Solve a visual odd-one-out question by comparing the same visible features across every figure, stating a simple rule that groups most of them, then checking that exactly one figure breaks it. Start with count, orientation, enclosure, connection, shading, or position; do not rely on a vague sense that one shape looks different. Use the answer choices to check your rule, not to invent one after choosing. If two equally simple rules identify different figures, the item may be ambiguous. This method makes your reasoning checkable, but research on related matrix puzzles does not prove one strategy works best for every odd-one-out item.
What should you compare first?
Begin with features you can name consistently: the number of parts, whether a line is open or closed, how many times lines meet, direction, shading, or the position of a mark. A visual feature is a property you can inspect in each figure using the same question. For example, ask “How many enclosed spaces are present?” of every figure, rather than describing one as “busy” and another as “simple.”
Imagine four original line drawings: three contain one closed loop, while the fourth is an open line with no enclosed space. “Has one enclosed region” is a candidate rule for grouping the first three. It is not automatically the intended rule; first check that the drawings do not differ in another equally salient way, such as number of strokes or orientation. The point is to compare like with like, not to treat this illustration as a test item.
A directly relevant study by Yousif examined odd-one-out judgments about geometric and topological features. In one experiment, children aged four to seven saw six line forms and chose the one that differed in network properties such as holes or junctions. A later experiment tested selected topological and geometric properties with children, alongside an adult comparison sample. Performance varied by feature, and the authors noted that stimulus details affected how discriminable some items were. This supports checking which property a particular set makes visible; it does not establish a universal adult solving recipe or show that any single puzzle measures general intelligence.
Should you state a rule before using elimination?
Usually, yes: make a provisional grouping rule before rejecting figures one by one. In research on matrix reasoning, constructive matching means forming a predicted answer from relations in the problem before inspecting choices. Response elimination means checking options in turn and ruling out those that do not fit. Matrix tasks ask for a missing figure; odd-one-out tasks ask which member breaks a grouping. They are related visual reasoning formats, not interchangeable evidence.
The practical reason to state a rule first is that it gives you something to test. Suppose three figures each have two joined parts and a fourth has three. The candidate rule is “the grouped figures have two joined parts.” Check all four figures against that same count. Then use elimination as a second pass: if the proposed exception is removed, do the remaining figures still share the stated feature? If the answer depends on changing the rule halfway through, pause and compare alternatives.
Niebaum and Munakata’s primary study makes the matrix evidence more specific. It used eye tracking while 38 six-year-olds, 43 nine-year-olds, and 51 college-aged adults completed matrix-completion problems; the researchers inferred constructive matching from scanning matrix rows and columns and response elimination from consulting answer choices. Across ages, more row-and-column scanning predicted better overall performance, while quicker and more frequent consultation of potential answers predicted poorer performance. As problems grew harder, participants increased matrix scanning; adults and nine-year-olds also consulted potential answers more. These are associations in a particular matrix task, not a causal test of teaching rule-first, and not an odd-one-out trial. The age groups completed different item sets, and some participants were excluded from particular analyses. So this supports a cautious comparison of strategies in matrix completion, not a claim that rule-first always wins in odd-one-out questions. Elimination can still be a useful second pass when the rule is simple or an initial grouping does not settle the answer.
How do you test whether your answer is actually supported?
A strong answer explains the whole set with one rule. Say the rule in a short sentence, apply it to every figure, and verify that only one figure fails it. Avoid a post-hoc explanation that describes only your chosen figure. For instance, if three symbols each have one enclosed space and a fourth has none, count enclosed spaces in all four. Do not switch to color or orientation only for the figure you want to call odd.
Next, try a competing rule of similar simplicity. A set might contain three figures with an even number of strokes and one with an odd number, while also containing three upright forms and one rotated form. If both rules are equally visible and point to different exceptions, the drawing alone may not specify a unique answer. A well-made item should constrain the intended distinction enough that one explanation fits the full set more cleanly than its rivals. Your uncertainty in that case may reflect the item, rather than a failure to spot a hidden truth.
The odd-one-out study is a useful caution here: even in controlled research, the particular visual feature and how clearly it appears in the stimulus matter. A separate adult eye-tracking study used a specific matrix instrument with university students; its authors describe limits from small samples and restricted generalizability. Those studies investigate task performance and strategy under set conditions. They do not validate a new puzzle, establish a score for an individual, or show that a browser quiz is equivalent to a professional assessment.
Sources: Children’s Understanding of Topological Relations; Eye Gaze Patterns during Reasoning Provide Insights Regarding Individual Differences in Underlying Cognitive Abilities
What can practice on these questions tell you?
Practice the observable steps: inventory features, compare the same feature across figures, state a grouping rule, test every figure, and check whether another rule fits just as well. When reviewing an answer, write down the feature that separated the exception and note whether you missed it, counted inconsistently, or found the item unclear. That record can guide practice on the operation you want to improve. Better performance on familiar or similar puzzles shows performance on those tasks under those conditions; it does not by itself establish a gain in general intelligence.
Verdict: rule-first comparison is a useful, auditable default, and elimination is a cross-check. The qualification matters: available strategy evidence is mainly from matrix-completion tasks, while direct odd-one-out research covers particular stimulus sets and populations. Neither justifies treating one puzzle as an IQ score or assuming that every item has a uniquely compelled answer.
For low-stakes practice, you can take Test IQ Free’s 50-question reasoning test and review its worked answers. Its free result includes a provisional estimated IQ range from an unvalidated conversion with no validated adult reference population; that band is not a confidence interval or an established comparison with the population. Treat it as provisional test performance, not identity, diagnosis, or fixed potential. The optional $9 report adds explanation using the same conversion, not validity. For a consequential educational or clinical decision, use a qualified professional assessment instead.
Sources: The Development of Relational Reasoning: An Eyetracking Analysis of Strategy Use and Adaptation in Children and Adults Performing Matrix Completion; Eye Gaze Patterns during Reasoning Provide Insights Regarding Individual Differences in Underlying Cognitive Abilities
Questions readers ask
What if two figures could be the odd one out?
State each candidate rule and check whether it applies consistently to every figure. If equally simple visible rules point to different exceptions and the item gives no way to choose between them, treat the question as ambiguous rather than inventing certainty.
Sources
- Children’s Understanding of Topological Relations
In an odd-one-out task with 100 children aged four to seven, participants were above chance across tested topological comparisons, but accuracy varied by feature: a one-hole difference was detected more accurately than differences of one or two T-junctions. This supports the narrower point that performance depends on the feature being compared; the study concerns its specified stimuli and child sample, not a universal adult solving method or an intelligence measure.
- The Development of Relational Reasoning: An Eyetracking Analysis of Strategy Use and Adaptation in Children and Adults Performing Matrix Completion
In an eye-tracking study of matrix-completion problems with six-year-olds, nine-year-olds, and adults, indices of constructive matching such as encoding and integration were generally associated with higher accuracy, while increases in answer-choice toggling were associated with lower accuracy. The study concerns matrix completion and strategy-performance associations; it does not test odd-one-out questions or causal instruction to state a rule first.
- Eye Gaze Patterns during Reasoning Provide Insights Regarding Individual Differences in Underlying Cognitive Abilities
This two-study investigation relates eye-gaze measures during a matrix-reasoning task to other reasoning and cognitive measures in university-student samples. Its discussion notes limits on generalizing from the specific task and samples, so it supports treating matrix-task gaze results as bounded evidence rather than validation of an individual puzzle or a general score.
Try the difference yourself.
Work through 50 adaptive reasoning questions, then see your estimated IQ band. Unlock a detailed personalized report after purchase.
Take the free test