中国科学院行为科学重点实验室知识库

Technological advancements in virtual reality challenge the human vision, especially stereopsis, a function, which characterizes how two eyes coordinate to form a unified three-dimensional (3D) representation of the external world and is found to be deficient in 30% of the normal population. Although a few previous studies have consistently found that the perceptual learning of stereopsis significantly improved stereoacuity, an underlying mechanism of stereo learning remains heavily debated. Here, we trained subjects with normal stereo vision (assessed with the FLY Stereo Acuity Test) to judge stereopsis containing three types of binocular disparity orders (i.e., zero-, first-, and second-order), aiming to systematically examine the characteristics and plasticity of stereo learning. Thirty subjects were randomly assigned to the three training groups (each for the zero-, first-, or second-order disparity separately). The disparity thresholds were measured before and after training. The disparity threshold was measured in 10 additional control subjects only at the pre- and post-training phase. Stereoscopic images were displayed through a shutter goggle, which is synchronized to a monitor. We found that the training significantly improved the zero-, first-, and second-order disparity threshold by 52.42, 36.28, and 14.9% in the zero-order training condition; 30.44, 63.74, and 21.07% in the first-order training condition; and 30.77, 25.19, and 75.12% in the second-order training condition, respectively. There was no significant improvement in the control group. Interestingly, the greatest improvements in the first- and second-order disparity threshold were found in the corresponding disparity training group; on the contrary, the improvements in the zero-order disparity threshold were comparable across all the three disparity training groups. Our findings demonstrated both general (related to the zero-order disparity) and specific improvements (related to the first- and second-order disparity) in stereo learning, suggesting that stereo training occurs at different visual processing stages and its effects might depend on the specific training sites.