We estimate the effect of malaria on settlement and land use patterns in the Brazilian Amazon, where potential settlers were randomly assigned to plots in a newly opened settlement area. The random assignment allows us to estimate the risk of malaria on each plot based only on its characteristics. Using survey data, we find that a high malaria risk significantly reduces the probability that a plot is inhabited. Using satellite images, we find that a high malaria risk does not reduce forest clearance or crop coverage on a plot. Non-resident farming substitutes for physical inhabitation when malaria risk is high.