For decades, neuroscientists looked at how sloppily our brains handle numbers and concluded we were stuck with the neural equivalent of a foggy windshield. The imprecision was just... there. A fixed tax on being biological. Then a pair of researchers at Columbia University flipped the script: what if that fog isn't a bug at all, but your brain carefully choosing where to spend its limited clarity?

Counting Without Counting

Before we get to the twist, let's set the scene. Somewhere deep in your parietal cortex, a cluster of neurons is doing something remarkable right now - estimating quantities without bothering to count. See a flock of birds overhead? You don't tick off each one. Your brain just... vibes it. "That's roughly thirty-ish." This is the approximate number system, or ANS, and every human (plus a surprising roster of fish, bees, and baby chicks) has one (Halberda et al., 2018).

The catch? It's noisy. Your internal estimate of 30 birds could easily be 25 or 37 on any given glance. And the bigger the number, the fuzzier it gets - a pattern that follows Weber's law, the same principle that explains why you can tell 1 kg from 2 kg but struggle with 50 kg versus 51 kg. Scientists have known about this sloppiness for a while. The question was always why.

The Brain as a Thrifty Landlord

Arthur Prat-Carrabin and Michael Woodford's new study in eLife proposes an answer that's surprisingly elegant: your brain is running a cost-benefit analysis on every number it processes (Prat-Carrabin & Woodford, 2026).

Think of it like an ecosystem allocating energy. A forest doesn't grow the same density of roots everywhere - it invests heavily where water is scarce and conserves where rain is plentiful. Your brain, the researchers argue, does something similar with its representational precision. It allocates neural resources based on what the current situation demands, not according to some fixed wiring diagram.

They tested this by giving people two tasks involving numerosity (the fancy word for "how many things are there"). In one, participants estimated the number of dots flashed on a screen. In the other, they compared sequences of numbers to decide which had a higher average. Crucially, the researchers varied the range of numbers people could expect - sometimes narrow, sometimes wide.

The Plot Thickens (Sublinearly)

Here's where it gets interesting. When the range of possible numbers widened, people's imprecision grew too - but not proportionally. It scaled sublinearly, meaning the brain didn't just throw its hands up when the range got bigger. It stretched its resources, getting somewhat fuzzier but fighting to keep useful accuracy. Like a photographer adjusting exposure for a wider landscape - you lose some detail, but you don't lose the whole picture.

Even more telling: this sublinear scaling was different between the two tasks. The estimation task and the discrimination task produced distinct patterns of imprecision, even when using the same number ranges. Same brain, same numbers, different goals - different precision allocation.

This is the key finding. If perceptual noise were just a fixed property of your neurons (like static on an old radio), switching tasks shouldn't change anything. But it does. Your brain is actively tuning how precisely it represents numbers based on what you're trying to accomplish with those numbers.

Not Lazy - Efficient

The theoretical framework behind this is called efficient coding, and it's one of the most compelling ideas in modern neuroscience. First formalized for sensory systems, it suggests that the brain encodes information in a way that maximizes usefulness while minimizing metabolic cost (Wei & Stocker, 2015). Your neurons are expensive to run - each spike costs energy - so the brain has evolved to be strategic about when and where it spends that currency.

Prat-Carrabin and Woodford's earlier work showed this principle applies to numerical cognition: people encode frequently occurring numbers more precisely and rare ones more loosely, just as an efficient code would predict (Prat-Carrabin & Woodford, 2022). The new study extends this by showing the task itself reshapes the code. Your brain isn't just adapting to what numbers are out there - it's adapting to what you need to do with them.

Why This Matters Beyond the Lab

If perception is endogenously determined (generated from within, not imposed from without), it changes how we think about everything from education to clinical diagnosis. A child struggling with math might not have a "worse" number sense in any absolute way - their brain might simply be allocating precision differently based on context, expectations, or prior experience. Understanding the knobs the brain is turning could eventually help us figure out how to turn them more effectively.

It also nudges us toward a more respectful view of the brain's apparent imperfections. That fuzziness in your number sense? It's not sloppiness. It's the signature of a system doing remarkable optimization under real constraints - like a wildlife reserve managing finite land to support the greatest possible biodiversity. Not perfect for any single species, but impressively balanced across the whole ecosystem.

Your brain, it turns out, isn't bad at numbers. It's just very, very good at budgeting.

References

1. Prat-Carrabin, A., & Woodford, M. (2026). Endogenous precision of the number sense. eLife, 13, e101277. https://doi.org/10.7554/eLife.101277

2. Prat-Carrabin, A., & Woodford, M. (2022). Efficient coding of numbers explains decision bias and noise. Nature Human Behaviour, 6(8), 1142-1152. https://doi.org/10.1038/s41562-022-01352-4

3. Wei, X.-X., & Stocker, A. A. (2015). A Bayesian observer model constrained by efficient coding can explain 'anti-Bayesian' percepts. Nature Neuroscience, 18(10), 1509-1517. https://doi.org/10.1038/nn.4105

4. Halberda, J. (2018). An Introduction to the Approximate Number System. Child Development Perspectives, 13(1), 41-48. PMCID: PMC6286047

5. Piazza, M., Izard, V., Pinel, P., Le Bihan, D., & Dehaene, S. (2004). Tuning curves for approximate numerosity in the human intraparietal sulcus. Neuron, 44(3), 547-555. PMID: 15504333

Disclaimer: The image accompanying this article is for illustrative purposes only and does not depict actual experimental results, data, or biological mechanisms.