Now Shipping! – The EyeLink 3; combined head and eye tracking at up to 1000 Hz.

SR Support Forum   ›   Learning Resources   ›   FAQs   ›   EyeLink Hardware FAQs   ›   FAQ: Which calibration model should I use?
FAQ: Which calibration model should I use?
Sam
Research Support Specialist
*******
Joined: Aug 2020
Posts: 1,015

#1
09-07-2020, 08:34 AM
The best calibration model is largely dependent on the experimental task, the participant population that is being tracked, and the eye-tracking mode (head-fixed vs head-free Remote mode).The EyeLink systems offer a range of calibration models - horizontal only H3, and horizontal / vertical HV3, HV5, HV9 and HV13 models (with the numeral referring to the number of calibration targets). As a general rule, using more calibration points will result in better spatial accuracy across the entire screen.
   


In general:
  • If the experiment is presenting text, images, and / or video stimuli that occupy most or all of the screen, then the HV9 or HV13 calibration models would be most appropriate.
  • If using Remote mode (target sticker), for best spatial accuracy we recommend the HV9 or HV13 calibration for full screen stimuli.
  • If the experiment is examining horizontal and vertical saccades or smooth pursuit, or presenting relatively small central stimuli (such as faces), or images arranged horizontally / vertically (as in a preferential looking or dot-probe type task) then the HV5 calibration model would be appropriate.
  • For single-line text reading experiments, then a simple horizontal only H3 calibration is often optimal. The vertical position of the H3 targets can be configured, so that they are presented at the same vertical position as the sentence. The H3 calibration will only record the "X" (horizontal) position of the eye. Any changes in gaze position in the "Y" (vertical) axis will not be recorded. 
  • For infants or other participants who may find it difficult to complete an HV9 or HV13 calibration, the HV3 or HV5 models would be more appropriate. 

If you are still not sure which model to use, here are some additional resources: