Analyzing Infant Eye-Tracking Data

Most eye-tracking software provide raw gaze data, with the following variables that are critical for the present analyses: (1) x- and y-coordinates for the point of gaze on the screen (separately for each eye), sampled at the specified temporal resolution (60–300 Hz in most eyetrackers used with infants), (2) time stamps for each data sample (e.g.,“Tobii Eye Tracking or “TETTime” provides the time stamps at microsecond accuracy), (3) information about the “validity” indicating the reliability of tracking at each time point (e.g., Tobii TX300 uses codes 0–4, with codes 0 or 1 typically considered to indicate technically reliable gaze tracking), and (4) additional time stamps to provide exact synchronization between eye tracking and stimulus presentation (e.g., a column specifying the stimulus that is currently on screen). The x-coordinates of the gaze location for one overlap SRT trial of a 7-month-old participant are shown in Fig. 2 (the y-coordinates were omitted from the visualization because these tend to remain relatively stable across time in paradigms in which the first and the second stimuli are aligned on the vertical axis). The visualization illustrates two common characteristics of eye-tracking data collected from infants (Wass et al., 2013 (link), 2014 ). First, the raw data includes occasional periods of missing or unreliable data (shows as gaps in the thick red line at the y = 0). Second, the point of gaze undergoes constant fluctuation at periods of fixation (a problem known as low precision of eye tracking). The visualization further shows that the x-coordinates show an abrupt change at the time of the saccade.Fig. 2

X-coordinates of gaze location as a function of time for one trial of a 7-month-old infant. The data were recorded in a paradigm involving a central stimulus (a picture of a face or a facelike pattern) and a lateral stimulus (a geometric shape). The lateral stimulus was presented at 1,000 ms. Raw values for the point of gaze are shown by the narrow green line, and interpolated and median-filtered values by the thick blue line. Saccade is indicated by an abrupt change in the x-coordinates ~1,700 ms from the start and is measured as the last sample before the point of gaze leaves the area of the first stimulus (indicated by an open circle)