A laboratory mouse model was used to investigate the criteria that have been suggested as differentiating between a maintenance host and an accidental host for a particular leptospiral serovar. The comparative studies were conducted with serovars ballum, pomona, balcanica and hardjo. The relative pathological response, ratio of serological to bacteriological prevalence, level of serological response, age-susceptibility to infection and demonstration of artificial intraspecies transmission were found to be inadequate criteria with which to differentiate maintenance and accidental hosts for a particular serovar. The demonstration of natural intraspecies transmission was considered to be the definitive criterion for differentiating such hosts. In the light of the results obtained from the laboratory mouse model and the results obtained from field studies, a maintenance host may be defined as an animal which is capable of acting as a natural source of leptospiral infection for its own species. A maintenance population may be defined as a population of a species of animal which acts as a continuous reservoir of a serovar in a specific ecosystem.