Neural Encoding Geometry of Human Voluntary Visual Selection
Year: 2026
Type: preprint
Abstract: Selecting relevant targets from competing alternatives is a core cognitive function. While human visual selection can be tracked with high spatiotemporal precision, it remains unsolved how selected targets and non-selected alternatives are neurally encoded to effectively separate them. Building on recent work revealing the neural encoding geometry underlying target selection in animals, we here examined how the representational format of selected targets and non-selected alternatives evolve over time during human voluntary selection. Initially, selected and non-selected objects showed equal spatially selective neural activity, yet the selected target was represented more strongly after 300-350 ms. Using cross-decoding, we found that selected and non-selected objects are encoded along a shared sensory axis, whereas selected targets are encoded along an additional spatially selective axis. This encoding geometry effectively separates selected targets from non-selected alternatives in neural space. Our results generalized across selection conditions: we found comparable results in instructed and free-choice selection, as well as in delayed oculomotor and covert visual selection. Together, we demonstrate that the human brain selectively encodes targets, but not distractors, along an additional axis, revealing a neural mechanism through which voluntary selection prioritizes behaviorally relevant information.
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7225729646
Cites:
Cited by:
Related to: