The present study mapped lesion location associated with the RHI in acute stroke patients to examine the role of PMv in generating the illusory feeling of hand ownership and asomatognosia.
Brain regions associated with RHIF were located exclusively subcortically. In particular, PMv did not emerge from the lesion-mapping approach. To interpret this finding, it is important to realize that the lesion-mapping approach has inherent limitations in stroke patients: (1) The likelihood of alerting patients to possible stroke varies widely across affected brain regions and might be highest if motor, somatosensory, or visual deficits are present. This results in a bias toward patients with deficits in motor or sensory capacities and might lower the detection rate of strokes within secondary cortical areas. (2) The probability of lesions is not distributed equally across the brain, but depends on the vascular architecture and differential vulnerability of brain regions (Rorden and Karnath, 2004
). (3) Patients with aphasia and severe deficits of tactile perception and attention were excluded from participation in the study. Therefore brain regions associated with these deficits, such as Broca’s area (which is located in the vicinity of premotor cortex), Wernicke’s area, and primary somatosensory cortex, were unlikely to be associated with RHIF. Together, these considerations imply that the lesion method applied here may have low or absent sensitivity for certain brain regions. However, it may be highly specific as it demonstrates unequivocally that disruption of certain brain regions is causally associated with a behavioral phenotype (Rorden and Karnath, 2004
The fact that premotor cortex did not emerge from the lesion analysis does not exclude the possibility that processing activity in PMv plays a decisive role in RHI. However, detecting this role might rather depend on the analysis of projecting fibers. Tractography analysis of fiber projections through the lesion voxels revealed consistent connections with unilateral PMv from voxels associated with failure of the RHI. This finding supports the hypothesis that PMv plays a causal role in the RHI and that lesions to PMv or its connections should be associated with RHIF (Ehrsson et al., 2004
). Though RHI was associated with bilateral activation of premotor cortex in the fMRI study by Ehrsson et al. (2004)
, the absence of transcallosal projections to homologous PMv from RHIF-associated lesion voxels in the present study suggests that disruption of one PMv might suffice for failure of the illusion. Furthermore, since RHIF was found after lesions to either hemisphere, there is no hemispheric dominance in generating the RHI.
Somatognosia, the continuous awareness about one’s own body parts, seems to be based on a variety of distributed systems built of specifically interconnected brain areas rather than one specialized region (Berlucchi and Aglioti, 2010
). Analyses of the functional anatomy of asomatognosia have implicated brain regions including the parietal (So and Schaüble, 2004
), temporoparietal (Arzy et al., 2006a
), insular (Baier and Karnath, 2008
), and premotor (Arzy et al., 2006b
) cortices. Lesions associated with loss of awareness about one’s body parts should be expected to overlap with those disrupting feelings of body ownership in the RHI paradigm. Indeed, the connections of asomatognosia-associated lesion voxels were partially congruent with those found for RHIF. However, we found two important dissociations between RHIF and asomatognosia: Behaviorally, RHIF was present in similar percentages of stroke patients with and without contralesional asomatognosia; in particular, a number of stroke patients with impairment of the ability to perceive their real hand as belonging to them easily integrated the plastic hand as their own. Anatomically, as outlined above, the fiber tracts affected by ischemic lesions in RHIF involved connections to PMv, which were not implicated in the functional anatomical analysis of the asomatognosia group.
The fact that somatognosia, unlike RHI, may be robust toward lesions of fiber connections with PMv, has important implications for understanding the role of PMv in body ownership. Somatognosia and RHI differ behaviorally, among other features, by the time scale of the perceptual experience, which is extended in the former, but highly dynamic and short-lived in the latter. We consider it likely that this behavioral difference holds a clue to explaining the anatomical dissociation, regarding the role of PMv. In line with a neurocognitive model of body ownership during the RHI (Tsakiris, 2010
), we propose that there is a sequence of critical comparisons between current perception and preexisting models of the body. While visual and postural properties of the hand are matched in distinct other brain regions, PMv is active during the comparison between the vision of touch and the felt touch and the respective reference frames, i.e., body-related multisensory integration preceding the onset of illusory body-ownership feeling. We thus consider PMv necessary for the detection of concurrent multisensory events challenging body representation, and for the resolution of a potential conflict with current body representation by assimilation of the latter. Lesion to the network sustaining continual body ownership experience might not necessarily block induction of the RHI as long as connections with PMv are intact. Lesions to premotor areas, including PMv, have been shown to be associated with denial of hemiparesis following right-hemispheric stroke (Berti et al., 2005
). This finding may indicate that the supramodal role of PMv is not restricted to the sensory domain but also relates to integration of action plans with actions.
There are a number of limitations to this study. First, RHIF and asomatognosia were assessed only in a subgroup of the stroke patients. Second, the RHI was not assessed with objective measures, but only with questionnaires. Another limitation results from the MRI protocol: We used DWI to visualize ischemic lesions at an early stage. DWI is particularly sensitive for the detection of hyperacute infarcts (Chong et al., 1998
; Schaefer et al., 2002
) and is able to predict final infarct volume (Schaefer et al., 2002
), but areas that appear intact are not necessarily functioning normally, as a result of abnormal perfusion or deafferentation and diaschisis. This may result in an underestimation of the “functional size” of the ischemic lesions and consecutively in an underestimation of the seed areas used for tractographic analysis. Finally, the fact that we acquired high angular resolution diffusion data for tractography in a separate group of healthy control subjects and not in the stroke patients could be seen as a limitation. However, this approach was taken as the aim of the tractography was to determine the fiber pathways that would normally traverse through lesioned tissue. Interpretation of tractography in patients with brain lesions is challenging and would not have provided a straightforward test of the specific hypotheses considered here.
Altogether, our findings are consistent with a role of PMv in dynamic assimilation of body representation following multisensory challenge. Although feelings of ownership can be manipulated easily by experimental procedures or simply by tool use, they might not be updated continuously. One strong prediction arising from our conclusions is that recovery from asomatognosia depends on the functional intactness of (at least one) PMv.