PURPOSE: Direct Observation (DO) has been used as a criterion measure of children’s physical activity (PA) since the early 1990’s. A limitation of these DO systems include researchers instantaneously recording only the highest activity level performed in a given time interval. Therefore, the purpose of this study is to compare the estimates of time spent in activity intensity categories using a novel video recorded DO methodology, compared to accelerometer data. METHODS: Children ages 6-10 years old participated in a simulated free play session. Toys, equipment, and a sibling or friend were available to facilitate a full range of activity intensities; sedentary (SED), light (LPA), moderate (MPA), and vigorous (VPA) for 30 minutes. Participants wore an ActiGraph GT3X+ (AG) over their right hip (AG-H) and non-dominant wrist (AG-W). The sessions were recorded with a GoPro™ video camera and, using Noldus™ behavioral observation software, an activity intensity category was assigned each time the child changed their activity. Two observers independently coded every 30-minute session. The percent of time spent in each PA category (calculated from the Noldus software) was compared between AG and DO using Spearman-Rank Correlations and Wilcoxon Rank-Sum Tests. RESULTS: Twelve children completed this study (8.9±1.2 yrs; 33% female). DO and AG-H were moderately correlated for all intensity categories (r=0.32 to 0.7), except for MPA (r=0.15). Mean values of percent of time spent in activity intensity categories were similar between DO and AG-H (p=0.11 to 0.83), except for MPA (DO 12%+5%, AG-H 18%+8%, p=0.02). DO and AG-W showed weak or negative associates for all intensity categories (r=-0.6 to 0.09). The AG-W underestimated percent of time spent in SED, LPA, and MPA (p<0.001 for all) but overestimated VPA (DO 13%±5%, AG-W 81%±10%, p=0.0001), compared to DO. CONCLUSIONS: Based on these results, the novel DO system tested here is promising but needs to be further refined to better distinguish moderate intensity activities from light or vigorous movement. Caution is warranted when interpreting wrist accelerometer data from free-living children.