Data from 4049 Romanov sheep belonging to a flock affected by natural scrapie were analysed by using survival-analysis techniques. Failure time was defined as the period of time between first exposure to infection and the date that animals left the flock with scrapie signs. Four hundred and forty-seven sheep were identified as ‘scrapie animals’. Several models, including level of exposure as a time-dependent effect, PrP genotype, sex, age at first exposure, litter size and factors related to vertical transmission, were tested. The best model was extended to a sire–dam frailty model, in order to estimate the polygenic variation in addition to that in the Prnp gene. A combined effect of rearing type and the dam's disease status was detected. Thus, only sheep with a low degree of exposure to infection as lambs (lambs reared artificially and born out of a healthy dam) showed less risk than others. Animals first exposed to infection at older ages seemed to be less susceptible to scrapie. In this Romanov population, new genotypes (AHQ/AHQ, AHQ/VRQ, ARR/VRQ and ARR/ARQ) were associated with risk, suggesting the effect of genotypes on the incubation period of animals. Polygenic variance was responsible for 21 % of the total genetic variability that was related to susceptibility to scrapie. Therefore, the genetic susceptibility to scrapie may be explained by the joint effect of point mutations at the Prnp major gene and a number of genes that modulate its effect.