Applied Human-Machine and Median Interaction Modeling: Why Trained Researchers (Experts) Get Best Eye-Tracking Results

The Best Eye-Tracking Expert is an Engineer, Optometrist, and a Psychologist

The best research expert for running an eye-tracking study embodies the characteristics of an Engineer, an Optometrist, and a Psychologist. This is because tracking the human eye to determine what a study participant is seeing is much more involving than simply putting them infront of an eye-tracker and having them look at a stimulus. There are many factors involved. First, the equipment never works as flawlessly as intended and the equipment requires constant adjustment to meet the variability of study participants. Engineering knowledge for this purpose has been a tremendous asset. Second, we have had to develop custom eye-tracker calibration procedures, and use ‘tricks’ to determine if study participants are actually ‘seeing’ where their eye is pointing. This has taken an understanding of human vision drawn upon from the field of Optometry. Third, we have had to draw on knowledge from the field of Psychology to help us design data collection paradigms that are not confounded by commonly unnoticed characteristics of the data collection session. This knowledge has helped us maintain the ethological plausibility of our study paradigms and has helped us reduced the effects of psychological variables such as stress, or variability of task instructions that might impact our data. Finally, the experience of an Engineer has provided us with the ability to remove noise from collected eye-tracking data and apply robust statistical analysis methods to obtain useful results. In all, it has been advantageous for use to combine these multiple areas of expertise in our research.

There are multiple factors to consider when using eye-tracking to evaluate visual media such as one would encounter while playing video games, surfing the web, or even watching billboard advertising while walking down the street. The remainder of this page is dedicated to describing issues related to various aspects of tracking eye-movement to identify where people are looking. Since this is a very broad topic, eye-tracking issues and solutions will be added to this page on an ongoing basis.

The Advantage of Experience: Strange Eye Behavior, Multiple Studies and Many Participants

The first eye that I ever tracked was my own. My collegues and I had built a reflective-type eye-tracking system and we were evaluating how well it could track where we were looking on the computer screen. The software we wrote to display where the eye-tracking system 'thought' we were looking did so by animating a small 1.5 cm red dot on the screen. The software was written to move the dot around the screen, following the motion of our tracked eye. To a degree, it did what was intended; however, there always seemed to be something strange happening that would make the dot stray from areas of the screen at which we believed we were looking. One of the strangest eye-movement behaviors is when the dot drops below what we beleive we are looking at. All we could say for sure at the time was that our 'attention' is where we believe we are looking, but the tracked part of the eye (indicated by the red dot) is below where our attention lay. We tracked our eyes while we viewed an animation of a small circle on the screen moving left and right across the screen. We found that very often, the tracked part of our eye (and hence the red dot) would move left and right with the animated circle, however about 1 or 2 cms below it-- even though we believed we were looking directly at the animated circle.

Since my first eye-tracking experience and discovering some of the strange behaviors of the eye, I've had extensive time to play with our eye-tracking system and the red dot on the screen. I've learned how to move the dot without any apparent change in what I see on the screen. I've learned how to move the tracked eye slightly up and slightly down, and still 'see' and attend to the visual stimuli of interest. I also seem to have been able to learn how to move my eyes independently (just a little, and a scary thought!). Since then, I have written new eye-tracking algorithms, and have learned how to compensate for some of these strange eye behaviors. I have tracked eyes in 3 studies totalling over 130 people and I have learned a great deal about the human eye and visual system.

Abnormal or Changing Visual Correspondence

Not all eyes and people are alike and their individual differences and personal vision characteristics must be identified and compensated for before data that is useable in a study can be collected. One of the main characteristics of vision that we have encountered and compensated for is abnormal visual correspondence.

What is it? : Visual correspondence is a measure of the similarity of how the image that a person is looking at is projected on the retina of each eye. Normal visual correspondence (or perfect visual correspondence) would have the visual image projected to the same part of the retina in both their left and right eyes. Abnormal visual correspondence means that the visual image is not projected onto the retina of each eye the same way.

Our Experience: We have found that a notable percentage of the population has some kind of abnormal (or non-perfect) visual correspondence. Many of the people who have walked through our door to participate in our studies occasionally suppress vision from one eye, either all the time, or only some of the time. For example, supression of one eye might be more likely towards the end of a study when the particpant becomes fatigued. Hence, we have had to make measurements to identify these people in the study and to make appropriate adjustments to the equipment to accommodate them.

Our Solution: Because abnormal visual correspondance is so common, the method of tracking both eyes while participants in a study perform a task and averaging the measured eye-positions is unwise. Our current suggested approach is to first, before the participant is seated for the study, determine which of the participant's eyes is dominant. This can be accomplished using a number of simple optometry tricks. Once eye-dominance has been identified, note the eye dominance in the behavioral data so that it can be looked up later, and then track both eyes separately during data collection. Calibration for each eye should be done while the other eye occluded. Once the data have been collected, evaluate the correspondance between eyes. If the correspondance is weak, then use the data from the dominant eye. If the eye-tracking equipment available does not track both eyes, track the dominant eye.