The paced auditory serial addition test (PASAT) is increasingly used in multiple sclerosis (MS) studies. Since these studies rely on repeated assessments with relatively short inter-test intervals, practice effects can be a confounding factor. We examined intra-session PASAT practice effects in 70 relapsing remitting (RR) and 40 secondary progressive (SP) patients. The average number of correct answers increased from 39.6+/-11.7 in the first PASAT run to 43.8+/-11.4 in the second run for the RR group, and from 39.1+/-11.6 to 41.8+/-13.3 in the SP group. PASAT scores showed a consistent decrease when comparing the second half of each test to the first half for both patient groups, and for both runs. Items for which the answer was a number greater than 9 had the same discrimination ability as other test items, but were significantly more difficult. A simulation of ;single-button' responses supported the use of the simplified scoring method which is currently used in fMRI studies. Our results demonstrate a within-session PASAT practice effect in MS, as well as a fatigability effect for both patient groups.