Accurate assessments of physical activity are essential for advancing many lines of physical activity research. Numerous physical activity assessment techniques have been developed, but continual refinement and evaluation of these techniques is important to further improve accuracy and precision. A machine-learning technique known as the Sojourn method has demonstrated promise for improving the accuracy of accelerometer-based physical activity monitors, such as the widely used Actigraph. However, fewer studies have validated this method under free-living conditions. Purpose: The purpose of this study was to examine the performance of the Sojourn method for estimating energy expenditure and time spent at various intensities of activity relative to estimates from an established monitor-based measure (Sensewear Armband) and an established report-based measure (ACT24). A secondary purpose is to examine the context-related factors that are captured with the Sojourn method by comparing activity patterns with parallel data from the Sensewear Armband (SWA) and ACT24. Methods: The study used data obtained through a large (ongoing) field-based evaluation of activity monitors conducted in the Physical Activity and Health Promotion Lab at Iowa State University. The data for the present study involved temporally matched data from a sample of 85 adults with complete data on these three measures (Sojourns, SWA, and ACT24). The study involved two laboratory assessment days split by a 24-hour period during which activity monitors were worn. The first meeting consisted of participants completing a demographic survey and anthropometric measurements. Participants were instructed to wear the Actigraph and the Sensewear monitor (along with 5 other monitors) for a full 24-hour period (midnight to midnight) under free-living conditions. On the day following monitor wear, participants returned to the lab to complete the ACT24. Correlations, mean percent error (MPE), mean absolute percent error (MAPE), and Bland-Altman plots were used to assess method agreement. Results: Correlational analyses revealed moderate-strength relationships between the Sojourns and SWA (r = 0.65) and strong associations between the Sojourns and ACT24 (r = 0.91) for capturing total daily energy expenditure. Additionally, correlational analyses revealed moderate-strength relationships between the Sojourns and each of the other methods for time spent in sedentary, vigorous intensity activity, and MVPA. Error analyses revealed modest amounts of error between the Sojourns and SWA (MPE: 3.5%, MAPE: 16.1%) as well as between the Sojourns and ACT24 (MPE: 6.6%, MAPE: 9.5%) for capturing total daily energy expenditure. Error between methods was lowest for sedentary time (with all values below 24%) but classification accuracy was higher for light and moderate intensity activity. Bland-Altman analysis revealed some bias between methods for all indicators. Conclusions: The Sojourn method has promise as a standardized method for estimating energy expenditure and time spent in physical activity. However, additional refinements are warranted to further improve the utility for field-based research applications.