Learning to be fast: Gain accuracy with speed

Anna Sterkin, Oren Yehezkel, Uri Polat

Research output: Contribution to journalArticlepeer-review

20 Scopus citations


Our recent neurophysiological findings provided evidence for collinear facilitation in detecting low-contrast Gabor patches (GPs) and for the abolishment of these collinear interactions by backward masking (BM) (. Sterkin et al., 2008; Sterkin, Yehezkel, Bonneh, et al., 2009). We suggested that the suppression induced by the BM eliminates the collinear facilitation. Moreover, our recent study showed that training on a BM task overcomes the BM effect, hence, improves the processing speed (. Polat, 2009). Here we applied training on detecting a target that is followed by BM in order to study whether reinforced facilitatory interactions can overcome the suppressive effects induced by BM. Event-Related Potentials (ERPs) were recorded before and after ten training sessions performed on different days. Low-contrast, foveal target GP was simultaneously flanked by two collinear high-contrast GPs. In the BM task, another identical mask was presented at different time-intervals (ISIs). Before training, BM induced suppression of target detection, at the ISI of 50. ms, in agreement with earlier behavioral findings. This ISI coincides with the active time-window of lateral interactions. After training, our results show a remarkable improvement in all behavioral measurements, including percent of correct responses, sensitivity (d'), reaction time (RT) and the decision criterion for this ISI. The ERP results show that before training,. BM attenuated the physiological markers of facilitation at the same ISI of 50. ms, measured as the amplitude of the negative N1 peak (latency of 260. ms). After the training, the sensory representation, reflected by P1 peak, has not changed, consistent with the unchanged physical parameters of the stimulus. Instead, the shorter latency (by 20. ms, latency of 240. ms) and the increased amplitude of N1 represent the development of faster and stronger facilitatory lateral interactions between the target and the collinear flankers. Thus, previously effective backward masking became ineffective in disrupting the collinear facilitation. Moreover, a high-amplitude late peak (P4, latency of 610-630. ms) was not affected by training, however its high correlation with RT (95%) before training was significantly decreased (to 76%), consistent with a lower-level representation of a trained skill. We suggest that perceptual learning that strengthens collinear facilitation results in a faster processing speed.

Original languageEnglish
Pages (from-to)115-124
Number of pages10
JournalVision Research
StatePublished - 15 May 2012
Externally publishedYes

Bibliographical note

Funding Information:
This research was supported by grants from the National Institute for Psychobiology in Israel, funded by the Charles E. Smith Family and the Israel Science Foundation (U.P. and A.S.), the Israel Science Foundation (U.P.), and the Nela Horovitz Foundation, TAU (U.P.).


  • Backward masking
  • ERP
  • Facilitation
  • Lateral interactions
  • Perceptual learning
  • Processing speed


Dive into the research topics of 'Learning to be fast: Gain accuracy with speed'. Together they form a unique fingerprint.

Cite this