Many previous studies of orientation discrimination learning often adopted real stimuli as experimental material. Until now, the specificity and generalization of learning is one of the focuses in debate. In the present study, we adopted an orientation discrimination task, using both real lines and imagery lines as experimental stimuli, and examined the characteristics of two kinds of learning and their relationship. In the first experiment, we used “real lines in the upper-left visual field” as training stimuli, “real lines in the lower-right visual field”, “imagery lines in the upper-left visual field” and “imagery lines in the lower-right visual field” as untrained conditions. We investigated whether the learning effect of real line could transfer to untrained location and imagery lines. In the second experiment, we used “imagery lines in the upper-left visual field” as training stimuli and used the other three types of stimuli as untrained conditions. We investigated whether the learning effect of imagery lines can transfer to untrained location and real lines. The results showed that, for both real lines and imagery lines, orientation discrimination training significantly decreased the subject’s thresholds. What’s more important, this learning effect could partly transfer to the untrained location and untrained stimuli type. In the third experiment, in which subjects only need to participate in the pre-test and the post-test (five days after the pre-test), we ruled out that the improvement of performance was just due to pre-test and proved that orientation discrimination learning indeed transferred to the untrained location and untrained stimuli type. In sum, our results strongly suggested that the two kinds of orientation discrimination learning might involve some non-retinotopic high brain areas and there are common neural mechanisms between them.