Investigators of schizophrenic cognition often produce 2 or more tasks of differing difficulty levels by manipulating a variable that affects the accuracy of both normal and schizophrenic individuals; the investigators find that the variable also affects the difference between the groups in accuracy and conclude that the variable taps a schizophrenic differential deficit. An alternative hypothesis is that task differences in true-score variance artifactually produce the finding. For free-response tasks, group differences tend to be larger when difficulty is near 50%. The authors illustrate a new method of controlling this artifact by selecting items for hard and easy tasks on opposite sides of 50% difficulty and equidistant from it. Using this design with an anagram task, they found that schizophrenic and normal individuals differ no more on hard anagrams than on easy ones, and they propose the design for testing hypotheses concerning schizophrenic deficit on tasks that differ in difficulty.