September 2011
Volume 11, Issue 11
Vision Sciences Society Annual Meeting Abstract  |   September 2011
Dissociable Perceptual Learning Mechanisms Revealed by Diffusion-Model Analysis
Author Affiliations
  • Alexander A. Petrov
    Ohio State University, USA
  • Nicholas M. Van Horn
    Ohio State University, USA
  • Roger Ratcliff
    Ohio State University, USA
Journal of Vision September 2011, Vol.11, 993. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Alexander A. Petrov, Nicholas M. Van Horn, Roger Ratcliff; Dissociable Perceptual Learning Mechanisms Revealed by Diffusion-Model Analysis. Journal of Vision 2011;11(11):993.

      Download citation file:

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

  • Supplements

Performance on perceptual tasks improves with practice. Most theories address only accuracy (or, conversely, threshold) data and tacitly assume that perceptual learning is a monolithic phenomenon. The response times (RTs) provide a wealth of additional data that can be used to probe the mechanisms of perceptual learning. The current study uses the diffusion model (DM, Ratcliff, 1978; Ratcliff & McKoon, 2008) to convert error rates and RT distribution statistics into estimated parameters of various processing components. DM characterizes the process of making simple two-choice decisions. Among its many advantages is the ability to account for speed-accuracy tradeoffs and to estimate the decision and nondecision contributions to the total RT.

Method: We measured the stimulus specificity of perceptual learning of motion-direction discrimination. The stimuli were moving filtered-noise textures presented for 400 ms. 27 observers trained to discriminate small deviations from a fixed reference direction for 4 sessions with feedback, accuracy-contingent bonuses, and “slow down” messages. Session 5 tested whether the learning effects transferred to the orthogonal direction.

Results and Discussion: The d' increased by 55% on average and the mean RT decreased by 27% by the end of training. Specificity indices were SI = 0.60 ± 0.10 for d′ and 0.37 ± 0.08 for mean RT (group-level data ±80% bootstrap CIs). DM achieved good fits to the RT distributions for each individual in each block. The learning curves of the DM parameters identified two distinct learning mechanisms with markedly different specificities. A stimulus-specific (SI = 0.68 ± 0.09) increase in the drift-rate parameter indicates improved sensory input to the decision process. A stimulus-general (SI = 0.00 ± 0.08) decrease in the nondecision-time variability parameter suggests improved timing of the decision-process onset relative to stimulus onset (which was preceded by a beep). The traditional d′ analysis misses the latter effect but the diffusion-model analysis identifies it in the RT data.


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.