We used a prospective seroepidemiological approach to investigate endemic abortion in a dairy herd in Arequipa, Peru, and its association with Neospora caninum and bovine viral-diarrhoea virus (BVDV) infections. Between January 2002 and March 2004, 1094 pregnancies were confirmed in 538 cows. Of these, 137 pregnancies (13%) in 121 cows ended in abortion. The serological status to N. caninum was assessed using a single serological screening, whereas BVDV status was assessed at the herd level through consecutive samplings of young stock. Cox proportional-hazards models were used to estimate the effect of N. caninum and BVDV on the hazard of early (between day 42 and day 100 in gestation), and late (after day 100) abortions, respectively. Serological status to N. caninum was included as a dichotomous variable, and the effect of BVDV estimated at the herd level, as a time-dependent seasonal effect. Because data from repeated pregnancies were included, we considered possible lack of independence between observations and included frailty effects into the models. Our models also considered the possible confounding by parity and animal origin. Only multiparity was associated with the hazard of early abortion (HR = 2.8 compared to nulliparous heifers). N. caninum seropositivity significantly affected the hazard of late abortion, but interacted with parity. The HRs for Neospora-positive animals were 6.4, 3.7 and 1.9, respectively, for nulliparous heifers, first-lactation cows and multiparous cows. Evidence of BVDV circulating (or not) among the young stock was not associated with abortions, but few cows in this herd were susceptible to incident infection.