April 2014
Volume 55, Issue 13
ARVO Annual Meeting Abstract  |   April 2014
Optimal stimulus placement for psychometric function estimation
Author Affiliations & Notes
  • Daniel Robert Coates
    Vision Science Graduate Group, UC Berkeley, Berkeley, CA
  • Susana T L Chung
    Vision Science Graduate Group, UC Berkeley, Berkeley, CA
  • Footnotes
    Commercial Relationships Daniel Coates, None; Susana Chung, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science April 2014, Vol.55, 3533. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Daniel Robert Coates, Susana T L Chung; Optimal stimulus placement for psychometric function estimation. Invest. Ophthalmol. Vis. Sci. 2014;55(13):3533.

      Download citation file:

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

  • Supplements

Purpose: Psychometric functions (PFs) are used to characterize many aspects of visual function, including acuity, reading performance, and stimulus detection. Accurate and efficient measurement of psychometric functions is crucial, and becomes particularly challenging in situations where testing duration is limited or subjects exhibit greater variability than normal. While adaptive procedures such as staircase methods or QUEST exist, they can be taxing on non-expert subjects, are overly sensitive to lapse errors, and can be difficult to implement. We evaluated a simple hybrid scheme that positions stimuli at arbitrary levels of a continuously-estimated PF.

Methods: We tested this adaptive method using Monte Carlo simulations and experiments with low-visions patients. We measured psychometric functions of percent correct vs. word duration in 3 patients with central field loss using a rapid serial visual presentation paradigm. Trials comprised 8-12 words, scored as percent correct. We tested several stimulus placement paradigms in blocks, including standard constant stimuli (6 repetitions at 5 stimulus levels) and two variations of the adaptive method, one with stimuli close to subject threshold, and one with more widely distributed stimulus levels. Testing order was randomized.

Results: The simulations confirmed previous studies showing that blocks repeating 2-5 carefully placed stimulus locations were effective in estimating known PFs. Within 30 trials, parameter values could be estimated to within 5% accuracy on 95% of the simulations. Even when unknown PFs were estimated online, 30-40 trials were sufficient across the parameter combinations tested. For the low vision subjects, the methods did not differ significantly, using the aggregation of all trials as the ground truth. Blocks with stimuli remaining near subject threshold did introduce a slight negative bias, which agreed with informal observations about the difficulty of those blocks.

Conclusions: This method provides a compromise between the flexibility and optimality of existing adaptive procedures, and the simplicity, robustness, and subject comfort of the method of constant stimuli. Experimental sessions needn't converge to repeatedly test near a subject's threshold in order to yield good adaptive estimates. Patient results also showed that parameter estimation was more accurate when trials were spread modestly across the PF, which differs from the statistical prediction.

Keywords: 640 pattern vision • 641 perception  

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.