September 2011
Volume 11, Issue 11
Vision Sciences Society Annual Meeting Abstract  |   September 2011
Participants know best: Influence of calibration method on accuracy
Author Affiliations
  • Kenneth Holmqvist
    Humanities Laboratory, Lund University, Sweden
  • Marcus Nyström
    Humanities Laboratory, Lund University, Sweden
  • Richard Andersson
    Humanities Laboratory, Lund University, Sweden
Journal of Vision September 2011, Vol.11, 503. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kenneth Holmqvist, Marcus Nyström, Richard Andersson; Participants know best: Influence of calibration method on accuracy. Journal of Vision 2011;11(11):503.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Calibration of an eye-tracker is associated with several challenges, both theoretical and practical. Theoretical challenges include finding a good mathematical model for the eye as well as a mapping function from eye- to gaze-positions. An important practical aspect is what Goldberg and Wichansky (2003) distinguish as calibration controlled by the experimenter, the system, and the participant. In the first case, the experimenter accepts a calibration target when she has the impression that the participant is fixating the target, additionally verifying that the participant's eye is stable in the video feed of the eye image. In the second, the system decides what raw data samples should be used for calibration. Participant-controlled calibration is when the participant clicks when looking at the target to calibrate. The current trend is that increasingly more control over the calibration procedure goes to the system; for instance, three of the largest eye-tracker manufacturers all use system controlled calibration as default [EyeLink manual, Tobii manual, and SMI manual]. We recorded 149 participants binocularly on the SMI HiSpeed at 500 Hz, in the three conditions: Automatic, operator, and participant controlled calibration. Accuracy was measured directly after calibration, and again after 15 minutes of participant reading a text. Points for accuracy measurements were identical to calibration points. Accuracy was defined as the minimal distance from a fixation (detected by the algorithm by Engbert and Kliegl (2003)) to the current point that participants were instructed to look at. The difference in accuracy between calibration conditions were tested using Kolmogorov-Smirnov and the Kruskal-Wallis tests. In both cases, participant controlled calibration showed to give significantly more accurate data while system-controlled calibration gives the poorest accuracy. The result will be discussed in relation to data recording practices, participant idiosyncracies and the types of data analysis that follow recordings.


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.