Purpose: If sufficiently valid, wearable activity trackers are promising tools for health-related research. Recent models claim to identify the type of activity being performed, but no studies have reported the validity of these features. The purpose of this study was to determine the accuracy of automatic activity detection on 4 wrist-worn, physical activity trackers; specifically, the correct type, duration, and start time of select activities. Methods: 69 healthy adults were recruited via flyer, email, or word of mouth to complete at least one of four activity modules, comprised of activities automatically detectable by the trackers. Module A, completed on a treadmill, consisted of 3 activities separated by a 10-min rest (15-min walk, 15-min run, and 25-minute continuous series of 5-min walk, 15-min run, 5-min walk). Module B, consisted of 3 activities separated by a 10-min rest (15-min outdoor walk, 15-min outdoor run, and 15-min bout on an elliptical trainer). Module C consisted of 15 min of outdoor cycling and Module D consisted of 15 min of freestyle swimming (only one tracker was used in Module D as others are not waterproof). The actual activity type, duration, and start/stop times were recorded, then compared to device data (recognized activity type, duration, and start times). Results: Participants (N=69) were 26.3 ± 8.7 years old, had a BMI of 23.8 ± 4.0 kg/m2, and were 60.9% female. Participants reported an average rating of perceived exertion of 11.9 (6 - 19). The percentage of correctly identified activities by the four trackers was 94.7% (93.5 - 97.1) for treadmill walking, 97.7% (93.8 - 100.0) for treadmill running, 45% (35.3 - 59.4) for running in the treadmill series, 97.8% (97 - 100) for outdoor walking, 100% for outdoor running, 70.4% (3.1 - 93.9) for the elliptical, 83.1% (44.1 - 97.1) for outdoor cycling, and 87.5% for swimming. Lower accuracy for the elliptical was driven down by a low accuracy of 3.1% for one tracker. Conclusion: The four wearable activity trackers were best at detecting outdoor running and worst at detecting running in the 25-minute treadmill series. Overall, the trackers were better at detecting ambulatory activities, apart from running in the treadmill series, than detecting swimming, cycling, or using an elliptical. This research was funded through startup funds through UW - Madison.