While viruses are often associated with particular diseases, proof of a causal connection between the virus and the clinical condition is frequently elusive. Pioneers in epidemiology such as Koch and Hill were among the first to present algorithms addressing proof of causality and their methods are still used. However, the issue of causality is especially problematic when animal models are controversial or unavailable, the virus cannot be isolated or undergoes mutations during in vitro propagation, or the presumed etiological agent is cleared from the host before disease onset. In this review, we use multiple sclerosis to exemplify this difficulty, since numerous causal pathogens in this disease have been described over the years but none has ever been confirmed. Giant cell arteritis is another case in point. We also consider biological and molecular virological criteria that may be more appropriate to prove, or at least strongly suggest, causality in the 21^st century.