The effect of transcranial random noise stimulation (tRNS) on visual perceptual learning has only been investigated during early training sessions, and the influence of tRNS on later performance is unclear. We engaged participants first in 8 days of training to reach a plateau (Stage 1) and then in continued training for 3 days (Stage 2). In the first group, tRNS was applied to visual areas of the brain while participants were trained on a coherent motion direction identification task over a period of 11 days (Stage 1 + Stage 2). In the second group, participants completed an 8-day training period without any stimulation to reach a plateau (Stage 1); after that, they continued training for 3 days, during which tRNS was administered (Stage 2). In the third group, participants completed the same training as the second group, but during Stage 2, tRNS was replaced by sham stimulation. Coherence thresholds were measured three times: before training, after Stage 1, and after Stage 2. Compared with sham simulation, tRNS did not improve coherence thresholds during the plateau period. The comparison of learning curves between the first and third groups showed that tRNS decreased thresholds in the early training stage, but it failed to improve plateau thresholds. For the second and third groups, tRNS did not further enhance plateau thresholds after the continued 3-day training period. In conclusion, tRNS facilitated visual perceptual learning in the early stage, but its effect disappeared as the training continued.