Slow/Imprecise Calibrations with Eyelink

hvw014

Hello,

We are currently implementing our first eyetracking study with Open Sesame. Our eyetracking setup is the Eyelink 1000 (SR Research), running on Mac OSX 10.12. We have noticed that calibration with Open Sesame is giving much higher values during validation then when we are running Experiment Builder studies and its difficult to get a calibration below the threshold set by the program. Additionally, the eyetracking computer seems to be lagging (taking a longer time for fixations to occur on the calibration points). We previously modified the Experiment Builder screen width and recorded it in the Eyetracker to fix a similar problem in Experiment Builder and are wondering if anyone knows how to do this in Open Sesame.

Thank you,

Hannah

sebastiaan

Hi Hannah,

The calibration procedure is always handled by the EyeLink PC, regardless of whether you're doing this through OpenSesame, Experiment Builder, or some other software. So in that sense, the software cannot affect the quality of the calibration.

However, the software can control how the calibration dots are shown on the display. So I'm guessing that's where the issue comes from. Are you using the same display resolution in OpenSesame and Experiment Builder (i.e. are the calibration dots in similar places)?

Cheers!

Sebastiaan

hvw014

Hi Sebastiaan,

Thank you for your answer to my question. I believe that the display resolution is the issue that we are having. During validation with the eye tracker it appears on the Eyelink computer that the participant is looking next to the dot when they are actually focused on it.

We tried changing the Eyelink host computer to the same resolution as the display computer but the issue still persisted. Our monitors are also different sizes, the host monitor is square and display is rectangular. Would this potentially be a cause for our issue? Please let us know if you have any suggestions on how to troubleshoot.

Thank you,

Hannah

sebastiaan

Hi Hannah,

The resolution of the display computer is the only thing that matters (the host display doesn't), and in principle there's no 'correct' resolution. However, if you have a really big screen and the calibration dots cover the entire screen, then you might be trying to calibrate at angles at which the EyeLink is not well able to pick up the eye anymore.

So that's one thing that comes to mind. And a solution might be to set the experiment resolution to something smaller than the display resolution, and to use a back-end (psycho or xpyriment) that doesn't actually change the display resolution (but uses a sort of virtual screen in the display center).

Cheers!

Sebastiaan