The composition of the landscape between patches (the matrix) can have important effects on movement rates that potentially outweigh the effects of patch size and isolation. We conducted a small-scale experiment with radiocollared meadow voles (Microtus pennsylvanicus) to quantify the effects of matrix habitat on movement behavior of voles. Habitat patches were created within an enclosure and connected by pathways of different vegetation heights designed to represent different levels of predation risk to the voles. Voles were fitted with radiocollars and tracked 4 times daily for 1–3 weeks. We quantified the probability that a vole would transition through a particular matrix type (i.e., cross a habitat edge) and estimated transition probabilities using multinomial random effects models. These methods along with permutation tests were used to test whether the transition probabilities of voles could be modeled by assuming common movement rules such as “random,” “static choice” when all voles always choose the same type of pathway, or “probabilistic” when voles are most likely to choose a pathway of tall vegetation and select other pathway types in inverse order of the relative risk associated with the pathway. Vole movement decisions in our experimental landscapes reflected a probabilistic approach to decision making in which pathways were chosen in proportion to risk associated with travel. Quantification of transition probabilities through different matrix types is necessary for modeling animal movements in complex environments.