Neurons in MT (V5) are selective for the direction of visual motion. In addition, many MT neurons are selective for the motion of complex patterns independent of their component orientations, a behavior not seen in earlier visual areas. We show that the responses of MT cells can be captured by a linear model that operates not on the visual stimulus, but on the afferent responses of a population of nonlinear directionally-selective V1 complex cells. We fit this cascade model to the responses of individual MT neurons to "hyperplaids" made by summing six randomly-chosen gratings. The model accurately predicts the separately-measured responses of MT cells to gratings and plaids, and captures the full range of pattern motion selectivity found in MT. Cells that accurately signal pattern motion are distinguished by having broadly tuned excitatory input from V1, strong motion opponent suppression, and a tuned normalization at the V1 stage that may correspond to suppressive inputs from the surround of V1 cells.