August 2012
Volume 12, Issue 9
Vision Sciences Society Annual Meeting Abstract  |   August 2012
Parameter distribitions of eye-movements based on 1,000,000 trials.
Author Affiliations
  • Aaron Johnson
    Psychology, Concordia University, Montreal.
  • John Brand
    Psychology, Concordia University, Montreal.
  • Bruno Richard
    Psychology, Concordia University, Montreal.
Journal of Vision August 2012, Vol.12, 1021. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Aaron Johnson, John Brand, Bruno Richard; Parameter distribitions of eye-movements based on 1,000,000 trials.. Journal of Vision 2012;12(9):1021.

      Download citation file:

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

  • Supplements

Eye movement metrics, such as fixation duration, frequency and saccade amplitude are often reported in terms of means per subject or condition, and are subsequently analyzed by parametric statistics. Yet, when such metrics are plotted as a frequency distribution, the resulting histogram is heavily skewed to the right, violating one of the primary rules of parametric testing; the data should be normally distributed. Here we explore how common skewed distributions are by analyzing eye movements during a number of cognitive experiments that have been conducted within the Concordia Vision Lab over the past few years. Eye position was monitored using an SR Research Eyelink 1000 with a minimum sample rate of 500Hz, with a minimum average calibration accuracy of .5 deg. (max. error of 1 deg.). Eye movement recordings from multiple tasks were collected from participants, whose ages ranged from 18 months to 78 years, using consistent definitions of fixations and saccades. Tasks included scene perception, visual search, face emotion recognition, and empathy response (all stimuli and subsequent eye movements are available to download at We find that all tasks show a similar skewed distribution in fixation duration, frequency and saccade latency and amplitude. Accordingly, we show that the median, and not the mean, is the measure of central tendency that best represents the distribution. Subsequent parametric analysis of these data can use there be conducted with median values. Finally, we show how more subtle changes in the distribution can be described using parameter fitting eye movement distributions with a Weibull curve (which shows the best goodness-of-fit measures to eye movements). We find little change in the shift parameter, but large differences in the scale parameter. Future investigations of eye movement recordings should therefore take further care in properly characterizing the data they have collected prior to their analyses.

Meeting abstract presented at VSS 2012


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.