Context The Australian sea lion population at Seal Bay Conservation Park, South Australia, was estimated to be declining at a rate of 1.14% per breeding season, on the basis of maximum counts of live pups in each of 13 breeding seasons (Shaughnessy et al. 2006). The reliability of the pup-production estimates used to identify this decline is uncertain. Aims Our aims were to obtain representative and repeatable estimates of pup production and to assess the current rate of decline. Methods We compared four estimates of pup abundance over five breeding seasons (200203, 2004, 200506, 2007, 200809), including the count of cumulative new births, the maximum live-pup count, the number of pups given passive integrated transponder (PIT) tags, and markrecapture methods using the Petersen estimate. Key results A total of 90% of pup births occurred over a mean of 124 days (s.d.≤14). Final estimates of pup production (from the largest of the four estimation methods used) in the five seasons were 227 (CL 221239), 288 (CL 273302), 219 (NA), 260 (CL 254272) and 268 (CL 268269). The average estimate of pup mortality was 28.6% (s.d.≤6.3%). The decline in the population at Seal Bay over 17 breeding seasons on the basis of maximum counts of live pups was 0.51% per year or 0.76% per breeding season. However, this trend was not based on best estimates of pup production. On the basis of final estimates for the last five breeding seasons, there is no declining trend. Conclusions The count of cumulative new births was the most reliable measure of pup production; the Petersen markrecapture estimate provided a check for accuracy and confidence limits about the estimate. Implications The actual rate of change and the expected trajectory of the Seal Bay population remain uncertain. Ongoing monitoring is a priority for this site, using the reliable methods of estimating pup production identified in the present study.