Bibliography: pages 130-139. / The utility of the National Adult Reading Test (NART) as a predictor of premorbid intellectual functioning is dependent on its ability to adequately predict IQ from irregular word reading ability. The primary aim of this study is the replication, on a South African sample, of the findings reported by Nelson (1977 unpub. manuscript) in her standardisation study. A total of 234 subjects are divided into groups according to language usage and availability/type of IQ score and utilised in a correlational study which investigates the psychometric characteristics of the NART and the degree of correspondence between predicted and observed I.Q. values. The research yields regression formulae for the prediction of IQ from performance on the NART. While correlation coefficients obtained for English speaking subjects do not differ significantly from those derived from Nelson's data (1977 unpub. manuscript), it is concluded that the use of the test for Afrikaans subjects is not justified. Satisfactory reliability and validity characteristics reported for the sample utilised in this study suggest that the instrument can be usefully applied under local conditions. An investigation of 24 potential additional items for the test does not yield encouraging results. The lack of an adequate model explaining the mechanisms underlying the functioning of the test is addressed, and a number of areas of interest for further research are identified.