A fundamental problem in clinical neuropsychology is the estimation of premorbid levels of cognitive functioning. Lezak (1983) described the so-called 'best performance method' of estimating premorbid abilities. Essentially, this method consists in using the highest test scores or the best performance in everyday tasks as the best estimate of premorbid ability. This paper describes three studies of the empirical consequences of using the 'best performance method': the first study demonstrates that application of this method on WAIS subtests leads to a gross overestimation of intelligence as measured by the IQ. The second study corroborates this finding and also shows that overestimation of intelligence leads to systematic errors in the expected performance in neuropsychological tests in normal subjects. Finally, the third study shows that this is also the case in patients with diffuse cerebral atrophy.