09-07-2020, 05:09 AM
When people talk about the accuracy of eye trackers they are typically referring to "spatial accuracy" - which can be defined as the difference between the gaze position measured by the eye tracker and the true (actual) gaze position. The difficulty here is that there is no way of knowing where the true gaze position is - because we have no "perfect" technique for measuring eye position. So accuracy is generally estimated as the difference between gaze position measured by the eye tracker, and the position that the participant "should" be looking at (e.g. the position of a target).
To estimate the accuracy simply perform a Validation from the Camera Setup screen (after first performing a calibration); the validation will report the level of error (i.e. the difference between the target location and the gaze position measured by the eye tracker based on the calibration model) for each target presented, as well as the average and maximum error values. Typically for adults with normal vision, average error values range from 0.25 - 0.50 degrees. We generally recommend aiming for average error values of < 0.5 degrees and a maximum error of < 1.0 degrees. See What determines whether a validation is "GOOD", "FAIR" or "POOR"? for how the validation criteria.
For further information about spatial accuracy and how to measure / maximize it, please see the relevant Set Up and Usage Training Videos:
To estimate the accuracy simply perform a Validation from the Camera Setup screen (after first performing a calibration); the validation will report the level of error (i.e. the difference between the target location and the gaze position measured by the eye tracker based on the calibration model) for each target presented, as well as the average and maximum error values. Typically for adults with normal vision, average error values range from 0.25 - 0.50 degrees. We generally recommend aiming for average error values of < 0.5 degrees and a maximum error of < 1.0 degrees. See What determines whether a validation is "GOOD", "FAIR" or "POOR"? for how the validation criteria.
For further information about spatial accuracy and how to measure / maximize it, please see the relevant Set Up and Usage Training Videos: