Our sense of numbers is worse than we think, and that might be exactly right
The human brain is bad at quantities, and that appears to be a feature, not a bug. A new study investigates why our ‘number sense’ is so imprecise, and why that imprecision might actually be the optimal strategy.
People are reasonably good at estimating whether there are more or fewer of something, but the precision of that judgment is surprisingly low. Estimate quickly: are there more than fifty dots on a page? Most people perform reasonably, but variability is high. This ‘number sense’, the ability to rapidly assess quantities without counting, has been studied in cognitive neuroscience for decades. What always stood out was its inherent imprecision.
The standard explanation was straightforward: the brain has limited processing capacity and therefore makes errors. But a new study proposes a different interpretation, drawing on efficient coding models from information theory. The core idea: if the brain wants to optimally use scarce representational resources, it should invest more precision in accurately representing quantities that occur most frequently in the environment. Rare quantities can afford to be represented less precisely, that’s more efficient.
Precision shifts with context
The study tested this experimentally. Participants judged a series of quantities while researchers varied the statistical distribution of those quantities. The prediction: if efficient coding holds, the precision of the number sense should track the statistical structure of the environment, better for common quantities, worse for rare ones.
That’s exactly what they found. Precision shifted in a predictable way with statistical context. The ‘noise’ in the number sense isn’t a fixed error in the system, it’s a flexible, context-sensitive adaptation. The brain doesn’t distribute its representational capacity evenly; it distributes it optimally.
What this means for cognitive aging
This might sound like abstract cognitive psychology, but the implications extend further. If the precision of cognitive representations isn’t fixed but depends on environmental statistics, then changes in the environment, or in internal processing, can affect precision.
In aging, both cognitive processing capacity and sensitivity to statistical patterns in the environment change. Whether the number sense changes accordingly, and whether that could serve as a diagnostic signal for cognitive decline, is a question this research raises without yet answering. It also offers a lens for reexamining other cognitive domains: how efficiently does the aging brain distribute its scarce representational capacity, and at what point does that optimization break down?