Effects of Practice on Aptitude and Achievement Test Scores

40 studies

Students taking aptitude and achievement tests, school-age through adult

Meta-analytic synthesis of studies in which the only manipulation is taking a practice form of the test. No instruction is involved anywhere in the treatment, which makes the design unusually clean for the question it answers - the effect cannot be a knowledge effect because no knowledge was supplied. Graded C for the usual pooling-of-mixed-primaries reason.

Students raise their scores on aptitude and achievement tests simply by taking practice forms of the tests, with no teaching at all. One practice trial on an IDENTICAL form buys +0.42 SD; one trial on a PARALLEL form buys +0.23 SD - about 6 versus 3 IQ-scale points, or 4 versus 2 grade-equivalent months. That gap is the finding, not a footnote: if the gain were growth in the underlying skill it would not care whether the practice form was the same one, and it cares a great deal. Roughly half the practice effect is specific to the particular test. Three moderators structure the effect and each is directly relevant to reading a child's repeat score. First, gains are LARGER when the identical form is used for practice and for the criterion, and smaller with parallel forms - so how much of a retest gain is real depends on how close the practice test was to the real one. Second, the effect increases with the NUMBER of practice tests taken. Third, and least expected, gains were LARGER for high-ability subjects than for low-ability ones - practice effects are not a floor phenomenon; the ablest children capture the most from them. That last result cuts directly against the intuition that a high scorer has nothing left to gain from familiarity.

Low. The manipulation is administering a practice test; the outcome is the same test later. There is no selection into treatment in the controlled subset.

Reproduced in the adult retesting literature and consistent with the school-age test-taking-skills trials; that scores rise on retest without instruction is not disputed.