PURPOSE: Motion direction learning (MDL) showed less location specificity when trained with a staircase method (Zhang & Li, 2010) than with a single-level method of constant stimuli (MCS) (Ball & Sekuler, 1982). Learning also transfers to an untrained direction with a TPE (training-plus-exposure) procedure when staircases are used (Zhang & Yang, 2014), but the transfer fails with the single-level MCS (Liang et al., 2015). We suspect that training with a single-level MCS, which uses a pair of fixed motion stimuli with a single-level direction difference, may allow learning specific local cues for direction judgments.
METHODS: MDL were trained with the single-level MCS using a pair of moving dot patterns. To disturb the potential local cues in training, we slightly jittered the directions of the stimulus pair every presentation within a small range while keeping the dificulty constant.
RESULTS: When trained with fixed stimuli (?dir = 10°), peripheral MDL transferred little to contralateral and diagonal quadrants with transfer indices (TI) of 0.27 ± 0.06 and 0.25 ± 0.09 respectively. However, when the directions of the stimulus pair were slightly jittered within a range of ± 4°,  the transfer significantly increased (TI = 0.65 ± 0.10 & 0.63 ± 0.13). Similarly, when trained with fixed stimuli pair, fovea MDL (?dir = 3°)   failed to tranfer to an orthogo-nal direction even with TPE paradigm(TI = 0.25 ± 0.16), again if the stimulus directions were jittered within ± 2°, significantly more-transfer was enabled by TPE training(TI = 0.82 ± 0.24).
CONCLUSIONS: These results are consistent with the local-cue learning hypothesis regarding MDL with the single-level MCS, which is responsible for part of the specificity in motion direction learning literature. They also demonstrate the importance of using appropriate psychophysical methods in training to reduce specificity in perceptual learning.