Neural representations of visual categories generalise across tasks and adapt to cross-task interference

Poster Session B - Sunday, April 14, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Uma Ajmeria¹, Cal Shearer², Jan Derrfuss¹, Nicholas Myers¹; ¹University of Nottingham, ²University of Oxford

The brain’s ability to group stimuli into relevant categories is a fundamental cognitive process. The neural basis of categorisation is not completely understood, with some studies suggesting the presence of multitasking neurons (encoding for multiple category distinctions), and others suggesting highly specialised neurons (encoding for a single category). This differentiation may be due to task demand, as less multitasking has been shown when two different category distinctions are in conflict. This suggests that neurons multitask by default but become specialised for one task in the presence of cross-task conflict. Little is known about this distinction in humans as most studies have examined non-human primates. This study investigated this proposal in humans using electroencephalography (EEG). Participants learned to perform two visual categorisation tasks on stimulus sets drawn from distinct parts of a continuous three-dimensional feature space (no conflict), but were later introduced to new stimuli in both tasks, creating conflict between the two categorisation rules. Category-specific neural patterns generalised across tasks for the no-conflict condition, suggesting the presence of multitasking neurons. In contrast, neural patterns became more independent across tasks when new stimuli were introduced in both tasks (conflict condition). Unexpectedly, the previously learned stimuli continued to be represented by the same generalised patterns. Taken together, our findings suggest that reduction in multitasking may be a general principle of interference resolution in neural coding.

Topic Area: EXECUTIVE PROCESSES: Working memory