Abstract
Performance has been shown to improve in texture discrimination tasks 6–8 hours after training in waking state (1) or following at least an hour day-time nap (2). However, the increase is the largest after a good night's sleep of at least 6 hours (3, 4). Here we studied the effect of time between testing sessions on the performance of normal human observers in a contour integration (CI) task. The CI task involves spatial integration at a longer range than in texture discrimination tasks.
Stimuli were composed of a background of randomly oriented Gabor patches, and a closed contour defined by Gabors of the same parameters (5.0 c/deg carrier frequency, 90% contrast). In a 2AFC paradigm, observers decided whether the contour, forming a smooth egg-shape, was pointing to the right or to the left. Angular difference between the path of the contour and contour elements was varied between 0–28° in 7 levels of increasing difficulty. Five 30 minutes sessions were completed within a day in the one-day group of observers, and one session was completed daily on five consecutive days in the five-day group.
The overall performance of the one-day group did not increase by the end of the experiment, although there was a slight but statistically not significant progress in CI by the third session. There was no time-of-day effect in the five-day group. The five-day group showed significant improvement between the 1st–4th and the 1st–5th sessions.
Our results are consistent with earlier studies of perceptual learning, and confirm a slow, sleep-dependent consolidation period in a task involving long-range spatial interactions.
(1)
KarniA.SagiD.Proc Natl Acad Sci USA 88, 4966–4970, 1991
(2)
MednickS. Vision Sciences Society, First Annual Meeting, 2001
(3)
StickgoldS. J Cogn Neurosci. 12(2), 246–254, 2000
(4)
GaisS. Nat Neurosci. 3(12), 1335–1339, 2000.
Funded by NSF BCS/CRI 0126151