CogAT Levels 13/14 common mistakes
CogAT Levels 13/14 · Understand the traps and how to avoid them.
On picture/verbal analogy items, the candidate picks an answer that is merely associated with the prompt pair (e.g., 'dog-bone' because dogs like bones) instead of one that reproduces the same relationship in the same direction.
Younger candidates reason from real-world story knowledge rather than from the abstract relation; the strongest-feeling association wins over the structurally correct one. The CogAT picture and verbal analogies reward relational matching, not familiarity.
What to do instead
Teach a two-step routine: first say the relationship aloud as a full sentence ('A ___ is a kind of ___', 'A ___ is used to ___'), then test each option against that same sentence. Before choosing, check direction - if the prompt goes whole→part, the answer must also go whole→part. Practice with pairs where the associative distractor is deliberately tempting.
On sentence-completion items the candidate inserts a word that sounds sophisticated but does not fit the grammar or the logic of the sentence, or picks a word matching only one half of a two-blank item.
Candidates substitute 'which word sounds most advanced' for 'which word makes the sentence coherent'. On double-blank items they evaluate each blank separately and stop at the first plausible fit instead of confirming both blanks work together.
What to do instead
Cover the options, read the sentence, and predict your own word first. Then check each option against the whole sentence, including signal words (although, because, therefore) that set up contrast or cause. For two-blank items, eliminate any option where either blank fails, and re-read the completed sentence once before marking.
On verbal classification items the candidate groups by a superficial feature - same first letter, same number of syllables, things that appear in the same place - rather than by the category the item is testing.
Beginning readers over-weight orthographic and phonological features because decoding is still effortful; 'all start with B' feels like a pattern. The test intends meaning-based categories (all are vehicles, all are emotions).
What to do instead
Drill a hierarchy of checks: meaning category first, then function, then physical attribute. Explicitly practise rejecting letter/sound groupings by asking 'would this group still make sense if the words were in another language?' If yes, it is a meaning category; if no, it is a surface trick.
The candidate answers the quantitative analogy by computing the prompt pair once, then choosing an option that 'looks similar' rather than verifying the same operation applies to the answer pair.
Number analogies require identifying an operation (×3, +7, square) and applying it to a new pair. Candidates often stop after finding any relationship in the prompt and never test the candidate answer, or they accept a pair that works only in reverse.
What to do instead
Force the written rule: 'the first number times three equals the second'. Then plug each option into that exact rule and reject any pair that needs a different operation. Include reverse-direction traps in practice so the candidate learns to check order, not just magnitude.
On number series the candidate finds one gap, assumes a constant difference, and commits before checking whether the pattern is alternating, increasing, or two interleaved sequences.
Constant-difference series are the first pattern taught, so it becomes the default hypothesis. Mixed series (e.g., +2, +4, +6 or two alternating rules) break that assumption, and candidates rarely test a second hypothesis after the first yields a near-miss.
What to do instead
Require a written difference row under every series and a second row of second differences when the first row is not constant. Teach the three common shapes - constant step, growing step, alternating/interleaved - and have the candidate name which shape fits before selecting an answer.
On quantitative classification the candidate picks the number that 'seems odd' by magnitude (largest, smallest) instead of testing a shared property such as divisibility, parity, or being a perfect square.
Without a systematic property checklist, magnitude is the most visually salient difference. Items are built so the outlier shares no obvious size feature, and the three group members share a non-obvious arithmetic property.
What to do instead
Run a fixed checklist on the three candidate group members: even/odd, divisible by 3 or 5, prime, perfect square, digit sum. The property that all three share identifies the rule; the option lacking it is the answer. Time the checklist so it becomes automatic.
On figure analogies and figure classification, choosing the option that looks most similar to the stem figures overall instead of applying the specific transformation shown.
Nonverbal items deliberately include options that share many visual features (same shape, same fill) but omit the one changed attribute; holistic 'looks alike' judgment favors these decoys.
What to do instead
Name the transformation in words before scanning options ('the outer shape is removed and the inner shape is shaded'). Then check each option against that single stated rule and eliminate any that violate it, regardless of overall resemblance.
On paper folding items, tracking the punched holes but not the fold lines, so the answer shows the right number of holes in the wrong arrangement.
Candidates mentally punch through the paper without reversing each fold; the reflection across the last fold is the step most often skipped, and symmetric distractors make a wrong arrangement look plausible.
What to do instead
Work backward from the last fold: unfold one crease at a time and mirror the hole set across that crease before moving to the previous fold. Count holes at each unfold and check the total against the options; a count mismatch eliminates options immediately.
On figure matrices, solving only the row rule and ignoring the column rule, or assuming the same rule applies in both directions.
Matrices are constructed so that one direction carries one rule (for example, add a side) and the other carries a different rule (for example, rotate); a candidate who finds one rule stops searching and picks an option that satisfies only that rule.
What to do instead
Always derive a row rule and a column rule separately, write both down, and accept an option only if it satisfies both. If only one rule is found, use it to eliminate and then re-examine the remaining options for the second dimension.