The Paced Auditory Serial Addition Test (PASAT) is widely used in the evaluation of multiple sclerosis (MS) patients’ cognitive performance, and also used as the sole measure of cognition in a recently developed assessment tool for MS clinical trials, the Multiple Sclerosis Functional Composite (MSFC). We analysed if MS patients and healthy controls have different patterns of responding in the PASAT, and whether different scoring methods influence the PASAT’s sensitivity and specificity in detecting disease-associated cognitive impairment. Forty-five relapsing-remitting MS patients and 48 healthy controls were evaluated using the PASAT and a comprehensive neuropsychological examination. Cognitively deteriorated MS patients compensated for their difficulties in PASAT by omitting rather than guessing answers. They skipped items intermittently, which reduces the difficulty of the task. Furthermore, towards the end of the PASAT’s 60-item series MS patients’ performance had a trend to fade whereas controls’ performance was more even throughout the task. The dyad score or the percent dyad score did not essentially improve the sensitivity or the specificity, but the accuracy improved when the answers at the end of the PASAT series were specifically emphasized. Using the combined score, 73% of the patients were correctly classified as cognitively impaired or unimpaired.