Tuesday, June 27, 2006
The Influence of Web Browsing Experience on Web-Viewing Behavior
Yoshiko Habuchi, Haruhiko Takeuchi, Muneo Kitajima
National Institute of Advanced Industrial Science and Technology (AIST)
1 Introduction
The World Wide Web has become an important source of
information, as much as traditional media like books, newspapers,
and television. While there have been many studies on Web
searching, research into Web-viewing behavior using eye-tracking
systems has only recently begun [Pan et al., 2004]. Josephson and
Holmes [2002] studied Web-viewing behavior focusing on the
category of Web page visual design. They suggested that eye
movements were affected by the following two factors: (1) visual
design of Web pages and (2) habitually preferred path across the
visual stimuli. However, these previous studies did not
sufficiently consider the user’s experience. The purpose of this
study is to investigate how past Web-browsing experience
influences Web-viewing behavior. We used a detailed
questionnaire to measure a user’s Web-browsing experience and
analyzed the eye-tracking data based on the user’s prior Web
experience.
2 Method
Four Internet users were recruited based on their Internet usage.
While they all used the Internet every day in their work, they
differed in terms of Web-browsing style. They were divided into
two groups, A and B, based on their Internet usage styles. Group A
consisted of comparatively heavy Internet users who browse and
use various portal sites. Group B consisted of comparatively
conservative users who use a regular portal site for fixed purposes.
The experiment consisted of two phases, observation and
recognition. In the observation phase, participants focused on
evaluation tasks. They were instructed to look at the Web pages
carefully and evaluate the usability of those pages. Participants
viewed 15 Web pages in three categories: Portal, News, and
Advertisement. After the observation phase, participants filled out
an Internet-usage questionnaire. In the recognition phase,
participants were asked to observe 30 Web pages including the
previous 15 pages and to determine if they had seen these pages in
the previous phase. Eye movement data were collected using an
eye-tracking system (Tobii Technology) during the experiment.
Each Web page was presented for 20 seconds during the
observation phase, and for 5 seconds in the recognition phase.
3 Results
The contents area of each web page was classified into six
functions based on Nielsen & Tahir [2002]. Table 1 presents the
mean percentage of gaze frequency in each condition. A framed
rectangle indicates the highest score within the specific condition.
The results indicated distinctive differences in Web-viewing
behavior between user groups. At the Portal page, Group A
exhibited consistent viewing behavior between the observation
and recognition phases. In contrast, Group B exhibited
inconsistent viewing behavior. At the News page, both groups
demonstrated consistent viewing behavior between observation
and recognition phases. At the Advertisement page, both groups
exhibited an instable viewing-behavior pattern because they
seldom browse advertisement pages. These results suggest that
prior Web-browsing experiences form an individual’s efficient
tracking method or mental model of how to view a Web site to get
information.
References
JOSEPHSON, S., AND HOLMES, M. 2002. Attention to
repeated images on the World-Wide Web: Another look at
scanpath theory. Behavior Research Methods, Instruments, &
Computers, 34, 539-548.
NIELSEN, J., AND TAHIR, M. 2002. Homepage usability 50
Websites deconstructed. Indianapolis: New Riders Publishing.
PAN, B. et al. 2004. The determinants of Web page viewing
behavior: An eye-tracking study. In Eye Tracking Research &
Applications (ETRA) Symposium, ACM, 147-154.
Copyright (C) 2006 by the Association for Computing Machinery, Inc.
National Institute of Advanced Industrial Science and Technology (AIST)
1 Introduction
The World Wide Web has become an important source of
information, as much as traditional media like books, newspapers,
and television. While there have been many studies on Web
searching, research into Web-viewing behavior using eye-tracking
systems has only recently begun [Pan et al., 2004]. Josephson and
Holmes [2002] studied Web-viewing behavior focusing on the
category of Web page visual design. They suggested that eye
movements were affected by the following two factors: (1) visual
design of Web pages and (2) habitually preferred path across the
visual stimuli. However, these previous studies did not
sufficiently consider the user’s experience. The purpose of this
study is to investigate how past Web-browsing experience
influences Web-viewing behavior. We used a detailed
questionnaire to measure a user’s Web-browsing experience and
analyzed the eye-tracking data based on the user’s prior Web
experience.
2 Method
Four Internet users were recruited based on their Internet usage.
While they all used the Internet every day in their work, they
differed in terms of Web-browsing style. They were divided into
two groups, A and B, based on their Internet usage styles. Group A
consisted of comparatively heavy Internet users who browse and
use various portal sites. Group B consisted of comparatively
conservative users who use a regular portal site for fixed purposes.
The experiment consisted of two phases, observation and
recognition. In the observation phase, participants focused on
evaluation tasks. They were instructed to look at the Web pages
carefully and evaluate the usability of those pages. Participants
viewed 15 Web pages in three categories: Portal, News, and
Advertisement. After the observation phase, participants filled out
an Internet-usage questionnaire. In the recognition phase,
participants were asked to observe 30 Web pages including the
previous 15 pages and to determine if they had seen these pages in
the previous phase. Eye movement data were collected using an
eye-tracking system (Tobii Technology) during the experiment.
Each Web page was presented for 20 seconds during the
observation phase, and for 5 seconds in the recognition phase.
3 ResultsThe contents area of each web page was classified into six
functions based on Nielsen & Tahir [2002]. Table 1 presents the
mean percentage of gaze frequency in each condition. A framed
rectangle indicates the highest score within the specific condition.
The results indicated distinctive differences in Web-viewing
behavior between user groups. At the Portal page, Group A
exhibited consistent viewing behavior between the observation
and recognition phases. In contrast, Group B exhibited
inconsistent viewing behavior. At the News page, both groups
demonstrated consistent viewing behavior between observation
and recognition phases. At the Advertisement page, both groups
exhibited an instable viewing-behavior pattern because they
seldom browse advertisement pages. These results suggest that
prior Web-browsing experiences form an individual’s efficient
tracking method or mental model of how to view a Web site to get
information.
References
JOSEPHSON, S., AND HOLMES, M. 2002. Attention to
repeated images on the World-Wide Web: Another look at
scanpath theory. Behavior Research Methods, Instruments, &
Computers, 34, 539-548.
NIELSEN, J., AND TAHIR, M. 2002. Homepage usability 50
Websites deconstructed. Indianapolis: New Riders Publishing.
PAN, B. et al. 2004. The determinants of Web page viewing
behavior: An eye-tracking study. In Eye Tracking Research &
Applications (ETRA) Symposium, ACM, 147-154.
Copyright (C) 2006 by the Association for Computing Machinery, Inc.
Location location location: Viewing patterns on WWW pages
Laura Granka
Google, Inc.
granka@google.com
Helene Hembrooke
Cornell University
hah4@cornell.edu
Geri Gay
Cornell University
gkg1@cornell.edu
1 Abstract
This study investigates which components of a web page are most ikely to both attract and maintain a viewer's attention. We measure these two aspects of viewing behavior—attention onset and maintenance—through an analysis of eye movements on three Web page components—page location, element size, and nformation density. More specifically, the present research addresses how the overall composition and structure of a Web page influences an individual’s ability to perceive content.
2 Introduction
Ocular data were analyzed with respect to location, element size, and information density (visual salience). For location analysis, each web page was divided into a grid of nine equal regions
measured by pixel coordinates and eye movements were measured relative to these regions. To analyze element size, ERICA’s Gazetracker software was used to partition areas of unique content (e.g. titles, pictures, links, content, and advertisement) using the LookZone feature. The area of these regions were then calculated based on pixel coordinates. Finally, we derived a metric of information density to measure of the relative salience of elements on the page. This was calculated by building on Shannon’s [1948] information theory algorithm for determining entropy, and for the case of Web stimuli, our formula specifically focused on the hue and image contrast between neighboring pixel values.
3 Methods
34 participants viewed three commercial homepages: Amazon, Cnet, and Ebay. These pages were selected due to their relative popularity [Gay et al. 2001] and because their homepage serves as the primary space to provide consumers with company-related promotions. While participants were given 15 seconds to browse the site at their leisure.
4 Results
Overall, we found that location most significantly impacts attention onset, while element size affects attention maintenance. Also, analyzing viewing patterns relative to the information
density (visual salience) of page elements produces behaviors representative of "banner blindness." These findings have direct implications for the placement of online content and advertising and can be used by designers to maximize the structure and organization of home pages. Analysis of location data indicated that certain regions of a web page are indeed significantly more likely to capture users' initial attention and be viewed first. Specifically the top left, mid-left and center were the top three regions where users first fixated.
Secondly, we found that the size of page elements did not significantly influence or help to attract a viewer's initial attention. We had expected that larger elements would more readily attract attention, but this was not the case. While not affecting attention onset, size was significantly related to attention maintenance, with larger elements attracting both more fixations and subsequently longer total viewing times. However, this result should be viewed with caution as element size may directly correlate with information density.
Finally information density, similar to element size, did not impact attention onset, but did significantly affect attention maintenance. Hierarchical multiple regression indicated that
location and size accounted for most of the variability in the model. The type of content or advertisement (text, image, graphic) did not contribute significantly to the outcome.
5 Implications and Conclusions
From an applied perspective, the research presented here has direct implications for advertisers. One very interesting result from this study is that size and information density did not contribute significantly to attention onset. One reason for this is that many of the large and "information dense" regions were isolated from the primary site content as advertisements. In spite of attempts to make these ads stand out by using bold contrasts and backgrounds, users are likely to “visually disconnect” from these regions because the elements appear extraneous to both the goals of the site and the user's own needs/motivations for accessing the site. While this explanation requires systematic investigation, we surmise that it may contribute to findings
reported by others concerning “banner blindness" [Benway 1998; Pagendarm and Schaumburg 2001]. Thus, our findings offer a position that the newer text-based ads, selected specific to page content, may indeed be more effective at maintaining a viewer's attention.
6 References
ENWAY B , J.P. 1998. Banner blindness: The irony of attention grabbing on the World Wide Web, Proceedings of the Human Factors and Ergonomics Society 42nd .
AGENDARM CHAUMBURG P , M. and S , H. 2001. Why Are Users Banner-Blind? The Impact of Navigation Style on the Perception of Web Banners, Journal of Digital Information,
http://jodi.ecs.soton.ac.uk.
ORMAN N , D.A. 1999. Commentary: Banner Blindness, Human Cognition and Web Design, Internetworking.
SHANNON, S.E. 1948. A mathematical theory of communication. Bell System Tech. J. 27, 379-423, 623-656.
GAY, G., STAFANONE, M., GRACE-MARTIN, M., and HEMBROOKE, H. 2001. The effects of wireless computing in collaborative learning environments. International Journal of Human
Computer Interaction, 13(2), 257-276.
Copyright (C) 2006 by the Association for Computing Machinery, Inc.
Google, Inc.
granka@google.com
Helene Hembrooke
Cornell University
hah4@cornell.edu
Geri Gay
Cornell University
gkg1@cornell.edu
1 Abstract
This study investigates which components of a web page are most ikely to both attract and maintain a viewer's attention. We measure these two aspects of viewing behavior—attention onset and maintenance—through an analysis of eye movements on three Web page components—page location, element size, and nformation density. More specifically, the present research addresses how the overall composition and structure of a Web page influences an individual’s ability to perceive content.
2 Introduction
Ocular data were analyzed with respect to location, element size, and information density (visual salience). For location analysis, each web page was divided into a grid of nine equal regions
measured by pixel coordinates and eye movements were measured relative to these regions. To analyze element size, ERICA’s Gazetracker software was used to partition areas of unique content (e.g. titles, pictures, links, content, and advertisement) using the LookZone feature. The area of these regions were then calculated based on pixel coordinates. Finally, we derived a metric of information density to measure of the relative salience of elements on the page. This was calculated by building on Shannon’s [1948] information theory algorithm for determining entropy, and for the case of Web stimuli, our formula specifically focused on the hue and image contrast between neighboring pixel values.
3 Methods
34 participants viewed three commercial homepages: Amazon, Cnet, and Ebay. These pages were selected due to their relative popularity [Gay et al. 2001] and because their homepage serves as the primary space to provide consumers with company-related promotions. While participants were given 15 seconds to browse the site at their leisure.
4 Results
Overall, we found that location most significantly impacts attention onset, while element size affects attention maintenance. Also, analyzing viewing patterns relative to the information
density (visual salience) of page elements produces behaviors representative of "banner blindness." These findings have direct implications for the placement of online content and advertising and can be used by designers to maximize the structure and organization of home pages. Analysis of location data indicated that certain regions of a web page are indeed significantly more likely to capture users' initial attention and be viewed first. Specifically the top left, mid-left and center were the top three regions where users first fixated.
Secondly, we found that the size of page elements did not significantly influence or help to attract a viewer's initial attention. We had expected that larger elements would more readily attract attention, but this was not the case. While not affecting attention onset, size was significantly related to attention maintenance, with larger elements attracting both more fixations and subsequently longer total viewing times. However, this result should be viewed with caution as element size may directly correlate with information density.
Finally information density, similar to element size, did not impact attention onset, but did significantly affect attention maintenance. Hierarchical multiple regression indicated that
location and size accounted for most of the variability in the model. The type of content or advertisement (text, image, graphic) did not contribute significantly to the outcome.
5 Implications and Conclusions
From an applied perspective, the research presented here has direct implications for advertisers. One very interesting result from this study is that size and information density did not contribute significantly to attention onset. One reason for this is that many of the large and "information dense" regions were isolated from the primary site content as advertisements. In spite of attempts to make these ads stand out by using bold contrasts and backgrounds, users are likely to “visually disconnect” from these regions because the elements appear extraneous to both the goals of the site and the user's own needs/motivations for accessing the site. While this explanation requires systematic investigation, we surmise that it may contribute to findings
reported by others concerning “banner blindness" [Benway 1998; Pagendarm and Schaumburg 2001]. Thus, our findings offer a position that the newer text-based ads, selected specific to page content, may indeed be more effective at maintaining a viewer's attention.
6 References
ENWAY B , J.P. 1998. Banner blindness: The irony of attention grabbing on the World Wide Web, Proceedings of the Human Factors and Ergonomics Society 42nd .
AGENDARM CHAUMBURG P , M. and S , H. 2001. Why Are Users Banner-Blind? The Impact of Navigation Style on the Perception of Web Banners, Journal of Digital Information,
http://jodi.ecs.soton.ac.uk.
ORMAN N , D.A. 1999. Commentary: Banner Blindness, Human Cognition and Web Design, Internetworking.
SHANNON, S.E. 1948. A mathematical theory of communication. Bell System Tech. J. 27, 379-423, 623-656.
GAY, G., STAFANONE, M., GRACE-MARTIN, M., and HEMBROOKE, H. 2001. The effects of wireless computing in collaborative learning environments. International Journal of Human
Computer Interaction, 13(2), 257-276.
Copyright (C) 2006 by the Association for Computing Machinery, Inc.
