How to practise reasoning and learn from mistakes

Choose an observable skill

Instead of aiming to become more intelligent in a week, choose a specific task: Recognise a rotation rule, organise information in a word problem or explain why you rejected an alternative. An observable goal lets you review your process without needing an IQ number.

Work with different examples and briefly record what you did. A sentence such as I checked the rows before the columns gives you something useful for the next session. Copying the whole solution can leave you uncertain about which part you understood and which part you merely recognised.

Try retrieving and explaining

In learning research, Roediger and Karpicke examined how retrieving information could improve later retention of texts compared with studying them again. That is evidence about learning materials. It is not a demonstration that practising matrices increases IQ.

A simple study routine is to explain an idea without looking, check what is missing and try again later. For a logic problem, describe the rule in your own words and see whether it also explains a new example. If it does not, revise your interpretation rather than memorising only the picture.

Distinguish improvement from transfer

Getting better at a game first demonstrates improvement at that game. A review by Simons and colleagues found substantially stronger evidence for gains on trained tasks than for broad gains in different activities. We therefore do not recommend programmes or products as a guaranteed way to increase intelligence.

If you are preparing for an examination or recruitment assessment, check the official format and use relevant exercises. Understanding numerical operations differs from learning vocabulary or getting used to a platform. Useful preparation recognises those differences rather than treating every cognitive task as interchangeable, or assuming that success in one area must carry over everywhere.

A short session with a purpose

Choose a small set of exercises, attempt them without consulting the solution and leave time to examine mistakes. Stop when the session is no longer useful. Next time, combine a familiar example with a new one and write down what changed in your strategy. This is a practical suggestion, not a validation protocol or a promise of progress. When reflecting on our assessment, preserve the distinction between practice and a first attempt. Note previous exposure so that a later comparison remains honest about what happened.

An original example: Change one feature

Draw three arrows pointing up, right and down. One possible continuation is left if the proposed rule is a 90-degree turn at each step. Check it with another sequence starting at right: It should continue down, left and up. This invented example demonstrates proposing and checking a hypothesis; it is neither a pilot item nor a measurement.

Next add another feature, such as alternating outline and solid fill. Examine orientation and fill separately before combining them. An explanation fitting only one step has not yet explained the whole sequence. When information is insufficient, several continuations may be plausible; do not treat an intuition as a unique rule before checking it.

An error log you can actually use

For each exercise, record four things: My initial rule; something it failed to explain; my revised rule; and a new example to test it on. Describe the error specifically: I confused the turn, forgot a component or misunderstood the instruction. That is more useful than writing I am bad at logic because it identifies something you can revisit.

At the next session, attempt a different problem before consulting your notes. If you only recognise the old answer, record recognition; if you can explain and apply the rule to another case, record that particular achievement. This log is an educational suggestion from the project, not a validated way of estimating IQ changes. The preparation guide addresses a different question: Arranging a session without instruction or device difficulties.