Decoding illusory self-location from activity in the human hippocampus

Abstract

Decades of research have demonstrated a role for the hippocampus in spatial navigation and episodic and spatial memory. However, empirical evidence linking hippocampal activity to the perceptual experience of being physically located at a particular place in the environment is lacking. In this study, we used a multisensory out-of-body illusion to perceptually 'teleport' six healthy participants between two different locations in the scanner room during high-resolution functional magnetic resonance imaging (fMRI). The participants were fitted with MRI-compatible head-mounted displays that changed their first-person visual perspective to that of a pair of cameras placed in one of two corners of the scanner room. To elicit the illusion of being physically located in this position, we delivered synchronous visuo-tactile stimulation in the form of an object moving toward the cameras coupled with touches applied to the participant's chest. Asynchronous visuo-tactile stimulation did not induce the illusion and served as a control condition. We found that illusory self-location could be successfully decoded from patterns of activity in the hippocampus in all of the participants in the synchronous (P < 0.05) but not in the asynchronous condition (P > 0.05). At the group-level, the decoding accuracy was significantly higher in the synchronous than in the asynchronous condition (P = 0.012). These findings associate hippocampal activity with the perceived location of the bodily self in space, which suggests that the human hippocampus is involved not only in spatial navigation and memory but also in the construction of our sense of bodily self-location.

Keywords: body perception; multisensory integration; perceptual illusion; self-consciousness; self-location.

FIGURE 1

Scanner room environment and visual stimulus. (A) A schematic drawing of the scanner room environment. The key spatial landmarks and the illusory self-locations (Locations A and B) are indicated. The red stick figures represent the two illusory self-locations, whereas the blue stick figure indicates the veridical location of the participant’s body inside the bore of the scanner. (B) The timing and five representative frames from the 3D visual stimuli (only the image from the left eye is shown) for one experimental block. The first and the last frame indicate the viewpoint from Location A and Location B, respectively.

FIGURE 2

Decoding results. (A) Individual peak decoding accuracies for the synchronous (illusion; red symbols) and the asynchronous (control; yellow symbols) conditions. The symbols indicate each participant (female participants are represented by triangles). A solid symbol represents a significant value (P < 0.05, corrected) and an open symbol represents a non-significant (n.s.) value (P > 0.05, corrected). At the group-level, the peak decoding accuracy was significantly higher in the synchronous than in the asynchronous condition (*P = 0.012). (B) Maps illustrating the localization of voxels informative about self-location during the synchronous (red) and asynchronous (yellow) conditions (P < 0.001, uncorrected) for each participant (P1–P6). Note that there were significantly fewer (or no) such voxels in the asynchronous condition (mean number of voxels ± SD: 2.0 ± 2.9) than in the synchronous condition (mean number of voxels ± SD: 32.3 ± 15.4; P < 0.05, two-tailed Wilcoxon signed-rank test). At the group-level, there were no significant differences among the hippocampal subregions (right anterior, right posterior, left anterior, left posterior; χ²(3) = 1.316, P = 0.725, Freidman test) in the number of informative voxels. (C) Permutation testing. The histograms indicate the participantwise distributions of peak decoding accuracies under the null hypothesis that there was no information regarding LOCATION A and LOCATION B encoded in the hippocampus for the synchronous (red) and asynchronous (yellow) conditions (permutation testing with 10,000 iterations). The frequency of permuted (randomly labeled) decoding maps (Y axis) is presented as a function of the peak decoding accuracy (X axis). The true peak decoding accuracy is indicated with a black cross (corresponding to the symbols shown in A). The probability of obtaining the true value under the null hypothesis (i.e., the P-value) is displayed for the synchronous and asynchronous conditions for each participant.

FIGURE 3

Subjective and neurophysiological evidence for successful illusion induction. (A) The average subjective ratings of questionnaire statements S1–S3 for LOCATION A and LOCATION B during the SYNCHRONOUS and ASYNCHRONOUS conditions, respectively. The SYNCHRONOUS condition generated significantly higher ratings for the illusion statements (S1–S3), compared with the ASYNCHRONOUS condition. There was no significant difference in the illusion strength between LOCATION A and LOCATION B. The error bars represent the SEM, and the statistical results refer to the main effect of the visuo-tactile stimulation mode in a 2 × 2 analysis of variance (ANOVA). *P < 0.05, **P < 0.01. (B) Examining the main effect of the visuo-tactile stimulation mode using standard univariate analysis revealed increased activity in key multisensory regions in five out of six participants (P2–P6) at the statistical threshold of P < 0.001 (uncorrected). One representative brain slice for each subject is shown for the activations in the premotor (PMC, white circles) and intraparietal cortices (IPS, blue circles), which have previously been associated with limb (Ehrsson et al., 2004) and full-body ownership (Petkova et al., 2011; Guterstam et al., 2015b).