Adaptation to Numerosity Changes Monotonic Responses to Image Contrast in Early Visual Cortex

1356 

Abstract Submission 

Liangyou Zhang¹, Evi Hendrikx¹, Yizhen Wang², Surya Gayet¹, Serge Dumoulin³, Ben Harvey¹

¹Utrecht University, Utrecht, The Netherlands, ²South China Normal University, Guangzhou, China, ³Spinoza Centre for Neuroimaging, Amsterdam, The Netherlands

Liangyou Zhang  
Utrecht University
Utrecht, The Netherlands

Evi Hendrikx  
Utrecht University
Utrecht, The Netherlands

Yizhen Wang  
South China Normal University
Guangzhou, China

Surya Gayet  
Utrecht University
Utrecht, The Netherlands

Serge Dumoulin  
Spinoza Centre for Neuroimaging
Amsterdam, The Netherlands

Ben Harvey  
Utrecht University
Utrecht, The Netherlands

Humans and many animals effortlessly perceive numerosity, the number of visual objects in an image. Numerosity perception is affected by adaptation (Burr & Ross, 2008): after viewing a particular adapter numerosity frequently or for a long time, other numerosities appear more different from this adapter than they are. Recent studies show two broad classes of neural responses to numerosity. First, parietal and frontal numerosity-tuned responses (Harvey et al., 2013; Nieder et al., 2002) show activation peaks at a specific (preferred) numerosity. We have recently shown that this preferred numerosity is affected by numerosity adaptation (Tsouli et al., 2021). Second, monotonic responses in the early visual cortex (DeWind et al., 2019; Park et al., 2015) increase in amplitude as numerosity increases. These monotonic responses are found in neural populations at the retinotopic location of the stimulus in early visual field maps (Paul et al., 2022). Furthermore, they follow spatial frequency domain image contrast more closely than they follow numerosity, though close relationships between contrast and numerosity may underlie numerosity's straightforward perception.
Here we ask whether effects of numerosity adaptation on perception and numerosity-tuned responses originate in early visual contrast representations.

We used ultra-high field 7T fMRI to record responses to numerosity in early visual field maps V1-V3, hV4, LO1-2 and V3A/B. Participants viewed dot groups of gradually changing numerosity to map numerosity preferences. We repeated this under three different conditions. The low numerosity adaptation condition alternated this changing numerosity display with displays containing one item. The high numerosity adaptation condition alternated the changing numerosity with twenty items. In the control condition, every display contained the changing numerosity.
In each condition, we used a general linear model (GLM) to capture the responses to the changing numerosity, where each voxel's response is predicted by the spatial frequency domain contrast energy (i.e. Fourier power) of each display. This determines the amplitude of the monotonic response to increases in Fourier power as the slope (beta) of the relationship between the display's Fourier power and each voxel's response.
We used visual field mapping to determine which visual field map each voxel lay in and its preferred visual field position (Figure 1a-c). For each visual field map, we selected voxels responding within the numerosity stimulus area and showing a monotonically increasing response to Fourier power in the control condition. We tested whether the mean response slope differed between low and high numerosity adaptation conditions. We then compared, across visual field maps, the proportion by which the response slope decreased between the low and high adaptation conditions.

The monotonic response's amplitude (slope or beta in the GLM) reduced during adaptation to high numerosities, compared to adaptation to low numerosities, in all visual field maps (Figure 1d). This is consistent with perceptual effects where high numerosity adaptation decreases perceived numerosity. Furthermore, the proportion by which response amplitude decreased from the low to high numerosity adaptation conditions became progressively greater from earlier visual field maps (~30% decrease in V1) to later visual field maps (~70% decrease in V3A/B and LO2) (Figure 1e).

These results imply that adaptation effects on numerosity-tuned neural populations' preferred numerosities, and on numerosity perception, may result from effects beginning in early visual contrast representations. These effects begin by V1 and increase through the visual processing hierarchy. Therefore, numerosity adaptation in part reflects early-stage adaptation to image contrast. However, the progressive increase in this effect suggests that adaptation may have further effects at later stages of numerosity processing.

Activation (eg. BOLD task-fMRI) ¹

Multivariate Approaches

Perception: Visual ²

Other - Numerosity adaptation, 7T fMRI, image contrast, visual cortex