Pattern Recognition in IQ Tests: What It Is and How to Practise It

Pattern recognition is the ability to spot a regularity, such as a repeating sequence, a consistent change or a hidden rule, and use it to predict what comes next. In IQ testing it is the core of most nonverbal reasoning items, from number series to matrix puzzles, because it shows how well you can work out something new rather than recall something you already know. This guide explains how those items work, what they measure, where the idea overlaps with artificial intelligence, and gives you six practice puzzles with answers.

What is pattern recognition?

At its simplest, pattern recognition means noticing that something familiar has turned up again. It might be a trend, a relationship between objects, or a rule that links one item to the next. Once you have noticed the regularity, you can use it to fill a gap or make a prediction.

The term is used in two quite different worlds, which is why searches for it return such mixed results:

• In psychology and IQ testing, it describes a mental skill: looking at shapes, numbers or letters and inferring the rule that connects them.
• In computing, it describes a branch of artificial intelligence and machine learning concerned with finding regularities, trends or structures in large data sets, used for things like facial and voice recognition.

Both meanings share the same basic idea of moving from raw information to a rule. A useful caution applies to both, too: a pattern you think you see is a starting hypothesis, not a proven fact. Part of doing pattern recognition well is checking whether the regularity is genuinely there or is just noise that happens to look like a trend.

Why do IQ tests rely on pattern recognition so much?

Intelligence tests aim to measure reasoning, not schooling. Pattern-based items are attractive because they can be built from abstract shapes or simple symbols that do not depend on vocabulary, reading or general knowledge. That makes them a fairly direct window onto what psychologists call fluid reasoning: solving a problem you have not seen before.

The distinction between fluid and crystallised ability matters here. Crystallised intelligence is the stock of knowledge and skills you have built up, while fluid intelligence is the capacity to reason through novel material. Pattern recognition tasks lean heavily on the fluid side, which is explained in more depth in our guide to fluid vs crystallized intelligence.

Major clinical tests reflect this. The fifth edition of the Stanford–Binet Intelligence Scales, released in 2003, measures five factors: fluid reasoning, knowledge, quantitative reasoning, visual-spatial processing and working memory. These factors are tied to the Cattell-Horn-Carroll model of cognitive abilities, and the test pairs verbal subtests with nonverbal counterparts. Several of those factors, especially fluid reasoning and visual-spatial processing, are exactly where pattern-spotting is assessed. If you want the background on that test, we cover the Stanford-Binet test separately.

This emphasis was not always there. The 1937 edition of the Stanford–Binet was already broadening the range of nonverbal abilities it covered and placing less weight on recall memory than the 1916 version, a shift towards reasoning over remembering.

What does a pattern recognition test look like?

A pattern recognition test, whether stand-alone or part of a wider IQ test, usually uses a handful of item types:

• Matrices. A grid of figures with one cell missing. You choose the piece that completes the grid from a set of options.
• Number series. A sequence such as 3, 6, 12, 24 where you supply the next term.
• Letter series. The same idea with letters, often using their positions in the alphabet.
• Odd one out. Several items share a property; one does not.
• Analogies and codes. A transformation applied to one item that you must apply to another.

The best-known matrix test is Raven's Progressive Matrices. It contains 60 multiple-choice questions that grow harder as you go, it is nonverbal, and it can be given to groups. The task is always to identify the missing element that completes the pattern, and it is designed to measure reasoning ability. A harder version, Raven's Advanced Progressive Matrices, has 36 items and is used to measure fluid intelligence: each item is a 3x3 matrix with one piece missing, chosen from eight options.

Notice two design features. First, difficulty rises steadily, so the early items teach you the format and the later ones separate stronger reasoners from the rest. Second, multiple-choice options are deliberately chosen so that several look nearly right; the wrong answers usually satisfy one rule but break another.

How do people and machines recognise patterns differently?

Because so many pages about pattern recognition are really about software, it is worth being clear on the overlap. In computing, pattern recognition has roots going back to the 1950s in engineering, and grew as machine learning and computing power improved. It is usually described as part of machine learning rather than a synonym for it.

A typical machine system works through stages:

• Collecting or sensing data, such as images, audio, text or sensor readings.
• Extracting features, the characteristics that carry useful information, such as edges in an image or word frequencies in text.
• Detecting patterns, such as clusters, trends or outliers.
• Classifying or deciding, for example labelling an email as spam or flagging a transaction as risky.

A person solving a matrix puzzle does something loosely similar: you scan the grid, pick out the features that change (shape, number, shading, rotation), test a rule, and choose an answer. The difference is scale and training. Software learns from very large numbers of labelled examples, and it can go wrong through overfitting, fitting its training data so closely that it performs poorly on new data. An IQ item does the opposite: it gives you only a few examples and asks you to generalise from them straight away. That is why matrix tests are used to measure novel problem solving in people.

What happens in the brain during pattern-based reasoning?

Researchers have paired scores on tests such as Raven's matrices with brain imaging to look for links between reasoning performance and brain structure or activity. This is an active area, and the findings are correlations rather than a complete explanation.

One example is a 2012 study from Washington University in St. Louis by Cole and colleagues. It found that a region of the lateral prefrontal cortex, an area linked to planning and cognitive control, showed high connectivity across the brain, and that this global connectivity had a selective relationship with individual differences in fluid intelligence. Other work has suggested that people with higher intelligence tend to use more efficient whole-brain network organisation.

The honest takeaway is modest. Pattern-based reasoning seems to draw on networks that coordinate control across many brain areas, rather than on one dedicated pattern centre. Imaging research does not give anyone a brain-scan IQ, and it does not tell you how to train a particular region. If you have concerns about memory, attention or reasoning that affect daily life, that is a matter for a doctor or a qualified psychologist, not an online puzzle.

Practice puzzles: can you spot the rule?

Try each one before reading the answer. Work out the rule in words first, then check it against every item, not just the last two.

1. The gaps are 4, 6, 8, 10, so the next gap is 12.

2. Letter series. A, C, F, J, O, ?

Answer: U. Using alphabet positions (1, 3, 6, 10, 15), the gaps grow by one each time: +2, +3, +4, +5, so the next is +6, giving 21, which is U.

3. Interleaved series. 1, 10, 3, 9, 5, 8, ?

Answer: 7. Two sequences are woven together. The odd positions run 1, 3, 5, 7; the even positions run 10, 9, 8. The next term belongs to the first sequence.

4. All the others are perfect squares (3x3, 4x4, 5x5, 7x7).

5. Code. If CAT becomes DBU, what does DOG become?

Answer: EPH. Each letter moves forward one place in the alphabet.

6. Described matrix. Picture a 3x3 grid. Row one: one circle, two squares, three triangles. Row two: two squares, three triangles, one circle. Row three: three triangles, one circle, ?

Answer: two squares. Each row and column contains each shape once, and each shape always appears with the same count: circles once, squares twice, triangles three times. Two rules are in play, which is typical of harder matrix items.

For more items in this style, our practice puzzles with worked answers go through the reasoning step by step.