Fig 1.
Example of loss of pupil and corneal reflections.
A participant whose corneal and pupil reflection were initially successfully tracked (A,B), then subsequently obstructed by the lower eyelid (C,D). The Eyelink system tracks head distance relative to the camera using a small target sticker placed on the infant’s forehead. If the infant shifts position or turns away from the camera, tracking of the target sticker can be lost. The eye-tracker view of the participant’s eye image (A) demonstrates that pupil reflection (dark blue) and corneal reflection (cyan) were initially tracked by the camera. Status indicators (B) confirm that the camera was tracking the target sticker in addition to the pupil and corneal reflections. Relatively small head movements resulted in loss of pupil and corneal reflection due to obstruction by the lower eyelid (C). Status indicators (D) confirmed loss of pupil and corneal reflection, but no obstruction of the target sticker, verifying that loss of corneal reflection is not due to significant head movement or the infant looking away from the camera.
Fig 2.
Static images of video stimuli.
Fig 3.
Modifications to standard eye-tracking setup.
Fig 4.
Participant gaze data from two different trials (A,B and C,D) are shown. Gaze positions in screen coordinates are shown in A and C. Fixation durations are indicated numerically next to Eyelink identified parsed fixations which are shown as light blue circles in B and D.
Fig 5.
Participant enrollment and data analysis flow chart.
A total of 54 mother-infant dyads were enrolled. *An additional 3 dyads were excluded from analysis due to inaccurate pre-enrollment screening. Eye-tracking was not attempted for 3 participants due to infant temperament and difficulty establishing a reliable corneal reflection. For the 51 participants for whom eye-tracking was attempted, data quality metrics are reported. Data from 32 of the 51 participants met or surpassed the 70% threshold for valid data across all 6 trials, and are reported in the analysis of oculomotor dynamics. ** For 2 of the 32 participants whose data met quality thresholds, their data was excluded from analysis of saccade amplitude and saccade velocity due to an error in screen resolution affecting those specific calculations, but was retained for all other analyses of saccade and fixation metrics.
Table 1.
Demographic information for included and excluded participants.
Fig 6.
Validation quality and monocular and binocular success rates.
Data is shown for participants for whom (A) eye-tracking was attempted (n = 51) and (B) whose data met quality thresholds for analysis of oculomotor metrics (n = 32). Although monocular data was used for analyses, binocular recording was attempted for all participants to maximize the probability of collecting high quality data from at least one eye. If calibration was repeated, only the final attempt is reported here. Infants for whom calibration was aborted or skipped after multiple unsuccessful attempts are represented by the “None” category.
Fig 7.
Histograms of calibration and validation error.
Average error and maximum error during the calibration and validation procedure for the participants for whom eye-tracking was attempted and calibration was obtained (n = 51) (A, B) and participants whose data met data quality thresholds for subsequent analysis of oculomotor metrics (n = 32) (C, D). Error values indicate the amount of offset between the center of the calibration graphic and the reported x,y coordinates of the gaze sample. If monocular data was collected, values are reported for the recorded eye. If binocular data was collected, values are reported for the eye with higher calibration quality.
Table 2.
Qualitative output results from calibration and validation procedures.
Fig 8.
Raw data are shown from different participants for whom decreasing levels (from top to bottom) of valid data were collected. X and Y eye positions are shown in blue and orange, respectively. Horizontal plateaus are periods where x and y coordinates of the eye remain relatively steady, indicating fixations. Vertical lines are periods where the eye position is changing rapidly, indicating saccades. Gaps in the traces represent periods of data loss. As the threshold for valid data decreases, there are more frequent gaps in the data and increased noise in the x,y traces.
Fig 9.
Histograms of different valid data thresholds.
The number of participants and number of trials that met or exceeded cutoffs for valid data at 50 (A), 60 (B), 70 (C), 80 (D), 90 (E) and 99 (F) percent thresholds are shown.
Fig 10.
Percent of valid data in thresholded and unthresholded trials.
The percentage of valid data across thresholded (orange) and unthresholded (purple) trials is shown. The unthresholded trials include data from all 50 participants who completed the eye-tracking task. Thresholded trials include the 32 participants whose data met or exceed 70% valid data across all 6 trials. In this population of young infants, there was no statistically significant loss of valid data across time in either the thresholded or unthresholded trials. Data presented as mean +/- 1 standard error of the mean (SEM).
Fig 11.
Fixation metrics across time for thresholded data.
There were no significant differences in fixation duration (A) or count (B) across the 6 trials of data collected. Data presented as mean +/- 1 SEM.
Fig 12.
Saccade metrics across time for thresholded data.
Saccades did not significantly change in duration (A), count (B), mean velocity (C), or amplitude (D) across time. Data presented as mean +/- 1 SEM.