A sampling strategy was evaluated in the Andean highlands of Peru to optimise the detection of Ralstonia solanacearum in seed tubers harvested from symptomless crops. A sensitive and specific serological method developed at CIP was used to detect the pathogen in latently infected tubers. Optimum sample size was evaluated for symptomless crops after analysing various numbers of composite samples and using a binomial distribution model to calculate the detection probabilities. R. solanacearum was detected in all lots from fields with visible symptoms, so validating the detection technique. About half of the seed lots from apparently healthy fields at altitudes of up to 3,100 m were found positive for the pathogen. R. solanacearum was detected with 99% probability in samples of 350 tubers from seed lots from symptomless crops. This number of seed tubers could feasibly be processed in a seed-health test without incurring too high a cost for labour and materials.