Clinically, tests of executive functions tend to be chosen on face validity. If such tests are to be used to evaluate a clinical population, their ability to measure executive functions should be reliably demonstrated in a normal population. In order to investigate the reliability of such tests, a sample of 130 normal adults (74 women, 56 men) ages 17 to 55 years were administered 4 tests purporting to measure planning/problem-solving: the Tower of London Test, the Six Element Test, the Twenty Questions Test, and the Rey Complex Figure Test. A structural equation modeling approach provided by the LISREL 8 program was used to evaluate three models hypothesized to explain the relationship among the test variables and the latent construct of planning/problem-solving. An adequate model was unable to be estimated, thus raising questions about the meaning of the latent construct planning/problem-solving and the psychometric structure of the Tower of London Test. (JINS, 1997, 3, 108–119.)