In the “Where’s Wally” game, we search a chaotic scene through eye movements to identify Wally. However, upon revisiting the same scene, we can quickly recall Wally’s location and direct our gaze accordingly. This example illustrates how visuospatial perception and memory are dynamically shaped by eye movements. Yet, it remains unclear whether the identity of an object (“what”) and its spatial location (“where”) are stored in memory as a conjunctive code or as separate representations.
To address this question, we leveraged the rare opportunity to record single-unit activity from patients with epilepsy to investigate the neural codes of identity and space in the human medial temporal lobe (MTL). During each mini-block, patients encoded a four-item visuospatial memory by fixating on four oblique locations, each presenting a different image in a gaze-contingent manner. They then re-explored the layout in the same gaze-contingent manner and were tasked with judging visuospatial match/mismatch, constituting the retrieval phase. We identified two distinct populations of cells: identity-selective units and location-selective units. These populations formed an orthogonal neural code for identity and space in the human MTL. Identity selectivity was most prominent in the amygdala and hippocampus, whereas location selectivity was confined to the parahippocampal cortex. Furthermore, the firing of both identity- and location-selective cells were expedited during retrieval, providing support for predictive processing models of MTL function.
These findings provide evidence for segregated neural coding of identity and space in the human MTL, offering new insights into how active vision supports visuospatial memory.