Continuous and discontinuous data-collection methods were compared in the context of discrete-trial programming. Archival data sets were analyzed using trial sampling (1st 5 trials, 1st 3 trials, and 1st trial only) and session sampling (every other session, every 3rd session, and every 5th session). Results showed that trial sampling systematically underestimated the number of sessions and days to mastery and overestimated the number of sessions and days to the 1st independent response. Session sampling systematically overestimated both sessions and days to mastery and sessions and days to the 1st independent response. A time-savings analysis was included to evaluate empirically how much time would be saved by using each sampling method. Results suggested that data sampling would produce relatively minimal time savings.