Tracking numerical and continuous magnitude processing by frequency-tagged neuromagnetic responses

Abstract

Event Abstract Back to Event Tracking numerical and continuous magnitude processing by frequency-tagged neuromagnetic responses Amandine Van Rinsveld1*, Vincent Wens2, 3, Mathieu Guillaume1, Anthony Beuel1, Wim Gevers1, Xavier De Tiège2, 3 and Alain Content1 1 ULB Neuroscience Institute, Free University of Brussels, Center for Research in Cognition and Neurosciences , Belgium 2 ULB Neuroscience Institute, Free University of Brussels, Laboratoire de Cartographie fonctionnelle du Cerveau, Belgium 3 Erasmus Hospital, Free University of Brussels, Magnetoencephalography Unit, Department of Functional Neuroimaging, Service of Nuclear Medicine,, Belgium The ability to handle approximate large quantities has been identified as a building block of mathematical skills but the mechanisms allowing to extract numerical magnitudes (e.g., numerosity) from environmental stimuli is still debated. Most authors agree that humans have an approximate number system that specifically processes numerosity. However, a set of objects is not only characterized by its numerosity but also by additional visual information related to its continuous dimensions (e.g., object size), leading to alternative theories to explain numerosity extraction. Until now opposite views could not be tested properly due to the intrinsic correlations between numerosity and continuous dimensions. We isolated the specific responses to numerosity and other continuous dimensions using a 10 Hz-rate presentation of dot patterns that varied randomly in all dimensions but one dimension of interest, which was varied systematically at 1.25 Hz5 (see Fig.1). We recorded the evoked brain activity as frequency-tagged steady-state visual responses measured with magnetoencephalography. Neuromagnetic data were recorded in twenty-one subjects with a 306-channel MEG Triux system. Data were preprocessed by signal space separation (SSS) with movement correction and realignment on default head position. The amplitude spectra for all channels were obtained through Fourier transformation on the average of four 40 s-long epochs (frequency resolution of 0.025 Hz). To assess the neural synchronization at 1.25 Hz, we extracted and summed spectra in 0.6 Hz-wide frequency intervals around the 7 first harmonics of 1.25 Hz. The resulting sum-based amplitudes (SBA) were then converted into statistical maps of individual t scores (df=19) comparing the SBA at 1.25 Hz with the SBA at the surrounding frequencies (i.e., 10 adjacent frequency bins on both sides of the 1.25 Hz bin). At the group level, significance of the t maps was assessed using one-tailed maximum-based permutation tests against the null hypothesis ts = 0. This non-parametric approach automatically corrects for multiple comparisons across all sensors. We found neural synchronization at the presentation rate of 10 Hz in all sequences (average t=10.73, permutation p<.001). Crucially, we observed significant SBA peak at 1.25 Hz, corresponding to the periodic change of one dimension, for three dimensions: total area, number, and convex hull (ps<.01), all emerging at medial occipital sensors (see Fig.2). No significant synchronization occurred for density or dot size. In conclusion, by using an objective measure of implicit discrimination of each magnitude change, we observed neuromagnetic responses synchronized on the periodic change of numerosity, which is coherent with previous EEG results but also of total area and convex hull of the dot collections. In contrast, no synchronization was observed for other continuous dimensions. Numerosity, Area, and Convex hull can thus be rapidly and independently extracted in the visual stream, supporting the idea of an early visual number sense. At this stage of the processing, the discrimination of numerosity changes cannot be reduced as resulting from continuous dimensions’ discrimination. Figure Captions: Fig 1. Sequences of dot patterns were characterized by five dimensions: dot size, total area, convex hull, density, and numerosity. a. Time series of values (200 stimuli) for the five dimensions considered. In this example, the numerosity varies periodically at 1.25 Hz and the other continuous dimensions vary randomly. b. Frequency spectra of the corresponding time series after Fourier transformation. c. Illustration of a subset of stimuli. Fig 2. Topographic maps of group t values for SBA at 1.25 Hz and its harmonics in the conditions that disclosed significant SBA according to the maximum-based permutation test (H0: t=0). The star indicates significant channels, whose SBA spectrum is depicted on the right. Figure 1 Figure 2