Early Numeracy: How Number Sense Develops in Babies and Toddlers

"I was always rubbish at maths" is one of the most common things parents say to their toddlers, and it's also one of the few things that will measurably hurt that toddler's later maths achievement. Maths anxiety is contagious — parents who self-deprecate around numbers raise children who do worse at numeracy, even controlling for parents' actual maths ability (Beilock et al., PNAS 2010). The flip side is encouraging: there's almost nothing you need to buy or set up. The home maths environment that matters is conversational, embedded, and runs on stairs, biscuits, and Lego.

The Healthbooq app tracks early development including the rough numeracy milestones — useful for noticing when something has clicked rather than for hitting a target.

What's Already There at Birth

Before any counting, babies already have what cognitive scientists call the approximate number system (ANS) — a rough, non-verbal sense of "more" and "less" that we share with rats, fish, and crows. It works on ratios: a four-month-old can tell 8 dots from 16 (a 1:2 ratio) but not 8 from 12 (a 2:3 ratio). By 9 months the ratio has tightened to about 2:3.

The strongest evidence for this is Karen Wynn's 1992 Nature paper. Five-month-olds watched a Mickey Mouse doll placed behind a screen, then a second doll added behind it. When the screen lifted, sometimes there were two dolls (correct) and sometimes just one (impossible). Babies stared significantly longer at the impossible outcome — they were tracking the addition.

This is not a precursor to maths in the abstract sense; it is maths, in its earliest form. Children with sharper ANS acuity at age 3–4 perform better in school maths assessments at age 7, even after controlling for vocabulary and IQ.

The practical implication for parents: babies are already doing quantity comparisons in everyday life. Two biscuits versus one. The big slice versus the small slice. Lots of grapes versus a few. Naming what they're already noticing — "this pile has more; that one has fewer" — gives them the language scaffold to upgrade an intuition into a tool.

How Counting Actually Develops

Toddlers will recite "one, two, three, four, five" as a string of sounds long before any of those sounds map to actual quantity. This is the count list — a piece of memorised vocabulary, no different from "ABCDE." A 2-year-old who counts to 10 is showing memory, not numeracy.

Real counting requires the five principles laid out by Gelman and Gallistel in 1978, and they don't all arrive together:

1. One-to-one correspondence — one word per object, no doubling up. Watch a 2-year-old "count" three blocks: they often skip blocks or tap the same one twice. This typically firms up between 2½ and 3.
2. Stable order — the words always come in the same sequence. By 3, most children have a stable count list to about 10.
3. Cardinality — the realisation that the last number you say is the size of the set. This is the big one. Ask a 3-year-old "how many?" after a careful count and you may get a re-count. Ask a 4-year-old and they'll just say the last number. The transition between these two states is the moment counting becomes useful for anything other than performance.
4. Abstraction — you can count anything, including ideas (jumps, sleeps until Christmas, claps). Solid by around 4.
5. Order irrelevance — it doesn't matter which object you start with; the count is the same. Around 4–5.

The cardinality jump is the developmentally interesting one. Susan Carey's "knower-level" research breaks the year before cardinality clicks into stages: the "one-knower" can hand you 1 thing but guesses for 2 or more; the "two-knower" can do 1 and 2; and so on through three- and four-knowers, before the cardinality principle generalises to all numbers in their count list. This sequence usually plays out between 2½ and 4 years.

Maths Talk: The Single Biggest Variable

The Levine et al. study (2010, Developmental Psychology) followed 44 families with toddlers, recording 90-minute home interactions every four months from 14 to 30 months. The total amount of "number talk" parents used varied roughly thirty-fold — from about 4 number words across all sessions to about 250. The amount of number talk at age 2 predicted children's cardinality knowledge at 4½, even after controlling for general parent talk, parent education, and the child's vocabulary.

Number talk isn't only counting. It includes:

• Cardinality questions — "How many spoons do we need?"
• Comparisons — bigger, smaller, more, fewer, longer, shorter, heavier, lighter
• Spatial language — above, below, behind, between, inside, on top of, halfway
• Shape language — round, square, pointy, flat, curved, edge, corner
• Pattern language — first, next, last, again, every other one, in a row

A 2014 University of Chicago study found that spatial talk in particular ("the wheel goes under the chair," "the L-shape goes next to the long block") had outsize effects on later block-building and mental-rotation skills, which in turn predict STEM success years on.

Practical hooks for the day:

• Stairs are the easiest counting prop in the house. One stair, one number, every time you climb.
• Snack time — count grapes, compare pile sizes, share equally ("if you have three and I have one, that's not fair — let's split them"), introduce halves and quarters with apples.
• Bath toys — submerging two ducks and asking how many are still floating teaches subtraction with no curriculum.
• The car — "how many red cars can you see at this junction?" Counting under time pressure builds cardinality.
• Setting the table — five people, five forks. The most natural one-to-one correspondence task in the world.

Shape, Space, and Why Lego Is Underrated

Spatial reasoning is a distinct cognitive skill from numerical reasoning, and a stronger predictor of STEM outcomes at university than maths SAT scores (Wai, Lubinski, Benbow, 2009, longitudinal data over 35 years). It's the ability to manipulate, rotate, and project shapes mentally.

What builds it in early childhood:

• Construction toys — Duplo, Lego, magnetic tiles. Free-build is fine; following a model picture is even better because it forces 3D-to-2D translation.
• Puzzles — jigsaws specifically. Children who do puzzles between 2 and 4 score higher on mental rotation tasks at age 4½ (Levine, Ratliff, Huttenlocher, Cannon, 2012).
• Shape sorters at toddler age, then tangrams from age 4.
• Drawing maps of familiar places — the bedroom, the route to nursery — from around age 4.
• Origami and cutting paper for older preschoolers.

Notice that none of these is "maths education." They look like play, and they should — formal sit-down maths instruction at this age has poor returns and risks souring the relationship with the subject.

What Early Years Settings Do (and What Parents Don't Need To)

The Early Years Foundation Stage (EYFS) framework in England, revised 2021, includes Mathematics as one of seven areas of learning. The Early Learning Goals at age 5 are modest by design: count to 10 reliably, recognise small quantities without counting (subitising), use number bonds within 5, recognise patterns and shapes. The framework is explicit that this should be delivered through play and embedded routines, not formal instruction. The same is true of US programmes that score well on the ECERS-R (Early Childhood Environment Rating Scale): incidental maths language threaded through play correlates with numeracy outcomes; structured maths "lessons" do not.

For parents this means the homework is being done for you in any decent nursery. What you can usefully add at home is more of the same — number talk, shape talk, spatial talk, pattern noticing — without converting it into instruction. A toddler who is asked "do you want one biscuit or two?" five times a day is doing maths.

What Genuinely Worries

Most differences in early numeracy are environmental rather than diagnostic. But the small group of children who go on to be diagnosed with dyscalculia (specific learning difficulty in maths, prevalence around 3–7 per cent) often show signs by age 4–5 that are visible in retrospect:

• Cardinality not clicking by 5
• Persistent difficulty with one-to-one counting at 4–5
• No grasp of "more" and "fewer" by 4
• Trouble recognising small quantities without counting (subitising) at 5
• Difficulty learning the count list past 5 by age 5

These features overlap heavily with delayed exposure and language delay, and a chat with the GP, school SENCO, or health visitor is usually the right starting point. Formal dyscalculia assessment is rare before age 7.

The other thing to flag, gently: a parent's own anxiety. If maths talk feels uncomfortable because of your own school history, the workaround is to lean on questions rather than instruction. "How many do you think?" rather than "Count them." "Which is bigger?" rather than "This one is bigger." Your child does the maths; you do the prompting.

Babies as young as five months notice arithmetic mistakes — Karen Wynn's classic Yale work shows infants stare longer when one toy plus one toy equals one. That early 'approximate number system' is the foundation maths sits on, and it's there before any teaching. The most reliable predictor of a child's maths skills at school entry isn't flashcards or apps; it's the amount of casual maths language they hear at home (Levine et al., 2010). Counting stairs, comparing pile sizes, and noticing shapes do more than any structured curriculum at this age. Cardinality — understanding that 'five' means a set of five things — typically clicks somewhere between 3 and 4 years, and that's the moment counting becomes a tool rather than a chant.