This dissertation focuses on the mechanisms and implications of perceptual learning of binocular interactions. Perceptual learning is an important means of adapting to the changing environment, demonstrating the possibility of neural plasticity in adults and providing a powerful approach to investigate dynamic processes in the mature perceptual system. Most studies on perceptual learning have focused on learning mechanisms that target excitatory circuits. However, we recognize that the inhibitory circuits also play a critical role in cortical plasticity, as shown by growing evidence from neurophysiological studies, and that the inhibitory connection is more dynamic than the excitatory connection in adult visual cortex. Thus, our goal is to design a psychophysical method that exploits the contribution of the inhibitory circuits to perceptual learning. This in turn helps us to implement more efficient learning paradigms for visual training. Our study capitalizes on properties of the binocular visual system, a good system for exploring both excitatory and inhibitory mechanisms. We first measured local Sensory Eye Dominance (SED) and showed that excessive SED can impede stereopsis ability. To reduce SED, a typical perceptual training paradigm (Push-only protocol) would only stimulate the weak eye to target the excitatory network. In contrast, we designed a novel Push-Pull training protocol to target both the excitatory and inhibitory networks. By presenting binocular rivalry stimuli to both eyes, the push-pull protocol can excite the visual pathway of the weak eye (push), while inhibiting the visual pathway of the strong eye (pull). We found that the push-pull training protocol, mainly affecting the early visual processes, is more effective than the push-only protocol in reducing SED and enhancing stereoacuity, even beyond the focus of top-down attention through a stimulus-driven mechanism. We further demonstrated that the perceptual learning induced by the push-pull protocol involves both feature-based and boundary-based processes, and that the learning effect can be generalized to other stimulus dimensions within early feature channels. Therefore, our psychophysical study demonstrates the important role of inhibitory synaptic circuits in neural plasticity of the adult brain, and that our push-pull training protocol can be a more effective clinical training paradigm to treat amblyopia.