003; Stenfors et al., 2003). These
studies propose that in the Web environment top-down control of
attention is capable of overriding the bottom-up attentional capture
by salient low-level visual features (e.g., color, orientation, luminance,
or motion). For example, Burke et al. (2005) reported that
ads containing only uniform gray color were looked at equally
often as colored commercial ads including graphics, and that there
were no differences between animated and static commercial ads.
In addition to bottom-up and top-down routes, the effects of
visual attention can be divided into overt and covert influences.
Covert attention allows us to attend to a specific location without
moving our eyes to that location. That is, covert processing of
advertisements occurs when an individual is exposed to ads in the
peripheral visual field while overt visual attention is allocated
around the foveal region. Previous research on covert processing
of online ads has provided ambiguous findings. Some studies have
demonstrated that ads interfere with the primary task. For example,
Burke et al. (2005) reported longer search times in the presence of
commercial banners compared to uniform gray banners, when the
task was to search for exact matches of given headlines. Likewise,
Hong, Tong, and Tam (2004) reported increased response times in
an online shopping task when a nontarget shopping item was
flashed. In contrast, Burke et al. (2005) demonstrated that when the
task was to locate single words from lists of words, peripheral ads
including both static and animated graphics decreased visual
search times compared to when uniform gray banners were presented.
Day et al. (2006) also found that online decision-making
became faster when peripheral flash banners were presented compared
to when no flash banners were presented. The authors
concluded that peripheral ads might also increase the participants’
level of arousal, which in turn motivates them to allocate additional
processing resources. All these studies, along with that of
Dre´ze and Hussherr (2003), report that participants rarely look at
the ads overtly, suggesting that covert attention to the ads is
responsible for the effects observed in the search and decision
times.
Experienced Distraction
Although some studies have reported that Web users are able to
avoid ads by top-down attentional control, other studies have
demonstrated that the presence of ads results in increased workload,
leading the users to experience the ads as intrusive and
distracting. For example, Burke et al. (2005) asked participants to
rate their impression of workload for different types of ads: uniform
gray, static, animated, or flashing text ad. The flashing text
elicited greatest subjective ratings of workload and was evaluated
as the most frustrating and demanding. Animated commercial ads
were rated slightly higher on workload than static ads. In the study
by Diaper and Waelend (2000), participants’ ratings suggested
more attention being paid to animated than static graphics. Moreover,
Yoo and Kim (2005) reported that subjective ratings of
attentional demand were the highest when the speed of animation
was fast compared to static ads or ads containing a moderate speed
of animation. In the experiment by Zhang (2000), 88% of the
participants reported that they would rather have no animated
graphics in the same page where they performed a string search
task. Animated color and string graphics were rated as the most
distracting types of animation. Gao, Koufaris, and Ducoffe (2004)
showed that the presence of continuous animation and pop-up ads
resulted in higher perceived irritation that was associated with
negative attitudes toward the Website. In addition, Hong et al.
(2004) demonstrated that users had unfavorable attitudes toward a
Web page when a nontarget item was flashed. Thus, previous
research suggests a discrepancy between the Web users’ performance
and their subjective experience, that is, the users seem to be
aware of the ads and they perceive them as distracting, yet they
rarely look at the ads directly.
Overview of the Experiments
The present experiments investigated how online ads attract
visual attention and affect reading of Web pages. By instructing
participants to read the texts to answer multiple choice text content
questions, attention was primarily directed toward the text, while
the ads were considered secondary stimuli to the reading task.
Based on earlier research, we expected that the ads might affect
reading in three possible ways: (a) ads are ignored and thus do not
affect reading, (b) ads attract visual attention and are overtly
fixated, causing disruption to the ongoing reading process, or (c)
covert attention to ads disrupts reading with the influence showing
up only indirectly in the reading performance. These three hypotheses
were tested in the three experiments.
175EFFECT OF ADS ON ONLINE READING
As regards the first hypothesis, various studies have demonstrated
that specific task demands may guide attention only to
those parts of the display that are relevant for the task at hand
(Theeuwes & Burger, 1998). Studies on online ad processing
(Diaper & Waelend, 2000; Stenfors et al., 2003) further suggest
that ads can be avoided by top-down attentional control. Thus, the
first hypothesis is that ads are not attentively processed during
online reading, but instead attention is directed to the text, which
is relevant for the given task.
The second hypothesis proposes that standard-sized ads (especially
the animated ones) are salient enough to direct attention in
a bottom-up manner, and thus result in overt fixations to ads and
interruptions to reading. Previous studies, which suggest that ads
do not capture attention, have used rather simple Web page layouts,
small ad sizes (in the range of 100 ϫ 100 pixels), and the
graphic contents may not have been comparable to authentic
online ads (Diaper & Waelend, 2000; Zhang, 2000). Moreover,
studies on visual attention indicate that abrupt changes (e.g., stimulus
onsets, offsets, and motion) are capable of capturing attention
involuntarily (Franconeri & Simons, 2003; Jonides & Yantis,
1988). Further support for the second hypothesis comes from
subjective reports on experienced distraction, suggesting that users
are aware of the ads, and they experience them as distractors to the
main task (Gao et al., 2004; Hong et al., 2004).
The third hypothesis is that users have learned to ignore ads, but
covert attention to ads results in an indirect processing cost, that is,
a drop in reading performance. As suggested by Itti and Koch
(2000), directing attention away from the most salient stimulus
requires extra effort. This strategy suggests that users are aware of
the processing cost associated with directing overt attention elsewhere,
and therefore attempt to avoid attentional capture by peripheral
distractors. This also implies that the user knows the
perceptual properties of the objects being ignored (see Theeuwes
& Burger, 1998). According to the central capacity theory of
attention (Kahneman, 1973), if users spend some of their attentional
resources on an interfering task, that is, covertly attending to
ads, they will be left with fewer resources for the primary task,
reading in our case. Moreover, because both the ads and text are
presented in the visual modality, they compete for the same specific
attentional resources (see Allport, 1980).
To test the three aforementioned hypotheses about allocation of
attention to online ads during reading, we recorded participants’
eye movements and analyzed the results relative to the text and ad
areas. Eye movements provide an excellent indicator of the locus
of overt visual attention, because attention and targeting of saccades
are tightly linked together (Deubel & Schneider, 1996;
Dore´-Mazars, Pouget, & Beauvillain, 2004). Well-established eye
movement metrics such as fixation durations and number of fixations
(Rayner, 1998) were used as indices of the online reading
process. Reading rates and responses to the text content questions
were collected to assess reading performance. As argued above,
avoidance of ads is observed if the participants do not fixate the
ads and the reading performance stays constant under different
online ad conditions. On the other hand, direct processing cost is
observed if the ads attract overt attention (i.e., participants fixate
them directly), resulting in interruptions in and slowing down of
the ongoing reading process. Finally, an indirect processing cost is
observed if the reading performance drops while the ads are not
overtly fixated.
The selected Web page layout comprised a central text and two
online ads located either above or to the right of the text. The ad
above the text was aligned horizontally and the ad to the right was
vertically aligned. The ad locations and formats corresponded to a
standard Web page layout. In the present study, ad location rather
than ad format was more important for making predictions about
allocation of attention to different parts of a Web page. Previous
studies (McConkie & Rayner, 1975; Rayner, 1998) indicate that
for readers of Western languages, the perceptual span is biased
toward the right of fixation, that is, they acquire more information
(about 15 letters) to the right of fixation than to the left (only about
3–4 letters). Therefore, we expected that because of its proximal
location, the ad to the right of the text would attract more attention
than the ad above the text. Prior research has shown location
effects on attention toward online ads. For example, ads positioned
in the lower visual field attract more attention than ads in the upper
visual field (Goodrich, 2010). Other studies have emphasized the
effect of format on ad perception and memory (Kuisma, Simola,
Uusitalo, & Öörni, 2010) as well as on attitudes toward online
advertising (Burns & Lutz, 2006).
The salient visual feature of the ads in Experiment 1 was motion
(animation), whereas Experiment 2 tested the effect of abrupt ad
onsets on reading. Finally, to determine the degree to which a
given task affects ad perception, Experiment 3 assessed the effect
of task-orientation on the processing of online ads by instructing
the participants to browse the pages according to their own inter-
ests.
Experiment 1
Advertisers incorporate fast moving images in online ads to
capture attention and to break through the information clutter on
Websites. Previous studies suggest that the human visual system
allows priority to behaviorally urgent events such as dynamic
discontinuities (e.g., abrupt onsets and movement) in the environment
(Franconeri & Simons, 2003; Theeuwes, 1994). In addition,
research has shown that attention can be directed efficiently to
moving items and that their presence is detected even if attention
is allocated elsewhere (McLeod, Driver, & Crisp, 1988; McLeod,
Driver, Dienes, & Crisp, 1991). Experiment 1 tested whether the
degree of peripheral motion inherent in animated ads affects attentional
allocation and online reading. The stimulus displays
included one ad above the text and another ad to the right of the
text (see Figure 1). Our expectation was that a high degree of
animation (both ads simultaneously animated) attracts attention
and distracts reading more than when only one of the ads is
animated. The smallest effects were expected when both ads are
static.
Method
Participants. Twenty-eight volunteers with normal or corrected
to normal vision participated in the experiment. Table 1
shows demographic information about the participants and their
previous experience with the Internet. All participants provided
written informed consent and they were paid to participate in the
experiment.
Apparatus. Eye movements were recorded by a Tobii 1750
remote eye-tracking system with a spatial accuracy of .5°. The
176 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
screen coordinates of both eyes were sampled at 50 Hz. For each
participant, the system was calibrated before the experiment using
a set of 16 calibration points shown one at a time and covering the
whole screen area. The stimuli were presented on a 17-inch display
with a screen resolution of 1024 ϫ 768 pixels. The display was
located on a table at the distance of ϳ60 cm from the participants’
eyes.
Materials. The stimuli comprised 32 Web pages each containing
a text, an ad above the text, and another ad to the right of
the text (see Figure 1). The text material contained 32 online
magazine articles of 90–118 words, with mean length of 104
words, and with an average word length of 6.8 characters. The
texts were modified from the text materials previously used in
Laarni, Simola, Kojo, and Na¨sa¨nen (2004) and were written in
Finnish in a 14-point Arial font. From the viewing distance of 60
cm, the average character height was .4° and the average width
was .3°. The texts were about music, religion, advertising, art,
business, and golf.
The advertisements consisted of 64 full-color ads depicting 16
different topics (7 cellular phone and 3 broadband subscriptions, 2
mobile games, a lottery, a magazine, a holiday, and a charity
campaign). Previous studies investigating the effect of animation
have been criticized for lacking an adequate control between the
static and the animated conditions, and for not providing comparable
content and amount of information, or for confounding the
effect of animation with other factors, such as interactivity (reviewed
in Tversky, Morrison, & Betrancourt, 2002). To control for
the effects of ad content, four different versions of each ad topic
Figure 1. A black-and-white version of the Web pages used in the study (the actual Web pages were colored)
with an example text and a four-choice question translated into English from the original Finnish version. For
illustration purposes, a larger font size for the questions is used here than in the actual experiments.
177EFFECT OF ADS ON ONLINE READING
were professionally designed so that the same topic was presented
above the text in static and animated versions, and also to the right
of the text in static and in animated versions. Thus, each participant
saw the same advertisements in four different versions during
the experiment, but different versions of the same topic never
appeared simultaneously on a Web page. The static ads consisted
of representative frames from the corresponding animated ads, and
the animations included horizontal, vertical and/or rotating movement,
appearing and fading graphics, or text objects.
The Web page layout was adopted from the former front page of
a commercial portal, and it included two ads meeting the international
standards. The ad above the text subtended 14.4 ϫ 2.0°
(468 ϫ 60 pixels), and the ad to the right of the text subtended
4.4 ϫ 11.6° (140 ϫ 350 pixels). Both ads were simultaneously
visible, and were presented under the following four conditions:
both ads were static (S ϩ S), the ad above was static and the ad to
the right was animated (S ϩ A), the ad above was animated and the
ad to the right was static (A ϩ S), and both ads were animated
(A ϩ A). A Latin square design was used so that each participant
read an equal number of texts in each condition (eight texts per
condition), while across participants each text was read in each
condition by an equal number of participants. In addition, the
assignment of different texts and ad combinations to different
conditions was randomized to prevent possible interactions between
the text and ad contents; however, the text and ad contents
were not related. The conditions were presented in a randomized
order, and the presentation was controlled by a Java-software
developed for the experiment. The Web pages were displayed
using Internet Explorer Web browser. The text and ads fitted in
one screen so no scrolling was necessary.
Procedure. Participants were told that the experiment concerned
online reading, and nothing was said about the advertisements.
They were instructed to read 32 texts to be able to
answer a multiple-choice content question presented after each
text (see Figure 1). Participants read the texts at their own pace
while their eye movements were recorded. After they finished
reading a text, they clicked on a link below, and a four-choice
question appeared. Answering the question prompted the next
stimulus to appear. After participants had finished reading all
the texts, they were asked to rate, on a scale from 1 to 5,
whether they had paid attention to the ads and whether the ads
had interfered with reading. The experiment lasted about 30
min.
Data analyses. Fixations, saccades and blinks were extracted
from the raw eye coordinate data with Tobii Clearview software
using a window-based algorithm with 40 pixel window and a
minimum fixation duration of 100 ms. To measure reading performance
as a function of the overall scanning on the Web page,
reading rates were calculated as the number of words covered in
the total time spent on the page (words per minute, wpm). Comprehension
accuracy was measured as the proportion of correct
answers in the multiple-choice questions. To test whether the ads
interfered with reading either overtly (fixations to the ads), or
covertly (no fixations detected on the ads, but a drop in reading
performance), we calculated eye movement measures separately
for the ad and the text regions (see Figure 1). For the ad regions
two frequency measures (number of entries and total number of
eye fixations) and two time measures (time of first entry and total
dwell time ϭ summed duration of all fixations) were calculated.
For the text region, in addition to the frequency measures and the
time of text entry, we calculated the mean fixation duration and the
number of saccades directed backward in the current text line (regressions),
because these measures provide a good indicator of reading
difficulty. Previous research has shown that text difficulty strongly
influences the number of regressions, and that regressions most likely
indicate comprehension failures, because when readers determine that
their interpretation of a sentence is wrong or they have failed to
adequately comprehend it, they often regress back to the relevant text
region to reprocess parts of the text (as reviewed in Rayner, 1998).
Moreover, longer fixation durations and more frequent fixations during
reading are associated with processing difficulty, for example,
longer eye fixations have been observed for misspelled, less common
or unpredictable words (Rayner, 1998).
Differences in reading performance and eye movement measures
concerning the text area were studied with a repeated measures
analysis of variance (ANOVA). The effect of Ad Type
(animated vs. static) was studied for both ad Locations (above vs.
right), because in the experimental design both ads were always
presented simultaneously. Furthermore, we studied the interactions
between Ad Type and Location. Around 40% of the participants
reported that they had paid attention to ads constantly or relatively
often, and 68% reported that animated ads interfered with reading
Table 1
Background Information About the Participants
Exp. 1 Exp. 2 Exp. 3A Exp. 3B
N 28 30 32 30
N of females 14 15 16 15
Mean age (SD) 35.1 (7.3) 34.1 (10.1) 37.6 (9.8) 29.4 (10.0)
Age range 19–49 20–58 18–53 19–54
Education
High school/vocational school 10 12 26 7
University/college 17 18 6 23
Computer and Internet experience in years (SD)
Computer experience 16.4 (4.4) 16.3 (4.2) 14.8 (6.8) 14.0 (4.0)
Internet experience 9.71 (3.0) 11.4 (2.3) 9.61 (2.6) 10.5 (2.5)
Internet use
Daily 24 30 28 30
Weekly 4 — 3 —
178 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
constantly or at least occasionally. We tested the relation of participants’
self-reports to eye movement measures by adding participants’
self-reported attention and distraction as binary betweenparticipants
factors to the ANOVA model. Furthermore, the
relation between comprehension accuracy and eye movements was
tested by adding the mean comprehension scores to the model.
The eye movement measures for the ads were subjected to a
Generalized Estimating Equations (GEE) model. The GEE-model
also tested the effects of participants’ self-reported attention to ads
and distraction by ads as well as comprehension accuracy on eye
movement measures concerning the ads. A negative binomial
distribution using log link was used for frequency measures and a
gamma regression using the inverse link was used for time measures.
A negative binomial distribution is typically used for analyzing
count measures, and a gamma distribution is often used to
model eye movement data (e.g., Wedel et al., 2008). The structure
of the within-participant covariance matrix was set to independent,
and least significant difference (LSD) tests were used for the post
hoc comparisons.
GEE (e.g., Hardin & Hilbe, 2003) is an extension to the standard
array of Generalized Linear Models (GLMs). As likelihood-based
models, traditional GLMs are based on the assumption that individual
subjects or observations are independent. GEE models do
not require such an assumption, and can thus handle situations
where responses are correlated. GEE models can also handle data
with missing values (e.g., in our data roughly 70–80% of the trials
did not contain eye movements to ads) more efficiently than the
traditional repeated measures ANOVA, which is based on the
assumption that all missing values can be omitted (Diggle, Heagerty,
Liang, & Zeger, 2002). The design of the present study is a
traditional repeated measures design where the distributions of
outcome variables are non-normal (e.g., variables such as the
number of fixations and fixation duration on ads are strongly
skewed to the right) and the structure of the missing value pattern
can be considered missing at random (Little & Rubin, 1987).
Therefore, the GEE model is well suited for this kind of design and
was used in the present study.
Results
Reading performance, eye movements, and subjective reports.
Animation of the ad to the right of the text did not affect
reading, whereas animation of the ad above the text resulted in
numerical differences in the mean number of regressions during
reading and the text entry times. That is, the mean number of
regressions was higher (19.37 Ϯ 8.68 SD) when the ad above was
animated than when it was static (18.77 Ϯ 8.78 SD), although this
difference did not quite reach significance (F(1, 23) ϭ 3.60, p ϭ
.070, ␩p
2
ϭ .135). Similarly, the text was entered somewhat later
when the ad above was animated (.65 Ϯ .69 SD) compared to
when it was static (.60 Ϯ .62 SD), but this effect did not reach
significance either (F(1, 23) ϭ 3.71, p ϭ .067, ␩p
2
ϭ .139). Table
2 shows the means of reading performance and eye movement
measures on the text region.
Participants who reported that they had paid attention to ads
entered the text later (.99 Ϯ 1.67 SD) than those who reported no
attention to ads (.39 Ϯ .42 SD) (F(1, 23) ϭ 4.98, p ϭ .036, ␩p
2
ϭ
.18). Reported distraction interacted with animation of the ad
above the text in number of fixations (F(1, 23) ϭ 6.28, p ϭ .020,
␩p
2
ϭ .214) made during reading, suggesting that those who did not
report distraction made fewer fixations (p ϭ .012) when the ad
above the text was static. Furthermore, reported ad distraction
(F(1, 23) ϭ 6.16, p ϭ .021, ␩p
2
ϭ .211) and comprehension
accuracy (F(1, 23) ϭ 4.15, p ϭ .053, ␩p
2
ϭ .153) affected reading
rates, indicating that reading was slow when the reported distraction
and comprehension accuracy were high.
Eye movements on the ad above the text. On average,
participants fixated the ad above the text in 21% of the trials.
Figure 2 illustrates the mean number of entries and fixations to ads
per trial. GEE model showed that the number of entries to the ad
above the text was higher when the ad to the right was static (.45 Ϯ
1.11 SD) compared to when it was animated (.25 Ϯ .86 SD)
(␹2
(1) ϭ 6.58, p ϭ .010) (see Table 3). Furthermore, the analysis
showed an Ad Type ϫ Location interaction for the dwell time on
the ad above the text (␹2
(1) ϭ 11.53, p ϭ .001), suggesting longer
dwell times when the ad above was static and the ad to the right of
the text was animated (S ϩ A) compared to all other conditions,
that is, when both ads were static (S ϩ S) (p ϭ .013), or animated
(A ϩ A) (p ϭ .012), or when the ad above was animated and the ad
to the right of the text was static (A ϩ S) (p ϭ .011) (see Table 3).
The total dwell time on the ad above the text was longer for
participants who reported attention to ads (.51 Ϯ 1.33 SD) compared
to those who did not report attention to ads (.42 Ϯ 3.46 SD)
(␹2
(1) ϭ 5.29, p ϭ .021). Participants’ self-reported distraction
was associated with the entry time on the ad above the text
(␹2
(1) ϭ 4.27, p ϭ .039), suggesting later ad entries for those who
reported distraction (16.81 Ϯ 20.00 SD) compared to those who
did not report distraction (10.38 Ϯ 15.74 SD).
Eye movements on the ad to the right of the text. Participants
fixated the ad to the right of the text in about 27% of the
trials. The ad to the right was entered and fixated more often than
the ad above the text (see Figure 2). The number of entries
(␹2
(1) ϭ 3.90, p ϭ .048) and the number of fixations (␹2
(1) ϭ
9.92, p ϭ .002) for the ad to the right of the text were higher when
it was animated compared to when it was static, but both effects
were qualified by an Ad Type ϫ Location interaction (see Table
3). These interactions suggest that the number of fixations on the
ad to the right was the highest in the S ϩ A condition, compared
with all other conditions, that is, the S ϩ S condition (p ϭ .005),
Table 2
Means of Reading Performance and Eye Movement Measures
on the Text in Experiment 1, as a Function of Ad Condition
Ad condition S ϩ S S ϩ A A ϩ S A ϩ A
Reading performance
Comprehension .88 .89 .85 .87
Reading rate (wpm) 216.91 220.13 197.41 194.40
Eye movements on the text
No. of entries 1.93 2.09 2.27 1.92
No. of fixations 93.81 98.17 100.60 95.56
No. of regressions 18.20 19.33 20.17 18.65
Time of entry (s) .58 .62 .70 .61
MeanFixDur (ms) 208.19 214.38 211.84 210.33
Note. Time of entry is calculated in seconds from the trial onset. S ϩ S ϭ
both ads were static; S ϩ A ϭ the ad above was static and the ad to the
right was animated; A ϩ S ϭ the ad above was animated and the ad to the
right was static; A ϩ A ϭ both ads were animated.
179EFFECT OF ADS ON ONLINE READING
the A ϩ A condition (p ϭ .035), and the A ϩ S condition (p ϭ
.011).1
An Ad Type ϫ Location interaction was also observed for
the dwell time on the ad to the right of the text (␹2
(1) ϭ 6.55, p ϭ
.010), indicating the longest dwell times in the S ϩ A condition
compared with all other conditions, that is, the S ϩ S condition
(p ϭ .025), the A ϩ A condition (p ϭ .034), and the A ϩ S
condition (p ϭ .017). Participants who reported attention to ads
(␹2
(1) ϭ 8.50, p ϭ .004) and distraction by ads (␹2
(1) ϭ 9.81, p ϭ
.002) fixated the ad to the right of the text more often (2.03 Ϯ 3.72
SD and 1.50 Ϯ 3.26 SD) than participants who did not report
attention or distraction (.50 Ϯ 1.75 SD and .26 Ϯ .92 SD).2
Discussion
The results of Experiment 1 showed a clear pattern of increased
attention to both ads in the condition where the ad above the text
was static and the ad to the right of the text was animated (the S ϩ
A condition). Our expectation was that the highest degree of
animation (both ads animated simultaneously) would attract the
most attention. However, contrary to our expectations, the results
suggest that a top-down control of attention to ignore the ads failed
most often when there was moderate amount of animation. This
finding was supported by increased dwell times on both ads as well
as by the greater number of entries and fixations toward the ad to
the right of the text in the S ϩ A condition. The results support the
hypothesis that the ads distract reading and that distraction by ads
occurs through overt fixations toward ads rather than as covert
processing of ads during reading.
The results also show that the ad to the right was attended more
often than the ad above the text. Moreover, when the amount of
animation was moderate (the S ϩ A condition), animation increased
attention toward the ad to the right, whereas no such effect
was found for the ad above (the A ϩ S condition). In contrast, the
ad above the text was attended more often when the ad to the right
was static, suggesting that when the ad to the right was animated,
less attention was directed to other parts of the Webpage. Thus, the
results indicate that animation presented in the proximity of the
text attracts more attention compared to when the animated ad is
presented in the periphery. This interpretation is based on the fact
that when reading the text from left to right, the eyes come very
close to the ad on the right each time the reader reaches the end of
a text line (see Figure 1). Therefore, it is likely that the ad on the
right, at least occasionally, enters the region of effective vision
(i.e., the perceptual span, see Rayner, 1998), resulting in occasional
fixations on the ad.
To further examine whether attention to ads occurred in a covert
or overt fashion, we included participants’ self-reported attention
and experienced distraction by ads to the analyses. The results
showed later text entries for participants who reported attention to
ads, and slower reading rates for those who reported distraction.
Furthermore, participants who reported being distracted by the ads
had longer dwell time and fewer fixations on the text when the ad
above was static. From the ad processing perspective the results
show that participants who reported paying attention to ads looked
at the ad above longer and entered the ad to the right more often
than participants who did not report attention to ads. Moreover,
participants who were distracted by the ads entered the ad above
the text later than participants who did not report distraction.
However, attention toward the ad on the right increased for those
participants who reported that they attended to the ads and were
distracted by them. This pattern of results possibly suggests that
the ad above the text was attended covertly whereas the ad on the
right was attended overtly.
The findings of Experiment 1 suggest that especially the animated
ads to the right of the text were overtly attended (i.e.,
fixated) during reading, even when they were irrelevant to the task
at hand. Participants’ self-reports further supported the finding that
ads attracted overt attention and also distracted reading. Thus, in
1
A similar pattern was observed for the number of entries toward the ad
to the right, suggesting a higher number of entries in the S ϩ A condition
compared to when both ads were static (S ϩ S) (p ϭ .014) or animated
(A ϩ A) (p ϭ .045).
2
The number of entries revealed a similar, statistically significant pat-
tern.
Figure 2. Number of entries and number of fixations to the ads in
Experiment 1. Brackets indicate significant differences in Wilcoxon
signed-ranks test. Error bars indicate SE.
Table 3
Eye Movement Measures for the Ad Above and the Ad to the
Right of the Text in Experiment 1, as a Function of Ad
Condition
Ad condition S ϩ S S ϩ A A ϩ S A ϩ A
Ad above text
No. of entries .40 .28 .50 .23
No. of fixations .86 .80 .96 .51
Time of entry (s) 15.35 16.98 15.72 13.10
Dwell time (ms) 924.39 1560.72 878.63 753.55
Ad to the right of text
No. of entries .48 .77 .64 .59
No. of fixations .74 1.55 1.02 1.03
Time of entry (s) 13.69 15.39 13.26 15.90
Dwell time (ms) 796.04 1462.82 895.71 872.80
Note. Time of entry is calculated as seconds from the trial onset. S ϩ S ϭ
both ads were static; S ϩ A ϭ the ad above was static and the ad to the
right was animated; A ϩ S ϭ the ad above was animated and the ad to the
right was static; A ϩ A ϭ both ads were animated.
180 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
contrast to previous studies (Dre`ze & Hussherr, 2003; Stenfors et
al., 2003), our results show that ads are not always efficiently
avoided, but they may interrupt the reading process by capturing
participants’ overt attention to ads. Thus, contrary to the hypothesis
that online ads can be efficiently ignored, the results of
Experiment 1 suggest that ads do attract attention and are overtly
fixated. The results also go against those of Diaper and Waelend
(2000) who concluded that Web page designers should not be
concerned about the effects of graphics on Web pages, because
typical ads are too small to interfere with the main task. On the
contrary, our results suggest that animation should be applied with
care, because it can capture attention and interfere with the online
reading process.
It is noteworthy, however, that participants fixated the ads rather
late, that is, after around 15 s from the trial onset, and that the ads
were fixated in less than one third of the trials. Previous studies
have shown that even though visually salient features can be
efficiently detected, it does not necessarily indicate that all salient
features capture attention. For example, objects exhibiting motion
do not necessarily draw attention when motion is uninformative
about the target location (Hillstrom & Yantis, 1994; Yantis &
Egeth, 1999). To increase the salience of ads, we delayed their
appearance in Experiment 2, as prior research suggests that taskirrelevant
abrupt stimulus onsets efficiently capture visual attention
(e.g., Theeuwes, Kramer, Hahn, Irwin, & Zelinsky, 1999;
Yantis & Jonides, 1984).
Experiment 2
Experiment 2 tested the effect of delayed ad onset time on
attention toward ads and online reading performance. Previous
research has shown that people are able to ignore some taskirrelevant
salient features, such as motion, color or brightness, but
that an abrupt appearance of a new object efficiently captures
attention in a bottom-up fashion (e.g., Hillstrom & Yantis, 1994).
These studies have demonstrated faster response times to abruptonset
targets than to targets that are present from the beginning of
the search trials (Jonides & Yantis, 1988; Yantis & Hillstrom,
1994; Yantis & Jonides, 1984), and that the response times to
abrupt-onset targets do not increase with the number of elements in
the array, whereas substantial array size effects are observed for
no-onset targets. Attentional capture by stimulus onset could not
be attributed to deviant or unique stimulus features such as luminance
increment (Yantis & Hillstrom, 1994), color (Jonides &
Yantis, 1988), or motion (Hillstrom & Yantis, 1994), indicating
that abrupt visual onsets per se are sufficient to capture attention.
On the basis of these studies, it is concluded that processing of
abrupt onsets differs fundamentally from the way other distinctive
features are processed, in that stimulus onset captures attention and
is examined first, whereas processing of salient items defined by
brightness or color does not differ from processing of other items
in the array.
Furthermore, Brockmole and Henderson (2005) showed that
abruptly appearing new objects capture attention also in realworld
scenes, especially when the new object appears during a
fixation and introduces motion in the visual field. Gaze was also
found to be directed to a new object appearing during a saccade,
when input of visual information is switched off, suggesting
that transient onset is not necessarily required for the new
object in a scene to influence gaze. Thus, new objects could also
be prioritized through top-down attentional control, but without
transient onsets observers need to rely on their memory representation
of the scene to guide attention and to localize the
changes.
To our knowledge, no earlier studies have examined the
effects of abrupt ad onsets on online reading and attention. The
ad appearance was delayed so that either the ad above the text
or the ad to the right of the text appeared on the screen after a
random delay of 0–12 s. The abrupt onset conditions were
compared to conditions, where either the ad above or the ad to
the right of the text was present throughout the trial. This
approach made it also possible to separate out the independent
effect of ad location on reading, which was not the case in
Experiment 1 where both ads were simultaneously visible
throughout the trial.
Method
Participants. Thirty volunteers gave an informed consent and
were paid to participate in Experiment 2 (see Table 1). All had
normal or corrected-to-normal vision. They were not aware of the
purpose of the study, and none of them took part in Experiment 1.
Procedure. The same texts, advertisements and Web page
layout were used as in Experiment 1, and the task was again to read
the texts for comprehension. The experimental procedure was
similar to Experiment 1, except for the ad conditions. To test the
effect of abrupt ad onsets on attention toward the ads, the following
conditions were used: (a) the ad above the text was present
throughout the trial, (b) the ad above the text appeared after a
random delay of 0–12 s, (c) the ad to the right of the text was
present throughout the trial, (d) the ad to the right appeared after
a random delay of 0–12 s.
Data analyses. To ensure compatibility of results with those
of Experiment 1, similar procedures for data preprocessing and
statistical analyses were used. Reading performance and eye
movement measures concerning the text area were subjected to a
2 ϫ 2 repeated measures ANOVA with Ad Location (above vs.
right) and Abrupt Onset (no-onset vs. onset) as within-participant
factors. Eye movements landing on both ad regions were analyzed
using the GEE model. Paired comparisons were performed for
fixations on the ad above the text between conditions where the ad
was present throughout the trial or when it appeared abruptly
(Conditions a and b). Similarly, fixations landing on the ad to the
right were compared between the two onset conditions (Conditions
c and d).
Around 53% of the participants reported that they had paid
attention to ads constantly or relatively often, and around 53% of
them reported that ads interfered with reading constantly or at least
occasionally. They formed the groups who reported attention to
ads and distraction by the ads. The relation of self-reports to eye
movements was tested by adding participants’ self reported attention
and distraction as between-participants factors to the statistical
models. Furthermore, the relationship between comprehension accuracy
and eye movements was tested by including participants’
answers to text content questions in both models.
To examine the effect of abrupt ad onset on attention toward the
ads, the ad onset time was included in the GEE model, and
correlations were calculated between the abrupt ad onset time and
181EFFECT OF ADS ON ONLINE READING
the time of first fixation landing on the ad. Spearman’s r was used
because the distributions were skewed and nonlinear. Moreover,
we calculated the number of fixations intervening between the ad
onset and the observers’ first fixation on the ad as a measure of
how quickly the onset captured attention. We first report how the
ad conditions affected reading and attention toward the ads (i.e.,
fixations on the text and ad regions) and then how abrupt onsets
captured attention.
Results
Reading performance, eye movements, and subjective reports.
Ad location affected comprehension accuracy (F(1,
27) ϭ 4.22, p ϭ .050, ␩p
2
ϭ .135), reflecting poorer comprehension
when the ad was presented to the right than above the text (see
Table 4). The main effect was qualified by an Ad Location ϫ
Abrupt Onset interaction (F(1, 27) ϭ 4.40, p ϭ .045, ␩p
2
ϭ .140),
suggesting that comprehension was somewhat poorer in the onsetright
than in the onset-above condition (p ϭ .026). However,
comprehension was better when the ad above the text appeared
abruptly compared to when it was present throughout the trial (p ϭ
.033). Ad location affected also reading rate (F(1, 26) ϭ 6.48, p ϭ
.017, ␩p
2
ϭ .199), suggesting that reading was slower when the ad
above the text was presented compared to when the ad to the right
was presented. Slower reading rates were further associated with
improved comprehension accuracy (F(1, 26) ϭ 5.68, p ϭ .025,
␩p
2
ϭ .179), but this effect interacted with ad location (F(1, 26) ϭ
5.93, p ϭ .022, ␩p
2
ϭ .186), suggesting that the effect was stronger
in the conditions where the ad to the right was presented. Improved
comprehension was also associated with greater number of fixations
(F(1, 26) ϭ 9.02, p ϭ .006, ␩p
2
ϭ .258) and regressions (F(1,
26) ϭ 6.29, p ϭ .019, ␩p
2
ϭ .195) on the text region.
Participants who reported paying attention to ads entered the
text more often (2.90 Ϯ 2.50 SD) (F(1, 26) ϭ 5.52, p ϭ .027,
␩p
2
ϭ .175) and they had longer mean fixation duration (286.86 Ϯ
49.07 SD) during reading (F(1, 26) ϭ 11.73, p ϭ .002, ␩p
2
ϭ .311)
compared to participants who reported no attention to ads (1.38 Ϯ
.36 SD and 235.41 Ϯ 36.50 SD).
Eye movements on the ad above the text. On average,
participants fixated the ad above the text on 14% of the trials.
Figure 3 shows the mean number of entries and fixations to the
ads. According to the GEE model, abrupt onset affected the
number of entries (␹2
(1) ϭ 12.09, p ϭ .001) and number of
fixations (␹2
(1) ϭ 10.23, p ϭ .001), suggesting higher number of
entries and fixations on the ad above the text when it appeared
abruptly compared to when it was present throughout the trial (see
Table 5).
Participants who reported attention to ads entered (␹2
(1) ϭ
11.26, p ϭ .001) and fixated (␹2
(1) ϭ 21.97, p Ͻ .001) the ad
above the text more often than participants who did not report
attention to ads (see Table 6). They also entered the ad above
sooner (␹2
(1) ϭ 13.42, p Ͻ .001) and had longer dwell times on
it (␹2
(1) ϭ 9.55, p ϭ .002) than participants who paid no attention
to ads. Further, participants who reported distraction by the ads
entered the ad above the text sooner (␹2
(1) ϭ 5.00, p ϭ .025) than
participants who did not report distraction. Moreover, there were
interactions between abrupt onset of the ad above the text and
reported attention to ads (␹2
(1) ϭ 7.43, p ϭ .006) and reported
distraction (␹2
(1) ϭ 9.27, p ϭ .002) (see Figure 4). Participants
who reported attention to ads had longer dwell times on the ad
above in the no-onset condition than in the onset condition,
whereas participant who did not report attention to ads had longer
dwell times in the onset condition. Reported distraction showed an
opposite pattern, that is, participants who reported distraction had
longer dwell times in the onset condition.
Eye movements on the ad to the right of the text. On
average, participants fixated the ad to the right of the text on 20%
of the trials. The number of entries and fixations were higher for
the ad on the right than for the ad above (see Figure 3). Abrupt
onset did not affect eye movements toward the ad on the right (see
Table 5). The self-report measures showed that participants who
reported attention to ads had a higher number of entries (␹2
(1) ϭ
18.24, p Ͻ .001) and fixations (␹2
(1) ϭ 27.45, p Ͻ .001) toward
the ad to the right than those who did not report attention to ads
(see Table 6). Further, participants who reported distraction by the
ads, fixated the ad on the right more often (␹2
(1) ϭ 5.23, p ϭ .022)
and had longer dwell times (␹2
(1) ϭ 6.79, p ϭ .009) on it than
those who did not report distraction.
The effect of abrupt ad onset. The time of first fixation
(entry time) on the ad above the text was not correlated with the ad
onset time, but such a correlation was observed for the ad on the
right (r ϭ .482, p Ͻ .001). The GEE model also revealed that
abrupt onset affected the entry times for the ad on the right
(␹2
(1) ϭ 8.00, p ϭ .005), whereas a similar, but nonsignificant
effect was observed for the ad above (␹2
(1) ϭ 3.53, p ϭ .060).
Figure 5 illustrates that both ads attracted more fixations during the
first five fixations after an ad onset compared to fixations occurring
later.
Discussion
The purpose of Experiment 2 was to test how abrupt onsets of
online ads affect reading and allocation of attention toward the
ads. The results showed an association between the ad onset
time and the time when the ad appearing to the right was fixated
for the first time. Even though no such effect was observed for
the ad above the text, the abrupt onset increased the number of
entries and number of fixations toward it. In addition, both ads
were fixated more often during the first five fixations after the
ad onset than during later fixations. The results suggest that
abrupt onset in the proximity of the text (i.e., to the right of it)
Table 4
The Effect of Abrupt Onset of Ads on Reading Measures in
Experiment 2
Ad above Ad to the right
Onset No-onset Onset No-onset
Reading performance
Comprehension .93 .89 .85 .89
Reading rate (wpm) 178.04 182.86 200.25 211.30
Eye movements on the text
No. of entries 2.16 1.97 2.22 2.40
No. of fixations 108.30 106.22 109.64 108.62
No. of regressions 23.25 22.70 24.20 23.36
Time of entry (s) .44 .46 .38 .47
MeanFixDur (ms) 260.66 260.56 265.77 264.40
182 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
captures more attention immediately, whereas abrupt onset in
the visual periphery captures attention less immediately, but
attention is nevertheless drawn toward the ad above the text
more often when it appears abruptly. Furthermore, abrupt onset
of the ad to the right of the text impaired comprehension
accuracy more than abrupt onset of the ad above. However,
comprehension was better when the ad above appeared abruptly
compared to when it was present throughout the trial. This is an
unexpected finding, possibly suggesting that an abrupt onset in
the periphery increases attention toward the text and results in
improved reading comprehension. Prior studies have reported
similar findings suggesting that peripheral ads might increase
users’ level of arousal and motivate them to devote increased
processing resources to the primary task (Burke et al., 2005;
Day et al., 2006).
Ad location also affected reading performance. Comprehension
was impaired more when an ad was shown to the right of the text
than above it. On the other hand, reading was slower when an ad
was presented above the text. Furthermore, slower reading was
associated with improved comprehension. This effect was even
stronger when an ad appeared to the right of the text. Improved
comprehension was also associated with an increased number of
fixations and regressions during reading, suggesting that a careful
reading strategy resulted in better comprehension than a more
superficial strategy. These results are generally in line with earlier
findings on the relationship between eye movements and text
memory (Hyöna¨, Lorch, & Kaakinen, 2002; Hyöna¨ & Nurminen,
2006).
Participants’ gaze behavior was consistent with their selfreports,
suggesting that participants who reported attention to ads
also fixated the ads more often and for longer time than participants
who did not report attention to ads. In addition, experienced
ad distraction was associated with earlier fixations on the ad above
as well as longer dwell time and greater number of fixations on the
ad to the right. Experienced attention and ad distraction also
interacted with the effect of abrupt ad onset, suggesting that abrupt
onset increased experienced distraction as reflected in the eye
movement measures (e.g., increased dwell time on the ad above
the text). In contrast, reported attention to ads was associated with
longer dwell time on the ad above in the no-onset condition,
whereas participants who did not report attention to ads had longer
dwell time on the ad above when it appeared abruptly. This
suggests that participants who paid attention to ads were not
affected by an abrupt onset, but instead looked at the ad above
longer in the no-onset condition (possibly because the ads were
visible for a longer time by being present throughout the trial).
However, participants who did not pay attention to ads had longer
dwell times on the ad above when it appeared abruptly, suggesting
that for them abrupt onset produced an automatic attention capture
resulting in longer dwell times on the ad above.
In Experiments 1 and 2, attentional allocation to online ads was
studied when the users had to read texts surrounded by ads.
However, the type of task users are engaged in may modulate the
effects of animation or abrupt ad onset on attentional allocation.
Experiment 3 examined the effect of task-orientation on the allocation
of attention to online text and advertisements.
Experiment 3
Previous studies suggest that processing of online ads varies as
a function of the difficulty of the primary task or task orientation.
Zhang (2000) showed that the effect of animation was related to
task difficulty so that animated ads influenced the performance of
simple tasks more than that of more difficult tasks. Burke et al.
(2005) also reported increased search times in the presence of
Figure 3. Number of entries and number of fixations to the ads in
Experiment 2. Brackets indicate significant differences in Wilcoxon
signed-ranks test. Error bars indicate SE.
Table 5
The Effect of Abrupt Onset of Ads on Eye Movements to Ads in
Experiment 2
Ad above Ad to the right
Onset No-onset Onset No-onset
No. of entries .36 .21 .48 .53
No. of fixations .76 .54 1.27 1.30
Time of entry (s) 15.94 12.49 15.69 13.92
Dwell time (ms) 765.86 873.56 1,125.72 1,263.47
Table 6
The Relationship Between Participants’ Self-Report Measures
and the Eye Movement Measures for Ads in Experiment 2
Attention Distraction
Yes No Yes No
Ad above text
No. of entries .28 .13 .26 .15
No. of fixations .65 .16 .57 .25
Time of entry (s) 13.88 28.10 16.26 19.96
Dwell time (ms) 817.88 437.74 760.22 644.32
Ad to the right of text
No. of entries .44 .11 .40 .15
No. of fixations 1.12 .17 1.05 .25
Time of entry (s) 14.41 21.50 15.39 16.83
Dwell time (ms) 1,180.20 507.78 1216.29 604.60
183EFFECT OF ADS ON ONLINE READING
commercial ads including colored graphics, when the task was to
search for exact matches of given headlines (easy task) in comparison
to when the sentence search task required semantic transformations
(difficult task).
Experiment 3 was conducted to also provide guidelines for
online advertisers as to what type of task conditions would benefit
from using animation or abrupt ad onsets and when it may be
better to avoid such design cues. Two tasks were compared to each
other; in one, participants read the texts to answer text content
questions (Experiment 3A) similarly to Experiment 1 and 2, while
in the other task participants were asked to browse the Web pages
according to their own interests (Experiment 3B). We expected
that the free browsing task is supposedly easier than reading for
comprehension, as it allows the participants to more freely distribute
their visual attention across the Web page. To test how taskorientation
affects attention to online ads, the following ad conditions
were selected as examples of different Web page styles: An
animated ad appeared above or to the right of the text after a
random delay between 0–12 s, or both ads were presented simultaneously
throughout the trial so that the ad above was static and
the ad to the right of the text was animated (identical to the S ϩ A
condition of Experiment 1). Moreover, to assess the effect of ad
presence, Experiment 3 included a baseline condition where no ads
were presented, which was not the case in Experiment 1 and 2,
where at least one ad was always present.
Method
Participants. Thirty-two volunteers in Experiment 3A and 30
volunteers in Experiment 3B gave an informed consent and were
paid to participate in the study (see Table 1). All had normal or
corrected-to-normal vision. They were not aware of the purpose of
the study, and none of them took part in Experiment 1 or 2.
Procedure. The same texts, advertisements and Web page
layouts were used as in Experiments 1 and 2. The task in Experiment
3A was to read for comprehension. In Experiment 3B,
participants were asked to browse the pages according to their own
interest and subsequently report the main content of the page with
a few words (i.e., we did not present text comprehension questions).
The following ad conditions were used: (a) no ads present
(baseline), (b) two ads present throughout the trial (the ad above
was static and the ad to the right of the text was animated, S ϩ A),
(c) only an animated ad above the text was present after a random
delay of 0–12 s (A ϩ blank), (d) only an animated ad to the right
of the text was present after a random delay of 0–12 s (blank ϩ A).
Data analyses. To ensure compatibility of results with Experiments
1 and 2, similar procedures for data preprocessing and
statistical analyses were used. The GEE model compared the
fixations on the ad above the text between Conditions b and c,
while the fixations on the ad on the right were compared between
Conditions b and d. In addition, the effect of abrupt ad onset on
attention toward the ads was tested similarly to Experiment 2.
Results of Experiments 3A and 3B were analyzed concurrently
using a mixed design. Eye movement measures on the text region
were subjected to repeated measures ANOVA with the task (i.e.,
experiment), participants’ subjective reports of attention to ads and
distraction by ads as between-participants factors and the ad condition
as a within-participant factor. Around 56% of the participants
in Experiment 3A and around 83% of the participants in
Experiment 3B reported that they had paid attention to ads constantly
or relatively often. Moreover, 69% of the participants in
Experiment 3A and 77% of the participants in Experiment 3B
Figure 4. Interactions between abrupt onset and participants’ self-reports for the dwell time on the ad above
the text in Experiment 2. Error bars indicate SE.
Figure 5. The probability of first fixating the ads at different ordinal
fixation positions after abrupt onset of the ads in Experiment 2. Brackets
indicate significant differences in Z test for two proportions.
184 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
reported that the ads distracted them constantly or relatively often.
They formed the groups who reported attention to or distraction by
ads. A similar design was used to analyze the results for the ad
regions with the GEE-model.
The samples in the two experiments differed from each other in
age (t(60) ϭ 3.22, p ϭ .002) and education (t(60) ϭ Ϫ3.06, p ϭ
.003), suggesting that participants in Experiment 3A were older
and less educated than participants in Experiment 3B (see Table 1).
The samples did not differ in computer and Internet experience or
in the frequency of Internet use. To control for the betweensamples
differences, age and education were included as factors in
the pooled analyses of Experiment 3A and 3B. We first report how
task and type of ad affected reading and attention toward the ads
(i.e., fixations on the text and ad regions) and then how abrupt
onsets captured attention.
Results
Reading performance, eye movements, and subjective reports.
Reading rates differed between the tasks (F(1, 56) ϭ
14.03, p Ͻ .001, ␩p
2
ϭ .200), suggesting faster reading when the
task was to browse the pages according to participants’ own
interest (Experiment 3B) compared to the reading-forcomprehension
task in Experiment 3A (see Table 7). Furthermore,
the tasks differed from each other in the number of
fixations (F(1, 56) ϭ 16.81, p Ͻ .001, ␩p
2
ϭ .231), and in the
number of regressions made in the text region (F(1, 56) ϭ 9.03,
p ϭ .004, ␩p
2
ϭ .139), with fewer fixations and regressions in
the free browsing than in the reading task. The text entry times
showed a Task ϫ Ad Condition interaction (F(1, 56) ϭ 4.99,
p ϭ .018, ␩p
2
ϭ .082), suggesting that the text was entered later
during free browsing than reading, but only in the condition
where the ad above the text was static and the ad to the right of
the text was animated (S ϩ A) (t(60) ϭ 2.41, p ϭ .023, d ϭ
.65). Participants’ self reports did not affect eye movements on
the text region.
Eye movements on the ad above the text. In Experiment
3A, participants fixated the ad above the text in 11% of the trials,
whereas in Experiment 3B the ad above was fixated in 32% of the
trials. Figure 6 shows the mean number of entries and fixations to
both ads. According to the GEE model, the ad above the text was
entered (␹2
(1) ϭ 16.57, p Ͻ .001) and fixated (␹2
(1) ϭ 12.13, p Ͻ
.001) more often during free browsing (Experiment 3B) than
during reading (Experiment 3A) (see Table 8). In addition, the
dwell time (␹2
(1) ϭ 9.89, p ϭ .002) on the ad above was longer
during free browsing than during reading. There was also a Task ϫ
Ad Condition interaction (␹2
(1) ϭ 6.20, p ϭ .013) in the entry
time, suggesting that the ad above was entered later during the free
browsing task when the ad was animated and appeared abruptly
compared to the S ϩ A condition (p ϭ .036), while in the reading
task the entry times on the ad above did not differ between the ad
conditions.
Self-reported attention to ads was associated with the number of
entries (␹2
(1) ϭ 18.75, p Ͻ .001) and number of fixations (␹2
(1) ϭ
20.80, p Ͻ .001) on the ad above the text. These observations
suggested more entries and fixations (.49 Ϯ 1.15 SD and 1.10 Ϯ
2.97 SD) on the ad above for participants who reported attention to
ads than for participants who did not report attention to ads (.13 Ϯ
.48 SD and .24 Ϯ 1.26 SD, respectively).
Eye movements on the ad to the right of the text. In
Experiment 3A, participants fixated the ad to the right of the text
in 23% of the trials, whereas participants in Experiment 3B fixated
the ad to the right in 33% of the trials. In both tasks, the number
Table 7
Means of Reading Performance and Eye Movement Measures in Experiment 3, as a Function of
Task (Exp. 3A: Reading vs. Exp. 3B: Free Browsing) and Studied Conditions
Ad condition Baseline S ϩ A A ϩ blank Blank ϩ A
Comprehension
Exp. 3A .81 .87 .84 .77
Reading rate (wpm)
Exp. 3A 173.73 169.16 169.74 163.60
Exp. 3B 282.37 291.39 283.59 287.09
No. of entries
Exp. 3A 1.95 2.02 2.00 2.54
Exp. 3B 2.63 3.31 2.77 3.43
No. of fixations
Exp. 3A 109.55 110.32 111.44 113.40
Exp. 3B 82.45 78.62 81.92 81.77
No. of regressions
Exp. 3A 23.33 23.00 23.14 24.12
Exp. 3B 17.31 16.49 17.52 17.52
Time of entry (s)
Exp. 3A .34 .40 .37 .36
Exp. 3B .39 .92 .52 .42
MeanFixDur (ms)
Exp. 3A 277.76 281.61 276.48 282.17
Exp. 3B 258.74 261.73 262.21 262.01
Note. Baseline ϭ no ads were presented; S ϩ A ϭ the ad above was static and the ad to the right was animated
(the same condition as in Experiment 1); A ϩ Blank ϭ only the ad above (animated) appeared after a random
delay of 0–12 s; Blank ϩ A ϭ only the ad to the right (animated) appeared after a random delay of 0–12 s.
185EFFECT OF ADS ON ONLINE READING
of entries and fixations were higher for the ad on the right than
above the text (see Figure 6). The number of entries (␹2
(1) ϭ 6.43,
p ϭ .011) and the number of fixations (␹2
(1) ϭ 12.15, p Ͻ .001)
toward the ad on the right differed between the tasks, suggesting
more frequent entries and fixations in the free browsing than in the
reading task (see Table 8). In addition, the dwell time (␹2
(1) ϭ
20.02, p Ͻ .001) on the ad on the right was longer during free
browsing than during reading. We also observed a Task ϫ Ad
Condition interaction in the number of entries (␹2
(1) ϭ 4.50, p ϭ
.025) and entry times (␹2
(1) ϭ 6.99, p ϭ .008) toward the ad to the
right of the text. These interactions suggest that in the S ϩ A
condition the ad on the right was fixated more often (p ϭ .009) and
earlier (p ϭ .032) during free browsing than during reading.
Self-reported attention to ads was associated with the number of
entries (␹2
(1) ϭ 7.13, p ϭ .008), number of fixations (␹2
(1) ϭ
15.97, p Ͻ .001), and dwell time (␹2
(1) ϭ 15.41, p Ͻ .001) on the
ad to the right of the text. These observations suggest more
frequent entries (2.08 Ϯ 1.47 SD), fixations (5.87 Ϯ 6.60 SD), and
longer dwell time (1799.41 Ϯ 2046.33 SD) on the ad on the right
for participants who reported paying attention to the ads compared
to participants who did not report attention to the ads (1.48 Ϯ 1.00
SD, 2.27 Ϯ 3.43 SD, and 621.77 Ϯ 718.45 SD).
The effects of abrupt ad onset. The time of the first fixation
landing on the ad above was not associated with the ad onset time
in Experiment 3A; however, in Experiment 3B the ad onset time
correlated with the time of first fixation (r ϭ .256, p ϭ .011). The
GEE model showed that in both experiments, the number of entries
(␹2
(1) ϭ 4.13, p ϭ .042) and the dwell time on the ad above the
text (␹2
(1) ϭ 4.58, p ϭ .032) were associated with its abrupt onset
time.
For the ad to the right, correlations between the ad onset time
and the time of first fixation were observed in both experiments
(3A: r ϭ .246, p ϭ .014; 3B: r ϭ .520, p Ͻ .001). Moreover, the
GEE model indicated that the ad onset time was associated with
the time of first fixation landing on it (␹2
(1) ϭ 7.64, p ϭ .006).
Figure 7 illustrates that the ads attracted more fixations during the
first five fixations after an ad onset compared to later occurring
fixations (except for the ad above in Experiment 3A).
To examine the effect of task on attentional capture by abrupt ad
onset, the number of intervening fixations between the ad onset
and observers’ first fixation on the ad was compared between
Experiments 3A and 3B. The task did not affect the number of
fixations between the ad onset and the time of first fixation on the
ad above. However, the ad on the right was fixated earlier after its
onset in the free browsing than in the reading task (␹2
(1) ϭ 11.11,
p ϭ .001). Moreover, the probability of first fixation on the ad on
the right occurring during the first five fixations (Z ϭ 1.66, p Ͻ
.05) or during the fixation sequence of 6–10 (Z ϭ 1.66, p Ͻ .05)
after its onset was higher in Experiment 3B than in Experiment 3A
Figure 6. Number of entries and number of fixations toward the ads in Experiment 3. Brackets indicate
significant differences in Wilcoxon signed-ranks test. Error bars indicate SE.
Table 8
The Effect of Ad Condition on Eye Movements to Ads in
Experiment 3, as a Function of Task (Exp. 3A: Reading vs. Exp.
3B: Free Browsing) and Studied Conditions
Ad above Ad to the right
A ϩ blank S ϩ A Blank ϩ A S ϩ A
No. of entries
Exp. 3A .19 .13 .68 .47
Exp. 3B .91 .84 1.42 1.37
No. of fixations
Exp. 3A .40 .29 1.25 .95
Exp. 3B 2.30 2.15 4.56 4.21
Time of entry (s)
Exp. 3A 12.27 20.62 19.25 20.69
Exp. 3B 17.30 11.56 14.11 8.90
Dwell time (ms)
Exp. 3A 680.96 1015.09 882.69
Exp. 3B 1,184.82 2,309.41 2,057.30
Note. A ϩ Blank ϭ only the ad above (animated) appeared after a
random delay of 0–12 s; S ϩ A ϭ the ad above was static and the ad to
the right was animated; Blank ϩ A ϭ only the ad to the right (animated)
appeared after a random delay of 0–12 s.
186 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
(see Figure 7). No task differences were observed when an animated
ad appeared abruptly above the text.
Discussion
Participants paid more attention to ads in Experiment 3B
where the task was to browse the pages according to their own
interest compared to Experiment 3A where the task was to read
for comprehension. This was shown by more frequent eye
entries and fixations as well as longer dwell times on the ads in
Experiment 3B than in Experiment 3A. Moreover, the task
instruction affected eye movements during text reading, suggesting
more superficial reading when the task was to browse
the pages compared to a more careful reading strategy required
in Experiment 3A, where participants were instructed to read
for comprehension. This was reflected in faster reading rates in
Experiment 3B and increased number of fixations and regressions
on the text region in Experiment 3A. Comparisons between
the experiments also showed that when both ads were
present throughout the trial (the S ϩ A condition), the text was
entered earlier in the reading-for-comprehension than in the
free browsing task. An early text entry in the reading task
makes sense because the task instructions made the text relevant
and the ads irrelevant for the task.
In the free browsing task, we observed a correlation between the
abrupt onset time of both ads and the time when the participants
first fixated an ad, whereas in the reading task such an effect was
observed only for the ad on the right. The result supports the
conclusion that during reading an abrupt ad onset in the proximity
of the text captures attention efficiently, whereas an ad onset does
not produce an immediate attention capture when the ad appears
further in the visual periphery. These results are consistent with
those obtained in Experiment 2 (also using a reading task), where
the abrupt onset of the ad on the right was associated with the time
of first fixations to it, while no such effect was observed for the ad
above the text.3
The results of Experiment 3 also suggest that participants’
self-reports are compatible with their actual eye behavior. That is,
participants who reported paying attention to ads had more frequent
entries and fixations on both ads, as well as longer dwell
time on the ad to the right than participants who did not report
attention to ads. Moreover, the results of Experiment 3 indicate
that the mere presence of ads did not interfere with reading,
because the baseline condition where no ads were present did not
prove to be better in any of the reading measures. This finding,
combined with other findings of the present study, supports the
hypothesis that distraction by ads primarily occurs through overt
attention to ads rather than through covert processing.
The ad’s location in relation to the text appeared to be a relevant
factor. An abrupt onset of the ad above the text did not distract
reading, whereas the onset of the ad to the right captured attention.
This is presumably because of the proximity of the ad on the right
with respect to the ends of text lines. Thus, when reading a text
participants frequently approach the ad on the right, thus increasing
the likelihood of the ad capturing attention. Or put it differently,
when Web users focus their attention on the reading task,
they are more capable of overriding the attentional capture by
abrupt ad onsets when an ad appears in a peripheral location than
when it appears in the proximity of the main task area. This
conclusion is inconsistent with the study of Stenfors et al. (2003),
who suggest that novel stimuli as such do not attract attention in
the Web environment, because participants are likely to learn the
ad locations and consequently do not pay attention to taskirrelevant
stimuli presented in predictable locations. It should be
noted, however, that Stenfors et al. studied only the effects of ads
appearing on the top of the screen in a visual cuing task where the
color of a central cue indicated the location of a forthcoming
lateral target. The top location of an ad in the present experiments
proved to attract less attention than the ads located on the right side
of the Web pages. However, the location effects may be qualified
3
It may be noted, however, that the comparisons between the abrupt
onset conditions and the S ϩ A condition are not as solid as in Experiment
2, where only one ad was present at a time and the effect of an abrupt onset
was compared with the same ad when present throughout the trial.
Figure 7. The probability of first fixating the ads at different ordinal fixation positions after the abrupt onset
of either the ad above or the ad to the right of the text. In Experiment 3A, the task was to read the texts for
comprehension, and in Experiment 3B the participants were asked to browse the Web pages according to their
interest. Brackets indicate significant differences in Z test for two proportions.
187EFFECT OF ADS ON ONLINE READING
by task conditions, that is, in some tasks the upper part of a Web
page may be task-relevant.
General Discussion
The present study investigated how Web users allocate attention
toward advertisements during text reading (and free browsing). We
manipulated the combination of animated and static ads presented
simultaneously in the periphery and in the proximity of the text
region. Moreover, the ads appeared on the screen together with the
text either in the beginning of the trial or abruptly after some delay.
Finally, we manipulated the nature of the primary task. In Experiment
1, 2, and 3A, the task was to read the texts for comprehension
and the comprehension accuracy was tested after each text. In
Experiment 3B, participants were instructed to browse the Web
pages according to their own interests and briefly report the main
content of the page. Allocation of attention was examined by
registering participants’ eye movements to measure the extent to
which ads capture attention during reading and whether the presence
of ads affects online reading behavior. The former case refers
to overt attentional capture by ads, while the latter case reflects
covert allocation of attentional resources. In the analyses, participants’
self-reports on experienced attention to ads and distraction
by them were compared with their actual eye movement behavior.
The main results of the present study may be summarized as
follows: (a) Contrary to some earlier studies, we found that online
ads are not ignored during reading, irrespective of whether the task
was engaging (reading for comprehension) or free browsing. In our
experiments, the ads were overtly and not only covertly attended.
This was shown in that the effects of ads on reading were always
accompanied by similar effects obtained for overt attention to ads.
(b) Animation distracted reading especially when a combination of
static and animated ad was presented. Such conditions led to
increased attentional capture relative to conditions where both ads
were static or animated. (c) Abrupt onset of ads during reading
captured attention, but this effect was modulated by ad location
and task-orientation. (d) Ads located immediately to the right of
the text region produced a greater attentional capture than ads
positioned above the text region. (e) The ads were attended to more
during a free browsing than during a reading task. Moreover,
attentional capture by abrupt onset of both ads was observed in the
free browsing task, whereas in the reading-for-comprehension
task, only the abrupt onset of the ad to the right of the text captured
attention. In the following, we discuss each finding in more detail.
The finding that an effect of ads on reading was always accompanied
by eye fixations on ads is in line with the view that
attentional capture by ads is primarily related to mechanisms of
overt attention. This finding runs counter to studies (Burke et al.,
2005; Day et al., 2006; Dre`ze & Hussherr, 2003), suggesting that
online ad processing occurs mostly peripherally via the covert
attention mechanism. The fact that the ad conditions had only a
slight effect on reading eye movements further supports the hypothesis
that distraction by ads occurs through overt fixations
toward the ads rather than as covert processing of ads. However,
the finding in Experiment 1 that only the ad above the text
influenced reading possibly reflects the fact that the ad to the right
was attended more than the ad above. In other words, the ad on the
right affected overt attentional allocation, while the ad above
possibly exerted its influence also via covert attentional allocation
(i.e., its influence was less robust). Corroborative evidence in
support of the overt attentional mechanism being significantly
involved comes from the participants’ retrospective verbal reports
that were consistent with their actual eye behavior. Furthermore,
comprehension accuracy was in line with eye movement results,
that is, accuracy improved by a slow reading rate, indicating that
a careful reading strategy (also indexed by frequent fixations and
regressions during reading) resulted in better comprehension than
a more superficial strategy.
The result that a combination of one static and one animated ad
increased attentional capture is an interesting but surprising finding,
as we expected two animated ads to produce the most attentional
capture. It is likely that when two animated ads are present
simultaneously, they are equally salient (both contain motion) and
thus compete with equal strength for users’ attentional resources.
Moreover, increased saliency of ads is likely to be accompanied by
increased attentional resources being invested in ignoring taskirrelevant
stimuli. As a consequence, two animated ads are ignored
by top-down attention as effectively as two static ones, but when
only one of the ads is animated, it is individually more salient and
is thus more likely to capture attention. This finding suggests that
to secure good reading performance, displaying one animated ad in
the presence of a static ad should be avoided, as such displays
prevent readers from ignoring distraction by task-irrelevant stimuli.
For advertisers the results suggest that to attract attention in
cluttered Web pages, several animated ads should not be presented
simultaneously, because an excessive use of animation seems to
help users avoid task-irrelevant information. However, the underlying
attentional mechanism, that is, whether suppression of taskirrelevant
ad information or enhanced processing of relevant text
information is involved, is beyond the scope of the present study
and needs more empirical evidence.
Another major finding was that abrupt ad onsets capture attention,
but the effect depends on the ad location and task-orientation.
This suggests that when engaged in a reading task, Web users can
ignore abrupt stimulus onsets in the periphery but not in the
proximity of the main task area. The use of an avoidance strategy
was probably boosted by the fact that the ads always appeared in
the same two locations, so their location was highly predictable
and thus easier to ignore (cf. Theeuwes & Burger, 1998). In future
studies, it would be interesting to study the effects of abrupt onsets
of ads appearing in multiple, less predictable locations.
The ads located close to the right border of the text were
attended to more than ads above the text. This suggests that motion
or abrupt onset attract users’ visual attention toward ads when they
frequently approach the ads with their eyes. This happens every
time readers fixate the ends of text lines located in the proximity
of the ad. When an ad is positioned above the text, readers’ gaze
does not approach it; instead, the further they advance in the text
the longer away their gaze is from the ad. Thus, to prevent users
from unintentionally attending to animated ads, the ads should be
positioned as far away as possible from the display area the users
are likely to gaze at. Ad positioning, however, depends on the
goals of the online content providers, because the Web page
publisher may also want to direct attention toward the ads. In that
case, the position to the right of the text is the most effective
location. Similarly, previous research has reported that ads located
close to the main task area or in the lower visual field are beneficial
for advertising efficiency measured in terms of perception of
188 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨
and memory for ads (Kuisma et al., 2010) or brand attitude
(Goodrich, 2010). It should be noted, however, that the design of
the present experiments could not disentangle the effects of ad
location and format from each other, because the ad to the right
was always vertical and the ad above was horizontal.
One of the main findings was that more attention was allocated
to ads when task constraints were less stringent (free browsing).
Furthermore, during free browsing we obtained evidence for attentional
capture by abrupt onsets of ads appearing in the two
studied locations. Combined with the previous findings, these
observations suggest that the current goal of Web users can exert
a strong top-down influence on attentional allocation during Web
browsing so that the presence of ads and their abrupt onsets attract
more attention during a free browsing than during a reading task.
This was also reflected in the participants’ verbal reports: over
80% of the participants in the free browsing task paid attention to
ads, whereas only around 40–50% of the participants in the
reading task reported that the ads had attracted their attention.
Thus, in line with other authors (Rayner et al., 2008; Wedel et al.,
2008), we suggest that advertisers should consider the users’ key
goals when selecting the format and placement for online ads. This
perspective is largely ignored by the advertising research that has
so far mainly emphasized how specific layout and content elements
of ads influence viewers’ gaze, assuming that only
bottom-up attentional patterns matter.
From the theoretical perspective, the present study demonstrated
that findings from basic attention research and the theories of
visual attention derived from them, especially research pointing to
the significance of visual saliency (Itti & Koch, 2000), the distinction
between bottom-up versus top-down control of attention
(Theeuwes, 1994; Theeuwes & Burger, 1998), the principles of the
central capacity theory (Kahneman, 1973), and the research related
to abrupt stimulus onsets (e.g., Theeuwes et al., 1999; Yantis &
Jonides, 1984) can be successfully applied also to the Web environment
when the influences of advertising are studied. This is
relevant, as arguments to the contrary have been made (Diaper &
Waelend, 2000; Zhang, 2000).
Finally, the present study provides important practical suggestions
for efficient Web design. The suggestions are in line
with the view proposed by Rosen and Purinton (2004) in that we
believe the development of effective and user-friendly Web
designs can be significantly assisted by utilizing perspectives
offered by cognitive psychology. Given a choice, content providers
and online advertisers should prefer ad formats and
context that only minimally disturb users’ main task because, as
previous studies have shown, distraction can lead into irritation
and unfavorable attitudes toward a Website (e.g., Gao et al.,
2004). As practical implications for minimizing the distraction
by ads, the findings from the present experiments can be summarized
as follows: If more than one ad is presented on the
screen simultaneously, they should all be either animated or
static, to allow users to efficiently ignore task-irrelevant ad
information. Preferably, the ads should not be placed in the
close proximity of the main task area, because ads presented
close to the main task area capture attention and result in
disruptions to the main task. Users’ task and variations in task
load should be considered when placing ads on Web pages. Our
results demonstrate that ads capture more attention during a free
browsing task than when reading for comprehension.
References
Allport, A. D. (1980). Attention and performance. In G. Claxton (Ed.),
Cognitive psychology: New directions (pp. 112–153). London: Routledge
& Kegan Paul.
Brockmole, J. R., & Henderson, J. M. (2005). Prioritization of new objects
in real-world scenes: Evidence from eye movements. Journal of Experimental
Psychology: Human Perception and Performance, 31, 857–868.
Burke, M., Hornof, A., Nilsen, E., & Gorman, N. (2005). High-cost banner
blindness: Ads increase perceived workload, hinder visual search, and
are forgotten. Transactions on Computer-Human Interaction, 12, 423–
445.
Burns, K. S., & Lutz, R. J. (2006). The function of format consumer
responses to six on-line advertising formats. Journal of Advertising, 35,
53–63.
Day, R.-F., Shyi, G. C.-W., & Wang, J.-C. (2006). The effect of flash
banners on multiattribute decision making: Distractor or source of
arousal? Psychology & Marketing, 23, 369–382.
Deubel, H., & Schneider, W. X. (1996). Saccade target selection and object
recognition: Evidence for a common attentional mechanism. Vision
Research, 36, 1827–1837.
Diaper, D., & Waelend, P. (2000). World Wide Web working whilst
ignoring graphics: Good new for web page designers. Interacting with
Computers, 13, 163–181.
Diggle, P. J., Heagerty, P., Liang, K.-Y., & Zeger, S. L. (2002). Analysis
of longitudinal data. New York, NY: Oxford University Press.
Dore´-Mazars, K., Pouget, P., & Beauvillain, C. (2004). Attentional selection
during preparation of eye movements. Psychological Research, 69,
67–76.
Dre`ze, X., & Hussherr, F.-X. (2003). Internet advertising: Is anybody
watching? Journal of Interactive Marketing, 17, 8–23.
Franconeri, S. L., & Simons, D. J. (2003). Moving and looming stimuli
capture attention. Perception & Psychophysics, 65, 999–1010.
Gao, Y., Koufaris, M., & Ducoffe, R. H. (2004). An experimental study of
the effects of promotional techniques in web-based commerce. Journal
of Electronic Commerce in Organizations, 2, 1–20.
Goodrich, K. (2010). What’s up? Exploring upper and lower visual field
advertising effects. Journal of Advertising Research, 50, 91–106.
Hardin, J., & Hilbe, J. (2003). Generalized estimating equations. London:
Chapman and Hall/CRC.
Hillstrom, A. P., & Yantis, S. (1994). Visual motion and attentional
capture. Perception & Psychophysics, 55, 399–411.
Hong, W., Thong, J. Y. L., & Tam, K. Y. (2004). Does animation attract
online users’ attention? The effects of flash on information search
performance and perceptions. Information Systems Research, 15,
60–86.
Hyöna¨, J., Lorch, Jr., R. F., & Kaakinen, J. K. (2002). Individual differences
in reading to summarize expository text: Evidence from eye
fixation patterns. Journal of Educational Psychology, 94, 44–55.
Hyöna¨, J., & Nurminen, A.-M. (2006). Do adult readers know how they
read? Evidence from eye movement patterns and verbal reports. British
Journal of Psychology, 97, 31–50.
Itti, L., & Koch, C. (2000). A saliency-based search mechanism for overt
and covert shifts of visual attention. Vision Research, 40, 1489–1506.
Jonides, J., & Yantis, S. (1988). Uniqueness of abrupt visual onset in
capturing attention. Perception & Psychophysics, 43, 346–354.
Kahneman, D. (1973). Attention and effort: Prentice Hall: Englewood
Cliffs, NJ.
Kuisma, J., Simola, J., Uusitalo, L., & Öörni, A. (2010). The effects of
animation and format on the perception and memory of online advertising.
Journal of Interactive Marketing, 24, 269–282.
Laarni, J., Simola, J., Kojo, I., & Na¨sa¨nen, R. (2004). Reading vertical text
from a computer screen. Behaviour & Information Technology, 23,
75–82.
189EFFECT OF ADS ON ONLINE READING
Little, S. R., & Rubin, D. B. (1987). Statistical analysis with missing data.
New York: Wiley.
McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus
during a fixation in reading. Perception & Psychophysics, 17, 578–586.
McLeod, P., Driver, J., & Crisp, J. (1988). Visual search for a conjunction
of movement and form is parallel. Nature, 332, 154–155.
McLeod, P., Driver, J., Dienes, Z., & Crisp, J. (1991). Filtering by movement
in visual search. Journal of Experimental Psychology: Human
Perception and Performance, 17, 55–64.
Radach, R., Lemmer, S., Vorstius, C., Heller, D., & Radach, K. (2003). Eye
movements in the processing of print advertisements. In J. Hyöna¨, R.
Radach, & H. Deubel (Eds.), The mind’s eye: Cognitive and applied
aspects of eye movement research (pp. 609–632). Amsterdam: Elsevier.
Rayner, K. (1998). Eye movements in reading and information processing:
20 years of research. Psychological Bulletin, 124, 372–422.
Rayner, K., Miller, B., & Rotello, C. M. (2008). Eye movements when
looking at print advertisements: The goal of the viewer matters. Applied
Cognitive Psychology, 22, 697–707.
Rayner, K., Rotello, C. M., Stewart, A. J., Keir, J., & Duffy, S. A. (2001).
Integrating text and pictorial information: Eye movements when looking
at print advertisements. Journal of Experimental Psychology: Applied, 7,
219–226.
Rosen, D. E., & Purinton, E. (2004). Website design: Viewing the web as
a cognitive landscape. Journal of Business Research, 57, 787–794.
Stenfors, I., More´n, J., & Balkenius, C. (2003). Behavioural strategies in
web interaction: A view from eye-movement research. In J. Hyöna¨, R.
Radach, & H. Deubel (Ed.), The mind’s eye: Cognitive and applied
aspects of eye movement research (pp. 633–644). Amsterdam: Elsevier.
Theeuwes, J. (1994). Stimulus-driven capture and attentional set: Selective
search for color and visual abrupt onsets. Journal of Experimental
Psychology: Human Perception and Performance, 20, 799–806.
Theeuwes, J., & Burger, R. (1998). Attentional control during visual
search: The effect of irrelevant singletons. Journal of Experimental
Psychology: Human Perception and Performance, 24, 1342–1353.
Theeuwes, J., Kramer, A. F., Hahn, S., Irwin, D. E., & Zelinsky, G. J.
(1999). Influence of attentional capture on oculomotor control. Journal
of Experimental Psychology: Human Perception and Performance, 25,
1595–1608.
Tversky, B., Morrison, J. B., & Betrancourt, M. (2002). Animation: Can it
falilitate? International Journal of Human-Computer Studies, 57, 247–
262.
Wedel, M., Pieters, R., & Liechty, J. (2008). Attention switching during
scene perception: How goals influence the time course of eye movements
across advertisements. Journal of Experimental Psychology: Applied,
14, 129–138.
Yantis, S., & Egeth, H. E. (1999). On the distinction between visual
salience and stimulus-driven attentional capture. Journal of Experimental
Psychology: Human Perception and Performance, 25, 661–676.
Yantis, S., & Hillstrom, A. P. (1994). Stimulus-driven attentional capture:
Evidence form equiluminant visual objects. Journal of Experimental
Psychology: Human Perception and Performance, 20, 95–107.
Yantis, S., & Jonides, J. (1984). Abrupt visual onsets and selective attention:
Evidence from visual search. Journal of Experimental Psychology:
Human Perception and Performance, 10, 601–621.
Yoo, C. Y., & Kim, K. (2005). Processing of animation in online banner
advertising: The roles of cognitive and emotional responses. Journal of
Interactive Marketing, 19, 18–34.
Zhang. (2000). The effects of animation on information seeking performance
on the world wide web: Securing attention or interfering with
primary task. Journal of the Association for Information Systems, 1,
1–28.
Received March 3, 2010
Revision received February 25, 2011
Accepted March 24, 2011 Ⅲ
Correction to Beck et al. (2010)
In the article “Measuring Search Efficiency in Complex Visual Search Tasks: Global and Local
Clutter,” by Melissa R. Beck, Maura C. Lohrenz, and J. Gregory Taflon (Journal of Experimental
Psychology: Applied, Vol. 16, No. 3, pp. 238–250), the copyright for the article was incorrectly
listed. The article is in the public domain.
DOI: 10.1037/a0023640
190 SIMOLA, KUISMA, ÖÖRNI, UUSITALO, AND HYÖNA¨