The new McDonald diagnostic criteria for multiple sclerosis (MS) incorporate detailed criteria for the interpretation and classification of magnetic resonance imaging (MRI) findings, but, in contrast, provide no instructions for the interpretation of clinical findings. Because MS according to the McDonald criteria is one of the primary endpoints in a large trial enrolling patients after the first manifestation suggestive for a demyelinating disease (BENEFIT study), it was decided to organize a centralized eligibility assessment for this trial. During this eligibility assessment it was observed that there were marked inconsistencies in the decisions of participating neurologists with respect to the classification of clinical symptoms as being caused by one or more lesions provoking discussions in about one in every five patients. This paper describes these inconsistencies and their sources, and recommends a systematic approach that attempts to reduce the variability in interpreting clinical findings.