In an attempt to identify “atypical” strains which could account for vaccination failures, 10 street, one intermediate (DR19) and 4 fixed rabies virus isolates from men, cattle, dogs, cats, mongoose and vampire bats in five countries (Argentina, Brazil, Chile, Cuba, and France) were studied by cross-protection tests in mice. For the purpose of this study, any virus that killed more than 20 % of the vaccinated mice challenged with that virus was considered “atypical”. When the suckling mouse brain rabies vaccine was used, two “atypical” isolates were found: one, from a human case in Chile (91, 125 mouse passages) and the other, from a vampire bat in Brazil (DR19, 22 mouse passages). However, when mice immunized with a cell culture vaccine (PV-BHK) of a higher antigenic value than the brain vaccine, were challenged with those same isolates, mortality was below 20%. The fact that these two isolates killed enough vaccinated mice to be considered “atypical” could be related to antigenic differences between these viruses and those included in the vaccine. However, since this mortality was observed only in the mice immunized with the vaccine with a lower antigenic value, reasons are given why it could be attributed to other biological characteristics of those strains than antigenic differences. Causes for vaccines failures other than immunological differences of rabies virus strains are also analyzed and discussed.