The Foundation of Number Sense: From Approximate Quantities to Symbolic Representation

The construct of “number sense” serves as the foundational pillar upon which all subsequent mathematical cognition is built, representing not merely a set of skills but an integrated, pre-verbal capacity for quantifying the environment. It is the primary target of early mathematical intervention because its integrity and calibration in early childhood predict later achievement in formal mathematics more reliably than any other early cognitive factor. This chapter establishes the theoretical and empirical boundaries of this construct, arguing that number sense emerges from the interplay of two evolutionarily ancient, domain-specific cognitive systems: the Approximate Number System (ANS) and the Object Tracking System (OTS). The ANS provides a continuous, ratio-dependent sense of magnitude for large sets, while the OTS precisely enumerates small sets (typically up to four items) via parallel, attention-limited indexing. Their neural substrates, developmental trajectories, and points of convergence form the biological bedrock from which symbolic number concepts—verbal counting and Arabic numerals—are subsequently constructed. Crucially, this transition is not a simple replacement but a calibration and externalization process, profoundly shaped by the multi-sensory, socially embedded inputs of everyday environments, which both refine the ANS and create the pragmatic necessity for a symbolic representational system.

The Dual-System Architecture of Pre-Verbal Quantity Representation

The consensus in contemporary cognitive neuroscience and developmental psychology is that human infants and non-human primates possess at least two dissociable systems for representing numerical information, each with distinct computational properties, neural loci, and evolutionary histories. This dual-system architecture is not a theoretical speculation but a robust finding replicated across behavioral, neuroimaging, and lesion studies.

The Approximate Number System (ANS): A Continuous Sense of Magnitude

The ANS is an analog magnitude representation that allows for the discrimination of differing quantities in a continuous, ratio-dependent manner. Its signature characteristic is adherence to Weber’s Law: the just noticeable difference (JND) in perceived magnitude is a constant proportion ($k$) of the baseline magnitude. This is formalized as $\Delta I / I = k$, where $\Delta I$ is the change in intensity (here, numerosity) required for detection, and $I$ is the initial intensity. Consequently, discrimination acuity improves as the ratio between two numerosities approaches 1:1. A classic demonstration involves presenting infants or adults with two arrays of dots (e.g., 12 vs. 24) and measuring looking time or reaction time. Discrimination is robust for a 1:2 ratio (12:24) but becomes difficult or impossible for a 8:9 ratio (16:18), even though the absolute difference (2) is identical. This ratio dependence is observed in infants as young as 6 months, in adults across different sensory modalities (vision, audition, touch), and even when controlling for non-numerical confounds like density, area, and contour length.

Neuroanatomically, the ANS is subserved by a distributed frontoparietal network. Key nodes include the intraparietal sulcus (IPS), particularly the horizontal segment (hIPS), which shows parametric modulation of activity with numerical magnitude regardless of stimulus format (dots, sounds, fingers). The IPS is thought to host topographic “number maps” where numerical quantity is represented as a neural population code along a continuous mental number line. Supporting this, single-unit recordings in non-human primates reveal neurons in the IPS that are tuned to specific numerosities, exhibiting Gaussian-like tuning curves—a neural signature of an analog magnitude system. The prefrontal cortex (PFC), especially dorsolateral regions, interacts with the IPS to support ANS-based comparisons and decision-making, integrating the approximate magnitude signal with task rules and working memory demands.

Developmentally, the ANS is present at birth or emerges extremely early. Newborns show a preference for arrays with a 2:1 ratio over those with a 3:2 ratio. Its acuity, measured by the Weber fraction ($k$), improves dramatically during the first year of life and continues to refine through childhood, asymptoting in adolescence at levels comparable to adults ($k \approx 0.10-0.15$). This improvement is not due to a qualitative change in the system’s architecture but to increased efficiency and precision of the underlying neural representations, likely driven by experience with quantities in the environment. Individual differences in ANS acuity (measured by tasks like the nonsymbolic comparison task) are stable over time and correlate with concurrent and future mathematical achievement, even after controlling for general intelligence and working memory, making it a critical marker for early identification.

The Object Tracking System (OTS): Parallel, Limited Enumeration

The OTS, often termed “subitizing,” is a fundamentally different mechanism for quantity representation. It allows for the immediate, accurate, and effortless enumeration of small sets of individual objects, typically up to three or four items for adults, and fewer for infants. Its properties are discrete, not continuous: it provides a precise, one-to-one representation of each item in a small set without any sense of ratio or approximation. The OTS is subject to a severe capacity limit of approximately 3–4 items, a constraint linked to the limited capacity of visual working memory or attention. This limit is not arbitrary; it aligns with the known capacity of the brain’s object-file system, a mechanism for tracking individual objects through time and space.

Neuroanatomically, the OTS relies on a partially overlapping but distinct network from the ANS. The IPS is again involved, but posterior parietal regions, including the superior parietal lobule and regions involved in spatial attention and object-based processing (e.g., the intraparietal sulcus’s posterior extent), show greater activation for subitizable arrays (1–3 items) compared to larger arrays that must be processed by the ANS. The OTS also engages the ventral visual stream (e.g., lateral occipital complex) more strongly, as it requires individuating and maintaining detailed object features. The capacity limit is thought to reflect constraints on the number of object files that can be simultaneously instantiated and attended.

Developmentally, the OTS capacity increases slowly in infancy. Newborns can discriminate 1 vs. 2 but not 2 vs. 3. By 6–12 months, infants can reliably discriminate 1 vs. 3, and by 2–3 years, children’s subitizing range extends to 4 items. This gradual expansion is likely due to improvements in attentional control, visual working memory, and the efficiency of object-based attention mechanisms. Unlike the ANS, which is ratio-dependent, OTS performance is all-or-none for sets within its capacity: accuracy is near-perfect and reaction times are flat for 1–3 items, then jump dramatically as set size exceeds the capacity, forcing a switch to an ANS-based or serial counting strategy.

Dissociation and Interaction: Evidence from Behavior and Brain

The two systems are dissociable. Patients with parietal lesions, particularly to the IPS, can exhibit “acalculia” or “dyscalculia,” often showing impaired ANS-based comparison while preserving OTS function for small sets, or vice versa. In typically developing children, the correlation between ANS acuity and OTS accuracy is weak to moderate, and they predict different aspects of later math: ANS acuity correlates with later symbolic magnitude processing and arithmetic fluency, while OTS capacity correlates more with early counting skills and working memory.

However, the systems are not entirely independent in everyday perception. For sets of 3–5 items, both