The National Adult Reading Test (NART) is widely used in clinical and research settings to estimate pre-morbid intellectual levels. The validity of the NART in estimating premorbid ability in Alcoholic Korsakoff Syndrome (AKS) is examined in the present study. Twenty AKS subjects were compared with 40 healthy controls. The validity of the NART as a pre-morbid measure in AKS was examined using four methods. AKS subjects made more NART errors than controls, had lower NART predicted IQ than demographically predicted IQ, made more NART errors than predicted by demographic variables and demonstrated NART performance which correlated with degree of memory impairment. It is concluded that NART performance is detrimentally affected by the AKS and that estimating pre-morbid intellectual level in Korsakoff's psychosis using the NART may be invalid. Furthermore, it is postulated that the impaired ability to pronounce correctly irregular words in AKS may reflect a failure in cognitive 'error checking' which may represent a consequence of frontal lobe dysfunction.