Neural representations of magnitude for natural and rational numbers

Melissa Dewolf, Jeffrey N. Chiang, Miriam Bassok, Keith J. Holyoak, Martin M. MontiNeuroImage 2016 · DOI

Introduction

Humans possess the unique ability to represent numerical magnitudes through various symbolic systems. We use natural numbers (integers like 50) and rational numbers (decimals like 0.50 or fractions like 1/2) interchangeably in daily life. However, a fundamental question in cognitive neuroscience remains: does the brain translate these diverse symbols into a single, abstract "mental number line," or does the format of the notation dictate how the information is neurally encoded?

This study investigates the neural signatures of integers, decimals, and fractions within the intraparietal sulcus (IPS), the brain's primary hub for mathematical cognition. By comparing these formats, researchers sought to determine if the brain organizes magnitude based on mathematical value or the structural properties of the notation (e.g., base-10 positional vs. bipartite).

Experimental Methodology

To isolate the neural representation of individual numbers, the researchers employed a sequential magnitude comparison task during fMRI scanning. Unlike traditional tasks where two numbers are presented simultaneously, this design showed a single number, followed by a jittered delay, and then a second number for comparison. This separation allowed for the analysis of the brain's response to a single stimulus before the cognitive process of comparison began.

The experimental design utilized a sequential presentation format to isolate the neural representation of individual numbers before comparison.

The study involved 16 healthy adults who evaluated 240 trials across three notation types: Integers, Decimals, and Fractions. Data analysis included both univariate methods to measure overall activation volume and Representational Similarity Analysis (RSA) to examine the fine-grained patterns of voxel activity.

Behavioral and Univariate Findings

Behavioral results indicated that processing fractions is significantly more cognitively demanding than processing integers or decimals. Participants were slower (averaging 1.91s for fractions vs. 1.30s for others) and less accurate (84% vs. 92%) when dealing with bipartite notation.

Univariate fMRI analysis mirrored these behavioral trends, showing increased activation in the left IPS and frontal regions for fractions. This suggests that the bipartite structure of fractions—requiring the integration of a numerator and a denominator—recruits more extensive neural resources than the relatively "automatic" processing of base-10 numbers.

Neural Architecture and RSA Results

The core of the study utilized RSA to map the "neural geometry" of numerical representations. The researchers tested several models to see which best explained the patterns of activity in the IPS.

RSA models testing the similarity of neural patterns across different notation formats.

The most significant finding was the "Base-10 Affinity." Despite the vast difference in absolute magnitude between an integer like "50" and a decimal like ".50," the IPS represented them using nearly identical neural patterns. In contrast, fractions evoked a distinct neural signature, even when their mathematical value was identical to a decimal. This indicates that the IPS codes magnitude on a relative scale and is highly sensitive to the notation format. The brain appears to group integers and decimals together due to their shared base-10 positional structure, while treating the bipartite structure of fractions as a separate category.

Discussion and Implications

These findings challenge the "fully abstract" view of numerical representation. Instead of a single mental number line, the brain maintains notation-specific representations organized by structural format. The shared neural language between integers and decimals suggests that our brains are optimized for the base-10 system we use most frequently.

Future research is needed to determine if this "fraction signature" is an inherent property of bipartite notation or a result of developmental factors, as fractions are typically learned later and practiced less than integers. Understanding these neural organizational principles provides critical insight into how humans bridge the gap between concrete symbols and abstract mathematical concepts.

Summary updated

BibTeX
@article{dewolf2016neural,
  author = {Melissa Dewolf and Jeffrey N. Chiang and Miriam Bassok and Keith J. Holyoak and Martin M. Monti},
  title = {{Neural representations of magnitude for natural and rational numbers}},
  journal = {NeuroImage},
  year = {2016},
  doi = {10.1016/j.neuroimage.2016.07.052},
}
Figure