Perception gives preference to the sensory modality best suited to the task at hand. (Welch, 1980). Sensory conflicts are resolved by subjugating the less reliable sense sense to the dominant sense, or possibly through sensory recalibration (see Kitagawa & Ichihara, 2002; Recanzone, 1998). Unisensory inputs can also be encoded into the task-appropriate modality (Guttman et al., 2005).
Such a hypothesis is possible in the first place because different sensory modalities are differentially specialized to different perceptual tasks. For instance, vision is dominant in spatial processing, and audition is dominant in temporal processing.