Numerosity processing in early visual cortex

Abstract

While parietal cortex is thought to be critical for representing numerical magnitudes, we recently reported an event-related potential (ERP) study demonstrating selective neural sensitivity to numerosity over midline occipital sites very early in the time course, suggesting the involvement of early visual cortex in numerosity processing. However, which specific brain area underlies such early activation is not known. Here, we tested whether numerosity-sensitive neural signatures arise specifically from the initial stages of visual cortex, aiming to localize the generator of these signals by taking advantage of the distinctive folding pattern of early occipital cortices around the calcarine sulcus, which predicts an inversion of polarity of ERPs arising from these areas when stimuli are presented in the upper versus lower visual field. Dot arrays, including 8-32dots constructed systematically across various numerical and non-numerical visual attributes, were presented randomly in either the upper or lower visual hemifields. Our results show that neural responses at about 90ms post-stimulus were robustly sensitive to numerosity. Moreover, the peculiar pattern of polarity inversion of numerosity-sensitive activity at this stage suggested its generation primarily in V2 and V3. In contrast, numerosity-sensitive ERP activity at occipito-parietal channels later in the time course (210-230ms) did not show polarity inversion, indicating a subsequent processing stage in the dorsal stream. Overall, these results demonstrate that numerosity processing begins in one of the earliest stages of the cortical visual stream.

Keywords: Calcarine sulcus; ERPs; Numerosity; Visual cortex; Visual processing.

FIGURE 1.

Procedure. Participants were asked to fixate on the central fixation cross while stimuli were displayed on the screen for 200 ms, either in the upper or lower portion of the screen (center-to-center eccentricity = 3.75 deg). Different stimuli were divided by blank periods where only the fixation point was displayed, with a variable ISI of 500–700 ms between successive presentations. Occasionally, a red dot-array was showed, and in such cases participants were instructed to press a button on a joypad as soon as possible.

FIGURE 2.

Time course of brain responses to stimuli collapsed across all stimulus conditions in the upper (upper panel) and lower (lower panel) visual field. In both cases, early activity showed maximum amplitude at around 75–125ms, with positive-polarity modulation for upper visual field stimuli and negative-polarity modulation for lower visual field stimuli. Later activity on the other hand showed maximum amplitude around 225–250ms, with positive-polarity modulation in both visual-hemifield conditions.

FIGURE 3.

Results of the linear mixed effects model analysis. (A-D) Beta values for the different regressors entered in the analysis as fixed effects, across several time windows. Posterior view of the time course of beta values for numerosity (A), size (B), spacing (C), and visual hemifield (D).

FIGURE 4.

The effects of magnitude (Euclidean norm of the beta vector comprising numerosity, size, and spacing) in the medial occipital (A) and bilateral occipito-parietal (B) channels. The arrows indicate the peak latencies selected for further analysis (55 ms and 88 ms in panel A, 213 ms and 231 ms in panel B).

FIGURE 5.

Brainwaves sorted for changes along numerosity, size, and spacing, at channel Oz and PO8i. (A, B and C) ERPs sorted for the different values of numerosity (A), size (B) and spacing (C), for the midline posterior channels Oz. (D, E and F) Contrast waves representing the weighted difference between responses elicited by different levels of numerosity (D), size (E) and spacing (F). Orange waves represent the contrast waves relative to upper visual field stimuli, while cyan waves represent the contrast waves relative to lower visual field stimulation. (G, H and I) Brainwaves sorted along the different values of numerosity (G), size (H) and spacing (I), recorded at the right occipito-parietal channel PO8i. (J, K and L) Contrast waves relative to the different levels of numerosity (J), size (K), and spacing (L) for upper (orange) and lower (cyan) visual field stimulation (channel PO8i).

FIGURE 6.

Angle analysis at early latency stages of magnitude processing. We assessed which one of the candidate dimensions best represents the beta estimates vector at the channels and latencies identified with the ROI selection and regression analyses, in the C1 range. Angle between the parameter estimate vector and all the candidate dimensions (A) for channel Oz at 55 ms, (B) for channel Oz at 88 ms. Values inside square brackets represent the Euclidean norm of the parameter estimate vector. The size of each polar plot is scaled according the norm of the parameter estimate vector.

FIGURE 7.

Angle analysis over the bilateral occipital channels in the P2p latency range. Angle between the parameter estimate vector and all the candidate dimensions (A) for channel PO7i at 231 ms, (B) for channel PO8i at 213 ms. Values inside square brackets represent the Euclidean norm of the parameter estimate vector. The size of each polar plot is scaled according the norm of the parameter estimate vector.