This study compared the RT3 to the R3D in both laboratory and field settings and tested the hypothesis that the RT3 records higher physical activity counts and smaller standard deviations, as compared to the R3D. Twelve men [age 23 ± 2 yr] and thirteen women [age 24 ± 2 yr] performed each of the following exercise conditions on a motorized treadmill: slow walking (4.8 km•h-1), fast walking (6.4 km•h-1), and jogging (9.7 km•h-1) at 0% grade; fast walking (6.4 km•h-1) at 5%, 10%, and 15% grades; and slow walking (4.8 km•h-1), fast walking (6.4 km•h-1), and jogging (9.7 km•h-1) on an outdoor field. Factorial ANOVA were performed to compare the vector magnitude and individual component (mediolateral, anteroposterior, and vertical) differences. Pearson’s r was used to determine the correlation between the activity counts of the RT3 and R3D. Statistical procedures for assessing agreement between two instruments (Bland & Altman, 1986) were also used. The RT3 assessed higher [F(8, 891) = 12.71, p < .001, h2 = .10] vector magnitude counts than the R3D. However, significantly (p = .008) higher mean standard deviations were found for the RT3 (527 ± 236) as compared to the R3D (417 ± 227). The analyses of RT3 and R3D components demonstrated differences for mediolateral [F (8, 891) = 42.26, p < .001, h2 = .28], anteroposterior [F (8, 891) = 49.00, p < .001, h2 = .31], and vertical [F (8, 891) = 12.84, p < .001, h2 = .10]. The overall association between the vector magnitude counts for the RT3 and R3D was high (r = .96, p < .001). The correlations for change in speed on the treadmill (r = .98, p < .001) and outside (r = .95, p < .001) at 0% grade were high whereas the correlation for change in grade on the treadmill (r = .70, p < .001) was moderate. In terms of agreement between the instruments, the RT3 might be 582 below or 1236 above (activity counts) the R3D in assessing physical activity. Differences in agreement between the RT3 and R3D did not vary in any systematic way over the range in testing conditions which substantiates that the RT3 and R3D accelerometers are sensitive on flat surfaces but are insensitive to changes in grade. These results do not establish that the RT3 is consistently measuring higher physical activity counts than the R3D.