[Purpose] We examined whether different mechanisms are involved in the learning of coherent motion detection and motion direction discrimination. [Methods] The experiment consisted of two test and one training phases. In the pre-training test, all subjects were shown a random-dot cinematogram and were asked to judge whether the stimulus contained coherently moving dots or whether all dots moved randomly. Coherent motion detection thresholds were measured for each of 9 directions within 90 degrees. Subjects were divided into 2 groups for training. One group (n=6) participated in the same motion coherence detection task as in the pre-test except that only 2 of 9 directions were used (difference of the 2 directions = 20 deg) and the signal/noise ratio was fixed at 18%. The other group of subjects (n=6) participated in the motion direction discrimination task, where they were asked to judge whether the global motion flow directions in two successively presented displays matched. The same stimuli (i.e., the same directions and the same S/N ratio) were presented an equal number of times as in the detection task. After training, all the subjects participated in the post-training test, with the same procedure as in the pre-training test. [Results] In the post-training test, the subjects trained on coherent motion detection showed significantly greater improvement for trained directions than for untrained directions. However, to our surprise, in the post-training test the group trained on motion direction discrimination showed significantly less improvement for trained directions than for untrained directions. [Conclusions] Despite the fact that both groups viewed the same stimuli equally often, they showed opposite patterns of improvement in their detection of presented motion directions. Thus, we conclude that the learning of motion detection may involve different mechanisms than the learning of motion direction discrimination.

Supported by NSF (BCS-9905194) to TW.