The present study estimates the longitudinal effects of prenatal cocaine exposure on indicators of sustained attention processing at 3, 5 and 7 years of age in an urban sample of full-term African-American children (235 cocaine-exposed, 207 noncocaine-exposed). The sample was enrolled prospectively at birth, with documentation of prenatal drug exposure status through maternal interview, urine and meconium toxicology assays. Sustained attention was measured at age 3 years using a standardized measure of task persistence during a challenging task [G.A. Morgan, N.A. Busch-Rossnagel, C.A. Maslin-Cole and R.J. Harmon, Individualized Assessment of Mastery Motivation: Manual for 15-36 Month Old Children, 1992.], and at ages 5 and 7 years using omission error scores from computerized continuous performance tasks (CPT) [L. Greenberg, R. Leark, T. Dupuy, C. Corman, C. Kindschi, M. Cenedela, Test of Variables of Attention (T.O.V.A. and T.O.V.A.-A.), 22, Universal Attention Disorders, Los Alamitos, CA, 1996; C.K. Conners, Conners' Continuous Performance Test (CPT), second ed., Multi-Health Systems, Canada, 1995.]. Findings from longitudinal GLM/GEE analyses of the three measured time points support a stable influence of prenatal cocaine exposure on indicators of sustained attention, after controlling for prenatal exposure to alcohol, marijuana, tobacco and over 20 additional medical and social-demographic covariates drawn from potentially confounding influences assessed at birth and later assessment visits (D=0.21; 95% CI=0.04, 0.38; P=.017). This effect was not mediated by fetal growth or gestational age and remained highly stable with increasing levels of covariate control. Separately, using the age 7 data, a structural equations model (SEM) was constructed combining all available self-report and bioassay data to measure magnitude of cocaine exposure in relationship to attention task performance. Results indicated a gradient of influence, with each standard deviation increase in the level of prenatal cocaine exposure relating to a 16% standard deviation increase in omission error scores at age 7. Overall findings support a stable cocaine-specific effect on indicators of sustained attention processing during the early childhood years. Results are discussed within the context of neurobiological and behavioral research linking prenatal cocaine exposure to long-lasting disruption of the brain systems subserving arousal and attention.