The present authors assessed a consecutive series of inborn extremely low birth weight children at 2 years and again at 51/2 years of age. The frequency of sensorineural impairment and the results of psychological testing at these ages were then compared to determine the predictive value of the 2-year psychological test scores and the stability of the diagnosis of sensorineural impairment over time. The study included 139 consecutive liveborn infants with birth weights between 500 and 999 gm, bom in the 31/2 calendar years beginning in January 1977. Forty-one per cent (57 of 139) survived the primary hospitalization, but two of them died later of sudden infant death syndrome. The mean (± SD) birth weight of the 55 2-year survivors was 864 ± 99 gm, and the mean gestational age was 27.2 ± 2.0 weeks.) Fifty-four (98.2 per cent) of the 55 survivors at 2 years were assessed at approximately 2 years of age. One child died accidentally. Fifty children were assessed at 51/2 years of age (at least). At 2 years of corrected age, the study children had a mean Bayley Mental Development Index of 91.1, significantly lower than the population mean. At 51/2 years of corrected age, there was a catch-up effect. The children no longer differed from the normal population on the Wechsler Preschool and Primary Scales of Intelligence (full scale, verbal scale, and performance scale). The extremely low birth weight children scored significantly below the test mean for only one of the 11 subtests measuring more specific intellectual abilities. The Bayley Mental Development Index was a significant predictor of Wechsler preschool and Primary Scales of intelligence full scale results (r = 0.700; P = 0.00001, variance explained at 49.0 per cent). Thirty-one children improved, however, between 2 and 51/2 years, as compared with only 14 who had numerically identical or lower scores at the later age. The authors attempted to identify factors responsible for improvement over time in psychological test scores. Increasing duration of maternal education accounted for 12 per cent of variance, and decreased Bayley Mental Development Index for a further 7.5 per cent of variance (P = 0.01). In other words, children with lower Bayley Mental Development Index Scores and with better educated mothers were more likely to improve. Most of the variables, such as birth weight and apnea did not reach significance in the regression analysis. Decreasing birth weight was not significantly associated with lower Wechsler Preschool and Primary Scales of Intelligence. By the age of 51/2 years, there was a considerable readjustment in the impairment assessment of the 50 children with sufficient data for reclassification. Four children were judged to be more severely impaired than at 2 years of age, 19 were unchanged, and 27 had improved. Of nine children considered to be unimpaired at 2 years of age, two had minor neurological or behavioral problems at 51/2 years of age. The dominant pattern was of either stability or improvement at 51/2 years of age. At 51/2 years of age, no new case of blindness or severe sensorineural deafness was diagnosed. Six children had spastic cerebral palsy at 2 years of age. This diagnosis remained stable for the two children with spastic quadriplegia with severe motor disability; but of four children with cerebral palsy with minor motor disability at 2 years of age, three improved, and the diagnosis could be sustained in only one child with spastic diplegia. Of the three children who “recovered” from mild cerebral palsy, one still had hyperactive deep tendon reflexes in the lower limbs, and another had difficulty in walking on the heels.