From counting to addition: how kindergarten math really works

A five-year-old I know — call her Maya — can recite "one, two, three" all the way to one hundred without a stumble. She does it in the car. She does it in the bath. Her grandparents are convinced she is gifted at math.

Last month her mother put seven wooden blocks on the table and asked, "Can you show me seven of these?"

Maya froze. Then she touched the first block and said, "One." Touched it again. "Two." She lost count at four, started over, gave up, and asked for a snack.

That gap — between reciting numbers and knowing what they mean — is the entire story of kindergarten math. And most of us, including most apps that claim to teach early math, do a fairly poor job of noticing the gap, let alone closing it.

Reciting is not knowing

There's a name for what Maya was doing in the car. Educators call it the counting sequence or rote counting: a memorized string of sounds, the same way a kid memorizes the alphabet song or the lyrics to a TV theme. Useful, but not yet mathematical.

What Maya was being asked to do at the table is different. It's called cardinality — understanding that the last number you say when counting a set tells you how many things are in the set. If you count "one, two, three, four, five" across five blocks, the answer to "how many?" is five. Not "five" the recited word — five the quantity.

A surprising number of kindergarteners can do the first thing fluently and the second thing not at all. They count past the end of the row. They start over each time. They give different answers to "how many?" depending on which direction you point. None of this means anything is wrong. It means they are exactly where developmental science says they should be: somewhere on the long, uneven road from sound to symbol to quantity.

If you want the canonical map of that road, the National Council of Teachers of Mathematics has been publishing it for decades. It is not a fast road.

Subitizing: the foundation almost no one teaches

Here is the thing most parents have never heard of, and it might be the single most important math skill in early childhood.

Subitizing is the ability to look at a small group — usually one to five items — and know how many there are without counting. You don't say "one, two, three." You glance and see three. Roll a die and you don't count the pips; you see "four" instantly.

Researchers distinguish between perceptual subitizing (recognizing one, two, three almost reflexively) and conceptual subitizing (seeing six as "three and three" or "five and one" without working through it). Both matter, and both are trainable.

Why does this matter? Because subitizing is the bridge between how many? and arithmetic. A child who can see "three and two" as five on a domino tile is doing addition with their eyes before they ever pick up a pencil. A child who can't has to count every single time, which means addition feels like climbing a ladder where every rung is wobbly.

Most early math apps skip this entirely. They jump from "count to ten" to "what is 2 + 3?" with nothing in between. The kid who hasn't subitized yet has no way to answer except to count on their fingers — which, as it turns out, is also where most apps go wrong.

Fingers are not embarrassing

There is a deeply weird cultural belief that finger-counting is something kids should grow out of quickly, ideally by first grade, definitely by second. Teachers shoo it. Parents worry about it. Apps reward speed in a way that punishes it.

The neuroscience says the opposite. Jo Boaler, the Stanford math educator who runs youcubed.org, has written extensively about research showing that finger use in childhood is correlated with later mathematical achievement, and that there are specific brain regions associated with finger representation that get recruited even when adults do arithmetic without moving their hands. Telling a five-year-old to "stop using your fingers" is, roughly, telling them to stop using their visual cortex.

What you actually want is for finger-counting to gradually become unnecessary because the child has internalized the quantities — not because someone shamed them out of it.

Counting-on, and the moment 4 + 3 changes forever

Watch a kindergartener add four and three for the first time. Most of them will count both groups from scratch: "one, two, three, four… one, two, three… one, two, three, four, five, six, seven."

Then, somewhere between five and seven years old, something shifts. The child puts four in their head and counts on: "four… five, six, seven." This is called the counting-on strategy, and it is a real cognitive leap. It requires holding "four" as a stable mental quantity while doing something else with three.

Counting-on is the doorway to fluent arithmetic. But — and this is the part most curricula miss — you cannot make it happen by drilling. A child who hasn't built the underlying number sense will memorize "4 + 3 = 7" the same way they memorized the count sequence: as a sound with no meaning. They'll get the worksheet right. They'll get next year's worksheet wrong. Their parents will be confused.

The progression that actually works looks roughly like this:

1. Subitize — recognize small quantities at a glance.
2. Count — say the sequence and match it to objects (cardinality).
3. Count-on — start from a known quantity and add to it.
4. Compare — know that seven is more than five, and how much more.
5. Combine and decompose — see seven as "five and two" or "four and three" or "six and one."
6. Recall — eventually, some facts become instant. Not all. Not on a deadline.

Steps four and five — comparing and decomposing — are where number sense becomes arithmetic. A kid who can decompose seven five different ways doesn't need to memorize 3 + 4. They already know it, because they can see it.

Why early drilling backfires

If you push arithmetic worksheets at a child who hasn't gotten through step three, you don't accelerate them. You build a parallel system in their head: a "school math" system made of sounds and rules, running alongside a "real math" system that hasn't developed yet.

For a while this looks like success. The kid gets the answers. Then, usually around second or third grade, the questions stop being memorizable. Word problems show up. Place value shows up. The child who never built number sense hits a wall, and now the wall has emotional weight: I used to be good at math and now I'm not.

This is the origin story of a lot of the math anxiety that follows kids — and adults — for the rest of their lives. Boaler's work and the broader research on math mindset suggests that the damage from premature drilling is real and durable. We are very good at producing fluent worksheet-solvers who quietly hate the subject.

There is a much better question than "can you do 4 + 3?" The better question is "how do you know?"

"How do you know?" is the whole game

A child who answers "seven" to 4 + 3 might be:

• Reciting a memorized fact (fragile).
• Counting on their fingers from one (still building).
• Counting on from four (good, building number sense).
• Seeing "four and three" as a known combination (excellent).
• Decomposing — "well, three is two and one, and four and two is six, so seven" (gold).

Same answer. Five completely different cognitive states. None of them are wrong. But only some of them mean the child is ready for what comes next.

The standard test — "is the answer right?" — cannot tell you which state your kid is in. And almost no software ever asks.

This is one of the reasons we built Lumi the way we did. Lumi's early-math work, which is rolling out after the reading beta, doesn't just check answers. It asks the child to explain. "How did you figure that out?" "Can you show me with the blocks?" "Is there another way?" It listens — actually listens, with the same low-latency voice stack we use for reading — and the next question depends on what the child just said, not on a pre-built skill tree. If you want to understand why that matters, the post on adaptive learning gets into the mechanics.

The point isn't that Lumi has solved kindergarten math. It is that the right question for a five-year-old is almost never "what is the answer?" — it's "what are you seeing?"

What to actually do at home

You don't need an app for any of this, and I'd argue you shouldn't lead with one. Some things that work, that any parent can do, in roughly the order they get useful:

• Roll dice and don't count the pips. Two dice. Race to call the total. You're training subitizing and conceptual subitizing in the same minute.
• Play with quantities, not numerals. Blocks, beans, beads, raisins. The numeral "7" is an abstraction. Seven raisins is a thing.
• Ask "how many more?" When there are three crackers on one plate and five on another, "how many more does she have?" is a comparison problem dressed up as a conversation.
• Let them use fingers. Don't comment. Just watch which fingers they use and how.
• Ask "how do you know?" Make it the most common question in your house. Not "are you sure?" — that signals doubt. Just curious: how do you know?

If you're a parent watching a kindergartener who can recite to one hundred but can't show you seven blocks, you have not failed. You have just noticed the gap that most of the educational software industry pretends doesn't exist. Closing it is slow work — months and months of small moments — and it is mostly done with your hands, not a screen.

When a screen helps, it should help by asking better questions, not by drilling faster.

Try the Lumi beta when our math module opens — or sign up to be told the day it does.

Image brief

• Hero image: A close-up of small hands sorting colored wooden blocks into groups of three and four on a wooden table, warm natural light, no faces.
• Inline image 1: A simple diagram showing six dots arranged as "three and three" versus six dots in a random scatter, captioned "which one can you see without counting?" — placed after the "Subitizing" section.
• Inline image 2: A clean, friendly progression diagram of the six steps (subitize → count → count-on → compare → combine/decompose → recall) as connected circles, placed after the "counting-on" section.

Internal link suggestions

• "What developmental science says about attention in early learners" — anchor: the same low-latency voice stack we use for reading (or link from the section on "how do you know?")
• "Adaptive learning isn't a setting — it's the whole product" — anchor: adaptive learning
• "The science behind Lumi's pacing" — anchor: optional second mention in the home-practice section, e.g. small moments add up

Editor's note

Two things to confirm before publishing. First, the Maya anecdote is a composite based on a real conversation with a parent in the beta cohort — Tim, please confirm you're comfortable framing it as "a five-year-old I know" rather than naming, or swap in a Remi-specific moment if you have one (the block-counting freeze, in particular, would land harder with Remi as the subject). Second, the Jo Boaler citation about finger representation and brain regions is paraphrased from her public writing on youcubed.org; the specific neuroscience claim (finger-region recruitment in adult arithmetic) is well-established but the exact study should be linked directly if we want to bulletproof it [VERIFY].