 

 

i
52.!
in”

I

i 3...

a 93.11.

éimhpww
-.

4:
.

tau-1!
m... 3H.“
0““
"our“ .1 :

"QM
1‘

gt.

.1:
35}
i3 .

s

.33..
‘.A......ﬂ€1hl
:~-‘. "”1.
I:
83:)? F‘kﬂ¢WtWWiﬁW l!

sir-avg

3‘

.931.
1.1.1:» x
.10”; Int!) Juuﬂﬂﬁey It...
hauw‘ﬁr .lfilnlltr
amt

£71W9 1::

din?! A! rill-5.5.373...

:11“... .A T}:
1” t)l le'l!“
“151.01}

 

 

 

 

2 LIBRARY
2 m z Michigan State
University

 

 

This is to certify that the
dissertation entitled

CONTROLLING DISTRACTION ON THE INTERNET:
AN INVESTIGATION INTO THE MECHANISMS INVOLVED
IN MINIMIZING THE INFLUENCE OF INTERNET ADS ON
AN INFORMATION SEARCHING TASK

presented by

ELIZABETH ANN HELDER BABCOCK

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in PsychologL

 

 

Klara/é

Major Professor’s Signature

46134457’ N, :IMS’

 

Date

MSU is an affirmative-action, equal-opportunity employer

_ _.—.—-.-.-.-.-.-u

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:lProIIAcc&Pres/ClRC/DateDue.indd

CONTROLLING DISTRACTION ON THE INTERNET:
AN INVESTIGATION INTO THE MECHANISMS INVOLVED IN MINIMIZING THE
INFLUENCE OF INTERNET ADS ON AN INFORMATION SEARCHING TASK

By

Elizabeth Ann Helder Babcock

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Psychology

2008

Copyright by
ELIZABETH ANN HELDER BABCOCK
2008

ABSTRACT
CONTROLLING DISTRACTION ON THE INTERNET:

AN INVESTIGATION INTO THE MECHANISMS INVOLVED IN MINIMIZING THE
INFLUENCE OF INTERNET ADS ON AN INFORMATION SEARCHING TASK

By
Elizabeth Ann Helder Babcock

Every day, millions of Internet users encounter advertisements designed to
capture their attention. Often, these advertisements are irrelevant to the task that
the Internet user is trying to complete, and as such, represent a major source of
potential distraction. Previous research indicates that although Internet ads are
effective in capturing Internet users' attention (as measured by their ability to
recognize previously presented ads), individuals engaged in an information search
are surprisingly good at minimizing the processing of ads, to the extent that few, if
any, distraction effects emerge in the presence of ads.

The experiments included in this dissertation were designed to investigate
the extent to which three features of the task environment are used to guide
selective attention and enable Internet users to preserve goal task performance
when distractors are present. It is hypothesized that visual attention can be guided
by perceptual features, semantic content, and / or spatial location.

Experiment 1 investigated the inﬂuence of perceptual and semantic features
by varying the semantic relatedness of ads and the likelihood that images contained
task relevant information. Participants were college aged adults and older adults
(60+). Recognition of images increased for both age groups as their semantic

relatedness to target content increased. No reliable inﬂuence of perceptual features

was found. Furthermore, no distraction effects from the presence of ads were found
in measures of search task performance, indicating that older adults were not
differentially susceptible to distraction in this environment.

Experiment 2 investigated whether reliable ad recognition in the absence of
reliable distraction effects could be attributed to the use of location cues. It was
hypothesized that spatial location could be used to guide attention toward probable
target regions, and avoid subsequent processing in regions containing distractors. In
order to test this hypothesis, this experiment manipulated the extent to which
participants needed to be attentive to the entire area of the web page by introducing
a secondary task to the primary information searching task. This experiment again
provided support for the role of semantic content in guiding attention, but no

evidence of spatial location guiding attention was found.

Dedicated to Case and my family—

If this dissertation '5 length were proportionate
to the amount of love, support and encouragement
you have given me over the years,
this would be a multi-vqume work!

ACKNOWLEDGEMENTS

I wish to thank my advisor, Dr. Rose Zacks for her invaluable mentoring and
training during my graduate studies. I also wish to thank the members of my
dissertation committee, Dr. D. Zachary Hambrick, Dr. Mark Becker and Dr. Richard
DeShon, along with past committee members for helping me deﬁne my research
questions and shape a vague curiosity into a successful line of research—null effects
notwithstanding. Several members of the Cognitive Aging Lab were extremely
instrumental in helping me with data collection and data entry, especially Andrew
Iarosz and Ashleigh Burgess. In addition, I am indebted to Dr. Scott VanderStoep and
the members of the Hope College Psychology Department, who permitted me to use
their facilities for data collection.

In addition, I'd like to thank the following: Mareike Weith and Emily Swensen,
for their friendship and support; Dr. John Shaughnessy, for his’hours of unofﬁcial
advising and mentoring since I left Hope College; my family, for never wavering in
their encouragement; and my husband, Case, who has heard nothing else but
dissertation descriptions from me for months and still manages to patiently and

insightfully comment on my thoughts.

vi

TABLE OF CONTENTS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LIST OF TABLES X
LIST OF FIGURES XII
INTRODUCTION 1
Measuring Attention to Internet Advertising 4
Eye Fixations on Internet Advertisements 4
Memory for Internet Advertisement; 6
Factors Inﬂuencing Attentional Capture 8
Perceptual Features 8
Involuntary Attentional Capture 8
Contingent Attentional Capture 11
Semantic Features 15
Location 17
ADVERTISEMENTS THAT SHARE PERCEPTUAL OR SEMANTIC FEATURES WITH
WEB PAGE CONTENT 21
Preliminary Study 21
Predictions 21
Attention to Advertisements 21
Goal Task Performance 23
Materials and Methods 24
Participants 24
Design 24
Materials 26
Procedure 31
Results 32
Manipulation Check 32
Attention to Advertisements 34
Goal Task Performance 38
Discussion 41
Attention to Advertisements 41
Goal Task Performance 42
Limitations 42
Experiment 1 44
Age Differences in Distraction 44
Older Adults and the Internet 44
Increased Susceptibility to Distraction 45
Factors Attenuating Age Differences in Attention 46
Predictions 49
Materials and Methods 50
Participants 50
Design 50

 

vii

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Materials 52
Procedure 59
Results 61
Manipulation Checks 61
Goal Task Performance 72
Reading With Distraction Task 77
Role of Experience 81
Discussion 84
Attention to Advertisements 84
Goal Task Performance 86
Age Differences 88
Conclusion 89

THE INFLUENCE OF LOCATION-BASED ATTENTIONAL FILTERS ON ATTENTION T0
INTERNET ADVERTISEMENTS 91
Predictions 92
Materials and Methods 94
Participants 94
Design 94
Materials 95
Procedure 99
Results 100
Manipulation Checks 100
Participants 100
Dual Task Performance 102
Reading with Distraction Task 104
Attention to Advertisements 106
Goal Task Performance 109
Web Page Viewing Time 110
Content Learning Questions 111
Discussion 113
Attention to Advertisements 113
Goal Task Performance 115
Conclusions 116
GENERAL DISCUSSION 118
Summary of Experimental Results 118
Role of Perceptual Features 118
Inﬂuence of Semantic Content 120
Inﬂuence of Spatial Location 123
Conclusions 126
Implications for Advertisers 127
Implications for Internet Users 128
APPENDIX A: SCREEN SHOTS FROM THE PRELIMINARY EXPERIMENT .................... 130

Baseline Conditions

130

Ad Conditions in the Uniform Presentation Condition 131

 

 

 

 

 

 

 

Ad Conditions in the Hybrid Presentation Mode ....... 132
APPENDIX B: SCREEN SHOTS FROM EXPERIMENTS 1 & 7 133
Baseline Conditions 133
Ad Conditions in Uniform Presentation Mode 134
Ad Conditions in Hybrid Presentation Mode 135
APPENDIX C: COMPUTER EXPERIMCE QUESTIONNAIRE 136
APPENDIX D: INTERNET EXPERIENCE QUESTIONNAIRE 138
APPENDIX E: EXIT QUESTIONNAIRE 140

 

REFERENCES ..... 141

ix

LIST OF TABLES

Table 1: Image types displayed in each presentation mode in the preliminary study.

 

Table 2: Means (and 95% conﬁdence interval) on the demographic variables

 

included in the preliminary study.

Table 3: Proportion of hits and false alarms (FAs), and their corresponding

discrimination index (dL) for image recognition in the preliminary study. ..............

Table 4: Mean time in seconds (and 95% conﬁdence interval) taken to make a

correct Old/ New decision for images in the preliminary study memory test. ........

Table 5: Mean time in seconds spent viewing web pages (and 95% conﬁdence
interval) in the preliminary study

Table 6: Mean accuracy (and 95% conﬁdence interval) on content learning

questions in the preliminary study ........

Table 7: Image types displayed in each presentation mode in Experiment 1 ...........

Table 8: Counterbalancing of passage topics and image type conditions in
Experiment 1

Table 9: Means (and 95% conﬁdence interval) on the demographic measures
included in Experiment 1.

 

Table 10: Mean time (and 95% conﬁdence interval) in seconds spent viewing web

 

 

25

33

36

38

39

....40

51

52

62

pages in Experiment 1 73
Table 11: Mean accuracy (and 95% conﬁdence interval) on content learning

question in Experiment 1 .---76
Table 12: Scores on the demographic measures included in Experiment 2 ............... 101

Table 13: Average number of gems collected per page in Experiment 2 ..................... 103

Table 14: Mean proportion correct on content learning questions in Experiment 2.
112

 

Table 15: Mean time (in seconds) taken to select a correct answer on the content
learning questions in Experiment 2 112

xi

LIST OF FIGURES

Figure 1: Mean discrimination indices for the images presented in the preliminary
study memory test . 35

Figure 2: Percent of images recognized by (a) young and (b) older adults in the
Experiment 1 image memory test 66

 

Figure 3: Percent of images recognized by participants in the Hybrid presentation
condition on the Experiment 1 image memory test 68

Figure 4: Mean time (in seconds) taken for (a) young adults and (b) older adults to
make a correct decision in the Experiment 1 image memory test 70

Figure 5: Mean time (in seconds) taken to correctly recognize images by
participants in the Hybrid presentation condition on the Experiment 1 image
memory test 71

Figure 6: Mean reading time in the Experiment 1 reading with distraction task ...... 78

Figure 7: The six gem shapes and colors used in the dual task component of
Experiment 2 ‘ 96

Figure 8: Representation of the gem materialization animation sequence in
Experiment 2, with timestamps (in ms) below each image 97

Figure 9: Page layout indicating image positions (open squares) and gem locations
(ﬁlled squares) for all six pages of the Health Information passage 98

Figure 10: Average RT (in seconds) to collect gems in Experiment 2 104

Figure 11: Mean passage reading time (in seconds) on the reading with distraction
task in Experiment 7 105

Figure 12: Proportion of images recognized in the Experiment 2 image memory test.
107

 

xii

Figure 13: Time taken (in seconds) to make a correct decision on the image memory
test in Experiment 2 109

 

Figure 14: Mean time (in seconds) spent viewing website pages in Experiment 2.110

 

 

 

 

 

 

 

 

Figure A1: Screen shot of the ﬁrst page displayed in the text only condition ........... 130
Figure A2: Screen shot of the ﬁrst page displayed in the graphics only

condition 130
Figure A3: Screen shot of the ﬁrst page displayed in the related ads condition ...... 131
Figure A4: Screen shot of the ﬁrst page displayed in the unrelated ads

condition I 131
Figure A5: Screen shot of the ﬁrst page displayed in the related ads condition ...... 132
Figure A6: Screen shot of the ﬁrst page displayed in the unrelated ads

condition 132
Figure B1: First page of the travel passage displayed in the text only

condition 133
Figure B2: First page of the travel passage displayed in the graphics only

condition 133
Figure 83: First page of the travel passage displayed in the related ads

condition 134
Figure B4: First page of the travel passage displayed in the unrelated ads

condition 134
Figure B5: First page of the travel passage displayed in the related ads

condition 135
Figure B6: First page of the travel passage displayed in the unrelated ads

condition

135

xiii

INTRODUCTION

Increasing numbers of Americans are turning to the Internet as a leading
source for information (“America Internet usage and population statistics, ” 2007;
Czaja & Lee, 2001 ; Morrell, Mayhorn & Bennett, 2002; Spiezle & Moulton, 2001;
Thorson, Duffy 8: Schumann, 2007; United States Department of Commerce, 2000).
The Internet offers a wealth of information in a single place, available to many
individuals in the comfort of their own homes. The ability to easily access vast
amounts of information on the Internet has revolutionized the way many people
pursue information gathering tasks.

The Internet is not only unique in that it conveniently offers access to a
nearly inﬁnite amount of information, it is also unique in its ability to deliver
requested information to people side by side with unrequested, often irrelevant
information. This irrelevant information can come in a variety of forms, from search
results that do not meet the searcher's needs, to professionally developed marketing
campaigns designed to attract attention and turn would-be information-seekers into
consumers. The focus of this dissertation is on the latter (and arguably the largest)
category of distractors on the Internet—Internet advertisements.

Advertisers have increased spending on Internet advertising drastically in
the last ﬁve years (annplugged, 2007: Internet Advertising Bureau), resulting in an
increased presence of advertisements on the Internet. Increasing Internet-based
advertising has proven to be lucrative to advertisement agencies. Revenue from

Internet advertising has shown steady growth over the course of the last 10 years

("America Internet usage and population statistics, " 2007), and shows no indication
of slowing down (“Internet advertising revenues grow 35% in ‘06", 2007; Li &
VanBoskirk, 2005). As ad spending increases, so do the number of Internet
advertisements vying for Internet users' attention.

Advertisers recognize that the Internet “is a medium that can uniquely affect
consumer behavior” (“America Internet usage and population statistics, " 2007). The
question is, “What are these unique effects of Internet advertising?" 0n the one
hand, increased spending on Internet advertising indicates that advertisers believe
that the Internet is an effective medium for capturing users' attention and
prompting them to engage in “consumer behavior." On the other hand, the resulting
increase in advertising has, in many cases, lead to public outcry, with critics claiming
that ads are intrusive and ultimately impair Internet users' abilities to use the
Internet as desired. As Barbara Holt and Roger Morrell (2002) describe, “The
majority of web sites are multi layered with advertisements, pop-up screens, reams
of pages of scrolled text, irrelevant graphics, and ﬂashing banners announcing soon-
to-pass purchasing opportunities” (p. 109).

By design, Internet ads are designed to attract the attention of potential
consumers engaged in non-consumer activities on the Internet. Nevertheless, many
of these potential consumers object to having their attention diverted by
advertisements. Since attention is commonly conceptualized as a limited resource, it
seems plausible that Internet users have a valid complaint. Factors that make ads

effective in attracting attention could also make it difﬁcult for Internet users to

efﬁciently carry out the task they intended to complete on the Internet. If Internet
users are unable to guide attention away from ads when attending to ads is not
relevant to the user's goal, these users may have increased difﬁculty attending to
and completing their goal. In this latter case, objections to increased advertising
volume on the Internet would seem justiﬁed.

The following dissertation stems from research investigating the abilities of
Internet users to control distraction as it occurs in the form of Internet
advertisements. I begin by reviewing research that assessed the effectiveness of
Internet advertisements, followed by a discussion of factors likely to contribute to
attention capture by advertisements in an Internet environment. This discussion
focuses on three features hypothesized to guide selective attention: perceptual
features, semantic content, and spatial location.

Studies have indicated that people attend to advertisements enough to later
remember them. Despite this evidence that people tend to remember ads, ﬁndings
of distraction effects (or performance decrements) due to Internet ads can be
characterized as mixed. Although there are some studies that report distraction
effects, in general distraction effects from Internet ads range from small to
nonexistent, especially when the goal task involves complex cognitive processing.
This is a rather surprising result, given evidence elsewhere (e.g. visual search
studies) that participants perform more poorly on a primary task when their
attention is captured by distractors. The experiments presented in this dissertation

were designed to explore the extent to which three features of the task environment

are involved in guiding selective attention and enabling participants to maintain
steady goal task performance.

It is hypothesized that Internet users may utilize distinctive perceptual
features, such as color, shape or size, to guide attention to target content. In
particular, it is predicted that users will be more likely to process items that share
perceptual features with the goal task. However, as an information searching task
requires that users process content for meaning, it is also proposed that Internet
users are likely to direct attention to items that are considered topic-relevant. Users
are predicted to process all web content for meaning, and ultimately make a
determination as to whether an individual item is task relevant or task irrelevant.
Attention can then be directed to task relevant objects for further processing.
Finally, Internet users could utilize spatial cues to direct attention away from items
deemed irrelevant and toward the semantically related, task relevant regions of a
page. I

Measuring Attention to Internet Advertising

The relatively young ﬁeld of research on Internet advertising indicates that
Internet ads are generally effective in capturing consumers' attention, as measured
by eye tracking technology, and various types of memory tests.

Eye Fixations on Internet Advertisements
There is generally assumed to be a direct link between what the eye is

looking at and ongoing cognitive processing (Hollingworth & Henderson, 2002 ;

Rayner, 1998).1 Eye-tracking studies have shown that participants engaged in a web
page reading task divert their attention from reading in order to look at task-
irrelevant ads (Burke et al., 2005; Day, Shyi, & Wang, 2006; Dreze & Hussherr, 2003;
Lewenstein, Edwards, Tatar, & DeVigal, 2004; Wang & Day, 200 7). In one such study,
67 participants who regularly visited Internet news sites were recruited, and asked
to do their news reading on web-pages of their choosing while wearing a head-
mounted eye tracker (Lewenstein et al., 2004). The investigators found that
participants’ eyes ﬁxated for an average of 1000ms on 45% of the displayed banner
ads. In this experiment, a total of 1,127 banner ads were displayed on 97 of the web
pages accessed by participants. Dreze and Hussherr (2003) report similar results. In
their study, all of their 49 participants directly ﬁxated at least one of the 8 presented
banner ad. On average, each banner ad had a 0.49 probability of being seen, or each
person looked at 3.96 banner ads.

A study by Burke and colleagues (Burke et al., 2005) reports that eye
ﬁxations occurred directly on commercial advertising banners in approximately
13% of the trials. This study used a visual search paradigm in which participants

were either cued with a news headline (verbatim match) or the ﬁrst sentence of a

 

1 Burke et al. (Experiment 1, 2005) claim that 94% of banner ads correctly
identiﬁed on their recognition test were not directly ﬁxated by the participant.
The authors do not discuss the signiﬁcance of this ﬁnding in terms of calling into
question the link between attention and eye movements, nor do they offer an
explanation for this extremely high percentage of unviewed, but recognized, ads.
Furthermore, as discussed in the following section on memory for ads, these
researchers did not obtain reliable recognition of advertiSements in their
experiment. As such, the implication and reliability of this ﬁnding is difﬁcult to
interpret.

news article (semantic match), and asked to ﬁnd the matching headline in a search
set of news headlines. The participants in this study were found to be twice as likely
to look at banner ads when they were engaged in effortful processing in the
semantic match condition, as opposed to merely matching the exact headline in the
verbatim match condition.

Finally, two recent studies by Day and Wang (Day, Shyi, 8: Wang, 2006; Wang
& Day, 2007) also record participant eye ﬁxations on banner ads. However, these
researchers report their ﬁndings in terms of percent of total ﬁxations directed
toward ads, rather than in terms of the proportion of ads ﬁxated. In their ﬁrst study
(Day, Shy & Wang, 2006), participants viewed two grayscale ads, positioned at the
top and the bottom of a matrix decision making task, that were set to alternately
onset and offset to create an impression of ﬂashing. Participants in this experiment
directed 3.62% of their ﬁxations toward these banners. In their second study (Wang
& Day, 2007), they report that ﬁxations on small square ads located in each corner of
the screen, and ﬁxations on banner ads located at the top and the bottom of each
screen accounted for just under 3% of total ﬁxations. These ﬁndings indicate that
although advertisements are looked at, they are not looked at as often as target
content on a website.

Memory for Internet Advertisements

Another measure frequently used as an indicator of attention to

advertisements is memory for a brand name, or memory for the ad itself. Studies

using these measures report that consumers are able to recall and /or recognize

previously viewed ad content (Bayles, 2000; Bayles, 2002; Calisir & Karaali, 2008;
Chung, 2006; Dahlen, 2001 ; Danaher & Mullarkey, 2003; Diao 8: Sundar, 2004; Dreze
8: Hussherr, 2003; Edwards, Li, 8: Lee, 2002; Hong, Thong, 8: Tam, 2004; Johnson 8:
Neath, 1999; Iones, Pentecost 8: Requena, 2005; Li 8: Bukovac, 1999; Moore,
Stammerhohan 8: Coulter, 2005; Sagarin, Britt, Heider, Wood, 8: Lynch, 2003; Sundar
8: Kalyanaraman, 2004; Yoo 8: Kim, 2005, but see Burke et al., 2005 Experiment 2 for
ﬁndings of unreliable ad recognition). Although both recall and recognition can be
used to measure attention to ads, there is some indication that recognition tests
provide more sensitive measures of ad memory than recall tests (Bayles, 2000;
Bayles, 2002 ; Danaher 8: Mullarkey, 2003; see Neath & Surprenat, 2003 for a more
extensive discussion of the differences between these two types of memory tests),
as recall tests for incidentally processed stimuli such as advertisements are more
prone to ﬂoor effects than recognition tests.

Few studies in this literature compare memory for non-advertising content
with memory for ads. The available studies provide clear indication that people
appear to allocate most of their attention to the content of the website, and not to
ads. Two studies indicate that although memory for brand names presented in
Internet advertisements is reliable, it is signiﬁcantly worse than memory for a web
site’s non-advertising content (johnson 8: Neath, 1999; Iones, Pentecost 8: Requena,
2005). This pattern is consistent with eyetracking studies that ﬁnd only a low
percentage of total eye ﬁxations being directed toward ads, with the majority of

ﬁxations being directed toward content (Day, Shyi 8: Wang, 2006; Wang & Day,

2007). Despite this apparent focus on web content, repeated ﬁndings of reliable
memory for advertisements does indicate that people (willingly or unwillingly) are
allocating some of their attentional resources toward processing ads.
Factors Inﬂuencing Attentional Capture

The amount of attention allocated to Internet advertisements is likely
inﬂuenced by several factors, such as the perceptual features of an advertisement,
whether or not the content of the advertisement relates to information that the
Internet user is seeking, and the spatial location of items on the web page.

Perceptual Features

An ongoing debate in the visual attention literature centers the degree to
which attention capture by perceptually distinct features in the visual display is
automatic. Some researchers have proposed that these features capture attention
automatically, without effort or intention on the part of the viewer: Such automatic
attentional capture is sometimes referred to as involuntary attentional capture.
Other researchers have proposed that participants' current attentional set directs
the likelihood that visually distinctive features will capture attention, often called
contingent attentional capture. According to these researchers, visual salience is
determined in part by the perceptual features of an item, and in part by the task
demands. These two views, as they apply to processing Internet ads, are discussed
in the following sections.
Involuntary Attentional Capture

A recent review of the visual search literature by Wolf and Horowitz (2004)

suggests that some perceptual features of a target stimulus are associated with very
efﬁcient visual searches. Searches are termed efﬁcient when the search slope is not
dependent upon the number of items in the search set—in other words, some
distinct or unique feature of the target object causes it to pop-out from the ﬁeld of
distractors and immediately become the object of attention. Features that
undoubtedly guide attention in this way include color; motion, orientation and size
(Wolf & Horowitz, 2004). The efﬁciency with which these features are able to guide
attention suggests that these features may be available pre-attentively. In keeping
with this notion, some researchers have suggested that attention capture by these
distinct perceptual features may be automatic.

Since advertisements are often more colorful than their surrounding page
content, and employ animation (motion) in comparison to static content elements, if
attention capture by these features is automatic advertisements should be highly
salient and should show evidence of being processed regardless of their relevance to
the goal task. A series of studies investigating the inﬂuence of Internet visually
salient advertisements on an information search task found indication of attention
to ads, as measured by ad recognition (Helder, under revision). However; although
participants in these studies showed evidence of having attended to ads, recognition
rates were quite low, and no performance decrements (as measured by reading rate
and content learning) were observed. These last two ﬁndings do not seem to be
consistent with involuntary attentional capture—rather, they suggest that

participants were able to avoid attention capture from visually salient features of

10

advertisements some of the time. Similar ﬁndings have been observed elsewhere in
the Internet advertising literature. In a study in which ads should have been visually
salient compared to text-based content, Sagarin, Britt, Heider, Wood, and Lynch

(2003) found no direct inﬂuence of Internet ads on the number of anagrams solved.

On the other hand, Burke and colleagues (Experiment 2; 2005) found results
that appear to support automatic attention capture when the search task requires
fewer cognitive resources, but not when the search task involves more complex
processing. In their study, the time taken for participants to complete a visual search
task containing graphics of varying salience was recorded. They found a distraction
effect attributable to salient (animated) ads when the task was a simple search task
involving matching headlines in a search set to a target pre—cued headline. However;
when the task became more cognitively demanding by cuing participants with an
opening sentence from an article, and asking them to choose the most likely
headline from the search set, the visual salience of the ad no longer inﬂuenced
search speed.

In conclusion, this pattern of results may suggest that distinctive perceptual
features on distractor ads are unlikely to automatically capture attention when
individuals are involved in cognitively demanding tasks, such as processing a verbal
for meaning. This interpretation is intuitively appealing, as performance on the task
itself is likely to be more strongly tied to semantic than basic perceptual processing.
On the other hand, involuntary attentional capture may occur in highly constrained

visual search type tasks, in which the task itself is deﬁned on the basis of perceptual

11

features.
ContingentAttentional Capture

In the information search tasks (Burke et al., 2005; Helder, under revision)
and anagram solving tasks (Sagarin et al., 2003) discussed in the previous section,
the bright colors and the use of motion in advertisements were inconsistent with
the set of perceptual features considered task-relevant. This set of features is
frequently termed the “attentional set". Participants in these experiments were
engaged in tasks in which the target information was presented in black text on a
white background. Therefore, any objects of a graphical nature, such as the
advertisements, should not have shared perceptual features with the attentional set.
The ﬁnding that participants were able to ignore advertisements in these paradigms
may indicate that attentional capture is contingent upon attentional set. Further
support for the contingent attentional capture hypothesis comes from studies on
attentional set and inattentional blindness outside the domain of advertising
research.

Inattentional blindness occurs when individuals fail to see things that are
present at the attended location. In perhaps the most famous of inattentional
blindness studies, 50 % of participants failed to spot a person in a gorilla suit
walking through the middle of a basketball game (Neisser 8: Becklen, 1975 ; Simons
8: Chabris, 1999). Since this study, experimenters have developed displays with
carefully controlled visual characteristics to investigate the source of inattentional

blindness. Several of these studies suggest that the degree of inattentional blindness

12

is determined by how similar the perceptual features of a distractor are to those in
the current attentional set (Most et al. 2001; Most, Scholl, Clifford 8: Simons, 2005).
These ﬁndings are consistent with contingent capture views of attention, which
propose that distractors sharing features with the current attentional set (meaning
they share features with the target) are likely to be noticed, while distractors that do
not match the current attentional set are not likely to be noticed.

This basic pattern of results holds even when the distractor should be salient
based on its color, movement, or size (Most, et al., 2001), contrary to the predictions
of involuntary attentional capture advocates. For instance, if participants are
attending to the motion of white objects and ignoring the motion of black objects,
even a visually salient red object is likely to go unnoticed by a large proportion of
participants because the red object does not share perceptual features with the
current attentional set. In summary, inattentional blindness studies have
successfully dissociated between focusing attention on the basis of what should be
distinctive visual features and focusing attention on the basis of attentional set. In
the case of attention to Internet ads, this view predicts that ads that do not share
perceptual features with the current attentional set may not be attended.

Similar results have also been found in experiments involving stimuli more
characteristic of those encountered on the Internet. The ﬁrst of these studies is a
visual search study that manipulated the degree of perceptual similarity between
distractors and targets. The author found that the degree of perceptual similarity

between goal task content and distractor content appears to be a strong predictor of

13

distractor inﬂuence (Zhang 2000).

In this task, participants were presented with a matrix consisting of
nonsense letter strings ranging from 1 to 4 letters in length. At the beginning of each
trial, participants were given a target string to search for. Trials ended when
participants indicated how many instances of the target string they had found in the
matrix. While participants were searching, an animated string of nonsense letters,
or an animated picture of an animal was presented around the periphery of the
search display. The critical finding was that the animated letter string (which shared
perceptual features with the target letter strings) was more disruptive than the
animated animal. As such, this study provides support for the idea that when
distractors match features of the current attentional set, they are likely to receive
more processing than distractors that do not match.

Two other studies (Hong, Thong 8: Tam, 2004; Miarmi 8: DeBono, 2007)
investigated the inﬂuence of Internet advertisements on goal task performance, and
report results that could be interpreted as supporting the role of attentional sets in
predicting distraction effects. Hong and colleagues (2004) investigated the inﬂuence
of “animation" on an Internet shopping task. They found than when the target item
was animated (ﬂashed), response time to ﬁnd the target decreased, supporting the
idea that animation increases visual salience. This increase in visual salience
purportedly not only captures attention initially, but leads to subsequent processing.
Furthermore, response time was signiﬁcantly longer when a non-target item was

animated compared to conditions without any non-target animation. Although

14

details on the target and non-target products were not provided, non-target items
were likely to share features with the target item since participants were engaged in
an online shopping task. Consequently, the animation may have attracted attention
initially to the non-target item, and the non-target may have received additional
processing since it shared perceptual features with the current attentional set.

Another study of distraction effects from ads found evidence that
advertisements appeared to disrupt goal task performance (Miarmi 8: DeBono,
2007). Participants in this study were asked to determine an appropriate sentence
length for a guilty offender. The researchers found that participants’ decisions were
affected by racial stereotypes when ads were present during the presentation of
evidence. In contrast, participants did not seem to rely on racial stereotypes when
no ads were present. These ﬁndings were interpreted as evidence that the presence
of ads diverted attention away from the sentencing task, resulting in the emergence
of racial stereotypes either as a mental shortcut, or as a suppression error when
reduced attentional resources failed to correct for the inﬂuence of native
stereotypes.

The central manipulation in this study involved showing participants a photo
of the offender, pictured as either a white male or a black male. Given this task, other
content containing human faces are likely to be considered task relevant. Although
the actual ads used in this study are not published, the authors’ description of ad
content indicates that at least half of the ads included in this study contained faces.

Therefore, the ads included in this study may have been processed as task relevant

15

information, resulting in the observed distraction effects.
Semantic Features

Increasingly, advertisers are harnessing the'power of search engines to make
sure that the ads appearing alongside Internet users’ search results are relevant to
the topic they are searching for (Li & Leckenby, 2007). There is some evidence from
research on selective attention supporting this marketing decision. Several studies
have indicated that attentional sets may also be sensitive to semantic matches
(Edwards et al., 2002; Koivisto 8: Revonsuo, 2007; Moore, Stammerjohan 8: Coulter,
2005; Shamdasani et al., 2001), and distractors that are semantically related to the
goal task are more likely to be processed than distractors that are semantically
unrelated.

Koivisto and Revonsuo (2007) report this basic effect, where items that were
semantically related, but did not share any. visual features with the target, were able
to capture attention. In one version of their inattentional blindness experiment,
participants were shown pictures of items arranged around a central ﬁxation point.
Four items appeared onscreen at once, with one item in each quadrant of the screen.
The task was to ﬁnd the picture of an animal (piece of furniture), and indicate in
which quadrant it appeared. On the critical trial, a distractor word naming an animal
(piece of furniture) appeared at ﬁxation. When the word was task relevant (ex. the
goal task was to look for animal picture, and the distractor word was “horse”), the
word was noticed signiﬁcantly more often than when the word was task irrelevant

(ex. the distractor word was “desk"). They replicated these results using an alternate

16

version of the experiment in which the distractor stimulus was a picture, and the
search display contained words. This study suggests that participants engaged in
reading a passage on a topic are likely to attend to images that are related to the
content—even if the images are not part of the goal task.

Additional evidence for semantic similarity inﬂuencing attention comes from
ﬁndings related to advertising effectiveness in an Internet environment.
Researchers have found that consumers have better attitudes toward aids and
brands that are related to the content of the web page as opposed to unrelated
(Moore, Stammerjohan 8: Coulter, 2005; Shamdasani, Stanland, 8: Tan, 2001).
Consumers also reported a higher intention to click on the related ads, or purchase
the products advertised in related ads. Similarly, Edwards, Li and Lee (2002) found
evidence that pop-up ads that related to the reading content were judged to be less
intrusive, but potentially more attention-grabbing.

Finally, Diaper and Waelend (2000) report results that seem to contradict the
proposed role of semantic similarity in capturing attention. In their experiment,
participants were asked to search short informational web pages for the answer to a
question appearing at the top of the web page. Search time was recorded, and web
pages were presented with and without animated, clip-art style graphics that were
related to the content of the web page. Contrary to expectations, they found no
decrements in search performance due to the presence of these content-related
graphics. However, the authors point out that despite being content-related, the

graphics never contained task-relevant information, and that participants were

17

likely able to learn this relationship quickly during the experiment. Consequently,
the participants in this experiment may have implemented attentional ﬁlters based
on perceptual features, rather than semantic features, in order to distinguish target
text from irrelevant graphics, as has been proposed to explain similar results in an
information search task (Helder, under revision).
Location

There is evidence that selective attention can be guided by location, such that
distractors in unattended locations are less likely to be processed. For instance, the
effectiveness of onset distractors in capturing attention is greatly reduced when
participants know in advance where a target will appear (Gronau, Sequerra, Cohen,
8: Ben-Shakhar, 2006; Lamy 8: Tsal, 1999; Theeuwes, 1991; Yantis 8: jonides, 1990).
This is relevant for predicting attention to ads, as task-relevant content often
appears in a predictable location near the center of the screen. However, in some
cases, distractors in unattended locations have been shown to still capture attention
(Beck & Lavie, 2005; Gronau, Sequerra, Cohen, & Ben-Shakhar, 2006), and in other
cases, distractors in attended locations have failed to attract attention, resulting in
inattentional blindness (Most et al., 2001; Most, Scholl, Clifford & Simons, 2005;
Neisser & Becklen, 1975; Simons 8: Chabris, 1999).

According to location-based models of selective attention, advertisements
that appear in close proximity to the attended area should be more disruptive than
advertisements that appear in more peripheral locations. This prediction was

directly tested in a study examining the inﬂuence of Internet ads on reading

18

performance in younger and older adults (Helder, under revision). In this
experiment, participants read short passages without any ads displayed, with ads
displayed around the periphery of the screen, or with ads included in the central
text region of the screen and passage text wrapping around the ads.

Contrary to the predictions of location-based models, neither older nor
younger adults showed any difference in content learning or in reading rate
between the central and the peripheral ad conditions. Furthermore, young adults
showed no evidence of any ad-related distraction effects in measures of reading rate
and content learning. The older adults read the passages containing ads (in either
central or peripheral locations) more slowly than passages without ads, but showed
no distraction effects on a measure of content learning.

Although location-based ﬁlters did not seem to be employed in this study, it
is important to point out that they may not have been necessary, as the ads used in
this study were unlikely to share perceptual or semantic features with the goal task
stimuli. In this study, target content was comprised of black font printed on a white
background, characteristics likely to be visually distinct from colored, graphical ads.
Furthermore, participants were reading passages written to be entertaining
descriptions of events in different characters’ days, which were not likely to share
semantic characteristics with advertisements for products or services.
Consequently, participants may have used perceptual or semantic features, rather
than location cues, to guide selective attention in this experiment.

However, in her visual search study, Zhang (2000) also failed to ﬁnd support

19

for a location-based model of attention. In her study, animations sharing perceptual
features with the letter strings involved in the primary task (animated letter strings)
as well as animations not sharing such features (animated animals) both appeared
in identical peripheral locations. Location-based models predict that both types of
animations should have been equally disruptive to performance on the string
counting task. Instead, similarity between the goal task and features of the
distractor appears to have been a stronger predictor of distractor inﬂuence.

These studies provide evidence that although it seems feasible for location to
guide selective attention in an Internet environment, location-based ﬁltering may be
likely to occur after perceptual and semantic ﬁltering. In addition, it has been
pointed out that web pages tend to be processed as a Gestalt (Moore, Stammerjohan
8: Coulter, 2005). In keeping with this characterization, it is likely that Internet users
who do employ location-based ﬁlters may select locations based on the overall
context of the page. This is type of ﬁlter is proposed to function similarly to the
ﬁlters used in visual marking studies.

In visual marking studies, some of the distractors appear onscreen slightly
before the rest of the display appears. In these experiments, participants are able to
efﬁciently exclude previously presented distractors from the search on the basis of
their location (Watson 8: Humphreys, 1997; Watson, 8: Maylor, 2002). In the case of
attention to Internet ads, ads do not generally appear on the page before the text on
most web pages. However, the participants in visual marking studies show

sensitivity to the overall location of distractors in the context of the entire page

20

layout. This ability is proposed to generalize to Internet environments, enabling
participants to visually mark the locations of objects determined to be task
irrelevant. Therefore, once an image has been determined to be task irrelevant,
participants may be able to guide selective attention away from that image by

visually marking that image's location in the context of the current web page.

21

ADVERTISEMENTS THAT SHARE PERCEPTUAL OR SEMANTIC FEATURES WITH
WEB PAGE CONTENT

Previous research indicates that in some cases, attention to ads appears to be
linked to performance disruptions, or distraction effects (Burke et al., 2005
[Experiment 2]; Miarmi 8: DeBono, 2007; Zhang, 2000), but not in all cases (Burke et
al., 2005 [Experiment 1] ; Diaper 8: Waelend, 2000; Helder; under revision; Hong,
Thong 8: Tam, 2004;). An explanation for this contradictory pattern of results has
yet to be proposed. One possibility is that attention to ads (and by extension, the
likelihood of ad-related distraction effects) can be predicted by the extent to which
the ads resemble the target content either perceptually or semantically.

None of the previously reviewed studies systematically varied both the
perceptual and semantic similarity of targets and distractors to study the interplay
between these two factors. The studies presented in this section were designed to
address this shortcoming. Results from a preliminary study are presented, followed
by the results of Experiment 1. Experiment 1 was designed to replicate the
preliminary study using more diverse material, and a more diverse participant
population by including older adults as well as college-aged adults.

Preliminary Study
Predictions
Attention to Advertisements
It is predicted that the current experiment will replicate previous ﬁndings in

which participants show evidence of attending to ads, even when the ads are task

22

irrelevant. All experiments reported in this dissertation use the results of an
unexpected image recognition test as a proxy for attention. It is assumed that for an
item to be correctly recognized, it must have been processed, and for it to have been
processed, it must have been attended.

In the reported experiments, participants' primary task was an information
search task. All target information was text-based, although participants were not
informed of this feature of the task environment. In addition to the target text, web
pages contained non-ad, semantically related graphics, semantically related ads,
and semantically unrelated ads. It is predicted that participants will show evidence
of processing all of these content items for meaning, and will guide attention to the
items based on the degree to which the image's content matches the target content.
Consequently, graphics should be recognized better than related ads, which in turn
should be recognized better than unrelated ads.

Even though ads should be visually salient on the web page, as compared to
the target text, it is predicted that participants are not likely to show evidence of
involuntary attentional capture by these features. If visually salient features are
involuntarily capturing attention, no difference should emerge in the recognition
rates of graphics, related ads, and unrelated ads. This result, if obtained, would be
contrary to the predicted effect of semantic content.

It is, however, predicted that participants may show evidence of contingent
attentional capture. Since participants were not informed that the targets of their

information search would not be contained in graphics, it is assumed that

23

participants are likely to view semantically related non-ad graphics as potentially
task-relevant items. Therefore, when graphics appear on a page, it is possible that
participants may include perceptual features common to both graphics and ads in
their attentional set. A pattern of results in which ads are more likely to be attended
when they appear on a page that contains graphics along with text, than when they
appear on a page that only contains other ads and text would be interpreted as
supporting the contingent capture hypothesis.
Goal Task Performance

Previous research has found mixed results when it comes to distraction
effects from ads—however, if only the results of studies using information searching
tasks are considered, virtually no distraction effects from ads have been found
(Burke et al., 2005 [Experiment 2, semantic search condition]; Diaper 8: Waelend,
2000; Helder; under revision). Accordingly, it is predicted that very few, if any,
distraction effects on measures of goal task performance will emerge in the current
experiment. However; if participants are assumed to use both semantic content and
perceptual features to guide attention, it is predicted that participants should be
least likely to show distraction effects from ads when viewing pages containing only
content-unrelated ads. This prediction is based upon the idea that a page containing
only ads should not result in the participant including perceptual features unique to
images in the attentional set. Furthermore, semantically unrelated ads are unlikely
to be processed beyond the level required to determine their basic meaning. On the

other hand, participants are predicted to be most likely to show distraction effects

24

when viewing pages containing a mix of content-related ads and graphics, as the
inclusion of graphics should activate an attentional set containing perceptual
features common to graphics and ads, while the semantically related content should
be considered relevant for further processing.
Materials and Methods

Participants

Ninety-one students enrolled in a psychology course at MSU, ranging in age
from 19 to 29 (M= 19.36, CI=18.76-19.81) participated in the experiment. The
experiment was conducted in two phases. The 45 participants (15 males, 30
females) in the first phase of the experiment were tested in a “Uniform"
presentation condition, and the 46 participants (7 males and 39 females) in the
second phase of the experiment were tested in a “Hybrid” presentation condition.2
Design

There were three different types of images used in this experiment: graphics,
related ads, and unrelated ads. All “graphics” were non-ad images related to the
content of the text. The graphics did not contain additional unique information
beyond that provided in the text. Consequently, participants did not need to look at
the graphics to successfully complete the task. This control was necessary in order
to keep the total amount of task-relevant information on each page constant,
regardless of the type of image presented on the page. In addition, this design

enabled a comparison to be made between participant responses to graphics and

 

2 Three participants were tested in the Uniform presentation condition following
the conclusion of testing on the Hybrid presentation condition.

25

ads. All ads used in this experiment were color advertisements containing a mix of
text and images. Some of the ads were animated, some were static. “Related ads”
were advertisements whose content related to the text, while “unrelated ads” were
advertisements whose content was unrelated to the text. As with the graphics,
participants did not need to look at either type of ad to successfully complete the
task.

Image type was manipulated within subjects. Each participant was presented
with one passage in each of four image conditions: Text Only, Graphics Only, Related
Ads, and Unrelated Ads. The passage presented in the Text Only condition contained
only text and no images. The passage presented in the Graphics Only condition
contained text mixed with graphics. These two conditions served as baselines,
enabling the comparison of web page viewing behavior with and without graphics
to the behavior observed in the ad-present conditions.

Table 1: Image types displayed in each presentation mode in the preliminary study.

Image Type

 

 

Texﬁnly “GraTJh—ics Only Related Ads Unrelatedjtds—

No graﬁs I —_1_-l4graphics 7 I 1:3—rElate'd—ads 71-3 uanlated

 

l

Hybrid Uniform

or ads ads

No graphics 1-4 graphics 1 related ad 1 unrelated ad

Presentation .

or ads 1-2 graphics 1-2 graphics

26

As shown in Table 1, the allocation of images to the Text Only and Graphics
only conditions was identical for participants in both phases of the experiment. On
the other hand, the allocation of images to the ad-present conditions varied across
the two phases. Image type was randomly selected for each participant for each
page of a 20 page website. Each participant saw ﬁve pages displayed in each of the
four image types.

Materials

Web page design and content questions. Participants viewed a 20 page
passage on the topic of interpreting health news as reported by mass media
channels. Individual pages contained an average of 246 (SD=14; range=218-274)
words. Each page of the passage was displayed with one of the four image types.
Individual pages were randomly assigned to image types for each participant.
Efforts were made to use styles and coloring consistent with professionally
developed websites. A total of 96 images was used across the various website
display conditions.

A total of 34 graphics was selected for use in the Graphics Only condition.
Pages displayed in this condition contained between one and four graphics, with six
pages displaying a single graphic, 13 pages displaying two graphics, and one page
displaying four graphics. Each ad-present condition used 31 ads, for a total of 62
ads. Pages displayed in the Related Ads condition contained ads related to health
care. Pages displayed in the Unrelated Ads condition contained ads drawn from a

variety of different categories. While most of the images a participant saw were

27

presented on only one page, there was one of each type of image that could appear
on up to three different pages for an individual participant, depending on the image
conditions to which these three pages were assigned for that participant.

The Text Only and Graphics Only pages were identical for participants in the
Uniform and the Hybrid presentation conditions. Participants in the Uniform
presentation condition viewed each ad-present page of the passage with only
related ads or unrelated ads present. In this presentation condition, pages displayed
in the Related Ads condition contained between one and three ads, with eight pages
displaying a single ad, 11 pages displaying two ads, and one page displaying three
ads. The allocation of ads to passage pages in the Unrelated Ads condition mirrored
that of the Related Ads condition, to control for the total number of ads per page.

Participants in the Hybrid presentation condition viewed ads mixed with
graphics. The Hybrid Related Ads and Unrelated Ads pages were created from the
page designs used in the Graphics Only condition, as this was’an existing baseline
condition. As in the Uniform condition, the allocation of ads to passage pages in the
Unrelated Ads condition mirrored that of the Related Ads condition to control for
the total number of images per page. On nine of the Graphics Only pages, a single
advertisement was added to the existing graphic layout, to increase the total
number of images on the page by one. On the remaining 11 pages of the Graphics
Only passage, the mix of graphics and ads was created by swapping one of the

graphics on the page for an ad, resulting in no change in the total number of images

per page.

28

This method resulted in two changes from the Uniform condition. First,
participants in the Hybrid condition saw more images in the passage (Uniform: M:
25.35, SD=1.90; Hybrid: M: 31.51, SD=1.16). Secondly, a greater proportion of the
total number of images displayed in the Hybrid condition were graphics than in the
Uniform condition (Uniform: 55.05%; Hybrid: 68.26%). The set of ads displayed in
the Uniform condition, but removed from the Hybrid condition, were not
counterbalanced. Screen shots from the two baseline conditions (Text Only and
Graphics Only), and the four ad-present conditions (Uniform: Related Ads and
Unrelated Ads; Hybrid: Related Ads and Unrelated Ads) are presented in Appendix
A.

The time spent by each participant on each page of the passage was recorded
in milliseconds by the computer to enable comparisons in viewing time across
conditions. In addition, participants were asked to complete two short answer
questions after they ﬁnished viewing each of the even numbered passage pages
(after page 2, page 4, page 6, etc.). Each question was designed to assess content
presented on one of the previous two pages. Questions ranged in difﬁculty from
11% correct (“What is the relative difference between $50. 00 and $100.00?”; 100%)
to 95% correct (“What sort of funding does the website www.medlineplus.gov likely
receive? ”; government).

Similarly, the set of acceptable answers for individual questions ranged from
a single correct answer (ex. “TRUE or FALSE ?: A study that reports effectiveness of a

medication using a range is less accurate than a study that reports effectiveness using

29

a single number. ”; Accepted: FALSE) to several potentially correct answers (ex. “List
at least two things you should lookfor in a website to make sure you can trust the
health information they are reporting”; Accepted: Can you easily see who sponsors
the website?; Is the sponsor a Federal agency or a medical school, or is it related to
one of these?; Can you ﬁnd the mission or goal of the sponsor of the website?; Can
you see who works for the agency or organization and who is the author? Is there
contact information?; Can you tell when the information was written?; Is your
privacy protected?; Does the website make claims that seem too good to be true?
Are quick, miraculous cures promised?). The author scored each participants'
answers for accuracy to provide a measure of content learning. Responses were
scored as either correct or incorrect—no partial credit was awarded. Questions that
were missing an answer were scored as incorrect.

Memory Test for Images. A memory test containing all 96 images selected for
the experiment was considered to be too long. Consequently, the images were
divided across four forms to create four different 26-item memory tests, one of
which was randomly assigned to each participant. As previously mentioned, three
images (one graphic, one related ad, and one unrelated ad) had the potential to be
"repeated” during the presentation portion of the experiment. Since the images
were included in the design of multiple pages, it was possible (although not certain)
that a participant could be exposed to that image multiple times, depending on the
image condition under which the pages containing the repeated image were viewed.

These three images were included on all four of the memory test forms to make sure

30

that each participant's memory for the image was tested. The remaining images
were equally divided across the four test forms. These images were split
approximately equally among the three image categories (graphics, related ads, and
unrelated ads). In total, three of the four forms contained 9 graphics, 9 related ads,
and 9 unrelated ads. The fourth form included 11 graphics, and 7 of each type of ad.

Participants were presented with images displayed one at a time on the
computer screen. Participants were asked to decide, for each image, whether they
recognized it from the web page viewing task (old), or whether the image was new
(new). Although the number of new images on the memory test varied by
participant (M= 18.01, SD=2.19, range=11-2 6), the memory test contained more
new images than old images on average. This bias reﬂected the bias in the overall
construction of the experiment.

On average, participants were exposed to approximately 28 images during
the web page viewing portion of the experiment, which means that the average
participant viewed less than 30% of the possible images. The random assignment of
pages to image conditions for each participant meant that the memory test would
either need to be custom-created for each participant in order to balance new and
old items, or else reﬂect this overall bias of new/ old images. For the purposes of a
preliminary study, the latter was judged acceptable. Accuracy in recognizing the
images presented on the memory test was recorded. The time taken (in ms) to
record each decision was also recorded.

Questionnaires. Participants completed a brief demographic questionnaire

31

generated by the author to assess participants' gender, age, and the number of years
of formal schooling they had completed. In addition, participants completed a 15-
item vocabulary test (shortened from Zachary, 1986), a computer experience
questionnaire (excerpted from a questionnaire used by Kubeck, Miller-Albrecht, and
Murphy, 1999; Appendix C) and an Internet experience questionnaire (generated by
the author; similar to the one in Appendix D). In addition to providing an overall
index of Internet experience, three questions from this questionnaire were used to
measure attitude toward Internet ads, and two additional questions were used to
measure amount of self-reported distraction from ads.

Procedure

Participants were tested in small groups, ranging from one to six participants
at a time. Each participant was seated at his or her own computer station.
Participants worked through the experiment independently, although the
experimenter monitored their progress and was available for assistance in
navigating the web pages.

The experiment began with an informed consent procedure, after which
participants were asked to enter their demographic information. They were then
given instructions on the website viewing task. Participants were told to do their
best "to read the article carefully and to learn the material being presented” as they
would "be asked to answer several questions testing [their] comprehension of the
material presented in the article." Once participants completed reading the

instructions, they were taken to the ﬁrst page of the website, and began the website

32

viewing task. Participants answered two content questions after every other
completed page. Participants were not informed of the question content prior to
viewing the passage, and could not return to a previous page once they had left it.
After participants had ﬁnished viewing the website, they ﬁlled out the computer
experience questionnaire, the vocabulary test, and the Internet experience
questionnaire. All three of these forms were administered via the computer.

The ﬁnal task in the experiment was the memory test for images.
Participants were instructed that they would be shown images on the following
screens, and that some of the images would be “advertisements or illustrations that
appeared in the article while [they] were reading it.” Participants were not told
ahead of time that their memory for images displayed in the experiment would be
tested, and were encouraged to guess if they were unsure of the correct response to
an image. The memory test did not advance to the next image until the participant
had indicated a Yes or No decision. After all testing was completed, participants were
provided with a written description of the experiment, and given the opportunity to
ask questions about the purpose of the experiment.

Results

Signiﬁcance testing was carried out with (1:05. Error bars on graphs indicate

the 95% conﬁdence interval around the mean.
Manipulation Check
Assignment of participants to presentation condition was blocked.

Consequently, the ﬁrst group of participants tested were assigned to the Uniform

33

condition, and the second group of participants were assigned to the Hybrid

condition. To ascertain that these two samples were equivalent, participants in each

presentation group were compared on several demographic variables.3 Summary

data on these measures are displayed in Table 2.

Table 2: Means (and 95% conﬁdence interval) on the demographic variables
included in the preliminary study.

Age
Yrs. Education
Vocabulary 6‘
Computer Exp b
Internet Exp C
Ad Attitude d

Ad Distraction 6

Uniform

19.29
(18.76-19.81)

14.00
(13.53-14.47)

9.30
(8.30-10.29)

38.17
(36.72-39.61)

9.81
(931-1031)

3.00
(2.51-3.49)

5.89
(5.31-6.47)

Hybrid

T-Test

 

519.43
(18.90-19.97)

13.69
(13.22-14.16)

9.26
(819-1032)

39.15
(36.99-41.32)

9.32
(854-1010)

2.58
(2.02-3.14)

5.97
(5.20-6.73)

t(89)<1
t(69)<1
t(66)<1
t(66)<1
t(66)=1.06,

p=n.s.

t(66)=1.11,
p=n.s.

t(66)<1

NOTE: (a) Maximum score 15 (b) Most experience=50 (c) Most experience=15 (d)
Positivity of attitude towards Internet Ads with most positive=9, most negative=1 (e)
Self-report measure of distraction from In ternet ads with 2=least distracted, 12=most

distracted.

There were no reliable differences on these variables between the participants in

 

3 Due to the nature of this preliminary study, demographic information was not
available for all participants tested.

the two phases of the experiment, indicating that these two samples appear to be
equivalent.
Attention to Advertisements

The proportion of incorrectly "recognized” new images (false alarms) and the

proportion of correctly recognized old images (hits) were computed, and used as a
measure of attention to advertisements. A logistic discrimination index (dL; derived
from signal detection theory) was used to evaluate participants' ability to
discriminate between new and old ads. A positive, non-zero dL value is an indication

of memory strength, after controlling for guessing behavior. Based on

recommendations by Snodgrass and Corwin (1988), the following formula was used
to calculate the discrimination index: dL=ln {[H(1-FA)]/[(1-H)FA]} where ln=natural

log, H=proportion of hits, and FA=proportion of false alarms. Hit rates and false
alarm rates equal to either 0 or 1 were adjusted by 0.01.

It was expected that graphics would be considered more task-relevant than
advertisements, and would consequently be processed more than ads, leading to
their increased recognition. Similarly, related ads were expected to be processed
more (and recognized more) than unrelated ads. It was also predicted that
participants might be more likely to process ads paired with graphics (Hybrid
condition) than ads paired with other ads. A 3 (Image Type: graphics, related ads,
and unrelated ads) X 2 (Presentation condition: Uniform or Hybrid) repeated

measures ANOVA was conducted using the discrimination index to test these

35

hypotheses. In order to control for differences in the actual images presented in
each of the presentation conditions, this analysis was conducted using only
recognition data from ads appearing in both the Hybrid and the Uniform
presentation conditions, and for graphics appearing in the Graphics Only condition

(held constant across presentation conditions).

I Uniform Hybrid

 

 

dL

 

   

  

0 ------ ~ —w———~
Graphics Related Ads Unrelated Ads

Image Type

Figure 1: Mean discrimination indices for the images presented in the
preliminary study memory test.

Contrary to predictions, there were no signiﬁcant main effects of image type
or presentation condition, Fs <1. The interaction between image type and
presentation condition was not signiﬁcant, F(2, 114)=2.81, p=.06, r12=.05, as is
shown in Figure 1. These values, along with the proportions of hits and false alarms

for each image type, are displayed in Table 3.

36

Table 3: Proportion of hits and false alarms (PAS), and their corresponding discrimination

index (dL) for image recognition in the preliminary study.

 

Uniform
A Hits .61 (so-.72)
O
8
3 FAs .08 (05-.12)
E
5"
5 dL 4.26 (3.14-5.37)
Hits .53 (41-64)
8
<
'3 FAs .14 (09-20)
E
34’ (IL 3.17 (1.97-4.37)
... Hits
‘5.“ FAs
.2
u:
3' d
U L
v Hits .45 (34-56)
<
'5
g PAS .15 (.08-.21)
T:
5 (IL 2.59 (1.42-3.77)
.. Hits
:5
U
u. FAs
.2
a:
B a
U L

Hybrid

 

545— (33-56)—
.15 (07-21)

2.48 (1.29-3.68)

 

.61—(.4844)
.14 (07-20)
4.00 (2.60-5.41)
.64 (.52-.76)
.15 (07-22)

4.61 (3.33-5.90)

.54 (40-68) _
.08 (.04-.12)
4.07 (2.76-5.39)
.54 (.40-.66)
.15 (07-22)

3.39 (1.95-4.84)

NOTE: Graphics (GO) denotes graphics presented in the Graphics Only condition, Graphics
(RA) denotes graphics presented with related ads in the Hybrid presentation condition, and
Graphics (UA) denotes graphics presented with unrelated ads in the Hybrid presentation

condition.

37

An additional analysis of image recognition was conducted on the data from
just the participants in the Hybrid condition. As shown in Table 3, there were 5
image conditions in the hybrid condition to account for the mix of ads and graphics
appearing in the ad-present conditions. The discrimination indices were compared
across these 5 image conditions in this additional analysis to look for differences in
the recognition of graphics presented with graphics as opposed to graphics
presented with either type of ad. This analysis, however, replicated the null results
of the previous analysis, in which there was no main effect for image type. The
discrimination indices and proportion hits and false alarms for this analysis are also
reported in Table 3.

The time taken for participants to make their decision was also recorded. The
average time taken for each participant to reach a correct decision (a hit or a correct
rejection) was computed. This average was computed using only the RT’s from the
ﬁrst 25 images, as the RT for the ﬁnal decision reﬂected both [decision time and the
time taken to advance to the next task. Outliers on the RT measure (decision times
in excess of 30s) were also removed. The decision times were predicted to follow
the same basic pattern predicted for the accuracy data, where increased processing
of graphics relative to ads was expected to lead to faster reaction times.

The decision times were analyzed using a 3 (Image Type) X 2 (Presentation
Condition) repeated measures ANOVA. Mean decision times are reported in Table 4.
There was again no signiﬁcant effect of image type, F(2, 178)=1.98, p=.14, and no

signiﬁcant effect of presentation condition, F<1. The interaction between image type

38

and presentation condition was also not signiﬁcant, F<1.

Table 4: Mean time in seconds (and 95% conﬁdence interval) taken to make a
correct Old / New decision for images in the preliminary study memory test.

 

 

 

 

Uniform Hybrid
T T'ErthTcsEtiT) T T230 (T201223? T T T2.T3T7(ioe-2.T67)TT
T EéIEteETAtTiTsTT 231672173288) T T T261 (2.19-3.03)T T
Graphics (RA) 2.22 (1.59-2.84)
T TEnEJaEdATdsTT 2.55T(2.16-T2T.94)T T T 2T.T47T(2.08-2§5T)T T
Graphics (UA) --- 2.20 (1.83-2.56)

NOTE: Graphics (GO) denotes graphics presented in the Graphics Only condition,
Graphics (RA) denotes graphics presented with related ads in the Hybrid presentation
condition, and Graphics (UA) denotes graphics presented with unrelated ads in the
Hybrid presentation condition.

 

As with the recognition accuracy data, an additional analysis of the decision
times made by the participants in the Hybrid presentation condition is possible. An
ANOVA was conducted to compare the 5 different image conditions present for the
participants in the Hybrid presentation condition. This analysis also failed to ﬁnd a
main effect of image type or of presentation condition, replicating the results of the
original analysis.

Goal Task Performance

Two measures of goal task performance were included in this experiment:

web page viewing time and content question accuracy. Previous experiments using

similar paradigms have shown few, if any, distraction effects. Consequently, large

39

distraction effects were not expected in this experiment either. However, it was
expected that if distraction effects were to emerge, they should be most likely in the
conditions in which attention was most likely to becaptured by the task irrelevant
ads. In particular, it was predicted that attention was most likely to be captured
when related ads were presented with graphics in the Hybrid condition.

Web page viewing time. The total time spent per page was averaged across
each of the ﬁve pages in the four image conditions. A 4 (Image Type) X 2
(Presentation Condition) repeated measures ANOVA was conducted on the measure
of time spent viewing the page. The means for time spent viewing web pages are
presented by image type and condition in Table 5.

Table 5: Mean time in seconds spent viewing web pages (and 95% conﬁdence
interval) in the preliminary study.

 

 

Uniform Hybrid
TEXTtTOTn-ly T T T 103.77%524-12233T T9081 (7248-10914)
Graphics Only 96.18 (8508-10728) 96.28 (8530-10726)
Related Ads 99.04 (8595-11212) 105.03 (9209-11798)
Unrelated Ads 87.94 (78.66- 97.22) 89.55 (80.37- 98.72)

There was no signiﬁcant main effect of image type, F(3, 267)=2.22, p=.09,
n2=.02. Although this was a small effect size, the pattern of the means observed may
support the hypothesis that distraction effects were more likely to occur on pages

containing related ads than unrelated ads, as participants spent signiﬁcantly more

40

time viewing web pages with related ads (M: 102.04; CI=92.83-111.24) than with
unrelated ads (M: 88.74; Cl=82.22-95.27). The time spent viewing pages containing
no images, (M: 97.29; CI=84.26-110.33), and pages containing graphics (M: 96.23;
Cl=88.42-104.04) were not signiﬁcantly different from either of the two ad
conditions. There were no other signiﬁcant effects.

Content question accuracy. Multiple-choice content learning questions were
administered, and given a score of 0 (either no answer was entered, or an incorrect
answer was entered) or 1 (a correct answer). Overall accuracy was computed for
each condition. A 4 (Image Type) X 2 (Presentation Condition) repeated measures
ANOVA was conducted with the question accuracy data to test for distraction effects
related to ads.

Table 6: Mean accuracy (and 95% conﬁdence interval) on content learning
questions in the preliminary study.

Uniform Hybrid
TéiTtTislyT THEM-.76) TIT him) "
Graphics Only .69 (.59-.78) .76 (.67-.85)
Related Ads .69 (.62-.76) .71 (.64-.78)
Unrelated Ads .69 (.61-.77) .61 (.54-.69)

As can be seen in the means shown in Table 6, there were no signiﬁcant
effects for image condition or for presentation condition, and no signiﬁcant

interaction between the two.

41

Discussion

The preliminary study was designed to test the hypotheses that participants
are likely to select content-related items for further processing, while only
minimally processing unrelated items. In addition, it was postulated that
participants may be more likely to process irrelevant advertisements when they are
presented with seemingly task relevant, graphics. This hypothesis was based on the
idea that the inclusion of graphics is likely to activate an attentional set that contains
perceptual features unique to images, but shared by both graphics and ads.
Attention to Advertisements

Attention to advertisements was measured by accuracy and decision time on
a Yes / N 0 image recognition test. No signiﬁcant differences were found in the
participants' accuracy in recognizing graphics, related ads, or unrelated ads. The
interaction between image type and presentation condition was not signiﬁcant,
although the effect size indicated that it accounted for approximately 5% of the
variation in image recognition rates.

The pattern of the means indicated that the predicted relationship across
image types may have been present for participants in the Uniform presentation
condition, who showed the best recognition for graphics, and the worst for
unrelated ads. On the other hand, participants in the Hybrid condition may have
been better at recognizing the two types of ads than they were at recognizing
graphics, and may have recognized more ads in general than the Uniform

participants. The time taken to make a correct recognition decision was also

42

analyzed, but no differences were signiﬁcant. In conclusion, although the
recognition data produced means that may have been consistent with the predicted
use of semantic content and perceptual features to guide attention, no signiﬁcant
results were found.
Goal Task Performance

Previous research indicates that Internet users do not appear to be
particularly susceptible to distraction in the form of ads, at least not while they are
engaged in an information search task. Consequently, few distraction effects were
predicted for the current experiment. However; it was predicted that the ads that
captured the most attention would also be most likely to produce distraction effects.
However, no signiﬁcant differences in recognition rates of the different types of ads
was found, and accordingly, no signiﬁcant differences in performance measures
were found.
Limitations

Experiment 1 represents an attempt to produce signiﬁcant effects supporting
the use of these environmental features to guide attention, as well as to address
several limitations of the preliminary study. First, the materials used in the
preliminary study were somewhat limited. One of these limitations dealt with the
distribution of images across the passage, as the number of images per passage page
was not held constant. In addition, the passage content (a long article on processing
health information) is not necessarily representative of the varied types of

information people seek on the Internet. To improve generalizablility, the

43

manipulations of presentation and image type should be carried out on multiple
topics. The inclusion of multiple topics may also increase the likelihood that
participants are interested in the search content. Exit interviews with 70 of the
participants from the preliminary study indicated that on a scale of 1 (Most
Interesting) to 9 (Most Boring), participants on average rated the health care
informational passage a 6.6. The inclusion of more interesting topics is likely to
improve the external validity of this research, as few Internet users are likely to
search for information on a topic in which they have no inherent interest.

Another limitation in the materials concerns the content learning questions.
Participants were only tested on one piece of information per page, and no
signiﬁcant differences in the content learning measure emerged. Increasing the
question density could improve this measure by increasing its sensitivity. In
addition, the content learning questions only assessed understanding of the target
passage text, and did not allow for assessment of intrusions related to processing
advertisements.

Finally, this experiment included only young adults. The Internet is a
resource used by adults of all ages, and in recent years, adults aged 50 and over have
been the fastest growing age group online. This age group has also been identiﬁed
as particularly likely to be susceptible to distraction that is semantically related to
the goal task (e.g. Connelly et al., 1991). Therefore, it is possible that older adults
may show larger distraction effects than younger adults when reading passages that

are displayed with Internet ads for products or services related to the passage

44

content.
Experiment 1
Age Differences in Distraction
Older Adults and the Internet

Studies investigating attention and ad‘processing on the Internet commonly
involve only young adults, as they represent the majority of Internet users (Pew
Internet and American Life Project, 2007), and are consequently the age group
targeted by most web site developers. Over 80% of adults aged 18-29 are using the
Internet, while less than 60% of adults over the age of 65 take advantage of this
resource. However; these numbers don’t tell the whole story. just as older adults are
the fastest growing age demographic in the United States (Wetle, 2002), they are
also the fastest growing age group on the Internet. The percent of adults online aged
50 and over grew 53% from 1998 to 2000 (United States Department of Commerce,
2000), and another 47% through 2004 (Fox, 2004).

As a group, older adults spend signiﬁcant time online, with almost half of the
age group logging between 11 and 33 hours each week (Strong, Walker 8: Rogers,
2002). Furthermore, this group spends more money online than any other age
group (Adams, 2003; N UA Internet surveys, 2000; University of Wisconsin-
Extension, 2003). Not surprisingly, advertisers have identiﬁed older adults as an
ideal audience to target in Internet advertising campaigns (Adams, 2003; The Media
Audit, 2001). Since older adults are just as likely to encounter irrelevant distraction

on the Internet as young adults, it is important to also include this age group in

45

studies that investigate mechanisms involved in controlling distraction on the
Internet.
Increased Susceptibility to Distraction

An extensive literature on age differences in cognition indicates that older
adults, relative to young adults, often have more difﬁculty ignoring irrelevant
distraction. Hasher and Zacks (Hasher, Lustig, 8: Zacks, 2007; Hasher & Zacks, 1988;
Zacks & Hasher; 1994) have proposed a general processing model to account for
age-related differences in attentional abilities. Their framework proposes that
attentional ability reﬂects the ability to control the processing of irrelevant
information.

One way in which individuals can control the processing of distracting
information is by ignoring the distraction in the ﬁrst place, which has been termed
the “access function” of attention regulation (Hasher et al., 1999, 2007). The access
function closely approximates selective attention abilities. According to Hasher and
Zacks, declines on complex cognitive tasks can be attributed to the access function
of attention regulation becoming less efﬁcient with advanced age, thus allowing
irrelevant information to be processed along with task relevant information.

Although age differences in selective attention abilities within Internet
environments have not been previously examined in the literature, age differences
in selective attention during a reading task have been investigated using the
“reading with distraction” task (Connelly et al., 1991; Darowski, Helder, Zacks,

Hasher; & Hambrick, 2008; Duchek, Balota, 8: Thessing, 1998; Dywan 8: Murphy,

46

1996; Earles et al., 1997; Salthouse, Atkinson, 8: Berish, 2003). In this task,
participants are asked to read paragraphs that contain embedded distractors. The
distractors can be words that are related to the content of the paragraph, words that
are unrelated, or strings of X5. Distractor words are distinguished from target words
by the use of a different font style (ex. upright font for distractors and italicized font
for target text). Older adults tend to be more disrupted by all types of distractors
than younger adults, but the age effects tend to be differentially larger for word
distractors—especially those that are related to the text content. Findings from this
task indicate that older adults are more likely to show performance decrements
when distractors are semantically related to the target text than are young adults.
Factors Attenuating Age Differences in Attention

Although older adults may be more susceptible to distraction than young
adults, there is evidence that these age differences are somewhat attenuated when
the distraction occurs in predictable locations, or when older adults rely on
automatic attentional capture rather than attempting to maintain a goal to ignore
the distraction.

For example, when the location of the distractors was ﬁxed in a variation of
Connelly et al.’s (1991) reading with distraction task, differences in distraction
between old and young adults were no longer observed (Carlson, Hasher, Connelly,
8: Zacks, 1995). In this study, the reading passage was displayed in columns, with
every other column containing distractors. Participants were instructed to read

across each row, ignoring the distractor words in the intervening columns. In this

47

experiment, older adults were not differentially susceptible to distraction when
compared with young adults.

In another experiment, Carlson et al. (1995) demonstrated that the key to
eliminating age differences in distraction was that the location of the distractors was
ﬁxed, rather than merely predictable. In this experiment, they again displayed a
paragraph divided into columns. However; instead of displaying a full column of
content followed by a full column of distractors, the order of content and distractors
switched from line to line. Thus, if the ﬁrst line is arranged in an ABA pattern, where
A is content and B is distractors, the second line would be arranged in BAB pattern.
When distraction was in this way predictable, yet not ﬁxed, the older adults again
showed differential distraction. Therefore, it is possible that since the location of
advertisements on a web page remain ﬁxed once the page has loaded, older adults
will not show an age-related increase in susceptibility to distraction.

Further support for older adults’ ability to use location cues to ﬁlter out
distractors comes from a study on visual marking (Watson 8: Maylor, 2002). In this
study, both older and younger adults were presented with visual search displays in
which one set of stimuli (containing only distractors) appeared onscreen slightly
before a second set of stimuli containing distractors as well as a target. Older adults
were just as able as young adults to mark the locations of the ﬁrst set of distractors
and exclude those distractors from search when the position of the distractors was
stationary. When the distractors were moving, however, older adults were not able

to exclude the old distractors from search, although young adults were. Although

48

older adults may not be as able as young adults to apply a location-based ﬁlter to
moving stimuli, most Internet advertisements appear in one stationary location
onscreen, indicating that older adults should be as able as young adults to apply a
location-based ﬁlter.

Additionally, studies by Kramer and colleagues (Colcombe, Kramer, Irwin,
Peterson, Colcombe 8: Hahn, 2003; Kramer, Hahn, Irwin 8: Theeuwes, 2000) ﬁnd
evidence that older adults perform comparably in selective attention tasks (as
measured by eye movements and RT) when the control of attention is automatic.
For example, when the distractor is not perceptually salient, older adults were just
as able as young adults to ignore the distractor. However, when the control of
attention is voluntary, based on task demands, older adults display more distraction
effects than young adults. In the case of attention to Internet advertisements, it is
unclear as to whether distinctive visual features of ads such as bright colors and
motion are likely to require older adults to intentionally ignore advertisements, or
whether some other element of the task such as ﬁxed location of the
advertisements, or ease of ﬁltering on the basis of distinctive perceptual features,
may contribute to the automatic ignoring of advertisements.

In summary, age differences in attention control abilities could render older
adults particularly susceptible to distraction that shares features with the goal task.
There is some indication from the reading with distraction task (Connelly et al.,
1991) that semantically related (as opposed to perceptually related) features are

most likely to cause this age-related deﬁcit. Accordingly, ads that are related to the

49

content of the web page could be expected to produce age differences in distraction
effects. On the other hand, older adults have exhibited equivalent performance as
young adults when distractors appear in ﬁxed locations (Carlson et al. 1995).
Therefore, if the distractor advertisements appear in ﬁxed locations on the page,
older adults may be able to screen them out even if they are semantically related to
the web page content.
Predictions

The predictions detailed in the presentation of the preliminary study were
also valid for Experiment 1. In addition, Experiment 1 added several predications
about the role of age differences in attentional control abilities. It was predicted that
the older adults may be particularly distracted when a task-irrelevant item is
semantically related to target content. In the current experiment, both graphics and
related ads were not necessary for the completion of the goal task, but were selected
to relate to the target text content. Consequently, it was predicted that older adults
may show greater recognition of these types of items (indicative of greater
processing) than the young adults. Furthermore, it was predicted that if older adults
did demonstrate differential susceptibility to distraction in goal task performance,
this was most likely to occur in the graphics and related ads conditions. Finally,
given general age-related slowing (e.g., Salthouse, 1996) older adults were expected to
show age-related declines in processing speed, resulting in signiﬁcantly slower

reaction times throughout the experiment.

50

Materials and Methods

Participants

A total of 64 young adults (age 18-22; 29 male) and 64 older adults (age
59-89; 14 male) were tested. The young adult participants were recruited from the
psychology department subject pool and compensated with partial course credit.
The older adult participants were community-dwelling adults compensated $10.00
an hour for their time. All participants were ﬂuent in English, had completed at
minimum a high school degree, and were sufﬁciently experienced with a computer
to be able to control the mouse themselves throughout the experiment. Four young
adult participants and three older adult participants reported learning disabilities.‘
Analyses were repeated with learning disabled participants excluded, with no
change to the results.
Design

Experiment 1 was a 2 X 4 X 2 mixed design that replicated the design of the
preliminary study with the addition of older adults. Consequently, the ﬁrst quasi-
independent variable was age group (young and old). The second independent
variable was image type and was manipulated within subjects. The same two ad-
absent baseline conditions used in the preliminary study (Text Only and Graphics
Only) comprised the ﬁrst two levels of the image type variable. The second two

levels of the image type variable were comprised of the two ad-present

 

4 All four young adults were diagnosed with Attention Deﬁcit Disorder; while one
older adult was diagnosed with dyslexia and the other two did not report the
nature of the learning disability.

51

experimental conditions used in the preliminary study (Related Ads and Unrelated
Ads). Each participant read one 6-page passage in each of the four image conditions.
Each passage was on a different topic (Health, Loans, Love and Travel). The third
independent variable was ad presentation condition (Uniform and Hybrid),
manipulated between subjects as in the preliminary study. As in the previous
experiment, the presentation condition affected only pages presented in the ad-
present conditions. The allocation of images to display condition is depicted in Table
7 below.

Table 7: Image types displayed in each presentation mode in Experiment 1.

Image Type

 

Text Only Graphics Only Related Ads ‘Unrelated Ads!

 

 

E No graphics 7 5 graphics 2 related ads 2 unrelated
l-

: 52 or ads ads

0 s: ,

'5 D

E

3 No graphics 2 graphics 1 related ad 1 unrelated ad

a; 'U

I- '—

9‘ "E. or ads 1 graphic 1 graphic
I

The order of the passage topics and the image conditions were
counterbalanced across 16 forms. Four passage topic orders were created, across
which each passage appeared in each possible position (ﬁrst, second, third, or
fourth) one time. Four image condition orders were also created, across which each
condition appeared once in each possible position. Then, passage order was crossed

with condition order to create 16 forms in which each passage appeared in each

52

condition in each possible order, as shown in Table 8 below.

Table 8: Counterbalancing of passage topics and image type conditions in
Experiment 1.

Order Topic FormA FormB FormC FormD

 

 

 

 

 

 

1 Health TO GO UA RA
1 Loan GO RA TO UA
1 Love RA UA GO TO
1 Travel UA TO RA GO
2 Loan TO GO UA RA
2 Love GO RA TO UA
2 Travel RA UA GO TO
2 Health UA TO RA GO
3 Travel " TO " GO UA RA _
3 Health GO RA TO UA
3 Loan RA UA GO TO
3 Love UA TO RA GO
4 Love TO GO UA RA
4 Travel GO RA TO UA
4 Health RA UA GO TO
4 Loan UA TO RA GO

NOTE: T0=Text Only; GO=Graphics Only; RA=Related Ads; UA=Unrelated Ads

Materials

Web page and content questions. Participants viewed four 6-page passages,
with approximately 282.96 (SD=48.81) words per passage page. The per-page word
count included words in the page title, passage outline, and navigational links which
were displayed on each of the six pages of the passage. Passages were on four

different topics: interpreting health information (health), selecting a student loan

53

(loan), a travel journal from a cruise to the Caribbean (travel), and the psychology of
love and attraction (love). Efforts were made to give each passage topic a unique,
but professional appearance and style. Graphics, related ads, and unrelated ads
followed the selection criteria used in the preliminary experiment, with the
additional stipulation that unrelated ads were selected to be unrelated to the topic
of any of the passages. Time taken to read each page, in milliseconds, was recorded
by the computer.

The allocation of images to image type conditions followed the method used
in the preliminary study, with the modiﬁcation that every image-present page had
exactly two images on it, and each passage page contained unique images.
Therefore, a total of 12 graphics, 12 related ads, and 12 unrelated ads were selected
for use in each of the four websites, for a total of 144 images used across the various
conditions of the experiment. Participants in the Uniform condition were exposed to
12 graphics, 12 related ads, and 12 unrelated ads during the experiment.
Participants in the Hybrid condition were exposed to 24 graphics, 6 related ads, and
6 unrelated ads during the experiment. Accordingly, twice as many ads were
displayed in the Uniform condition as in the Hybrid condition. The set of 12 ads
displayed in the Uniform condition, but removed from the Hybrid condition, were
the same for each participant, meaning that they were not counterbalanced.

Half of the ads displayed in the ad-present conditions were animated.
Therefore, in the Uniform condition, each page of an ad-present passage contained

one animated ad and one static ad. In the Hybrid condition, half of the passage’s six

54

pages contained animated ads, while the other half contained static ads. None of the
graphics were animated.

Participants answered 12 multiple choice questions per passage to test their
learning of the passage content. To provide an even covering of the passage, two
questions per passage page were generated. Each question had four answer
alternatives. One answer alternative was correct based on the passage text, one was
a lure that was incorrect based on the content of the passage text, but corresponded
to the content of one of the related advertisements on that page, and two were
incorrect distractor answers.

The lure was included as an indicator of related advertisement content
processing. If including related ads results in lower content learning scores for one
or both groups of participants, it is not possible to determine whether the content of
the ads was actually processed, or whether the visual characteristics of the related
group of ads coincidentally attracted more attention than the unrelated ads. The
inclusion of the lure allowed for an additional test of any potential differences
between the related and unrelated ads. Since the lure corresponded to the content
of the related ads, an increase in lure selection rates in the Related Ads condition
would be interpreted as evidence that ad content is being processed. In addition to
the selected answer choice, time taken, in milliseconds, to select an answer choice
was also recorded.

Memory test for images. Over the course of the experiment, each participant

was exposed to 36 images. Consequently, the memory test consisted of 36 trials, and

55

tested all 36 of the images presented to an individual participant. Each trial
contained one previously presented (old) image and one new image. Consequently,
participants in the Uniform condition were presented with 12 old graphics, 12 old
related ads, and 12 old unrelated ads. Participants in the Hybrid condition were
presented with 24 old graphics, 6 old related ads, and 6 old unrelated ads.

New images were undisplayed images selected from the same passage as the
old image. For example, if participants viewed the health information passage in the
related graphics condition, they would see a related graphic paired with an image
presented in one of the ad-present conditions of that passage—in this case, either a
health-related ad or a health-unrelated ad. In this way, the new and old images were
counterbalanced across participants. The pair of images being tested were shown
together onscreen, one pair at a time. Pairs were dynamically selected for each
participant, based on the ads displayed to that participant during the experiment.
Since some of the topic-related images were old and some were new, simply
choosing only topic-related images would not ensure high accuracy on the test.
Furthermore, in many cases, participants needed to make a choice between a
related graphic and a related ad, rendering a selection strategy based on topic-

relevance ineffective.S Images from only three of the passages were tested, as the

 

5 To check for the possibility that participants nevertheless attempted to
implement such a strategy, image recognition rates for each type of topic-related
image (graphics and related ads) were compared between decisions involving
two topic-relevant images and decisions involving a new unrelated ad. There
were no signiﬁcant differences in either the young adult recognition rates, or the
older adult recognition rates, indicating that recognition rates of topic-related
images should not be inﬂated by the use of a strategy involving selection of only
topic-relevant images.

56

passage viewed in the Text Only condition was not displayed with any images.
Selection accuracy was scored, and time taken, in milliseconds, to classify an image
as old or new was also recorded.

Questionnaires. The ﬁrst questionnaire assessed demographic information,
the second assessed computer experience (Appendix C), and the third assessed
Internet experience (Appendix D). The demographic questionnaire was designed
and generated by the author to record participants' gender, age, handedness, and
presence of a learning disability, in addition to the variables used to compare the
older and younger participants, such as self-reported health status, and years of
education. As in the preliminary study, the computer experience questionnaire was
excerpted from one used by Kubeck et al. (1999), while the Internet experience
questionnaire was developed by the author.

A measure of attitude toward Internet advertising was computed using the
following questions from the Internet Experience questionnaire: Have you ever
purchased an item based on Internet Advertising? (Y/ N) ; Do you think Internet
advertising works? (Y/ N) ; Do you ﬁnd Internet advertising distracting? (Y/ N) ; Do you
ﬁnd Internet advertising to be annoying? (Y/ N). This measure of attitude toward
Internet advertising, and an additional question asking participants to indicate

whether they believed they were distracted by Internet ads, were only scored for

57

participants who completed all of the questions involvedé, and had at least some
experience with the Internet.

At the end of the experiment, participants were asked to ﬁll out an exit
questionnaire (Appendix E). The exit questionnaire assessed participants’ interest
in the passages included in the reading task, their awareness of advertisements
during the experiment, and whether they noticed any relationship between the
advertisements and the content of the articles.

Measures of processing speed. A common ﬁnding in aging research is that
older adults tend to perform tasks more slowly than young adults (e.g. Salthouse,
1996). Therefore, it is possible that older adults may perform more slowly than the
young adults in the web page viewing task due to age-related slowing, or due to
distraction effects. Participants in Experiment 1 completed two speeded
comparison tasks, Pattern Comparison and Letter comparison (Babcock &
Salthouse, 1990) to enable individual differences (and consequently, age
differences) in processing speed to be factored out of the website viewing time
measure.

In the pattern comparison task, participants were shown pairs of abstract
line drawings and asked to determine as quickly as possible whether the two

patterns were the same or different. In the letter comparison task, participants were

 

6 One young adult participant and 11 older adult participants did not answer all of
the items involved in this measure. No reasons for skipping these items were
reported, although in the case of the older adults, it seems possible that low
levels of Internet experience resulted in uncertainty in their opinions about
Internet ads.

58

shown pairs of nonsense letter strings ranging in length from three to nine letters
long, and were asked to determine whether the two strings were the same or
different.

In both comparison tasks, participants recorded their decision for each pair
by writing an “S” for patterns that were the same, and a “D" for patterns that were
different on a line separating each member of the pair. Participants completed two
trials of each comparison task, using a different set of stimuli for each trial. Each
trial was 305 long.

Reading with distraction task. This task was included as a well-established
and reliable measure of susceptibility to distraction (Darowski et al., 2008). In this
task, participants read four short narrative passages, each approximately 125 words
in length, selected from the materials used by Connelly et al. (1991). Two of these
passages were presented in a High Distraction condition in which four words or
short phrases that were semantically related to each story’s topic were repeated
approximately 15 times each. These distractors were randomly interspersed
throughout each passage, appearing on average every two to three words for an
approximate total of 60 distractors per passage. The remaining two passages were
presented in a Low Distraction condition in which the distracting words or phrases
from the high distraction condition were replaced by strings of X5 of equivalent
length in the same locations within the text.

In both conditions, target text was distinguished from distractors by a font

difference. Target text always appeared in an italicized font, while distractors

59

appeared in a regular upright font. Typically, this task is administered by instructing
participants to read the target text out loud. However, since most Internet users
read Internet content silently, participants in the current experiment were
instructed to read the target text to themselves. This methodological change also
enabled multiple participants to be tested in group sessions. Participants were
instructed to read only the italicized text of each passage, and to ignore the
distractors appearing in upright font. When participants had completed reading a
passage, they pressed the spacebar key to advance to the next screen.

Each passage was followed by a set of four 6-option multiple choice
questions, in which one answer was correct based on the target passage text, and
another answer served as a lure, corresponding to a distracting word or phrase in
the High Distraction condition. After two short example passages (one in each
condition), participants completed one Low Distraction passage, two High
Distraction passages, and a fourth Low Distraction passage. .

Procedure

Participants were randomly assigned to one of the presentation modes
(Uniform or Hybrid), and one of the 16 experimental forms (4 passage orders X 4
image condition orders) prior to arriving at the lab. Participants were tested in
small groups, ranging from one to ﬁve participants at a time. Upon arrival at the lab,
participants were greeted by the experimenter, completed an informed consent
procedure, and were seated at a computer. Participants received the following

instructions prior to beginning the web page viewing portion of the experiment:

60

“In the ﬁrst task in this experiment, you will be asked to ﬁnd speciﬁc
information on web pages You will be searching through several different
websites on four different topics to ﬁnd your answers. When you begin each
topic, you will see a list of questions. These are the questions you will be
asked after you have ﬁnished viewing the website. Once you click the 'View
Website' button, you will be directed to the website, and you will not be able
to review the content of the questions again. Once you have viewed all of the
pages in a particular topic, you will be asked to answer each of the pre-
viewed questions by selecting the most appropriate answer choice.”

During the web page viewing task, participants worked independently at a
pace that was comfortable to them. The experimenter monitored their progress and
was available to answer questions about the task procedure or navigating the
website. At the beginning of each passage, the 12 content learning questions
(without answer choices) were displayed on the screen for participants to
familiarize themselves with. When they were ready, they clicked on a hypertext link
to advance to the ﬁrst page of the website viewing task, and continued viewing each
page of the passage sequentially. Once a participant had navigated beyond a passage
page, he could not return to it again.

When participants completed reading the passage, they were directed to a
page where they proceeded to answer each of the previously viewed questions.
Questions were answered one at a time. This procedure of reading questions,
viewing a passage, and answering questions was completed for each of the four
passage topics.

When the content learning questions had been completed for the last

passage, the unexpected memory test for images was administered. Participants

were instructed to do their best to select the image that they recalled seeing earlier

61

in the experiment by using the mouse to click on the recognized image. If they were
unsure as to which image was old, they were instructed to guess. The image test did
not advance from one image to the next until a response was made.

Following this test, participants ﬁlled out the questionnaires on
demographics, computer experience, and Internet experience. While participants
ﬁlled out these forms, the experimenter interrupted them twice. During the ﬁrst
interruption, participants completed Pattern Comparison. During the second
interruption, participants completed Letter Comparison. Once the participants had
completed the questionnaires and the speeded comparison tasks, they began the
Reading with Distraction task on the computer. The experiment was concluded with
the paper/ pencil administration of the vocabulary test and the exit questionnaire.
After all testing was completed, participants were provided with a written
description of the experiment, and given the opportunity to ask questions about the
purpose of the study.

Results
Manipulation Checks

Sample selection. The predictions for the current experiment assume that the
samples of older and younger participants selected for the Uniform and the Hybrid
presentation conditions were equivalent. This assumption was checked by
comparing the older and younger adult samples across presentation conditions on

several demographic variables. These means are presented in Table 9.

62

Table 9: Means (and 95% conﬁdence interval) on the demographic measures
included in Experiment 1.

Age

Yrs Education
Health8

Vocabularyb
Computer ExpC
Internet Expd

Ad Attitudee

Ad Distractionf
Pattern Comparisong

Letter Comparisong

Young Adults
Uniform Hybrid
T 193? ’ # 19.28
(18.86-19.76) (18.81-19.75)
12.81 12.95
(12.37-13.25) (12.50-13.40)
3.56* 3.19“
(3.37-3.75) (2.94-3.44)
9.91 9.88
(911-1071) (887-1089)
41.43 40.22

(40.16-42.70)

16.38
(15.44-17.32)

1.64
(1.33-1.95)

64.52
(47.39-81.65)

41.56
(39.33-43.79)

24.28
(22.81-25.75)

(39.08-41.36)

16.81
(16.10-17.52)

1.66
(1.37-1.95)

68.25
(56.66-79.84)

41.00
(38.77-43.23)

24.09
(22.25-25.93)

 

(74.67-99.41)

2822*
(26.26-30.18)

18.12
(16.91-19.33)

Older Adults
Uniform Hybrid
AILH70156_— 70.88
(68.29-72.83) (68.37-73.39)
15.73 16.37
(14.71-16.75) (15.53-17.21)
3.16 3.28
(2.94-3.38) (3.03-3.53)
13.69 13.12
(13.22-14.16) (12.51-13.73)
31.79 32.44
(29.54-34.04) (30.64-34.24)
11.41 12.38
(9.55-1327) (10.93-13.83)
1.86 1.46
(1.55-2.17) (1.15-1.77)
87.04 66.67

(49.51-83.83)

31.31“
(29.39-33.23)

18.44
(17.10-19.78)

NOTES: (*) denotes difference between presentation conditions is significant (0)

Rating based on self-report: 1=worst health; 4=best health (b) Maximum score 15 (c)
Most experience=50 (d) Most experience=22 ( e ) Positivity of attitude towards Internet
Ads with most positive=3; most negative=0 0‘) Percent of participants claiming that
they are distracted by Internet ads (g) Number of correct comparisons completed in

6 0s.

63

Overall, there were only two signiﬁcant differences between presentation
condition samples within the same-age group. Young adults in the Uniform
condition rated themselves as signiﬁcantly healthier than the young adults in the
Hybrid condition, t(62)=2.28, p=.03. Older adults in the Uniform condition
completed signiﬁcantly fewer comparisons in the Pattern Comparison task than did
the older adults in the Hybrid condition, t(62)=-2.20, p=.03. Despite these small,
isolated differences, the overall pattern indicates equivalence of the groups assigned
to the Uniform and Hybrid conditions.

In addition to testing for equivalence across the presentation conditions, the
demographic measures were also used to compare the younger and older adult
samples. Participants were given a vocabulary test as a means of comparing verbal
ability between the two age groups. As is frequently found in aging research, the
older adults in this experiment scored signiﬁcantly higher on the vocabulary test
than the younger adults, t(104)=-9.18, p<.01, indicating stronger verbal ability in
the older group. In addition, the older adults had completed signiﬁcantly more years
of education than the younger adults, t(89)=-8.53, p<.01, although it should be
noted that the young adults were current students who had not yet completed their
formal education. The two age groups did not differ in terms of their self-rated
health status, t(126)=1.30. These three variables indicate that the older adults
should not be disadvantaged in comparison to the young adults in terms of their
verbal abilities, their education level, or their overall health status.

Since previous research has indicated that experience may play a role in

64

participants' ad processing (Bruner & Kumar, 2000; Dahlen, 2001; Raman &
Leckenby, 1998), participants' previous computer and Internet experience was
assessed. Not surprisingly, the young adults had signiﬁcantly more experience with
computers than the older adults, t(103.11)=10.19, p<.01. Similarly, the Internet
experience questionnaire indicated that the young adults had signiﬁcantly more
experience with the Internet than the older adults, t(92.60)=6.98, p<.01. Since
signiﬁcant age differences were found in these measures of experience, additional
exploratory analyses of the task performance measures were carried out to test the
hypothesis that age differences in performance may be due to different experience
levels, rather than an actual age-related change in attention abilities.

No age difference was found in attitude toward Internet advertising,
t(114)<1, p=n.s. Additionally, there was no age difference in participants’ Yes/No
rating of whether they were distracted by ads, t(118)<1, p=n.s. Therefore, despite
differences in experience levels, the participants reported similar attitudes toward
Internet advertising. Consequently, it was assumed that the perceived relevance or
irrelevance of advertisements to the web page viewing task should be equivalent
across age groups, and that participants in both age groups should be similarly
motivated to attend to ads (or to ignore ads, as the case may be).

Finally, older adults completed signiﬁcantly fewer comparisons than the
young adults on both speeded comparison tasks (Pattern comparison:
t(126)=10.69, p<.01 ; Letter comparison: t(126)=7.85, p<.01), indicating that the

typical ﬁndings of age-related slowing were present in these samples of older and

65

younger adults. In order to control for this age-difference, a speed composite score
was created, and used as a covariate in analyses in which age differences in
susceptibility to distraction appeared.

Attention to Advertisements

Experiment 1 employed a different type of recognition test than the one used
in the preliminary study as a measure of attention to advertisements. Although no
signiﬁcant difference was found in recognition rates of the different types of images
in the preliminary study, the overall pattern of the means supported the prediction
that recognition would be inﬂuenced by the degree to which an image was task
relevant. In the current experiment, the adoption of a forced-choice design for the
memory test may result in improved sensitivity of the test. Accordingly, it was again
predicted that recognition would be best for graphics and worst for unrelated ads.
Older adults were predicted to demonstrate proportionately higher recognition
rates of graphics and related ads than of unrelated ads, when compared to the
pattern of recognition found in young adults.

To test for these effects, each participants' accuracy in choosing old images
was computed separately for the three image types: graphics, related ads and
unrelated ads. To control for the differences in images presented across the two
presentation conditions, accuracy in recognizing graphics was computed using only
the recognition of graphics presented in the graphics only condition, and accuracy in
recognizing ads was only computed for ads that appeared in both the Hybrid and

the Uniform presentation conditions.

66

A 3 (Image Type) X 2 (Presentation Condition) X 2 (Age Group) repeated
measures ANOVA was conducted to compare accuracy across conditions. The means

used in this analysis are depicted in Figure 2.

a) I Uniform Hybrid
1 0 0 °/o

90% T
80% __-
70%
60%
50% -« ~~~~~ -- . . '-—-.

Graphics Related Ads Unrelated Ads

 

 

 

1
l

Percent
Recognized

    

 

100%
90%
80%
70%
60%
50%

 

 

Percent
Recognized

 

 

Graphics Related Ads Unrelated Ads
Image Type

Figure 2: Percent of images recognized by (a) young and (b) older adults
in the Experiment 1 image memory test.

67

A main effect of image type was found in the expected direction, F(2,
248)=37.46, p<.01, n2=.23. Planned comparisons between the accuracy in the three
image type conditions indicated that graphics (M: .85; CI=.83 - .88) were recognized
signiﬁcantly more accurately than related ads (M: .79; CI=.75 - .82), which in turn
were recognized signiﬁcantly more accurately than unrelated ads (M: .65; CI=.61
- .70). This ﬁnding is consistent with the predictions. No other effects attained
signiﬁcance.

An additional analysis was performed with the data from only the
participants in the Hybrid presentation condition. Since participants in this
condition viewed graphics presented with other graphics (graphics [GO]), as well as
graphics presented with ads (graphics [RA] and,graphics [UA]) a comparison can be
made in the recognition rates of graphics paired with ad as opposed to graphics
paired with other graphics.

A 5 (Image type: Graphics [GO], Related Ads, GraphicsIRA], Unrelated ads,
and Graphics [UA]) X 2 (Age Group) repeated measures ANOVA was conducted.
There was a main effect for image type, F( 4, 248)=10.64, p<.01, 112:.15, as shown in
Figure 3. Comparisons between the means indicated that the main effect of image
type found previously was replicated. More speciﬁcally, Graphics [GO] were
recognized signiﬁcantly more than any other type of image, and unrelated ads were
recognized signiﬁcantly less than any other type of image. Related ads, Graphics
[RA], and Graphics [UA] did not differ from each other, and were recognized less

than Graphics [GO], but more than unrelated ads. There was no signiﬁcant main

68

effect for age group, nor was the interaction signiﬁcant.

 

 

 

 

100%
3 S 800/ I '
E» ° j j l
a: 60% em..- ---..-._.-__- -_-__-. W...- .---.-._-

 

Graphics Related Graphics Unrelated Graphics
(GO) Ads (RA) Ads (UA)

Image Type

Figure 3: Percent of images recognized by participants in the Hybrid presentation
condition on the Experiment 1 image memory test.

Finally, it was hypothesized that a participant's interest in the topic of the
passage could inﬂuence the degree to which he/she attended to semantically related
images. In order to investigate this possibility, participants' recognition rates were
computed separately for related images in each topic. These topics were then
ordered according to the rank provided by each participant (1=most interesting,
4=least interesting). Paired t-tests for related image recognition rates for each rank
were conducted. Critically, a signiﬁcant difference between the recognition of
related images in the most interesting topic and the least interesting topics
emerged, t(18)=3.40, p<.05. This ﬁnding indicated that more related images were

recognized when a participant was interested in the passage, (M = .81; CI = .74-.88)

69

than when a participant was not interested in the passage topic, (M = .67; CI = .
59-.74). In addition, a signiﬁcant difference also emerged between the second most
interesting passage (M = .83, CI = .76-.90) and the least interesting passage (M = .75,
CI = 65-84), t(16)=2.20, p<.05. Although interest is not one of the main features
being investigated in this study, this ﬁnding indicates that this feature merits further
investigation in future work, as it may also be key to guiding attention.

The average time taken for each participant to reach a correct decision was
also computed. These decision times were predicted to follow the same pattern
found in the preliminary study, in which participants showed a trend to be faster
making decisions on graphics than on ads. In addition, it was also predicted that
older adults would take longer to make their decisions than young adults. These
predictions were tested using a 3 (Image Type) X 2 (Presentation Condition) X 2
(Age Group) repeated measures ANOVA. The overall results on this measure are
shown in Figure 4. The predicted main effects of image type, F(2, 248)=27.52, p<.01,
n2=.18, and age group, F(l, 124)=47.63, p<.01, 112:.23, emerged. Decisions were
made signiﬁcantly faster for graphics (M= 3.90; CI=3.44-4.35) than for either type of
ad, which did not differ from each other (related ads: M: 4.82; CI=4.31-S.32;

unrelated ads: M: 5.20; CI=4.70-5.69).

70

m
w

l Uniform - Hybrid
10 --- --.--...-..--_- ,_.._.-.-..-__- .

 

 

 

 

 

 

..‘

Graphics Related Ads Unrelated Ads

Decision Time (in s)
o :0 4s o 00

a
V

I Uniform Hybrid

 

 

 

 

 

“-1“..-

Graphics Related Ads Unrelated Ads
Image Type

Decision Time (in s)

Figure 4: Mean time (in seconds) taken for (a) young adults and (b) older adults
to make a correct decision in the Experiment 1 image memory test.

Older adults took signiﬁcantly longer to make a decision on all three types of
images (M: 6.17; CI=5.55-6.79) than did the younger adults (M: 3.10; 2.48-3.73).

The interactions between age group and image type, and age group and

71

presentation condition, were not signiﬁcant.

An additional 5 (Image Type) X 2 (Age Group) repeated measures ANOVA

was conducted to compare the time taken to make decision for all image types in the

Hybrid condition. These means as shown in Figure 5.

I

ime (m s

T

isnon

Dec

I Young Old

 

 

         

    

._..T...... ._....- .._,-........

-. . “‘1’ _____ '_-_-..‘

Graphics Related Graphics Unrelated Graphics
(GO) Ads (RA) Ads (UA)

Image Type

Figure 5: Mean time (in seconds) taken to correctly recognize images by
participants in the Hybrid presentation condition on the Experiment 1
image memory test.

There was a signiﬁcant main effect ofimage type, F(4, 248)=8.87, p=.01,

112:.12, and a signiﬁcant main effect of age group, F(l, 62)=40.78, p<.01, 112:.40. The

pattern of these main effects replicated the pattern found in the original analysis. In

this case, there was no signiﬁcant differences in the time taken to make a decision

on a graphic based on whether the graphic had originally been presented with other

graphics or with ads. Therefore, decisions on all graphics (graphics [GO], graphics

[RA], and graphics [UA]) were signiﬁcantly faster than decisions on related or

72

unrelated ads. The interaction between age group and image type was not
signiﬁcant, F(4, 248)=1.27, p=.28.
Goal Task Performance

Four measures of goal task performance were included in this experiment:
web page viewing time, content question accuracy, time taken to answer the content
questions, and rate of lure selection. As the preliminary experiment replicated
previous ﬁndings of little or no distraction effects in response to the inclusion of ads
on web pages, few distraction effects were predicted for this experiment. However,
in addition to the young adults tested in the preliminary study, older adults were
included in this experiment. Since older adults have been found to be more
susceptible to distraction than young adults on a variety of laboratory tasks, it
seemed possible that the older adults in this experiment could show distraction
effects where the young adults in the preliminary study did not. In particular, it was
expected that older adults may have more difﬁculty ignoring relevant ads than
irrelevant ads. In addition, it was expected that older adults would take longer than
the young adults to complete tasks, as indicated by elevated reaction times.

Web page viewing time. Average time spent per page was computed for each
image condition, and used to detect whether participants spent longer on the web
page viewing task when ads were present. A 4 (Image Type) X 2 (Presentation
Condition) X 2 (Age Group) repeated measures ANOVA was conducted to compare
the average web page viewing times. There was a signiﬁcant main effect of image

condition, F(3, 372)=S.36, p<.01, 112:.04, however, this effect was in an unexpected

73

direction. The means indicated that participants spent signiﬁcantly less time on
websites displayed in the related ads condition (M= 68.28; CI =64.53 - 72.03) than
those displayed in any of the other conditions (Text-Only: M: 72.80; CI=69.00 -
76.59; Graphics Only: M: 73.86; 61:69.29 - 78.43; Unrelated Ads: M: 72.84;
CI=68.66 - 77.02). The mean time spent viewing pages containing graphics or
unrelated ads did not differ. Means on this measure are shown in Table 10.

Table 10: Mean time (and 95% conﬁdence interval) in seconds spent viewing web
pages in Experiment 1.

Younger Adults Older Adults

7 Uniform ﬂ Twin 7 06E};— “ifyirﬁ —
-TextOnIy T P 69.95 ‘ 355—14 ‘ 8E _____ — ‘Eai T”

(52.34-67.53) (47.54-62.73) (81.68-96.86) (79.25-94.44)
Graphics Only 60.35 57.46 91.74 _ 85.90

(51.21-69.49) (48.32-66.60) (8260-10089) (76.75-95.04)
Related Ads 52.46 53.91 88.43 78.31

(44.96-59.96) (46.41-61.41) (80.93-95.93) (70.81-85.81)
Unrelated Ads 57.83 58.17 92.75 82.63

(49.46-66.19) (49.81-66.54) (84.38-101.11) (74.26-90.99)

There was also a signiﬁcant effect of age group, F( 1, 124)=66.56, p<.01,
112:.35, indicating that older adults (M: 86.98; CI=81.82 — 92.14) spent more time

per page than did young adults (M: 56.91; CI=51.75 - 62.07). This ﬁnding was

74

consistent with predictions. No other comparisons obtained signiﬁcance.

The analysis of page viewing time in Experiment 1 does not replicate the
preliminary ﬁndings of a marginally signiﬁcant effect of image type where pages
with related ads were viewed somewhat longer than pages in the other image
conditions. In fact, the pattern of results in this experiment is in the opposite
direction. To further explore this unexpected result, additional analyses
investigating possible practice effects were conducted.

The current experiment presented image conditions blocked by passage
topic, while the preliminary study presented image conditions randomized by page.
Therefore, it could be that inﬂuence of related ads on web page viewing time
changes from page to page of a blocked passage. To explore this possibility, a 4
(Image Type) X 2 (Presentation Condition) X 2 (Age Group) repeated measures
ANOVA was conducted on page viewing times from only the ﬁrst page in a passage,
instead of on the average page viewing time in the passage. In this analysis, the main
effect for image type was only marginally signiﬁcant, F(3, 372)=2.33, p=.07, 112:.02.
However, the pattern of results was the same, in which pages in the Related Ads
condition (M: 72.03; CI=66.79-77.28) were viewed for signiﬁcantly less time than
pages presented in the Text Only (M: 79.17; CI=74.14-84.20) or the Unrelated Ads
(M: 78.66; CI=72.52-84.80) conditions. Pages in the Graphics Only condition (M=
77.88; CI =71.82-83.94) were not signiﬁcantly different from any of the other image
conditions. Since the pattern of results from the ﬁrst pages of the passages

replicates the overall pattern of results, it seems unlikely that practice effects across

75

individual pages of a passage masked early distraction effects.

An alternative possibility is that practice effects could have emerged after the
ﬁrst passage was completed. To investigate this, a 4 (Image Type) X 2 (Presentation
Condition) X 2 (Age Group) ANOVA was conducted using the average page viewing
times from only the ﬁrst passage encountered by participants. It is important to
note that in this case, image type becomes a between subjects variable, since each
participant viewed the ﬁrst passage in only one of the possible image type
conditions. This analysis found only a main effect for age group, F( 1, 112)=56.80,
p<.01, 112:.34. None of the other effects obtained signiﬁcance.

Content learning questions. Mean accuracy on the content learning questions
was computed and compared in a 4 (Image Type) X 2 (Presentation Condition) X 2
(Age Group) repeated measures ANOVA . These data are shown by age group and
condition in Table 11. As in the preliminary study, no signiﬁcant distraction effects
emerged in this measure of goal task performance. Furthermore, older adults

performed just as well on this measure as young adults.

76

Table 11: Mean accuracy (and 95% conﬁdence interval) on content learning
question in Experiment 1.

 

Younger Adults . Older Adults

015101;: “ﬁg-bid T 63:63.? Hybrid T
TextTOTnIy ‘ T87T T84T TBET T TEETT

(83-92) (79-88) (81-90) (84-94)
Graphics Only .86 .85 .83 .87

(.81-.91) (.80-.90) (.78-.88) (.82-.92)
Related Ads .87 .87 .84 .88

(83-92) (83-91) (80-88) (83-92)
Unrelated Ads .85 .84 .83 .88

(.80-.90) (.79-.89) (.78-.88) (.82-.92)

In addition to scoring the content learning questions for accuracy, the time
taken for participants to select a correct answer was also recorded, and separate
averages were computed for age groups, image types, and presentation conditions.
The 4 (Image Type) X 2 (Presentation Condition) X 2 (Age Group) repeated
measures ANOVA found only the predicted main effect for age group, F( 1,
124)=72.71, p<.01, n2=.37, indicating that older adults (M= 16.89; CI=15.78-17.99)
take longer than young adults (M: 10.15; CI=9.04-11.26) to answer the questions.
No other comparisons attained signiﬁcance.

Lure selection. Finally, the content learning questions each contained a lure

77

choice that corresponded to the content of one of the related ads. Therefore, if
participants were choosing the lures due to confusion between the target text
content and the distractor ads' content, lure selectiOn rates should be elevated in the
related ads condition. However, the rate of lure selection was extremely low
(occurring 3% or less of the time), and there were no cases in which the lure
selection in the Related Ads condition was signiﬁcantly higher than the lure
selection rate in the Text Only, Graphics Only, or Unrelated Ads conditions (all ts<2,
ps>.05). As a result, the data from this measure was considered uninformative and
was not examined further.
Reading With Distraction Task

The reading with distraction task was included as an independent measure
of susceptibility to distraction. Since the participants were randomly assigned to the
Uniform and the Hybrid manipulation conditions, there was no reason to predict a
V difference between the two presentation conditions on any of the reading with
distraction measures. Nevertheless, independent samples t-tests for older and
younger adults assigned to the Uniform and the Hybrid conditions were conducted
to verify that there were no signiﬁcant differences between the two presentation
conditions. The two presentation groups were equivalent on measures of reaction
time, and accuracy for both the low and the high distraction conditions (all ts<2,
ps>.05). Consequently, presentation condition was excluded as an independent
variable in the following analyses.

The reading with distraction task was selected for its reliability in detecting

78

age differences in susceptibility to distraction (Darowski et al., 2008). The older
adults tested in this study did not show any increase in susceptibility to distraction
from Internet ads relative to their younger counterparts on the website viewing
task. This ﬁnding could indicate that older adults in general are just as able to ignore
Internet ads as young adults. However, it is also possible that the sample of older
adults selected for this study are higher performing, and consequently, less
susceptible to distraction than their average age-matched peer. If this is the case, the
typical interaction between age and distraction type should not emerge in the
reading with distraction task.

Reading time. A 2 (Distraction type: High or Low) X 2 (Age Group) repeated

measures ANOVA was conducted on the reading time means shown in Figure 6.

I Young Old

I

 

 

 

 

 

m

= 140 ---~--

'43 3

<9

lg

[-

on

IE

'6

G

a

Z

=

a

é

Low Distraction High Distraction
Passage Condition

Figure 6: Mean reading time in the Experiment 1 reading with
distraction task.

This analysis found a signiﬁcant main effect for age, F(1, 126)=77.55, p<.01,

79

n2=.38, and a signiﬁcant effect of distractor type, F (1, 126)=422.99, p<.01, n2=.77.
However, both of these main effects were qualiﬁed by the predicted interaction
between age and distraction type, F (1, 126)=19.40, p<.01, n2=.13. This interaction,
indicated that the older adults were more slowed by the high distraction passage
than were the young adults. This ﬁnding replicates the pattern of results typically
reported in this task, in which older adults are more susceptible to distraction than
the young adults.7

Content learning questions. Accuracy on the content learning questions was
computed, and compared in a 2 (Distraction type) X 2 (Age Group) repeated
measures ANOVA. There was a signiﬁcant age group effect, F( 1, 126)=9.91, p<.01,
n2=.07, in which older adults (M: .75; CI=.72-.78) were less accurate than young
adults (M: .81 ; CI=.78-.84). There was also a signiﬁcant main effect for distraction
type, F( 1, 126)=97.63, p<.01, n2=.44, indicating that participants were less accurate

in the high distraction passages (M: .70, CI=.67-.73) than in the low distraction

 

7 Since the older adults included in this study were signiﬁcantly slower than the
young adults on the processing speed tasks, it is appropriate to correct for
slowing prior to interpreting this interaction. As there are disputes in the
literature regarding the appropriate way to correct for slowing (e.g. Speiler,
Balota & Faust, 2000), two method of correction were. First, this analysis was
repeated “Sing a speed comPOSite score [(2 Pattern Comparison T Z Letter Comparison)
/ 2] as a covariate, with no change to the overall pattern of results. Secondly,
Standardized diStraCtion effeCts [(M High Distraction ' M Low Distraction) / M Low
Distraction] were compared between age groups. In contrast to the results of the
ﬁrst analysis, this method resulted in no signiﬁcant difference between age
groups, t(128)=-.08, p=n.s. Nevertheless, I have chosen to present this interaction
as signiﬁcant, believing that the covariate method of correcting for generalized

slowing may be a better method as it utilizes an independent measure of
processing speed.

80

passages (M: .87; CI=.85-.89). The interaction between distraction type and age
group, however, was not significant, indicating that older adults' accuracy was not
differentially affected by distraction when compared to younger adults' accuracy.

Predicting distraction control in the web viewing task. Although no overall
performance decrements (deﬁned as longer viewing times or lower accuracy) on
the website viewing task were observed in the presence of advertisements, it is
possible that some individuals who are particularly susceptible to distraction may
have demonstrated decrements. In order to explore this possibility, a measure of
susceptibility to distraction was created for each participant by subtracting the time
taken to read Low Distraction passages from the time taken to read High Distraction
passages. Larger differences between the two distraction conditions generated
higher scores on this measure, signifying greater susceptibility to distraction
(Darowski et al., 2008). Within each age group, participants were ranked by their
susceptibility to distraction scores. The third with the lowest scores were termed
Low distractible, the middle third were termed Middle distractible, and the third
with the highest scores were termed High distractible.

For each age group, a 3 (Susceptibility to Distraction: Low, Middle, High) X 4
(Image Type) X 2 (Presentation Condition) repeated measures ANOVA was
conducted on the measure of time spent viewing web pages. The analysis of the
young adults' data replicated the main effect of image type found earlier, F(3,
174)=5.95, p<.01, n2=.09. No other signiﬁcant effects were found in the young adults

analysis, nor in the older adult analysis. A similar analysis was conducted on the

81

measure of content question accuracy for each age group, but no signiﬁcant effects
were found. This pattern of results seems to indicate that susceptibility to
distraction on the Reading with Distraction task does not predict performance on
the website viewing task.8

Even if susceptibility to distraction did not predict performance on the web
page viewing task, the possibility remains that participants who were more
susceptible to distraction may have processed more images, as measured by image
recognition accuracy. To test this possibility, a 3 (Susceptibility to Distraction) X 3
(Image Type) X 2 (Presentation Condition) repeated measures ANOVA was
conducted on the image recognition data for each age group. This analysis replicated
the main effect of image type in both the young adult data, F(Z, 116)=22.02, p<..01,
112:.28, and the older adult data, F(Z, 116)=15.46, p<..01, n2=.21. No other
differences were significant. In conclusion, then, distraction scores on the Reading
with Distraction task, although reliable, do not appear to predict the degree to
which participants will attend to Internet advertisements.
Role of Experience

Other researchers (Bruner & Kumar, 2000; Dahlen, 2001; Raman 81 Leckenby,

 

8 In addition, the correlation between performance on the information searching
task and the residuals from the reading with distraction task were examined,
with no signiﬁcant relationship found. Two difference scores were used as
measures of information searching performance: Time spent viewing RA pages
minus time spent viewing TO pages and accuracy on T0 pages minus accuracy on
RA pages. The residuals did signiﬁcantly correlate with the raw measures of web
page viewing time in all four image type conditions. However, this effect was
attributed to the inﬂuence of reading speed, rather than distractor interference.

82

1998) have found that experience with the Internet appears to modulate the impact
of Internet ads. Although the older adults tested in this experiment had signiﬁcantly
less experience with the Internet than did the young adults, the majority of the older
adult sample had previous experience with the Internet. If any experience with the
Internet is sufﬁcient for enabling participants to efﬁciently control attention in this
environment, then a comparison between older adults with virtually no previous
Internet experience and older adults with substantial previous Internet experience
should reveal a difference in distraction control abilities.

In order to investigate this possibility, 7 participants who reported little to no
experience with the Internet on the Internet experience questionnaire (0-3 points;
M=1.14) were age-matched to 7 participants reporting more extensive experience
with the Internet (12-17 points; M=15.28). The two groups were not significantly
different in age, t(12)<1, but were signiﬁcantly different on the selection variable,
Internet experience, t(12)=-18.60, p<.01. The high Internet experience group also
had significantly more education(High: M=17.00; Low: M=14.16, t(11)=-2.48,
p=.03), and were signiﬁcantly faster on the speeded comparison tasks, as measured
by the speed composite score (High: M=-0.78; Low: M=-1.50; t(12)=-2.54, p=.03).

There was a signiﬁcant main effect of Internet experience on content
question accuracy, F(1,12)=5.79, p=.03, n2=.32, indicating that the High Internet
experience group (M=.89; CI=.82-.95) outperformed the Low Internet experience
group (M=.78; CI=.72-.85). However, as the High Internet experience group had

signiﬁcantly more education, this ﬁnding could reﬂect differences in reading skill as

83

opposed to differences in susceptibility to distraction.

In order to determine whether the Low Internet group was speciﬁcally
disadvantaged in ad-present passages, separate t-tests were conducted to compare
the groups on passages in each of the image conditions. Signiﬁcant differences
between the groups on content accuracy in the two ad-present conditions (Related
Ads: t(12)=-2.22, p=.05; Unrelated Ads: t(12)=-2.27, p=.04) indicating that the group
with High Internet experience (Related Ads: M=.89; Unrelated Ads: M=.92) was
signiﬁcantly more accurate on the ad-present passages than the Low Internet
experience group (Related Ads: M=.80; Unrelated Ads: M=.70). There was no
signiﬁcant difference between the two groups' accuracies for the graphics pages,
t(12)=-.59 or the text only pages, t(12)=-1.27. It is possible that this pattern of
results indicates that participants with little Internet experience are more disrupted
by the presence of advertisements than individuals with substantial Internet
experience.

However; there is an alternative explanation for this pattern of results that
cannot be ruled out. These samples were selected on the basis of their Internet
experience scores, and were not selected to be evenly distributed across the 16
different experimental forms. Consequently, the counterbalancing of passage topics
across presentation conditions is not likely to be complete in such a small sample.
Consequently, the difference between the two Internet experience groups on the ad-
present passages could be a materials effect. Analyses of overall accuracy for each of

the four passage topics (using the full data set of participants) indicates that the

84

content learning questions were of varying difﬁculty (Health: M: 81%; Loan: M:
89%; Love: M: 82%; Travel: M: 92%), which could have resulted in the accuracy
differences displayed in this small sample.
Discussion

Experiment 1 was designed to replicate the preliminary study, which
produced some (albeit non-signiﬁcant) indication in the measure of image
recognition, and in the measure of web page viewing time, that participants may be
using semantic and perceptual features to enable attentional ﬁltering of ads. In
particular, Experiment 1 was predicted to show that images related to the content of
the goal task were signiﬁcantly more likely to be processed, and consequently, were
more likely to result in distraction effects when present. Furthermore, it was
predicted that ads may be more likely to be processed when presented along with
graphics, than when presented with other ads. In addition, Experiment 1 extended
the ﬁndings from the preliminary study by testing older adults as well as young
adults. Older adults were primarily predicted to show age differences when images
were related to the target content, as was the case with the graphics and related ads.
Attention to Advertisements

Attention to advertisements was measured by accuracy and decision time on
a two alternative forced choice recognition test. A signiﬁcant difference was found in
the participants' accuracy in recognizing different types of images, where graphics
presented in the Graphics Only condition were better recognized than related ads,

which in turn were better recognized than unrelated ads. A similar pattern was

85

found in the time taken to make a correct decision, where decisions on graphics
were made signiﬁcantly faster than decisions on either type of ad. These ﬁndings
were consistent with predictions, and as such, are interpreted as an indication that
participants were utilizing semantic ﬁlters.

In addition, the pattern of image recognition seems to support the idea that
distinctive perceptual features are not likely to result in attention capture when they
are not part of the current attentional set. If attention to images was controlled only
by the presence of visually salient features, both types of ads should have resulted in
higher recognition rates than the graphics, as half of the ads were animated and
none of the graphics were animated.

It was also predicted that older adults may have more difﬁculty than young
adults in ﬁltering out images that were semantically related to the target content.
Although older adults took signiﬁcantly longer to make a decision in the image
recognition test, there were no signiﬁcant differences in the overall pattern of
recognition results between the two age groups.

It was also predicted that recognition of ads would increase when the ads
were presented with graphics as opposed to with other ads, thereby supporting the
hypothesis that participants may be more likely to use perceptual features to guide
attention when they are consistent with the current attentional set. Unfortunately,
there was no evidence of this type of an effect in the current experiment, as no
signiﬁcant differences between presentation conditions emerged in either the

recognition accuracy or recognition speed data.

86

In conclusion, the current study only found support for the use of semantic
ﬁlters. Participants appear to be more likely to attend to images in general, and ads
in speciﬁc, when they are semantically related to the target content. The ﬁnding that
graphics were better recognized than related ads can be viewed as an indication
that participants judged the graphics to be more task relevant than the related ads,
even though neither type of image contained information needed for the completion
of the goal task. Although it is likely that this effect truly reﬂects the inﬂuence of
semantic content, the lack of image counterbalancing across conditions leaves open
the possibility that this pattern of results could be caused by a materials effect in the
selection of images. In order to test this possibility, it is important to design future
experiments in such a way as to enable complete counterbalancing of images across
conditions.

Goal Task Performance

Goal task performance in the current study was measured by the average
time spent per passage page, as well as by several measures of content learning
(speed and accuracy in answering content questions and selection rate of lure
choices). It was predicted that, if observed, distraction effects would most likely
occur in the Related Ads condition. It was also predicted that older adults would be
somewhat more likely than younger adults to show distraction effects in this
condition. However, both the current study and the preliminary study supported the
results of earlier research (Helder, under revision) that failed to ﬁnd substantial

distraction effects from the inclusion of Internet ads in an information search task.

87

The preliminary study found a possible distraction effect in the measure of
web page viewing time, as participants in this experiment appeared to spend
somewhat more time viewing web pages in the Related Ads condition than they
spent viewing pages in any of the other conditions. However, Experiment 1 failed to
replicate this ﬁnding, and in fact produced a signiﬁcant result in the opposite
direction, indicating that overall, participants spent signiﬁcantly less time on pages
in the Related Ads condition than on pages in any of the other conditions. The
reversal in the pattern of results between the preliminary study and Experiment 1
was explored by testing for practice effects across the individual pages in a passage,
and across subsequent passages in Experiment 1. No evidence of practice effects
across individual pages in a passage, nor across the four passages, were found.
Consequently, practice effects do not seem to be a likely explanation for this pattern
of results.

A more likely explanation involves a materials effect, as alluded to in
conjunction with the results of the recognition test. Very few of the images used in
the preliminary study were also used in Experiment 1. Therefore, the selection of
images of differing visual salience and intrinsic interest across experiments could
account for the different pattern of viewing times. Furthermore, just as the images
were not held constant across experiments, the images used in the different image
conditions were not counterbalanced. An image that served as a related ad in the
health passage in Experiment 1 did not serve as an unrelated ad in the other

passages. Consequently, the images selected to serve as related ads in Experiment 1

88

may have been less interesting or less salient than the images selected as graphics
and as unrelated ads, resulting in a decrease in page viewing time in the related ads
condition. However, it is interesting to note that if this overall decrease in Related
Ads page viewing time resulted from a decrease in time spent viewing the actual
ads, this reduction did not translate into a reduced recognition rate of related ads.

The preliminary study evaluated content learning using short-answer format
questions, with a question density of one question per page. No signiﬁcant
differences between any of the conditions were found. Efforts were made to
improve the measure of content learning in Experiment 1. Participants answered 4-
alternative multiple choice questions, with a question density of two questions per
page. In addition, the answer set for each question contained a lure that
corresponded to information contained in one of the related ads displayed on that
page of content. This addition was included as a means of measuring the degree to
which ad content had been processed. Finally, time taken to select a correct answer
was also recorded. Despite these efforts to improve the measure of content learning
in Experiment 1, no signiﬁcant differences were observed in the content learning
accuracy, reaction time, or lure selection measures for either young or older adults.
Age Differences

As with previous studies including both young and older adults, it was
predicted that older adults could be more susceptible to distraction from
advertisements. No support for this prediction was found in Experiment 1. The

volunteer nature of the older adult sample, however, could be be associated with

89

better than average distraction control abilities. The Reading with Distraction task
(Connelly et al., 1991) was used to test for age differences in susceptibility to
distraction. The typical interaction between distraction type and age group emerged
on this measure, indicating that the older adults' ability to ignore distraction in the
form of Internet ads did not appear to be associated with age atypical abilities to
control distraction on this task. An additional exploratory analysis of performance
based on Internet experience was conducted, but no conclusive evidence of an
experience effect emerged.

Conclusion

The null results obtained on measures of goal task performance are similar
to those obtained in previous experiments (Helder, under revision), and appear to
support the prediction that Internet users are not disadvantaged in their search for
information when ads appear with target informational content. Nevertheless,
participants did display reliable recognition of ads in general, even when exposure
to the ads was incidental, and not relevant for task completion. Consequently, this
set of experiments can be viewed as providing encouraging results for both Internet
advertisers, and Internet users.

The data from the image recognition test indicates that Internet users are
likely to attend to ads enough to facilitate their later recognition. Furthermore, ads
are equally likely to be recognized whether they appear in competition with other
ads, or when they appear paired with graphics. Finally, it appears that advertisers

would be wise to choose to advertise primarily on websites that are related to the

90

product or service being advertised, particularly since Internet users seem to be
more likely to process images when they are related to a topic they ﬁnd intrinsically
interesting.

On the other hand, Internet users do not appear to show signs of consistent
signiﬁcant performance decrements in an information search task, as measured by
search speed and accuracy, when ads appear alongside the target information. Given
the results of reliable ad recognition, this ﬁnding indicates that participants are able
to allocate attention to ads, without impairing their ability to attend to the
information search task. This ﬁnding is especially encouraging for older adults, who
have been found to be differentially susceptible to distraction on a large number of
laboratory tasks. Experiment 1's ﬁnding of no signiﬁcant age differences in
distraction control abilities in an Internet environment, rife with distractors, could
indicate that older adults may be better able to control distraction in real world
tasks than in laboratory tasks.

However, there is a caveat to this interpretation. Participants in the current
study were engaged in a highly focused, well-constrained information search task.
Participants in this experiment may have more focused, and possibly more
motivated, to complete the task as quickly as possible than the average real world
Internet searcher. Consequently, although distraction effects did not emerge in this
experiment, it is possible that Internet users carrying out self-directed searches

outside the lab may be more susceptible to distraction from Internet ads.

91

THE INFLUENCE OF LOCATION-BASED ATTENTIONAL FILTERS ON ATTENTION TO
INTERNET ADVERTISEMENTS

The location of items onscreen may be key to enabling participants to direct
attention to target content and away from irrelevant ads. It is predicted that Internet
users may utilize spatial cues to help guide attention in this manner, similar to the
way participants in visual marking paradigms are able to use item location reduce
the processing of items previously determined to be distractors. The use of spatial
cues is likely to be in conjunction with processing items for semantic content, as
the relevance of individual items for task performance needs to ﬁrst be determined
prior to the marking of distractors. In particular, once the page loads, participants
could quickly process the page layout, and determine whether each image is task-
relevant or task-irrelevant. Participants could then mark the location of the
irrelevant images, such as ads, on the page and avoid re-processing them as they
focus on processing relevant page content.

The initial attention directed toward the ads as they are processed for
relevance, then marked for location, could account for previous ﬁndings that some
ads may be better recognized than others. It could be that ads that hold interest for
the person, whether that is intrinsic interest, or interest stemming from the
particular task demands at hand, may undergo a greater degree of processing. This
hypotheses was supported in the results of Experiment 1's image recognition test.
The efﬁciency with which spatial cues can be used to guide attention could account

for the lack of distraction effects on measures of reading performance in the

92

preliminary study and Experiment 1.

To test this hypothesis, this experiment introduces a secondary task designed
to interfere with participant's abilities to limit processing in regions of the screen
determined to hold task irrelevant items. This was done by incorporating a
secondary task that presented stimuli at various locations onscreen, sometimes
coinciding with the location of primary task relevant items, and sometimes not. The
need to complete this secondary task is likely to result in fewer attentional
resources being available in the primary information search task. Consequently,
performance decrements on the web page viewing task are likely to be observed
simply due to the dual task nature of this condition, and not necessarily due to any
task-speciﬁc interference with a location-based ﬁlter. To control for this inﬂuence, a
second dual task condition was used in which participants are involved in a similar
secondary task, but the stimuli appeared in just one ﬁxed location onscreen.
Performance in these two dual-task conditions was compared to performance in a
no dual task baseline condition.

Predictions

As in Experiment 1, it was predicted that evidence of attention to ads in
general would emerge on the recognition test. More speciﬁcally, it was predicted
that participants in this experiment would show evidence of using spatial cues to
guide attention. Participants completing a secondary task in which stimuli appeared
in multiple (variable) onscreen locations were expected to be less able to limit

processing in regions of the screen containing irrelevant images. Consequently, they

93

were predicted to show an increase in recognition of images in general, due to
increased processing of irrelevant images. On the other hand, it was predicted that
participants would be least likely to process images in a dual task condition in which
stimuli appeared in only one (ﬁxed) location onscreen. The rationale behind this
prediction is that participants engaged in a secondary task should have fewer
attentional resources available for processing images. This decrease in available
attention, coupled with a dual task condition which should not interfere with
location cuing, was expected to lead to lower image recognition rates. Image
recognition for participants in the no dual task baseline condition was expected to
fall between the two dual task conditions.

In addition, participants were predicted to again show evidence of guiding
attention based on the semantic relatedness of an object to the target content, with
attention to semantically related graphics and ads being greater than attention to
unrelated ads. A possible interaction between location cues and semantic content is
predicted as well, in which the recognition of unrelated ads increases more
substantially than the recognition of graphics or related ads under dual task
conditions. This dual task effect is not predicted to occur when secondary task
simuli appear in only one (ﬁxed) location onscreen.

It is also predicted that the introduction of a secondary task will result in
decrements in performance of the primary search task, as measured by search
speed and search accuracy, as it will result in less attentional resources being

available for the primary task. Furthermore, these decrements are expected to

94

increase as the difficulty of the secondary task increases. Therefore, greater
distraction effects are expected in the variable location condition of the dual task
than in the ﬁxed location condition.
Materials and Methods
Participants

The participants in Experiment 2 included the young adults tested in the
Hybrid condition of Experiment 1 plus 64 more young adults from the same
source(the psychology department subject pool), for a total of 96 young adult
participants (age 18-23; 29 males and one participant who chose not to report
gender). Testing on Experiment 1 and Experiment 2 occurred concurrently, and
participants were randomly assigned to experiments to reduce the likelihood that
selection biases would occur between the participants used in the baseline group
and in the two dual task groups. As in Experiment 1, all participants were ﬂuent in
English and had completed at minimum a high school degree. Six participants
reported learning disabilities .9 Analyses were repeated with learning disabled
participants excluded, with no change to the results.

Design

Experiment 2 was designed to replicate the Hybrid presentation condition of

Experiment 1, with the addition of two dual task conditions. Consequently, the data

from the 32 young adult participants in the Hybrid presentation condition of

 

9 Five diagnoses were for Attention Deﬁcit Disorder or Attention Deﬁcit-
Hyperactivity Disorder, one was of Obsessive Compulsive Disorder, and one was
for Dyslexia.

95

Experiment 1 were included as a no dual task baseline in this experiment to
conserve participant testing time and resources. The 64 additional subjects were
tested in two conditions involving a dual task. With the exception of this dual task,
the materials and procedures for these two additional groups of participants were
exactly the same as those used for Experiment 1.

Considering all three groups, Experiment 2 had a 4 X 3 mixed design. The
ﬁrst independent variable was image type (Text Only, Graphics Only, Related Ads,
and Unrelated Ads) and was manipulated within subjects. Each participant read one
passage in each of these four image conditions, with the order of the passage t0pics
and image conditions counterbalanced as in Experiment 1. The second independent
variable was the dual task condition with the following three levels: None, Fixed,
and Variable. The dual task conditions were manipulated between subjects.

Materials

The same passages used in the Hybrid presentation condition from
Experiment 1 were used in Experiment 2. The time spent viewing each page was
recorded. The content learning questions used in Experiment 2 were also identical
to those used in Experiment 1. Overall accuracy on the content learning questions,
rate of lure selection, and the the time taken to answer each question were recorded
and used to compare content learning across conditions.

The images used in Experiment 2 were the same as those used in the Hybrid
presentation condition in Experiment 1. Participants viewed the four websites in

one of three dual task conditions: None, Fixed, and Variable. As mentioned before,

96

the task in the None dual task condition was the same task used in the Hybrid
presentation condition in Experiment 1, and as such, used the data from the young
adults in this condition in Experiment 1. The Fixed and Variable dual task conditions
required participants to carry out a secondary task while they were engaged in the
website viewing task. In these conditions, participants were given the secondary
task of “collecting gems" as quickly as possible when a gem icon appeared on the
screen.

Gems were represented by six different gem icons. Each gem icon was
contained within an invisible 40 x 40 pixel square. The gems were different colors
and different shapes. The pairing between color and shape remained constant
throughout the experiment. Gray-scale images of the gems are depicted in Figure 7

below, with the actual gem color indicated in the text below the gem.

White Green Blue Orange Red Yellow

 

Figure 7: The six gem shapes and colors used in
the dual task component of Experiment 2.

Each gem was an animated ﬁle that appeared onscreen gradually by
randomly adding pixels to the gem until the image was complete. This animation
involved 55 frames, each of which lasted 150ms. Consequently, gems took 8100ms
to go from no visible icon to a complete icon. This transition is shown at

approximately 1s timestamps in Figure 8.

97

:f-‘v-i.',=‘,j:-,_', 1' 23-7-33 ﬁtting m
5.. .“ 1'. .3 E: .314 2% ‘ . t’: 1 it;

. '10:" 'u .I 5"" - I § '.."' (E: I
n - '.; ~Iqb ‘61-:l0- _5- 1': *TﬁACHLJ 0‘ : Sui-‘5'; .L 28.01);

 

0 1050 1950 3000 4050 4950 6000 7050 8100

Figure 8: Representation of the gem materialization animation sequence in
Experiment 2, with timestamps (in ms) below each image.

The background of the gem was transparent, allowing the background of the
web page to show through until blocked by a gem-pixel. This procedure reduced the
likelihood that the abrupt onset of the gem image could result in automatic
attention capture. Instead, it was assumed that participants would have to maintain
attention to areas of the screen in which the gems could appear.

For each of the four passage topics, four potential gem locations were
identiﬁed. These gem locations were selected to be distributed across the area of the
screen in which content and images appeared, as shown in Figure 9. The same set of
four gem positions was used throughout a passage, but varied across passages.
Within each image-present condition of a passage (Graphics Only, Related Ads and
Unrelated Ads), six of the images could potentially have a gem land on it, and six
images were never landed on by a gem. Efforts were made to select gem locations
that would enable a gem to potentially land on an image on every page of a passage,
but this was not always possible. There was one page in the loan passage, and two
pages in the love passage in which none of the images aligned with any of the

possible gem positions.

98

 

 

 

 

 

 

 

 

 

 

 

 

- . -.- ~..—a,........ ~w.._--.. -..

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 9: Page layout indicating image positions (open squares) and gem locations
(ﬁlled squares) for all six pages of the Health Information passage.

To collect the gems, participants pressed the zero (0) key‘ on the number pad
of the keyboard, indicating that the gem was perceived. Participants were instructed
to press the key as soon as they saw a gem onscreen Once the zero key was pressed
to indicate that the gem has been perceived, it disappeared from the screen.
Participants were encouraged to collect gems quickly by providing them with
feedback on their average gem collection speed at the end of each passage. Each trial
(marked by the loading of a new web page) began without a gem onscreen. The ﬁrst
gem began appearing 4-7s after the page had completed loading. The interval

between the offset of one gem and the onset of the next gem varied between 4000

99

and 6999ms, averaging 5482ms (SD=863ms).

In the Fixed condition, all gems appeared in the same location on the screen
for an entire passage, resulting in a dual-task condition in which the location of
secondary task stimuli was predictable. Since the location of the gems in this
condition was predictable, allocating attention to the gem collecting task in this
condition was not predicted to interfere with spatial location cues. In the Variable
condition, the location in which the gems appeared onscreen was randomly selected
from the set of four possible locations for each passage. Because gems were likely to
appear in several onscreen locations in this condition, allocating attention to the
gem collecting task was predicted to interfere with participants’ ability to direct
attention according to spatial location cues.

The total number of gems collected on each page, and the time taken to
collect each gem, was recorded for each participant. These measures were used as
indicators of the difﬁculty of the two dual task conditions, as well as of participant's
involvement in the secondary task

The remaining materials used in the experiment—the memory test for
images displayed in the experiment, the questionnaires, the processing speed tasks,
and the reading with distraction task—were identical to the ones used in
Experiment 1.

Procedure
The procedure for Experiment 2 followed the same task order as the

procedure in Experiment 1. However, participants who completed the experiment in

100

either the Fixed or the Variable dual task conditions completed a brief gem-
collection practice immediately after listening to the website viewing task
instructions. The practice entailed collecting ﬁve gems displayed one at a time, using
the same timing parameters as the experimental conditions. The average collection
speed for the ﬁve practice gems was reported, and participants were encouraged to
try to maintain a similar collection speed throughout the experiment. After
completing the gem collection practice, participants continued to follow the
procedure of Experiment 1 by previewing a list of 12 content questions, and viewing
the ﬁrst passage. Participants in the Fixed and the Variable dual task conditions
needed to monitor the screen for the appearance of gems while viewing the web
page.
Results
Manipulation Checks
Participants
Participants were given the same 15-item vocabulary test, demographic
questionnaire, processing speed tasks, computer experience questionnaire, and
Internet experience questionnaire as participants in Experiment 1. The means on
these demographic measures, separated by dual task group, are presented in Table
12. These measures were used to establish that the three dual task groups were
equivalent. No signiﬁcant differences between any of the dual task groups were

found on any of the demographics measures, Fs between 0.11 and 2.23, ps>.11.

101

Table 12: Scores on the demographic measures included in Experiment 2.

Yrs Education

Healtha

Vocabularyb

Computer Expc

Internet Expd

Ad Attitudee

Ad Distractionf

Pattern
Comparisong

Letter
Comparisong

None

19.28
(19.79-19.77)

12.95
(12.48-13.42)

3.19
(2.92-3.45)

9.88
(882-1893)

40.22
(39.03-41.41)

16.81
(16.07-17.55)

1.66
(1.36-1.95)

71.88
(55.41-88.34)

41.00
(38.67-43.32)

24.09
(22.18-26.00)

Fixed

 

19.77
(19.28-20.27)

13.31
(12.83-13.79)

3.34
(3.13-3.56)

9.56
(865-1048)

39.65
(38.66-40.64)

16.03
(15.22-16.84)

1.82
(1.49-2.13)

78.13
(62.98-93.27)

41.78
(39.40-44.16)

24.44
(23.40-27.48)

Variable

 

19.22
(18.74-19.70)

12.62
(12.17-13.08)

3.47
(3.26-3.67)

9.66
(8.71-10.60)

39.11
(37.92-40.30)

16.22
(15.46-16.98)

1.75
(1.42-2.08)

68.75
(51.77-85.73)

40.53
(38.41-42.65)

25.19
(23.45-26.93)

NOTES: (a) Rating based on self-report: 1=worst health; 4=best health (b) Maximum score 15
(c) Most experience=50 (d) Most experience=22 (e) Positivity of attitude towards Internet Ads
with most positive=3; most negative=0 (f) Percent of participants claiming that they are
distracted by Internet ads { g ) Number of correct comparisons completed in 605.

Dual Task Performance

For the gem collecting task to have the potential to impact performance,
participants must consistently attend to both the primary web-viewing task and the
gem collecting task. The total number of gems collected per passage by participants,
as well as their reaction time in collecting the gems were used as indicators of
attention to the gem collection task.

On average, participants collected one gem every 9.90s (Cl=9.48-10.49s)
while viewing a web page, with the average collection rate ranging from one gem
every 7.50s to one gem every 19.825 across participants. The gem collection rate
indicates that participants were attending to the gem collection task appropriately.
The average gem collection time across participants and conditions was 2.375
(CI=2.17-2.58$). Since gems took approximately 8s to fully materialize, this
collection speed indicates that participants were vigilant and quick to respond to
gems.

The number of gems collected, and the speed with which they were collected,
can also be used to test the difﬁculty of the two dual task conditions. It was assumed
that the Variable condition of the gem collection task would be more difﬁcult than
the Fixed condition. This was tested by comparing the number of gems collected per
page in a 4 (Image Type) X 2 (Dual Task) repeated measures ANOVA. The means
used in this comparison are reported in Table 13. The predicted main effect of dual
task emerged, F( 1, 62)=7.04, p=.01, n2=.10, indicating that significantly more gems

were collected per page in the ﬁxed condition (M=7.34; CI=6.57-8.12) than in the

103

variable condition (M=5.89; CI=5.11-6.66). There was no signiﬁcant main effect of
image type on the number of gems collected per page, although the direction of the
means indicates that the more gems were collected on Text Only pages (M=6.80;
CI=6.13-7.46), than on Unrelated Ads pages (M=6.42; CI=5.84-7.01). The interaction
was also not signiﬁcant.

Table 13: Average number of gems collected per page in Experiment 2.

 

Fixed Variable
TIT-Romy T T T T7.T4T8T(T6T.Ts-T4T-8.T42T) T TT 6.11 (5.17-7.05)T T T
Graphics Only 7.62 (6.60-8.65) 5.85 (4.83-6.88)
Related Ads 6.92 (6.09-7.75) 6.08 (5.25-6.91)
Unrelated Ads 7.34 (6.51-8.17) 5.51 (4.67-6.34)

The average time taken to collect a gem was also compared in a 4 (Image
Type) X 2 (Dual Task) repeated measures ANOVA. The predicted dual task effect
emerged on this measure as well, F(1,62)=30.55, p<.01, n2=.33, indicating that
participants in the Variable location condition took signiﬁcantly longer to collect
gems than did participants in the Fixed location condition. A main effect of image
type was also found, F(3,186)=18.92, p<.01, n2=.23. However, these main effects
were qualified by a signiﬁcant interaction between image type and dual task
condition, F(3,186)=3.12, p=.03, 112:.05. As shown in Figure 10 below, the
interaction indicated that the dual task condition was more inﬂuential on gem
collection speed when there were images onscreen, than in the Text Only condition.

In summary, these measures support the prediction that the Variable location

104

condition should be more difﬁcult than the Fixed location condition.

$1.3: Fixed I Variable

 

 

 

 

.... 4 3“ — -

g 3.5

E 3 1

a, 2.5

3‘ 2 3

5 1.5 7‘

”t5" 1

2 0.5

3 o .
TextOnly Graphics Related Ads Unrelated

Only Ads
lmageTVpe

Figure 10: Average RT (in seconds) to collect gems in Experiment 2.
Reading with Distraction Task

The Reading with Distraction task was also included in Experiment 2. In
Experiment 1 it served as a manipulation check to make sure that the older adults
displayed the expected age-related increase in susceptibility to distraction. Similarly,
in this experiment, scores on the reading with distraction task can be used as a
manipulation check to ensure that there were no signiﬁcant differences in
distractibility between the three randomly assigned dual-task groups. A 2
(Distraction Level) X 3 (Dual Task) repeated measures ANOVA was conducted to
compare the three dual task groups on the difference scores for passage reading
time. The main effect of dual task, as predicted, was not signiﬁcant. There was a
signiﬁcant main effect of distraction level, F(1,93)=337.08, p<.01, n2=.78, indicating

that high distraction passages were read more slowly than low distraction passages.

105

lNone Fixed EEVariable

 

100 ~ ~-

 

 

 

Passge Reading Time (in 5)

High Low

Distraction Level

Figure 11: Mean passage reading time (in seconds) on the reading
with distraction task in Experiment 2.

Critically, however, there was a signiﬁcant interaction between distraction
level and dual task, F(2,93)=3.71, p=.03, n2=.07. As shown in Figure 11, this
interaction was due to the participants in the Fixed condition being somewhat less
affected by the high distraction condition than participants in the other two
conditions. In other words, this ﬁnding seems to indicate that the participants
assigned to the Fixed location condition may be somewhat less susceptible to
distraction than the participants assigned to the other dual task conditions. This
selection bias is somewhat problematic, as the predicted effect of the dual task
manipulation is that the participants in the Fixed location condition are less likely to
show distraction effects than participants in the Variable location condition.

In order to further explore the cause of this difference, the 2 (Distraction

level) X 3 (Dual Task) repeated measures ANOVA was repeated, with the

106

participants' processing speed composite score included as a covariate to control for
individual differences in processing speed. There was again a signiﬁcant effect of
distraction level, F(1,106)=415.43, p<.01, n2=.80, and no signiﬁcant effect of dual
task. However, the inﬂuence of processing speed was also signiﬁcant, F(1,106)=7.59,
p<.01, n2=.07. As a result, the interaction between distraction level and dual task
condition was no longer signiﬁcant, F(2,90)=2.67, p=.07, 112:.05. In conclusion, the
results of this analysis indicate that the three dual task groups should be
approximatley equivalent in their susceptibility to distraction, as measured by the
Reading with Distraction task, once individual differences in processing speed have
been controlled.
Attention to Advertisements

It was predicted that overall, recognition rates would reﬂect the inﬂuence of
semantic content, replicating Experiment 1. In addition, it was predicted that the
use of location cues would be most difﬁcult in the Variable location dual task
condition, in which participants needed to monitor several locations onscreen in
order to efﬁciently collect gems. This predicted difﬁculty in using location cues was
expected to manifest in elevated image recognition rates in the Variable location
condition.

A 3 (Image Type: Graphics [GO], Related Ads, and Unrelated Ads) X 3 (Dual
Task) repeated measures ANOVA was conducted to test these hypotheses. A main
effect of image type was found, F(Z, 186)=26.74, p<.01, n2=.22. The direction of this

effect was consistent with the predictions. Graphics presented with other graphics

107

in the Graphics Only condition (M=.85, CI=.83-.88) were recognized signiﬁcantly
better than related ads. In turn, related ads (M=.73; CI=.68-.77) were recognized

signiﬁcantly more than unrelated ads (M=.65; CI=.60-.70).10

INone Fixed ﬁVariable

 

 

 

Proportion Recognized

Graphics Related Ads Unrelated Ads

Image Type

Figure 12: Proportion of images recognized in the Experiment 2 image
memory test.

Contrary to predictions, there was no effect of dual task condition on image
recognition, F<1. There was also no signiﬁcant interaction between image type and
dual task condition, 112 = .03. Although the interaction was not signiﬁcant, it is worth
noting that the general pattern of the means provided some indication that the dual
task manipulation may have had more inﬂuence on the recognition rate of related
ads than on the recognition of either graphics or unrelated ads. This pattern is
shown in Figure 12.

10 There was no signiﬁcant difference between recognition of graphics presented
with related ads (M=.74; CI=.70-.79) and graphics presented with unrelated ads
(M=.77; CI=.73-.81), although both types of graphics were recognized signiﬁcantly

more than unrelated ads, and signiﬁcantly less than graphics presented with other
graphics.

108

It was predicted that recognition should increase when the locations of gems
were likely to draw attention to images. Although this effect did not emerge in the
overall data, it is still possible that recognition of speciﬁc images on which gems
appeared may be better than the recognition of images on which gems did not
appear. To test this prediction, overall recognition rates of images on which gems
appeared were compared to overall recognition of images on which no gems
appeared. This analysis included only participants in the gem-present dual task
conditions. Contrary to predictions, no signiﬁcant difference in image recognition
rates was found as a factor of gems appearing on an image, t(62)=.2 7, p=n.s. (Images
on which gems appeared: M=.76, CI=.70-.82; Images on which no gems appeared:
M=.76, CI=.74-.79).

The average time taken for each participant to reach a correct decision was
also computed. The pattern of decision time results was predicted to replicate those
found in Experiment 1, in which participants were faster to make decisions for
graphics than for ads. In addition, it was predicted that if participants in the Fixed
location dual task condition were least likely to process images, these participants
might take the longest time to make their decisions.

To test these hypotheses, the correct decision times were analyzed using a 3
(Image Type) X 3 (Dual task) repeated measures ANOVA. The predicted main effect

ofimage type emerged, F(Z, 184)=21.27, p<.01, 112:.19.“ Planned comparisons

 

11 One participant in the Variable location condition failed to correctly recognize
any of the unrelated ads, and was thus excluded from this analysis.

109

indicated that participants were signiﬁcantly faster to recognize graphics (M: 2.45,
CI=2.27-2.63) than either of the two types of ads.12 The time taken to recognize
related ads (M= 2.96, CI =2.77-3.14) and unrelated ads (M= 3.09, CI=2.86-3.33) did
not differ from each other. The main effect for dual task, and the interaction between

image type and dual task did not obtain signiﬁcance (Fs<1.5). These results are

 

 

 

 

depicted in Figure 13.
I None Fixed iii-Wariable

A 4
:9,
0
,§ 3
.—

2 _.._
L5.
6 1 '-
8

0 -._.- . _. _

Graphics Related Ads Unrelated Ads
ImageType

Figure 13: Time taken (in seconds) to make a correct decision on the
image memory test in Experiment 2.

Goal Task Performance
As in Experiment 1, four measures of goal task performance were included:

web page viewing time, content question accuracy, time taken to answer the content

12 Participants were also signiﬁcantly faster to recognize graphics presented with
ads (Graphics [RA]: M=2.35, CI=2.19-2.51; Graphics [UA]: M=2.30, CI: 2.11-2.49)
than to recognize either type of ads.

110

questions, and rate of lure selection. Once again, few distraction effects were
predicted for this experiment. However, it was predicted that distraction effects
would be most likely to emerge in the Variable location dual task condition, and in
particular, when related ads were onscreen.
Web Page Viewing Time

Average time spent per page was computed for each image type, and used to
detect whether participants spent longer on the web page viewing task when ads
were present. A 4 (Image Type) X 3 (Dual Task) repeated measures ANOVA was
conducted to compare the average web page viewing times. The overall pattern of

results is depicted in Figure 14.

INone - Fixed EVariable

 

 

 

 

 

 

'5‘
.E
75
.§
)—
u
C
E
0)
S
0)
iii
0.
5
§ ‘""‘"
Text Only Graphics Related Ads Unrelated
Only Ads
Image Type

Figure 14: Mean time (in seconds) spent viewing website pages in
Experiment 2.

111

There was a significant main effect of image condition, F(3, 279)=4.66, p<.01,
n2=.05, indicating that passages with related ads (M: 57.07; CI=53.48-60.66) are
viewed signiﬁcantly less than passages in any of the other image conditions (Text
Only: M=61.01, CI=57.22-64.81; Graphics Only: M: 62.89; CI=58.69-67.09;
Unrelated ads: M: 60.30; CI=56.70-63.89). This effect replicated the ﬁndings of
Experiment 1, although it was not anticipated in the predictions developed for the
preliminary study.

The dual task conditions were predicted to result in an increase in time spent
on task, as attention should be divided between the information gathering task and
the gem collecting task. Inspection of the means shows that in general, participants
engaged in the gem collection task to took longer than participants not engaged in a
secondary task, as shown in Figure 14. However; there was no signiﬁcant main effect
for dual task condition, F (2,93)=1.83, p=n.s., nor was there a signiﬁcant interaction
between image type and dual task condition, F(6,279)=1.59, p=n.s.

Content Learning Questions

Mean accuracy on the content learning questions was computed and
compared in a 4 (Image Type) X 3 (Dual task) repeated measures ANOVA. As in the
previous experiments, no signiﬁcant differences were found (the main effect for
image and the interaction between image and dual task produced Fs<1). Means on

this measure are shown in Table 14.

112

Table 14: Mean proportion correct on content learning questions in Experiment 2.

None Fixed
TethToﬁly” T T084T(To.79-T0.89) T T T0.8T2(T0'77:0.T8T7) ' T
Graphics Only 0.85 (0.8-0.91) 0.83 (0.77-0.88)
Related Ads 0.87 (0.82-0.92) 0.80 (0.75085)
Unrelated Ads 0.84 (0.78-0.9) 0.78 (0.72-0.85)

Variable

 

T039 674-084)
0.78 (0.72-0.84)
0.80 (0.75-0.85)

0.79 (0.72-0.85)

In addition to scoring the content learning questions for accuracy, the time

taken for partcipants to select a correct answer was also recorded, and separate

averages were computed for image types and dual task conditions. A 4 (Image Type)

X 3 (Dual Task) repeated measures ANOVA was conducted on the content question

RT measure, with no signiﬁcant differences between any of the conditions. These

means are shown in table 15.

Table 15: Mean time (in seconds) taken to select a correct answer on the content

learning questions in Experiment 2.

None Fixed

 

TethBnly T 10.3?(8813180) 11.28 (9.78-12.77)
Graphics Only 10.01 (862-1139) 10.89 (950-1228)
Related Ads 9.74 (842-1 1.05) 10.95 (9.64-12.27)

UnrelatedAds 10.57(8.81-12.32) 11.53(9.78-13.29)

Variable

 

'TT11.15 (9.66-12.65)

10.68 (9.29-12.07)
10.2 1 (8.90-1 1.52)

10.70 (894-1245)

Finally, the content learning questions each contained a lure choice that

corresponded to the content of one of the related ads. Therefore, if participants

were choosing the lures due to confusion between the target text content and the

113

content of distractor ads, lure selection rates should be elevated in the Related Ads
condition. However, as in Experiment 1, the rate of lure selection was extremely low
(less than 4% of the time), and there were no cases in which the lure selection in the
Related Ads condition was signiﬁcantly higher than the lure selection rate in the
other three conditions. As a result, data from this measure was considered
uninformative and was not considered further.
Discussion
Experiment 2 introduced a secondary gem-collection task, which
participants completed in conjunction with the Hybrid presentation web page
viewing task from Experiment 1. This secondary task was designed to interfere with
participants' abilities to selectively attend to only regions of the screen containing
task-relevant images or content, and to enable investigate whether participants may
utilize spatial cues in conjunction with processing items for semantic content.
Attention to Advertisements
It was predicted that the memory test results would replicate the results
from Experiment 1 in which recognition of graphics was best, followed by
recognition of related ads and then unrelated ads. This prediction was supported in
both the recognition accuracy and the recognition speed data. Central to the design
of Experiment 2, however, was the dual task manipulation. The Variable location
condition of the dual task was predicted to interfere with participants' abilities to
restrict processing to relevant regions of the screen. Consequently, participants in

this condition were predicted to show higher recognition rates than participants in

114

the no dual task baseline, indicative of greater image processing. On the other hand,
it was thought that the Fixed location condition of the dual task would result in the
participants having fewer attentional resources available to allocate to the
processing of ads, resulting in lower recognition rates in this condition, compared to
the baseline.

Unfortunater no inﬂuence of dual task condition was found on either the
recognition accuracy or the recognition speed data. It is, however, worth noting that
the recognition rates of related ads may have increased slightly from the None dual
task condition to the Fixed location condition to the Variable location condition,
while the recognition rates of graphics and of unrelated ads remained fairly stable
across dual task conditions. This ﬁnding, although not statistically signiﬁcant, may
indicate that the related ads were more likely to beneﬁt from participants' inability
to conﬁne their attention to speciﬁc content-relevant regions of the page.

The results of Experiment 1, interpreted as evidence of semantic content
guiding processing, showed that related ads were more likely to be recognized than
unrelated ads. If attention is guided to regions of the screen based on the semantic
content of items, then participants were predicted to use spatial cues to avoid
returning attention to semantically unrelated items. In other words, the disruption
to the use of spatial cues caused by the secondary task was expected to most beneﬁt
the recognition rate of unrelated ads, since these ads were expected to be marked as
distractors and subsequently ignored, while related ads were expected to be marked

for further processing.

115

Since the gems did not appear on each image presented in the experiment, it
is possible that the inﬂuence of the dual task would only emerge in the recognition
rates of images on which gems appeared. Accordingly, this analysis was conducted.
Although the recognition rates for images presented in a dual task condition on
which a gem actually appeared was numerically higher than for images on which no
gems appeared, this difference was not signiﬁcant. In conclusion, there was little
evidence in the image recognition data that the introduction of this dual task
resulted in any measurable change in image processing.

Goal Task Performance

The current study used the same measures as Experiment 1 to evaluate goal
task performance. It was predicted that distraction effects (previously unobserved
in similar paradigms) were likely to occur in both of the dual task conditions, in
which fewer attentional resources should be available for the primary information
search task. Distraction effects in these two conditions were predicted to be
attributable to the inﬂuence of that secondary task, rather than the inﬂuence of ads
as distractors.

In keeping with the ﬁndings of Experiment 1, no decrements in goal task
performance were observed across the different image type conditions. In fact, the
measure of website viewing time indicated that the pages presented in the Related
Ads condition were actually viewed for signiﬁcantly less time than the pages
presented in any of the other image conditions. This result replicates the

unexpected pattern of results obtained in Experiment 1, and as previously noted,

116

could reﬂect a materials effect as the same materials were used in both Experiment
1 and Experiment 2.

Contrary to predictions, no inﬂuence of the Secondary gem collection task
was found, as no performance decrements were observed across dual task
conditions. Unfortunater the failure to ﬁnd a secondary task effect in this
experiment may indicate that the goal task performance measures are not
sufﬁciently sensitive to detect changes in search performance, as there is a
substantial body of evidence pointing to dual task effects on measures of accuracy
and RT. This difﬁculty with the design of the current experiments will be discussed
in more detail in the general discussion section.

Conclusions

In summary, the recognition data in the current experiment supported the
ﬁndings in Experiment 1 of semantic ﬁlters, but none of the predicted dual task
effects were found. In addition, no distraction effects from the type of images
included on the web pages were observed, although pages were viewed signiﬁcantly
less in the Related Ads condition than in any of the other image conditions. As was
the case in Experiment 1, these results may be attributed to a materials effect.

Surprisingly, the introduction of a secondary task did not result in any
observable changes in goal task performance, or in the measure of attention to ads.
On the one hand, this may be due to using global measure of time spent on page, and
content accuracy, rather than more ﬁne-grained measures such as eyetracking. On

the other hand, if performance changes in the information search task performance

117

cannot be detected using these global measures, it is unlikely that Internet users
would be able to detect appreciable differences in their ability to carry out their goal
task when Internet ads were present onscreen.

As such, this experiment again provides support for the idea that Internet
users are not likely to be at a disadvantage when it comes to completing tasks online
in the presence of ads. Furthermore, the non-signiﬁcant inﬂuence of a secondary
task supports the idea that Internet users may be able to carry out a secondary task,
such as chatting online, with few repercussions to primary task performance.
However, the participants in this experiment were all college students, an age group
that arguably has the most experience dividing their attention across multiple
online tasks. A follow-up study testing the generalizability of these ﬁndings with a
group of older adults seems appropriate prior to making a strong case for
unimpaired primary task performance in conjunction with a secondary task. In
addition, as mentioned in the Experiment 1 discussion, the possibility also exists
that distraction effects may emerge in search performance when participants are
engaged in less constrained, self-directed searches outside of a laboratory

environment.

118

GENERAL DISCUSSION

Summary of Experimental Results

The aim of this work was to investigate mechanisms enabling Internet users
to control their attention on the Internet and avoid unwanted processing of
distractor stimuli, such as task irrelevant advertisements. Based on a review of
previous research on attentional control mechanisms, it was hypothesized that
Internet users should be particularly likely to use perceptual features, semantic
content, and spatial location cues to guide their attention on the Internet.

A preliminary study, and Experiment 1, explored the relative inﬂuences of
distinctive perceptual features and semantic content on attention to ads and on
search task performance. Experiment 2 was designed to explore the relative
inﬂuences of semantic content and location cues on attention control abilities
during an information searching task. As in previous studies (Helder, under
revision), ad recognition rates were used as a measurement of attention to ads in all
the reported experiments. Measures of goal task performance involved the time
taken to complete the information searching task, and the accuracy with which the
target information was obtained from the experimental web pages.

Role of Perceptual Features

It was predicted that participants would show evidence of using distinctive
perceptual features to guide attention in the preliminary study, and in Experiment 1.
In particular, it was predicted that distinctive perceptual features would be likely to

capture attention when they shared features with the current attentional set,

119

supporting a contingent capture hypothesis rather than an involuntary capture
hypothesis. This effect was predicted to take the form of increased recognition of
ads presented with graphics in the Hybrid presentation condition, as compared to
the recognition rates of ads presented with other ads in the Uniform presentation
condition. This prediction was based on the idea that when graphics appeared on
the web page, the attentional set should contain perceptual features unique to
images, but shared by both graphics and ads. On the other hand, when only ads
were presented onscreen, none of the images should be considered task relevant,
and thus the attentional set should not contain perceptual features unique to
images. However, the difference in recognition rates between the two presentation
conditions in both the preliminary study and in Experiment 1 did not reach
signiﬁcance. Consequently, no consistent evidence supporting the use of perceptual
features to capture attention

Although support for a contingent attentional capture hypothesis was not
obtained, support for an involuntary attentional capture hypothesis was also not
obtained. According to this hypothesis, distinctive and visually salient features
should capture attention regardless of their consistency with the current attentional
set. In general, images should be more visually salient than text, due to their use of
bright colors and large size. Furthermore, animated ads should be more visually
salient than static graphics, and there should be no difference in salience between
related and unrelated ads. Therefore, the contingent attentional capture hypothesis

would predict that recognition would be best (and equivalent) for the two types of

120

ads, followed by graphics. This pattern of results was not observed.

Although these experiments did not support the hypothesis that perceptual
features are involved in controlling attention in an Internet environment, this
possibility cannot be completely ruled out. The manipulation of perceptual
relatedness in these experiments was confounded with the manipulation of
semantic relatedness. In these experiments, it was assumed that the appearance of
semantically related graphics on a web page would alter the characteristics of
attentional set, resulting in an attentional set that contained perceptual features _
common to both graphics and ads. However, it is not certain that the graphics were
treated as task relevant by participants, and consequently, the attentional set may
not have ever contained perceptual features common to both ads and graphics. A
stronger test of the role of perceptual features would involve a direct manipulation
of the primary t ask to ensurethat perceptual features common to ads and a target
stimulus are contained in the attentional set.

Inﬂuence of Semantic Content

It was predicted that participants would show evidence of using semantic
content to guide attention to items. Overall, this prediction was supported. Although
the preliminary study did not ﬁnd a signiﬁcant effect of semantic content on image
recognition rates, both Experiment 1 and 2 did. In both of these experiments,
graphics were better recognized than related ads, which in turn were better
recognized than unrelated ads. These ﬁndings were interpreted as support for the

participants' use of semantic ﬁlters to control the processing of potentially task

121

irrelevant images.

Previous studies (Helder, under revision) have failed to produce any reliable
distraction effects from ads on the measures of goal task performance. However,
these studies only included ads that were unrelated to the web content used in the
information search task. Consequently, it was predicted that participants in the
experiments included in the current study may show distraction effects when ads
displayed were related to the target content.

None of the experiments reported herein found any evidence of ad-related
distraction effects. Unexpectedly, Experiment 1 and Experiment 2 produced results
in the opposite direction in the web page viewing time measure. In these
experiments, participants were signiﬁcantly faster to process Related Ads pages
than pages in any other image condition (which did not differ from each other). This
ﬁnding was not expected, and is difﬁcult to interpret. Faster processing of web
pages containing related ads could be indicative of a facilitation effect, in which the
presence of ads actually aids the participant in searching for information on the
page. This explanation seems unlikely, as there does not seem to be any reason to
assume that content-related ads would beneﬁt processing any more than content-
related graphics.

Another possibility is that images were processed less when they were
related ads than when they were related graphics. This explanation can account for
the ﬁnding that participants spent longer on the pages displayed in the Graphics

Only condition—however, it seems to break down when it comes to accounting for

122

the results obtained in the Unrelated Ads condition. If the reduced viewing time in
the Related Ads condition reﬂects a reduction in image processing, then the ﬁnding
that processing time for pages in the Unrelated Ads condition did not differ from
processing time of pages in the Graphics Only condition would imply that unrelated
ads were attended similarly to graphics. Yet the recognition tests in both
Experiment 1 and Experiment 2 show that recognition of unrelated ads is
signiﬁcantly lower than both graphics and related ads.

A ﬁnal explanation involves a materials effect, as proposed previously in the
individual experiment discussions. One of the difﬁculties in attempting to conduct
“practically relevant research” (Czaja & Sharit, 2003) is balancing the need to
replicate real world conditions with the need to exert experimental control. In the
current study, the unrelated ads were selected based on their dimensions and ability
to ﬁt in the same area of the page occupied by a graphic or a related ad in the other
conditions, thus controlling for image size and position across image conditions.
This decision was made at the expense of counterbalancing the ads by using ads that
were considered a related ad in one passage and an unrelated ad in another passage.
It was thought that the resulting greater variety of categories represented by the
unrelated ads was more similar to the variety of ads Internet users were likely to
encounter during a real world web search. Nevertheless, since the images were not
counterbalanced across conditions, there is no way to determine where the
inﬂuence of individual images ends and where the inﬂuence of the image's semantic

content begins. Therefore, the results of these experiments would beneﬁt from an

123

additional study in which the semantically related and unrelated ads were

counterbalanced. This experiment would also improve the strength of the

interpretation of the recognition test results, as image recognition rates are also

susceptible to materials effects when counterbalancing is not complete.
Inﬂuence of Spatial Location

Although Experiment 1 showed evidence that attention toward ads appears
to be inﬂuenced by the extent to which an ad's content resembles the target content,
no consistent distraction effects were found in this experiment. These ﬁndings,
along with previous research (Helder, under revision), have been interpreted as
indication that the majority of attentional resources are allocated toward processing
target content, rather than advertisements. Experiment 2 was designed to explore
whether participants may be utilizing both semantic content and spatial location to
guide attention and avoid distraction effects. It was hypothesized that in order to
successfully complete the task, participants would need to process all web page
images for content. However; to conserve attentional resources, it was predicted
that participants may employ a visual marking strategy to enable them to avoid
subsequent processing of images deemed to be task irrelevant.

To test this hypothesis, a secondary task was designed to interfere with
participants' abilities to inhibit subsequent processing of screen regions containing
irrelevant images. Speciﬁcally, participants were asked to "collect gems" that
appeared onscreen as quickly as they could. This secondary task was performed

concurrently with the primary information searching task. The key dual task

124

condition involved the presentation of gems in variable locations on the web page.
Since the location of the gems could not be predicted by participants in advance, it
was assumed that participants would have to remain vigilant to the entire area of
the screen, rather than avoiding returning to locations previously determined to
contain task irrelevant images.

It was predicted that interference from this dual task condition would result
in increased recognition of images, as well as an overall decrease in information
search task performance. A second dual task condition, in which the location of the
gem was ﬁxed (and consequently predictable) was included in order to compare
differences in search task performance due to the inclusion of a secondary task, and
those due to the inability of participants to use location cues to guide attention.

Unfortunately, no signiﬁcant dual task effects emerged. Not only did image
recognition rates not differ across between the baseline and the dual task
conditions, a separate analysis comparing the recognition rates of images on which
a gem had actually appeared to the recognition rates of images on which no gem
appeared failed to ﬁnd any signiﬁcant differences. Furthermore, goal task
performance remained constant across the dual task conditions, indicating that
there was no difference in page viewing time or in content learning in the dual task
conditions as compared to a baseline (no dual task) condition.

This null effect is disappointing on multiple accounts. Central to the
motivation for this study, it is disappointing to ﬁnd no evidence for location-based

ﬁlters. However; it is more disappointing to ﬁnd no general dual task effect, as this

125

calls into question both the effectiveness of the dual task manipulation, as well as
the sensitivity of the performance measures.

In order for the dual task to have inﬂuenced participants' abilities to set
location-based ﬁlters, participants needed to be engaged in the dual task. If
participants were not dividing their attention between the secondary gem collection
task and the primary search task in Experiment 2, it would not be surprising to ﬁnd
no results from the dual task manipulation. However; measures of gem collection
performance indicated that participants appeared to be attending to the secondary
task in addition to the primary search task. Participants on average collected more
than ﬁve gems per passage page, indicating that gem collection was ongoing while
participants were viewing a page. Even more indicative of attention to this
secondary task is the ﬁnding that the average time taken to collect a gem was less
than 35. As gems were animated to appear gradually over the course of 85, this
ﬁnding indicates that the majority of gems were collected before they had fully
materialized. Consequently, it does not seem likely that the failure to ﬁnd dual task
effects can be attributed to a lack of attention directed to the secondary task.

The explanation that remains, then, is that the measures of goal task
performance were not sufﬁciently sensitive to detect performance differences
between the baseline and dual task conditions. As in Experiment 1, performance in
Experiment 2 was measured primarily in terms of average time spent viewing each
web page, and average accuracy in answering content learning questions, although

the time taken to answer content learning questions correctly, and the proportion of

126

lures selected were also recorded and analyzed. In comparison to experiments that
have measured performance using more ﬁne grained methods, such as reaction time
in a simple search task, or tracking of eye movements, these measures are global,
and subject to more variability both within and across individual participants.

In order to address this particular methodological problem, it is
recommended that future research be conducted in which external validity is
sacriﬁced for increased sensitivity of goal task performance. One way of achieving
this would be to substantially reduce the amount of information presented on each
web page, and reﬁne the task to more closely resemble a visual search task than an
information search task. In particular, the methodology used by Burke et al. (2005)
in their headline visual search task could be adapted to investigate the inﬂuence of
semantic and location-based ﬁlters.

Conclusions

The current study ﬁnds evidence for the use of semantic ﬁlters, but not for
the use of perceptual or location-based ﬁlters in controlling attention in Internet
environments. Although no evidence for perceptual or location-based ﬁlters was
found, it is possible that these mechanisms may still play a role in guiding selective
attention on the Internet. Future experiments designed to address methodological
issues present in the current study are required before any deﬁnitive conclusions
can be made about the presence or absence of these mechanisms. Nevertheless,
consistent with previous research, this study offers good news to both advertisers

and Internet users.

127

1m plications for Advertisers

It was expected that participants would show reliable overall recognition of
advertisements presented during the experiment—even though participants were
not aware that their memory for advertisements would be tested. This overall effect
was observed in the current study, as well as in previous studies (Helder; under
revision). Accordingly, this represents good news for advertisers whose goal is to
obtain the attention of potential consumers on the Internet. Of note is the fact that
the current experiments ﬁnd reliable recognition of a fairly large number of
advertisements by participants, as compared to some studies in the advertising
literature that test recognition of just a few advertisements (Raman & Leckenby,
1998; Yoo 8: Kim, 2005). Experiment 1 presented participants with either 12 or 24
ads (Hybrid and Uniform conditions, respectively) and Experiment 2 presented
participants with 12 ads. In addition, graphics were presented in the experiment,
which likely resulted in competition for attention to advertisements—yet
recognition rates were still quite good.

Another implication of this study, relevant to Internet advertisers, is the
ﬁnding that the semantic content of an ad can inﬂuence the amount of attention
(and consequently, the rate of later ad recognition) directed to an ad. This ﬁnding
supports the recent trend for Internet advertisers to focus their marketing efforts
around key word searches (Li & Leckenby, 2007). Advertisements that are closely
matched to the interests and current goals of an Internet user are likely to pay off

with increased attention directed toward the advertisements.

128

The ﬁnal implication of this study for advertisers is that reliable recognition
of ads can be achieved without resorting to forced exposure techniques. All ads used
in these experiments were non-interactive banner ads. Target web content was
never obscured by an ad, and participants were never required to click on or
otherwise interact with an ad to complete their goal task. The implication of these
ﬁndings is that creating progressively more intrusive advertisements may not be
necessary in order to gain Internet users' attention.

Implications for Internet Users

A recent review of Internet advertising effectiveness notes that the majority
of Internet users who choose to block advertising base their choice on the belief that
the advertisements impede their ability to carry out a goal task on the Internet (Cho
& Cheon, 2004). This belief is echoed in the participants tested in these
experiments, the majority of whom believe that they are distracted by Internet
advertisements to some degree. Nevertheless, these experiments did not produce
any signiﬁcant differences in the measures of search task performance. Participants
in this study did not spend signiﬁcantly longer per web page when ads are present,
nor were they signiﬁcantly less able to gather the information they sought.

Although these performance measures may not be the most sensitive used in
the literature on attention and distraction, they are representative of the measures
that are likely to be most important to the average Internet users. Internet users are
not likely to be aware of whether the pattern of their eye ﬁxations varies between

web pages with and without ads, but they are likely to be aware of (and frustrated

129

by) an inability to direct their attention as needed to obtain the sought-after
information. The present research indicates that there are no real decrements likely
to be observed on goal task performance, when measured globally. Therefore,
although Internet users may continue to feel as though their attention is co-opted

and coerced by advertisements, the data indicate otherwise.

130

APPENDIX A: SCREEN SHOTS FROM THE PRELIMINARY EXPERIMENT

Baseline Conditions

 

.l l..-

 

_. '8 c w: “ . o... x-u ,
health Inlumiotlon arAllebuo m the media and walk. Every day in Uta newspaper at on ,w “m l a" u. soot-Au"
mm s-on we we stores about new medical Ilnd nos Part1 ups we hen: that a return ding

as e 3005 cu three-‘00 locrelu n stto hes. The t‘s a large increase" It sounds story But,

 

wwwmanmmw.

 

 

ifa nu know that in every 10 000 people not telling the drug. them are two $0,011.42“: men a “rum: mun-m u - rm. WWW
:Nee Inld Increase really ml, means six more strong. Mayne hethnt‘ ts not quite so ‘nah'enlm
m. 9m and Cm'mvonv
.. . w H6001 my "CUM
prevents I devestatlnq niecucn. a It doosn'L How are ~e to make sense of such stories? How ma.» nu: cum and um. sum say:
Wk "hr [mm sun/.mrwmumntmlwmcm

- rm, ertv. e see 5 - .._ cl l-al (.1 p. r "m“, rant-18mm mun ru- Over um:
" of outlets from the evening news to web sltes or. the Hntemet a...» wan-ye. m5: shoot 51: . I...

mm W m in: soup, sum, mm

.1- min...- aw-

 

 

 

 

 

 

 

age-ear; agree-r 'sliéE arde “ﬁrsteagea’igpiayed‘n the dreamy
condition.

 

 

.1. .l E

"elm-=5 Eu Her-arm

 

3; that ;r-:'

 

‘qr' M11 ta Car..- hm

hm mu 5 Day w saw Any

rum 9- “ MI nut“ (in
mm W L: a car- war-1; ay-
tt- Sn» Adm (awe-n

Ola-m Ml cumin LUN Twin. Sum Sm
smor wan Fb- mu no: Cuttakﬂ Lu“
Mn \nl rum mum KAL Cw! Once!
paw Wan-v. rm sou su .. an
roan noun he not so to. study w

.h- auu-m m,

 

 

 

 

 

eesboves no
musesa aloo; 1:1 than ‘0. a mouse -n stun-res That‘s a large lnueaso- It soundsscary .But I! you um that In every 10. 000 “cool that
[thing the drug, that If! two strokes. the“ a " ‘- t-rty .May be that s Incl do u so m'ngtl: enlnq.
. A . . 1 (does

 

 

 

Now are we to mete“!l sense of such stories? How do we km: what to be! eve?

 

 

1r. ._ . , . . . . x- u. .
» ‘ "" ' ‘ " "‘ Van'qvl n‘. tun-Luz“:

   

wit:
a...“

 

 

 

Figure A2: Screen shot of the ﬁrst page displayed in the graphics only
condition.

131

Ad Conditions in the Uniform Presentation Condition

   

at“. ..i “in... man. Siluriudmnntlcu nun no: I- much-Nun: ’mr

«r. Mm-.rﬂ,...m- , 2 ' .17.!

cl:' 4 n .1-

. _ ._ .. ,. t. _. ., .. .t. .. .. ....-._.,...........‘,...M. u............._........a.w.._.“—.4-

‘Sta I‘ May tn Cant-r hrthrtur

 

Tue (Jr-nu J 0J1 Iietp Stolut In."

 

. Wasting the Rm at a Duvet” Curr.
E 5 ' a : :5. -. *. Advances in science. and .‘rxhnclogy continue to increase the amount of

E hearth inionnaticn available to the media and public Even, day in. the lie-«spasm or on
:elevson we see szorres about new medical lrnd ngs. Pen-ans we hear that a certain drug
Muses a 300% nr three-‘0 (1 mouse in strokes. That's a Inigo. increase—t sounds scary. But.
i. ii vcu know that in every 10,000 people not Isl-mg the drug, there are two strokes. than a

E three-felt! inc-“.550 really mlv rr'urans sit more strokes Maybe that". not quite so t'rrghzminq.
'1 it’s also confusing that sometimes stories seem tc report opposite resultswa new vaccine . m w you“; 9mm," ”I a...“ my
; prevents a devastating rriettiorr, or it doesn't. How are we to make sense of such stories? How iv
:‘ do we know what to believe?

P1:- mt-sn'. Hawthorn b‘ a (INC! Vim-rig 5”,.

The Sm rut-39 Controversy

(Hour. Not carting ind Ynn, ‘31th Raw:

{MUJV‘ tkil tit-0.1 I)!!!" Om !_ :1." (N00 (DINO!

haunt. Warn-19!. You: Shark 5m 1 "in

Home updm ﬁsh Not So "oh, Stu-1» tint!

‘ This artirie seeks to help consu'ners evaluate health and scronfiiic 'nfmmaiion from a variety

' it». .1 «1‘- ‘AI. .... to-

 

 

 

 

; ‘. of outlets-4mm the evening news to web sites on the internet. 1_

i .

r

i ~33 ‘.‘. ‘ _‘_' - u 2' '-'.1'4 : > '2-_'.“ i ' 7 t.‘ '-'I'I.r- e". .'.' r'o ‘.".'."_‘.-'n-"‘ i:'.’-.f. v_.'“' r.Y.'.' ant-‘Jﬁ-hn. u- ‘-.< 9.2. v ' bur-3.x sun-1”Mn-r-!---r=-.‘J.~.'_v.s-.--- -._--Ibgr—ul¥-'r!
: '.'-'), ‘."‘. . r r“. .‘t'.:‘ a ' ﬁnk”. .3 I".'.‘.‘ N. - r's‘.'.‘,'. ,>._-".'- ~ -- -.”‘.‘. . Urn“:- "...'.."'-~Av.A;‘.‘----‘-'O.-s .9A;-"-’.': ‘:'.‘i'r"'-n ,,6,- ‘_ ,".'.>:'_ ' ;'_I; .' 'rp' :,€:1.I;q._l-r.'._‘:l-“L'-;II1I§I,
I

i

i

'1

Figure A3: Screen shot of the ﬁrst page displayed in the related ads
condition.
my.“ II—IkD'Ib-Ah lrirmndm-lh “I -’apl dammed-cc ,,- i";-i.=‘x2§'.9*~'\2':' .;.‘-"-‘»' . "2 . V :JZJE

   

 

1;.“ ”N Jtrg"“,r3.§ q

3’ ”Mi"! trir: rm‘afi -ri.’

 

".iLa-s' Mn. bu Canto: Iii-alum!

 

Yno Limits J Day (one strata Au"

 

"erupting the Rash ol 3 Cuba!” Cr!-
-' ‘ ’ =; n; ' - rr Advances in science and tedrnolcgy continue tc increase the amount of
heel: In information available to tire med 3 and public. Everv dav rn the newspaper or on

‘ teieusion we see smut-s nhmrt new medical ﬁndings. Perhaps we hear that a tartan drug
causes a 300% or L‘uee-loid increase in strokes That‘s a Earqe |ﬂLI865£"II sounds nary. But.

i
l
l
r ”warm! marmm is a Cunt Wynn; (up
i
i

. r.’ you know that In every 10,000 people not taking the drug, there are two strokes, men a i
I
3
l

"a Sikh Adina Corinna-11

Quinn Moi (.tIUKJ Lind rem, Study buy-9'

thrwiold increase really cniv means sit more strokes Mame that's not quite so nightenmg. “'3" "'9' M“ W" "‘ "1 5‘" C°'°"‘ “M"
it's also confusrng that sometimes stones seem to report opposite rowits-‘a new vaccrne

j prerents a devastatrng infedion. or it doesn't. How are we to make sense at such stories? How ‘
do we lurow wnat to behave?

No U1; me ﬁelwnm RN. Mn! (mat
Instant warn-m, Tanr, ’3er Col a urn:
For" Radon Rah rm. 90 HQ“. Shady Hurts

‘ This article seeks to help consumers evaiuate health and scientiﬁc Information from a variety

in! .v “a, - 00A {om

 

 

of outietsnfrom the evening news to web srtes on the internet. L
’ ".'."‘.‘ - f '3'. ‘-'v ~ '~,-,__'j':*.':';‘.'..-"'-'-" .__:’- "-."r .,-"r'.w::'=-.rm|--IJ"
._;,.. nu .-v-A.1_-.-.-... I-r «tarsus-raw n -.-.-- -Ir-+-u‘.‘r'-.-a 125-"1.5 .- \...-u ‘I-Il‘ﬂll’i’5‘x'.:'=')“'lv14:0 'lll . ’ - -‘ --" 'id-‘I , ~ . , . ~ - -- i- s I ﬁre-v.5 n-W’l‘l‘m

 

 

 

 

FigureA4: Screen shot of the first page displayed in theunrelated ads
condition.

132

Ad Conditions in the Hybrid Presentation Mode

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

u»..o.-i.~.urulh itauv-‘m-uMW-Oapl- "’7'“ i ”:ng ' " ' ' IL]!
Oven-zineimeu '6‘; His-alt“ ii i;‘:i‘:m=.tii;vi,’
"'. 0.". “VJ n I 515053.“; “‘-’-‘”"' ‘ " ‘ 1 'Stas' Mn tn Lemur Dual-Lt“
'3:
L , g _ _ , J . 1.. A . j Tm; Drinks 3 my (my snot; AMI;
_ _. .- __L, ,7 -~— 7.5”“ Weim rr-n Ink 0! a outwit.“ Cite
I“ i_ .i,_‘- a ___ -“' -—-.> ‘d-
, “-._“u g: ‘k '1, m 52:? pi‘ln\¢til insult-an h a Comm Wan—1g that
oil )"\~',~_U\ ~. 5. a a. ,.
V - ' - d— J—ki .- ~
g ‘. ‘ ‘ $2-1" . - j l '} 1m .ude Airtug Connoymv
gnu. -- l,‘ .a 74...; I. e. 5 _ .'
.33. n :17 l' 34-, .~ ’r'l Chidh‘n mv Corn!) lama um, 9mm,- say:
..'_'.... -r ..e.. ['5‘ f" '
':v;:'j __ (“'4 ,-v ~ K may H-ai- roe oumi Don't CU! cum Canon
4 I.“ (1",: ' "‘9." '3' ‘
. ;',.‘.Z“-\ ~- ' . _:_ l‘?’ ’l f {_' j_ No tnk ‘omd Between-tat (toast Lancet
-’_~ - - ‘ , - < J."-
:BZJ “1% fig, {772:1 ”j—w ' Dewitt.- Wotmrqg, Tout Sticck Sci! a lulu
”ﬂ- ' ﬂ ’ /‘{ _Z' A ‘ 1"}
" s ‘ ‘IQ . —- z
. \‘R‘ —‘ , C. w) Whom ma NNSOHQH Shiny unis
f _, _ \’ ‘17:: 4...”
'_ 5. ;' «‘2. S‘gﬁL— 9‘..- .u 1.1-. - on. new
--l-’ n--. a» ~.—~. ..O‘. n..-——~
"1;, '8. ' ﬁt .1 we Advances in science and technology continue to Increase the amount of health information avaliablo to the media and
public. Even; day in the newspaper or on telewsion we see storms atom new medrca' findings. Devhaps we hear that a certain drug muses a
300% or tinee--’oid increase m sue-(es. That‘s a huge incveese~it sounds Stacy. But, if you know that in even 10,000 people not taking the
duo, thete are two Sth‘xL‘ﬁ, then a {"IL‘O‘fCﬂd increase really only means six men: strokes. Maybe that's not quite so frightening. it's ago
ccn‘usnng that sometimes stories seem to repovt opposite results —a new vaccine prevents a devastating iniection. or it doesn't. How are we
to make sense of sud! stones? How do we knew what to believe? ‘
This attécie seeks to help consumers evaluate heal-1h and scientific Information from a variety 0! outlets--from the evening new to web sites
on the intemet.
' ' l d ' h l d d
I
Figure A5. Screen shot of the ﬁrst page disp aye in t e re ate a s
I 0
condition.
Ciu_i.._.i_«-4-..;v-4Ii.lawman-ilk“ I‘l-mm ”i ‘ - , mmﬁﬁfg~HRNIF ' ’ LL]:

  

' L')vei‘~vh2?3'i?t~;-d 8‘; Health arzi'omiatiesni."

Wé‘xw “-'I'll ‘ Ely-IT"; ‘

"'7 ‘ 'V

.————__. ‘11—.‘_- '(i‘ :7-v._.~_

 

  

0“ ' ' —’ ' —~.. _
‘ '9.“ Eu".-
v-I. Immune : a. .‘v .1" A- a- _
g-Utany-o‘re, --. hm.-.
‘Ap“'\_l ’1‘.-. ‘.__ 7‘...

.:.';?~_ .. .. -.. ‘.
L...L‘ II~\. '
».i.L’,-_ --.. '

 

to make sense of such stones‘ How do we know what to believe?

 

"“99. 39‘? .9139???” ..

‘Scan' may be Canter Weakly

Two E-v-Iu a Gay keep Strut. Away

WW the Hub 0! a Cuban (we
i‘evmtent ﬂeeiuwm l: a Cancel Win-n9 Sign
The Sade tutu; Convent“

Chddmn Hot Genny Lead Test}. cum Say!
Study: N9?! tun not! Dcn't Lut mason tar-(m
he LN: F'o-Jid Between Fat, Brent Came!
new." Wan-gs. To“: Shiv) 9m 0 him
Hun-v H.-.im Inn-A not 50 ugh, study Minn

:ﬁ-n*%1'<u—-o§,ﬂl.

 

 

 

’2‘; ‘.-:t a c: Alb/Bikes in science and [ethnology continue to Incense the amount of heaitn information available to the media and
public. Every day in the newspaper or on (Hermit-n we w». stones about new mod (a: findings Domain we beat that a certain ding causes a
100‘.» or thee-told anriease in strikes. That's a large. incense-4t. scunds scary. But, if you know that l" every 10,000 people not taking the
diuq, time are twc smokes. then a twee—low increase tealiy only means sir mute strokes? Maybe that‘s wt quite so incinerator; it's 3190
ccn‘using that sometimes stones seem to report opposite results--a new name prevents a devastating infection, or it doesn't. How are we

This article seeks to help consumew evaluate health and scientiﬁc lniounatlon irom a variety oi outlets-item the evening news to web sites

 

 

[5 igure A6: SEreen shot 6? "iii; ﬁrst page displayed ind-itiieiinrelated ads

condition.

APPENDIX B: SCREEN SHOTS FROM EXPERIMENTS 1 8; 2

Baseline Conditions

1“ 61%

Spring Break zoorﬁidh n Cruise ----- I 1

nun-m VII-MM!"

.. A V- .t_......_u_.1_x.;3.£i .

 

 

 

 

 

 

 

 

 

 

 

,ﬁ-mmﬂllJnn-v'wmﬂll

 

 

    

 

Figure F 1: 7 First page hi the travel passage displayed in the text only
condition.

III-d'MNINV launching-m NJ:- MlmJ’uuuvnv-w-ur
. run-'-

nmhwwM‘huuuumulmunwn-o‘vm
. > ' v1! ‘.~..-.r.»..1|.._‘ .. ‘ - -

 

 

Figure BZ: First page ofthe travel passage displayed in the graphics
only condition.

134

Ad Conditions in Uniform Presentation Mode

1

6 ":4: ix 3’“)
v w ‘9)
Spring Break 2007:39a‘gbbaa‘h Cruise

._ Mum

 

' -‘l‘hﬂllﬁdﬂmumuhrubwwmwlﬁ

Hassle

 

 

 

 

 

 

 

_. muons-m Munhnlaluwh“
n—n -

 

    
    

 

Figure BB: First page of the travel passage displayed in the related ad
condition.

 

.
.. n'vlvmhwm 3
uumrmmh

 

ﬁmum-mmw-um-wvw I’Imnmn‘ww mvnnmnuu ...‘ ..... ‘4............3

 

 

 

 

'. nmun-u-un Inmoeumm

 

mu rm I an" In». mm u mun '- M

 

Figure B4? First page bi’ili‘éﬁéiiel passage displayaa' inithe'nnrelated ' ”
ads condition.

135

Ad Conditions in Hybrid Presentation Mode

Spring Break 2007 -

11» mm m “mum

 

I. I... r-une-qwu m. ~a..«mm~.,m w,nm_uy,rmmummwuv
. ' run.
': nwunmsmulwnoumwum-um.

........................................... 3.. in mm

.Exaegia. 5;

«no-w an.
_.‘ «nun-1.2”“-
noun—:(‘umu

 

 

 

 

uh”-..- up»

 

 

 

 

 

 

mundane-nu. Ilma-‘EUI|I'OIW~~I

. n n'.

 

  
    

 

 

iii—£155: First page of the travel passage displayed in the related ads
condition.

‘3 a
Spring Break 2007:5'3 _

lhl' mm- In Ililrm

 

 

 

 

 

 

 

 

 

 

F1gil~reB6 First page of the travel passage displayed in the uanTéYéE'w
ads condition.

136

APPENDIX C: COMPUTER EXPERIMCE QUESTIONNAIRE

NOTE: Based on a questionnaire used by Kubeck, Miller-Albrecht, and Murphy (1999).
Items original to that questionnaire are indicated using an asterisk (’9. All other items
have been updated by the author. '

Please answer the following questions about your experience with computers.

* How well do you think you understand computers and their applications?

0 O O
Very well Well Somewhat Not very well Not at all
* Do you feel anxious about using computers?
0 O O O 0
Very anxious Anxious Somewhat Not very Not at all

* Are you interested in learning about or using computers?
0 O O O 0
Very interested Interested Somewhat Not very Not at all

* How difﬁcult or easy do you feel it would be for you to learn to use computers (or

how difﬁcult or easy was it for you to learn to use computers)?
0 O O O 0
Very hard Hard Somewhat Easy Very easy

* Have you used a computer before?
OYes 0N0

* If YES, what have you used computers for?

 

(Check all that apply)
UEntertainment DWord Processing
ULibrary . . .
I]
Research Busmess Appllcations
UEmail UPersonal Applications
Ulnternet DOtherzl l

 

* How long have you used computers for these activities?

 

Years

 

* Do you know how to use a mouse?
OYes OSo-so 0N0

137

* How often do you use computers?

0 O 0
Daily Weekly Several times Once a month Rarely, if at all
a month

* Do you own a computer?
OYes 0N0

* Have you taken any courses using computers?
OYes 0N0 If yes, please describe:

* How comfortable are you with using a computer?

0 O O O 0
Very Comfortable Somewhat Not very Not at all
Comfortable

* When you want to find information about a topic, what KINDS or TYPES of sources
do you typically use? Please rate the following in order of frequency/importance.

Where do you typically look FIRST, SECOND, etc.
Ask an Expert
QNewspaper
1“ Internet
Library >
1' Encyclopedia
Magazines Q
iBooks
Databases
'; Scientific Journals
3 Television
:Ask a lriendfrelative
':_ Other:
—

 

 

Please specify "other", if selected: I

 

When you have completed this form, please click the button below.

Record My Answers Q

138

APPENDIX D: INTERNET EXPERIENCE QUESTIONNAIRE

Please answer the following questions about your experience with the
Internet.

* If you had to brieﬂy define the following terms to someone who had never heard of
them, what would you say? In one or two sentences, describe what you know about:

1. the Internet:

2. being online:

3. e-mail:

4. the world wide web:

5. using online search engines:

How often do you use the Internet?

0 O O O 0
Daily Weekly Several times a Once a month Rarely, if at all
month

If you are a daily Internet user, approximately how many hours do you estimate you
spend online a day?

0 O O O 0
Not Applicable Less than Between Between More than
1 hour 1 and 2 hours 2 and 4 hours 4 hours

If you are an Internet user, approximately how many years have you been using the
Internet?

0 O O O 0
Not Applicable Less than More than More than More than
1 year 1 year, less than 5 years, less 10 years
5 years than 10 years

Are there speciﬁc Internet websites that you tend to visit on a regular basis?
OYes ONo

If so, what types of sites do you read on a regular basis?

139

(Check all that apply)

 

UHometown News [National News
USports DWeather
DSociaI networking (Ex: Myspace
D 1
Personal web pages / Blogs Facebook)
UOnline Games UEmail
UPortal page (Ex: yahoo.com,
msn.com) DOther: Q J

 

Have you ever purchased a product based on an Internet advertisement?
OYes 0N0

Do you think Internet advertising works?
OYes O No Why or why not?

Do you ﬁnd advertisements on the Internet to be distracting?
OYes ONo Why or why not?

Do you ﬁnd advertisements on the Internet to be annoying?
OYes O No Why or why not?

Do you pay attention to advertisements on the Internet?
OYes ONo Why or why not?

In your opinion, what makes (or could be done to make) Internet advertisements
1. eye catching?:

2. appealing?:
3. memorable?:
4. distinctive?:

When you have completed this form, please click the button below.

I
Record My Answers Q

140

APPENDIX E: EXIT QUESTIONNAIRE

Internet Searching 2008: Exit Questionnaire

1. Please rank the following topics that you read a passage on in order
from most interesting (1) to least interesting (4):

Understanding Health Information in the Media

Student Loans: What You Need to Know

The Psychology of Love and Attraction

Spring Break 2007: Caribbean Cruise

 

 

2. For each of the following four topics, please indicate on a scale of 1
(Extremely Interesting) to 7 (Extremely Boring) how interesting you
found the topic to be:

a. Understanding Health Information in the Media
b. Student Loans: What You Need to Know

c. The Psychology of Love and Attraction

d. Spring Break 2007: Caribbean Cruise

(The following rating scale was duplicated for each of the four passages.)

1 2 3 4 5 6 7
Extremely Very Interesting Neither Boring Very Extremely
Interesting Interesting Interesting Boring Boring

nor Boring

3. Did you pay attention to the images presented on the-websites? YES NO
4. Did you pay more attention to certain types of images than others? YES NO
3. If so, which types?

5. Did you pay attention to the advertisements presented on the websites?

YES NO
6. Did you pay more attention to certain types of advertisements than others?
a. YES NO

b. If so, which types?
7. Did you notice any relationship between the advertisements presented in

this experiment and the content of any of the passages? YES NO
3. If 50, describe the relationship you noticed.

141

REFERENCES

Adams, R. (2003). www.advertising: advertising and marketing on the World Wide
Web. New York: Watson-Guptill Publications.

annplugged. (2007). The share of search in total ad spending in 2006. Remixing the
routes. Retrieved August 13, 2007, from
http: / / linemarketing.wordpress.com / 2 00 7 / 0 S / 24 /the-share-of-search-in-
total-ad-spend-2006/

Babcock, R., & Salthouse, T. A. (1990). Effects of increased processing demands on
age differences in working memory. Psychology and Aging, 5, 421-428.

Bayles, ME. (2000). Just how “blind" are we to advertising banners on the web?
Usability News, 2. Retrieved July 17, 2008, from

http://wwwsurl.Qrg/usabilitymews/ I Q I /.

Bayles, ME. (2002). Designing online banner advertisements: Should we animate?

Usability News, 4. Retrieved from W.

Beck, D. M., 8; Lavie, N. (2 005). Look here but ignore what you see: Effects of
distractors at ﬁxation. journal of Experimental Psychology-Human Perception
and Performance, 31 (3), 592-607.

Bruner, I. & Kumar, A. (2000).Web commercials and advertising hierarchy of effects.
journal of Advertising Research, 40, 35-42.

Burke, M., Hornof, A., Nilsen, E. & Gorman, N. (2005). High-cost banner blindness:
Ads increase perceived workload, hinder visual search, and are forgotten.
ACM Transactions on Computer-Human Interaction, 12, 423-445.

Calisir, F, 8: Karaali, D. (2008) The'impacts of banner location, banner content
and navigation style on banner recognition. Computers in Human Behavior, 24, 535-
543

Carlson, M. C., Hasher, L., Connelly, S. L., & Zacks, R. T. (1995). Aging, distraction, and
the beneﬁts of predictable location. Psychology and Aging, 10(3), 427-436.

Chung, Y. (2006). Processing advertisements on the web: the effects of animation
and arousing content on orienting, arousal, encoding, and storage.
Dissertation Abstracts International Section A: Humanities and Social Sciences,
68, 2763.

142

Colcombe, A. M., Kramer, A. F., Irwin, D. E., Peterson, M. S., Colcombe, 5., & Hahn, S.
(2003). Age-related effects of attentional and oculomotor capture by onsets
and color singletons as a function of experience. Acta Psychologia, 113(2),
205-225.

Connelly, S. L., Hasher, L., 8: Zacks, R. T. (1991). Age and reading: The impact of
distraction. Psychology and Aging, 6(4), 533-541.

Craik, EL, 8: McDowd, ].M. (1987). Age differences in recall and recognition. journal
of Experimental Psychology: Learning, Memory, and Cognition, 13, 474-479.

Czaja, 8.1., 8: Lee, C. C. (2001). The Internet and older adults: Design challenges and
opportunities. In N. Charness, D. C. Parks 8: B. A. Sabel (Eds), Communication,
technology and aging: Opportunities and challenges for the future. New York,
NY, US: Springer Publishing Co.

Czaja, 8.]. 8: Sharit, I. (2003). Practically relevant research: Capturing real world
tasks, environments, and outcomes. The Gerontologist, 43, 9-18 .

Dahlen, M. (2001). Banner advertisements through a new lens. journal of
Advertising Research, 41 (4), 23-30.

Danaher, P. 1., & Mullarkey, G. W. (2003). Factors Affecting Online Advertising Recall:
A Study of Students. journal of Advertising Research, 43(3), 252-267.

Darowski, E., Helder, E., Zacks, R. T., Hasher, L., 8: Hambrick, D. Z. (2008). Measuring
distraction control with the reading with distraction task Accepted.

Day, R. F., Shyi, G. C. W., 8: Wang, 1. C. (2006). The effect of ﬂash banners on
multiattribute decision making: Distractor or source of arousal? Psychology 8:
Marketing, 23(5), 369-382.

Diao, F. E, 8: Sundar, S. S. (2004). Orienting response and memory for web
advertisements: Exploring effects of pop-up window and animation.

Communication Research, 31(5), 537-567.

Diaper, D. 8: Waelend, P. (2 000). World wide web working whilst ignoring graphics:
Good news for web page designers. Interacting with Computers, 13, 163-181.

Dreze, X., 8: Hussherr, E-X. (2003). Internet advertising: Is anybody watching?
journal of Interactive Marketing, 1 7(4), 8-23.

Duchek, ]. M., Balota, D. A., 8: Thessing, V. C. (1998). Inhibition of visual and

143

conceptual information during reading in healthy aging and Alzheimer's
disease. Aging Neuropsychology and Cognition, 5(3), 169-181.

Dywan, 1., 8: Murphy, W. E. (1996). Aging and inhibitory control in text
comprehension. Psychology and Aging, 1 1(2), 199-206.

Earles, 1. L., Connor, L. T., Frieske, D., Park, D. C., Smith, A. D., 8: Zwahr, M. (1997). Age
differences in inhibition: Possible causes and consequences. Aging
Neuropsychology and Cognition, 4(1), 45-57.

Edwards, S. M., Li, H. R., 8: Lee, 1. H. (2002). Forced exposure and psychological
reactance: Antecedents and consequences of the perceived intrusiveness of
pop-up ads. journal of Advertising, 31(3), 83-95.

Fox, S. (2 004). Older Americans and the Internet. Washington DC: Pew Internet and
American Life Project.

Gronau, N., Sequerra, E., Cohen, A., 8: Ben-Shakhar, G. (2006). The effect of novel
distractors on performance in focused attention tasks: A cognitive-

psychophysiological approach. Psychonomic Bulletin 8: Review, 13(4),
570-575.

Hasher, L., Lustig, C., & Zacks, R. (2007). Inhibitory mechanisms and the control of
attention. InA. R. A. Conway, C. Jarrold, M. 1. Kane, A. Miyake 8:1. N. Towse
(Eds.,) Variation in working memory. (Vol. xix, pp. 227- -.249) New York, NY,
US: Oxford University Press.

Hasher, L., 8: Zacks, R. 'I‘. (1988). Working memory, comprehension, and aging: A
review and a new view. In G. H. Bower (Ed.), The psychology of learning and
motivation: Advances in research and theory (Vol. 22). San Diego, CA, US:
Academic Press.

Helder, E. (under revision). Inﬂuence of Internet Banner Advertisements on Reading
in Older and Younger Adults.

Hollingworth, A. 8: Henderson, 1.M. (2002). Accurate visual memory for previously
attended objects in natural scenes. journal of Experimental Psychology:
Human Perception and Performance, 28(1):113-136.

Holt, B. 1., & Morrell, R. W. (2002). Guidelines for Web site design for older adults:
The ultimate inﬂuence of cognitive factors. In R. W. Morrell (Ed.), Older
adults, health information, and the World Wide Web. Mahwah, N], US:
Lawrence Erlbaum Associates.

144

 

Hong, W., Thong, 1. Y. L., 8: Tam, K. Y. (2004). Does animation attract online users'
attention? The effects of ﬂash on information search performance and
perceptions. Information Systems Research, 15(1), 60-86.

Internet advertising revenues grow 35% in '06. (2007). Retrieved August 13, 2007,
from http: / /www.iab.net/ news / pr_2007_0 5_23.asp

Johnson, C. R., 8: Neath, l. (1999). The World Wide Web: Exploring a new advertising
environment. CyberPsychology and Behavior, 2(3), 195-212.

Jones, M.Y., Pentecost, R., 8: Requena, G. (2005). Memory for advertising and
information content: Comparing the printed page to the computer screen.
Psychology and Marketing, 22, 623-648.

Koivisto, M. 8: Revonsuo, A. (2007). How meaning shapes seeing. Psychological
Science. Vol 18(10), 845-849.

Kramer, A. F., Hahn, S., Irwin, DE, 8: Theeuwes, 1. (2000). Age differences in the
control of looking behavior: Do you know where your eyes have been?
Psychological Science, 11, 210-217.

Kubeck, 1. E., Miller-Albrecht, S. A., 8: Murphy, M. D. (1999). Finding information on
the World Wide Web: Exploring older adults' exploration. Educational
Gerontology, 25(2), 167-183.

Lewenstein, M., Edwards, 6., Tatar, D., 8: DeVigal, A. (2 004). Retrieved lanuary 28,
2004, from http://www.poynterextra.org/et/i.htm

Li, H., 8: Bukovac, 1. L. (1999). Cognitive impact of banner ad characteristics: An
experimental study. journalism and Mass Communication Quarterly, 76(2),
341-353.

Li, H. & Leckenby, JD. (2 00 7). Examining the effectiveness of Internet advertising
formats. In Schumann, D.W. 8: Thorson, E. Internet Advertising: Theory and
Research. Lawrence Earlbaum Associates, Mahwah, M. 203-224.

Li, C. 8: VanBoskirk, S. (2005, May 2). US online marketing forecast: 2005-2010—
dollars will follow consumers online, driving spending to $26 billion by 2010.
Forrester Research. Retrieved from

no A 11" min ‘°-. 1|- u'i .‘oi ,o4.-H

m1.

145

Miarmi, L., & DeBono, K. G. (2007). The impact of distractions on heuristic
processing: Internet advertisements and stereotype use. journal of Applied
Social Psychology, 37(3), 539-548.

Morrell, R., Mayhorn, C., 8: Bennett, J. (2002). Older adults online in the Internet
century. In R. Morrell (ed.) OlderAdults, Health Information & the World Wide
Web, Mahwah, M1: Lawrence Erlbaum Associates

Moore, RS, Stammerjohan, C.A., & Coulter, RA. (2005). Banner advertiser—web site
context congruity and color effects on attention and attitudes. Journal of
Advertising, 34, 71-84.

Most, S.B., Scholl, 8.1., Clifford, E.R., 8: Simons, DJ. (2005). What you see is what you
set: Sustained inattentional blindness and the capture of awareness.
Psychological Review, 112, 217-242.

Most, S. B., Simons, D. J., Scholl, 8.1., Jimenez, R., Clifford, E., & Chabris, C. F. (2001).
How not to be seen: The contribution of similarity and selective ignoring to
sustained inattentional blindness. Psychological Science, 12(1), 9-17.

Neath, I., & Surprenant, A.M. (2003). Human memory: An Introduction to research,
data, and theory (2nd edition). Belmont, CA: Wadsworth/Thompson Learning

Pew Internet and American Life Project (2007). Tracking Survey. Retrieved August
25, 2007, from
http: //www.pewlnternet.org/trends/User_Demo_6.15.07.htm

Raman, N. V., & Leckenby, J. D. (1998). Factors affecting consumers' "Webad" visits.
European journal of Marketing, 32(7 / 8), 737.

Sagarin, B. 1., Britt, M. A., Heider, J. D., Wood, 5. E., 8: Lynch, 1. E. (2003). Bartering our
attention: The distraction and persuasion effects of on-line advertisements.
Informational Technology, 8 2.

Salthouse, TA. (1996). The processing-speed theory of adult age differences in
cognition. Psychological Review, 103, 403-428.

Salthouse, T. A., Atkinson, T. M., 8: Berish, D. E. (2003). Executive functioning as a
potential mediator of age-related cognitive decline in normal adults. journal
of Experimental Psychology: General, 132(4), 566-594.

Simons, D. 1., & Levin, D. T. (1998). Failure to detect changes to people during a real-
world interaction. Psychonomic Bulletin & Review, 5(4), 644-649.

146

Spieler, D. H., Balota, D. A., 8: Faust, ME. (2000). Levels of selective attention
revealed through analyses of response time distributions. journal of
Experimental Psychology: Human Perception and Performance, 26, 506-526.

Sundar, 8.5. 8: Kalyanaraman, S. (2004). Arousal, memory, and impression-formation
effects of animation speed in web advertising. journal of Advertising, 55, 7-17.

T horson, E., Duffy, M., 8: Schumann, D.W. (2007). The Internet waits for no one. In
Schumann, D.W. & Thorson, E. Internet Advertising: Theory and Research.
Lawrence Earlbaum Associates, Mahwah, N.J., 3—13.

United States Department of Commerce (2000). Falling through the net: Toward
digital inclusion., from http://www.ntia.doc.gov/ntiahome/digitaldivide

Wang, J.-C., 8: Day, R.-F. (2 007). The effects of attention inertia on advertisements on
the WWW. Computers in Human Behavior, 23(3), 1390-1407.

Watson, D. G., & Humphreys, G. W. (1997). Visual marking: Prioritizing selection for
new objects by top-down attentional inhibition of old objects. Psychological
Review, 104(1), 90-122.

Watson, D. G., & Maylor, E. A. (2002). Aging and visual marking: Selective deﬁcits for
moving stimuli. Psychology and Aging, 1 7, 321-339.

Wolf, 1.M. 8: Horowitz, T5. (2 004). What attributes guide the deployment of visual
attention and how do they do it? Nature Reviews, 5, 1-7.

Yoo, C. Y., 8: Kim, K. (2005). Processing of animation in online banner advertising:
The Roles of Cognitive and Emotional Responses. journal of Interactive
Marketing, 19(4), 18-34.

Zachary, R. A. (1986). Shipley Institute of Living Scale. Los Angeles: Western
Psychological Service.

Zacks, R. T., 8: Hasher, L. (1994). Directed ignoring: Inhibitory regulation of working
memory. In D. Dagenbach 8: T. H. Carr (Eds), Inhibitory processes in attention,
memory, and language. San Diego, CA, US: Academic Press.

Zhang, P. (2 000). 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 (March), 1-2 8.

147

       

     

MICHIGAN STA IE

llill/lll111/l1111111111111ll

3 129 02956 7124

 

.......