Many people now carry multiple mobile devices on a daily basis. Wearables, smartphones, tablets, and laptops all have their different advantages, but collectively they can increase a user's device management burden. Management problems include leaving a device behind accidentally, receiving notifications on the wrong device, and failing to secure all of the devices as needed. Reducing this burden requires detecting which of a user's devices are "personally collocated" -- those devices he currently wears, carries, or has under his immediate physical control. We present a lightweight method to detect personal collocation by comparing accelerometer-based footstep signatures across the devices over time. Through several experiments, we demonstrate that the technique is lower latency and lower power than state-of-the-art RSSI-based collocation techniques. We describe other advantages and limitations of our method and also provide several examples of higher-layer applications and services that can make use of personal collocation information.