Purchase this article with an account.
Jacob Paul, Tuomas ten Cate, Ben Harvey; Monotonic responses to numerosity in human primary visual cortex reflect canonical visuospatial computations. Journal of Vision 2020;20(11):472. doi: https://doi.org/10.1167/jov.20.11.472.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Humans and many animals have neurons tuned numerosity, the number of items in a display. Numerosity tuned neurons respond selectively, decreasing in response amplitude with distance in either direction from a peak preferred numerosity. How numerosity-tuned responses are derived from visual images remains unclear. Computational models suggest an initial monotonic stage where response amplitude increases with numerosity. Recent evidence suggests this monotonic stage may occur in primary visual cortex. Here we used ultra-high-field 7T fMRI and population receptive field (pRF) modelling to characterize the location and nature of these responses. During scanning, we displayed visual stimuli containing black circles whose numerosity changed between one and seven in ascending and descending order. Different stimulus configurations controlled for covarying visual features such as luminance, edge density and item size. We quantified monotonic responses with a general linear model that increased response amplitude proportionally to log(numerosity). We used pRF visual field mapping to identify visual field map borders and quantify eccentricity preferences for each voxel. In all stimulus configurations, a monotonic response model closely predicted responses in primary visual cortex (V1-V3). Explanatory power was highest in V1 and progressively decreased into extrastriate cortex. Model fit depended critically on the eccentricity preference of each voxel, with goodness-of-fit progressively decreasing as pRFs covered less of the stimulus area. Normalized stimulus luminance and contrast predict very similar responses to numerosity, and capture V1’s responses even more closely. Our findings indicate monotonic responses to numerosity are already established by V1 and inherited by subsequent processing stages. We discuss how basic neural computations in primary visual cortex (such as divisive normalization or surround-suppression) could drive such monotonic responses. These canonical operations may provide a simple, biologically plausible mechanism linking established low-level response properties of V1 with numerosity perception and higher-level numerosity-tuned responses in association cortices.
This PDF is available to Subscribers Only