An accurate estimate of the prevalence of scrapie infection in the Great Britain (GB) sheep flock is essential when assessing any potential risk to human health through exposure to sheep transmissible spongiform encephalopathies (TSEs). One method for assessing the prevalence is to sample sheep intended for human consumption using a diagnostic test capable of detecting infected animals prior to the onset of clinical signs. An abattoir survey conducted in Great Britain in 1997-1998 tested brain samples from 2809 apparently healthy sheep of which none was found to be positive for scrapie by histopathology or immunohistochemistry (IHC) although 10 were positive for scrapie-associated fibrils (SAF). Subsequently, the tonsils from a subset of the animals sampled were examined using IHC, one of which tested positive. To interpret these results we use a likelihood-based approach, which accounts for the variation in the prevalence of infection with age and test sensitivity and specificity with stage of infection. Combining the results for all of the diagnostic tests yields an estimate of the prevalence of scrapie infection in the GB sheep flock of 0.22% (95% confidence interval: 0.01-0.97%). Moreover, our analysis suggests that all of the diagnostic tests used are very specific (greater than 99%). Indeed, only SAF detection yields a specificity estimate of less than 100%, which helps to account for the high number of samples found to be positive for SAF.