Currently, the dominant view of the organization of the human brain postulates a series of parallel, hierarchically organized, sensory modality specific systems: a visual system, an auditory system, a tactile system, and so on. Sensory systems are generally characterized as having peripheral receptor systems that transduce information to pre-cortical neural relay stations. These relays (for instance, the nuclei of the thalamus) direct sensory signals to unimodal sensory cortical areas, thought to be responsible primarily for processing of data from a single modality. Unimodal sensory areas seem arranged in an orderly hierarchy of increasingly complex functional significance, from primary, through secondary, to unimodal association areas (Mesulam, 1998). Only after this series of steps, in which sensory information is believed to remain isolated by modality, is information thought to merge into higher order multimodal association areas of the cortex (Calvert, Campbell, & Brammer, 2000; Mesulam, 1998; Stein & Meredith, 1993). These multimodal association areas of the cortex contain
multisensory cells, which provide a neural mechanism for integrating sensory experiences, modulating the saliency of stimuli, assigning experiential and affective relevance, and providing the substrate for perceptual experience. However, such a model of brain organization appears excessively simplistic and raises several questions. For example, cortico-cortical and cortico-subcortical connections are generally arranged in feed-forward and matching feed-back loops. If such is the case between unimodal and multisensory areas, we ought to expect that the activation of multisensory areas by one sensory input would affect activity in all other sensory systems. Such interactions are likely to be quite specific, depending on the precise pattern of reciprocal connections, the cortical layers targeted by them, and so forth. In any case, if connections are indeed reciprocal, we ought to expect that given the appropriate task and circumstance, activity in multisensory areas would affect and modulate the activity and presumably the behavioral contribution of unimodal areas, including primary unimodal areas (de Gelder, 2000; Driver & Spence, 2000). Several other chapters in this volume address this idea of feed-back, crossmodal interactions in detail. Furthermore, it is conceivable that given such a structure of parallel, sensory-modality specific systems, connections might exist across unimodal sensory areas (Falchier et al., 2001; Rockland & Ojima, 2001). Such connections (whether cortico-cortical or through subcortical structures such as the thalamus) could be present during development and be expected to normally degenerate except in cases of early loss of a sensory modality (Innocenti, 1995). In certain instances, for example in the case of congenital or early blindness, such connections would allow for the primary sensory cortices normally associated with the deprived sensory modality to become colonized by the remaining sensory modalities (Bavelier & Neville, 2002). On the other hand, persistence of such developmentally transient connections into adulthood could be viewed as a reason for the normal experience of crossmodal interaction and, if uncontrolled, even synesthesias (Baron-Cohen & Harrison, 1997; Cytowic, 1989; Grossenbacher & Lovelace, 2001). These two mechanisms, feed-back influences from multimodal areas or feed-forward connections across unimodal areas, are not mutually exclusive, nor all exhaustive. In any case, appropriate tasks or interventions, for example temporary complete visual deprivation, might unmask the functional significance of such connections and shed new light onto the organization (at systems level) of the nervous system.