Seven rule families cover almost every number series item on an IQ test, with a fixed check order, worked examples and a 20-minute drill built on 2019 research.
Most people lose number series items in the first fifteen seconds. They read 3, 6, 11, 18, 27, ?, decide the item is a riddle, and wait for a pattern to announce itself. Nothing announces itself. Then the timer runs out.
The item was not built to be glimpsed. It was built from a short list of rules by someone who also wrote the tempting wrong answers. Once you know that list and check it in a fixed order, most sequences stop being puzzles and start being lookups.
Why Number Series Are Easier Than They Look
Cognitive accounts of series completion describe the same four moves every solver makes: detect the relation between neighbouring terms, work out how often the pattern repeats, describe the pattern to yourself, then extrapolate to the missing term. That framework was developed for letter series and later extended to numbers; a 2019 Frontiers in Psychology study on how difficult number series items are built summarizes it before going on to generate new items. The part that matters for anyone sitting a test is what the framework leaves out: none of the four moves requires mathematical talent.
The same paper is blunt about the arithmetic. Its item writers limited themselves to simple operations, because the goal was to build markers of fluid reasoning rather than items of mathematical proficiency. The digits are symbols. Swap them for shapes and the task barely changes.
The other useful admission is that the rule inventory is small. Beyond straightforward arithmetic and geometric progressions, the paper cites four rule families that item generators draw on: addition of earlier terms, Fibonacci-style recursion, interpolation, and multiplication. Publishers are not inventing a new species of sequence for your test session. They are recombining a short list.
That is the whole basis for a checklist.
The 7 Rule Families
Each family below covers a distinct mechanism. If a sequence is not in one of them, it is a combination of two, which is a separate problem and a later one.
1. Constant Difference
Subtract each term from the next. If the gaps are identical, the rule is "add the same amount every time."
7, 12, 17, 22, 27, ? — gaps of 5 throughout, so the answer is 32.
This is the family to check first because subtraction is the cheapest operation available and the gap column is self-terminating: the moment it goes constant, you have the shape of the rule before you have the answer.
2. Constant Ratio
Divide each term by the one before it. A constant quotient means multiplication.
3, 6, 12, 24, 48, ? — each term doubles, so the answer is 96.
If the ratios are not constant but form a tidy sequence of their own, that sequence is the rule. 2, 6, 24, 120, ? has ratios 3, 4, 5, so the next ratio is 6 and the answer is 720.
3. Second-Order Differences
When the first differences drift in a straight line, take the differences of the differences.
3, 6, 11, 18, 27, ? — gaps are 3, 5, 7, 9, rising by 2 each time. The next gap is 11, so the answer is 38.
A constant second difference means a square rule underneath. Here every term is n² + 2: 1 + 2, 4 + 2, 9 + 2, 16 + 2, 25 + 2. Seeing the square form is faster than extending the difference table, and it gives you a second way to check the answer.
4. Two Series in One
If a sequence jumps up and down without settling, read the odd positions and the even positions as separate rows.
2, 40, 4, 35, 8, 30, 16, ? — the odd terms double (2, 4, 8, 16) and the even terms fall by 5 (40, 35, 30). The missing term sits in an even position, so it is 30 − 5 = 25.
This is a periodicity of two: two rules interleaved. If splitting by parity does not produce two clean rows, test a period of three before you abandon the family.
5. Recursive Sums
When the gaps grow roughly in proportion to the terms themselves, check whether each term is built from the two before it.
3, 4, 7, 11, 18, ? — 3 + 4 = 7, 4 + 7 = 11, 7 + 11 = 18, so the answer is 11 + 18 = 29.
Fibonacci is the famous version, but the mechanism is broader: the rule can be "sum of the previous two," "sum plus one," or "previous term plus its own index." The tell is that no fixed gap or ratio survives the first three terms, yet the numbers keep growing smoothly.
6. Index-Linked Terms
Stop comparing terms with their neighbours and compare each term with its position number.
2, 9, 28, 65, 126, ? — these are n³ + 1: 1 + 1, 8 + 1, 27 + 1, 64 + 1, 125 + 1. The answer is 216 + 1 = 217.
The same family includes squares, factorials, the primes in order, and any of those shifted by a constant. 2, 3, 5, 7, 11, 13, ? is just the primes, and the answer is 17. When terms climb far faster than any constant ratio explains, this is the check to run.
7. Digit-Level Rules
Last resort: the rule acts on how a number is written rather than what it is worth.
12, 15, 21, 24, 30, ? — each step adds the sum of the digits of the current term. 12 + (1 + 2) = 15, 15 + (1 + 5) = 21, 21 + (2 + 1) = 24, 24 + (2 + 4) = 30. The next step is 30 + (3 + 0) = 33.
Reversing digits belongs here too: 12, 21, 13, 31, 14, ? alternates a number with its mirror, so the answer is 41. Digit rules are the least portable of the seven, since they assume base-ten notation.
The Check Order That Saves You Time
Order the families by how cheap they are to test, and stop at the first that survives every term.
| Check | What it detects | Stop when |
|---|---|---|
| First differences | Constant step | Gaps identical |
| Second differences | Square or polynomial rule | Second row constant |
| Ratios | Multiplication or factorial growth | Quotients constant or tidy |
| Interleaving | Two rules in one row | Each parity reads cleanly |
| Index links | n², n³, primes, factorials | Terms match position values |
| Digit operations | Digit sums, reversals | Written form explains the step |
Two habits keep this from turning into a fishing expedition.
The first is a self-test: the rule has to be sayable in one sentence. "Add five." "Double it." "The primes." If your candidate rule needs a subordinate clause and an exception, it is not the intended rule — go back one check.
The second is knowing why the item works at all. Any finite list of numbers can be fitted by infinitely many rules, so a sequence never has a unique mathematical continuation. Items function because everyone accepts a convention: the intended answer is the one the simplest fitting rule produces. "Simplest" is a decision rule here, not a matter of taste.
Why Some Items Still Beat You
Difficulty in number series is not about harder arithmetic. The same 2019 study reviews four features that item writers manipulate, following Arendasy and Sommer: periodicity (how many repeats you have to notice), rule span (how many operations link two terms), the number of rules in play, and rule complexity. A two-rule item with a period of two is harder than a one-rule item with bigger numbers.
Working memory is the mechanism underneath. A 2014 study by Little, Lewandowsky and Craig, indexed on PubMed, found that the harder a reasoning item gets, the more working-memory capacity separates stronger from weaker performers — a relationship that only becomes visible once the overall association between capacity and fluid ability is high enough to detect it. Long strings and multiple rules are expensive because you must hold several candidate patterns in mind while testing a new one.
The practical consequence: if you keep failing the last two items in a section, the problem is probably not that you cannot see the rule. It is that you are carrying too many hypotheses at once. Writing the gap column on paper exists precisely to move that load off your head.
A 20-Minute Practice Routine
Three passes, ten items each. Volume is not the point.
Pass one — classify, do not solve. Give yourself five seconds per sequence and say the family out loud: ratio, interleaved, index-linked. You will be wrong sometimes. Being wrong fast is the skill being trained.
Pass two — write the rule before the answer. One sentence, no hedging. If you cannot write it, you guessed. A guess that lands on the right number is worse practice than a clean failure, because it teaches you nothing you can reuse.
Pass three — build two items yourself. Pick a family, choose a starting term, generate six terms, then remove the last one. Building an item forces you to see what the writer was optimizing: a single defensible rule, plus distractors that punish the wrong check. This is the pass most people skip and the one that moves scores fastest.
For reps, the number sequence puzzle game on this site runs the same classify-then-extend loop, and its clock is not there to punish you. If your failures cluster on the working-memory side rather than the rule side, the number sequence memory game trains the other half of the job. The 60-question Raven's matrices test applies identical logic with shapes instead of digits, which is a good check on whether your weakness is numerical or general. For the wider map of item formats, our guide to IQ test question types covers the other seven.
What the Score Does and Does Not Tell You
A number series subtest feeds a fluid-reasoning score, and that score is a standardized comparison, not an absolute measurement. Mensa's explanation of what IQ is puts it plainly: the norming sample is set at 100, and the standard deviation is defined per test, at 16 or 24 points depending on the instrument. Mensa's own entry bar is the 98th percentile, which means 132 or above on the Stanford-Binet and 148 or above on the Cattell.
Two cautions follow. Scores are not portable: the same qualifying standard lands at 132 on one instrument and 148 on another, so comparing your number to someone else's is meaningless without knowing the test. And practice tests are practice. Mensa UK's testing page separates its free online workout, which produces no iq score at all, from the supervised tests that actually qualify a candidate.
So treat a number series result as what it is: a reading on one narrow task, taken on one day, under one set of conditions. It is useful for spotting whether your rule vocabulary or your working memory is the bottleneck. It is not a verdict on how well you think.
