The rodent model for multiple sclerosis, experimental allergic (autoimmune) encephalomyelitis (EAE), has been used to dissect molecular mechanisms of the autoimmune inflammatory response, and hence to devise and test new therapies for multiple sclerosis. Clearly, artificial immunization against myelin may not necessarily reproduce all the pathogenetic mechanisms operating in the human disease, but most therapies tested in multiple sclerosis patients are nevertheless based on concepts derived from studies in EAE. Unfortunately, several treatments, though successful in pre-clinical EAE trials, were either less effective in patients, worsened disease or caused unexpected, severe adverse events, as we review here. These discrepancies must, at least in part, be due to genetic and environmental differences, but the precise underlying reasons are not yet clear. Our understanding of EAE pathogenesis is still incomplete and so, therefore, are any implications for drug development in these models. Here, we suggest some potential explanations based on new thinking about key pathogenic concepts and differences that may limit extrapolation from EAE to multiple sclerosis. To try to circumvent these rodent-human dissimilarities more systematically, we propose that pre-clinical trials should be started in humanized mouse models.