Ionic–electronic coupling in a lattice strongly influences the memory and learning process by synaptic weight update in electrochemical synaptic transistors. In particular, anisotropic crystal symmetry offers a highly anisotropic diffusion process, which leads to facilitated ion migration and efficient coupling in synaptic devices. Here, we report all-solid-state VO2 synaptic transistors in which the proton diffusion under gate bias can be tuned by utilizing different crystal facets in anisotropic VO2 channels. Synaptic weight update (i.e., excitatory post-synaptic current) by a proton (H+) in the VO2 channel was sensitively tuned depending on the empty tunnel alignment of VO2 layers. By emulating synaptic functions using diffusion-pathway-controlled transistors, the alignment of a facile ionic pathway with gating direction increases the retention of H+ in VO2 lattices by locating H+ into the deep regions from the interfaces, and thus strengthens long-term memory in artificial synaptic devices. These results demonstrate that the control of field-driven ionic redistribution guided by crystal anisotropy provides an opportunity to manipulate the learning and forgetting behavior in artificial synaptic devices.