The Paced Auditory Serial Addition Test (PASAT), most used neuropsychological test in people with multiple sclerosis (pwMS), is available as PASAT-3 and more difficult PASAT-2. We determined PASAT performance in 105 pwMS and 105 matched controls and assessed the utility of a shortened PASAT-3 version. Qualitative and quantitative analyses were performed, the latter involving assessment of dyads (two consecutive correct responses) and singles (correct response between two omitted responses). Based on provisional results we then calculated the sensitivity of the first 20 items of PASAT-3 in relation to the full version. Mean PASAT-2 and 3 total score, dyads and singles differed significantly (P < or = 0.0002) in pwMS and controls, due to more omissions in pwMS, number and type of incorrect answers did not differ. Differences between pwMS and controls were highly significant for the first 50, 40, 30 and 20 responses of the PASAT-3. The sensitivity of the first 20 items of PASAT-3 was 87% (95% CI 0.70-1.00). PASAT-3 and PASAT-2 effectively discriminate pwMS from controls. Use of the first 20 items of PASAT-3 appears as a highly useful PASAT-3 version as it retains the discriminative value and sensitivity of the full version yet is quicker to administer and less demanding.