Power Tests

Definition and Application

The term power test, together with the opposite concept of speed test, was first proposed by Harold Gulliksen in his 1950 book, Theory of Mental Tests. To correctly measure the ability of the test takers and assess individual differences among them, it is crucial to apply a power test with an adequate difficulty level for the group under assessment. In a power test, most test takers should be unable to solve a large proportion of the items, and a few items should be solved only by a few test takers, or even not solved by anyone. Ideally, none of the test takers should be able to correctly answer all of the items because this may indicate a ceiling effect for those who do.

One example of a pure power test would be an English vocabulary examination in which a list of words is presented to a group of sixth-grade students who then have to find a synonym for each word without using any external help. If the words are adequately chosen, most students should be easily able to find synonyms for some of them, others should be moderately difficult for the students, and most students should be unable to find a synonym for the most difficult words. Importantly, the students should have a generous time limit for completing the test (e.g., 2 minutes per word on average), so the results yield a pure measure of English vocabulary knowledge, uncontaminated by individual differences in word fluency, processing speed, and reaction time.

The Raven’s progressive matrices test is often used as an example of a power test for evaluating mental ability. It is intended to measure inductive and visual reasoning, and it is composed of a list of items of increasing difficulty. Because there are different versions for children, the normal adult population, and the above-average adult population, the exact number of items and time restrictions depends on the test version. But, under standard conditions of application, examinees are given more than 1 minute per item in every version. Even if the test takers have generous time restrictions, the fact that the last items are very difficult and the test is often completed sequentially implies that many test takers do not have time to consider all items. Thus, although there is a consensus on considering it a “nonspeeded” test, individual differences in speed may still have a minor effect on the examinees’ scores.

...