This paper considers, from a theoretical point of view, two measurement approaches used in measuring success and failure in skills tests in physical education. The first, "fixed length" (FL) testing, entails counting the number of successful performances in a fixed number of trials. The second, "trials-to-criterion" (TTC) testing, involves counting the number of trials required to achieve a specified number of successes. TTC measurement results in high measurement error variance for individuals with low probabilities of success on a single trial. Error variance declines as the probability rises. If there are many more people with low probabilities than there are with high probabilities, which is the case for a positively skewed distribution, the TTC approach will result in less reliable measurement than will the FL approach. Under the latter, error variance is largest for people with a probability of .5. Individuals lower and higher will have smaller error variances. Two generalizations based on these results can be made with regard to skills testing: (1) If the skills test task is one on which most untrained individuals perform poorly, FL testing would be the better choice; and (2) If the test scores tend to be negatively skewed, then TTC testing would be more efficient and reliable for the same total testing time. Two formulas are presented for estimating the reliability of TTC measures. (JM)