Eye Tracking the User Experience. Aga Bojko

remote eye tracker may be suitable for your study setup. For example, if you would like to use a remote system to study how drivers scan the dashboard while driving, some remote eye trackers won’t work because they are too large and awkward to be installed in a car. However, other remote systems have been designed specifically for in-vehicle research.

      By the same token, even if you decide on a wearable eye tracker, not all devices will be able to accomplish what you have in mind. For example, a wearable system may be needed for studying the use of smartphones by doctors in hospitals. But if the scene camera (i.e., the camera that records the participant’s view) is fixed to the frame of the eye tracking glasses and can’t be angled down, it may miss some of the doctors’ interactions with the phones. A more appropriate system for this study would require a flexible scene camera that could be tilted down, or a camera that could capture wide vertical gaze angles.

Technical Specs You’d Probably Rather Not Know About

      Once you have selected the type of eye tracker that is appropriate for your study, you will probably want to compare the available models. There is at least one full page of technical specifications for every eye tracker out there. This section, however, will cover only the few specifications that may have the most impact on your decision to use an eye tracker for your research.

       Sampling Rate

      Sampling rate is one of the most important features of an eye tracker. Measured in hertz (Hz), it is the number of times the eye tracker registers the person’s gaze location per second. This means that every second a 120 Hz eye tracker collects 120 data points for each tracked eye. That’s over 200,000 data points recorded during a 30-minute eye tracking session! But don’t worry, you probably won’t have to deal with all of these raw data—they will be nicely combined into meaningful fixations for you by the software.

      The range of sampling rates currently available is quite wide—from 25 Hz to 2000 Hz. If you only want to determine where people are looking, a higher sampling rate doesn’t do a better job at identifying gaze location than a lower sampling rate. However, the higher the sampling rate, the more precisely you can measure when a fixation started, and the smaller the error in fixation duration. Fixation durations measured by 25–30 Hz systems can have an error of +/– 20 ms, while durations measured by 250 Hz eye trackers can have an error of only +/– 2 ms.¹

      Therefore, if you are interested in measuring not only where people look but also for how long (and as a UX researcher, you might be), you should be aware of this error because it can increase noise in your data. To reduce the error, you should aim for higher sampling rates. Alternatively, you can collect more data (for example, by testing more participants) to wash away the noise.

      Considering the fact that systems with higher sampling rates cost more, how high do you really need to go? The general rule of thumb for UX research is 50–120 Hz. These sampling rates will produce a fixation duration error of +/– 10 ms or less. Keeping in mind that a typical fixation lasts between 100 and 500 ms (half a second), an error of this magnitude is generally acceptable in our field.

      Sampling rates of 250 Hz and higher are required for research measuring characteristics of saccades (for example, their speed), as well as micro eye movements, which can be of interest to other fields, such as neuroscience.

       Accuracy and Precision

      The accuracy of an eye tracker is the average difference between what the eye tracker recorded as the gaze position and what the gaze position actually was. You should want this offset to be as small as possible, but it is unrealistic to expect it to be equal to zero.

      Accuracy is measured in degrees of visual angle (see Figure 3.6). Typical eye tracking accuracy values fall in a range between 0.5 and 1 degree. To give you an idea of what that means, one degree corresponds to a little less than half an inch (1.3 cm) on a computer monitor viewed at a distance of 27 inches (68.6 cm). In other words, the actual gaze location could be anywhere within a radius of 0.47” from the gaze location recorded with an eye tracker with one degree of accuracy (see Figure 3.7).

      

FIGURE 3.6 To determine the visual angle of an object, extend lines from the eye to the top and bottom of the object. The farther (or smaller) the object, the smaller the visual angle.

FIGURE 3.7 The “X” indicates the gaze location the eye tracker recorded. Assuming tracking accuracy of 1°, 1680 x 1050 px screen resolution, and 27” viewing distance, the participant’s actual gaze location could have been anywhere inside of the circle.

      Accuracy values reported in eye tracker manuals are measured under ideal conditions, which typically include testing participants with no corrective eyewear and taking the measurement immediately after calibration. During “real research,” the difference between the reported and actual gaze locations can be larger for participants wearing glasses or contact lenses, or those who moved at some point following the calibration procedure.

      Precision is a measure of how well the eye tracker is able to reliably reproduce a measurement. Ideally, if the eye is in the same location in two successive measurements, the eye tracker should report the two locations as identical. That would be perfect precision. In reality, precision values of currently available eye trackers range from 0.01 to 1 degree. These values are calculated as the root mean square of the distance (in degrees of visual angle) between successive samples. Because the precision values reported by manufacturers are measured using a motionless artificial eye, tracking real eyes will exhibit less precision.

      Accuracy and precision are not the same but they are often confused with one another. Table 3.3 should clarify the distinction between the two. The cross in each cell of the table indicates the actual gaze location, while the dots are gaze locations reported by the eye tracker.

TABLE 3.3 THE RELATIONSHIP BETWEEN EYE TRACKING ACCURACY AND PRECISION

       Head Box Size

      Head box indicates the freedom of head movement that a remote eye tracker allows (see Figure 3.8). As long as the participant’s head remains in this imaginary box, the eye tracker can do its job and track the position of the eyes.

FIGURE 3.8 Eye tracker head box indicating the allowed range of head motion.

      It is impossible for people to be completely motionless during eye tracking sessions and, in UX research, you typically don’t stabilize participants’ heads using chinrests or other contraptions. Therefore, the larger the head box, the less data will be lost due to head movement. The size of the head box depends on the eye tracker, but the typical ranges are 12–17 inches (~30–44 cm) in width, 7–9 inches (~17–23 cm) in height, and 8–12 inches (~20–30 cm) in depth.

      Wearable eye trackers don’t have a specified head box. Because the eye tracker moves