In an important attempt to integrate different literatures, Dijkerman & de Haan (D&dH) provide a scholarly review on somatosensory processing. In analogy to the visual system, they propose two distinct cortical somatosensory streams subserving perception and action. They argue convincingly, albeit somewhat in contrast to the notion of separate streams, that tactile object recognition requires the active exploration of the objects with finger and hand movements in order to capture properties like shape and texture. Hence, recognition requires close coordination between action-related and recognition-related processes, as well as temporal and spatial integration of tactile information, and therefore depends on an intricate collaboration of the posterior parietal and the insular system (sect. 3.1). This proposal goes beyond the idea of two strictly separate pathways for perception and action, and raises three related questions which I pursue here: How separate are perception and action in vision? Is the posterior parietal cortex involved in visual recognition and categorization? Do perception and action rely on different processing principles?

First, given their close interrelation in somatosensory processing, are perception and action really separate in vision? In a similar way as in tactile recognition, active exploration strategies are used in visual recognition. For instance, observers systematically explore the visual scene with eye movements (e.g., Yarbus Reference Yarbus1967). More generally, seeing has been regarded as a way of acting, of exploring the environment (O'Regan & Noë Reference O'Regan and Noë2001). Moreover, there is abundant evidence for perception-action couplings in vision (Hommel et al. Reference Hommel, Müsseler, Aschersleben and Prinz2001; Prinz Reference Prinz, Neumann and Prinz1990, Reference Prinz1997; Sommerville & Decety Reference Sommerville and Decety2006). Studies investigating visuomotor priming have demonstrated that observing manipulable objects automatically potentiates actions (e.g., Craighero et al. Reference Craighero, Fadiga, Rizzolatti and Umilta1999; Pavese & Buxbaum Reference Pavese and Buxbaum2002; Tucker & Ellis Reference Tucker and Ellis1998; Reference Tucker and Ellis2001; Reference Tucker and Ellis2004). In accordance, premotor and parietal areas are active when manipulable objects are perceived that afford motor interactions, like tools (e.g., see Chao & Martin Reference Chao and Martin2000; Grèzes et al. Reference Grèzes, Tucker, Armony, Ellis and Passingham2003). More direct evidence for interactions from perception for action to the object-recognition system comes from a study showing that knowledge about motor interactions with objects improves recognition performance even in a naming (recognition) task (Helbig et al. Reference Helbig, Graf and Kiefer2006). Thus, there seems to be a strong coupling between perception and action not only in tactile, but also in visual recognition.

Second, is the posterior parietal cortex (PPC) involved in visual recognition, as well? The target authors claim that posterior parietal activity is specific for tactile recognition. According to the proposals of two separate visual streams, object recognition and categorization occur in the ventral stream and do not include parietal cortex (Goodale & Milner Reference Goodale and Milner2004; Milner & Goodale Reference Milner and Goodale1995; Ungerleider & Haxby Reference Ungerleider and Haxby1994). However, there is evidence from neuropsychological patients, monkey lesion studies, and neuroimaging studies that the visual recognition of misoriented objects involves the parietal cortex (for a brief review, see Graf Reference Graf2006). A recent experiment using transcranial magnetic stimulation confirmed that the parietal cortex is involved in object recognition (Harris & Miniussi Reference Harris and Miniussi2006). The parietal cortex is also active in the visual categorization of motion directions (Freedman & Assad Reference Freedman and Assad2006; see also Ferrera & Grinband Reference Ferrera and Grinband2006). Similarly, the categorization of distorted (dot pattern) prototypes involves both lateral occipital and parietal activations (Seger et al. Reference Seger, Poldrack, Prabhakaran, Zhao, Glover and Gabrieli2000; Vogels et al. Reference Vogels, Sary, Dupont and Orban2002).

Third, are processing principles and reference frames in perception and action really so different? The notion of two separate streams is connected with the assumption that the streams are based on different processing principles (Milner & Goodale Reference Milner and Goodale1995). Visuomotor control requires coordinate transformations, because receptor surfaces and motor effectors rely on different coordinate systems (e.g., eye centered vs. hand centered). Object recognition, in contrast, is thought to rely on the detection of enduring object properties. Similarly, it is assumed that visuomotor control implies egocentric reference frames, whereas recognition relies on object-centered representations (e.g., Milner & Goodale Reference Milner and Goodale1995). Therefore, recognition should be achieved irrespective of viewpoint (sect. 5 of the target article). However, the large majority of findings indicate that visual recognition performance is orientation (viewpoint) and size dependent (for reviews, see Graf Reference Graf2006; Tarr Reference Tarr, Peterson and Rhodes2003; Tarr & Bülthoff Reference Tarr and Bülthoff1998), suggesting that object representations are coded in viewer-centered (i.e., egocentric) coordinates.

Interestingly, recognition performance is not only orientation and size dependent, but shows orientation and size congruency effects. The ability to identify a misoriented object is facilitated if the object is preceded by a different object shown in the same orientation (e.g., Graf et al. Reference Graf, Kaping and Bülthoff2005). Therefore, it has been proposed that object recognition is based on an adjustment of a perceptual coordinate system (coordinate transformation) that aligns memory and input representations. Recognition is facilitated if the coordinate system is already adjusted in the correct orientation or size (because of the presentation of the previous object). According to this approach, both object recognition and visuomotor control rely on coordinate transformations (Graf Reference Graf2006; Graf et al. Reference Graf, Kaping and Bülthoff2005); coordinate transformations can be regarded as a common processing principle for visuomotor control and object recognition (Graf Reference Graf2006; Salinas & Abbott Reference Salinas, Abbott and Nicolelis2001; Salinas & Sejnowski Reference Salinas and Sejnowski2001).

Similar processes seem to underlie tactile recognition. Tactile recognition, too, is orientation dependent (Newell et al. Reference Newell, Ernst, Tjan and Bülthoff2001; Pasqualotto et al. Reference Pasqualotto, Finucane and Newell2005), which suggests viewer-dependent reference frames. Moreover, the representations can be updated with observer movement or scene context, both for visual and for tactile recognition (Pasqualotto et al. Reference Pasqualotto, Finucane and Newell2005). This suggests spatial updating processes with observer movement in tactile and visual recognition, consistent with the notion that recognition relies on transformations of a spatial coordinate system, defined by scene context (Graf Reference Graf2006).

To summarize, visual and somatosensory processing are even more similar than proposed in the target article by D&dH. Both in somatosensory and visual processing, perception for action and perception for object recognition are not strictly separate, but seem to interact extensively. There is evidence that the PPC is involved not only in tactile recognition, but also in visual recognition and categorization. Moreover, the two streams seem to rely on similar processing principles. Overall, it may be more appropriate to assume two cortical streams that serve complementary purposes and tend to interact to overcome the mutual deficiencies of each stream (Grossberg Reference Grossberg2000).