This study examines the utility of current accelerometer threshold definitions and epoch durations for physical activity intensity classification in preschool-aged children. Using video footage of children engaged in active play, directly observed 1-sec epoch physical activity intensity scores were derived from a modified version of the Children's Activity Rating Scale, averaged into 15-sec epochs (corresponding to the accelerometer scale), and compared to accelerometer-measured classifications. Overall, 6,540 individual 1-sec epochs were captured and classified from six children, yielding 429 time-matched 15-sec epoch physical activity intensity classifications from both methods. Children participated in 25 distinct physical activity behaviors and spent 75% of free-play in one intensity level for less than or equal to 5 sec. The maximum amount of time spent participating in an activity prior to changing activity intensity was 98 sec. Agreement between methods was generally poor (overall: K = 0.22; child range: K = 0.00 to 0.66), and the difference between methods was significantly asymmetrical (McNemar's test overall: p = 0.01). Accelerometer thresholds and epoch durations used to classify preschoolers' physical activity intensity yielded biased estimates in the current study. Further investigation is necessary to establish optimal physical activity measurement methods with this population. (Contains 3 tables and 1 figure.)