Historically, the classification of sedentary behaviors (SB) with accelerometers has used either single axis or 1-minute cut-points developed in laboratory. PURPOSE: To develop 3-axis Vector Magnitude cut-points per second (cts/sec) to classify SB based on movement in free-living conditions. METHODS: Two researchers recorded on a tablet free-living movement in 20 adults for >6 consecutive hours on two separate days. Movement types were recorded as: sitting, reclining, standing, walking, running, and kneeling. Participants wore 5 accelerometers collecting data at 100Hz (ActiGraph: right- and left wrists, and right hip; GENEActiv: right- and left wrists). Accelerometer data were integrated into 1-sec epochs. Observations in which both researchers had 100% agreement were included in the analyses. Observation data were categorized as sedentary and non-sedentary and randomly divided into training (50%) and testing (50%) datasets. Using the training dataset, receiver operating characteristic curve analyses were conducted using ROCPLOT macro (SAS v. 9.4) to determine the optimal sensitivity (SE)/specificity (SP) of cts/sec for SB. Kappa (κ) was computed in the testing dataset to compare the time in SB from the accelerometer-determined cts/sec with time spent in SB from direct observation. RESULTS: Adults (N=20; 50% female), Mage= 30.25 ± 6.43 years, normal BMI (85%) gave informed consent to participate. SE/SP analyses identified different cts/sec to detect SB for each accelerometer placement: ActiGraph [left wrist 5 cts/sec (SE, .66/SP,.64); right wrist 14 cts/sec (SE,.69/ SP,.66); right hip 0 cts/sec (SE,.88/ SP,.44)]; GENEActiv [left wrist 2 cts/sec (SE,.65/ SP,.66); right wrist 3 cts/sec (SE,.71/ SP,.70)]. Kappa’s were modest but statistically significant (p<.005) for all devices and placement sites: ActiGraph [left wrist κ=.29 (95% CI, .29, .30); right wrist κ=.32 (95% CI,.32, .33); right hip κ=.35 (95% CI,.34, .35) ]; GENEActiv [left wrist κ=.29 (95% CI,.28, .29); right wrist κ=.31 (95% CI,.30, .31) ]. CONCLUSION: Based on SE/SP to detect SB and agreement with direct observations, the best locations for the ActiGraph and GENEActiv accelerometers were the right hip and right wrist, respectively. Machine-learning algorithms may improve SB detection and future studies should confirm these findings.