A prospective study was undertaken on a California dairy from 1984 to 1987 to examine factors associated with contact transmission of bovine leukemia virus in cows. Two approaches were used to model the probability of infection. First, the expected number of new infections per pen-month was assumed to follow a binomial distribution. The probability of infection was modeled, using logistic regression, as a function of prevalence of infection in pen, presence of lactating cows, proportion of pregnant cows, presence of an infected bull, and proportion of infected cows with a lymphocyte count above 10,000 cells/microliters of blood and/or with Mr 24,000 protein (p24) antibodies. The probability of infection was significantly associated with the prevalence of infection and at the limit of significance for presence of lactating cows in pen. Second, the Cox model with time-dependent covariates was used to analyze time from first parturition or from the beginning of the study to infection. Factors examined were age, breed, lactation number, whether the cow was pregnant and/or lactating, prevalence of infection, presence of an infected bull, and density of cattle in the pen holding the cow. Non-pregnant cows were 2.9 times more at risk of infection than were pregnant cows, and risk of infection significantly increased as prevalence of infection in a pen increased.