EPSO Abstract Reasoning common mistakes
EPSO Abstract Reasoning · Understand the traps and how to avoid them.
Tracking only the element that moves most obviously (usually the largest shape) and assuming the smaller marks follow no rule, so the sequence appears to break when the small element actually progresses by a fixed step each frame.
Attention is drawn to the most salient change; candidates form a hypothesis from one element and then try to force the remaining frames to fit it, rather than cataloguing every element separately.
What to do instead
Before choosing an answer, list each distinct element (large shape, small dot, arrow, internal line) and write its state frame by frame. Confirm each element has its own consistent rule; if one element's rule fails on one frame, that answer option is eliminated.
Assuming a rotation is a constant 90 degrees when the figure actually rotates by 45 degrees, or alternating clockwise/counter-clockwise, and then rejecting the correct option because it does not match the assumed constant step.
The most familiar step size is over-applied, and small orientation changes on near-symmetric shapes are hard to see without a reference point.
What to do instead
Anchor orientation to an asymmetric feature (a notch, a dot, a short stem) and measure the change between consecutive frames. Test whether the step is constant, increasing, or alternating in direction before matching the next frame.
Solving a 3x3 matrix row by row and applying that rule to the third row, when the true rule operates down the columns or across the whole grid (for example, each row and each column contains one of each attribute).
Rows are read first by habit, and a rule that fits the first two rows often appears to fit the third by coincidence on a single attribute.
What to do instead
Derive the rule from rows, then verify it on columns, then on diagonals if the grid allows. Only accept a rule that holds for every complete line of the grid, not just the line you used to find it.
Classifying figures by overall visual similarity ('these all look pointy') instead of testing a countable or structural attribute, and picking the odd one out on a superficial impression.
Global shape gestalt is processed faster than counting sides, intersections, or enclosed regions, so the first impression dominates.
What to do instead
For odd-one-out items, run a fixed checklist across all figures: number of sides, number of enclosed regions, number of intersections, symmetry axes, open versus closed lines, straight versus curved. The odd figure must be the only one that fails a property shared by all others.
Treating an analogy as 'apply the same visual change' without isolating which specific transformation links A to B, so a distractor that changes the wrong attribute (for example, size instead of position) is chosen.
Multiple changes usually occur between A and B, and candidates match the most noticeable one rather than the complete set.
What to do instead
Describe the A-to-B change as an explicit ordered list (rotation by X, element added, fill inverted, position swapped). Apply exactly that list to C and check the result against each option attribute by attribute.
In superposition items, adding overlapping lines instead of cancelling them, or cancelling lines that merely touch, producing an answer with the wrong number of segments.
The default assumption is that overlaying means combining; the cancellation convention must be inferred from the example and is easily misapplied at junctions.
What to do instead
Establish the convention from the worked pair first: do coincident lines disappear, or do only lines present in one figure survive? Then process the target line by line, marking each segment as present in both, in one only, or in neither.
Mistaking a reflection for a rotation because the shape is nearly symmetric, and accepting an option that is a mirror image when the rule requires rotation (or the reverse).
For shapes with one or more symmetry axes, a rotated copy and a reflected copy can look almost identical at small step sizes.
What to do instead
Check chirality: pick an asymmetric detail and note whether it is clockwise or counter-clockwise from another feature. Rotation preserves that handedness; reflection reverses it. Use this test before judging angle.
Counting sides or enclosed regions only on the outer boundary of a composite figure and ignoring internal lines that create additional regions or intersections.
The outline is perceived as the figure, while internal structure is treated as decoration rather than data.
What to do instead
Count systematically: trace the outer boundary for sides, then count every enclosed area including small ones formed by internal lines, then count intersections. Write the numbers down rather than holding them in memory.
Spending a disproportionate amount of time on one hard series or matrix and then rushing the remaining items, producing errors on easier questions that would have been answered correctly with normal pacing.
Sunk cost on a single item plus the pressure of a timed, negatively scored or tightly paced test; the perceived difficulty of one question distorts time allocation for the whole section.
What to do instead
Set a hard per-item time cap (for example 60-75 seconds). If no rule is confirmed by then, mark a provisional answer, move on, and return only if time remains. Practise with a timer so the cap becomes automatic.
Checking only whether the chosen option satisfies the rule and not whether it is the only option that does, so a distractor that also fits the partial rule is selected.
Confirmation bias: once a plausible rule is found, the first matching option ends the search, and remaining options are not tested against the rule.
What to do instead
After selecting an option, apply the rule to every other option and confirm each one violates it. If two options satisfy the rule, the rule is incomplete and must be refined with an additional attribute.