If you keep drilling a CAT topic you have already mastered, the extra practice buys you almost nothing by the time the exam actually arrives. Rohrer, Taylor, Pashler, Wixted and Cepeda (2005, Applied Cognitive Psychology, 19, 361-374) ran two experiments on 218 college students and found that "overlearning" - continuing to study material after you have already nailed it once - produced a real recall advantage at one week, and that advantage declined by roughly two-thirds by the next retest. In the second experiment, students who spread the same study time across more material, mastering each item less completely, recalled more items in absolute terms than the overlearners, at every retention interval tested. CAT prep is a 3-to-26-week project. A benefit that survives one week is not a strategy for it.
What the study actually did
The paper defines overlearning precisely: once you reach one perfect trial on a piece of material, any further study in that same session is overlearning. It is a genuinely common study behaviour - re-solving the arithmetic set you already get right, re-reading the RC passage type you already read well - and the paper's own introduction notes it is "widely advocated" in education and training literature, generally on the assumption that continued practice "prevents significant losses in retention."
Experiment 1 had participants learn city-country pairs (geography facts) to a criterion of one perfect trial. One group ("Hi Learners") then kept studying for roughly four times as long as a second group ("Lo Learners"), who stopped near criterion. Both groups returned for a cued-recall test one, three, or nine weeks later.
| Group | Recall at week 1 | Recall at week 9 | Decline |
|---|---|---|---|
| Hi Learners (overlearned) | 70% | 24% | ~two-thirds |
| Lo Learners (stopped near criterion) | 31% | 17% | less than half |
The Hi Learners recalled more at every interval - overlearning is not worthless. But the gap between the two groups was statistically significant at one week (t = 4.42) and no longer significant by the final test (t = 0.69 at week nine in one analysis, t = 0.80 in another restricted to the strongest and weakest learners specifically). Four times the study effort bought a real, measurable edge that had essentially disappeared by the time a CAT-length prep window would have run its course.
Experiment 2: the more striking result
The second experiment removes any ambiguity about what "more effort" actually buys you. Participants had a fixed amount of study time. One group used it to overlearn a 10-word list of unusual word-definition pairs (e.g. "cicatrix - scar"); the other group spread the identical time budget across a 20-word list, reaching a lower degree of mastery per word. Recall was tested one and four weeks later.
| Group | Studied | Words recalled, week 1 | Words recalled, week 4 |
|---|---|---|---|
| Overlearners | 10 words | 6.4 | 2.2 |
| Broader coverage | 20 words | 7.5 | 3.5 |
The group that covered twice the material, at a lower mastery threshold per item, recalled more items in absolute terms at both retention intervals - not just proportionally competitive, actually ahead in raw count. The paper's own summary is blunt: "the underlearning of 20 words proved superior to the overlearning of 10 words at both retention intervals." Same time budget, more topics touched, better outcome months later - or in this case, weeks later, which is the more alarming part.
What this means for a CAT prep cycle specifically
CAT prep runs from three to six months for most aspirants, and the retention intervals this study tested - one to nine weeks - sit squarely inside that window rather than at its edge. If you have a topic you already answer correctly under time pressure - a favourite Arithmetic sub-type, a DILR set format you have solved a dozen times, an RC passage genre you read fast - the marginal hour spent re-drilling it is competing against the same hour spent on a genuinely weak area, and this data says that trade loses by the time the exam is months away, not days.
This is a different claim from interleaving quant practice, which is about mixing problem types within a single practice session so you have to select a strategy rather than execute one on autopilot. Overlearning is about a longer-horizon allocation question: given a fixed number of hours across your whole prep timeline, how many of them should go to topics you have already cleared versus topics you have not. The two findings point the same direction - stop rehearsing what already works - for different reasons at different timescales.
Where this genuinely does not apply
- Short-horizon needs are the explicit exception. The paper itself says overlearning "might be ideal for learners who seek only short-term retention" - its own example is "a student with an exam later in the day." If you are the night before a mock or the morning of the actual CAT, cramming an already-strong topic for a same-day payoff is exactly the situation this data supports, not contradicts.
- This is not the same claim as "mastery doesn't matter." The paper distinguishes overlearning (a study strategy - continued practice past one perfect trial) from mastery itself, which it notes elsewhere in the literature does boost long-term retention when reached efficiently. The finding is about the marginal extra rehearsal past that point, not about whether reaching competence matters at all.
- Both experiments used isolated factual material - geography pairs and word definitions - learned and tested in a controlled lab setting, not CAT-style applied problem-solving under time pressure across a real multi-month syllabus. The generalisation to "don't over-drill a mastered CAT topic" is a reasonable extension of the mechanism, not something the study tested directly on exam-style material.
- The paper reports one contrary study in its own literature review (Kratochwill, Demuth & Conzemius, 1977) that found "virtually no benefit" of overlearning even at short intervals - the effect is not universal even in the window where it is strongest.
What to actually do with a weak-topic list
The practical version of this finding is closer to a resource-allocation rule than a technique: once a topic type reliably clears your own accuracy bar under timed conditions, redirect the next study block to whatever is still below it, rather than to more repetitions of what already clears it. A single error log across mocks - which topic, which mistake, how often it recurs - is what actually tells you which side of that line each topic sits on; guessing from how a topic feels under study conditions is exactly the failure mode this paper's numbers describe, since the overlearners in both experiments felt more prepared and tested better within a week of feeling that way.
Frequently asked questions
End of essay
- Anish Guruvelli