Purpose To evaluate the reliability and variability of eight RT3 accelerometers.Methods The RT3 were subjected to two repeated trials of six activities: rest, walking (4 and 6 km·h−1), running (8 and 10 km·h−1), and sit-stand position (20 min). One person performed all trials (female: age 24 yr, height 158.0 cm, mass 48.2 kg). Each activity lasted 12 min. The middle 10 min were taken from each 12-min trial and used as the output measure (cts·min−1). Data were analyzed for activity (6), monitor (8), and trial (2) effects using four three-way ANOVA: vector magnitude, X (vertical), Y (anterioposterior), and Z (mediolateral) axes.Results Intermonitor coefficient of variation was <6% during locomotive activities, however, increased to 8–25% during sit-stand. A three-way interaction was found for vector magnitude (F35,315=88945.7, P < 0.001) and Y (F35,315=978435.6, P < 0.001) and Z axes (F35,315=103802.8, P < 0.001). A two-way activity × monitor interaction was found for the X axis (F35,315=1037787.0, P < 0.001). Follow-up tests revealed no differences between trials 1 and 2 for vector magnitude, X and Z axes. One monitor recorded significantly lower activity counts in trial 1 compared with trial 2 along the Y axis. Intermonitor differences were evident at 4, 6, 8, and 10 km·h−1 for the Y and Z axes, and at 6, 8, and 10 km·h−1 for the vector magnitude and X axis. Variability between monitors at each activity increased as intensity increased.Conclusion Reliability of the RT3 is good; however, intermonitor variability exists. The vertical axis of the RT3 accelerometer showed the least variability and was the most reliable. It is recommended that intermonitor variability and reliability of RT3 on each axis be assessed before use.