The education community regularly leverages psychometric measurements of test quality to evaluate the quality and reliability of multiple choice tests in educational settings. Using a broad collection of actual multiple-choice tests offered at Trent University over the last few years, we examine the standard "item-total correlation" of these tests, purported to measure the discriminatory power of individual questions on the test. This measure is known to be biased, and we discuss the bias and the implications for the rule-of-thumb breakpoints used in most published work in the educational testing community.