Do STEM Toys Actually Work?

What we will and will not claim

This site publishes what manufacturers publish and does arithmetic on it. Where we say what a toy teaches, we mean what it mechanically requires the child to do — sequence steps, complete a circuit, plan two moves ahead — because that we can point at.

We do not claim educational outcomes. Nobody selling a toy is in a position to tell you it will make a child better at mathematics, and we are not either.

The three things that do transfer

Sequencing. Breaking a goal into ordered steps, then finding that the order matters. Botley 2.0 publishes sequences of up to 150 steps; Code Master publishes 60 levels whose later stages introduce conditionals and loops.

Prediction. Saying what will happen before it happens. Gravity Maze publishes 60 challenges where a child predicts a marble’s path and then watches whether they were right.

Debugging. The transferable one. A robot that drives into a wall is a clearer error message than most software produces, and finding the one instruction that was wrong is the habit that carries furthest.

What almost certainly does not transfer

Subject knowledge, mostly. A child who builds 305 Snap Circuits projects learns that a circuit must form a complete loop. They do not learn Ohm’s law, because nothing in the format requires it.

Career direction. No evidence a toy sets one, and the claim is made constantly.

General intelligence. A logic game makes a child better at that logic game, reliably. Anything beyond that is a much larger claim.

The question worth asking instead

Almost every review site in this category answers “which is best?” Very few answer “will it still be out of the cupboard in March?” — which is the question a gift-buyer actually has.

A mediocre toy used forty times teaches more than an excellent one used four times. That is why we check four published properties on every product: whether it is open-ended, whether anything runs out, whether there is room to grow, and whether an adult is needed.

How we pick sets the method out in full.

Where the marketing goes wrong

The word STEM on a box means nothing on its own. It is applied to products with a published 150-step program ceiling and to products whose format is a storybook. Both descriptions can be accurate; the word does not distinguish them.

Educational-outcome language is unfalsifiable. “Builds problem-solving skills” cannot be checked. “Publishes 60 challenges graded beginner to expert” can.

Big numbers are often the wrong numbers. A project count tells you how long a manual is. Projects per part tells you how much each component is doing — Elenco publishes 305 projects from 60+ parts on Snap Circuits Classic, which is 5.1 each.

What a good STEM toy is, honestly

A toy that gets played with repeatedly and happens to require thinking. That is a much smaller claim than the packaging makes and it is achievable, checkable, and worth paying for.

The toys that manage it share published properties: no finished state, no consumables, a ceiling high enough not to be hit, and no requirement for an adult who is never free.

Why we will not cite studies at you

It would be easy to gesture at research showing that play supports development. We do not, for a specific reason: none of it tells you whether a particular box on a particular shelf is worth buying for a particular child.

The gap between “construction play supports spatial reasoning” and “buy this 100-piece set” is enormous, and it is exactly the gap marketing copy exploits. A general finding about a category is not evidence about a product.

What we can do instead is report what a manufacturer publishes and reason about what the format requires. That is a smaller claim and it has the advantage of being checkable.

The one measurable thing

There is a single outcome a buyer can actually observe: whether the toy is still being used in three months.

That is measurable in a way “develops critical thinking” is not, and it is the thing every other claim depends on. A toy that has been in a cupboard since January has taught nothing regardless of what the box promised.

So the four properties we check are chosen for their effect on that one observable outcome. Is it open-ended, does anything run out, is there room to grow, does it need an adult. Those predict March.

Where to go next

Toys that actually teach applies the four-part test in detail. What are STEM toys covers the definition, and how we pick covers our method and its limits.