The Human Genome Initiative has increased significantly the rate at which disease-causing genes are being mapped and sequenced. New cost-effective methods to locate the genes and to characterize disease-causing mutations require robust, reproducible, and accurate protocols for measuring DNA fragment lengths. Capillary array electrophoresis (CAE) offers rapid, high-resolution separations, high throughput, and sensitive detection. To assess the utility of CAE for the accumulation of genetic information, we tested both sizing accuracy and reproducibility using 48-capillary prototype systems. Two multiplex PCR allelic ladder standards and several CA-repeat markers were analyzed in > 100 runs. Reproducibility in typing > 8000 genotypes reveals a standard deviation of less than 0.2 bp on these systems under optimized conditions. However, sequence-dependent migration anomalies were observed at most simple sequence loci even when analyzed under denaturing conditions, resulting in a systematic bias in estimated fragment sizes. We show here that, by normalizing results to known typing controls, one can obtain locus-averaged accuracies of < 0.06 bp and normalized results within 1 bp of actual. We detect as little as a 1:30,000 dilution of a DNA quantitation standard stained with highly sensitive intercalating dyes, indicating an 80-zeptomole sensitivity limit. However, to obtain reproducible electrokinetic injection, approximately 200 attomoles of fluorescein-labeled DNA is required. These sensitivity limits, sizing precision, and accuracy, together with the 1-hr run times for 48-96 samples, indicate that CAE is a viable method for high-throughput genetic analysis of simple sequence repeat polymorphisms.