University students' self-reports indicate that they often solve basic subtraction problems (13 - 6 = ?) by reference to the corresponding addition problem (6 + 7 = 13; therefore, 13 - 6 = 7). In this case, solution latency should be faster with subtraction problems presented in addition format (6 + _ = 13) than in standard subtraction format (13 - 6 = 3. In Experiment 1, the addition format resembled the standard layout for addition with the sum on the right (6 + _ = 13), whereas in Experiment 2, the addition format resembled subtraction with the minuend on the left (13 = 6 + _. Both experiments demonstrated a latency advantage for large problems (minuend > 10) in the addition format as compared with the subtraction format (13 - 6 = _), although the effect was larger in Experiment 1 (254 msec) than in Experiment 2 (125 msec). Small subtractions (minuend < or = 10) in Experiment 1 were solved equally quickly in the subtraction or addition format, but in Experiment 2, performance on small problems was faster in the standard format(5 - 3 = _) than in the addition format(5 = 3 + _). The results indicate that educated adults often use addition reference to solve large simple subtraction problems, but that they rely on direct memory retrieval for small subtractions.