There is a task-specificity in the visual-tactile interaction for perception: The polymodal posterior parietal cortex is related to the comparison of the shapes coded by different sensory modalities, whereas the lateral occipital complex is the part of the network for multimodal shape identification. These interactions may be mediated by some latent pathways potentiated by sensory deprivation or learning.

We review evidence for multifaceted functional specialization of somatosensory information processing, both within and outside classical somatosensory cortex. We argue that the nature of such specialization has not yet been clarified adequately to regard the proposed action/perception dichotomy as being established. However, we believe this is a good working hypothesis that can motivate further work.

Neonatal behavioral data support the argument that multiple relations exist between vision and touch. Looking at an object triggers the motion of a neonate's arm and hand towards it. A textured surface that is seen can be recognized tactilely, but not a volumetric shaped object in cross-modal transfer tasks. These data are supported by adult neuroimaging data.

Our commentary addresses two issues that are not developed enough in the target article. First, the model does not clearly address the distinction among external objects, external body parts, and internal bodies. Second, the authors could have discussed further the role of body schema with regard to its dynamic character, and its role in perspective and in imitation.

A general shortcoming of the localist, decompositional, approach to neuroscientific explanation that the target article exemplifies, is that it is incomplete unless supplemented with an account of how the hypothesized subsystems integrate in the normal case. Besides, a number of studies that show that object recognition is proprioception dependent and that cutaneous information affects motor performance make the existence of the proposed subsystems doubtful.

The model presented in the target article includes feature processing and higher representations. I argue, based on neuropsychological evidence, that spatial representations are also involved in perceptual awareness of somatosensory events. Second, there is an asymmetry, with a right-hemisphere–based bilateral representation of the body. Third, the specific aspect of bodily awareness concerning motor function monitoring involves a network that includes the premotor cortex.

The commentaries have raised important points regarding different aspects of our model. Some have queried the nature of the proposed dissociations, whereas others have requested and provided further details regarding aspects we had glossed over. Here we suggest that our approach to identify major processing streams based on the processing goal does not preclude interactions between them. We further specify details regarding body representations, haptic object recognition, and crossmodal processing, but are also aware that several features of the model require further filling in.