The experiment described here was designed to find out which of two methods of testing--blank-filling and translation--gave the more accurate guide to linguistic ability. Two tests were devised to elicit identical correct answers using different means (blank-filling and translation). They were administered to two different classes of Guinean lycee pupils, who were not given notice, or told that the two tests were the "same." To offset the possibility of memory interference with the second test, given one week after the first, the translation was given to Class A first but to Class B second. The results of the tests were plotted on the same graph as the half-year results which had been established as standard for classroom work. From the graphs it was determined how far each pupil had deviated from his half-year standard in each test. The translation scores were roughly twice as near to the half-yearly scores as those of the blank-filling test. Blank-filling, the author contends, cuts down the risk of making mistakes because it aims to test only a very narrow slice of language. He suggests a more thorough exploration of the use of translation as a testing technique. (AMM)