Behavioral experiments reveal that motion in cardinal (vertical or horizontal) directions is represented more strongly in the visual system than motion in oblique directions. For example, discrimination of direction of motion is better for cardinal than oblique stimuli. The goal of this study was to determine whether human MT+ shows the neural bias predicted by these behavioral findings. We first localized human MT+ and area V1 in four subjects using standard methods. We then compared cortical responses to two types of stimuli. In the first, two component gratings were oblique (spatial freq 0.8cpd, temporal frequencies 2-6Hz, contrast 0.4, orientations 45° and 135°) and formed a plaid pattern that moved at 90°. In the second stimulus, the components were cardinal (0° and 90°) and formed a plaid that moved at 45°. Patterns were presented in two Gaussian windowed patches (sigma=2°) at 5° eccentricity, to the left and right of fixation. To control for effects of attention, subjects performed speed discriminations on the stimuli in a two-interval forced choice paradigm. Change in fMRI signal (3T BOLD, EPI) in human MT+ was reliably greater for patterns moving in cardinal than oblique directions. V1 activity showed opposite effects: change in fMRI signal was reliably greater for patterns moving in oblique directions (containing cardinal components). Our results indicate that MT+ responds best to principal pattern motion even though the components are oblique. These findings suggest that human MT+ contains large populations of pattern-selective cells, whose distribution is biased towards cardinal directions. Such a bias in the neural circuitry could explain the perceptual oblique effects for moving stimuli.