Tobii Whitepaper Studying Web Pages Using Eye Tracking

Whitepaper

Studying web pages using Eye Tracking

Written by Kirk Ewing, Tobii Technology August 2005

Abstract Eye tracking has been investigated and “toyed with” for many years by researchers and commercial usability professionals. There are examples of high quality results being obtained in both of these areas using eye tracking techniques. Too often though these studies and commercial ventures have been hampered by inadequate hardware and software tools. These inadequate tools have caused the scope and goals of studies to be limited in their ability to provide interesting and powerful results without undue resources being required. Many researchers have become discouraged by the previous generation of eye tracking solutions and branded the technique “interesting but practically useless”.

New hardware and software tools have been developed by companies such as Tobii Technology to overcome all of the outstanding objections to the collection of high quality eye tracking data. Modern, customizable and relatively automated analysis tools have also been developed. These tools do not interfere with traditional techniques and do not require many extra resources.

Many new techniques and therefore interesting and powerful results are now available that were previously inaccessible. These results, available with limited additional resources, make eye tracking commercially viable prospect in usability testing, and web site testing in general, for the first time. There are quite a number of commercial and academic labs running highly successful programs today testing hundreds, and in some cases thousands, of subjects per year and effectively analyzing this data to produce highly useful results.

Table of contents

ABSTRACT 2

INTRODUCTION 4

PREVIOUS WORK AND PREVIOUS PROBLEMS 4

NEW DEVELOPMENTS SOLVE ISSUES 5 Adequate tools to record data 5 New analysis tools 6

ADDITIONAL INSIGHTS TO TRADITIONAL TECHNIQUES 7 Complementing verbal protocols 7 Stunning Deliverables 8 Answering typical questions 8

What happened in the breakdown event? 8 What do users see? 9 Where should valuable content be placed? 9 Do available visual cues drive users effectively? 10 How effectively do users complete the task? 10

LIMITATIONS OF EYE TRACKING 11 Data easily influenced 11 They looked at it but did they really “see” it 11 How people use peripheral vision 11

CONCLUSION 12

REFERENCES 13

Introduction Eye tracking as a tool to discover problems in user interfaces has been probed many times over the last 15 years. The larger problem of trying to infer cognitive processes of a subject by analyzing their eye movements has been studied for more than a century. Eye tracking techniques have always been discussed as having great promise but there has been only sporadic adoption of it as a usability tool, especially in the commercial field. Jacob & Karn (2003) talk of eye tracking as “‘Rising from the Ashes’ rather than ‘Taking off like Wildfire’”. What has caused this to be the case? When so many researchers and cutting-edge commercial ventures have examined eye tracking, and in some cases used it successfully, why is it not more widespread?

There seem to be three areas which have held the field back. The first issue investigators struggle with is hardware that is to complex and difficult to use. This has caused many studies to be restricted in their scope and resources by the amount of effort it has taken to simply collect good eye tracking data. The second is the mountainous task it has been to match the eye tracking data to meaningful stimulus. This has caused investigators to use “static” or highly simplified versions of the interfaces they are examining. This results in data and conclusions which are extremely difficult to generalize and also makes commercial use of eye tracking highly problematic. Software with the ability to map the eye tracking data back to useful stimulus such as application user interfaces and web pages is an example of the tools missing. The third area is the ability to analyze the enormous amount of data that eye tracking generates. Investigators have often struggled to make sense of the quantity of data they have collected and draw powerful conclusions from this.

Modern innovations resulting in new hardware and software tools seem to solve many, if not all, of the outstanding issues holding back widespread us of eye tracking in both the research and commercial fields. New hardware which is “plug-and-play”, accurate and extremely easy to use allows data to be collected painlessly and new software tools allow the data that is collected to be powerfully analyzed.

Previous work and previous problems There have been many studies over the last fifteen years studying the use of eye tracking in usability studies. The difficulties in collecting and analyzing data had previously been prohibitively difficult and time consuming. Many investigators have struggled with difficult tools from both a software and hardware point of view (Aaltonen, 1999; Cowen, 2001; Golberg, Stimson, Lewenstein, Scott & Wichansky, 2002). For a comprehensive history of eye tracking as it relates to usability and discussion of these problems see Jacob & Karn (2003) as well as in Duchowski (2003).

Much of the previous research would lead an investigator to believe that eye tracking holds promise as all these quotes arranged by Jacob and Karn (2003) attest:

• “eye tracking in human-computer interaction remains a very promising approach” (Jacob & Karn, 2003)

• “For a long time now there has been a great need for a means of recording where people are looking while they work at particular tasks. A whole series of unsolved problems awaits such a technique.” (Mackworth & Thomas, 1962, p.713).

• “…[T]he eyetracking system has a promising future in usability engineering” (Benel, Ottens & Horst, 1991, p.465).

• “…[A]ggregating, analyzing, and visualizing eye tracking data in conjunction with other interaction data holds considerable promise as a powerful tool for designers and experimenters in evaluating interfaces” (Crowe & Narayanan, 2000, p.35).

• “Eye-movement analysis does appear to be a promising new tool for evaluating visually administered questionnaires” (Redline & Lankford, 2001).

• “Another promising area is the use of eye-tracking techniques to support interface and product design. Continual improvements in… eye-tracking systems… have increased the usefulness of this technique for studying a variety of interface issues” (Merwin, 2002, p.39).

The investigator might also conclude that, although the insights available are amazingly useful, a high quality, proven solution to painlessly collect and analyze eye tracking data does not seem to be readily available.

Many different groups have designed their own software and/or hardware to try and solve some of the limitations of the available solutions (Golberg, Stimson, Lewenstein, Scott & Wichansky, 2002; Schiessl, Duda, Thölke & Fischer, 2003; Redline & Langford, 2001). Some of these studies have been reasonably successful though none seem to have successfully developed a strong, flexible solution. It is not apparent from the available literature, that a strong hardware and/or software solution is available for off-the-shelf studies.

Eye trackers have traditionally been difficult to setup and run along. They have often required head-mounted cameras or other equipment. They also have been unable to track a significant proportion of the population; 15-20% according to Jacob and Karn (2003). Others have placed extreme limitations on head movement, often requiring chin rests or other forms of head stabilization. Adjustments for each participant have added to the difficulty and time required to run a study, not to mention increasing the complexity for any potential researcher.

Software that makes studies powerful and easy also doesn’t seem to have been available. Such software solutions which automatically and powerfully deal with normal events on web pages such as Page transitions, scrolling and the like have only been developed in the last few years.

New developments solve issues

Adequate tools to record data As discussed there was a desperate need for new tools to collect data. A new generation of eye tracking systems have recently become available, such as the Tobii 1750, which are fully automatic and measure eye gaze without any requirements for specialist technical expertise. They provide a completely normal subject environment without any head mounted equipment or undue restraints on subject movements. This takes the pain out of collecting eye tracking data, and also enables eye tracking to take place in parallel with the collection of other interesting data such as talk-aloud audio and video feeds. These solutions track 95% or more of the population with high accuracy. Calibration is now a 30 second process with no adjustments required by the operator. These improvements remove all of the major hurdles from a broad-based adoption of eye tracking hardware. Commercial and academic labs are currently testing hundreds of subjects in an extremely efficient way with Tobii eye tracking systems. The removal of the need for an expert operator allows many more staff members to run the equipment and therefore maximizes the efficiency with which a study can be run.

There are also a range of new software solutions that have become available which solve the recording and analyzing issues we have already discussed. Modern software tools such as the Tobii ClearView software solution combine collection of eye tracking data with sophisticated logging of the contents of web pages, video feeds of user behavior and capture of screen contents at full resolution. Straightforward analysis tools make it relatively pain-free to analyze data and present the results in a meaningful way. These results allow an accurate estimation about the behavior of a set of subjects to be made. From this estimation it is then possible to answer questions that investigators and, more importantly, their customers consider critical to creating a successful design. These conclusions are also very easy to justify using the built in visualization or statistical tools. Simple, powerful and fast hardware and software tools along with answers to important questions all add up to an all important Return On Investment (ROI) improvement for the overall usability study.

One of the key benefits of solutions such as the Tobii Technology system is they can be easily added into a standard usability lab or test procedure with minimal additional burden. This allows

a practitioner to continue to use familiar and trusted techniques and protocols while incorporating the additional power and insight offered by an eye tracking solution.

New analysis tools Analyzing eye tracking data can be done in a multitude of different ways. One can use visual tools to get a qualitative representation of the viewing of a single subject or multiple subjects. These qualitative measures have a great deal more objectivity than traditional usability measures as they represent the actual eye movements of subjects rather than opinions or subjective observations. One can also use a variety of statistical measures to get a quantitative and thoroughly objective measurement of the behavior of a group of subjects.

Using modern software solutions it is possible to use visualizations to gain clear insight into the behavior of a single test subject. A tool that is useful is one which captures the contents of the subject’s screen as they view it so that they eye tracking data can be displayed over the stimulus the subject was viewing. This can be done in real-time or during post-analysis. This becomes even more useful when combined with video of the user’s expressions and audio of talk-aloud reactions. This allows you to understand the thought process of your subject in a very real and insightful way. This knowledge can be used either during the test sessions, to allow you to ask better questions and direct the study more efficiently, or in post analysis. This also provides context of any feedback and statements you obtain from your subject.

There are statistical tools that can also be used to extract certain key metrics that provide insight into the behavior of a single user. An important metric is a sequence of the user’s fixations, known as a scan-path. Modern software has the ability to automatically analyze eye movement data and generate a scan-path. This identifies points on the web page that the user fixated on, giving great insight into the user’s cognitive process.

Other tools provide an understanding of the behavior of a group of subjects using visualizations and metrics based on total or average results from entire groups of users. Typical examples include “hotspot” or “heatmaps” and certain statistical metrics. This can be even more interesting when you analyze the results of two or more distinct user populations (this might be subjects who were successful in solving a task compared to those who were unsuccessful). These hotspot tools overlay the viewing behavior of a group of subjects onto the viewed web-pages. Various different views are available such as total viewing time, normalized viewing time, number of fixations (independent of fixation length) and the percentage of subjects viewing an area.

There are also a number of statistical measures which are interesting in analyzing eye tracking data. These include statistics about the viewing of specific, practitioner defined areas of the interface. These areas are often referred to as Areas of Interest (AOI). Some of the statistics which can be of interest here are the total and average viewing time, number of repeated examinations, transition statistics between areas and many others. There are other statistics such as scan path length, searching/fixating ratio and others which are independent of AOIs which can be of interest also. A comprehensive history of the use of these metrics is available in Jacob and Karn.

All of these analysis techniques, both those involving visual techniques and those looking at statistical measures are supported by modern software tools such as Tobii Technology’s ClearView software. These results and tools are being successfully used both in industry and academia to provide new insights into the design of pages. The new techniques available are providing a measurable impact on both the actual and perceived value of user testing. They are also providing measurable improvements in the effectiveness of the sites that have been studied.

Additional insights to traditional techniques

Complementing verbal protocols Verbal protocols, also known as think- or talk-aloud protocols, are commonly used in usability testing today. Papers in the field, and the practice of commercial usability labs, show this is often considered to be the usability technique. These techniques could be said to be the industry standard method of user testing. Jakob Nielsen and many others, who are often seen as leading lights in the area of usability, are tireless supporters of these techniques. It appears to be an effective tool to discover usability issues (Ebling & John, 2000 amongst many others). The take-up of this technique in commercial labs shows this to be the case.

Though the think-aloud protocol is a useful technique, the fact that there are some inherent problems with the technique cannot be ignored. There is a range of in-depth research into these issues which is comprehensively examined in Edgar (2005), a paper comparing talk-aloud, retrospective and eye movement data cued retrospective protocols.

In the examination of the previous research into talk-aloud, Edgar (2005) identified the following issues as problems with the think-aloud protocol:

• Inaccuracy – Subjects generate causal theories to explain their behavior rather than reporting their actual mental process

• Incompleteness – Subjects are not able to put thought processes into words. This is especially true of the processes we are most interested in studying (e.g. they will become silent when struggling with an issue). Subjects edit their speech based on what they anticipate the moderator wishes to hear (this can be greatly increased with biased prompting that is often used in practical testing)

• Interference – Where the actual thought process of the subjects is changed by the process of talking as the subject thinks.

The results in Edgar (2005) show the use of eye movement data cued retrospective protocol seemed to uncover a much higher number of usability problems thus suggesting a lower lever of inaccuracy and incompleteness. The subjects were much more likely to complete the tasks they were given and reported a preference for the protocol, feeling that there was less interference from the testing process. This would show that interference is much less of an issue when using this protocol.

Interference is arguably one of the most critical issues in think-aloud testing. This occurs when subject behavior is effected by the requirement to verbalize their thoughts, searching strategies etc as they are performing a task. Both the talk- and think-aloud processes can also slow a subject’s performance or alter their spontaneous responses. These protocols also place a high cognitive load on the subject, often when they are struggling to solve a problem they will become silent, in the very time the practitioner needs them to be talking most.

Many of these issues may not be severe in a particular test but the issues should not be ignored. It does seem that some of the cognitive processes practitioners are looking to gain a window into using verbal protocols are very clearly shown in eye tracking data. Especially when this data is viewed in real-time the requirement on the subject to constantly verbalize, especially in the periods requiring high cognitive load such as searching, should be reduced dramatically. As is shown in Edgar (2005) this does not need to replace verbal protocols and is very much complementary.

One possibility might also be the use of what could be thought of as a fusion of the think-aloud and retrospective protocols. One of the goals of using think-aloud may be to ensure that problems can be understood, both how the user got into the problem and the nature of the problem itself. Using eye tracking one could allow the subject to act without having to verbalize every step or reaction until the subject seemed to be in a problematic situation. When viewing the eye tracking information in real-time it is normally quite obvious when a subject is at an impasse.

The moderator could then prompt the subject to explain the problem they are having and explore how it developed. If this is a major problem a replay of the session with eye tracking information may be useful. This is especially true when the subject has a specific task and you are aware of what that task is. Such a technique then provides the best of both worlds, allowing the power of understanding allowed by the talk-aloud protocol with faster and more natural behavior experienced without the need for constant verbalization.

Stunning Deliverables An issue with current usability testing report is the struggle to communicate their message effectively. They often lack strong tools such as stunning visuals. Having the best results and analysis will not always convince a customer or the customer’s development team. Having objective deliverables such as video of the web session with eye gaze overlaid or hotspots of multiple subjects etc. can provide not only an impression of quality but also the conclusive and objective proof that what is being reported is factual. With previous testing methodologies often the only proof is provided after an extensive redesign and then months later the click-through results and sales show the results. With eye tracking there is the possibility of doing rapid testing to show objective results of any changes required.

It is often difficult to choose a visual design for a website. Individuals in the customer’s organization, designers and developers often have strong opinions on what will work or won’t work in the visual design of a specific site. These opinions are just that, opinions. They are often not backed up by any proof at all. It is extremely difficult to convince someone that their opinion of how a page should be laid out is incorrect. Eye tracking provides conclusive, measurable and repeatable performance measures of a particular visual layout. Even with small sample sizes and an early prototype page you should be able to effectively choose and justify the visual designs to your customer.

Answering typical questions Although eye tracking alone does not offer all answers to a usability study, it can add clear value in answering certain key questions with a greater amount of understanding. A couple of straight-forward examples of this are described below.

What happened in the breakdown event? One of the most important events to fully understand in usability testing is what truly occurs when a subject finds a problem or has some sort of issue - what is often called a “breakdown event”. This is often the very event a usability practitioner is most interested in. Sometimes the reason for an issue or problem may be unclear or the information the subject is able to provide may be ambiguous. Quite often, test subjects tend to pause in the verbalizations using the think-aloud during such events. Having the eye tracking information as a supplement to think-aloud and video of the user’s expressions can provide a level of understanding that is difficult to replicate without the eye tracking data. One could think of the eye tracking information as providing a context to the expressions and think-aloud information one is trying to understand. You can see exactly which piece of text or button is causing them to frown, what they are looking at when explaining their frustration. This can provide vital clues to understanding the breakdown event and proposing an optimal solution to the issue.

What do users see? One of the most obvious questions answered by eye tracking is “What did people see?” This can be a very important issue if the element you are interested in is a key message or offer your customer considers vital. Design and marketing teams spend vast resources developing and promoting messages and it is critical to find out if these messages are getting across and if not, why not. The level of subject recall can be determined by questioning them after they exit the page. This can be problematic as there are numerous possible reasons a subject may not recall the information. The information may not be interesting to them or, after being attracted to the area of the screen containing the information, they read the first piece of text and moved on. Either of these reasons for non-recall may lead you to change the graphical design or layout unnecessarily, a change unlikely to bring any improvement, where a change of copy, likely to be a cheap and effective solution, is what is required.

The meaning and scope of the question “what did people see” could be a huge range of different possibilities. As such the answer may be best illustrated best using scan paths and hotspot visualizations or basic and simple eye tracking metrics such as “total time spent looking at an object”. In some instances a combination of these different tools might illustrate the answer with greater clarity.

Where should valuable content be placed? Eye tracking data obviously points to the attractiveness and/or the level of attention people focus on the various visual elements or sections of a webpage. This allows you to focus valuable development and design resources on the areas of the webpage which will provide the greatest return on investment. Effort placed on high traffic areas may provide measurably higher results than those on unused sections. Conversely one could wish to try and spread attention to other sections of the page as you may feel you are not effectively using the pages valuable “real estate”. The eye tracking information will easily show when a page has text or visual elements which are “dead” or unused space. The hotspot tool shows the spread of attention and areas of “dead space” extremely effectively and with great clarity.

Gaze time

0.00

200.00

400.00

600.00

800.00

1000.00

1200.00

1400.00

1600.00

Benetton logo Nipple Nipple Spoon Arm WebsiteAddresses

WFP Logo Leftside Arm Chest Area

AOI

Gaz

e tim

e Mean

Figure 3 - Hotspot of a watch commercial

Figure 2 - Graph of gaze time statistics

Do available visual cues drive users effectively? Information on what is attended to could also be used to determine if subjects find information that may influence a decision, such as a purchase. A useful tool often used in a study may be a pre-study interview or screening process to ensure a subject is interested in the message or products you are pushing to them. If they report being very open to your product or message then they way they read or scan the information on a page will be extremely interesting and should illustrate well how you are communicating you’re your target audience regardless of if they clicked on or even remembered the information.

Figure 4 - Plot of fixations

How effectively do users complete the task? Another important question is how effectively users are able to find the information they need. Also when they find information do they correctly identify it as the information they are looking for?

In many usability studies, it is common to measure “time to success” of a task. This is one of the few “objective” measurements practitioners have today, but one could think of many cases where this may be misleading. For example, not all tasks have a distinct finishing point defined by a mouse-click. Also, many “time to success” measures may be misleading if the subject carefully reads relevant information before completing the task. In this case, the subject has successfully found the right information quickly but their time to completion may be much longer than a subject who searches quickly in 4 or 5 wrong places before finding the relevant information but then completes the task quickly.

Finding how long it takes a subject to find the correct information is answered in a simple and powerful way by using basic eye tracking metrics. One example is to measure the time from the start of a task to when the subjects looks at the relevant information. For a closer examination, you could determine if they have correctly identified the information simply by looking at a video recording of screen contents with eye gaze overlaid to identify whether they have then read the information or simply scanned it and moved on.

Again this could be a time when think-aloud might be used in conjunction with eye tracking information. When viewing the eye tracking information in real-time one would see the person go to the correct area and then leave without a seemingly appropriate response. This would prompt you to ask the subject what their thought processes were in continuing to search when they had found the correct information.

Time to first fixation

0.00

500.00

1000.00

1500.00

2000.00

2500.00

3000.00

3500.00

4000.00

4500.00

Benetton logo Nipple Nipple Spoon Arm WebsiteAddresses

WFP Logo Leftside Arm Chest Area

AOI

Tim

e to

firs

t fix

atio

n

Mean

Figure 5 - Time to first fixation

Limitations of eye tracking

Data easily influenced When running eye tracking studies it can be very easy to influence the eye tracking data you collect. Pages or interfaces that are known or previously experienced have a very different viewing pattern, even after minimal exposure than pages or interfaces that are completely new. This first experience of a new page or interface can often provide a huge amount of information to the designer about where subject expect to find a piece of information. It is therefore important to provide all the information to the subject on the task required etc before the stimulus is shown. It is also important not to interrupt or introduce secondary cognitive processes during this initial period.

They looked at it but did they really “see” it Although eye tracking information does tell us the direction a person’s eyes are pointed in this does not directly translate into a conscious and complete cognitive understanding of the object their eyes were pointed at. It can be problematic to assume that because someone saw something they consciously processed it or if they processed it they understood what they were looking at.

When we look at eye movements on a “Where’s Waldo” picture for example some subjects look directly at Waldo two or three times before processing the visual correctly to “find” him. The same is true for users of a web page. Just because their eyes were aimed at an area of the screen does not, in all cases, allow you to draw the concrete conclusion that this information went into their conscious mind and was correctly processed. The likelihood is often quite high that this occurred, the longer they have looked at an area the higher this is, but it is problematic to draw definitive conclusions on an individual basis. It is possible though to draw conclusions on groups of people. For example if most people looked at an area with a simple visual message then there is a high likelihood that this information would be passed to a majority of users.

It is often important to examine the next target of the subject’s gaze or examine any action they took on the basis of their eye movements. Here transition matrix statistics are extremely useful as well as the comparison of eye movements to mouse clicks etc.

How people use peripheral vision When looking for information or attempting to solve a task a subject will look around the screen by making a series of saccades. They choose where they will saccade to using both their expectation and their peripheral vision. If the subject is looking for navigation for example they will saccade to something in their peripheral vision which has the appearance of a navigation area to their peripheral vision. Our peripheral vision in of much lower resolution or acuity than our main or foveal vision thus a subject will not be able to “know” what is in an area they are saccading to but will get a “feel” for it. An example of this is people avoiding ads, especially banner ads as these are superfluous to their task and they can easily identify these by their look and their expected location. (http://www.poynterextra.org/eyetrack2004/advertising.htm)

As you would expect the more focused a subject’s task is the more specific they will be able their visual search and thus the more important their peripheral vision becomes. An example of this would be asking someone to find the products section of a website. They will most often saccade directly to the navigation areas or what they expect to be navigation areas. As such columns of a different color on the left hand side of the page or a row of a different color at the top of the page would expect to get an extremely high number of fixations.

People use their peripheral vision to determine what areas seem to be of interest and which are clearly not of interest, especially when they are searching for specific information. They use this area of vision to define targets for their subsequent eye movements or saccades. As a result they often deliberately avoid areas which seem to be “useless” to them, for example advertisements. It has been shown in a number of studies that users have a very highly developed skill of avoiding all ads, be they in-line, banner ads or a variety of other formats.

It is obvious that any eye tracking system will not be able to determine what parts of the peripheral vision a subject was driven by. By definition the eye tracker reports where the subject is looking not the areas where they are not looking.

A visual design could fall down simply because a visual element looks too much like an ad and people then take active steps to avoid the area based on the visual information they collect with their peripheral vision. After you have documented the behavior of subjects avoiding an area, using the gaze plot or gaze replay, it may then be useful to interview them to determine why they avoided an area. Again here it is important to emphasize eye tracking is one of the many tools in the usability arsenal. Interviews and think-aloud often provide the depth of understand of an issue that eye tracking identifies.

Conclusion Modern solutions of the type provided by Tobii Technology remove the issues that have traditionally held back eye tracking. There have been many issues that have previously forced eye tracking into the “interesting but commercially useless” category. Long setup time, high level of expertise required, low reliability of both equipment and data and poor product packaging have been some of these. It has been shown in many commercial labs now that these issues have now been overcome. Modern solutions provide hassle-free collection of data and tools to allow a practitioner to easily analyze this data.

The value added by eye tracking in traditional usability studies is clearly significant. One can answer more questions, answer them with greater depth and conviction and develop deliverables that clearly illustrate these results. This is shown in research such as Edgar (2005), the Eye Track III study (http://www.poynterextra.org/eyetrack2004) as well as the business growth from companies running eye tracking tests. The questions which eye tracking can answer are also one which simply cannot be answered by traditional techniques. All of this will provide clear advantages in marketing as well as improved ROI for your customers.

Modern eye tracking solutions provide easy to use and effective tools which can be used to compliment traditional usability techniques. Running eye movement data cued retrospective protocol allows some of the more problematic issues of traditional usability testing to become a thing of the past. There is also the possibility to run more ambitious studies looking to gain even greater insights from a larger subject base than the traditional 6-15 subjects that are often run in usability studies.

The ability to run traditional studies, gathering all the data that they collect, plus have the added value eye tracking provides at a minimal additional cost and effort adds enormous value to a study. This combined with the less problematic methodology ensure a traditional usability test can provide better results with more confidence. The extra information and insight that can be gained using eye tracking ensures that customers are provided the maximum benefit from their investment. It is only a matter of time until eye tracking protocols become a standard part of usability testing and website design in general.

References Aaltonen A., (1999). Eye Tracking in Usability Testing: Is It Worthwhile? In CHI'99 Workshop The Hunt for Usability: Tracking Eye Movements, Pittsburgh, PA, May 1999.

Cowen, L. (2001). An Eye Movement Analysis of Web-page Usability. http://www.cs.ucl.ac.uk/staff/J.McCarthy/pdf/library/eyetrack/laura_cown.pdf

Ebling, M.R. & John, B.E. (2000). On the contributions of different empirical data in usability testing. In Proceedings of the conference on Designing interactive systems: processes, practices, methods, and techniques 2000.

Eger, N. L. (2005). Using Eye-movement Data to Cue Retrospective Protocols in Online Usability Testing.

Golberg, J.H., Stimson, M.J., Lewenstein, M. Scott, N. & Wichansky, A.M. (2002). Eye tracking in web search tasks: design implications. In Proceedings of the Eye Tracking Research & Applications Symposium 2002. Pg 51-58. New York, ACM.

Jacob, R. & Karn, K. (2003). Commentary on Section 4: Eye tracking in human-computer interaction and usability research: Ready to deliver the promises. In The Mind’s Eye, Elsevier, North Holland

Neilsen, J., Clemmensen, T. & Yssing, C. (2002). Getting access to what goes on in people's heads?: reflections on the think-aloud technique. In Proceedings of the second Nordic conference on Human-computer interaction. Pg 101-110 New York, ACM.

Redline, C.D. & Lankford, C.P. (2001). Eye-movement analysis: a new tool for evaluating the design of visually administered instruments (paper and web). Paper presented at 2001 AAPOR Annual Conference, Montreal, Quebec, Canada, May 2001. In Proceedings of the Section on Survey Research Methods, American Statistical Association.

Schiessl, M., Duda, S., Thölke, A., Fischer, R. (2003). Eye tracking and its application in usability and media research. In Sonderheft: Blickbewegung in MMI-interaktiv Journal - Online Zeitschrift zu Fragen der Mensch-Maschine-Interaktion. ISSN 1439-7854. Gastherausgeber: Katharina Seifert & Matthias Rötting. 12.03.03, Ausgabe Nr. 6.

http://www.poynterextra.org/eyetrack2004/advertising.htm