Referral bias is a significant problem affecting the generalizability of clinical studies conducted in a university setting. To examine referral bias in our university-based multiple sclerosis referral center, we analyzed the characteristics of referral center patients compared to the population-based group of multiple sclerosis patients from which the referral center patients originated. The referral center patient group differed from those that remained in the population-based group in the following important ways: (1) they were younger, (2) they had more mobility impairment for their age, (3) disabled females were overrepresented compared to disabled males, (4) they more often reported recent disease worsening, (5) they had a higher frequency of early diagnosis supported by laboratory tests, and (6) they more often relied on neurologists and therapists for routine care of their disease. The multiple sclerosis referral center setting would appear to be ideal for the conduct of intervention trials, but inadequate for collecting representative natural history data.