Fig. 1: Neural population responses to small and large numerosities.

a, b Cortical surface rendering of the right hemisphere shows a similar network of numerosity maps in both presented ranges. Preferred numerosities of cortical recording sites, estimated from responses to the small range (a) and the large range (b) for recording sites with over 30% of variance explained by the neural response model. Black lines outline the lateral borders of individual numerosity maps. The borders denoting the lowest and the highest preferred numerosities in each map are marked by white lines. An example fMRI recording site in anterior NTO shows different fMRI time courses (dots) for small (c) and large (d) numerosity ranges. Both time courses are similarly well captured by the predictions (coloured lines) of similar neural response models. Dots represent mean response amplitudes; error bars represent the standard errors over repeated measurements (n = 4). The presented numerosities are indicated at the top of the graph. e, f An example fMRI recording site in posterior NTO shows a higher preferred numerosity. This response does not reach a maximum in the small numerosity range (e). Dots represent mean response amplitudes; error bars represent the standard errors over repeated measurements (n = 4). g, h For both numerosity ranges, the anterior NTO site’s response is predicted by similar neural response models. i, j For the large numerosity range, posterior NTO site’s response is well predicted by a neural response model (j). However, this sites’ preferred numerosity is above the small numerosity range, so it could not be determined accurately (i), produces only low-amplitude responses and yields poorer model fits (e) with this range. Preferred numerosity is indicated by the highest response amplitude in the neural model, and tuning width is indicated by the full width at half maximum (FWHM). The neural response model within the presented range is shown with solid lines, outside the range with dashed lines.