Abstract
The classic texture discrimination task (TDT) (Karni & Sagi, 1991) is used in numerous psychophysical and brain imaging studies on perceptual learning. An observer reports the global orientation (H/V) of a 3-oblique-bar array imbedded in a background of horizontal or vertical bars, which was flashed briefly and masked by a field of randomly oriented Vs at various SOAs. Practice shortens the SOA threshold, but the learning is reportedly specific to the trained location and background orientation. Here we show that a majority of TDT learning is relevant to temporal resolution, not texture discrimination. Observers first learned (7–8 1.5-hr daily sessions) to discriminate the orientation of a letter “C” (flipped/no-flipped) on a clear screen which was flashed (13-ms) at the same location of the 3-bar target in a TDT task and then masked by a field of circles. This task only shared temporal and location properties with TDT. They then practiced TDT (4–5 sessions). The TDT improvement after initial “C” orientation learning accounted for >80% the overall TDT learning, and later TDT training accounted for the remaining <20% learning. Jittering the stimulus onset after training did not change the performance, indicating that what is improved is not the timing, but temporal resolution. In addition, TDT learning was found to be specific to the trained location in new observers, but further Vernier training in the diagonal quadrant (200_msec on a clear screen) enabled significant transfer of TDT learning, consistent with our previous double-training results (Xiao et al., 2008). Moreover, temporal learning as revealed by a similar “C” orientation task accounted for about half the learning of the famous feature detection task (Ahissar & Hochstein, 1997). These results suggest that temporal learning accounts for most TDT learning, and that the location specificity cannot be used to support a V1 site of TDT learning.
Supported by Natural Science Foundation of China grants (30725018).