The neural signature of numerosity by separating numerical and continuous magnitude extraction in visual cortex with frequency-tagged EEG

Significance Approximating large quantities has been identified as a building block of mathematical cognition. Despite its importance from both a fundamental and an educational perspective, the mechanism behind this ability is still not fully understood. The remaining gap comes from the fact that a set is not only characterized by its number of objects but additionally by nonnumerical features naturally correlating with numerosity (e.g., a more numerous set occupies a larger surface). The current frequency-tagging EEG paradigm isolates specific responses to number and to nonnumerical dimensions. We demonstrate that when numerosity and the other dimensions are totally decorrelated in stimulus sequences, changes of each are automatically discriminated in early visual cortex, highlighting the status of numerosity as a primary visual feature. The ability to handle approximate quantities, or number sense, has been recurrently linked to mathematical skills, although the nature of the mechanism allowing to extract numerical information (i.e., numerosity) from environmental stimuli is still debated. A set of objects is indeed not only characterized by its numerosity but also by other features, such as the summed area occupied by the elements, which often covary with numerosity. These intrinsic relations between numerosity and nonnumerical magnitudes led some authors to argue that numerosity is not independently processed but extracted through a weighting of continuous magnitudes. This view cannot be properly tested through classic behavioral and neuroimaging approaches due to these intrinsic correlations. The current study used a frequency-tagging EEG approach to separately measure responses to numerosity as well as to continuous magnitudes. We recorded occipital responses to numerosity, total area, and convex hull changes but not to density and dot size. We additionally applied a model predicting primary visual cortex responses to the set of stimuli. The model output was closely aligned with our electrophysiological data, since it predicted discrimination only for numerosity, total area, and convex hull. Our findings thus demonstrate that numerosity can be independently processed at an early stage in the visual cortex, even when completely isolated from other magnitude changes. The similar implicit discrimination for numerosity as for some continuous magnitudes, which correspond to basic visual percepts, shows that both can be extracted independently, hence substantiating the nature of numerosity as a primary feature of the visual scene.