Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
• Applying visibility adjustments to billboard circulation brings into play each billboard’s location
attributes, creating the opportunity to more accurately measure real audience exposure.
• The physical attributes of out-of-home structures and their relative position to the driver significantly
impact the likelihood that an advertisement on the unit will be noticed.
• The biggest advantage comes from great location coupled with great creative content.
based solely on the location of each piece assessed the impact location has on notic- action beyond being outside and moving
of advertisement inventory. TAB makes ing, using a subset of TAB primary research around (Ephron et al., 2003).
no adjustment to allow for the impact data. Second, the authors added in the vis- To potentially grow the business, the
of strong creative content or the visual ual saliency variable to approximate crea- industry needed a more refined measure
saliency of a billboard. This is primarily tive execution. To do this, they used a novel that took into account the actual noting of
because of difficulties in quantifying vis- software program, called Saliency Toolbox advertisements. Accordingly, TAB devel-
ual saliency and the expense of tracking (Walther and Koch, 2006), from the conflu- oped a more robust measurement system
each advertisement execution. The current ence of computational and cognitive neuro- that incorporated a Visibility Adjustment
research offered a method to objectively science to mimic the human visual attention Index used to reduce out-of-home’s
quantify visual saliency’s measure and process (Itti and Koch, 2000; Itti, Koch, and opportunity-to-see counts. This adjustment
evaluate its importance in comparison to Niebur, 1998). The software identified provided a metric that more accurately
and in conjunction with billboard location objects that likely would have received reflected the number of people who likely
attributes. By conducting two research attention due to visual saliency. It offered would see the advertisement. To formulate
studies, the authors found that a bill- a significant methodological improvement the adjustment, TAB conducted a study to
board’s location, indeed, drives attention opportunity for the advertising research identify the appropriate variables that affect
to advertising. Moreover, strong creative field (Milosavljevic and Cerf, 2008), and the noticing of roadside advertising. The
content, as defined through the research- software like it has been used in several results of that study found that roadside
ers’ visual salience measure, is most effec- studies to date (Pieters and Wedel, 2004; advertising more likely would be noticed
tive when the advertisement is properly van der Lans, Pieters, and Wedel, 2008; Wil- if it were located closer to the road, located
located within the driving environment. son, Baack, and Till, 2015). on the right-hand side of the road (in the
According to the Advertising Research United States and other right-side driving
Foundation, attention to or noticing of STUDY 1 countries), larger in size, viewable from the
advertising is the first stage at which con- Media Measurement center of windshield, and angled appro-
sumers react to the message found within And Out-of-Home Advertising priately to the road (Traffic Audit Bureau,
an advertisement (Ephron et al., 2003). The majority of today’s media measure- 2010). TAB collects, calculates, and records
TAB’s research used eye-tracking stud- ment is still based on people’s proximity the inputs of these variables for each out-
ies to determine whether consumers had to programs, editorials, and advertisement of-home advertisement location and then
focused on the advertising message. But content. This means there is no indication gives an appropriate Visibility Adjustment
eye-tracking studies have a limited scope: whether the advertisements are actually Index calculation for each advertisement.
Although they can determine what an noticed by consumers. This standard often There is a separate pedestrian study of vis-
individual looked at, they cannot deter- is called “opportunity to see.” Opportunity ibility in use as well, but that portion of the
mine why the advertisement captured to see is really a proximity measure and not TAB study fell outside of the purposes of
attention. Was a billboard noticed because an actual measure of noticing advertising. the current study, which focused on road-
of its location attributes or because of its Although not the only measure of expo- side advertising visible to drivers.
visual saliency? In many ways, the effects sure to a medium, opportunity to see is a In its effort to develop the Visibility
of an out-of-home message are signifi- standard measure across all media. Print Adjustment Index, TAB consulted earlier
cantly confounded with the effects of the measures exposure to its content, and tel- industry research conducted by the Out-
medium (Ephron et al., 2003). evision primarily measures exposure to its door Advertising Association of America
Incorporating visual saliency as a meas- programming. The out-of-home billboard (OAAA). These studies proved that bill-
ure of an advertisement’s creative execution version of opportunity to see is circulation. board attributes do have an impact on
into out-of-home attention research and Circulation is defined as the number of peo- noticeability (OAAA, 1983) and that crea-
teasing out its effects can help answer why ple passing by each billboard. tive content has an impact as well (OAAA,
certain advertisements are noticed. A grow- Many within the out-of-home advertis- 2000). The latter two studies also proved
ing number of researchers have made this ing industry believed its proximity meas- that color and luminance have an impact
issue a priority within the industry (Taylor, ures were insufficient. This is because on noticing.
2012; Wilson and Till, 2012). In attempting out-of-home exposure occurs spontane- TAB also reviewed work done in the
to answer these questions, the authors first ously and does not require any overt United Kingdom. The U.K. out-of-home
industry research body Route, formerly acknowledged the need for a further inves- scan the environment. Drivers have been
known as POSTAR, was the first to take tigation on the noticing of advertisements trained, for example, to look for roadside
out-of-home audience measurement into due to visual saliency. markers and directional signs on one par-
the realm of actually seeing the advertise- ticular side of the road (Cole and Hughes,
ment. They used static imagery to deter- Location and Attention to Billboards 1984; Shinoda, Hayhoe, and Shrivastava,
mine noticing of different out-of-home Academics and practitioners have identi- 2001). Scanning the environment for task-
media and developed a Visibility Adjust- fied a number of billboard attributes that relevant information along the roadside
ment Index model for the United King- influence the amount of attention given to creates the possibility for incidental expo-
dom. TAB was able to benchmark Route/ roadside advertising (Taylor, 2010; Traf- sure to billboard advertising, and research
POSTAR’s work, and their results were a fic Audit Bureau, 2010; Wilson and Till, has shown that advertisements found
guide for TAB’s own results. 2012). These attributes are primarily loca- on the right-hand side of the road in the
TAB made available its Visibility Adjust- tion based and include a billboard’s dis- United States and the left in the United
ment Index scores for the current article tance from the road; the side of the road Kingdom have a greater likelihood of
(Traffic Audit Bureau, 2010). In this 2010 in which it appears; whether it is view- being noticed because the advertisement
study, TAB collected the initial noticing able from the center of the windshield at falls within a driver’s shifted line of sight
and reexamining scores for each billboard any point during viewing, called a center (Donthu, Cherian, and Bhargava, 1993;
studied, but only used noticing scores in its approach; its size; and the amount of time Young et al., 2009).
2010 research, as the purpose was to assess in which it is visible to those who pass by Also promoting increased attention are
whether the board was seen or not seen. it, called dwell time. larger sized advertisements, which are
Reexamination scores were collected in For obvious reasons, drivers concentrate noticed more frequently than those that are
case something could be gleaned in future a majority of their attentional resources on smaller in scale (Hughes and Cole, 1984;
research. Noticing scores ranged from their forward field of vision (Crundall, Thomas-Smith and Barnett, 2010). One
the single digits to more than 80 percent, Underwood, and Chapman, 2002). Task- eye-tracking study of roadside advertising
depending on the quality of the location. relevant objects within this field, such as found that larger billboards received 0.65
The primary purpose of TAB’s visibility other vehicles and traffic signs, receive the glances per sign per subject, compared
research program was to determine the most attention with only occasional glances with 0.06 glances per sign per subject for
drawing power of out-of-home structures to task-irrelevant objects, such as roadside smaller billboards (Beijer, 2002). And, in
like bulletins or posters and their physical advertising (Chapman and Underwood, a transit advertising environment, larger
position, not the power of the advertising 1998). When drivers do attend to roadside advertisements garnered higher levels of
on the structure. TAB’s measurement sys- advertising, research indicates that 97 per- recall and recognition than smaller bill-
tem focuses specifically on the location, cent of glances are within 25 degrees of a boards (Wilson and Till, 2008).
and not the creative content, because it driver’s forward field of vision and 75 per- These first four variables are part of
is neither feasible nor practical to have a cent are within 10 degrees (Beijer, 2002). TAB’s Visibility Adjustment Index and
measure for creative execution, particularly Similar results have been found in other were analyzed previously in a larger data-
since there is no way to predict creative outdoor advertising environments, such set and found to be highly predictive of an
quality. In its 2010 study, TAB controlled for as airports and shopping malls (Thomas- ad’s potential to be noticed at least once by
creative quality by making sure that a wide Smith and Barnett, 2010). Taken together, drivers (Traffic Audit Bureau, 2010). The
array of advertisement executions were these studies highlight the importance current authors explain shortly that they
included (Traffic Audit Bureau, 2010). TAB for billboards to be positioned within a used a smaller portion of the dataset in
also graded the creative execution of each driver’s forward and narrow line of sight. this study. On the basis of the previously
advertisement included in the study based Billboards located closer to the road or discussed combination of academic and
on feedback from a panel of senior execu- having a center approach more likely will TAB work, the current authors offered the
tives. These grades served as benchmarks be noticed. following hypotheses:
that ensured that the creative content was Drivers make occasional digressions
neither overtly poor nor grand and that it from a forward line of sight to other areas H1: Billboard advertising more likely
was captured appropriately by the video. of the driving environment for additional will be noticed if it is closer to
By controlling for creative content, TAB task-relevant information or to simply the road.
H2: Billboard advertising in the U.S. experts in eye-tracking technology and randomly during the video loop. Video
more likely will be noticed if it have a long history of measuring out-of- had been recorded using a high-definition
is on the right-hand side of the home advertising. In that larger study, video camera.
road as opposed to the left-hand attention to out-of-home advertising was A variety of roads in Connecticut, Phoe-
side of the road. assessed across both pedestrian panels and nix, and Chicago had been selected for
roadside billboards. To narrow the scope of filming to be representative of a number
H3: Billboard advertising more likely their research and to minimize the number of driving experiences, including urban/
will be noticed if it has a center of potential confounding variables, the cur- suburban, old/new cities, and different
approach. rent authors held the advertising format billboard formats, including old or new
constant, using only billboards visible to structures and pole, building, or wall
H4: Billboard advertising more likely those driving. They also added dwell time mounts. Every attempt had been made
will be noticed if it is larger in to the existing mix of Visibility Adjustment to ensure the driving video was as natu-
size. Index variables from the 2010 TAB study. ral as possible with the vehicle traveling
Dwell time is an additional dependent vari- at the same speed as surrounding traffic
TAB’s original research (Traffic Audit able not previously analyzed in the original and in the center lane if multiple lanes in
Bureau, 2010) did not specifically explore 2010 TAB research. one direction were present. Filming had
the impact of dwell time or the amount of occurred at midday in sunny or partly
time that a billboard is in a driver’s field of Subjects sunny weather conditions to ensure that
vision. In 2013–2014, however, TAB subse- The 2010 TAB study recruited a total of 312 billboard views were comparable and free
quently conducted another visibility study subjects in the first stage of the study who of glare across video clips.
that focused on the impact of dwell time on were equally distributed across 10 U.S. cit- Video clips varying in length from one
the noticing of both standard and digital ies: Akron, OH; Boston, MA; Chicago, IL; to three minutes were carefully merged
out-of-home advertising. TAB released this San Antonio, TX; Detroit, MI; Enfield, CT; to create three unique driving sequences.
enhancement in 2014 as Ratings 2.0 (Traffic Fort Lauderdale, FL; Hicksville, NY; Phoe- Each video ranged from eight to 12 min-
Audit Bureau, 2014). In the 2014 TAB study, nix, AZ; and San Diego, CA. Participants utes in length and contained between
dwell time was measured in the field by were recruited by telephone and mall eight and 11 clips. To minimize hypoth-
driving respondents at different speeds and intercept, screened to have normal vision, esis guessing, some clips contained lim-
in different traffic congestion patterns past and to be above the age of 18 years. TAB ited billboards with long spacing between
billboards. It was confirmed that slower pas- had attempted to recruit an equal split of advertisement locations. Each clip faded to
sages, which create longer dwell times, do male and female participants as well as a black before the next clip faded in to cre-
increase the noticing of all billboards. Dwell wide array of ages. ate smooth transitions. A two-second black
time is now part of the permanent audience frame appeared in between each clip. A
measurement for all TAB-measured inven- Stimuli total of 119 billboards were included across
tories. Accordingly, the current authors also Video clips for the TAB study had been all three driving sequences.
offered the following hypothesis: obtained by mounting a camera at a
72-degree angle inside a vehicle from the Procedure
H5: Billboard advertising more likely driver’s point of view as per independent The 2010 TAB study brought participants
will be noticed if it has a longer research recommendations (Chapman and in the ten previously mentioned testing
dwell time. Underwood, 1998; Wallis and Bülthoff, sites into a testing facility at a local mall in
2000). The video had been filmed from a their respective city. The facility included
Study 1 Methodology location in the vehicle where its hood was a waiting room area and a separate room
In Study 1, the authors reanalyzed a portion not visible and when there was no reflec- where the study took place. Participants
of the findings from the earlier eye-tracking tion from the dash. Because multiple cars randomly were shown one of the three driv-
study performed by TAB and its vendor had been used across several filming loca- ing sequence videos. They were instructed
Perception Research Services (Traffic Audit tions, TAB did not want the viewer to be to watch it as if they were the driver of the
Bureau, 2010). Perception Research Ser- distracted by the hood changing and hav- vehicle. Participants were told the purpose
vices, based in Teaneck, N.J., are worldwide ing sun glare appearing and disappearing of the video-clip-viewing exercise was to
assess normal driving behaviors, which ear- participants were removed for inattentive- center approach and 1 for having a center
lier pilot phase participants had indicated ness, leaving 306 participants. approach (Hypothesis 3). A billboard was
was what they believed the purpose of the considered to have a center approach if it
study to be. No pilot participant believed Variables was visible from the center of the wind-
the purpose of the study was the noticing TAB defined the noticing of billboards as shield at any point during viewing, which
of roadside advertising. both the initial noting of a billboard and was typically within ten degrees of a driv-
Videos were displayed on a new large the subsequent reexamining of a billboard. er’s direct forward-looking glance (five
high-definition television at a distance of These two items formed the dependent degrees to the right and left of the viewer’s
about 10 feet, so as to approximate the variables used in the current study. The eye). Size was measured in square feet of
proper field of view given through a wind- initial noting score was measured as the the billboard’s display area (Hypothesis
shield. Participants’ eye movements were percentage of subjects attending to a bill- 4). A 30-sheet poster would, for example,
recorded using an ISCAN ETL-324 eye- board once for at least 0.25 seconds while be commonly listed as 300 square feet (12′
tracking system that was mounted to the the reexamining score was measured as × 25′). Finally, the amount of time a bill-
floor. No headgear was worn. Participants the percentage of participants noting a board was visible on the video, an indica-
were recorded as fixating on a billboard if billboard a second time for at least 0.25 tor of dwell time, was measured in seconds
they attended to it for at least 0.25 seconds. seconds. All location variables needed for (Hypothesis 5). The authors calculated the
This minimum threshold was identified each hypothesis were coded separately in dwell-time measure, as it was not part of
by an onsite eye-tracking expert and con- the TAB study. the 2010 TAB study.
firmed by consulting academic research For the independent location variables,
(Sereno and Rayner, 2003). a billboard’s distance from the road was Study 1 RESULTS
To verify a driver simulation, 20 stimuli measured in feet from its base to the closest A linear regression analysis was used to
across the three driving sequences were lane from which it was visible (Hypothesis determine how location attributes influ-
included as objects participants must see to 1). The side of the road in which a billboard ence the initial noting and subsequent
be included in the study, for example, a low- appeared was coded as a 0 for the left-hand reexamining of billboard advertising.
flying airplane on the horizon, a passing side of the road and a 1 for the right-hand Pearson correlations among the independ-
car, and so forth. Additionally, participants side of the road (Hypothesis 2). Whether ent variables were not greater than 0.37;
were observed behind a double mirror to a billboard was visible from the center of variance inflation factors (VIFs) for all
ensure participants remained attentive. Six the road was coded as 0 for not having a variables in the regression were less than
Table 1
Pearson Correlation Matrix (N = 119)
1 2 3 4 5 6 7 8 9 10 11
1. Noting —
2. Reexamining 0.95*** —
8. Visual saliency (VS) 0.12 0.02 –0.06 0.17 –0.06 0.18* –0.26** —
10. VS × Side of road 0.29** 0.19* –0.07 0.57*** –0.03 0.11 –0.29** 0.77*** 0.52*** —
11. VS × Center 0.23* 0.28** –0.07 0.05 0.75*** 0.15 0.16 0.06 –0.01 0.07 —
Billboard size (sq ft) 0.256** 0.264** study and added dwell time as an addi-
tional independent variable and reexam-
Dwell time 0.245** 0.303***
ining as an additional dependent variable.
Visual saliency 0.102 0.134 The initial TAB study also presented a mul-
Visual saliency × Distance –0.419** –0.253* tivariate solution while the hypotheses the
current authors tested were univariate.
Visual saliency × Side of road 0.412** 0.196
Visual saliency × Center approach 0.188* –0.087
STUDY 2
Adjusted R2 0.509 0.201 0.522 To determine visual saliency’s role in the
ΔR 2
0.530*** 0.228*** 0.029 attention capture of billboard advertising,
the current authors used visual attention
Model F 25.487*** 8.423*** 15.324***
theory from cognitive neuroscience to
Note: Standardized Beta coefficients are presented in all models. Sq ft = square feet. inform their hypotheses.
*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
Visual Attention Theory task-driven, top-down factors take hold in light and dark (the intensity dimen-
Visual attention theory suggests that (Parkhurst, Law, and Niebur, 2002). sion); and that are arranged along a real
two factors explain why people attend to Top-down factors are related to cogni- or imaginary axis (the orientation dimen-
objects in a visual scene (Treisman and tion and are associated with a person’s sion), more likely will receive involuntary
Gelade, 1980; Wolfe, 1994, 1998). The two existing knowledge and expectations eye movements. Visual saliency excludes
factors commonly are classified as bottom- about a visual scene (Corbetta and Shul- the noticing of objects within a billboard’s
up and top-down. These factors are not man, 2002). Objects are attended to if they creative content because of cognitive or
only applicable to broader visual contexts possess task-relevant features (Theeuwes, top-down factors, like existing knowledge
but also to more specific advertising con- 2004; Yarbus, 1967). In a driving context, of a scene or the semantic meaning of
texts (Milosavljevic and Cerf, 2008; Pieters attention related to top-down processing words, pictures, or graphics.
and Wedel, 2004). is often is focused on objects containing Objects that are more visually salient
Bottom-up factors refer to the characteris- information that aids in the driving task, tend to be noticed early. This is because
tics of objects that are prominent within the such as traffic signs, traffic signals, and attention is distributed broadly over a vis-
visual field. People attend to these objects objects in the intersection and highway ual field so the environment can be assessed
reflexively and involuntarily, often because exits (Chapman and Underwood, 1998). and relevant objects identified. With the
of an object’s size, motion, curvature, orien- Unlike bottom-up processing, which is passage of time, or as tasks become more
tation, color, and luminance. Objects with done in parallel, top-down processing is demanding, however, top-down guidance
these characteristics pop out and are gen- performed serially and objects are attended of attention increases in strength and influ-
erally the first objects to be noticed when to in a particular order depending on the ence and attention becomes much more
individuals orient themselves to a new task (Wolfe, 1994). focused, often described as a spotlight or
environment. In a driving context, objects window of attention (Chen and Zelinsky,
larger in size tend to be more easily noticed Visual Saliency 2006). As a result, the role of visual sali-
than smaller objects (Hughes and Cole, And Attention to Billboards ency in attracting attention becomes more
1984), as are objects that quickly move in Three bottom-up features typically are limited, but it does not disappear entirely
and out of a driver’s field of vision, such referred to as visual saliency: color; lumi- (Theeuwes, 2004). Although objects out-
as pedestrians or other vehicles (Shinoda, nance, which can be defined as intensity; side the window of attention become too
Hayhoe, and Shrivastava, 2001). Moreover, bright and dark contrasts; and orienta- far removed from focal attention to ben-
objects aligned along a real or imaginary tion, as in objects aligned along an axis. efit from visually salient properties, visu-
axis that possess contrasting colors and that They traditionally are grouped together ally salient objects within the window of
are brighter or darker than other objects in and researched collectively because they attention can still attract attention despite
the visual field more likely will be noticed represent the best approximation of the the predominance of top-down process-
(Underwood, Humphrey, and van Loon, visual features detected early in human ing (Mortier, Donk, and Theeuwes, 2003;
2011). In a driving situation, these might visual search. Humans also tend to Theeuwes, 1991; Yantis and Jonides, 1990).
include a red sign against a blue sky or a respond to these factors as a group rather The narrowed window of attention is
line of orange construction barrels. than individually (Borji, Sihite, and Itti, quite relevant to the roadside advertis-
Bottom-up processing is done quickly 2013; Itti and Koch, 2000; Le Meur and ing environment. In this context, drivers
and objects are processed in parallel, Chevet, 2010). primarily focus their attention in their
meaning that multiple objects can be pre- With respect to out-of-home advertis- forward-looking visual field where future
attentively processed simultaneously. ing, visual saliency is defined as the abil- traffic hazards are likely to be found
Bottom-up factors are thought to have ity for a billboard’s creative content, not (Chapman and Underwood, 1998). From
the greatest influence on attention imme- its physical structure, to be noticed by pas- a top-down perspective, other vehicles
diately after their onset as people orient sersby because of its reflexive or bottom- and pedestrians likely will capture their
themselves to the situation or task (Donk up properties, meaning its color, intensity, attention. Visually salient objects, how-
and Soesman, 2010; Le Meur, Le Cal- and orientation. Objects, or group of pix- ever, such as billboard advertising, also
let, Barba, and Thoreau, 2006). As time els, within the advertisement that have may capture attention. This is especially
progresses, however, attention to objects contrasting color combinations (the color true for billboards with a center approach
due to bottom-up factors decreases as dimension); that have stark differences as it places the advertisement within the
driver’s forward-looking window of atten- H7: Visually salient advertising on a Within each map, the software analyzes the
tion. Center-approach billboards with con- billboard will show a significant pixelated image using a center-surround
trasting colors, dark and bright areas, and difference in noticing when it is methodology. That is, it analyzes a pixel,
variation in object orientation likely will located on the right-hand side of or group of pixels, and compares it to its
capture attention. the road in the U.S. neighboring pixels. Areas with significant
Although drivers need to concentrate differences in color, intensity, and orienta-
their narrowed window of attention on H8: Visually salient advertising on a tion are thus extracted from the image as a
their forward-looking visual field, they billboard will show a significant potential object that is likely to be attended
may also need to scan the environment for difference in noticing when it to for stimulus-driven reasons.
street signs and other directional informa- has a center approach. Within the color conspicuity map,
tion. Drivers have been trained to look for objects are noticed based on contrasting
task-relevant signage close to the road and Study 2 Methodology color combinations of red–green and blue–
typically on one side of the road (Trick, Procedures yellow. For the intensity conspicuity map,
Enns, Mills, and Vavrik, 2004). Billboards To test Hypotheses 6 through 8, the objects that are expected to attract atten-
with visually salient advertising located authors calculated the visual saliency of tion are identified through a group of pix-
within a driver’s shifted window of atten- advertising from billboards in Study 1 els with dark centers and bright surrounds,
tion may then capture this attention. using a computational model for visual or vice-versa. The orientation conspicuity
Results from the first study confirmed attention. The model comes from the con- map is created by searching for pixels that
that billboards placed within the driver’s fluence of cognitive and computational are aligned along 0°, 45°, 90°, and 135° axis.
window of attention more likely would be neuroscience and successfully mimics To determine which object, that is, which
noticed. Specifically stated, placing a bill- human visual attention behavior (Koch group of pixels, is the most salient across
board closer to the road, on the right-hand and Ullman, 1985). The most widely used the entire image, the saliency model line-
side of the road (in the United States), or computational model is the one developed arly sums the three conspicuity maps into
with a center approach will help position by Itti and colleagues (Itti and Koch, 2000; one saliency map. The group of pixels with
the billboard within a driver’s window of Itti et al., 1998). It relies on computer algo- the highest value within the saliency map
attention and thus create the most optimal rithms to emulate human visual attention, is selected as the object that most likely
situation for noticing. A billboard’s ability to and its prediction of visually salient objects would be attended.
attract attention in these circumstances may in images has been shown to strongly cor- To determine the next object that likely
be greatly enhanced by the visual salience relate with actual human behavior and would receive attention, the saliency
of the creative message it carries. Billboard eye-tracking studies (Elazary and Itti, 2008; model inhibits the first attended to item,
advertising that possesses contrasting Le Meur and Chevet, 2010; Parkhurst et al., a process that has been demonstrated to
colors, bright and dark areas, and variation 2002; Peters, Iyer, Itti, and Koch, 2005). A occur naturally in human behavior (Pos-
in object orientation within the window of Matlab algorithm, called Saliency Toolbox ner and Cohen, 1984). This prevents the
attention will have an increased possibility (Walther and Koch, 2006), was used as first object from immediately being rese-
of being noticed (Mortier et al., 2003; Theeu- an interface to Itti and colleagues’ com- lected as salient. The object is inhibited for
wes, 1991; Yantis and Jonides, 1990). Higher putational model (hereinafter referred to approximately 500–900 milliseconds (ms)
noticing increases the amount of real pro- as simply the visual saliency model). The of simulated time depending on the static
cessing. Visual saliency outside the window Saliency Toolbox is a collection of math- image’s complexity. Once inhibited, the
of attention was not expected to increase the ematical functions and scripts used to cal- saliency model recalculates the conspicuity
noticing of a billboard’s advertising. As a culate the visual saliency in an image. and saliency maps to identify the second
result of these theories, the authors offered The visual saliency model analyzes a most salient object, a process that takes
the following hypotheses: static image for the presence of contrast- about 30–70 ms.
ing colors, bright and dark areas, and vari- The program continues this process until
H6: Visually salient advertising on a ations in object orientation. The analysis all salient objects are identified and the first
billboard will have a significant occurs separately for each of the three attended-to object is reselected. Through
difference in noticing when it is bottom-up features by generating a conspi- the selection process of salient objects, the
located close to the road. cuity map for each variable (See Figure 1). saliency model creates a scan-path plot,
Variables
Based on color, intensity, and orientation,
the visual saliency model was set to iden-
tify salient objects within a circular focal
area having a radius of 1/16 of the image’s
Color Intensity Orientation width (Elazary and Itti, 2008). To deter-
mine whether billboard advertising was
visually salient, still images from the video
from the original TAB study were captured
every two seconds. All images containing a
target billboard then were analyzed using
the visual saliency model and assigned a
visual saliency score. If any portion of the
billboard’s display area was within the focal
area and within the first five objects within
Saliency Map
the scan-path plot, the billboard’s adver-
tising was deemed visually salient within
that image. Advertising saliency for each
billboard then was calculated by summing
Figure 1 Conspicuity Maps for Color, Intensity, and up the number of images in which a par-
ticular billboard’s advertising appeared as
Orientation and Overall Saliency Map
visually salient and dividing it by the total
number of images containing the advertise-
ment. An advertisement visible on screen
for 20 seconds, for example, would have
10 images associated with it (one for every
3 two seconds on screen). If an advertisement
was visually salient in four of the 10 images,
then the advertisement was given a visual
1 saliency score of 0.4 (four divided by ten).
Using the first five objects as visually sali-
ent within the scan-path plot has support
2
in the visual attention literature. First, it is
consistent with other research that suggests
bottom-up saliency is greatest just after
stimulus onset (Donk and Soesman, 2010;
Le Meur et al., 2006; Parkhurst et al., 2002).
Second, the first five locations identified
by subjects as visually salient are the most
similar across subjects, whereas later fixa-
Figure 2 Abbreviated Scan-Path Plot tions are not (Tatler, Baddeley, and Gilchrist,
2005). Third, comparisons using the visual Billboard advertising more likely would be
saliency model with what people physically
label as generally the most interesting object reexamined if the advertising was visually
in outdoor scenes correlates the strongest
with the first five objects identified as visu- salient and located closer to the road.
ally salient (Elazary and Itti, 2006).
To specifically test Hypotheses 6 through
8, three of the location variables were mul-
tiplied by the visual saliency score for each
billboard advertisement creating three the road (Hypothesis 6, p < 0.01), visually Billboard advertising that is visually sali-
interaction variables: Visual saliency × Dis- salient and on the right-hand side of the ent and on the right-hand side of the road
tance, Visual saliency × Side of the road, road (Hypothesis 7, p < 0.01), and visually (Hypothesis 7, p = 0.17) and visually sali-
and Visual saliency × Center approach. salient with a center approach (Hypothesis ent with a center approach (Hypothesis 8,
8, p < 0.01). Advertising visual saliency by p = 0.39) were not supported. As expected,
Study 2 RESULTS itself was not significant (p = 0.856). advertising visual saliency by itself was
A linear regression analysis was used to To address the research question of also not significant (p = 0.39).
assess the impact the visual saliency of a whether visual saliency adds any addi- The results for the full model for the
billboard’s advertising had on its poten- tional value beyond the location-based subsequent reexamination of billboard
tial to be initially noticed and subse- attributes, the authors used an additional advertising were significant, adjusted R2 =
quently reexamined. Pearson correlations set of linear regressions combining all vari- 0.514, F(9, 109) = 14.867, p < 0.00, but its
among the independent variables were ables into one regression equation. One adjusted R2 was not significantly improved
not greater than 0.77 (See Table 1). Multi- regression was used to assess the variables from the previous adjusted R2 of 0.530 for
collinearity appears not to be an issue as associated with all eight hypotheses for the only location-based attributes (p = 0.639)
the VIFs for all variables in the regression initial noticing of billboard advertising and (See Model 1 and 3, Table 3). Only four
were less than 2.5 (Hair et al., 2005). another to assess the subsequent reexami- hypotheses were supported. Billboard
The results for the initial noting of a nation of billboard advertising. In the new advertising more likely would be subse-
billboard’s advertising were significant, models, the VIFs did increase from earlier quently reexamined due to its distance
adjusted R = 0.201, F(4, 114) = 8.423, p <
2
models to 6.0, but they were still below the from the road (Hypothesis 1, p < 0.00),
0.00 (See Model 2, Table 2). All hypoth- level of ten recommended in marketing side of the road (Hypothesis 2, p < 0.01),
eses were preliminarily supported. Bill- research (Mason and Perreault, 1991). billboard size (Hypothesis 4, p < 0.01), and
board advertising more likely would be The results for the full model for the ini- dwell time (Hypothesis 6, p < 0.00). Center
noticed initially if it was visually salient tial noting of billboard advertising were approach (Hypothesis 3, p = 0.06), Visually
and located closer to the road (Hypoth- significant, adjusted R2 = 0.522, F(9, 109) = salient × Closer to the road (Hypothesis
esis 6, p < 0.01), visually salient and on the 15.324, p < 0.00, but its adjusted R was not
2
6, p = 0.23), Visually salient × Side of the
right-hand side of the road (Hypothesis 7, significantly improved from the previous road (Hypothesis 7, p = 0.32), and Visually
p < 0.01), and visually salient with a center adjusted R of 0.509 for only location-based
2
salient × Center approach (Hypothesis 8, p
approach (Hypothesis 8, p < 0.05). Adver- attributes (p = 0.142) (See Model 1 and 3, = 0.93) were not supported. As expected,
tising visual saliency by itself was not sig- Table 2). Six of the eight hypotheses were advertising visual saliency by itself was
nificant (p = 0.532). supported. Billboard advertising more also not significant (p = 0.97).
The results for the subsequent reexami- likely would be initially noticed due to To better understand these latter regres-
nation of a billboard’s advertising were its distance from the road (Hypothesis 1, sion results and determine which specific
significant, adjusted R2 = 0.164, F(4, 114) p < 0.00), side of the road (Hypothesis 2, variables contributed the most to atten-
= 6.803, p < 0.00 (See Model 2, Table 3). p < 0.01), center approach (Hypothesis 3, tion capture, a relative weight analysis
All hypotheses were preliminarily sup- p < 0.05), billboard size (Hypothesis 4, p < was performed. A relative weight analysis
ported. Billboard advertising more likely 0.01), dwell time (Hypothesis 5, p < 0.00), is useful in determining the set of predic-
would be reexamined if the advertising and whether it was both salient and located tor variables that maximize the amount
was visually salient and located closer to closer to the road (Hypothesis 6, p < 0.05). of variance explained in the regression
equation, which is different than what DISCUSSION of Studies 1 and 2 In the roadside advertising arena, there-
regression coefficients provide (Tonidan- A billboard’s location and proximity to the fore, there is no true definition of audience
del and LeBreton, 2011). Regression coef- driver’s window of attention was shown without understanding each billboard’s
ficients explain how a one-unit increase in here to be of primary importance. Visual specific location. In addition to the origi-
a predictor variable influences the depend- saliency becomes important only when nal variables TAB assigned to its Visibility
ent variable while holding all other vari- these location criteria are met. Adjustment Index (Traffic Audit Bureau,
ables constant. A relative weight analysis Location-based attributes appeared to 2010), the current authors added dwell time
explains the contribution each of the pre- have contributed the most to the current (the amount of time spent in range of see-
dictor variables has on the variance of the authors’ understanding of the issue. In fact, ing the board) to the list of variables having
dependent variable. Relative weights were the F value for the full model decreases a significant impact. It was dwell time that
calculated using the procedure outlined by quite significantly from the model con- connected the contact zone of the respond-
Johnson (2000) (See Table 4). taining only the location-based variables, ent to the location attributes of the billboard.
Although the order of importance for which suggests the superiority of location- This investigation also went beyond the
the following variables varies across based attributes in predicting attention to initial noticing of the advertisement and
dependent variables, the relative weight billboard advertising. therefore the original purpose of the TAB
analysis shows that distance from the The initial noticing of billboard adver- research. The authors also examined the
road, side of the road, dwell time, and tising was dependent on its distance from environmental and visual saliency aspects
billboard size were by far the most impor- the road, the side of the road in which against the reexamination of billboards.
tant variables in explaining both the it appeared, its size, it having a center Understanding reexamination is noteworthy
initial noting and the reexamining of bill- approach, and it being visible for longer from an audience measurement perspective
boards. Collectively, these four variables periods of time. Thus, the use of location- when it comes to applying these findings to
explained 72 percent and 75 percent of based attributes was necessary to develop digital billboards. Digital billboards have
the total variance in noting and reexam- an accurate and meaningful definition of more than one advertisement on the same
ining billboard advertising, respectively. out-of-home media audiences in a road- location. The rate of reexamining is thus a
No other variable individually contrib- side context. It is a higher level definition critical input to applying ratings to individ-
uted more than an additional ten percent in that it is based on the likelihood of notic- ual spots within one digital structure.
toward this explanation. ing advertisements. The four location-based attributes that
impacted the initial noticing of the adver-
Table 4 tisement also significantly contributed to
it receiving a subsequent glance. Distance
Relative Weights of Predictor Variables1
from the road, side of the road, dwell
Noting Reexamining
time, and billboard size are the factors that
Variable RW2 Percent3 RW2 Percent3 assisted drivers in the current study to take
Distance from road 0.153 27.5 0.136 24.5 another look at billboard advertising. One
Side of road 0.099 17.8 0.093 16.7 variable, center approach, had no effect on
reexamination in the full model. A possible
Center approach 0.030 5.3 0.041 7.4
explanation for this is that advertising with
Billboard size (sq ft) 0.061 11.0 0.062 11.2
a center approach may be noticed and suf-
Dwell time 0.089 15.9 0.124 22.3 ficiently processed in the first glance due
Visual saliency 0.022 3.9 0.012 2.1 to its centrality within the forward field of
Visual saliency × Distance 0.036 6.4 0.031 5.5 vision thereby negating the need to reex-
amine it again.
Visual saliency × Side of road 0.054 9.7 0.033 6.0
Visual saliency was important within the
Visual saliency × Center approach 0.014 2.4 0.024 4.3
framework of well-positioned locations, but
Notes: not as important as location-based attrib-
1
Full model (Model 3 from Tables 2 and 3)
utes. This seems to suggest that bottom-up
2
Relative Weight (RW)
3
Percent represents each individual RW’s contribution to the sum of RW factors have a limited influence on drivers’
circulation brings into play each billboard’s et al., 2015; Wilson and Till, 2011).
see is just the first part. Applying visibility based on the TAB ratings system, includ- promotion. Wilson’s published work has appeared in the
adjustments to billboard circulation brings ing dwell time, can make a difference in Journal of Advertising Research, Journal of Advertising,
into play each billboard’s unique location overall campaign impact. International Journal of Advertising, Journal of Current
attributes to more accurately measure real Issues and Research in Advertising, Journal of
audience exposure. No other media meas- LIMITATIONS and FUTURE RESEARCH International Marketing, and International Marketing
ures actually seeing advertising (Ephron A limitation of this research is that road- Review, among other journals.
et al., 2003). The current article reaffirms side advertising effectiveness was meas-
that the physical attributes of out-of-home ured using attention given to billboard Jeffrey Casper is svp, director of marketing for the Traffic
structures and their relative position to the advertising. The research did not take Audit Bureau of Media Measurement, Inc. (TAB, renamed
driver significantly impact the likelihood into account whether the message was Geopath in September 2016), the official audience
that an advertisement on the unit will be processed, retained in memory, or persua- measurement bureau for out-of-home advertising in the
noticed. Furthermore, this article verified sive. Although attention to an advertise- United States. Prior to his current role, Casper served
TAB’s recently completed study that dwell ment is certainly a necessary condition for as svp, director of audit operations and research at
time also has a significant impact on notic- message processing, it is not sufficient to TAB, where he codeveloped the TAB eyes on audience
ing advertising. TAB studies to date verify ensure that processing occurs (MacInnis measurement system. Casper has worked as a media
that this location premise along with dwell and Jaworski, 1989). research executive at MEC and JWT. He also has served
time do matter with quantitative facts. An additional limitation is that this as an adjunct assistant professor of marketing at Pace
Great creative execution as a measure research used video rather than an in-field and Fordham Universities in New York.
of impact was not part of the original TAB study, which makes it difficult to take into
study. The current research demonstrates account driver speed and the actual dis-
that identifying good locations is very tance a billboard comes into view. This ACKNOWLEDGMENTS
important for noticeability. There is an may be less of a concern, however, consid- The authors thank the Traffic Audit Bureau for
incremental finding, however, that comes ering other research has found that simula- Media Measurement, Inc. (TAB) for providing
from the power of creative content, opera- tions produce similar visual attention and the data analyzed in this article and acknowl-
tionalized in this study as visual saliency. task engagement results as compared to edge Perception Research Services, Inc. for their
Visual saliency shows that content and field studies (Wang et al., 2010). TAB’s 2014 work in providing the eye-tracking research.
execution can have a significant impact field study of dwell time also confirmed The data were shared so that TAB could learn
if care is taken to place the advertisement the positive impact of this variable (Traffic more about ad noting and visual saliency.
properly. To get the biggest advantage, Audit Bureau, 2014).
great location coupled with great crea- Future research could manipulate vari-
tive content is necessary. It is, therefore, ous top-down factors to determine their REFERENCES
possible and worthwhile to evaluate each influence on the window of attention. For
unit not only on its location specifics but example, is it possible to shift the win- Beijer, D. D. Driver Distraction Due to Roadside
also on whether the advertisement is con- dow of attention by altering top-down Advertising. (Master’s thesis), Graduate Depart-
sidered affective within the environment goals? Also, research should consider an ment of Mechanical and Industrial Engineer-
where it is placed. Testing before locations additional measurement for advertising ing, University of Toronto, Toronto, Ontario,
are selected and then selecting billboards effectiveness beyond attention with out of Canada, 2002.
Borji, A., D. N. Sihite, and L. Itti. “What Stands Ephron, E., B. Harvey, D. Maroney, B. Moran, J. Model Bottom-Up Visual Attention.” IEEE Trans-
Out in a Scene? A Study of Human Explicit Sali- Spaeth, and P. Brandon. “Making Better Media actions on Pattern Analysis and Machine Intelligence
ency Judgment.” Vision Research 91 (2013): 62–77. Decisions.” New York, NY: Advertising Research 28, 5 (2006): 802–817.
Foundation, 2003.
Chapman, P. R., and G. Underwood. “Visual MacInnis, D. J., and B. J. Jaworski. “Information
Search of Dynamic Scenes: Event Types and the Evanschitzky, H., C. Baumgarth, R. Hubbard, and Processing From Advertisements: Toward an
Role of Experience in Viewing Driving Situa- J. S. Armstrong. “Replication Research’s Dis- Integrative Framework.” Journal of Marketing 53,
tions.” In Eye Guidance in Reading and Scene Per- turbing Trend.” Journal of Business Research 60, 4 4 (1989): 1–23.
ception, G. Underwood, ed. Oxford, England: (2007): 411–415.
Elsevier Science, 1998.
Mason, C. H., and W. D. Perreault, Jr.“Collinear-
Hair, J. F., Jr., W. C. Black, B. J. Babin, R. E. Ander-
ity, Power, and Interpretation of Multiple Regres-
Chen, X., and G. J. Zelinsky. “Real-World Visual son, and R. L. Tatham. Multivariate Data Analy-
sion Analysis.” Journal of Marketing Research, 28, 3
Search Is Dominated by Top-Down Guidance.” sis with Reading, 6th ed. Upper Saddle River, NJ:
(1991): 268–280.
Vision Research 46, 24 (2006): 4118–4133. Prentice Hall/Pearson Education, 2005.
Cole, B. L., and P. K. Hughes. “A Field Trial of Hughes, P. K., and B. L. Cole. “Search and Milosavljevic, M., and M. Cerf. “First Attention
Attention and Search Conspicuity.” Human Fac- Attention Conspicuity of Road Traffic Control Then Intention: Insights From Computational
tors 26, 3 (1984): 299–313. Devices.” Australian Road Research Board 14, 1 Neuroscience of Vision.” International Journal of
Elazary, L., and L. Itti. “Interesting Objects Are Le Meur, O., and J. Chevet. “Relevance of a the Role of Salience in the Allocation of Overt
Visually Salient.” Journal of Vision 8, 3 (2008): 1–15. Feed-Forward Model of Visual Attention for Visual Attention.” Vision Research 42, 1 (2002):
Ephron, E., and J. C. Philport.“The Creative Her- Transactions on Image Processing 19, 11 (2010):
esy in Audience Measurement.” In Worldwide 2801–2813. Pieters, R., and M. Wedel. “Attention Capture
Audience Measurement 2005—Online and Outdoors/ and Transfer in Advertising: Brand Pictorial, and
Out-of-Home, D. S. Fellows, ed. Amsterdam, the Le Meur, O., P. Le Callet, D. Barba, and D. Tho- Text-Size Effects.” Journal of Marketing 68, 2 (2004):
Netherlands: ESOMAR, 2005. reau. “A Coherent Computational Approach to 36–50.
Peters, R. J., A. Iyer, L. Itti, and C. Koch.“Compo- Generator, D. S. Fellows, ed. Amsterdam, the Between In-Vehicle Information Interfaces: A
nents of Bottom-Up Gaze Allocation in Natural Netherlands: ESOMAR, 2010. Comparison With Field Testing.” Ergonomics 53,
Images.” Vision Research 45, 18 (2005): 2397–2416. 3 (2010): 404–420.
Tonidandel, S., and J. M. LeBreton. “Relative
Posner, M. I., and Y. Cohen. “Components of Importance Analysis: A Useful Supplement to Wilson, R. T., and B. D. Till.“Airport Advertising
Visual Orienting.” In Attention and Performance Regression Analysis.” Journal of Business Psychol- Effectiveness.” Journal of Advertising 37, 1 (2008):
X: Control of Language Processes, H. Bouma and D. ogy 26 (2011): 1–9. 59–72.
G. Bouwhuis, eds. Hillsdale, NJ: Erlbaum, 1984.
Traffic Audit Bureau.EYES ON Visibility Research. Wilson, R. T., and B. D. Till. “Effects of
Sereno, S. C., and K. Rayner. “Measuring Word (Unpublished research). New York, NY: Traffic Outdoor Advertising: Does Location
Recognition in Reading: Eye Movements and Audit Bureau for Media Measurement, 2010. Matter?” Psychology & Marketing 28, 9 (2011):
Event-Related Potentials.” Trends in Cognitive Sci- 909–933.
ence 7, 11 (2003): 489–493. Traffic Audit Bureau. Ratings 2.0. (Unpublished
research). New York, NY: Traffic Audit Bureau for Wilson, R. T., and B. D. Till. “Managing Non-
Shinoda, H., M. M. Hayhoe, and A. Shrivastava. Media Measurement, 2014. Traditional Advertising: A Message Processing
“What Controls Attention in Natural Envi- Framework.” In Advertising Theory, S. Rodgers
ronments.” Visual Research 41, 25–26 (2001): Treisman, A. M., and G. Gelade. “A Feature- and E. Thorson, eds. New York, NY: Routledge,
3535–3545. Integration Theory of Attention.” Cognitive Psy- Taylor & Francis Group, 2012.
chology 12, 1 (1980): 97–136.
Tatler, B. W., R. J. Baddeley, and I. D. Gilchrist. Wilson, R. T., D. W. Baack, and B. D. Till.“Crea-
“Visual Correlates of Fixation Selection: Effects Trick, L. M., J. T. Enns, J. Mills, and J. Vavrik. tivity, Attention, and the Memory for Brands: An
of Scale and Time.” Vision Research 45, 5 (2005): “Paying Attention Behind the Wheel: A Frame- Outdoor Advertising Field Study.” International
643–659. work for Studying the Role of Attention in Driv- Journal of Advertising 34, 2 (2015): 232–261.
ing.” Theoretical Issues in Ergonomics Science 5, 5
Taylor, C. R. “Innovation in Traditional Media: (2004): 385–424. Wolfe, J. M.“Visual Search 2.0: A Revised Model
Keeping Our ‘Eyes On’ an Innovative New Meas- of Visual Search.” Psychonomic Bulletin & Review
urement System for Out-of-Home Advertising.” Underwood, G., K. Humphrey, and E. van Loon. 1, 2 (1994): 202–238.
International Journal of Advertising 29, 4 (2010): “Decision About Objects in Real-World Scenes
521–525. Are Influenced by Visual Saliency Before Their Wolfe, J. M. “Visual Search.” In Attention, H.
Inspection.” Vision Research 51 (2011): 2031–2038. Pashler, ed. Hove, England: Psychology Press,
Taylor, C. R.“Editorial: Back to the Future: Some 1998.
Topics We Should Not Forget About in Advertis- Van der Lans, R., R. Pieters, and M. Wedel.
ing Research.” International Journal of Advertising “Competitive Brand Salience.” Marketing Science Yantis, S., and J. Jonides. “Abrupt Visual Onsets
31, 4 (2012): 699–702. 27, 5 (2008): 922–931. and Selective Attention: Voluntary Versus Auto-
matic Allocation.” Journal of Experimental Psy-
Theeuwes, J. “Exogenous and Endogenous Con- Wallis, G., and H. Bülthoff. “What’s Scene and chology: Human Perception and Performance 16, 1
trol of Attention: The Effect of Visual Onsets and Not Seen: Influences of Movement and Task (1990): 121–134.
Offsets.” Perception & Psychophysics 49, 1 (1991): Upon What We See.” Visual Cognition 7, 1/2/3
83–90. (2000): 175–190. Y arbus , A. L. Eye Movements and Vision. New
York, NY: Plenum, 1967.
Theeuwes, J.“Top-Down Search Strategies Cannot Walther, D., and C. Koch. “Modeling Attention
Override Attention Capture.” Psychonomic Bulle- to Salient Proto-Objects.” Neural Networks 19, 9 Young, M. S., J. M. Mahfoud, N. A. Stanton,
tin & Review 11, 1 (2004): 65–70. (2006): 1395–1407. P. L. Salmon, D. P. Jenkins, and G. H. Walker.
“ Conflicts of Interest: The Implication of
Thomas-Smith, C., and Barnett, G. “Seeing Is Wang, Y., B. Mehler, B. Reimer, V. Lammers, L. Roadside Advertising and Driver Distrac-
Believing: Viewing Engagement in Place-Based A. D’Ambrosio, and J. F. Coughlin.“The Validity tion.” Transportation Research Part F 12 (2009):
Media.” In Your Audience = Media Consumer + of Driving Simulation for Assessing Differences 381–388.