Creating an environment to facilitate collection of accurate eye tracking data

During the pilot phase the researcher established that a range of factors could negatively affect the reliability of the eye tracking data. These are now described and discussed in turn.

92

3.6.4.1

Calibration of the eye-tracker

Before starting to collect data the eye-tracker needs to be calibrated to each individual participant. Calibration involves asking participants to follow a dot which moves around the screen, whilst the eye tracker collects and analyses data about the participants eyes. The number of points that the dot moves to and the speed at which the dot moves can be selected by the researcher. The researcher found that using a slow nine-point calibration (the highest number of points permitted) gave the most accurate calibration. During calibration the moving dot pauses at nine different points: three rows (top, middle and bottom of the screen), each containing three calibration points (left, middle and right of the screen). During calibration the eye tracking software compares both light and dark pupil measurements to identify which readings are most accurate given the light conditions and the physical characteristics of the participant’s eyes.

After the calibration the researcher performed a check to review the accuracy of the calibration. This involved re-displaying the calibration points and asking the participants to focus on the points one at a time. The researcher was able to see a screen with both the calibration points and a moving dot, representing the participant’s gaze point as recorded by the eye tracker. This enabled the researcher to check the accuracy of the calibration. If the researcher was not satisfied with the accuracy, the calibration exercise was repeated.

3.6.4.2

Lighting

The overwhelming majority of eye tracking equipment (the Tobii X2-60 included) rely on tracking the reflection of light on the participants’ cornea. If the lighting in the room alters during the session, the accuracy of the eye tracking data can be affected. For example, during the day, if the sun comes out and increases the light levels in the room this can affect the

93

accuracy of the eye-tracker. When the researcher conducted the pilot a room with no windows was used. However, this room was not always available and / or participants were not necessarily able to travel to the campus where this room was located. Therefore, during the main study the researcher conducted several sessions where, as lighting conditions altered, the accuracy of the eye-tracker diminished and data from that session was unusable. Therefore, it is advisable to have a room where there is no natural daylight or where blinds or curtains can be used to minimise changes in levels of daylight entering the room. The room needs to be well lit; poor levels of lighting cause the participants’ pupils to dilate and this can reduce the accuracy of the data gathered. The type of artificial lighting in the room also has an impact on the quality of the eye tracking data.

Although Holmqvist et al. (Holmqvist et al., 2011:17) suggest using fluorescent lighting, the researcher has found that if a fluorescent light is flickering this reduces the accuracy of the data, especially when the room is illuminated by a single fluorescent bulb. The researcher had greater success in a room lit with a series of LED lights which provided bright, evenly lit conditions. Data was gathered most successfully in a room with closed blinds to reduce the effects of natural daylight and multiple LED lights.

3.6.4.3

Stability of the equipment

It may seem obvious but the surface on which the eye-tracker is placed needs to be extremely rigid. For this research the participants needed to be able to type on a keyboard without disturbing the eye-tracker mounted on the monitor. Once again during the pilot study and during early data collection sessions the researcher discovered that even small vibrations of the monitor, to which the eye-tracker is attached, proved sufficient to interfere with the accuracy of the data, once again leading to inadmissible data. The researcher suggests that as

94

a rule of thumb if, when you place a glass of water a desk or table and proceed to gently thump the desk the surface of the water in the glass is disturbed, the desk is not a sufficiently sturdy surface on which to place the eye-tracker.

Holmqvist et al. (Holmqvist et al., 2011) suggest that the floor on which the desk or table stands should be solid concrete to prevent interference from the movement of people in the room. The researcher had the greatest success when the equipment was mounted on an extremely heavy oak table, standing on a concrete floor.

3.6.4.4

Geometry of the arrangement of the equipment

Experience from the pilot study taught the researcher that participants seated in chairs fitted with castors can lead to participants moving around excessively during data collection sessions. Therefore, participants were seated in comfortable chairs with no casters to reduce the movement of participants during data collection. The monitor was positioned 650mm from the participant and adjusted so that the centre of the screen was level with the participants’ eye line. The screen on the 25’ monitor measured 532mm wide and 300mm high, with a resolution of 1920 x 1080 pixels. Figure 7 illustrates the set up.

95

Figure 7: Arrangement of eye tracking equipment for data collection sessions.