A path model was used to investigate the effects of some variables on bovine leukemia virus (BLV) transmission between herds. Data were obtained from 88 outbreaks of BLV infection and 92 herds free from BLV in Catalonia (Spain). The LISREL program was used to carry out the path analysis. This program enables the use of latent variables (they are treated by means of a factorial analysis). The interest and effort devoted to the improvement of farm management is a latent variable and has been quantified through four observable variables (official records of lactation, milk yield, percentage of imported semen, and frequency of reproductive control visits). These four variables explained 73% of the latent variable. Two models were elaborated, allowing explanation of 47 and 69%, of the variation in the presence or absence of the virus in the herd. The introduction of new animals into the herd had a significant effect on the outbreaks of BLV: the risk due to the introduction of positive animals (defined from the number of animals introduced, their age and the prevalence in the sites of origin) had a path coefficient of 0.83. When the variables, number of animals introduced and BLV prevalence in the herd of origin were considered separately, both factors were important: the number of new animals had a coefficient of 0.42, and prevalence in the herd of origin a coefficient of 0.47. The variables associated with possible iatrogenic transmission had negligible effects. The size of the herd had important effects on the presence of BLV (0.34); the model explains this indirect effect through the number of animals introduced into the herd. The hypothesized relationship between improvement of farm management and the leukosis outbreaks seems to be correct (coefficient of 0.17), being an indirect causal relationship through the prevalence in the site of origin of introduced animals. © 1990.