We used transcranial magnetic stimulation (TMS) to investigate how the visual context provided by viewing one's own body influences somatosensory processing. In the visual enhancement of touch (VET) effect, viewing the body enhances tactile acuity relative to viewing a nonbody object. Single-pulse TMS was delivered over anterior intraparietal sulcus (aIPS), which is crucial for integrating visual and somatosensory information related to the body, during the interval between a brief glimpse of the arm, or an object, and tactile stimulation. TMS to aIPS just after visual stimulation abolished VET, while TMS at the time of touch itself did not. Disrupting nearby areas just anterior or posterior to aIPS left the VET effect intact. Viewing the arm may activate multisensory areas in aIPS, which may then rapidly modulate somatosensory circuits. We suggest that this enhancement of touch by vision involves feedback signals from aIPS to unimodal somatosensory cortex. Our study provides causal evidence of a specific multisensory mechanism that modulates tactile processing in the human brain.
Cortical sensory pathways show several kinds of multisensory interaction. Each anatomical pathway from the periphery is organized to bring a single modality of sensory information to the cortex, notably via thalamic relay nuclei, which are effectively unimodal. However, even the “primary” sensory cortices in each modality can be activated by other modalities. Thus, Zhou and Fuster (2000) reported single units in primary somatosensory cortex, which responded to “visual” stimuli, while the human visual cortex was activated by purely tactile tasks (Sathian and Zangaladze 2002). In addition, activity in one modality can “influence” activation and function of early cortical areas in a second. For example, expecting an auditory stimulus can modulate activation of visual cortex (Bueti and Macaluso 2010).
The neuronal basis for these cross-modal effects remains unclear. Anatomical studies suggest that thalamocortical input is segregated according to modality, suggesting that cross-modal activations of early cortical areas may instead reflect a top-down influence from higher cortical areas. Studies of intermediate-level association areas, such as those surrounding the intraparietal sulcus, show a strong multisensory convergence, with many bimodal or trimodal neurons that respond to input in more than one sensory modality (Avillac et al. 2007). One possible mechanism of cross-modal influence in early cortical areas would be recurrent feedback from these multisensory areas into “unimodal” cortex (Driver and Noesselt 2008). To investigate contributions of early somatosensory cortex and of the intraparietal sulcus areas, we have focused on a specific form of multisensory interaction that appears to involve a top-down contextual modulation of one modality by another, rather than a simple feed forward integration of different modalities of information (Ernst and Banks 2002).
In the visual enhancement of touch (VET) effect, vision of the body boosts tactile processing (e.g. Kennett et al. 2001; Taylor-Clarke et al. 2002, 2004; Press et al. 2004; Serino et al. 2007), relative to viewing a neutral object. Evidence from ERPs (Taylor-Clarke et al. 2002; Cardini et al. 2011), TMS studies (Fiorio and Haggard 2005), and somatotopic specificity (Serino et al. 2007) suggest that VET involves visual modulation of early somatosensory processing. In a typical VET experiment, tactile acuity is measured by asking participants to judge the orientation of tactile gratings. In one condition, participants view their hand, but care is taken to ensure that they do not see the orientation of the grating. In a control condition, participants view a nonbody object appearing at the same location, along with a visual event corresponding to the grating approaching the hand. Thus, vision of the body provides no information about tactile orientation nor about the time of tactile stimulation, yet tactile performance is significantly better in the view-body condition than in the view-object condition. Because vision provides no information relevant to touch, multisensory interactions based on feed forward integration of vision and touch (Ernst et al. 2000) cannot readily explain the VET effect. In contrast, VET could potentially be explained by multisensory feedback mechanisms. Visual signals from body-specific occipital areas might be relayed to multisensory parietal areas, from where recurrent feedback connections could influence early somatosensory processing. However, direct evidence for parietal multisensory feedback was lacking. Here, we investigate this question using single-pulse transcranial magnetic stimulation (TMS).
Materials and Methods
Subjects were naïve about the aim of the studies and gave informed written consent for participation in the experiments, which were approved by the Local Ethical Committee.
During the experimental sessions, subjects sat in a dark room to the left side of a semisilvered mirror. Subjects rested their right hands in a fixed position on the right side of the mirror. A red LED was taped to their right middle fingers and was just bright enough to permit fixation throughout, without allowing vision of the hand. In the view-hand condition, subjects viewed their right hand through the mirror. In the view-object condition, they viewed a wooden block, similar in size and position to their hand, via the mirror reflection. The object was located on the left side of the mirror, behind black cardboard, which prevented direct sight of it. Computer-controlled lights on either side of the mirror allowed either the hand or the object to be viewed on each trial.
At the beginning of each trial, subjects viewed either their hand or the object for 100 ms, followed by 600 ms darkness. TMS was delivered during this interval, at 0, 300, or 600 ms after the visual offset (Fig. 1). Tactile gratings were delivered to the right middle finger for 1000 ms, beginning 600 ms after visual offset. The dark interval separated visual from tactile stimulation and ensured that vision was always noninformative about the tactile stimulus. Participants reported grating orientation by pressing 1 of 2 keys with their left hand. Participants were instructed to emphasize accuracy rather than speed and were given a 3000-ms response interval.
The tactile stimuli were square-wave gratings consisting of alternating grooves and ridges (Van Boven and Johnson 1994). The gratings could be oriented either across or along the finger and were automatically delivered using a robot. Orientation was varied randomly from trial to trial. The ridge width of the grating used in the experiment was determined for each subject during a selection procedure to be just above their individual discrimination threshold (0.4, 0.6, 0.8, 1, or 1.2 mm). Starting with the largest ridge width, the orientation of the grating was randomly chosen, and the number of correct answers was averaged over 20 trials. A smaller ridge width was chosen when the responses were above 65% correct, because the benefit of viewing the body for touch may occur only close to perceptual limits (Press et al. 2004). Once a grating giving performance below 65% was found, it was used throughout the experiment. The mean ridge width across subjects was 0.8 mm ± 0.3 mm (N = 56).
Ten subjects participated in the first experiment (right-handed, 4 males; age: 18–5 years). The experiment consisted of 200 trials, subdivided into 4 blocks. This experiment investigated whether the VET effect could be found with randomized presentation of brief visual glimpses, and with computer-controlled tactile stimulation.
Eighteen new subjects participated in the second experiment (all right-handed, 12 males; age: 20–32 years). The site of TMS was varied in a 2 (view: object, hand) × 4 (TMS site: sensorimotor hand area [SMHA], anterior intraparietal sulcus [aIPS], posterior intraparietal sulcus [pIPS], No-TMS) factorial design. Each condition consisted of 100 trials, subdivided into 4 blocks. The stimulation site was randomized across blocks. The No-TMS condition was designed to confirm whether orientation judgments were more accurate when viewing the hand than when viewing the object. This was confirmed (hand: 64 ± 6%; object: 57 ± 6%, respectively, P < 0.001), and the results were not further analyzed.
Localization of TMS Sites
Sensorimotor Hand Area
To localize SMHA, the cortical excitability threshold for each subject was determined by delivering single TMS pulses (Magstim 200, Magstim, Whitland, Dyfed, UK) using a figure-of-eight coil over scalp location C3 in the left hemisphere, tangential to the scalp with the handle pointing backwards at an angle of 45° from the midline. Starting at 30% of maximum stimulator output, TMS intensity was increased in increments of 5% until motor twitches of the resting hand were detected by report of the subject and observation by the experimenter. To verify that TMS targeted the “sensorimotor hot spot,” surrounding cortex was systematically explored by moving the coil anterior-posteriorly and mediolaterally in 1 cm steps, to find the location producing the largest twitches. The mean TMS intensity required to elicit motor twitches was 45 ± 5% of maximum. Stimulation at the motor hotspot location is assumed to influence both primary motor and primary somatosensory cortices, and is a favored location for disrupting somatosensory processing (e.g. Kanda et al. 2003; Johnson et al. 2006). We reasoned that any motor twitches occurring during the experiment could slightly displace the finger, thus altering effective tactile stimulation. We therefore rotated the coil so that the handle pointed downward, and decreased intensity in 5% steps until muscle twitches were no longer evoked. Across subjects, the mean experimental intensity was 41 ± 4%.
Anterior Intraparietal Sulcus
Recent fMRI studies suggested that the human ventral intraparietal area (hVIP) is located in aIPS (Konen and Kastner 2008a). While several fMRI studies have reported activations in the IPS in response to retinotopic visual stimulation, few of these studies have considered multisensory interactions, and no established functional localizer exists for the specific visual–tactile interaction of interest for our study (but see Sereno and Huang 2006). Therefore, in the present study, presumed hVIP was identified by its anatomical location relative to SMHA. We determined the cortical distance between hVIP and SMHA in a previous dataset (Konen et al. 2009). Those results showed that hVIP was located 0.8 cm ± 0.7 cm medial from SMHA, 0.9 cm ± 0.8 cm superior, and 2.5 cm ± 1.4 cm posterior. Given the small and variable medial and superior deviations, we chose to localize hVIP in aIPS by simply moving the TMS coil 2.5 cm posteriorly from SMHA. The level of stimulation over aIPS was 60% throughout the experiment.
Posterior Intraparietal Sulcus
We moved the coil 5 cm posteriorly from SMHA to have a suitable stimulation site within pIPS. This location was chosen on the grounds of being the same distance from the aIPS site as was SMHA. It thus provided a suitable control for any spatial spread of TMS unrelated to underlying functional neural connections. The pIPS area is involved in the encoding of saccades and objects (Schluppeck et al. 2006; Levy et al. 2007; Konen and Kastner 2008a, 2008b), but was used here as a spatially adjacent control site, rather than being targeted as a functional area.
The spatial resolution of TMS stimulation (∼1 cm2, Maccabee et al. 1990) coupled with the 2.5-cm distance between areas indicated that each TMS location should influence distinct regions of parietal cortex. The level of stimulation over aIPS was 60% throughout the experiment.
Experiments 3 and 4
Eighteen new subjects participated in the third and fourth experiments (right-handed, 10 males; age: 18–29 years).
Experiment 3 investigated whether different effects of SMHA and aIPS stimulation could relate to stimulation intensity. In experiment 2, we used 60% intensity over aIPS, but lower, subject-specific intensities over SMHA. This arrangement prevented muscle twitches that would occur at higher TMS intensities, but involved a confound between TMS site and TMS intensity. Therefore, the conditions of experiment 2 were replicated in a 2 (view: object, hand) × 2 (TMS site: SMHA, aIPS) factorial design with 100 trials each, divided into 4 blocks. The coil was tangential to the scalp with the handle pointing backwards at an angle of 45° from the midline. We chose an intensity, which 1) generally avoided motor twitches, 2) was comparable to the intensity of experiment 2, despite the different motor thresholds of the 2 groups of participants. Thus, TMS intensity was 45% throughout experiment 3. Over SMHA, this level of stimulation was below the motor threshold of the majority of subjects (N = 15) and evoked muscle twitches in only 4 subjects and 11 ± 5% of trials across subjects, as detected by report of the subject and observation by the experimenter.
Experiment 4 aimed to confirm our localizations of SMHA and aIPS. Specifically, we wanted to confirm that SMHA stimulation could influence basic tactile perception (independently of visual context), while aIPS stimulation could not. The first result would suggest that SMHA, which we identified by “motor” responses to TMS, would also involve stimulation of the primary somatosensory cortex. The second result would rule out the possibility that effects of aIPS stimulation in the main experiment were simply due to spread of the magnetic field to primary somatosensory cortex. The TMS pulse was delivered 20 ms after tactile stimulation, at which time it should disrupt tactile perception (Cohen et al. 1991; Fiorio and Haggard 2005). A 2 (view: object, hand) × 2 (TMS site: SMHA, aIPS) factorial design was used, with 100 trials each, subdivided into 4 blocks. The level of stimulation was 60% throughout experiment 4.
Ten new subjects participated in the fifth experiment (right-handed, 5 males; age: 19–23 years). TMS was applied over aIPS at 3 different time points between visual and tactile stimulation (0, 300, and 600 ms after offset of visual stimulation). Note that tactile stimulation began 600 ms after visual stimulus offset. A 2 (view: object, hand) × 3 (TMS timing) factorial design consisted of 100 trials each, divided into 4 blocks. The timing of stimulation was randomized across trials. The level of stimulation was 60% throughout the experiment.
Experiment 1 (Fig. 1) revealed that orientation judgments were more accurate when viewing the hand than when viewing the object (62 ± 6% vs. 54 ± 6%, respectively, P < 0.01). Thus, even brief, noninformative glimpses of the hand enhanced tactile performance relative to viewing a neutral object. This replicates and extends previous results, and strongly suggests that VET depends on interactions between body-specific multisensory areas, and the early somatosensory areas that underpin tactile acuity.
This rapid effect of viewing the arm on touch suggested a possible role of aIPS and more specifically, hVIP. This site of multisensory convergence serves as the hub of multimodal integration and peripersonal space representation in macaque monkeys (e.g. Colby et al. 1993; Duhamel et al. 1998; Bremmer et al. 2002; Schlack et al. 2002; Graziano and Cooke 2006).
In experiment 2, TMS was delivered during a brief dark interval after viewing the hand or the object and before touch. A no-TMS condition replicated the basic VET effect found in experiment 1, with orientation judgments being more accurate when viewing the hand (64 ± 6%), than when viewing the object (57 ± 6%, P < 0.001). Our core design, however, involved comparing VET effects following stimulation of 3 different sites in parietal cortex: SMHA, aIPS, and pIPS (see Materials and Methods section for localization of TMS sites). Therefore, the main analyses compared the different TMS sites, and the no-TMS condition was not analyzed further.
Two-way repeated measures ANOVA showed a significant main effect of view (P < 0.001) and TMS site (P < 0.01). The interaction between view and TMS site was also significant (P < 0.005). Figure 2 shows decreased accuracy in the view-hand condition for TMS over aIPS when compared with TMS over anterior and posterior sites.
We explored this interaction using simple effects testing to compare accuracy between hand and object conditions for each stimulation site. Significant VET effects remained after SMHA and pIPS stimulation (P < 0.005 and P < 0.001), but were abolished by aIPS stimulation (P > 0.05). AIPS TMS largely abolished VET, reducing it to a nonsignificant 1% increase in accuracy (P > 0.05). A significant (8%, P < 0.005) VET effect remained after SMHA stimulation and also after pIPS (9%, P < 0.001) stimulation. Thus, the 2-way interaction was due to the reduction of VET effects following aIPS stimulation.
Thus, TMS over aIPS interfered with VET. This interference was not due to nonspecific TMS effects, or disruption of adjacent regions, for 2 reasons. First, disruption of SMHA and pIPS had no effects on tactile perception, even though these areas were just anterior and posterior to aIPS. This excludes the possibility that our effects arose not from disrupting aIPS itself, but from disruption of a nearby region. Second, the effects were specific to the view-hand condition (Fig. 2). These findings indicate that aIPS plays a causal role in the modulation of somatosensory acuity by vision of the arm.
In this experiment, SMHA stimulation used lower intensities than aIPS stimulation, to avoid possible motor interference with touch. Experiment 3 showed that this change in TMS intensity could not explain our results. When both SMHA and aIPS were stimulated at equal, lower intensities comparable to those for SMHA in experiment 2, similar results were obtained (Fig. 3). Specifically, ANOVA showed a significant main effect of view (P < 0.005) and TMS site (P < 0.005). The interaction between view and TMS site was significant (P < 0.05). Simple effects testing showed significant effects of TMS site (P < 0.05) on VET, because aIPS TMS largely abolished VET, reducing it to 2%, relative to SMHA TMS (8%). VET effects differed between aIPS and SMHA stimulation (P < 0.05). The VET effect remained present at SMHA (P < 0.001) but not at aIPS (P > 0.05). This replicates the results of experiment 2.
In experiment 4, we sought behavioral evidence that SMHA stimulation did indeed interfere with primary tactile processing, while aIPS stimulation did not interfere with primary tactile processing. TMS over SMHA and aIPS was applied immediately after tactile stimulation. Two-way repeated measures ANOVA showed significant main effects of view (P < 0.01) and TMS site (P < 0.001) as well as a significant interaction between these factors (P < 0.05). The main effect of TMS site suggests that SMHA stimulation disrupted primary tactile perception, while aIPS stimulation did not (Fig. 3). SMHA stimulation immediately after touch abolished VET, reducing it to just 0%, relative to aIPS (8%) stimulation. Follow-up testing showed that VET effects differed between aIPS and SMHA stimulation (P < 0.05). Similarly, comparing accuracy between view-hand and view-object conditions at each stimulation site showed significant VET effects at aIPS (P < 0.005), but not at SMHA (P > 0.05).
Experiment 4 confirmed that SMHA TMS impaired primary tactile perception, suggesting disruption of primary somatosensory cortex. Conversely, statistical comparison between aIPS TMS in this control experiment and experiment 1 confirmed a main effect of view (P < 0.05) but no significant differences between the TMS/No-TMS conditions (P > 0.05) and no interaction (P > 0.05). Thus, aIPS TMS had no effect on primary tactile perception, suggesting that forward spread of disruption to primary somatosensory cortex did not occur. Moreover, the findings indicated that SMHA and aIPS play different roles in VET: SMHA appeared to be solely involved in perceptual processing after tactile stimulation, while aIPS seemed to be crucial before tactile stimulation. SMHA TMS after tactile stimulation effectively abolished VET and reduced overall tactile acuity, while aIPS TMS after touch left the VET effect intact. Thus, the contribution of aIPS to the VET effect is linked to the onset of viewing the arm, and involves modulating somatosensory circuits that process subsequent tactile stimuli.
The time course of aIPS was further investigated in experiment 5. Response latencies to visual and/or somatosensory stimulation in monkey VIP range from 40 to 200 ms (Avillac et al. 2007). Accordingly, TMS was applied synchronously with the offset of visual stimulation (0 ms), 300 or 600 ms after visual stimulation.
The results showed a significant main effect of view, confirming VET (P < 0.005), but no main effect of TMS timing (P > 0.05). The interaction between view and TMS timing was significant (P < 0.001). Figure 4 suggests that accuracy in the view-hand condition improved with increasing delays of TMS subsequent to visual stimulation. Comparing VET across the different times of stimulation showed that the effect was significantly greater at 600 ms than at 0 ms (P < 0.001) or at 300 ms (P < 0.05) after visual stimulation. Conversely, VET effects 0 and 300 ms after visual stimulation did not differ (P > 0.05).
To summarize, our findings showed that transiently disrupting activity in aIPS immediately after a brief, noninformative glimpse of the hand abolished the enhancement of tactile discrimination when viewing the arm. TMS over aIPS immediately after viewing the arm interfered with VET. Furthermore, the results showed that the critical period for aIPS to modulate somatosensory circuits extended up until 300 ms after visual stimulation, but had finished by 600 ms. This time window is compatible with the latency range of multisensory neurons with visual responses in monkey VIP (Avillac et al. 2007).
This temporal specificity is consistent with the hypothesis that the VET effect involves a visually triggered signal, or wave of neural processing, passing from early visual areas, to multisensory parietal areas, and then on to early somatosensory cortex. That is, it may represent a feedback signal reaching somatosensory cortex from the multisensory parietal cortex. An alternative hypothesis is that aIPS is “directly” involved in somatosensory processing for tactile acuity, in parallel with primary somatosensory cortex. VET might therefore take place within aIPS itself, as well as in primary somatosensory cortex. We assume that the specific area targeted by our aIPS TMS was hVIP. Indeed, the view that hVIP might directly participate in modulating tactile acuity receives some support from the recent report of somatotopic tactile representation in hVIP (Sereno and Huang 2006). However, this hypothesis cannot explain why hVIP's contribution depends on the body-specific “content” of visual input, not merely on the presence of a visual signal (Ro et al. 2004). Similarly, the alternative hypothesis cannot explain why hVIPs contribution to visual modulation decreases, rather than increases, with the time elapsed after viewing the arm, and before tactile stimulation begins. In contrast, this latter point is consistent with the view that hVIP provides a multisensory modulation of the early somatosensory areas that underlie tactile acuity. Therefore, we conclude that aIPS is causally involved in modulating touch, but its contribution is linked to the influence of viewing the arm on touch, rather than to tactile processing itself. In particular, when vision of the arm and tactile stimulation were separated by a short dark interval, aIPS TMS around the time of visual stimulation effectively abolished VET, while later TMS around the time of touch did not. AIPS is therefore not itself the site of the visual–tactile interaction in the VET effect, but the source of a visual signal that contributes to a visual–tactile interaction occurring elsewhere (presumably in primary somatosensory cortex, Taylor-Clarke et al. 2002).
Thus, aIPS TMS at either 0 or 300 ms after viewing the hand reduced the TMS effect, while TMS 600 ms after viewing the hand did not. The disruptive effect of TMS at 0 ms is fully consistent with the 40–200 ms latencies of multisensory signals in IPS of nonhuman primates, given the assumption that a single TMS pulse disrupts the normal processing of the underlying cortical area for around 100–200 ms (Day et al. 1989). However, the disruptive effect that we observed at 300 ms may appear less consistent with the monkey data: the multisensory processing in monkey aIPS would be finished by this time, and the modulatory signal would by then have passed through aIPS en route to destination areas in somatosensory cortex. However, peak neural latencies in visual, auditory, and somatosensory systems of monkeys are around 3/5 of those calculated from evoked potentials in humans (Schroeder et al. 2004). Thus by extrapolation, disruption of VET by TMS 300 ms after vision in humans would fall within the range expected from the monkey literature. Finally, the failure to disrupt VT with TMS 600 ms after visual stimulation is consistent with the view that visual information transits through aIPS in a feed forward sweep, and that the contribution of aIPS to VET has already occurred by 600 ms. Thus, our results show a clear time-specificity of the aIPS contribution to the VET effect. However, future studies with a closer sampling of TMS timings may be required to identify exactly when aIPS signals exert their specific modulatory influence on tactile processing.
In contrast, previous studies of primary somatosensory cortex TMS showed quite different temporal specificity. TMS just before or after touch was reported to mask tactile perception (Seyal et al. 1997). Similarly, VET was abolished by primary somatosensory stimulation when applied immediately before touch, and 2500 ms after viewing the hand (Fiorio and Haggard 2005). Thus, TMS disruptions of VET during the interval between vision and touch decrease with elapsed time for aIPS stimulation, but appear to increase with elapsed time for SMHA stimulation. Previous TMS studies disrupting primary somatosensory cortex in the interval between vision and touch already suggested that VET might involve visual presetting of primary somatosensory cortex. However, the source of the presetting signal remained unclear. Our results suggest that disrupting aIPS immediately after viewing the hand may interfere with the presetting signal on its path from visual to somatosensory areas. Multisensory feedback-based modulation of unimodal cortex thus depends on processing in aIPS. Indeed, connectivity studies in macaques confirmed strong connections between aIPS and primary somatosensory cortex (Lewis and Van Essen 2000).
TMS can have remote effects as well as local effects. In principle, our aIPS results could arise because TMS over aIPS might disrupt an additional area involved in VET. Note that this argument is not simply about spread of the induced activity to adjacent brain areas, but about transmission of TMS-evoked neural activity along functional pathways to brain areas that may be quite distant. Experiments 3 and 4 rule out the most obvious possibility, namely, that TMS over aIPS simply has remote effects because of conduction to primary somatosensory cortex. We showed that aIPS TMS and SMHA TMS have different time courses. Any explanation based on remote conduction would predict that aIPS TMS should show the same effects as SMHA TMS, though perhaps to a lesser extent. However, we cannot exclude the possibility that other areas functionally connected to aIPS are also involved in these effects.
Our pattern of results shows that aIPS plays a causal role in the VET effect, although it is not involved in primary tactile discrimination processing per se. The results are consistent with aIPS providing a descending feedback signal to primary somatosensory cortex. This feedback might modulate the gain of somatosensory processing circuits underlying tactile acuity, thus enhancing touch (Cardini et al. 2011). Our data therefore strongly support the “Critical Role of Feedback Circuitry” in multisensory interactions, recently proposed by Driver and Noesselt (2008). According to this account, multisensory effects depend not only on feed forward convergence of unimodal signals in multimodal cortex, but also on modulation of unimodal cortex via feedback projections from multimodal cortex. Indeed, this model explicitly predicts that disrupting multisensory areas should alter unisensory perceptual performance. Our study provides direct support for this model.
The visual input in our VET paradigm is uninformative about touch and is delivered prior to touch. In contrast, feed forward multisensory integration is based on combining signals carrying information about a common source (Ernst and Banks 2002; Gori et al. 2008), and is sensitive to temporal discrepancy (Meredith and Stein 1983; Avillac et al. 2007). We are therefore confident that our aIPS intervention disrupted a feedback rather than a feed forward pathway. Interestingly, TMS over parietal cortex was previously reported to abolish the interaction between synchronized visual and tactile information about location in a feed forward integration paradigm (Pasalar et al. 2010). However, no previous TMS study has, to our knowledge, investigated the role of parietal cortex in feedback, as opposed to feed forward multisensory interactions.
Previous connectivity studies based on fMRI data have provided indirect evidence for feedback projections from multimodal to unimodal cortex (Macaluso et al. 2000, 2002; Macaluso and Driver 2005). Here, we provide direct evidence for these feedback signals by selectively disrupting their putative source using spatially and temporally selective TMS within a precisely timed tactile perceptual task. Recent studies have suggested an important role of unimodal feedback connections from higher visual areas to early visual cortex, in both monkeys and humans (Moore and Armstrong 2003; Ruff et al. 2006). Thus, our data contribute to a growing body of work showing that sensory processing is not solely characterized by serial processing in a hierarchical fashion, but may also involve parallel or feedback projections. We extend this principle to multisensory interactions for the first time.
Visual Body-Related Processing in aIPS
Our interpretation assumes that visual information specifically related to the arm is processed in aIPS, where hVIP is located. In monkeys, area VIP is adjacent to the anterior intraparietal (AIP) and lateral intraparietal area (LIP). All 3 areas receive somatosensory as well as visual signals (Grefkes and Fink 2005). However, areas AIP and LIP are mainly involved in the transformation from sensory input into motor output enabling sensory-guided arm and eye movements, respectively, while area VIP may be more specifically involved in representing the body (Graziano and Cooke 2006). Thus, our results would suggest that we targeted hVIP in aIPS rather than area AIP or LIP in aIPS. Responses to viewing the arm itself have not been studied to our knowledge, though monkey premotor and area 5 neurons were reported to respond when viewing the arm (Graziano et al. 2000).
As VIP is part of the dorsal pathway, the above interpretation may appear contrary to a strong version of the traditional distinction between a ventral visual pathway, specialized for object vision, and a dorsal pathway, specialized for spatial vision (Ungerleider and Mishkin 1982; Goodale and Milner 1992). Recent studies in monkeys and humans, however, indicated object-selective responses in the dorsal pathway (Sereno and Maunsell 1998; Konen and Kastner 2008b). Our results suggest that hVIP may be similarly selective for body parts as a special category of objects. One putative candidate for projecting body-related information to hVIP is the extrastriate body area in the ventral pathway, which is selective for bodies and body parts (Downing et al. 2006, 2001). On this interpretation, hVIP would be a key node on a visual-multisensory-somatosensory modulatory circuit, providing visual modulation of touch, but only when viewing the arm. This would explain why our TMS over aIPS had an effect on tactile discrimination only when viewing the arm—the signals in the modulatory circuit would only be present when the appropriate visual stimulus for this tuned circuit is present, and would therefore only be disrupted by TMS in that case.
Several pathways for this body-related crosstalk between the dorsal and ventral pathways are possible. Physiological studies in macaque monkeys have shown that even though the majority of visual inputs to area VIP originate from the dorsal pathway (Lewis and Van Essen 2000), direct connections to the ventral stream exist (Ungerleider et al. 2008). Alternatively, VIP may receive body-related visual information via area LIP, which is also strongly connected to the ventral stream (Lewis and Van Essen 2000). Future experiments should aim to further specify the visual pathway(s) that enable hVIP to facilitate somatosensory processing subsequent to viewing the body.
This research was supported by a LeverhulmeTrustMajorResearchFellowship to PH, and by EUFP7 project grant VERE.
Conflict of Interest: None declared.