ArticlePDF Available

High-cost banner blindness: Ads increase perceived workload, hinder visual search, and are forgotten

Authors:

Abstract and Figures

The seeming contradiction between "banner blindness" and Web users' complaints about distracting advertisements motivates a pair of experiments into the effect of banner ads on visual search. Experiment 1 measures perceived cognitive workload and search times for short words with two banners on the screen. Four kinds of banners were examined: (1) animated commercial, (2) static commercial, (3) cyan with flashing text, and (4) blank. Using NASA's Task Load Index, participants report increased workload under flashing text banners. Experiment 2 investigates search through news headlines at two levels of difficulty: exact matches and matches requiring semantic interpretation. Results show both animated and static commercial banners decrease visual search speeds. Eye tracking data reveal people rarely look directly at banners. A post hoc memory test confirms low banner recall and, surprisingly, that animated banners are more difficult to remember than static look-alikes. Results have implications for cognitive modeling and Web design.
Content may be subject to copyright.
________________________________________________________________________
This work was supported by the National Science Foundation through Grant IIS-0308244 to the University of
Oregon with Anthony Hornof as the principle investigator. Authors’ addresses: Moira Burke, Human-Computer
Interaction Institute, Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, PA 15213, moira@cmu.edu;
Anthony Hornof, Computer and Information Science Department, University of Oregon, Eugene, OR 97403,
hornof@cs.uoregon.edu; Erik Nilsen, Psychology Department, Lewis & Clark College, Portland, OR, 97219,
nilsen@lclark.edu; Nicholas Gorman, Department of Preventative Medicine, University of Southern California,
1000 South Fremont Ave., HAS 5235, Alhambra, CA 91803, ngorman@usc.edu.
To appear in ACM Transactions on Computer-Human Interaction (TOCHI). Copyright 2005 Burke, Hornof,
Nilsen, and Gorman.
High-Cost Banner Blindness: Ads Increase
Perceived Workload, Hinder Visual Search, and
Are Forgotten
MOIRA BURKE
Carnegie Mellon University
ANTHONY HORNOF
University of Oregon
ERIK NILSEN
Lewis & Clark College
and
NICHOLAS GORMAN
University of Southern California
________________________________________________________________________
The seeming contradiction between “banner blindness” and Web users’ complaints about distracting
advertisements motivates a pair of experiments into the effect of banner ads on visual search. Experiment 1
measures perceived cognitive workload and search times for short words with two banners on the screen. Four
kinds of banners were examined: (1) animated commercial, (2) static commercial, (3) cyan with flashing text,
and (4) blank. Using NASA’s Task Load Index, participants report increased workload under flashing text
banners. Experiment 2 investigates search through news headlines at two levels of difficulty: exact matches and
matches requiring semantic interpretation. Results show both animated and static commercial banners decrease
visual search speeds. Eye tracking data reveal people rarely look directly at banners. A post hoc memory test
confirms low banner recall and, surprisingly, that animated banners are more difficult to remember than static
look-alikes. Results have implications for cognitive modeling and Web design.
Categories and Subject Descriptors: H5.2 [Information Interfaces and Presentation]: User Interfaces –
Graphic user interfaces (GUI); Screen design; H1.2 [Information Interfaces and Presentation]:
User/Machine Systems – Human information processing
General Terms: Design, Experimentation, Human Factors
Additional Key Words and Phrases: Animation, visual search, banner ads, eye tracking, World Wide Web
________________________________________________________________________
1. INTRODUCTION
Web designers routinely animate advertisements in an attempt to make them
more conspicuous. Yet few empirical studies explore the effect of animation on a
concurrent visual task, and few if any examine it in an ecologically realistic context.
Early research suggested that Web users are functionally blind to rectangular graphics
that they perceive to be advertisements [Benway and Lane 1998], but more recent studies
indicate that people do notice ads, dislike them, and that site credibility suffers [Fogg et
al. 2001]. Ad-blocking software, such as the products summarized by Rowe et al. [2001],
are popular. However, the HCI community lacks data that conclusively demonstrate
whether animated ads impede common visual Web tasks.
Animation ostensibly aids memory, but little evidence supports this theory when
applied to Web advertising. Advertisers want to leave a lasting impression: favorable
brand recall and viewer attitude matter as much as other effectiveness metrics (i.e., click-
through) [Interactive Advertising Bureau 2001]. How animation affects subjective user
experience and whether it aids memory have yet to be thoroughly investigated. Is
animation beneficial enough to advertisers to outweigh its negative reputation? We seek
to answer these questions through two experiments with animated banners: one that
measures search time and participants’ impressions of task workload, and another that
tests ad recall. The research presented here confirms that animated banner ads interfere
with common Web tasks and yet are no more memorable than static ads.
1.1 Animation and Attention Capture
Humans assign attentional priority via two biases: (a) exogenous, also known as
stimulus-driven or bottom-up, in which elements attract attention regardless of the task,
and (b) endogenous, also known as goal-directed or top-down, in which people attend to
elements based on a task strategy [Yantis 2000]. Separating the two has proven
challenging even under highly-controlled experimental design [Bacon and Egeth 1994,
Folk and Remington 1998]. Web tasks are even more complicated because viewers’
goals are not always well-defined, and ads with tempting words like “FREE” may attract
attention through semantic appeal. However, traditional studies of endogenous and
exogenous attention capture provide clues as to how animation affects Web users. There
are two prevailing schools of thought: the first contends that certain forms of animation
always attract attention from the bottom up, whereas the second argues that people
unintentionally create task-completion (top-down) strategies that make irrelevant stimuli
relevant, and thus distracting.
Motion attracts attention, though researchers disagree on the extent to which this
occurs automatically. Yantis and colleagues [Yantis and Jonides 1990, Jonides and
Yantis 1988, Hillstrom and Yantis 1994] found that abruptly-appearing stimuli capture
attention in a purely exogenous manner. However, people do not involuntarily look at
other forms of animation, such as oscillation and simple shape changes. Motion per se
does not attract attention, but rather abrupt appearances create new perceptual objects and
these new objects tend to attract attention. Mere luminance changes were ruled out as a
confound [Enns et al. 2001]. Even when the abruptly appearing items were known to be
irrelevant to the task, people still had trouble ignoring them [Remington et al. 1992].
However, Franconeri and Simons [2003] refuted the “new perceptual object” theory,
proposing instead that attention is delegated to events of behavioral urgency. They
maintained that with more salient stimuli, other kinds of motion will capture attention as
strongly as abrupt onset. Looming and disocclusion (moving from behind another object)
were highly distracting, while receding and unique-coloring were not. Combined, these
experiments suggest that some kinds of motion will inherently attract attention regardless
of the viewer’s intent.
The second prevailing theory is that no stimulus truly captures attention
automatically; instead, people adopt task-completion strategies that make them
susceptible to certain stimulus properties [Folk et al. 1992, Folk and Remington 1998,
Remington et al. 2001, Bacon and Egeth 1994, Yantis and Egeth 1999]. For example,
observers searching for a particular singleton (an object unique from its neighbors in one
dimension, such as shape) will be delayed by an irrelevant singleton in another dimension
(color, orientation) [Pashler 2001, Theeuwes 1991, Folk and Remington 1998]. This
effect was originally attributed to exogenous capture. Bacon and Egeth [1994] instead
concluded that, though the additional singleton was not informative of target location,
searchers were in singleton detection mode, deliberately attending to anything that
“popped out,” because it required less effort than consciously filtering for the singleton in
the relevant dimension. When the same target was no longer a singleton (i.e., there were
multiple instances of the target object), the irrelevant singleton no longer distracted.
More broadly, in the contingent involuntary orienting hypothesis, Folk et al. [1992]
asserted that distractors sharing task-critical properties with the target (such as singleton
status) will have an effect, while other highly-salient but task-irrelevant distractors (such
as abrupt onsets) will not. Pashler [2001] supported the theory, with the surprising result
that participants searched faster in the presence of irrelevant (flashing, twinkling, and
shimmying) distractors. Prinzmetal et al. [2005], attempting to tease apart endogenous
and exogenous effects, found that irrelevant spatial cues did not affect accuracy, although
the results suggest different attentional mechanisms may control accuracy and response
time performance. Endogenous theories of attention assert that the attractiveness of
certain stimuli may depend on the viewer’s mindset rather than any intrinsic power of
animation.
Though the debate over the relative influence of goal-directed and stimulus-
driven attractors continues, there has been little effort to explore these issues in common
Web tasks. Does a car suddenly appearing within a banner advertisement create a new
perceptual object? If underlined blue text on a Web page is task-relevant to someone
searching for a link, what if the blue text is within a banner ad? Would a lone ad be
considered a singleton on a typical, cluttered Web page?
A Web-like environment has been used in at least two visual search studies. In
Zhang [2000], participants searched and counted random text strings on a Web page with
an animated distractor irrelevant to the task. Distractors included images (i.e., a blinking
eye or a waving robot) and big letter strings that alternately loomed and receded (a
motion shown to capture attention in some situations [Franconeri and Simons 2003] but
not in others [Hillstrom and Yantis 1994]). Zhang found that both kinds of distractors
slowed the primary search task, but the degree depended on the difficulty of the task.
Participants were worse at counting both short and long strings in the presence of
animation, but short-string tasks were more adversely affected by animation than long-
string tasks. The results are difficult to interpret in part because speed and accuracy were
not reported separately, but instead combined into a single metric of “performance,” in
which fast but inaccurate counts were potentially scored the same as slow but accurate
ones. Static versions of the distractors were not tested, so the effect could be due to the
presence of large graphic singletons rather than animation.
In Diaper and Waelend [2000], participants answered questions based on blocks
of text adjacent to animated graphics. Two levels of text length (short, long) and three
versions of graphics (none, static, animated) were tested. Participants also rated the
complexity of the six conditions “at a glance.” Animation did not have an effect on
either search time or perceived complexity. Search times were greater for the longer text
blocks¸ as commonly observed. The study concluded that the amount of text on a page
contributes to task difficulty far more than animation does, but provided no statistically
significant support for this conclusion. The results are difficult to interpret: Participants
were not given incentive to search quickly, and complexity was rated on an unmarked
visual scale. The experiments of Zhang [2000] and Diaper and Waelend [2000] bridge
traditional attention capture and Web research, but methodological issues hinder their
usefulness.
1.2 Animation and Memory
Some studies have explored the memorability of banner ads. Whether animated
ads capture attention in a strict sense, they may imprint some features strongly enough to
achieve subsequent recognition. Bayles [2002] addressed this issue by posting static and
animated versions of two novel banners on a modified Library of Congress Web page.
After four information-gathering tasks, participants were presented with an unanticipated
recall test in which they were asked to draw the layout and contents of the page from
memory. They were also given a page of twelve ads and asked to select the ones they
had seen. No correlation was found between animation and recall, and more than half the
participants did not remember the presence of the banners at all. Animation also did not
affect recognition. One detail in the design of the experiment is that the ten distractor
banners in the recognition task included several that were very similar to the two banners
shown in the information-gathering task. It is not clear whether people failed to
recognize the banners altogether or just specific design details. Furthermore, only two
banners were used in the experiment. Without a larger pool of banner designs, Bayles’s
results cannot be easily generalized.
Pagendarm and Schaumburg [2001] suggest that banners are more memorable to
casual browsers than goal-driven searchers. A group of “aimless browsing” participants
explored a 55-page Web site “according to their own interest,” and a group of “goal-
directed” participants navigated the site to answer a list of questions. Both groups were
exposed to 16 animated ads. Aimless browsers recalled motifs and products in 3 to 17
times as many ads as did goal-directed searchers, but the performance was still very low,
with details recalled from an average of just one or two ads. Participants also rated their
confidence recognizing the banner ads when presented with them again. Confidence was
higher for aimless browsers. However, the test did not include any banners that the
participants had not seen, and thus confounded confidence with recognition.
The two previous studies examined banner recognition as a secondary task
immediately following a primary browsing task but, because of details in the
experimental designs, in both cases it is difficult to conclude whether people retained
knowledge of the ads that appeared during the primary task.
2. EXPERIMENTS
The present experiments test whether standard animated banner ads affect Web
users’ (a) visual search speed, (b) perceived workload, (c) memory, and (d) gaze patterns.
As Pagendarm and Schaumburg [2001] propose that “banner blindness” may
occur especially in a goal-directed task, this is the context in which we focused our
examination of the phenomenon. Zhang [2000] observed that simpler tasks were more
adversely affected by animation, and so we assessed three levels of task complexity. We
also sought to extend Bayles’s [2002] analysis of banner memory; in that study, only 2
different banners were shown during the primary task, so we increased the pool to 100
banners. Lastly, unlike the Diaper and Waelend [2000] study, when we failed to find a
significant effect for search time in our first experiment, we looked further, exploring
other search tasks and distractor types.
In Experiment 1, participants searched for short words while two banners
appeared within the search area. Banners included (a) animated commercial, (b) static
versions of the commercial banners, (c) novel cyan banners that flashed big text, and (d)
blank (invisible) banners. After the timed trials, participants ranked their impressions of
workload for each type of banner. In accordance with the contingent involuntary
orienting hypothesis [Folk et al. 1992], we expected search times to be greatest for the
“big text” banners because their text was similar to the search target. We also predicted
that the animated versions of commercial banners would increase search times more than
their static counterparts. Though irrelevant to the task, the animated banners contained
graphics that appeared abruptly or grew in size, dynamic events that have been shown to
capture attention [Hillstrom and Yantis 1994, Franconeri and Simons 2003]. Finally, we
expected participants to report greater workload under the big text and animated banner
conditions: If these banners do capture attention, they should require more effort to
ignore.
Experiment 2 extended the first study with eye tracking and a more ecologically
valid task: searching for news headlines. Participants performed two kinds of searches:
exact, in which the target headline text was known, and semantic, in which the first few
sentences of a full story appeared and the best-matching headline had to be found. A post
hoc recognition test determined which banners participants remembered having seen.
Consistent with previous findings that animation affects simpler tasks more than harder
ones [Zhang 2000, Diaper and Waelend 2000], we expected animated banners would
prolong both the exact and semantic search tasks, but to different degrees. We also
predicted banner memory would be low, in accordance with Bayles [2002] and
Pagendarm and Schaumburg [2001]. The eye tracking would reveal under what
conditions participants looked at the banners.
3. EXPERIMENT 1
The primary aim of Experiment 1 was to determine whether a simple visual
search task, finding a single word, would be affected by animated ads. Confounds such
as those introduced by reading were removed from the experimental design.
3.1 Method
Participants. Twelve adults (six female) with a mean age of 27 participated in
the experiment for compensation. All were experienced with graphical user interfaces
and had normal or corrected-to-normal vision.
Materials. Each experimental trial presented one target object amid 19
distractors. Targets and distractors were capitalized four-letter words in 18-point
Helvetica enclosed in rectangles with 1-point borders. Roughly 700 words were used.
They were collected from the first two levels of Wired.com, filtered for profanity, and
limited to one or two syllables to facilitate vocal repetition during the search.
Experimental software was written with Macromedia Director.
Two banners appeared among the targets and distractors. Figure 1 presents the
four banner types tested: (a) blank, (b) animated commercial, (c) static commercial, and
(d) flashing text. One hundred ten animated banners were selected from reputable
commercial Web sites including the New York Times, AOL, and Alta Vista. Static
banners consisted of a representative frame from each animated banner. Flashing text
banners were introduced as an extreme variety of animation: Large black text alternately
appeared on the left and right halves of a cyan rectangle every 150 ms. Flashing text
banners cycled asynchronously, offset from each other by 80 ms. Text for these banners
was randomly generated from the target and distractor words for a given trial. Figure 2
shows the screen layout. The target and distractors were arranged in three columns of
eight rows. Banners spanned two columns each, removing four distractors per trial.
Banners emerged in random rows, analogous to the unpredictable placement of ads on the
Web, and targets appeared in all 24 positions across the experiment.
After finishing the timed search trials, participants completed Task Load Index
(TLX) surveys. Developed by NASA [Hart and Staveland 1988], TLX measures
perceived workload, defined by the following factors: mental, physical, and temporal
demand; effort; frustration; and impression of performance. For each banner type,
participants rated these factors from 1 to 100 and then indicated for all possible pairs
which factor contributed more to the overall workload. A combination of these values
would reveal the relative importance of each factor, providing a metric with which
participants’ subjective responses could be compared.
Design. Trials consisted of two stages: precue and search. During the precue
stage, four randomly-ordered words appeared, one of which was the target. After
studying the four words, the participant would click a box, hiding the precue and
initiating the search stage. Search and selection times were separated using the Point
Completion Deadline (PCD) [Hornof 2001]. In short, participants were instructed not to
Fig. 1. The four banner types tested.
use the mouse until they visually located the target, at which point they should quickly
click on it. Upon moving the mouse more than five pixels from its original location,
participants had a limited amount of time, scaled by Fitts’ law, to reach the target. By
performing quickly and accurately, participants could increase their baseline pay of $10
to $15. Each trial had a potential bonus of seven cents with one cent deducted per
second. Clicking the target earned the bonus and a 150 ms chime, but clicking anything
else or exceeding the PCD warranted a five-cent penalty and a 350 ms buzzer.
Each participant completed 96 trials, one for each unique combination of target
location (24) and banner type (4). The trials were divided into four blocks and
randomized, with blocks counterbalanced across participants through a Latin square.
Banner-target combinations in error trials were repeated, shuffled into the remainder of
the block. Thus, participants correctly completed all trials for all combinations.
Procedure. Participants positioned themselves 56 cm from the screen with the
precue at eye level; eye-to-screen distance was reestablished before each block.
Participants were allowed an unlimited number of practice trials from the first block type
to become accustomed to the PCD. When they were ready, the software was reset and
the data collection began. Five additional practice trials initiated every block.
For each trial, participants studied the precue words as long as necessary. Then
they clicked the box to dismiss the precue and display the layout. After visually locating
the target, they selected it quickly; its colors would invert briefly and either the chime or
buzzer would sound.
After the timed trials, participants reported their experiences with the four
banner types through a TLX survey. They completed a short “reminder” block with a
single banner type and then provided TLX weights for the workload factors of that
condition, repeating the process for each banner type. Blocks were again counterbalanced
by a randomized Latin square.
Fig. 2. Layout of target and distractors in Experiment 1.
Table I. Mean Search Times for Each Banner Type in Experiment 1
Banner Type
Search time (ms)
Standard deviation (ms)
Blank
5,831
1,675
Flashing Text
5,234
1,116
Animated Commercial
4,795
1,010
Static Commercial
5,155
1,238
Note. Times are averaged across target positions and participants.
After completing the TLX evaluations, participants were interviewed and asked to
describe their overall impressions and search strategies.
3.2 Results
Overall Workload. Participants reported tasks with flashing text banners to have
the greatest perceived workload. A repeated measures ANOVA revealed a significant
difference between banner types, F(3, 36) = 6.52, p < .001, and a Fisher’s PLSD post hoc
test showed the flashing text banners were perceived as more workload-intensive than the
others (Mean = 160.71). Animated banners also had greater overall workload, though
non-significantly (M = 121.44). The static and blank banners ranked approximately the
same (M = 107.83 and 110.54, respectively).
Workload Factors. Participants perceived the flashing banners to be more
frustrating and mentally demanding than the other banner types, F(3,32) = 3.50 and 1.62,
respectively, p < .05. Marginally greater temporal demand and effort were reported, as
well, p < 0.1. Figure 3 shows how each of the six TLX factors contributed to the overall
workload by banner type.
Search Times and Error Rates. Table I shows the mean search time for each
banner type. Error and practice trials were excluded from analysis, and were not
significantly different across banner types. Results from a repeated measures ANOVA
fail to show a significant difference in search time across banner types. A position effect
was observed as expected: Participants found targets in the upper-left positions much
faster than those in the lower-right, F(23,72) = 5.08, p < .0001. One significant effect
was that when a target was sandwiched between two flashing test banners it took an
average of 75% (4.3 seconds) longer to find than if the banners were located elsewhere,
F(1, 298) = 7.0, p < .01.
3.3 Discussion
Workload. TLX measures participants’ conscious impressions of task workload,
though not necessarily the underlying mechanisms that control visual attention.
Participants reported that searching in the presence of the flashing text banners was the
most frustrating and mentally demanding (two of the six factors measured by TLX).
However, the same results were not observed for the commercial animated banners,
suggesting several possibilities. Comparing the two banner types, participants may have
felt the animation of the commercial banners to be more subdued, and thus no more
workload-intensive than the static banners. Or, the words in the flashing banners may
have been similar enough to the targets to impose a cognitive burden, as predicted by the
contingent involuntary orienting hypothesis [Folk et al. 1992]. Perhaps participants
believed themselves capable of ignoring typical advertisements, whether animated or
static, consistent with the findings of Benway and Lane [1998] and Bayles [2002].
Animated commercial banners were rated slightly higher than static banners for all
workload factors, suggesting that for another task or with a different set of commercial
banners, a significant effect might be observed.
Participant Comments. Participants discussed their search strategies and overall
impressions in the post-experiment interview. In general, they reported being able to
“tune out” the banners, although some found the flashing text and brightly-colored ones
difficult to ignore. Many participants explained that the layouts with blank banners were
the easiest to search, but others preferred the presence of banners because they helped to
divide the screen into smaller search regions, and thus helped to structure the search
space. We addressed this in Experiment 2 by replacing blank (invisible) banners with
gray boxes.
Search Times and Error Rates. Participants found targets just as quickly in the
presence of animated banners as in all other conditions. This suggests that banner
animation does not necessarily capture attention in a relatively simple visual search task.
However, like the goal-driven participants in Pagendarm and Schaumburg’s [2001]
experiment, our participants also had an incentive not to look at the banners—the bonus
pay decreased every second. Like Diaper and Waelend [2001] and Zhang [2001], we
propose that animation’s power to distract is dependent on the nature of the task. Thus,
in Experiment 2, we introduced a more challenging and ecologically realistic task, hoping
to elicit more conclusive search-time results.
4. EXPERIMENT 2
Experiment 1 suggested that animated banners might distract people performing
visual search tasks, but might not necessarily slow them down. Previous studies [Zhang
2000, Diaper and Waelend 2002] indicated that the nature of the search task influences
the strength of the distraction, and so we introduced a different task in the second
experiment—searching through news headlines that appeared as links on a web page. To
ensure ecological validity, we modeled the layout and format of links after several news
sites, including CNN, Google News, and Yahoo! News.
Pilot tests for Experiment 2 revealed that many participants had difficulty with
the point-completion deadline (PCD) on this task. Several participants wanted to use the
mouse as a visual placeholder as they read, but the PCD would not allow it. Timeout
errors were frequent. Researchers (such as Sears and Shneiderman [1994]) posit that
people use the mouse as a visual placeholder when searching menus. Brumby and
Howes [2004] observed this behavior, as well. Though participants do require a number
of trials up front to grow accustomed to not moving the mouse before finding the target,
the PCD seems to work fine, and participants have low timeout errors, in tasks that are
closer to a laboratory visual search task, such as finding single words in Experiment 1.
However, in the current experiment, participants did not grow accustomed to the PCD,
and timeout errors remained high. This suggests that, in a real-world setting, moving the
mouse may be an integral part of the visual search. More empirical work is needed to
explore this possibility. We removed the PCD, letting participants move the mouse
freely, as they would on the Web.
To determine whether participants actually looked at the banners, we added eye
tracking to collect fixation data and examine common search patterns.
4.1 Method
Participants. Twenty-four undergraduate students from Lewis & Clark College
(sixteen female) with a mean age of 21 participated in the experiment for compensation.
All participants were experienced with graphical user interfaces and had normal or
corrected-to-normal vision.
Materials. Three hundred twenty-nine headlines were gathered from humorous
news sites like CNN’s Offbeat News (http://www.cnn.com/offbeat) from April to
September 2003. All were displayed in underlined, blue, 12-point Arial, a common
format for Web links. One hundred twenty-four of the headlines were used as targets; the
remaining headlines were distractors. Target headlines appeared once per participant;
distractor headlines appeared up to three times.
The animated and static commercial banner types from Experiment 1 were
reused in Experiment 2. The flashing text banners were removed for their lack of
ecological validity. Gray rectangles were used in place of blank banners because of
participants’ previous suggestions that the banners helped to partition the search space.
Equipment. Eye movements were recorded using the LC Technologies Eyegaze
system. During data collection, participants used a chin rest to keep their heads relatively
still. A small, unobtrusive camera was mounted below the computer monitor. Two
separate computers were used in the data collection, one to collect the gaze position
(represented by a small yellow plus sign on the screen) and one to run the experimental
software. Both computers were connected to the same monitor with a two-way switch.
Participants viewed the output from the Eyegaze computer during a short calibration.
When they performed the experimental tasks, the monitor was switched to display the
output from the computer running the experimental software. The software used to
display the stimuli was directly derived from that used in Experiment 1.
The video signals from both computers were sent to a digital video mixer
(Videonics model MX-1) where the plus sign showing the participant’s gaze point was
superimposed over the screen from the computer running the experiment using a chroma
key effect. This composite image was then recorded to digital video and later transferred
to Quicktime format for data analysis.
Design. To manipulate the mental workload of the search task, two precue
conditions were used. In the exact precue condition, the precue contained the text of the
target headline, word for word. For example, both the precue and target headline might
be “Drop-outs doing just fine, thanks.” In the semantic precue condition, a sentence or
two from the beginning of the news article was used. If key content words in the
headline appeared in the semantic precue, synonyms found in the article were substituted.
For example, the semantic precue for “Drop-outs doing just fine, thanks” was as follows:
New research debunks the common belief that leaving school before completing
year 12 diminishes a teenager’s chance of a successful career.
In the semantic precue condition, participants could not merely look for a
keyword in the target headline. Instead they had to read the headlines and compare them
to the precue to find sufficient overlap in meaning to make the match. All precues were
written in black, non-underlined, 14-point Arial so that they would have a slightly
different overall appearance than the target and distractor headlines. This difference in
font size and color would prevent participants in the exact precue condition from simply
matching letter shapes.
Figure 4 shows the screen layout. Target and distractor headlines were arranged
in two columns of six rows each. Each trial contained two banners. One always
appeared at the top of the screen, directly above the area where the headlines were
displayed. This location was selected to ensure that a participant’s gaze would pass over
a banner on every trial, and to mimic a common position of banner ads on the Web. The
second banner was randomly placed in one of the six rows of the headline search area,
spanning both columns. For each trial, both banners were of the same type (static,
animated, or gray). Participants never saw both the animated and static versions of the
same commercial banner.
Two blocks of trials were presented in a counterbalanced order. One block was
the exact match condition; the other was semantic match. Each block consisted of 5
practice trials followed by 36 data collection trials, 12 trials each containing animated,
static, or gray banner ads. The target headline appeared in a different position for each of
the 12 trials. The type of banner presented was randomized across trials and within
blocks. Banner and target combinations in error trials were repeated, shuffled into the
remainder of the block.
Procedure. Search trials proceeded in the same manner as in Experiment 1.
Participants studied the precue for an unlimited period and when ready, clicked a box to
make the precue disappear and the search area appear. Unlike in Experiment 1,
participants were allowed to move their mouse while searching. The cursor changed
from an arrow to a hand over the headlines, as it would for links on the Web. After
locating the target, participants selected it, at which point the headline would briefly turn
magenta and either the reward chime or penalty buzzer would sound. The payoff matrix
was the same as in Experiment 1, except that exact search trials started with a potential
bonus of 9 cents and semantic search trials with a bonus of 14 cents. One cent was
deducted per second, and errors imposed a five-cent penalty.
After the visual search tasks, participants were given a short break and then
asked to view and identify banners that were shown in the study. This was the first
mention of the banners to the participants. It was explained that they would see some
banners that had been in the study and others that had not. The banners were displayed
on the screen one at a time, and participants responded by clicking a “yes” or “no” button
at the bottom of the screen. Each click triggered the presentation of the next banner. A
total of 60 banners were presented (30 animated and 30 static). Of these, 40 banners had
appeared during the visual search tasks and 20 had not. Participants were not given
feedback on accuracy for this memory task and speed was not recorded or emphasized.
4.2 Results
Search time. The type of precue (exact vs. semantic) produced the strongest
effect in the experiment. Search times for the exact precue condition (M = 2134 ms.,
Standard Deviation = 299 ms.) were much faster and less variable than for the semantic
precue condition (M = 6129 ms, SD = 1567 ms.), F(1, 23) = 231, p < .0001. Due to
overwhelming differences and the unequal variance in the search times for these precue
conditions, the remaining search time analyses are broken down by precue condition.
Fig. 4. Screen layout for a literal precue trial with a zoomed-out
view of three headlines. The precue at the top disappeared
when the layout appeared.
Table II. Mean Search Times for Each Banner Type in Experiment 2
Banner type
Search time (ms)
Standard deviation (ms)
Exact precue
Gray
2,040
289
Static
2,169
300
Animated
2,193
297
Semantic precue
Gray
6,065
1,614
Static
6,210
1,736
Animated
6,110
1,397
Note. Times are averaged across target positions and participants.
Table II shows the mean search time for each banner type. For the exact precue
condition, a repeated measures ANOVA revealed a significant difference among the
banner types, F(2,46) = 5.5, p < .007. Post hoc paired t tests showed that both the static
and animated banners resulted in slower search times compared to the gray banners (p <
.005 for both comparisons), but equivalent search times when compared to each other (p
= .65). Static banners were 6.3% slower than the gray; animated, 7.5%.
Significant differences in search time as a function of banner type do not persist
in the semantic precue condition. A repeated measures ANOVA revealed no significant
differences, F(2,36) = 0.18, p = .83. While the general pattern is that banner ads appear
to have slowed the search process even in the semantic precue condition, the high
between-subject variability overwhelms the statistical significance of the difference.
The error rates in this experiment were uniformly low in both precue conditions
(4.6%). There was no significant correlation between speed and accuracy, r = -.103, p =
.63.
Memory. Participants responded “yes” or “no” to each ad during the recognition
memory test. A “yes” to a banner that did appear in the experiment is a hit and to a
banner that did not appear is a false alarm. The number of “no” responses to banners that
did appear (misses) and that did not (correct rejections) can be derived from the number
of hits and false alarms.
The hit and false alarm rates are used to assess the participants’ memory for the
banners. Overall, memory for the banner ads was poor, with a hit rate of only 20.1% and
a corresponding false alarm rate of 20.2%. Perfect performance would have been 100%
and 0%, respectively. The hit rates did not differ by precue condition (exact = 20.0%,
semantic = 20.2%), χ2(1, N = 24) = .008, p = .94. Though the increased mental workload
of the semantic precue search task greatly increased visual search time, it did not affect
memory for the banner ads.
Though recognition memory for the banners was low overall, signal detection
theory can be used to show that recognition memory was better for the static banners.
The hit and false alarm rates were transformed into a single measure of memory strength
known as d [Green and Swets 1966]. A positive, non-zero d value is an indication of
memory strength, controlling for guessing behavior and decision strategies that
participants might adopt. One-group t tests showed that the d value for the static banners
(M = 0.67) is significantly higher than zero, t(23) = 2.66, p = .01, while the d for the
animated banners (M = -.07) is not, t(23) = -0.3, p = .77. A paired t test further revealed
that the d score for the static banners is significantly higher than for the animated
banners, t(23) = 2.14, p = .04. This shows that when we correct for their guessing
strategies, the participants have significantly worse memory for the animated banners
than for the static.
Memory for banner ads was examined as a function of the location of the ad on
the screen. For each search trial, one ad was placed in the same location at the top of the
screen, between the precue and the topmost headlines. The second ad was randomly
placed in one of the six rows of the headline search area, spanning both columns. By
combining the hit-rate data across all 24 subjects we are able to determine that screen
location affected the recognition memory for these banners. Overall, there was a trend
for the top banner to be better remembered, χ2(1, N = 24) = 2.83, p = .09. Breaking the
data down by banner types indicated that the top ad was remembered significantly better
than the randomly-placed ad for the static banners (p = .01), but not for the
animated banners (p = .82). Note that the d analysis is not used here because too many
participants had hit and false alarm rates of 0 in this combination of factors, thus making
d undefined.
Eye tracking. A digital video composite was created by superimposing the
screen output from the two computers used in the experiment: one that collected gaze
position (represented by a small plus sign) and another that presented the experimental
software. Fixations were encoded by watching the video. For each trial, it was noted
whether the first saccade (rapid, ballistic eye movement) from the precue occurred before
or after the appearance of the banners and headlines (hereafter collectively called
“items”). Each time the gaze landed on a new item, the item number was recorded.
Multiple saccades within a single item, such as those necessary to read a headline, were
not counted. Revisits to an item were counted, as long as the gaze moved to another item
and then returned.
Participants looked at nearly twice as many items in the semantic condition (M
= 8.35) as in the exact condition (M = 4.92), F(1, 16) = 120.86, p < .0001. Across
blocks, the number of items viewed was not affected by banner type, F(2,32) = 1.11, p =
.34, although in the exact precue block, a marginally significant difference was observed,
F(2,34) = 2.96, p = .07. The mean gaze time per item (calculated by dividing the search
time by the number of items fixated in a trial) was also greater in the semantic condition
(M = 786.73) than the exact condition (M = 452.14).
There was a strong positive correlation between the number of items viewed and
the search time, r = .845, p < .0001. Table III shows the correlations for each
combination of banner type and precue type. One exception was noted: There was no
correlation in the semantic precue condition when animated banners were present.
Participants looked at banners in 11.7% of the trials, regardless of banner type,
F(2,32) = 1.28, p = .29. They looked at gray banners (M = 8.8%) almost as often as the
static (M = 13.0%) and animated (M = 13.2%) ones. Of the trials in which participants
looked directly at banners, 70% of the banner gazes occurred during the participant’s first
eye movement. The precue type (exact versus semantic) did not affect whether
participants looked at banners, F(1,16) = .137, p = .72. Of the 164 banners that
participants correctly remembered in the memory test (hits), only 10 received direct gazes
during the search portion of the experiment. Thus, participants did not directly look at
93.9% of the banners they “remembered.”
Table IV shows how precue type and banner position affected whether
participants looked at banners. Banners that appeared in a fixed location between the
precue and topmost headlines (top banners) were viewed less often than banners that
were positioned in a random row within the search area (inside banners), F(1,16) = 6.73,
Table III. Correlation Between Number of Items Viewed and Search Time
Banner Type
Correlation
Exact precue
Gray
.749**
Static
.536*
Animated
.790**
Semantic precue
Gray
.702**
Static
.549*
Animated
.372
*p < .05 ** p < .001.
Table IV. Number of Trials with Direct Fixations on Top and Inside Banners
Top banner
Inside banner
Precue type
Total
On first fixation
Total
On first fixation
Exact
10
6
58
47
Semantic
36
25
38
21
p = .02. In the exact condition, participants rarely looked at the top banner but did look
at the inside banners. In the semantic condition, gazes at the top and inside banners were
approximately equal.
Participants frequently made anticipatory fixations: eye movements from the
precue to the search area before the headlines and banners appeared. Anticipatory
fixations occurred in approximately 40% of the trials, irrespective of precue type (M =
40.4% for semantic and 42% for exact), p = .76. Of the trials in which participants
looked directly at banners on their first eye movement, 54% were anticipatory gazes, in
which case the banners appeared after participants had moved their eyes to the location.
Regardless of whether the eye movement away from the precue occurred before
or after the onset of the layout, the gaze was almost always near the top of the screen
when the layout appeared. In 73% of all trials, at the moment that the layout (and thus
the top banner) appeared, the gaze was either on the precue, on the top banner position, or
in position for the top row headlines.
4.3 Discussion
Search time. Consistent with the findings of Zhang [2000] and Diaper and
Waelend [2000], the nature of the task influenced the distracting effect of the banner ads.
When searching for the simpler exact match, participants were more adversely affected
than when searching for the harder semantic match. This can be seen in the significant
increase in search times in the exact match condition when commercial banners were
present, and the non-significant difference in search times in the semantic match
condition. One possible explanation is that simple searches require less attention, and
thus other stimuli can be perceived and processed, slightly delaying the primary task. It
may be that difficult searches necessitate greater focus, leaving less processing available
for irrelevant objects, which are thus ignored. Or, the semantically-driven search may be
more controlled, so that participants were able to avoid banners. The high between-
subject variability in the semantic condition suggests an alternative methodological
explanation: By their very nature, difficult tasks result in a wider range of response
times, making the banners’ effect harder to detect.
Contrary to the prevailing opinion of designers, animated commercial banners
affected performance no differently than static commercial banners. Search times were
slower for the static and animated banners than for the gray banners, but were equivalent
between static and animated. Given that the animated banners had graphics roughly
identical to their static versions, attention capture could be attributed to the images,
colors, and text within the banners, rather than their motion. Even in the semantic
condition, where participants were exposed to an average of six seconds of animation per
trial, the animation did not affect search times. In fact, though the difference was not
significant, searches tended to be slightly faster with animated banners than static in the
semantic condition, loosely analogous to the findings of Pashler [2001].
Participants looked at twice as many items per trial when searching for a
semantic match than when searching for an exact headline. This result is consistent with
the nature of the task. In the semantic condition, the best match could rarely be
determined in isolation; a potential target had to be compared to the remaining headlines.
Return visits to a headline were common. Additionally, gaze time per item was greater in
the semantic condition, as participants considered the content of each headline.
The strong positive correlation between search time and the number of items
viewed is expected, but the lack of such a correlation when animated banners were
present in the semantic condition is surprising. Some other factor must influence these
variables. Perhaps participants consider headlines for a much wider range of times. This
is speculation, and one of many possible explanations, but perhaps the animation
increases this variance.
Fixations on banners. Graphics and animation in the commercial banners did
not attract participants’ gazes: Direct fixations occurred on the gray banners as often as
the commercial ones. Though the static and animated banners did increase search time,
the increase cannot be attributed to participants looking directly at the banners and thus
processing their detailed content. Instead, the delay might be caused by graphics and
animation viewed peripherally. This is consistent with Prinzmetal et al. [2005], in which
attention is allocated to a region away from the current gaze point.
Participants’ initial eye movements further suggest conscious efforts to avoid
banners. The majority (70%) of fixations on banners occurred in the first eye movement.
Half of these (54%) happened when the eyes moved to a location before the banner
appeared there. A maximally efficient search strategy would avoid banners altogether,
and it is possible that at the time of the first eye movement, participants had not yet
encoded the locations of the banners, and thus, any banner fixations were accidental.
Participants looked at the top banner roughly one-third (27.8%) as often in the exact
condition as they did in the semantic condition, which supports the idea of improved
banner avoidance in that condition. However, participants also spent about one-third as
long on the exact trials as they did on the semantic trials, suggesting that top banner
fixations simply increased proportionately with time on task.
Just as the nature of the search task (exact versus semantic) influenced the
degree to which different banner types delayed search, it also influenced which banner
positions were fixated. In the semantic condition, top banner fixations occurred as often
as inside banner fixations. Perhaps the semantic processing assisted in the formation of a
visual search strategy that focused on the headlines. This is further supported by the
substantially fewer fixations on inside banners for semantic condition (38 compared to 58
for exact) which is particularly surprising considering the increased time on task in the
semantic condition. Perhaps participants were better inured to the inside banners in the
semantic task.
Memory. The simplest explanation for participants remembering certain banners
would be that they looked directly at those banners. However, the eye-tracking data do
not support this explanation in that 94% of the banners correctly identified in the
recognition test had not been fixated by the participant. Perhaps peripherally salient
graphical elements aided memory, but the high false alarm rates refute this interpretation.
The low memory rate is consistent with Bayles [2002] and Pagendarm and Schaumburg
[2001]: People simply do not remember banners that are irrelevant to their goals.
Recognition also did not vary by length of exposure. Banners in the semantic condition
appeared three times longer than those in the exact condition, but participants did not
remember them any better.
Static banners fared somewhat better than animated banners in the recognition
test. Neither Bayles [2002] nor Pagendarm and Schaumburg [2001] had compared the
memorability of static to animated banners, but advertisers employ animation so often
that we were surprised to find it to be an ineffective memory aid. Though neither type of
banner was very memorable, participants did correctly identify a small percentage of
static banners. Animated banners were recognized no better than chance. In part, this
may be due to the changing appearances of these banners: A slogan or logo that
disappeared or moved may not have been visible or obvious during the brief period in
which participants processed the banners. This is not to say that participants were not
exposed to a full cycle of animation in each trial, but simply that they did not pay
attention to the entire cycle. They may have observed too few details to recognize the
banners later. Static banners, in which the message appeared continuously throughout the
trial, were more memorable.
Banners positioned at the top of the screen were remembered more often than
the inside banners. This is surprising, given that participants looked directly at the top
banners less frequently. However, as was mentioned, the gaze was usually near the top
of the screen when the layout and thus the top banner appeared. Perhaps attention was
slightly drawn to the onset of the top banner and, since that banner was relatively close to
the gaze at that time, a small amount of visual information was processed.
Like the goal-directed searchers in Pagendarm and Schaumburg [2001], the
participants in the present experiment had little incentive to look at the banners.
Nonetheless, the commercial banners slowed visual search responses. In trials where the
task was easy enough to allow participants to formulate a search strategy, they
intentionally avoided looking in places where banners were known to appear. On Web
sites where banner ads are unrelated to page content and viewers’ goals, the same results
may be expected: Ads will increase visual search times even though people will avoid
looking at them.
5. GENERAL DISCUSSION
The popular notion of “banner blindness” suggests that people just ignore
banner ads. Nonetheless, may Web users still dislike them. Motivated by the seeming
contradiction between “banner blindness” and Web users’ distaste of “distracting” ads,
we discovered results consistent with both schools of thought. Typical commercial
banner ads hinder searches through lists of links, even if the searchers never look directly
at the banners. Given the prevalence of animation on the Web, it was surprising to find
little disparity between the effects of animated and static versions of the same ads. We
had hypothesized that search times and perceived workload for animated commercial
banners would be greater than those for static banners. However, people do not look at
animated banners more often than static ones, and they can search equally quickly under
both. Extreme forms of animation, such as the flashing cyan banners from Experiment 1,
do increase frustration and perceived mental demand, but mainstream animated banners
performed no worse than their static equivalents, at least when compared head-to-head
with extreme flashing banners. The one surprising difference between animated and
static banners was that animation makes ads less memorable. Contrary to widespread
practice, animation may not provide a benefit to advertisers; in fact, it may even cost an
ad’s success.
In this task, there is evidence of both goal-directed (endogenous) attentional
control and stimulus-driven (exogenous) attentional capture. Though people rarely
looked at the banner ads, when they did it was independent of banner type. This suggests
that participants adopted search strategies that enabled them to ignore irrelevant
distractors, consistent with the notion put forth by Folk et al. [1992] and others that
people adopt task strategies that prevent involuntary shifts of attention to at least some
stimulus properties.
Nonetheless, there is some evidence for stimulus-driven capture. In Experiment
2, participants searched longer in the presence of commercial banners than in the
presence of gray rectangles. This suggests that some graphical elements may have
captured attention regardless of the participants’ strategies. Interestingly, it is unlikely
that the stimulus-driven attentional capture was due to animation, as animated banners
did not have greater search times than static banners.
Unlike the previous studies of animated distractors on the Web [Bayles 2002,
Diaper and Waelend 2000, Pagendarm and Schaumburg 2001, Zhang 2000], the present
experiments include eye tracking data that reveal the underlying behavior of people
searching in the presence of the distractors. People rarely looked directly at banners, and
adding graphics did not appear to matter. Gray banner “placeholders” were fixated as
frequently as commercially-designed banners. In fact, most banner fixations in
Experiment 2 occurred on participants’ first eye movements, perhaps before they encoded
the banner locations so as to avoid them. The infrequent fixations on top banners in
Experiment 2 and the “sandwich effect” from Experiment 1 (in which targets sandwiched
between two rows of flashing banners took longer to find) further support the idea that
people intentionally avoid looking in locations where they expect banners, at least once
the search has begun.
Clearly, the nature of the primary task strongly interacts with the attention-
getting capacity of the banners. Across the two experiments, three kinds of simple visual
search tasks with only subtle differences between them were tested, and they resulted in
markedly different outcomes. The negative effects of banner ads are subtle and not
always easy to directly measure. Experiment 1 may not have been ecologically realistic
enough to elicit a search-time disparity. The semantic condition in Experiment 2 led to
high between-subjects variability, again potentially concealing a time effect. Diaper and
Waelend [2000] present yet another study in which the effect could not be detected.
Even in a highly controlled environment, the interaction between banner factors, such as
color and semantic appeal, and task factors, such as participants’ reading abilities, make
quantifiable results difficult to detect.
The negative effects discussed here apply directly to experienced Web searchers
who know exactly—or nearly exactly—for what they are looking. To them, banners are
irrelevant. Many Web surfers fall into Pagendarm and Schaumburg’s [2001] “aimless
browser” demographic. The effect of banner ads on this population is still unclear.
Practical implications. Web designers and site owners should post ads closely
related to page content if they hope to attract their viewers’ attention. Participants in the
present studies had an overriding incentive not to look at banners, and no amount of
banner manipulation increased their pull. Longer exposure time, animation, and the
presence of images did not make the task-irrelevant ads more conspicuous. Connecting
advertising to viewers’ goals may make ads more successful; Yahoo! received positive
feedback when it deployed ads related to page content (see Rohrer and Boyd [2004] for a
discussion of user experience and advertising).
Banners positioned at the top of the screen may be more memorable, although
this effect could be due to the specific screen layout of the present experiments. Top
banners were favorably situated between the precue and the content to be searched.
Browser address bars and standard site navigation areas appear roughly in the same
screen region as our precue; perhaps Web users’ eyes might follow paths similar to those
observed experimentally. Designers should be wary, nevertheless, of habituating viewers
to predictable banner locations: People avoid looking in areas where they expect to find
ads. People’s success in avoiding banners may be dependent on the cognitive complexity
of their tasks; top banner avoidance was only clearly observed in the exact precue
condition of Experiment 2, but the Web presents numerous tasks of varying complexity.
To lessen the spread of banner blindness to critical page elements, usability guidebooks
(i.e., Nielsen and Tahir [2002]) advise against placing site navigation above banner ads.
The “sandwich effect” from Experiment 1 supports this advice.
Further investigation is needed into all aspects of visual search on the Web.
Traditional attention-capture studies, such as those discussed in Section 1.1, explain some
search behaviors, but the myriad of interacting Web factors should be explored in a more
ecologically valid context. Additional studies of animated distractors are needed for
more involved tasks, such as multi-page surfing and form-filling. Though the results are
too premature to report here, we noticed dramatically different gaze paths over the same
headline layouts in Experiment 2, depending on whether the precue had been exact or
semantic. That people may scan identical screens differently based on mental load has
implications for cognitive modelers, especially those seeking to predict eye movements.
Future empirical studies into other tasks may lead to a more comprehensive
understanding of visual search on the Web. In the present experiments, the implications
are clear: Banner ads degrade visual search and are quickly forgotten.
ACKNOWLEDGMENTS
We would like to thank Robin Jeffries, our colleagues at the University of Oregon and
Lewis & Clark College, and the three anonymous reviewers for their feedback.
REFERENCES
ABRAMS, R.A., MEYER, D.E., AND KORNBLUM, S. 1990. Eye-hand coordination: Oculomotor control in
rapid aimed limb movements. Journal of Experimental Psychology: Human Perception and Performance 16,
248-267.
BACON, W.F., AND EGETH, H.E. 1994. Overriding stimulus-driven attentional capture. Perception &
Psychophysics 55, 485-496.
BAYLES, M. 2002. Designing online banner advertisements: Should we animate? In Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems, Minneapolis, MN, April 2002, ACM Press,
New York, 363-368.
BENWAY, J.P., AND LANE, D.M. 1998. Banner blindness: Web searchers often miss “obvious” links.
Internetworking 1.
BRUMBY, D.P., AND HOWES, A. 2004. Good enough but I’ll just check: Web-page search as attentional
refocusing. In Proceedings of the Sixth International Conference on Cognitive Modeling, Pittsburgh, PA,
August 2004, 46-51.
DIAPER, D. AND WAELEND, P. 2000. World Wide Web working whilst ignoring graphics: Good news for
Web page designers. Interacting with Computers 13, 163-181.
ENNS, J.T., AUSTEN, E.L., DI LOLLO, V., RAU SCHENBERGER, R., AND YANTIS, S. 2001. New objects
dominate luminance transients in setting attentional priority. Journal of Experimental Psychology: Human
Perception and Performance 27, 1287-1302.
FOGG, B.J., MARSHALL, J., LARAKI O., OSIPOVICH, A., VA RMA, C., FANG, N., PAUL, J.,
RANGNEKAR, A., SHON, J., SWANI, P., AND TREINEN, M. 2001. What makes Web sites credible? A
report on a large quantitative study. In Proceedings of the SIGCHI Conference on Human Factors in
Computing Systems, Seattle, WA, April 2001, ACM Press, New York, 61-68.
FOLK, C.L., AND REMINGTON, R.W. 1998. Selectivity in distraction by irrelevant featural singletons:
Evidence for two forms of attentional capture. Journal of Experimental Psychology: Human Perception and
Performance 24, 847-858.
FOLK, C.L., REMINGTON, R.W., AND JOHNSTON, J.C. 1992. Involuntary covert orienting is contingent
on attentional control settings. Journal of Experimental Psychology: Human Perception and Performance 18,
1030-1044.
FRANCONERI, S.L., AND SIMONS, D., J. 2003. Moving and looming stimuli capture attention. Perception
& Psychophysics 65, 999-1010.
GREEN, D.M., AND SWETS, J.A. 1966. Signal Detection Theory and Psychophysics. Wiley, New York.
HART, S., AND STAVELAND, L.E. 1988. Development of NASA-TLX (task load index): Results of
empirical and theoretical research. In Human Mental Workload, P.A. HANCOCK AND N. MESHKATI, Eds.
Elsevier Science Publishers B.V., Amsterdam, 139-183.
HILLSTROM, A.P., AND YANTIS, S. 1994. Visual motion and attentional capture. Perception &
Psychophysics 55, 399-411.
HORNOF, A.J. 2001. Visual search and mouse-pointing in labeled versus unlabeled two-dimensional visual
hierarchies. ACM Transactions on Computer-Human Interaction, 8, 171-197.
INTERACTIVE ADVERTISING BUREAU. 2001. IAB/Dynamic Logic ad unit effectiveness study.
Available from http://www.iab.net/resources/dynamic_lbrandrl.asp
JONIDES, J. AND YANTIS, S. 1988. Uniqueness of abrupt visual onset as an attention-capturing property.
Perception & Psychophysics 43, 346-354.
NIELSEN, J. AND TAHIR, M. 2001. Homepage Usability: 50 Websites Deconstructed. New Riders,
Indianapolis, IN.
PAGENDARM, M. AND SCHAUMBURG, H. 2001. Why are users banner-blind? The impact of navigation
style on the perception of Web banners. Journal of Digital Information 2.
PASHLER, H. 2001. Involuntary orienting to flashing distractors in delayed search? In Attraction,
Distraction, and Action: Multiple Perspectives on Attentional Capture, C. FOLK AND B. GIBSON, Eds.
Elsevier Science B.V., Amsterdam, 77-92.
POST, R.B., WELCH, R.B., AND BRIDGEMAN, B. 2003. Perception and action: Two modes of processing
visual information. In Visual perception: The influence of H. W. Leibowitz, J. ANDRE AND D.A. OWENS,
Eds. American Psychological Association, Washington, DC, 143-154.
PRINZMETAL, W., McCOOL C., AND PARK, S. 2005. Attention: Reaction time and accuracy reveal
different mechanisms. Journal of Experimental Psychology: General 134, 73-92.
REMINGTON, R.W., FOLK, C.L., AND McLEAN, J.P. 2001. Contingent attentional capture or delayed
allocation of attention? Perception & Psychophysics 63, 298-307.
REMINGTON, R.W., JOHNSTON, J .C., AND YANTIS, S. 1992. Involuntary attentional capture by abrupt
onsets. Perception & Psychophysics 51, 279-290.
ROHRER, C. AND BOYD, J. 2004. The rise of intrusive online advertising and the response of user
experience research at Yahoo!. In Proceedings of the SIGCHI Conference on Human Factors in Computing
Systems, Vienna, April 2004, ACM Press, New York, 1085-1086.
ROWE, N.C., COFFMAN, J., DEGIRMENCI, Y., HALL, S., LEE, S., AND WILLIAMS, C. Automatic
removal of advertising from Web-page display. In Proceedings of the Second ACM/IEEE-CS Joint Conference
on Digital Libraries, July 2002, Portland, OR, 406.
SCHUMACHER, E.H., LAUBER, E.J., GLASS, J.M., ZURBRIGGEN, E.L., GMEINDL, L., KIERAS, D.E.,
MEYER, D.E. 1999. Concurrent response-selection processes in dual-task performance: Evidence for adaptive
executive control of task scheduling. Journal of Experimental Psychology: Human Perception & Performance
25, 791-814.
SEARS, A . AND SHNEIDERMAN, B. 1994. Split menus: Effectively using selection frequency to organize
menus. ACM Transactions on Computer-Human Interaction 1, 27-51.
THEEUWES, J. 1991. Cross-dimensional perceptual selectivity. Perception & Psychophysics 50, 184-193.
YANTIS, S. 2000. Goal-directed and stimulus-driven determinants of attentional control. Attention and
Performance 18, 73-103.
YANTIS, S. AND 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. AND JONIDES, J. 1990. Abrupt visual onsets and selective attention: voluntary versus automatic
allocation. Journal of Experimental Psychology: Human Perception and Performance 16, 121-134.
ZHANG, P. 2000. The effects of animation on information seeking performance on the World Wide Web:
securing attention or interfering with primary tasks?. Journal of the Association for Information Systems 1, 1-
28.
Receiv ed September 2004; revised May 2005; acc epted June 2005
... Moreover, although online advertisements have the advantage to embed such unexpected salient features as motion and abrupt onset to compete for consumers' attention, these kinds of digital tools could be intrusive and disturbing, leading to high workload and negative attitudes (Burke, Hornof, Nilsen, and Gorman, 2005;Diaper and Waelend, 2000;Gao, Koufaris, and Ducoffe, 2004). To discover the possible effect of digital disruption, the authors also investigated the effect of unexpected advertisements on consumers' attitude toward the advertising formats and perceived intrusiveness Li, Edwards, and Lee, 2002). ...
... Prior studies have shown that online advertisements are more likely to be noticed when consumers perform simple-rather than difficult-tasks (Burke et al., 2005;Diaper and Waelend, 2000;Simola et al., 2005Simola et al., , 2011Zhang, 2000). To that end, the authors compared consumers' attention toward unexpected advertisements during both free browsing (simple task) and reading for comprehension (difficult task). ...
... The effects of online advertisements on consumers' attention depends on the difficulty of the primary task. Commonly, prior studies have demonstrated greater impact with simple and less demanding tasks than with difficult, more demanding tasks, with findings reported for visual attention (e.g., number of fixations, dwell time) and self-reported questions (e.g., recall, comprehension accuracy) (Burke et al., 2005;Hong, Thong, and Tam, 2007;Lai, Kuan, Hui, and Liu, 2009;Simola et al., 2011). ...
... Accordingly, a recall advantage for emotionally-arousing ad-viewing experience with the presence of 3D animated agents seems to be expected. However, the study of Burke, Hornof, Nilsen, and Gorman (2005) compared animated web ads to static look-alikes and found that animated ads are more difficult to remember. Hong, Thong, and Tam (2004) reported that flashed animation on a website may cause a lower recall. ...
... Nevertheless, unlike web animation ads that usually take up most of the screen, interrupt the audiences' ongoing webpage reading or information searching task and split their attention (Burke, Hornof, Nilsen, & Gorman, 2005;Rohrer & Boyd, 2004;Yoo, Kim, & Stout, 2004), in a VR context, the ads and the animated 3D models can appear while the audiences are waiting for the contents to load. The cognitive load of the audience is relatively low during the waiting time, and the audience may even feel bored and await being aroused. ...
... Considering the positions and the order of the ad and animated agents appearing in the J o u r n a l P r e -p r o o f virtual environment in Study 1, this result echoes the assumption that humans have a limited capacity for cognitive processing of information (Lang, 2000). Many studies have suggested that the use of animations as an attention-grabbing tactic for website ads will decrease the audiences' memory of the ad content because the animations may split the audiences' attention and weaken the effectiveness of information processing (Burke, Hornof, Nilsen, & Gorman, 2005;Hong, Thong, & Tam, 2004;Rohrer & Boyd, 2004;Yoo, Kim, & Stout, 2004). There are much more dimensions and details to be explored in a VR environment compared with a web page environment. ...
Article
With the popularity of virtual reality (VR) devices, advertisers are starting to seek the potential advertising tactics and effects in VR. This study investigates the influence of introducing 3D animated agents on video ad effectiveness in a VR context (i.e., brand/product attitude, brand/product recall, purchase intention, and ad-skipping behavior). Findings from a between-subjects experiment in Study 1 confirm that when 2D video ads are played in a VR environment, the presence of 3D animated agents will make the audiences enjoy the ad-viewing experience more and further promote the audiences’ positive attitudes toward the brand and the product and their purchase intention. In Study 2, employing a large industry dataset of VR users, we found that the presence of 3D animated agents can also decrease the ad-skipping rate while the 2D ads are playing. This study provides empirical support for the use of animated agents as an effective tactic in a VR context. How these findings link to previous research and their practical implications are discussed as future directions.
... Search engine advertising has become very common and users are accustomed to encounter ads in their SERP. As a result, they sometimes develop a tendency to disregard the ads; skipping over them as if they do not exist, a phenomenon often referred to as banner blindness [6,13,28]. Nevertheless, marketing research shows search engine advertising is an effective tool advancing various consumer metrics, such as brand awareness and brand consumption, even among users who do not click on the ads [40]. ...
Preprint
Full-text available
Advertisements (ads) are an innate part of search engine business models. Advertisers are willing to pay search engines to promote their content to a prominent position in the search result page (SERP). This raises concerns about the search engine manipulation effect (SEME): the opinions of users can be influenced by the way search results are presented. In this work, we investigate the connection between SEME and sponsored content in the health domain. We conduct a series of user studies in which participants need to evaluate the effectiveness of different non-prescription natural remedies for various medical conditions. We present participants SERPs with different intentionally created biases towards certain viewpoints, with or without sponsored content, and ask them to evaluate the effectiveness of the treatment only based on the information presented to them. We investigate two types of sponsored content: 1. Direct marketing ads that directly market the product without expressing an opinion about its effectiveness, and 2. Indirect marketing ads that explicitly advocate the product's effectiveness on the condition in the query. Our results reveal a significant difference between the influence on users from these two ad types. Though direct marketing ads are mostly skipped by users, they can tilt users decision making towards more positive viewpoints. Indirect marketing ads affect both the users' examination behaviour and their perception of the treatment's effectiveness. We further discover that the contrast between the indirect marketing ads and the viewpoint presented in the organic search results plays an important role in users' decision-making. When the contrast is high, users exhibit a strong preference towards a negative viewpoint, and when the contrast is low or none, users exhibit preference towards a more positive viewpoint.
... The related studies can be divided into two groups based on whether the research media is static abstract images or abstract videos with changing backgrounds (Rea et al., 2017). The static images used to be applied in natural conditions, and the videos are usually used in complex scenes with free movement (Chun, 2000;Burke et al., 2005). ...
Article
Full-text available
Construction equipment teleoperation is a promising solution when the site environment is hazardous to operators. However, limited situational awareness of the operator exists as one of the major bottlenecks for its implementation. Virtual annotations (VAs) can use symbols to convey information about operating clues, thus improving an operator’s situational awareness without introducing an overwhelming cognitive load. It is of primary importance to understand how an operator’s visual system responds to different VAs from a human-centered perspective. This study investigates the effect of VA on teleoperation performance in excavating tasks. A visual attention map is generated to describe how an operator’s attention is allocated when VAs are presented during operation. The result of this study can improve the understanding of how human vision works in virtual or augmented reality. It also informs the strategies on the practical implication of designing a user-friendly teleoperation system.
... In a series of eye-tracking studies (Nielsen, 2007;Pernice, 2018), the Nielsen Norman Group stated that the banner blindness phenomenon is a type of selective attention and that users who navigate websites have learned how to pay attention to elements that can be helpful and how to ignore the one's void of information (Pernice, 2018). To increase the user attention toward the ad and deal with the banner blindness phenomena, strategies such as animation can be adopted (Griffith et al., 2001), although it could lead to an increase in the cognitive effort required for processing the ad and negatively affect memory performance associated to the ad recall (Bayles, 2002;Hong et al., 2004;Burke et al., 2005), along with a wide variety of paid advertising formats, more difficult to identify by the user (Mitra et al., 2008;Campbell et al., 2014). Regarding the introduction of new digital advertising formats, the so-called "native" format matches the look, feel, and function of the media format in which it appears (Wojdynski, 2016a; and such "deception" makes it more difficult to recognize by users (Most et al., 2001;Wasserman, 2013;Anni, 2014;. ...
Article
Full-text available
Display and native ads represent two of the most widely used digital advertising formats employed by advertisers that aim to grab the attention of online users. In recent years, the native format has become very popular because it relies on deceptive features that make harder the recognition of its advertising nature, reducing avoiding behaviors such as the banner blindness phenomena, traditionally associated to display advertising, and so increasing its advertising effectiveness. The present study, based on a forefront research protocol specifically designed for the advertising research on smartphone devices, aims to investigate through neurophysiological and self-reported measures, the perception of display and native ads placed within article webpages, and to assess the efficacy of an integrated approach. Eye-tracking results showed higher visual attention and longer viewing time associated with native advertisements in comparison to traditional display advertisements, confirming and extending evidence provided by previous research. Despite a significantly higher rate of self-reported advertising intent was detected for articles containing display ads when compared to articles containing native ads, no differences have been found while performing the same comparison for the neurophysiological measures of emotional involvement and approaching motivation of for the self-reported measures of pleasantness and annoyance. Such findings along with the employment of an innovative research protocol, contribute to providing further cues to the current debate related to the effectiveness of two of the most widely used digital advertising formats.
... However, in the ever growing ecology of information and signs in public spaces, digital public displays compete for attention. Similar to the "Banner Blindness" [19], where users have been accustomed to see uninteresting content and ads in specific place of a web page, the use of public displays for advertisement developed "Display Blindness" [70] for passers-by. The study presented by Dalton et al. [26] leveraging eye-tracking offers deeper insight on this blindness. ...
Thesis
Public libraries are lively third places welcoming people from all levels of society. In addition to offering extensive collections of documents, they host a wide range of events: from public readings, clubs, to hands-on participatory activities. Each of these events sparks the curation and creation of a variety of content by librarians and patrons: articles, bibliographies, playlists, etc. Documenting these events also leads to the creation of new content such as video or audio recordings of conferences. This diverse set of documents related to events is a complement to the preexisting collections. They provide a valuable trace of the history of the library, and opens opportunities for novel ways to communicate about events. Unfortunately, both the events and their related resources do not have a clearly defined place in the library. They vanish once the event is over. To anchor event-related content in a meaningful way in the library space, I followed a place-centric design approach. Public displays appear to be promising mediums for place-centricity. Yet, in an already crowded information space, their use and placement raises questions both on their design and their aim. In order to get a more precise idea of how we could anchor digital information about events in public libraries, we partnered with three public libraries in Sweden, Denmark, and France. We conducted a series of design workshops to imagine how future public displays could be used for events and leverage the places of libraries. This process led to the co-creation of 18 design concepts. From these I draw a design space of public displays in libraries and identify that public displays are perceived as an opportunity for the library to give access not only to new categories of content: live videos, websites, or playlists, but also to content created by the patrons themselves. While conducting these workshops, we reflected on the impact of design tools and activities on the creation of prototypes. To better understand how design tools could shape place centric design, I conducted a study comparing paper and digital sketching tools in situated enactments. I developed Ébauche, a collaborative sketching tool for public displays that supports bodystorming and compared it to paper in a study involving 16 interaction designers. Using both tools, designers were able to roam in a library while designing information systems for public displays. I found that paper led to broader exploration of ideas, and a deeper physical integration in the environment. Whereas, Ébauche encouraged refined and more sophisticated interactive enactments. Finally I present and discuss a case study of a public display technology probe: Explore, that conveys information about events in various places of libraries. Explore was deployed over a total of nine months in the public libraries of Lyon, Aarhus, Goteborg. The different versions of the probe displayed information about events of the library and additional content coming from librarian curation or patron participation. The use of the probe highlighted resources curated and created by librarians that could add value to existing collections of the library. This long field test of Explore also allowed us to identify that public displays were anchored in the place in Lyon by integrating with the mediation activities of the librarians, while their integration in the event’s space and time proved to be particularly relevant in Aarhus.
Article
Breakthroughs in new communication technologies were influenced parties of convergence differently over time. In this study, new ideas created by these relevant technological developments were described through the user and publisher's side of view. The literature review was used to explore the main technical outputs for both traditional and new media environments. The study is important to understand the main effects of horizontal and vertical integration processes on the structures that use publishing activities involved in new online environments. Therefore, this study was aimed to review the possible connections that emerged through the effects of convergence and structures of integrations. The findings obtained from the literature review have shown that horizontal integration in the context of convergence has critical importance for online publishing media organizations. Another important finding is that notions such as banner blindness problem, freemium product or frequency capping, were emerged through convergence that enabled different interactions between user and publishing mechanism.
Article
Full-text available
Modern theories in affective science postulate that emotional stimuli can affect subject’s attention. Emotional stimuli can guide and capture visual attention, which may be related to evolutional importance of quick reactions for emotional objects in the real life. The study examined the influence of valence and arousal of advertisement on the banner blindness phenomenon—ignoring the advertisement or interface details similar to it on the website. In two experiments participants were asked to find the information on the website, where different banners were placed. In the first experiment banners had the same valence, but different arousal. In the second experiment, the banners had different valence, but equal arousal. Contrary to the classical studies in affective science, we found that banners with neutral valence were recognized better as compared to negative and positive ones. The results are discussed in terms of user experience contributing to banner blindness occurrence.
Article
Full-text available
Under certain circumstances, external stimuli will elicit an involuntary shift of spatial attention, referred to as attentional capture. According to the contingent involuntary orienting account (Folk, Remington, & Johnston, 1992), capture is conditioned by top-down factors that set attention to respond involuntarily to stimulus properties relevant to one’s behavioral goals. Evidence for this comes from spatial cuing studies showing that a spatial cuing effect is observed only when cues have goal-relevant properties. Here, we examine alternative, decision-level explanations of the spatial cuing effect that attribute evidence of capture to postpresentation delays in the voluntary allocation of attention, rather than to on-line involuntary shifts in direct response to the cue. In three spatial cuing experiments, delayed-allocation accounts were tested by examining whether items at the cued location were preferentially processed. The experiments provide evidence that costs and benefits in spatial cuing experiments do reflect the on-line capture of attention. The implications of these results for models of attentional control are discussed.
Article
Selective visual attention to objects and locations depends both on deliberate behavioral goals that regulate even early visual representations (goal-directed influences) and on autonomous neural responses to sensory input (stimulus-driven influences). In this chapter, I argue that deliberate goal-directed attentional strategies are always constrained by involuntary, "hard-wired computations, and that an appropriate research strategy is to delineate the nature of the interactions imposed by these constraints. To illustrate the inter- action between goal-directed and stimulus-driven attentional control, four domains of visual selection are reviewed. First, selection by location is both spatially and temporally limited, reflecting in part early visual representations of the scene. Second, selection by feature is an available attentional strategy, but it appears to be mediated by location, and feature salience alone does not govern the deployment of attention. Third, early visual seg- mentation processes that parse a scene into perceptual object representations enable object-based selection, but they also enforce selection of entire objects, and not just isolated features. And fourth, the appearance of a new perceptual object captures attention in a stimulus-driven fashion, but even this is subject to some top-down attentional control. Possible mechanisms for the interaction between bottom-up and top-down control are discussed.
Article
An experiment investigates (1) how the physical structure of a computer screen layout affects visual search and (2) how people select a found target object with a mouse. Two structures are examined---labeled visual hierarchies (groups of objects with one label per group) and unlabeled visual hierarchies (groups without labels). Search and selection times were separated by imposing a point-completion deadline that discouraged participants from moving the mouse until they found the target. The observed search times indicate that labeled visual hierarchies can be searched much more efficiently than unlabeled visual hierarchies, and suggest that people use a fundamentally different strategy for each of the two structures. The results have implications for screen layout design and cognitive modeling of visual search. The observed mouse-pointing times suggest that people use a slower and more accurate speed-accuracy operating characteristic to select a target with a mouse when visual distractors are present, which suggests that Fitts' law coefficients derived from standard mouse-pointing experiments may under-predict mouse-pointing times for typical human-computer interactions. The observed mouse-pointing times also demonstrate that mouse movement times for a two-dimensional pointing task can be most-accurately predicted by setting the w in Fitts' law to the width of the target along the line of approach.
Article
Visual perception and action have traditionally been assumed to involve a sequential relationship in which one first visually perceives an object and then responds (e.g., reaches, grasps) with respect to it on the basis of this perceptual representation. However, a different view has recently been proposed and developed. In this alternate view, early vision is the source of parallel inputs to both visual perception (and therefore cognition and memory) and motor behavior. Thus, the relationship between the perceptual-cognitive and sensorimotor systems is one of common input rather than identity. Perceptual-motor dissociation has been reported in normal humans for a variety of situations. These include size-contrast illusions, the Mueller-Lyer illusion, misperception of depth intervals, induced motion, saccadic eye movements, estimates of hill pitch, the Roelofs effect, and judging eye level in pitched visual environments. The review of apparent dissociation in normal observers has yielded three instances in which the evidence of perceptual-motor dissociation meets the specified criteria for the valid study of this phenomenon. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Yantis and Jonides (1984) demonstrated that the detection of a target in visual search was markedly enhanced when the target was presented as an abruptly onset character embedded among other characters whose presentation was not characterized by abrupt onset. This effect was attributed to a shift of attention caused by abrupt onset. In the present article, we report experiments investigating whether abrupt onset is simply one member of a large class of stimulus characteristics, all of which are capable of capturing attention. To test this possibility, we compared abrupt onset with differences in stimulus luminance and hue to determine whether these also could elicit shifts of attention. They could not.