The Netherlands' secondary education system is highly differentiated, with four different school types for four scholastic ability levels. Final examinations must accommodate these four levels, and require a test-independent definition of the intended final ability levels as well as a sample-free evaluation of the range of ability levels at which a particular test will measure with sufficient accuracy. A method for locating, on a single scale, the ability distribution of a given population as well as the test's optimal reliability level is proposed and illustrated. The method is demonstrated using standardized tests of foreign language listening comprehension for the two highest levels of secondary education in the Netherlands. The languages tested are French, German, and English. Statistical analyses of the test results for each group are presented and discussed, and implications for test use are examined. (MSE)