The target article integrates an impressive amount of the relevant literature into a model on how the brain processes somatosensory information in two distinct streams. However, it fails to disentangle the functional distinction from the structural distinction of the two somatosensory streams.

In vision, two functional streams can be distinguished: One, a visual-spatial processing stream for the control of spatial motor actions (how), such as saccading, grasping, or pointing at external target objects. The other, a visual-spatial processing stream for perception, for example, for object and scene recognition (what). From a functional and computational perspective, these processing streams have to deal with quite different constraints (e.g., timing, invariance) and produce quite different neurocomputational solutions. For example, consider “how” functions such as keeping the fovea on a moving target, versus “what” functions such as recognizing an object at the basic level from various viewpoints. Milner and Goodale (Reference Milner and Goodale1995) introduced the functional differentiation between what and how in vision. They also equated this differentiation with a structural distinction between dorsal and ventral processing in the primate brain. Recently, this structural equation has become quite controversial (e.g., see Franz et al. Reference Franz, Fahle, Bulthoff and Gegenfurtner2001; Glover Reference Glover2002; Schenk Reference Schenk2006). For example, relevant processing for some actions (e.g., memory-based – not online – saccades) is carried out by both dorsal and ventral brain areas (Özyurt et al. Reference Özyurt, Rutschmann and Greenlee2006) – there is a heavy interaction between areas from both structural streams for performing a single function!

Given that the structural distinction has come under sharp attack with regard to vision, we should beware of a corresponding distinction in the case of somatosensation. The somatosensory system should not be divided into two structural streams – whether dorsal versus ventral or parietal versus insular, whatever – but divided into functional streams of how and what. The evidence provided in the target article speaks the same language – including the patient examples at the very core of the article: Numbsense patients, as “action-without-perception” cases par excellence, suffer from lesions that according to theory belong to structures shared by the streams (and require the postulate of additional connections). Therefore, they convincingly support the functional, but not the structural, distinction. Even more obvious, the “perception-without-action” patient G.L. suffers from lesions that are necessarily shared by both structural streams, namely, peripheral ones. Functionally, however, we can distinguish between G.L. locating cold points within a visual body image (what), but not being able to point to them (how). Note, thereby, that effects of G.L.'s lesions include proprioception, but exclude temperature perception (e.g., see Stenneken et al. Reference Stenneken, Prinz, Cole, Paillard and Aschersleben2006). Using temperature perception, G.L. is able to locate targets on the body. Precise pointing, however, requires the integration of the target location on the body with the actual position of the executing and the targeted limb in space. Lacking proprioception, G.L. is not able to perceive the actual position of her limbs in space and, so, G.L. is not able to point to a location on her own body. Thus, G.L.'s dissociation is probably a better example of different processing requirements than of different structures in how and what streams.

Haptic perception, at the very core of somatosensory what-functions, requires strong interactions between structural areas – and, as recognized by Dijkerman & de Haan (D&dH), it cannot even be conceptualized without assuming tight links between sensory processing and control of exploratory movements. The attempt to designate different structural areas as being more responsible for one or the other part is understandable, but this should not mislead us into ignoring that the function, first of all, emerges from structural interplay. In this context, it is also significant that structural networks similar to those observed to be involved in (exploratory) haptic perception (Binkofski et al. Reference Binkofski, Buccino, Posse, Seitz, Rizzolatti and Freund1999a) have recently been seen as responsible for a somatosensory-guided action (how); that is, for adaptive grip-force control during lifting (e.g., Schmitz et al. Reference Schmitz, Jenmalm, Ehrsson and Forssberg2005). Even if, up to now, differences between the structural interplay in haptic perception and in lifting actions might not have been resolved in detail, a great overlap in networks casts additional doubts on the usefulness of a structural distinction between action and perception streams.

Note, besides, that adaptive grip-force control represents a class of actions that is defined but otherwise neglected in the target article: the manipulation of external objects. This includes several actions that require fast and permanent online processing of somatosensory information from the objects – as in throwing or lifting. Differences between the online processing of sensory information for time-constraint somatosensory-guided action in comparison to (probably) more precision-oriented haptic exploration (e.g., in the case of lifting versus weight estimation), are definitely worth future investigations in terms of functional dissociation. In contrast, what the target article does discuss extensively are functional dissociations between perception and cases of action for which we have quite well established models, especially regarding how limits in somatosensory processing are dealt with. Efference copies, well-learned motor programs, and anticipatory parameter control represent ways in which (coarse, as opposed to fine) action control circumvents such limits (e.g., Drewing et al. Reference Drewing, Stenneken, Cole, Prinz and Aschersleben2004), and, in functional terms, may partly explain why deafferented patients can perform several spatially oriented movements (such as pointing to a body location) but are not able to sense stimuli with the deafferented limb. For example, the patient cited in section 2 (para. 5) had a complete loss of sensibility, but could still perform a “pincer grip and exploratory movements.” Note that this patient's “exploratory movement patterns … of course were never performed in an adequate relation to the explored object” (Pause et al. Reference Pause, Kunesh, Binkofski and Freund1989, p. 1611), and her pincer grip was guided by vision (and was successful in 70%, but not 100%, of trials). The aforementioned well-known “shortcuts” of the how-stream can easily explain this performance because they need not refer to actual somatosensory information. To take another example from the target article, the weaker so-called vibro-tactile illusions (that relate to proprioceptive receptors in the muscle) in pointing, as compared to in perception, and the vanishing of this dissociation after some seconds: these findings agree well with the use of the how-stream–related shortcuts (e.g., fast updating of limb position by efference copy) until sufficient processing time has passed to include actual sensory information (i.e., the illusory information) – but these findings do not agree with dissociated structures.

Taken together, the target article provides convincing evidence of a functional dissociation between “how” and “what” streams while failing to disentangle the functional description from structure. The evidence also raises an urgent question that is only marginally addressed in the article: How is the interplay of the two streams in somatosensory processing coordinated? From experimental work in vision, we know that the two functional processing streams of how and what are coordinated by visual attention. They do not work independently. Evidence for this claim comes from behavioral work on the role of attentional processes in perception and spatial-motor action control (Deubel & Schneider Reference Deubel and Schneider1996; Schneider & Deubel Reference Schneider, Deubel, Prinz and Hommel2002). Several studies have shown that the selection process of an external target object for spatial movements, such as saccading or grasping (how), is tightly coupled to the selection of the same object for visual (conscious, reportable) perception (what). When humans prepare a movement to an object in space (i.e., select a movement target there), visual perceptual processing resources are focused on that target object. It is plausible to assume that a similar coupling holds for somatosensory processing. Up to now, there are no corresponding studies in somatosensory processing. This holds even more with regard to the more complex, but highly relevant question of coordinating different processing streams from distinct modalities.