Introduction to Eyetracking: Seeing Through Your Users’ Eyes

By Matteo Penzo

Published: December 6, 2005

This article is the first in a series of articles on eyetracking that will appear in UXmatters. Over the coming months, I’ll use eyetracking to evaluate a lot of world-renowned user interfaces—including Web sites like Amazon.com®, Google™ News, and eBay®; Rich Internet Applications (RIAs); and desktop applications—and analyze quantitative eyetracking data to provide best practices for designing user interface elements like navigation systems, menus, and forms, and for effective ad placement.

“Eyetracking offers unique benefits and provides a practical alternative to conventional discount usability testing.”

For some time, usability professionals have evangelized the term discount usability testing. Discount usability testing was a product of the early years of the Internet. Its techniques promised to provide a simple, fast, and relatively economical way of conducting usability studies and improving users’ experience of the Web and other software user interfaces. However, such studies are mainly qualitative and subjective. The data reflect users’ conscious thoughts and feelings as well as the observers’ impressions. Some think this is the best, even the only method of conducting usability studies, but there are other—in some situations, perhaps better—ways of evaluating user interactions. Eyetracking offers unique benefits and provides a practical alternative to conventional discount usability testing.

As John Elvesjo—founder of the firm Tobii Technology, which produces eyetracking technology—said, “The eye is the mirror of the soul, and the soul is the mirror of our thoughts.” This quotation expresses wonderfully the magic of human sight and the manifold possibilities that the study of users’ eye movements makes possible.

Adoption of Eyetracking Methods

“Eyetracking can show which parts of your user interfaces users see and which parts seem to be invisible to them—not just by observing users and gathering qualitative data, but also by analyzing their gaze plots and other quantitative data.”

Usability gurus have written little on eyetracking, which is odd considering what an effective technique it is. Back in 2000, Jakob Nielsen wrote an Alertbox column about an eyetracking study of readers’ behaviors when reading news on the Web, and Jared Spool’s User Interface Engineering published an article on using eyetracking to test the usability of Web sites back in 1998. These early articles came out before recent advances in eyetracking technologies and techniques. The near silence of usability gurus on the subject of eyetracking since then is quite surprising to me. Eyetracking is not rocket science. It is a simple and effective means of evaluating Web pages and user interfaces.

Fortunately, things are now changing. If you take a look at Steve Krug’s book Don’t Make Me Think, you will see a graph that resembles a gaze plot. He derived that graph from years of experience observing user interactions with Web pages. Eyetracking now makes it possible to accurately measure the kinds of gaze patterns that Krug observed. Eyetracking can show which parts of your user interfaces users see and which parts seem to be invisible to them—not just by observing users and gathering qualitative data, but also by analyzing their gaze plots and other quantitative data.

What You Need to Know About Human Sight

Human sight has a visual field of about 120 degrees, encompassing three degrees of visual acuity: foveal, parafoveal, and peripheral vision. We primarily take in visual data from the outside world through the fovea, which provides the greatest visual acuity. We move our head and eyes to focus the fovea on objects of interest that we want to see.

Eye movement has two states:

  • saccade—A saccade is the fastest movement of which the human body is capable—taking only about 30 milliseconds—and centers content within the foveal area.
  • fixation—A fixation occurs when this movement stops, permitting the eye to acquire content.

During saccadic activity, we cannot see at all. We perceive the world visually only through fixations. The brain virtually integrates the visual images that we acquire through successive fixations.

How Eye-tracker Works

The Tobii Eye-tracker, shown in Figure 1, is an instrument that is capable of capturing data about both saccadic activity and fixations of the foveal area.

Figure 1—Eye-tracker

Eye-tracker

To accomplish this, it uses an infrared light source to illuminate the eyes, a CCD (Charge Coupled Device) sensor to capture a reflection of the user’s eyes, as shown in Figure 2, and eye-gaze analysis software to process the data.

 

Figure 2—Eye-tracker in operation

Eye-tracker in operation

By using a remote, digital eyetracker, we can record saccades and fixations, the length of each fixation, the distance to the eye, the pupils’ diameter, and a lot of other useful data within an accuracy of half a degree. Thus, eyetracking can, for example, tell us what a user is looking at and for how long.

 

Eye-gaze analysis software produces various graphs that are useful for data interpretation, as follows:

  • Hotspots—Generalize the behavior of a group of test subjects. They’re very similar to heat maps, as shown in Figure 3.
  • Gaze plots—Provide a comprehensive image of all the eye-gaze data from a single eyetracking test. Figure 4 shows a gaze plot.
  • Gaze replays—Provide both real-time and slow-motion replay of the paths a user’s eyes followed during an eyetracking test.

Figure 3—Hotspots

Hotspots

 

Figure 4—Gaze plot

Gaze plot

Both gaze plots and gaze replays are much more informative than hotspots. They permit an evaluator to see through each individual user’s eyes, displaying the gaze path a user’s eyes followed while exploring a Web page or interacting with a user interface.

To facilitate analysis of a user’s interactions at a deeper level, you can export all of these data to an Excel® file, as shown in Figure 5.

Figure 5—Data exported to an Excel file

Data exported to an Excel file

Analyzing the angles between saccades or the distance between two fixations can reveal a lot of information about a user’s thoughts while interacting with a Web page or user interface. Eye-tracker provides a powerful set of tools that lets you define areas of interest, or boxes of content, on a Web page such as an ad, then determine a user’s attentiveness to them by evaluating the percentage of time the user spends viewing each area of interest relative to time spent viewing other areas of a page during normal page browsing.

Evaluating Usability Using Eyetracking

“A very effective test methodology combines standard usability evaluation techniques—such as the think-aloud protocol—with the latest eyetracking capabilities.”

A very effective test methodology combines standard usability evaluation techniques—such as the think-aloud protocol—with the latest eyetracking capabilities, as shown in Figure 6. Eyetracking introduces quantitative measurement to the field of usability evaluation, which has typically provided mostly qualitative data. Whether you are videotaping or logging data during a usability test session, the think-aloud protocol lets you collect qualitative data such as a user’s mood through tone of voice and facial expressions, while Eye-tracker gathers and records quantitative data such as pupil diameter, fixation coordinates, fixation length, saccade angles, and more. The combined data of these two methods provide a broad overview of the problems a user encounters in a user interface while performing a task.

Figure 6—Combining the think-aloud protocol with eyetracking

Combining
the think-aloud protocol with eyetracking

We can now leave expert evaluators’ interpretations and de facto standards behind and instead determine things like the optimal position for the labels of fields in a form, the best placement for a navigation bar, or the most visible location for a logo or an advertisement by evaluating gaze data that shows the percentage of users actually seeing them.

Eyetracking technology lets us observe the behaviors of users and provides quantitative, objective data that lets us develop a deeper understanding of how users interact with Web pages and user interfaces. Armed with this knowledge, we can improve the usability and usefulness of Web pages and software products.

Eyetracking in New Interaction Paradigms

“Usability and marketing studies are just two of the possible applications for eyetracking technology.”

Usability and marketing studies are just two of the possible applications for eyetracking technology. In addition to using eyetracking to analyze Web pages and software user interfaces, through my research at Consultechnology, I am exploring how we can take full advantage of eyetracking capabilities and working to discover better interaction paradigms. For example, by capturing the screen coordinates of a user’s gaze, eyetrackers

  • allow people to use their eyes as input devices, thus creating new and more natural and ecological ways for them to interact with systems
  • can provide the means for a system to know where a user’s attention is focused while reading or interacting with a user interface and even to anticipate what a user wants to do next
  • can analyze a driver’s level of attentiveness while driving and prevent drowsiness from causing accidents

What’s Next?

Eyetracking provides a relatively quick and straightforward way of conducting studies of any kind, making it possible for me to present reports on eyetracking studies every month in UXmatters. Through these studies, I will evaluate specific user tasks and interactions with Web pages and user interfaces. My reports will demonstrate best design practices through the analysis of quantitative test results.

Next Month—“Evaluating the Usability of Search Forms Using Eyetracking: A Practical Approach

Bibliography

Krug, Steve. Don’t Make Me Think! A Common Sense Approach to Web Usability. Indianapolis, Indiana: Que, 2000.

Nielsen, Jakob. “Eyetracking Study of Web Readers.” Alertbox, May 14, 2000. Retrieved December 6, 2005.

Shroeder, Will. “Testing Web Sites with Eye-Tracking.” User Interface Engineering, September 1, 1998. Retrieved December 6, 2005.

30 Comments

Eh eh… this article was written some months ago; when the ubiquitous Nielsen wasn’t yet thinking on eye tracking technologies (Nice move Jakob! :-) Well, this is just another (big) brick added to the amount of buzz on E.T. I see on the Web today. And this will hopefully lead to better UIs and interactions.

Actually, I have been writing about eyetracking since 1993; see my paper on non-command user interfaces (scroll to section 3.1). Eyetracking is also discussed in several sections of my book Usability Engineering from 1993.

I used my first eyetracker at the Naval Research Laboratory around 1990.

And our current project (which will be presented at the Eyetracking Web Usability seminar in 2006) started in April 2005 (and has been planned since mid-2004), even though I didn’t write about it in public until this month. It takes a long time to do all the research for a full-day seminar, but there’s no need to alert the competition about what we are working on until the day we have to announce the conference program :-)

I think every usability guy is fully aware of your deep knowledge in the usability field, Jakob. While writing this article, I googled your company site for eyetracking, and sadly, excluding the 2006 event, nothing came out—no commercial services based on eyetracking. I’d say that ONE Alertbox on eyetracking in 10 years is quite few, but I have to respect your own editorial choices. My humble opinion is that eyetracking is one of the most powerful tools HF studies have—shall I say the most powerful tool? :-) Eyetracking offers much more than paper prototyping, expert evaluations, cognitive walkthroughs, etc. That’s why I was so surprised to find the big guys weren’t using it at all. My objective is to have eyetracking tests widely used in order to build better interfaces and interactions. That’s why I was so happy to see your Usability Week 2006 conference schedule included eyetracking. :-)

Any thoughts on the effect of think-aloud protocol and moderator interaction on eyetracking?

Josh, think-aloud can heavily affect the test execution/results, because it may continually remind a user that she’s being tested.

That said, I think or - better - from experimenting, judge that eyetracking and an adapted version of think-aloud can go together. You’re going to use a less invasive think-aloud methodology, because it’s obvious to the Eyetracker what a user is looking at, even though she doesn’t mention it.

We use both (i) videotaping and (ii) gaze recording of the user behaviors (i) and interactions (ii) in a way that makes it possible for us to have a deep quantitative (mainly ii) and qualitative (mainly i) analysis.

So, for example, we can say that while a user was upset (i) about not finding a Search field, she actually looked at it (ii) several times with no understanding of the field’s function. This kind of data analysis would be impossible without the synchronization of an eyetracking test and think-aloud.

But the eye-tracking specialist seldom interacts or forces a user to think aloud. It just depend on the user’s mood: if she feels comfortable speaking, we let her do so; if not, we just record and analyze facial expressions and gaze paths.

Is there any general research or guidelines that has come out, using eyetracking ? Any research you know about extracting general pattern using it ?

Yet, I’m performing eyetracking studies in the Netherlands. Customers are really enthusiastic about the gaze replays. During studies, I ask test participants to think aloud too. Many test participants do not say a lot. I have no problem with that, because the eyetracker registers eye movement continuously.

Audne: You can find interesting studies at Eyetrack III. There are a lot of studies coming out. Most of them are commercial products that you can easily google for.

Robert-Jan I agree that gaze replays are a powerful pre-sales/marketing tool and a nice deliverable, too. (I’m working to have some of them published here - darn Flash video :-)

Speaking of think-aloud, what I was saying is that sometimes think-aloud can invalidate the results. I’m thinking about the lab setup we’ve built to conduct the search form tests.

To avoid skewing the results of an eye-tracking usability test, can we simply eliminate the think-aloud component? Perhaps replaying the video after the test and questioning (e.g. “why did you look at that?”) the participant would be more sound? Any thoughts on this approach? Has it been done that way?

Hi Do you have some information about eyetracking regarding non-online media such as TV? And what about eyetracking measuring the emotions the viewer might be experiencing at the time he/she is viewing an ad.

Scott

You really have a good point by asking the test participant questions while replaying the video. It could be useful when participants conduct a so-called free session. When they perform a task like ‘order article X’ in a Webshop they concentrate on ordering and are less focused on design and images.

It seems to me a really interesting discussion is going on here. I have to apologize for the delay in my reply to the last comments.

Scott I’d say that it really depends. In the lab setups I’ve developed for the articles here on UXmatters, we aren’t performing think-aloud at all. And the results are still bright.

What I could suggest is not to force the user to speak, but let her behave naturally. So that IF she wants to speak then let her do so; you’re going to have so much eyetracking-based data to analyze afterward, and the “qualitative” part of the tests could really help.

I don’t agree with what you and Robert-Jan are suggesting; we never show gaze path videos to the users, since most (I’d say 50%) of the eye movements we record are unconscious. Thus, they’ll never be able to explai, the why.

Juan-Jose We have tested TV ads, and the results are as brillant as in the Web case. In fact, testing a TV ad is not very different from testing a Flash-based site, technically speaking. In my research projects, I’m also using eyetracking to let a user control 3D films on TV using just her eyes.

It’s very difficult to measure emotions (are emotions even measurable?) using eyetracking, which is very good at letting you understand cognitive processes. What we’re working on is testing with another machine developed by one of our partners that lets you measure users’ biodata, and then pairing the two types of data.

Thanks for your comments, Robert-Jan, Matteo, & Juan. Very interesting, indeed. I am currently engaged in post-graduate academic research at a university in New Zealand. Specificially, I am looking at Web usability (for the elderly user) and its relationship to user satisfaction. How does one measure satisfaction, you say? Now, this is a very good question…

BTW, I just got my copy of Prioritizing Web Usability/. Anyone read it yet?

LOL, we spell it “prioritising”.

Does someone know who are the main companies that provide eyetracking services for advertisement research? I’m doing my thesis, and it would be great if i could add a chapter with their services and differences. Thanks!

Francisco

The company I previously worked for sells this type of services. Go to http://int.consultechnology.com for more information.

In the US, you might get in touch with www.usercentric.com. That’s a company I have known, and they use eyetracking for tests—also on communication—using the same technology I have used for these tests.

For other countries, you might want to contact the Eyetracker manufacturer tobii, at www.tobii.com and ask for their local partners.

Hope this helps.

Hello Thanks for this great article! Are there any studies showing that eyetracking helps usability practitioners find more usability problems with a site or tool? I have looked though several academic databases, but I can’t find any. I’d love to follow any leads you can give me.

Thanks, Josh

Hi, I have been reading the comments and find everything very interesting. I am an undergraduate and am thinking about doing my dissertation using an eyetracker. Ideally, I wanted to compare how much attention people pay to/between advertising overlays and normal advertising boards during sports fixtures on a television. Does anyone have any ideas or information on this? It would be much appreciated.

I am a student in computers and information. Eyetracking is my graduation project, so I want to know from your experience, which is better to use in eyetracking for easy algorithms and better results: a Web camera or an infrared camera? Thanks for your help. - Amal

Amal, you can prototype an eye gaze-based product using a Webcam, but I strongly suggest you move to infrared equipment to improve the quality of your results. Webcams are just too slow and imprecise.

Hope this helps.

Dear all, Thank you for this interesting article and discussion. I plan to use eyetracking in an ongoing project, aiming to explore how science students use graphics during problem solving. Only for piloting purposes, I would like to use a low-cost eyetracker, such as the ones I have found at http://swik.net/eyetracking that utilise a Webcam—for example, the Starburst. Unfortunately, I have been unable to locate information on the actual camera setup. Any clues or information on other —preferably Windows-based—eyetrackers?

I am pursuing my degree and doing a project on eyetracking, using a Kalman filter-based eyetracker. I want to know much more information on this topic. If anyone knows of any links to more information, please let me know.

Hi Everyone

Very interesting issue. I’m a psychologist and participate in a neurosciences laboratory in Chile. I’d like to take a masters in the USA, and I’m very fond of this matter. Can anyone help me by giving me some information about investigation centers or university programs leading this kind of topic and social psychology?

Cheers

Hi Aldo

With regard to your question about university programs that study eyetracking… I assume that you mean eyetracking and cognition—as opposed to hardware or software development.

In the U.S., groups that use eyetracking to study Web experiences and user behavior are generally part of an Ergonomics department.

If you are more interested in scene search or high-level visual cognition, I’d recommend looking to Psychology departments and interdisciplinary vision programs. The Psychology departments at UC Berkeley, USC, and University of Illinois at Urbana spring to mind immediately, but there are a number of excellent departments. UC Berkeley’s Vision Science Program would be a great place to start looking. Research groups associated with optometry schools and neuroscience departments—like UCB’s Vision Science Program—also study low-level visual cognition. Many different eyetracking technologies are used in oculomotor control research, for example.

If you are interested in studying or developing new uses for eyetracking or computer vision technologies, computer science departments have ongoing work in these areas. I’d also recommend looking into MIT’s Media Lab. They do a lot of research into social computing, developing technologies, etcetera.

I know all this information is very broad, but I hope it helps. The nice thing about grad school is that, in the right lab, you can tailor your own research to your interests.

Hello,

This is an interesting discussion here. For those who are trying to use ordinary Webcams for eyetracking, I recommend you add an infrared light source at the height of the user’s eyes—just as explained in this article. Even ordinary Webcams, which are especially cheap, should be able to take advantage of IR light as it does not get filtered. So a panel of high-intensity IR-LEDs in addition to your Webcam could produce bright images without blinding the user. Those panels are cheap and, if your camera shows IR-light, you can easily test it by pointing at the panel with a remote control and pressing some knobs! Also, if you google “Webcam” and “Infrared,” you’ll find several sets of instructions for making an IR-only-camera out of your Webcam by adding some dark photo negative material to the lens—which, I guess, would not produce bright enough pictures for eyetracking purposes.

Just some thoughts! :-)

I am a student of VES Institute of Technology, India. I, along with my team, am making an eye tracker system as a part of our final year degree project. We would like to know which components we can use to make an eye tracker, so we can optimize on accuracy and cost and the software we could use for it.

What is up with this eye tracking stuff???? I know people see through my eyes, and I’ve known this for years, but isn’t this an invasion of privacy???? I know that people will continue this anyway, but why?

Ahhh, this technology is amazing. I can’t wait to see what will become of it.

Hi. This is really interesting post. Thank you! I have just subscribed to Your rss!

Best regards

Hi Everyone,

For those who are interested in low-cost and fast eyetracking, I suggest you have a look at the computer vision-based approach. It combines saliency maps and human visual attention mechanisms.

See how real eyetracking compares to computer vision-based eyetracking.

Cheers

Hi Everyone—BMW have tried many eyetracking systems from around the world, but recently decided to use Dikablis exclusively, where mobility and ease of movement is essential—like in testing cars and equipment.

If you would like to know more, please feel free to contact me.

Kind regards, Robert Wilson

Join the Discussion

Asterisks (*) indicate required information.