Lead
"She's three and still can't count" is usually not a cause for concern — and the reason is more interesting than it might sound. Hearing "one, two, three" come out of a child's mouth does not necessarily mean that child is counting, and conversely, a child who never says those words aloud may be quietly building a genuine understanding of quantity.
Number concept development runs considerably deeper than the surface skill of reciting a sequence. It begins in infancy with an intuitive grasp of amount, works its way through the gradual acquisition of counting principles across ages three and four, and arrives — in a moment researchers have spent decades trying to characterize — at the understanding that the last number you say when counting a group represents the whole group. This article lays out what the evidence currently shows about how that journey unfolds.
At Five Months: "1 + 1 = 1 Looks Wrong"
The research on infant numerical cognition has a canonical starting point: Karen Wynn's 1992 experiment, published in Nature [1].
Wynn used the violation-of-expectation method: an infant research technique measuring looking time as a proxy for surprise; infants look longer at outcomes that contradict their expectations with five-month-old infants. A stuffed animal was placed behind a screen; a second was visibly added. When the screen dropped to reveal only one animal, infants looked significantly longer than when two were revealed — the "that's not right" response. When the correct number appeared, they looked away sooner.
Wynn concluded that infants form expectations about the results of small additions and subtractions — that the beginnings of numerical reasoning are present in the first half-year of life [1]. The paper ignited a debate in developmental psychology that has continued for more than three decades, and its core finding has been replicated many times.
Two Core Systems
Following Wynn's work, a theoretical framework for understanding how infants and young children represent number has developed. The most widely accepted formulation is the two-core-systems model, laid out by Lisa Feigenson, Stanislas Dehaene, and Elizabeth Spelke in Trends in Cognitive Sciences in 2004 [2].
The first system handles small numbers — roughly one to four — with precision and speed. This is subitizing: the rapid, accurate perception of small quantities (typically 1–4) without counting, present from infancy and distinct from the approximate number system: the ability to know instantly that there are two objects, or three, without counting. The mechanism underlying subitizing is thought to be an object file system, a form of attention that tracks individual objects simultaneously [2]. This capacity persists throughout life.
The second system is the approximate number system: an innate, language-independent capacity to compare and estimate large quantities by ratio rather than by exact counting (ANS), which handles larger quantities with ratio-based precision rather than exact counting. It is what allows you to tell at a glance that twenty objects are more than ten, while being unable to distinguish ten from eleven [2]. The ANS responds to ratios: a 1:2 difference is readily detected; a 10:11 difference is not. This system, too, is present from early infancy.
Both systems function before language. They become the substrate onto which verbal counting — learning number words — is later mapped, and through that mapping, a more abstract conception of number gradually emerges.
The distinction between the two systems matters practically. Subitizing operates in the range where errors are rare and immediate: hold up three fingers and most adults, children included, know at once that it is three — no counting required. The ANS operates at scales where counting would take too long and provides a workable approximation instead. A preschooler who confidently declares that "the red pile has more" when comparing a group of twenty to a group of ten is not guessing randomly; they are deploying a real perceptual system. What they cannot yet do is tell you how many more, and that gap — between approximate comparison and exact representation — is precisely the gap that language-based counting fills over the years between two and five.
What "Understanding Numbers" Actually Requires
A three-year-old who can recite "one, two, three, four, five" may be doing something that looks like counting but is not yet counting in the relevant sense. The distinction matters.
Mathieu Le Corre and Susan Carey's 2007 study in Cognition analyzed the process by which children acquire the meaning of number words in detail [3]. They sorted children aged two to four into stages: one-knowers (who can correctly give one item when asked for "one" but grab a random handful when asked for "two"), two-knowers, three-knowers, and finally CP-knowers — children who have grasped the cardinality principle.
The cardinality principle: the understanding that the final number word in a count represents the total size of the set, typically grasped between ages 3.5 and 4 is the insight that the last word in a count represents the total quantity of the set. Before grasping it, a child who has just counted five blocks, when asked "how many blocks are there?", will recount rather than simply say "five." After grasping it, the child answers immediately: "five" [3]. This understanding, on average, crystallizes between ages three and a half and four — but individual variation is wide.
The broader conceptual framework for counting was laid out by Rochel Gelman and Charles Gallistel in their 1978 book, The Child's Understanding of Number [4]. They described five counting principles: one-to-one correspondence (each object gets exactly one count), stable order (the count words always come in the same sequence), cardinality (the last word gives the total), abstraction (any set of objects can be counted), and order irrelevance (counting can begin with any item in the set). A child's verbal counting behavior can look correct on the surface while the underlying principles — particularly cardinality — have not yet been grasped. The gap between apparent and genuine mastery is often substantial [4].
What Parents Can Actually Observe
There is no need for parents to test their children's number concepts. But watching, and occasionally recording, how a child engages with quantities in everyday life opens an interesting window.
Examples of numerically meaningful behavior in young children:
- Matching cups one-to-one with spoons (one-to-one correspondence, practiced)
- Responding to "how many are there altogether?" with an immediate answer rather than recounting (a sign that cardinality is taking hold)
- Comparing quantities: "this pile has more" (ANS in action)
Julie Sarama and Douglas Clements's work on mathematics education research, synthesized in their 2009 book Early Childhood Mathematics Education Research, makes clear that the accumulation of everyday quantitative experience shapes a young child's mathematical learning trajectory significantly [5]. Counting the stairs, dividing up a snack, lining up blocks — these lived encounters with quantity build the conceptual foundation. Flashcards and drills are not necessary; the material is already present in ordinary life.
Measuring a child against "how many can they count to at age X" is a less useful frame than noticing in what contexts they engage with quantity and what they do with it. A log in an app like Memori — a quick note when a child spontaneously counts their toys or insists on equal shares — captures the developmental arc in a way that periodic benchmarks do not.
Summary
Number concept development begins earlier than verbal counting does. Subitizing — precise, immediate apprehension of small quantities — and the approximate number system are both present in infancy and form the foundation onto which language-based counting is eventually mapped [1,2]. Number-word learning proceeds one word at a time, with the cardinality principle — the understanding that the last count word represents the whole set — typically consolidating between ages three and a half and four, with wide individual variation [3,4].
There is a wide range of normal within this process. Everyday encounters with quantity — sharing food, counting objects, matching items — provide the most natural and effective scaffolding [5]. Rushing the teaching tends to be less useful than noticing what the child is already doing with number, and making space for more of it.
References
- Wynn K. Addition and subtraction by human infants. Nature. 1992;358(6389):749–750. doi:10.1038/358749a0. PMID: 1508269.
- Feigenson L, Dehaene S, Spelke E. Core systems of number. Trends Cogn Sci. 2004;8(7):307–314. doi:10.1016/j.tics.2004.05.002. PMID: 15242690.
- Le Corre M, Carey S. One, two, three, four, nothing more: an investigation of the conceptual sources of the verbal counting principles. Cognition. 2007;105(2):395–438. doi:10.1016/j.cognition.2006.10.005. PMID: 17208214.
- Gelman R, Gallistel CR. The Child's Understanding of Number. Cambridge, MA: Harvard University Press; 1978.
- Sarama J, Clements DH. Early Childhood Mathematics Education Research: Learning Trajectories for Young Children. New York: Routledge; 2009.