A perfect score that may not be relational: diagnosing the look-alike strategy in Reasoning by Analogy
Mr. Delgado works one-to-one with a five-year-old on Reasoning by Analogy. The format is 'A is to B as C is to what?', with four answer choices. Five items sit on the table in increasing difficulty. On items 1 to 4 the child answers correctly in under ten seconds. Item 5 shows a red circle as A. B is the same red circle with one wedge removed, like a pie missing a slice. C is a blue square. The choices are: a blue square with a wedge removed, a red circle with a wedge removed, a plain blue square, and a blue square with a wedge stuck onto the outside. The child points at the red circle with a wedge removed and says, 'because it looks like that one.' Delgado checks the key: the child's four earlier correct answers all happened to be the choice that most resembled B, so a pure look-alike strategy would also have scored four out of four. He has eight minutes left.
1. What is the correct choice for Item 5, and what sequence of comparisons produces it?
Read the worked answer
Extract the rule from the A–B pair before looking at the choices.
A is a red circle. B is that same red circle with one wedge taken out of its outline. The change is a removal at the perimeter; colour, size and the basic shape are untouched.
Apply the same change to C, a blue square: the answer is a blue square with one wedge removed from its outline. Only then compare against the choices, and expect three of them to fail for different reasons.
- The red circle with a wedge removed reproduces B instead of transforming C. It answers 'which picture looks most like the middle one,' which is not the question.
- The plain blue square applies no change at all.
- The blue square with a wedge stuck on the outside reverses the direction of the change - the subtlest error on the page, because a child who registers 'wedge' without registering 'removed' will accept it.
A child who scans the choices first and takes the most familiar one will always land on the B-lookalike.
2. How much weight should Delgado give to the four earlier correct answers?
Read the worked answer
Less than the raw score suggests, given how those items were built.
If on each of the first four items the correct choice was also the choice most similar to B, then a child who never forms a rule but always picks the picture resembling the model scores four out of four. That is a scoring artefact: the items are confounded between 'applies the relation' and 'matches by resemblance,' and the four-item streak cannot separate them.
The check is to re-present one earlier item with the B-lookalike replaced by a neutral choice and see whether the child shifts. If she now picks the correct but less similar choice, the relational strategy was there and Item 5 is a genuine failure on a harder rule. If she stalls or reverts to resemblance, the earlier score was carried by the strategy Item 5 exposed.
Either way, Item 5 is more informative than the four items before it, and Delgado should record the pattern rather than a total.
3. Which specific misunderstanding does the wedge-added distractor isolate, and why is it more informative than the plain-square distractor?
Read the worked answer
The plain square catches 'no change applied' - a child who has not engaged with the A–B pair at all, and whose error is one of attention rather than reasoning.
The wedge-added square catches a child who has engaged. She has noticed that a wedge is the operative feature and has even tracked its position on the perimeter, but she has the direction of the operation reversed: added rather than removed. That is a far narrower error, and it separates a partial rule from no rule.
In practice it tells Delgado to slow the comparison step. Ask her to describe in her own words what happened to A's outline, then hide the choices and have her draw what C should become. A child who draws a wedge removed when prompted but selects the wedge added when the choices are visible is being pulled by the choices rather than by the rule - precisely the behaviour the item was designed to surface.