| Reference | Downloads | Github

Eye tracking data is shifted

Dear all,

I programmed an experiment based on Iohub selectTracker demo.
Most of the data look correct, however, for some runs and some trials, the eye tracking data is shifted from where it should be. The picture shows the wrong data in accumulated eye-tracking data for many trials.

for some runs the data looks perfect, the problem is randomly occurring.
Anyone has an idea why this is happening?


My guess is it’s likely to be the eye-tracker rather than psychopy. That’s the kind of error you get if an eye-tracker loses one eye while doing binocular tracking or misplaces a corneal reflection onto glasses or something like that. For the data to be off like that on psychopy’s side would be almost impossible, psychopy just gets very basic x, y information and does no transformations on it that I know of.

If you can, try to record a run of this on both psychopy and on the eye-tracker at the same time, and see if those mystery points also show up in the eye-tracker’s raw data. My bet is that they will.

1 Like

I’ll play Devil’s advocate and say that it looks like a simple linear geometrical error to me. Is it possible that the values for the screen distance or physical size can be altered across runs when specifying the monitor configuration to ioHub?

e.g. Check in the HDF5 file hierarchy like so:

Data Collection -> Experiment -> Message Event and examine that table for messages about the screen dimensions and the calculated pixels per degree. Do they vary?