The X-zero distinction has been used as the operational definition of MMPI item subtlety in previous research. A consistent finding has been that the X items, which are obvious in terms of pathological content, can be faked whereas the more subtle zero items cannot be faked. The present study examined the relationship between instructional sets and endorsement patterns, using a more refined, comprehensive measure of item subtlety. Sixty (30 male, 30 female) subjects completed two MMPI protocols, one under a standard instructional set and the other under either a fake-good or fake-bad response set. Order of administration and sex were counterbalanced. In general, the findings for the sublte-obvious dimension paralleled the previous findings with the X-zero distinction. Endorsement of obvious items was a direct function of instructional set, whereas endorsement of sublte items was inversely related to instructional set. Because subtle items did not appear to measure the constructs for which they had been originally intended, other uses for subtle items were discussed.