The results of a survey conducted during 1993-2000 to study the spread of bovine viral diarrhoeal virus (BVDV) among Estonian cattle are presented. The BVDV infection status of a representative random sample of cattle herds housing 20 or more dairy cows was established to estimate the prevalence of herds with active BVDV infection [potentially having persistently infected (PI) cattle--suspect PI herds]. The herds investigated comprised approximately 70% of all Estonian dairy cows. The BVDV infection status was established in 315-350 herds (making the sampling fraction about 20%) during three sampling periods: 1993-95, 1997-98, 1999-2000. BVDV antibodies were detected in herd bulk milk samples and/or sera from young stock by a liquid-phase-blocking enzyme-linked immunosorbent assay developed in the Danish Veterinary Institute for Virus Research. The results of the survey demonstrate the reduction in the prevalence of herds with active BVDV infection in the studied fraction of the Estonian cattle population. During the first sampling period (1993-95) a prevalence of 46% (+/- 5%) for suspect PI herds was observed, during the second sampling period this prevalence was 16% (+/- 3%) and in the third period it was 18% (+/- 3%). As there is no control programme for BVDV in Estonia, the observed changes reflect the natural course of the infection in the study population. A possible cause for these changes is the decreased trade in breeding animals as a result of the economic difficulties present in cattle farming during the study period. The farming practices (most large herds are managed as closed herds) and the low density of cattle farms have obviously facilitated the self-clearance of herds from the BVDV infection, diminishing the new introduction of infection into the herds.