2. Feature processing in the anterior parietal cortex

The main processing pathways of the somatosensory system from peripheral receptors to the cortex that are concerned with touch and proprioception are well known. Input from peripheral receptors ascends through the dorsal column in the spinal cord and subsequently arrives in the medulla. The fibres then decussate in the medial lemniscus and terminate in the ventral posterior lateral nucleus (VPL) of the thalamus (Martin & Jessell Reference Martin, Jessell, Kandel, Schwartz and Jessell1991; Mountcastle Reference Mountcastle and Brookhart1984). In addition, there are projections to the ventral posterior inferior nucleus (VPI) and the posterior nuclei group of the thalamus (Mountcastle Reference Mountcastle and Brookhart1984). A second ascending system, the anterolateral system, mainly deals with thermal and noxious stimuli, but also relays some pressure information. The anterolateral system also projects to the VPL in addition to smaller projections to VPI and centromedian (CM)/parafascicular complex and the intralaminar nuclei (Berkley Reference Berkley1980; De Vito & Simmons Reference De Vito and Simmons1976; Sinclair Reference Sinclair1981).

Most somatosensory information enters the cerebral cortex through projections from the VPL to the anterior parietal cortex (APC) (Jones & Powell Reference Jones and Powell1970; Jones et al. Reference Jones, Wise and Coulter1979; Whitsel et al. Reference Whitsel, Rustioni, Dreyer, Loe, Allen and Metz1978). This area was originally referred to as the first somatosensory cortex (SI), but more recently it has been suggested that only Brodmann area (BA) 3b can be considered to be the homologue of the primary area SI in non-primates (Kaas Reference Kaas1983; Reference Kaas, Paxinos and Mai2004). In addition, there are projections from the ventroposterior superior nucleus (VPS) to areas 3a and 2 (Cusick et al. Reference Cusick, Steindler and Kaas1985). Finally, small projections exist between the VPL and other thalamic nuclei to the secondary somatosensory cortex (SII), the posterior parietal and insular cortex (Burton & Jones Reference Burton and Jones1976; Friedman & Murray Reference Friedman and Murray1986; Jones et al. Reference Jones, Wise and Coulter1979; Whitsel & Petrucelli Reference Whitsel and Petrucelli1969).

The organisation of the APC can be characterised by several principles. First, it consists of four different Brodmann areas, BAs 1, 2, 3a, and 3b, containing several somatotopic maps of the contralateral half of the body (Chen et al. Reference Chen, Friedman and Roe2005; Kaas et al. Reference Kaas, Nelson, Sur, Lin and Merzenich1979). Neurophysiological studies suggest that one somatosensory aspect tends to dominate the input to each area. In area 3a, the dominant input originates in the muscle receptors (Phillips et al. Reference Phillips, Powell and Wiesendanger1971; Tanji & Wise Reference Tanji and Wise1981), although the hand and digit areas of BA 3a also contain a significant number of cutaneous neurons. In BA 3b (Tanji & Wise Reference Tanji and Wise1981) and BA 1, the main input originates in cutaneous receptors (Whitsel et al. Reference Whitsel, Dreyer and Roppolo1971). Within these areas, specific cortical domains are activated by different types of stimulation, such as vibration, pressure, or flutter (Friedman et al. Reference Friedman, Chen and Roe2004). Lesions in each of these areas cause perceptual impairments on tasks that require processing of the relevant modality. In monkeys, area 3b appears to be involved in most tactile discrimination tasks. Removal of area 3a severely impairs performance on several discrimination tasks, including hard-soft, roughness, concave-convex, and square-diamond discriminations (Randolph & Semmes Reference Randolph and Semmes1974). The neurons in area 1 are predominantly cutaneous, but respond to more complex stimuli, such as movement detection and the direction sensitivity (Gardner Reference Gardner1988; Hyvärinen & Poranen Reference Hyvärinen and Poranen1978). In humans, area 1 contains larger representations (Overduin & Servos Reference Overduin and Servos2004). Removal of area 1 impairs discriminations involving texture (e.g., roughness, hard-soft) (Randolph & Semmes Reference Randolph and Semmes1974). The main input to BA 2 originates in deep receptors (joint and muscle afferents; Merzenich et al. Reference Merzenich, Kaas, Sur and Lin1978). Indeed, removal of this area in primates results in impaired performance on tasks involving kinaesthetic input, such as discrimination of concave versus convex and diamond versus square shapes (Randolph & Semmes Reference Randolph and Semmes1974). Area 2 also contains modules of cutaneous neurons with complex receptive fields and response properties in the hand and digit areas (Gardner Reference Gardner1988; Hyvärinen & Poranen Reference Hyvärinen and Poranen1978; Pons et al. Reference Pons, Garraghty, Cusick and Kaas1985). In the early stages of cortical processing, the neuronal responses represent the characteristics of stimuli applied to peripheral nerves relatively accurately (Phillips et al. Reference Phillips, Johnson and Hsiao1988). Neurons situated further away from the thalamic input have more complex response properties, which suggests that advanced processing occurs. For example, electrophysiological studies showed that direction-sensitive neurons were found less commonly in area 3b, but more densely in areas 1 and 2 (Gardner Reference Gardner1988; Hyvärinen & Poranen Reference Hyvärinen and Poranen1978). Indeed, lesions of areas 3b, 1 and 2 cause impairments in distinguishing the speed of tactile movement (Zainos et al. Reference Zainos, Merchant, Hernandez, Salinas and Romo1997). In humans, lesions affecting the postcentral gyrus cause deficits in two-point discrimination; position sense and point localisation; object size, shape, and texture discrimination (Corkin et al. Reference Corkin, Milner and Rasmussen1970; Kaas Reference Kaas, Paxinos and Mai2004; Roland Reference Roland1987).

The combined lesion and neurophysiological literature suggests that the APC is important for the processing of simple somatosensory features related to both the stimulus and the part of the body that has been stimulated. Moreover, recent optical imaging of a tactile illusion suggests that the APC codes the perceived rather than physical location of peripheral stimuli (Chen et al. Reference Chen, Friedman and Roe2003). This finding suggests that neural processing is related to what the information is processed for (e.g., perception) rather than the stimulus characteristics. Thus, it is consistent with the idea that the purpose is at least as important when discussing the neural basis of sensory processing. Indeed, certain types of input may be more important for certain tasks, with proprioceptive input contributing more to action-related processes, and the skin receptors providing more information for perceptual purposes. However, this mapping of different somatosensory submodalities to output is by no means absolute.

With respect to somatosensory processing for the guidance of action, the findings of neuropsychological studies suggest that damage to the APC does not necessarily abolish accurate pointing movements. Several investigators have reported patients with damage to the primary somatosensory cortex who showed severe impairments in tactile perception while their motor deficits were surprisingly mild (Brochier et al. Reference Brochier, Habib and Brouchon1994; Halligan et al. Reference Halligan, Hunt, Marshall and Wade1995; Pause et al. Reference Pause, Kunesh, Binkofski and Freund1989; Volpe et al. Reference Volpe, LeDoux and Gazzaniga1979). In a seminal study, Volpe et al. (Reference Volpe, LeDoux and Gazzaniga1979) reported four patients with tactile and proprioceptive deficits following a stroke, who were nevertheless able to perform spatially oriented movements with the de-afferented hand. In the study by Pause et al. (Reference Pause, Kunesh, Binkofski and Freund1989), the patient, who had a total loss of sensibility and tactile recognition with the contralesional hand, remained able to perform several motor acts, including the pincer grip and exploratory movements. Brochier et al. (Reference Brochier, Habib and Brouchon1994) observed that their patient could touch the thumb of the insensate hand with each finger individually (similar observations were made also in a patient assessed by Halligan et al. Reference Halligan, Hunt, Marshall and Wade1995) and could crumble a piece of paper with this hand, while having severe difficulties in recognizing the direction of movements on her skin and letters drawn on her hand.

Spared sensorimotor guidance to targets on the impaired arm that were not perceived, has been reported in other studies. In a first description of this kind, Paillard et al. (Reference Paillard, Michel and Stelmach1983) reported a patient with a left posterior cortical lesion who could point to tactile targets on her right hand that she was unable to detect. They suggested that there were striking similarities with blindsight. A similar dissociation was reported by Rossetti and colleagues (see Rossetti et al. Reference Rossetti, Rode and Boisson1995b; Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001), who investigated a patient with a lesion affecting the thalamic nucleus VPL. They assessed his ability to use touch and proprioception for verbal and pointing localisation responses. Localisation of targets by using both touch and proprioception was above chance only when a pointing response was made aimed directly at the target (see Fig. 2). Verbal responses or pointing responses on a drawing of the arm were at chance. Delaying the motor response also reduced performance to chance levels. Rossetti et al. argued that their patient showed a dissociation between the what (object recognition) and the how (sensorimotor) systems and coined the term “numbsense” for this phenomenon. Another study, by Aglioti et al. (Reference Aglioti, Betramello, Bonazzi and Corbetta1996), reported similar findings. In addition, Aglioti et al. found above-chance performance when subjects pointed with their insensate hand to the location of stimulation applied to their normal hand.

Figure 2

Performance of numbsense patient J.A. on two pointing tasks. In the simple pointing task, the patient pointed with his left hand directly towards a tactile stimulus on the impaired right arm or hand (left picture). He clearly performed this task at above chance levels. In the second task, he was asked to indicate the position of the tactile target by pointing to its location on a drawing of the right hand (right picture). Performance on this task was not different from chance. Reproduced by permission of Oxford University Press (www.oup.com) from Rossetti et al. (Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001), Fig. 15.4, p. 275.

Thus, lesion studies suggest that somatosensory information can be used for the guidance of movements after lesions to the APC or the VPL, although the stimuli cannot be detected consciously. While some “numbsense” studies assessed pointing to proprioceptive or tactile targets on the insensate arm (Aglioti et al. Reference Aglioti, Betramello, Bonazzi and Corbetta1996; Rossetti et al. Reference Rossetti, Rode and Boisson1995b), others showed that movements with the impaired arm were still possible despite the absence of proprioceptive feedback (Brochier et al. Reference Brochier, Habib and Brouchon1994; Pause et al. Reference Pause, Kunesh, Binkofski and Freund1989; Volpe et al. Reference Volpe, LeDoux and Gazzaniga1979).

Regarding the neural substrate of these unconscious residual sensorimotor abilities, several possibilities have been suggested. Brochier et al. (Reference Brochier, Habib and Brouchon1994) and Rossetti and colleagues proposed that thalamic projections to the PPC, bypassing the APC and the VPL, may be responsible (Rossetti Reference Rossetti1998; Rossetti et al. Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001). Brochier et al. (Reference Brochier, Habib and Brouchon1994) and Rossetti et al. (Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001) identified projections from the posterior lateral nucleus and medial portion of the posterior complex to the PPC (cf. Jones et al. Reference Jones, Wise and Coulter1979; Pearson et al. Reference Pearson, Brodal and Powell1978) as possible substrate. A second possibility may be that direct projections from the VPL to the motor cortex are involved (Jeannerod et al. Reference Jeannerod, Michel and Prablanc1984), although this may be less likely in the case of Rossetti et al. because their patient's lesion primarily affected this thalamic nucleus. A final suggestion is that small ipsilateral pathways to the intact hemisphere may be responsible (Rossetti et al. Reference Rossetti, Rode and Boisson1995b; Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001). Supportive evidence for this notion comes from studies with hemispherectomized patients who usually do retain some crude somatosensory function on the side contralateral to the removed hemisphere (Dijkerman Reference Dijkerman1996; Holloway et al. Reference Holloway, Gadian, Vargha-Khadem, Porter, Boyd and Connelly2000; Muller et al. Reference Muller, Kunesch, Binkofski and Freund1991). Whether these pathways are also involved in certain basic aspects of conscious perception, such as stimulus detection, remains as yet to be determined.

The main lesson to be learned for the issue at hand is that – as is the case for the visual system – the execution of a motor action towards a spatially defined target does not necessarily depend on conscious awareness of that target. This observation supports the idea of two separate somatosensory pathways for action and conscious perception.

3. Higher somatosensory cortical processing

Processing of somatosensory input beyond the APC occurs in several cortical areas. These include the secondary somatosensory area (SII), the insula, and the PPC. Overall, somatosensory projections involving these areas are characterised by serial, as well as parallel, processing (Iwamura Reference Iwamura1998; Knecht et al. Reference Knecht, Kunesch and Schnitzler1996). The APC maintains reciprocal connections with the SII (Barbaresi et al. Reference Barbaresi, Minelli and Manzoni1994; Disbrow et al. Reference Disbrow, Litinas, Recanzone, Padberg and Krubitzer2003; Friedman et al. Reference Friedman, Jones and Burton1980; Pons & Kaas Reference Pons and Kaas1986), although the projections from the APC to the SII are more important than those from SII to the APC (Pons et al. Reference Pons, Garraghty, Friedman and Mishkin1987). Neurons in the SII have greater stimulus selectivity, larger receptive fields, reduced modality specificity, and respond to ipsilateral, as well as contralateral, stimulation (Disbrow et al. Reference Disbrow, Roberts, Poeppel and Krubitzer2001; Lin & Forss Reference Lin and Forss2002; Ruben et al. Reference Ruben, Schwiemann, Deuchert, Meyer, Krause, Curio, Villringer, Kurth and Villringer2001; Sinclair & Burton Reference Sinclair and Burton1993). The SII is reciprocally connected with granular and dysgranular fields of the insula (Friedman et al. Reference Friedman, Murray, O'Neill and Mishkin1986). Neurophysiological recordings from the granular insula in rhesus monkeys showed that a major portion of this area is exclusively devoted to somatic processing (Schneider et al. Reference Schneider, Friedman and Mishkin1993).

The SII has additional projections to the posterior parietal area 7b, both ipsilaterally and contralaterally, and the premotor cortex in the same hemisphere (Disbrow et al. Reference Disbrow, Litinas, Recanzone, Padberg and Krubitzer2003; Friedman et al. Reference Friedman, Murray, O'Neill and Mishkin1986). In monkeys, the PPC also receives direct connections from the APC. Area 5 (superior parietal cortex) receives input from areas 1 and 2 (Pearson & Powell Reference Pearson and Powell1985; Pons & Kaas Reference Pons and Kaas1986), whereas area 7b (inferior parietal cortex) receives direct input from area 1 (Pons & Kaas Reference Pons and Kaas1986). Several thalamic nuclei also project directly to the SII (Disbrow et al. Reference Disbrow, Litinas, Recanzone, Slutsky and Krubitzer2002) and to different parts of the PPC (Friedman & Murray Reference Friedman and Murray1986; Jones et al. Reference Jones, Wise and Coulter1979). Major cortical outputs from the PPC project back to the SII and to the premotor cortex, the limbic cortex, and the superior temporal sulcus (Kaas Reference Kaas, Paxinos and Mai2004).

An important question is whether two processing streams can be discerned in this network of connections, analogous to the visual system (Milner & Goodale Reference Milner and Goodale1995; Ungerleider & Mishkin Reference Ungerleider, Mishkin, Ingle, Goodale and Mansfield1982). Indeed, it has been suggested that the projections from the APC via the SII to the granular and dysgranular fields of the insula are involved in tactile perception and learning (and thus would constitute the somatosensory equivalent to the visual “what” pathway; see sect. 3.1) (Friedman et al. Reference Friedman, Murray, O'Neill and Mishkin1986; Mishkin Reference Mishkin1979). The tactile “dorsal stream” would involve projections from the APC to the PPC (Westwood & Goodale Reference Westwood and Goodale2003), either directly or through the SII. In a study of anatomical connections of two areas within the SII (the SII proper and the parietal ventral area [VP]) in monkeys, Disbrow et al. (Reference Disbrow, Litinas, Recanzone, Padberg and Krubitzer2003) suggested that the interconnections between these two areas in the SII overlapped considerably. They suggested that this was inconsistent with the idea of two separate processing streams. Indeed, Disbrow et al. (Reference Disbrow, Litinas, Recanzone, Padberg and Krubitzer2003) indicated that the SII may be involved in somatosensory processing for both perception and action. In our view, the data of Disbrow et al. are not at all inconsistent with the idea of separate processing pathways for perception and action, as segregation may not occur until after the SII.

In a recent fMRI study, Reed et al. (Reference Reed, Klatzky and Halgren2005) investigated whether a “what” versus “where” dissociation also exists for somatosensory processing. They compared tactile object recognition with tactile object localisation while controlling for differences in exploratory finger movements. Differential activation patterns were observed with tactile object recognition activating frontal, as well as bilateral, inferior parietal areas. In contrast, the tactile object location task was associated with activation in bilateral superior parietal areas. Note that these authors link their investigations with to the distinction between spatial and object vision made by Ungerleider and Mishkin (Reference Ungerleider, Mishkin, Ingle, Goodale and Mansfield1982). Indeed, Reed et al. (Reference Reed, Klatzky and Halgren2005) specifically controlled for differences in the action component, e.g., exploratory finger movements.

In the present review, instead of the “what” versus “where” distinction, we discuss evidence from patient, functional imaging, and neurophysiology studies for separate processing for somatosensory perception and action (“what” vs. “how”). As action and perception are both broadly defined functions that can include many different aspects, a further subdivision is made between processing of external target features (e.g., for object recognition) and somatosensory input about the body (e.g., the body representations). Indeed, functional differences have been found between judgements pertaining to the position of the stimulus on the body and in external space (Kitazawa Reference Kitazawa2002). Differences in perception and action are therefore reviewed separately for internal and external somatosensory information, respectively.

3.2. Cortical processing of somatosensory information pertaining to the body

Perhaps an even more important function of the somatosensory system is informing us about the position of our different body parts with respect to one another. To achieve this, tactile and proprioceptive input needs to be integrated with visual and vestibular input into a representation of the body. Evidence concerning body representations comes from different sources. Studies of neurological patients show that a variety of lesions to the peripheral and central nervous systems can result in changes of body representations. For example, some limb amputees report that their phantom limb can change in size and form over time (Berlucchi & Aglioti Reference Berlucchi and Aglioti1997). Lesions to the central nervous system can produce disorders such as anosognosia (denial of symptom) for motor and sensory deficits after a stroke (Berlucchi & Aglioti Reference Berlucchi and Aglioti1997; Levine et al. Reference Levine, Calvanio and Rinn1991), denial of ownership of a body part, and misplegia (hatred of hemiparetic limbs) (Berlucchi & Aglioti Reference Berlucchi and Aglioti1997; Moss & Turnbull Reference Moss and Turnbull1996). These disorders usually, but not always, are found after right hemisphere lesions and may be accompanied by reports of supernumerary limbs (Halligan & Marshall Reference Halligan and Marshall1995), suggesting that negative and positive syndromes share common neural mechanisms. Left posterior parietal brain lesions can result in other impairments of body representations, such as autotopagnosia (inability to localise body parts), finger agnosia, and left-right disorientation (Denes Reference Denes, Denes and Pizzamiglio1999).

The important question for the present review is whether different representations of the body are used for the guidance of movements, as compared to perceptual judgements about the spatial relations of the different body parts. Several authors have suggested that this indeed may be the case. For example, Paillard (Reference Paillard, Gantchev, Mori and Massion1999) distinguished between body schema and body image. Body image was described as a “perceptual identification of body features” related to an internal representation of sensory and motor input of corporeal origin. This internal representation would be accessible to conscious experience. In contrast, the body schema refers to the location of different body parts in a sensorimotor map of body space, which is not accessible to consciousness. In addition to the two body representations described by Paillard, other authors also distinguished a third representation containing conceptual and semantic knowledge about the body (Buxbaum & Coslett Reference Buxbaum and Coslett2001; Guariglia et al. Reference Guariglia, Piccardi, Puglisi Allegra and Traballesi2002; Schwoebel et al. Reference Schwoebel, Coslett and Buxbaum2001; Sirigu et al. Reference Sirigu, Grafman, Bressler and Sunderland1991).

Evidence for separate body representations from neuropsychological studies was first described almost a century ago (Head & Holmes 1911–1912). More recently, Paillard (Reference Paillard, Gantchev, Mori and Massion1999) suggested that the “numbsense” patients mentioned earlier in this review have a specific deficit in the perceptual representation of target (body image), while the sensorimotor representation (body schema) remains unaffected. He described the opposite dissociation in a patient who suffered from peripheral deafferentation, but with an intact motor system. She was able to verbally identify the location of tactile stimuli, but was poor at pointing towards the stimulus (see Fig. 3). This would be consistent with an impairment in body schema, with preserved body image. A similar pattern (impaired pointing in combination with intact verbal report) has also been reported by Halligan et al. (Reference Halligan, Hunt, Marshall and Wade1995).

Figure 3

Performance of peripherally de-afferented patient G.L. on three pointing tasks. Cold tactile stimuli were applied to various locations on her left hand. In the “without vision” condition, GL was greatly impaired in pointing towards the stimuli. Performance was considerable better when allowed vision of her left hand. Remarkably, her performance was similar to the “vision” condition when asked to point to the location of the stimulus on a picture of the left hand. From Paillard (Reference Paillard, Gantchev, Mori and Massion1999) with permission from Academic Publishing House, Sofia, Bulgaria.

Other evidence for separate body representations involved in the guidance of action and perceptual recognition of objects comes from patients with autotopagnosia. Buxbaum and Coslett (Reference Buxbaum and Coslett2001) described a patient who was unable to point to any of his body parts, or to those of another person, yet could perform visually guided grasping movements. The latter suggests that his impairment was unlikely to be the consequence of an impairment in body schema. Furthermore, he was able to point to objects attached at different locations to the body (see also Sirigu et al. Reference Sirigu, Grafman, Bressler and Sunderland1991). Buxbaum and Coslett suggested that their patient had an impaired “system of structural descriptions of the body and its parts which defines the position of body parts relative to one another in a perceptual . . . format” (p. 302). In a recent study of body representation disorders in a group of 70 stroke patients, Schwoebel and Coslett (Reference Schwoebel and Coslett2005) observed a triple dissociation between measures of three putative body representations (body schema, body structural description, and body semantics). They linked the left temporal cortex to structural and semantic knowledge, and the dorsolateral frontal and posterior parietal lesions to body schema.

Other authors have implicated a network that includes the APC, the PPC, and the insula to be involved in bodily awareness and perception (Berlucchi & Aglioti Reference Berlucchi and Aglioti1997; Melzack Reference Melzack1990). Each area is supposed to play a different role. Thus, lesions to the APC result in tactile and proprioceptive impairments, but not in higher-order body-awareness deficits. The SII appears to be important for integration of information from the two body halves (Hari et al. Reference Hari, Hanninen, Makinen, Jousmaki, Forss, Seppa and Salonen1998; Lin & Forss Reference Lin and Forss2002). Lesions to the PPC can result in alterations in higher-order body awareness, as mentioned previously, with lesions to the left and the right PPC producing different types of deficits. Functional imaging studies also suggest that the PPC is involved in body representations. A positron emission tomography (PET) study showed activation in the superior PPC, intraparietal sulcus, and adjacent inferior parietal lobule during mental transpositions of the body in space (Bonda et al. Reference Bonda, Petrides, Frey and Evans1995) In a more recent study, Ehrsson et al. (Reference Ehrsson, Kito, Sadato, Passingham and Naito2005b) used the vibrotactile illusion to study the neural basis of body representations and, specifically, body image. Participants were asked to hold their waist while vibrotactile stimulation was applied to the wrist extensors. They experienced predictable changes in waist size. This illusion was related to activation in the left postcentral sulcus, but also in the anterior part of the left intraparietal sulcus, suggesting that this area is involved in the perceptual experience of body size.

In addition to the PPC, the insula may be involved in corporeal awareness. Patients with insular lesions may experience somatic hallucinations (Roper et al. Reference Roper, Levesque, Sutherling and Engel1993) or somatoparaphrenia (Cereda et al. Reference Cereda, Ghika, Maeder and Bogousslavsky2002). A recent lesion overlap study implicated the posterior insula in bodily awareness (Karnath et al. Reference Karnath, Baier and Nägele2005). Functional imaging studies also suggest that the insula activation may be related to a sense of ownership and agency (Farrer et al. Reference Farrer, Franck, Georgieff, Frith, Decety and Jeannerod2003). Others have related the insula to subjective awareness and affective processing of bodily signals (Craig Reference Craig2002, Reference Craig2004). Craig has proposed a separate pathway relaying input regarding the ongoing condition of different organs of the body (viscera, muscles, joints, skin, etc.). This pathway mainly relays information through small-diameter afferent fibres to the spinal dorsal horn and the solitary nucleus in the medulla as part of the autonomic nervous system. It not only provides input as part of homeostatic mechanisms to the hypothalamus and brain stem, but also projects through the ventral medial nucleus of the thalamus to the posterior insula (see also Olausson et al. Reference Olausson, Lamarre, Backlund, Morin, Wallin, Starck, Ekholm, Strigo, Worsley, Vallbo and Bushnell2002). Further projections exist from the posterior insula to the anterior insula. Craig has suggested that this pathway is related to our subjective awareness of our body and bodily emotions (“how you feel”). Thus, in this model, emotional processing is related to activity of the autonomic nervous system. Craig (Reference Craig2005) has additionally suggested different roles for the left and the right insulae. The right insula is suggested to be related to “aroused” emotions linked to the sympathetic system. In contrast, the left insula is considered to be involved in “parasympathetic” or “enrichment” emotions.

Overall, the studies reviewed here suggest that the insula is concerned with higher-order somatosensory processing of the body that is either related to a sense of ownership or to emotional experience. In contrast, the posterior parietal cortex may be more concerned with metric aspects of the body, such as its spatial configuration and size. Although the data are not entirely consistent, the overall picture that emerges is that the anterior part of the intraparietal sulcus is consistently involved.

With respect to the neural basis of action-related body representations, again a network of different neural structures appears to be involved. Activity related to arm movement occurs already early in the somatosensory system. Neurons in the postcentral gyrus have been found to modulate their responsiveness to arm movement, depending on whether the arm movement was made passively or actively (Prud'Homme & Kalaska Reference Prud'Homme and Kalaska1994). Furthermore, different sensorimotor channels can be detected in the APC. Iwamura and Tanaka (Reference Iwamura and Tanaka1996) reported that certain neurones in area 2 were activated only by self-generated hand actions to reach or to grasp objects. These neurones were preferentially active during reaching, a precision grip, or a whole hand grasp. A recent fMRI study confirmed the involvement of the APC in somatosensory-guided movements in humans (Wenderoth et al. Reference Wenderoth, Toni, Bedeleem, Debaere and Swinnen2006). These areas may provide kinaesthetic information to the multimodal visuomotor neurones in the PPC, although another possibility may be the direct connections to the motor cortex (Johnson et al. Reference Johnson, Ferraina, Bianchi and Caminiti1996; Marconi et al. Reference Marconi, Genovesio, Battaglia-Mayer, Ferraina, Squatrito, Molinari, Lacquaniti and Caminiti2001; Rizzolatti et al. Reference Rizzolatti, Luppino and Matelli1998) Of course, we also know that information about the movement of the arm can bypass the postcentral gyrus (see sect. 2).

The SII also seems to play a role in proprioception. Activation of the SII is modulated by isometric muscle contraction (Lin et al. Reference Lin, Simoes, Forss and Hari2000) and activation after electric median nerve stimulation is enlarged bilaterally when accompanied by exploratory finger movements (Huttunen et al. Reference Huttunen, Wikstrom, Korvenoja, Seppalainen, Aronen and Ilmoniemi1996).

A third area that plays a particularly important role is the PPC. Neurophysiological studies suggest that particularly area 5 is involved in somatosensory processing concerning the body during goal-directed arm movements. Neurones in this area respond to somatosensory stimulation that is reach-related (Colby Reference Colby1998; Gregoriou & Savaki Reference Gregoriou and Savaki2001), whereas others have been found active during prehension (Debowy et al. Reference Debowy, Ghosh, Ro and Gardner2001; Gardner et al. Reference Gardner, Ro, Debowy and Ghosh1999; Ro et al. Reference Ro, Debowy, Ghosh and Gardner2000). Furthermore, integration of tactile, and especially proprioceptive, input about the movement of the arm with sensory information about the target probably occurs in the posterior parietal lobe. There is considerable evidence that the PPC is involved in the transformation of visual signal into motor commands (Jeannerod et al. Reference Jeannerod, Arbib, Rizzolatti and Sakata1995; Kalaska et al. Reference Kalaska, Scott, Cisek and Sergio1997; Milner & Dijkerman Reference Milner, Dijkerman and Milner1998). Neurophysiological studies (Colby Reference Colby1998; Sakata et al. Reference Sakata, Takaoka, Kawarasaki and Shibutani1973; Savaki et al. Reference Savaki, Raos and Dalezios1997) and neuroimaging studies (Clower et al. Reference Clower, Hoffman, Votaw, Faber, Woods and Alexander1996; Kertzman et al. Reference Kertzman, Schwarz, Zeffiro and Hallett1997) suggest particular involvement of the area 7 in the integration of visual target information with proprioceptive limb information (see also sect. 6).

4. Behavioural dissociations in studies of healthy subjects

While the review so far suggests that there is considerable evidence for dissociated neural processing for somatosensory perception and action, relatively few studies have investigated possible behavioural dissociations in healthy participants. Westwood and Goodale (Reference Westwood and Goodale2003) assessed the effect of a haptic version of a visual size-contrast illusion on perceptual size matching and grasping responses. Subjects were required to explore with the left hand a flanker object placed underneath a table and subsequently hold a target object positioned adjacent to the flanker object. The flanker object could be smaller, larger, or identical to the target object. They were then asked to either perceptually estimate the size of the target object by varying the index finger–thumb distance of the right hand, or grasp an object identical to the target object on the table with that hand. Although the size estimates were influenced by the flanker object (smaller for the larger flanker object), no effect was found for the maximum grip aperture during grasping.

Whereas Westwood and Goodale (Reference Westwood and Goodale2003) investigated somatosensory processing of external objects, Kammers et al. (Reference Kammers, van der Ham and Dijkerman2006) used a vibrotactile illusion to explore differences between matching and reaching involving the position of a body part. The vibrotactile illusion is evoked by repetitive stimulation of a tendon of a muscle in one of the extremities. This stimulation induces the subjective experience of a movement of that extremity congruent with relaxation of that muscle. Two conditions were used. The direct condition, in which the biceps brachii tendon of the dominant arm was vibrated, created an illusory extension of the underarm. The indirect condition, wherein the ipsilateral knee was held with the vibrated arm, caused an illusion of lowering of the leg. In each condition, subjects were asked to make a reaching response (point with the index finger of the non-stimulated arm to the felt location of the ipsilateral index fingertip or top of the kneecap), as well as a matching response (mirroring the position of the non-stimulated arm or knee to the perceived position of the stimulated arm or kneecap). The illusion was significantly larger for the matching as compared to the reaching response, with the largest difference observed in the direct condition, suggesting that body representations underlying perception and action may be differentially sensitive to the illusion in healthy individuals. In an earlier study, Sittig and colleagues observed that subjects continued to reach to the correct target position when vibrotactile stimulation was applied to the moving arm (Sittig et al. Reference Sittig, Denier van der Gon, Gielen and van Wijk1985). In a more recent study, Marcel (Reference Marcel, Roessler and Eilan2003) also observed a difference in sensitivity to the vibrotactile illusion between the perceptual report and motor response. In addition, he reported that, after several seconds, the motor responses also became influenced by the vibrotactile illusion. This intriguing observation suggests that, with time, cognitive perceptual representations of limb position become dominant.

Overall, the findings of these behavioural studies are consistent with the idea that somatosensory processing for perception can be dissociated from those underlying action, although the Marcel (Reference Marcel, Roessler and Eilan2003) study suggests that interactions between the two representations can occur.

5. Comparison with other sensory modalities

As mentioned before, the idea that the guidance of action requires different sensory processing than does recognition is not new and was first proposed about 15 years ago for the visual cortical system (Goodale & Milner Reference Goodale and Milner1992; Jeannerod & Rossetti Reference Jeannerod, Rossetti and Kennard1993; Milner & Goodale Reference Milner and Goodale1995). Separate processing streams have been also been proposed for the auditory system (Belin & Zatorre Reference Belin and Zatorre2000; Rauschecker Reference Rauschecker1998). These similarities in proposed cortical organisation of the different sensory systems may suggest a common plan of how sensory input is processed by the brain (Belin & Zatorre Reference Belin and Zatorre2000; Rauschecker Reference Rauschecker1998). In this section, we compare the cortical organisation of the somatosensory system with that of other sensory systems. Although there are many similarities, some important differences exist, as well.

One of the central premises of the “separate visual cortical processing streams for perception and action” model is that neural processing is related to the way in which we use this information – for example, to store for later recognition or to program a motor action (Milner & Goodale Reference Milner and Goodale1995). Support for this idea, rather than for a distinction in terms of input characteristics (e.g., spatial vs. object vision), was based on mainly two lines of evidence. First, monkey neurophysiology suggested that posterior parietal regions also process non-spatial characteristics (size) for guidance of the hand (Sakata et al. Reference Sakata, Taira, Murata and Mine1995). Second, it has been argued that patients with neurological lesions indicate a double dissociation between visual processing for perception and for action. Patient D.F., who suffers from visual form agnosia, was impaired when required to give a perceptual judgement about the size or orientation of a visual stimulus, but was able to use the same stimulus characteristics for the guidance of hand movements (Goodale et al. Reference Goodale, Milner, Jakobson and Carey1991). In contrast, optic ataxic patient A.T. was impaired when required to grasp an object, while remaining able to judge the size of the object perceptually (Jeannerod et al. Reference Jeannerod, Decety and Michel1994). More recent fMRI evidence also seems consistent with this idea (Culham et al. Reference Culham, Danckert, DeSouza, Gati, Menon and Goodale2003; James et al. Reference James, Culham, Humphrey, Milner and Goodale2003). Milner and Goodale (Reference Milner and Goodale1995) further suggested that perceptual and action-related responses require different processing characteristics. The visuomotor system requires information about the position of the target in relation to the observer that is continuously updated. As a consequence, dorsal stream processing is characterised by egocentric reference frames and real-time computation with an inability to store the input for longer than a few seconds. In contrast, the perceptual system is able to recognize objects irrespective of its viewpoint, and it stores this information over long periods of time.

The visual ventral stream also plays a role in visuomotor control when it involves aspects that are characteristic for ventral stream processing, such as holding visual information during a delay or retrieval of object knowledge (Goodale Reference Goodale2001). Indeed, visual form agnosic patient D.F. was impaired when required to wait for as little as 2 seconds after stimulus presentation when grasping an object (Goodale et al. Reference Goodale, Jakobson and Keillor1994). In contrast, optic ataxic patients improved their performance after a delay (Milner et al. Reference Milner, Dijkerman, Pisella, McIntosh, Tilikete, Vighetto and Rossetti2001; Reference Milner, Dijkerman, McIntosh, Rossetti and Pisella2003), consistent with the idea that such patients are able to use to intact ventral stream to overcome their visuomotor deficit when a delay is introduced. A second example is that optic ataxic patient A.T. improved in ability-to-grip scale when grasping familiar as compared to unfamiliar objects (Jeannerod et al. Reference Jeannerod, Decety and Michel1994), whereas visual agnosic patient D.F. failed to take into account stored knowledge about the characteristics of well-known objects when programming her grip (Carey et al. Reference Carey, Harvey and Milner1996). Other studies suggest that the ventral stream is also active when subjects must consciously identify the visual context and decide on the appropriate action (Passingham & Toni Reference Passingham and Toni2001). Together, these findings are consistent with the idea of two separate, but interacting, visual cortical streams. The dorsal stream is involved in the visual guidance of immediate goal-directed hand and arm movements. The visual ventral stream is primarily associated with visual perception and recognition; however, it is also involved in certain aspects of the visual guidance of movement that require delayed action, object knowledge, or conscious decision making.

To what extent are similar processing characteristics applicable to the cortical somatosensory system as proposed here? As already described by others, for both the somatosensory and the visual systems, response characteristics become increasingly more complex the further away the neurones are from the thalamic input into the cortex. This involves increasing receptive fields (Lin & Forss Reference Lin and Forss2002; Ruben et al. Reference Ruben, Schwiemann, Deuchert, Meyer, Krause, Curio, Villringer, Kurth and Villringer2001; Sinclair & Burton Reference Sinclair and Burton1993), more complex stimuli required to activate the neurones (Gardner Reference Gardner1988; Hyvärinen & Poranen Reference Hyvärinen and Poranen1978), and so on. A second similarity is that residual unconscious processes can occur after lesions affecting primary cortical areas or the main thalamic relay station. Thus, implicit processing of tactile stimuli for action in the numbsense patients is very similar to the findings observed in the visual equivalent, that is, blindsight (Perenin & Rossetti Reference Perenin and Rossetti1996; Weiskrantz Reference Weiskrantz1996) – as was indeed observed by the original investigators (Paillard et al. Reference Paillard, Michel and Stelmach1983; Rossetti et al. Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001). Third, there is some evidence that introducing a delay between stimulus presentation and response has similar effects in the visual and the somatosensory system. Rossetti et al. (Reference Rossetti, Rode, Boisson, de Gelder, De Haan and Heywood2001) reported that performance of their numbsense patient reverted to chance levels when a delay was introduced. Similar findings were reported for action blindsight (Perenin & Rossetti Reference Perenin and Rossetti1996). In a study with healthy participants, Zuidhoek et al. (Reference Zuidhoek, Kappers, van der Lubbe and Postma2003) observed that the introduction of a delay reduced errors on a haptic parallel-setting task. They attributed the improved performance to a shift from the egocentric towards the allocentric reference frame during the delay period. These findings are consistent with a shift from dorsal to ventral stream processing in the visual system.

Perhaps one of the most striking differences between the two-visual-cortical-streams model and the present proposal is the inclusion of body-related representations. Whereas the two-visual-cortical-streams model deals only with visual input concerning external stimuli, the somatosensory system first and foremost provides information about our own body. A model of somatosensory cortical processing, in our opinion, would be incomplete without incorporating ideas about the neural basis of body representations. In contrast, a model of the visual cortical processing does not necessarily include body representations, as the optical array can provide direct information about the structure and position of external stimuli. Nevertheless, visual input about the observer's body is an important source of information, both during the guidance of movements and for perceptual awareness and recognition of your body. With respect to the neural basis of visual body representations, there is ample evidence for egocentric coding of external targets in a reference frame linked to specific body parts (Andersen Reference Andersen1997; Colby Reference Colby1998; Graziano et al. Reference Graziano, Cooke and Taylor2000); however, the neural correlates of visual representations of the observer's own body has received less attention. Recent functional imaging studies have reported activation in the extrastriate body area (EBA) when body stimuli have been viewed (Downing et al. Reference Downing, Jiang, Shuman and Kanwisher2001). While activation in this area is modulated by the position of the observer's body (Arzy et al. Reference Arzy, Thut, Mohr, Michel and Blanke2006b) and movement with the observer's (unseen) arm (Astafiev et al. Reference Astafiev, Stanley, Shulman and Corbetta2004), activation responds mainly to pictures of other people's body and is therefore not specifically involved in own body representations.

A second possible difference between the visual and somatosensory systems pertains to the degree of separation in the two cortical processing streams and the amount of spatio-temporal integration that is required. In the visual system, processing for action and for perception is clearly linked to separate cortical processing streams. Although more recent work suggests that the two processing streams are probably more interconnected than originally was thought and the ventral stream plays a (specific) role in several aspects of visuomotor processing, functional characteristics can, nevertheless, be related to separate neural processing streams. Our review of the literature suggests that the separation between action-related and perception-related processes may be less distinct in the somatosensory system. We found evidence for the involvement of both the PPC and the insula in perception and again the PPC for action-related processes. The involvement of posterior parietal processing for somatosensory perception may be related to a fundamental difference between somatosensory and visual systems. Whereas the visual array enables simultaneous processing of different stimuli and stimuli features, tactile exploration of stimulus features occurs in a more sequential manner. Increased temporal and spatial integration may be required in order to be able to integrate the tactile input into a coherent representation of the stimuli. The inferior parietal cortex seems particularly important for such integration (Saetti et al. Reference Saetti, de Renzi and Comper1999). The role of the PPC in somatosensory perception may be related to these increased demands on spatio-temporal integration. Interestingly, higher-order visual spatial functions have also been related to inferior posterior parietal processing. This has also been described as the ventro-dorsal stream (Pisella et al. Reference Pisella, Binkofski, Lasek, Toni and Rossetti2006; Rizzolatti & Matelli Reference Rizzolatti and Matelli2003) and has been related to disorders such as hemispatial neglect (after right hemisphere lesions) and apraxia (after left hemisphere lesions) (Milner & Goodale Reference Milner and Goodale1995; Pisella et al. Reference Pisella, Binkofski, Lasek, Toni and Rossetti2006; Rizzolatti & Matelli Reference Rizzolatti and Matelli2003). Several investigators have noted that somatosensory deficits often co-occur with impairments in higher spatial processing such as neglect (Semmes Reference Semmes1965; Vallar Reference Vallar1997). This indeed may be the consequence of increased higher-order spatial processing that may influence visual, as well as somatosensory, spatial representations (Saetti et al. Reference Saetti, de Renzi and Comper1999; Vallar Reference Vallar1997).

A third sensory modality for which separate processing streams have been proposed concerns the auditory system. These suggestions are based on recent neurophysiological data discerning separable cortical areas for the processing of auditory input. These areas involve a core, a belt, and parabelt regions arranged in a concentric manner in the superior temporal gyrus (Kaas et al. Reference Kaas, Hackett and Tramo1999). An important model of the functional architecture of auditory processing was proposed by Rauschecker and Tian (Reference Rauschecker and Tian2000), who argued that a ventral pathway running from the auditory core area in a forward direction is involved in analysing the stimulus characteristics. This route is important for object identification (e.g., a voice) and hence called the “what” pathway. A second, more posterior dorsal, pathway is thought to be sensitive to spatial location (the “where” path). This distinction is further supported by patient studies (Clarke et al. Reference Clarke, Bellmann, Meuli, Assal and Steck2000) and neuroimaging experiments (e.g., see Romanski et al. Reference Romanski, Tian, Fritz, Mishkin, Goldman-Rakic and Rauschecker1999). This model has been challenged by Zatorre and colleagues (e.g., Belin & Zatorre Reference Belin and Zatorre2000), who agree with the conceptualisation of the ventral “what” pathway. However, they object to the idea that the posterior dorsal route is mainly involved in spatial localisation, because, among other reasons, there is little evidence for spatial maps in the auditory system. In contrast, they claim that the dorsal pathway is primarily involved in perceiving the “evolution in time of the signal emitted by one or several auditory objects” (Belin & Zatorre Reference Belin and Zatorre2000, p. 965). Hence, the processing characteristic is likened to that of the visual area V5 or MT, processing time-related change in the signal. According to Zatorre and colleagues, the dorsal pathway processes the verbal message, whereas the ventral route is responsible for the recognition of the voice. In addition, they propose a lateralisation of function with the left hemisphere system being better tuned for speech perception (time sensitivity) and the right hemisphere system for music perception (pitch sensitivity; Zatorre et al. Reference Zatorre, Belin and Penhune2002).

Overall, there are clear similarities between the organisation of the auditory and the somatosensory systems. These systems share the increasing complexity of the information processing while moving away from the primary cortical areas (APC and the auditory core [Heschl's gyrus]). There is a reasonable agreement in the literature that two distinct pathways can be discerned in the auditory system. The equivalent of the “what” pathway is comparable in that it is involved in the recognition of external objects (as in vision), but there is no active exploration of the objects as is the case in tactile object recognition. The status of the “where” pathway is controversial, with different competing views. So far, there is little evidence for spatially organised cortical maps in the auditory system and a link with action-related processes.

Taken together, there appears to be substantial support for the idea that the different cortical sensory systems share overall organisational principles (Belin & Zatorre Reference Belin and Zatorre2000; Rauschecker Reference Rauschecker1998). In addition, there are clear indications for a common organisational principle in the development of the morphology, architecture, and connections in the different modalities (e.g., see Pandya & Yeterian Reference Pandya, Yeterian, Scheibel and Wechsler1990). These include specialisation of function, subserved by separate processing routes. This distinction is, however, relative, because there are also clear differences resulting from the input (nature of the sensory signal) and the output characteristics (perception, action) of the particular modality.

6. Crossmodal interactions

Traditionally, the study of sensory systems has focused on the processes and structure within a single modality. More recent studies have demonstrated that the different senses work closely together and strongly influence one another. Neuroanatomy supports the notion of integration and mutual influence. An example at a relative basic level concerns the structure of the superior colliculus with close proximity and common organisation among visual, auditory, and somatosensory processing (May Reference May2005).

Crossmodal processes have been found to affect a variety of tasks. Early perceptual processing in one modality may be modulated by input from another modality presented shortly prior to or simultaneously with the first stimulus. A dramatic example of this phenomenon is the observation that a single light flash is perceived as two flashes when the subject simultaneously hears two short auditory stimuli. The effect of the auditory stimulus alters processing in the primary visual cortex (Watkins et al. Reference Watkins, Shams, Tanaka, Haynes and Rees2006). There is now a substantial literature on crossmodal links between auditory, tactile, and visual stimuli in spatial attention (Maravita et al. Reference Maravita, Spence and Driver2003; Schmitt et al. Reference Schmitt, Postma and de Haan2001). There are indications, however, that the links between vision and audition are perhaps stronger than between these modalities and touch (Eimer et al. Reference Eimer, van Velzen and Driver2002). Furthermore, cross-modality presentation may influence the “experience of ownership.” Studies have shown that normal subjects experience a rubber hand as their own when it is stroked in a synchronous, but irregular, manner with their own unseen hand (Botvinick & Cohen Reference Botvinick and Cohen1998; Tsakiris & Haggard Reference Tsakiris and Haggard2005). Functional imaging studies suggest that the ventral premotor cortex, the intraparietal sulcus, and the lateral cerebellum are involved in this illusion (Ehrsson et al. Reference Ehrsson, Spence and Passingham2004; Reference Ehrsson, Holmes and Passingham2005a).

In another study, subjects experienced the hand seen on a screen as their own when their hand and the one on the screen are touched simultaneously. Using magnetoencephalography (MEG), it was demonstrated that the activity in the APC was modulated depending on whether the subject experienced ownership of the hand on the monitor (Schaefer et al. Reference Schaefer, Flor, Heinze and Rotte2006a).

Crossmodal interactions have also been observed for somatosensory tasks in which non-informative vision was provided. Non-informative vision has been found to influence performance on a variety of perceptual tasks, ranging from spatial acuity (Kennett et al. Reference Kennett, Taylor-Clarke and Haggard2001) to size constancy (Taylor-Clarke et al. Reference Taylor-Clarke, Jacobsen and Haggard2004) and parallel setting of bars (Newport et al. Reference Newport, Rabb and Jackson2002; Zuidhoek et al. Reference Zuidhoek, Visser, Bredero and Postma2004). The visual information provided varied, from a view of the stimulated body part (Kennett et al. Reference Kennett, Taylor-Clarke and Haggard2001; Taylor-Clarke et al. Reference Taylor-Clarke, Jacobsen and Haggard2004), to the environment excluding the stimuli (Newport et al. Reference Newport, Rabb and Jackson2002; Zuidhoek et al. Reference Zuidhoek, Visser, Bredero and Postma2004). Fewer studies have investigated crossmodal interactions with respect to sensorimotor action. Newport et al. (Reference Newport, Hindle and Jackson2001) showed that impaired proprioceptive target information in a patient could be ameliorated during a pointing movement through non-informative vision of the surrounding environment.

The question relevant for this review is whether a certain specificity in the visual-tactile interaction can be observed. That is, are visual influences on somatosensory processing for action different from visual-tactile interactions during perceptual recognition? At a behavioural level, this topic has received little attention, even though neurophysiological and functional imaging studies suggest that these crossmodal interactions involve different neural processes. For example, a functional imaging study by Prather et al. (Reference Prather, Votaw and Sathian2004) showed that mental rotation of tactile forms activates the visual dorsal stream, whereas tactile form discrimination is associated with ventral stream activation. Furthermore, several fMRI studies showed activation of the lateral occipital complex (LOC, part of the visual ventral stream) during visual and tactile object recognition (Amedi et al. Reference Amedi, Malach, Hendler, Peled and Zohary2001; James et al. Reference James, Humphrey, Gati, Servos, Menon and Goodale2002; Reed et al. Reference Reed, Shoham and Halgren2004). On the other hand, neurophysiological studies and functional imaging studies show multimodal activation in the PPC that is related to motor action (Clower et al. Reference Clower, Hoffman, Votaw, Faber, Woods and Alexander1996; Kalaska et al. Reference Kalaska, Scott, Cisek and Sergio1997; Kertzman et al. Reference Kertzman, Schwarz, Zeffiro and Hallett1997; Savaki et al. Reference Savaki, Raos and Dalezios1997). Indeed, it could be argued that the two processing streams in both modalities project to the same higher-order cortical areas that are involved in multimodal sensory integration for the guidance of action (PPC) or perception (LOC).

With respect to the PPC in action, different areas within the superior parietal cortex and the intraparietal sulcus appear to be related to different visuomotor channels, including grasping, reaching, saccade, and pursuit eye movements (Hyvärinen & Poranen Reference Hyvärinen and Poranen1974; Milner & Dijkerman Reference Milner, Dijkerman and Milner1998; Mountcastle et al. Reference Mountcastle, Lynch, Georgopoulos, Sakata and Acuna1975). Some of these areas contain neurones that have bimodal response properties and are active during reaching (Colby Reference Colby1998). Multimodal processing in the PPC also appears to be related to distinct reference frames. For example, the bimodal responsive neurones in area VIP appear to code stimuli particularly in a head-centred reference frame (Duhamel et al. Reference Duhamel, Colby and Goldberg1998), while bimodal activity in area MIP is related to arm-centred spatial representations (Colby Reference Colby1998; Duhamel et al. Reference Duhamel, Colby and Goldberg1998; Graziano et al. Reference Graziano, Cooke and Taylor2000). In humans, the PPC appears to be active when a conflict is created between visual and proprioceptive signals during a reaching movement (Clower et al. Reference Clower, Hoffman, Votaw, Faber, Woods and Alexander1996). Overall, these findings suggest the PPC to be involved in multimodal coding of body-related and arm-related configurations used for the guidance of action.

With respect to perceptual recognition of external targets, activation of area LOC during visual as well as somatosensory object recognition has been found. Several possible explanations have been put forward for this finding. Participants may use visual imagery when performing a tactile object recognition task (Deibert et al. Reference Deibert, Kraut, Kremen and Hart1999). A second possibility is that LOC is a multimodal area related to higher-order perceptual representations of objects (Amedi et al. Reference Amedi, Malach, Hendler, Peled and Zohary2001; James et al. Reference James, Humphrey, Gati, Servos, Menon and Goodale2002). Amedi et al. (Reference Amedi, Malach, Hendler, Peled and Zohary2001) observed that activation in LOC during visual imagery was less than during either tactile or visual object recognition Furthermore, Pietrini et al. (Reference Pietrini, Furey, Ricciardi, Gobbini, Wu, Cohen, Guazzelli and Haxby2004) observed similar activation during tactile recognition in congenitally blind subjects, which suggests that visual imagery is less likely to be the principal cause for the involvement of LOC in tactile object perception. Together, these findings suggest that LOC is a multimodal area involved in perceptual representation of object-form features.