Odd-one-out / classification of figures by shared or distinguishing attributes
What this domain tests
You see five figures and must identify the one that does not belong with the other four. The shared property is never stated, so the task is hypothesis generation plus verification: propose a property, check it against all five, and reject it unless it holds for exactly four.
Use a fixed checklist
Work through attributes in a set order so you do not wander:
- Number of sides, vertices, enclosed regions, intersections, separate shapes.
- Straight versus curved elements; open versus closed outlines.
- Lines of symmetry and order of rotational symmetry.
- Handedness: are the figures rotations of one another, or is one a mirror image?
- Shading: filled versus outlined, number of shaded parts, position of shading.
- Positional relations: dot inside or outside, shapes touching or overlapping. Start with counts, because integers decide quickly and unambiguously.
Worked example
Five arrow-like figures, each with a notch cut into one side of the shaft. Four of them are the same arrow shown at 0, 90, 180 and 270 degrees of rotation; their notches sit on the same side relative to the arrowhead when you follow the outline. The fifth arrow has the notch on the opposite side: it is the mirror image, not a rotation. It is the odd one out. The distractors work because the mirrored arrow looks superficially similar at a glance, so you must trace an asymmetric feature rather than compare overall silhouette.
Traps that cost marks
- Accepting a rule that fits all five figures; it cannot be the discriminator.
- Accepting a rule that fits only three; recount before committing.
- Mistaking a rotated figure for a mirrored one, which is the commonest error in this item type.
- Counting the outer background as an enclosed region.
- Ignoring a second attribute that splits the set differently from the one you found first, which usually signals you miscounted.