The main objective of this paper is to estimate and compare the return to investments in education for men belonging to different parts of the ability distribution. Our ability measure is the achievement test score from the Swedish Military Enlistment Battery. By exploring the measurement error in the test score, i.e. the deviation from the true latent ability level of the individual, we try to predict the expected biases in the "ability specific" returns to education, when using an achievement test as a proxy for ability. We find that a higher score on the Swedish Enlistment test is associated with a higher return to schooling. The relationship between the return to schooling and the test score does however seem to be decreasing in the test score. Thus, it it primarily the ability specific return to schooling for the lower test groups that divert from the average return to schooling. In general, the measurement errors in the test score do not seem to bring about any major biases in the ability specific returns to schooling.