Abstract
Previous studies have shown that the discrepancy between performance on "fluid" and "crystallized" intelligence measures may serve as an indicator for intellectual decline. The validity of this procedure in older persons is unknown. The present study developed a multiple regression equation, to predict the Raven Advanced Progressive Matrices (APM) score from the National Adult Reading Test (NART) score and demographic variables in a large sample of healthy older persons (n = 270). The discrepancy between the predicted and observed Raven APM scores was transformed into a percentile distribution as an indicator of intellectual decline, which can be used in clinical practice. The validity of the procedure was further examined by comparing the proportion of persons with a significant decline (at the -1 and -1.65 SD level) between two older patient samples (87 patients with cerebral stroke and 387 patients with diabetes mellitus) by means of χ(2) tests. There was a significantly higher rate of intellectual decline at the -1 SD ("below average") and -1.65 SD ("impaired") cutoff levels for patients with stroke compared with patients with diabetes (stroke, 34% and 14%; diabetes, 16% and 5%, p < .05). These findings suggest that the Raven APM-NART discrepancy may be a useful measure of intellectual decline in older persons.