While the standard 200-item version of the Paced Auditory Serial Addition Task (PASAT) is a sensitive neuropsychological instrument, it can be quite aversive to some patients due to its length and progressively increasing difficulty. We present demographically-corrected norms for 50 and 100-item short-form versions in a sample of 560 neurologically normal adults. Age, education, and ethnicity (but not gender) were found to be significant predictors of performance. In a clinical sample of 786 HIV-infected adults, diagnostic accuracy of the 50, 100, and 200-item versions was essentially equivalent (using clinical ratings of a comprehensive neuropsychological battery as the gold standard, overall classification rates of the three PASAT versions were 71%, 74%, and 73%, respectively), with better specificity (89-92%) than sensitivity (46-53%). The 50-item version showed moderate ceiling effects, but the 100-item test did not. In a mixed clinical sample of 40 subjects, the 50-item version was administered more than twice as fast as the 200-item version, and was tolerated better (discomfort rating of 4.0 vs. 5.9 on a 10-point scale, p < .05). We conclude that in many cases the PASAT-50 and PASAT-100 provide equivalent diagnostic accuracy with a significant reduction in administration time and patient discomfort.