Wood deposited in streams provides a wide variety of ecosystem functions, including enhancing habitat for key species in stream food webs, increasing geomorphic and hydraulic heterogeneity and retaining organic matter. Given the strong role that wood plays in streams, factors that influence wood inputs, retention and transport are critical to stream ecology. Wood entrapment, the process of wood coming to rest after being swept downstream at least 10 m, is poorly understood, yet important for predicting stream function and success of restoration efforts. Data on entrapment were collected for a wide range of natural wood pieces (n = 344), stream geomorphology and hydraulic conditions in nine streams along the north shore of Lake Superior in Minnesota. Locations of pieces were determined in summer 2007 and again following an overbank stormflow event in fall 2007. The ratio of piece length to effective stream width (length ratio) and the weight of the piece were important in a multiple logistic regression model that explained 25% of the variance in wood entrapment. Entrapment remains difficult to predict in natural streams, and often may simply occur wherever wood pieces are located when high water recedes. However, this study can inform stream modifications to discourage entrapment at road crossings or other infrastructure by applying the model formula to estimate the effective width required to pass particular wood pieces. Conversely, these results could also be used to determine conditions (e.g. pre-existing large, stable pieces) that encourage entrapment where wood is valued for ecological functions.