The detection of coherent motion embedded in noise has been widely used as a measure of global visual motion processing. Animal studies have demonstrated that this performance is closely linked to the responses of direction-sensitive neurons in the macaque middle temporal (MT) and medial superior temporal (MST) areas. Despite the strong similarities between the visual cortex of human and that of non-human primates, the human middle temporal complex (area MT+), located in the posterior part of the inferior temporal sulcus and presumably comprising both area MT and area MST, has not consistently been found to share the functional hallmark of MT and MST neurons, i.e. their preference for coherent rather than incoherent visual motion. In order to search for such preferences in human area MT+, blood oxygen level-dependent responses to random dot kinematograms presented in the right visual hemifield were studied here as a function of stimulus size and dot density. The stimulus extensions were varied in such a way as to cover an area either equaling, exceeding or falling below the mean receptive field size of macaque area MT. Unlike the posterior part of human area MT+, the anterior part and its right-hemisphere homolog showed significantly stronger responses to coherent than to incoherent motion. These differences were only present for large stimuli that presumably exceeded the receptive field size of neurons in area MT. Our results suggest that functional magnetic resonance imaging may reveal stronger responses to coherent visual motion in human area MST, provided that the stimulus allows for sufficient summation within the receptive fields. In contrast, functional magnetic resonance imaging may fail to reveal the same dependency for human area MT.