Connectivity and functional diversity of different temporo-occipital nodes for action perception

1 Connectivity and functional diversity of different temporo-occipital nodes for 1 action perception 2 Baichen Lia, Marta Poyo Solanasa, Giuseppe Marrazzoa, and Beatrice de Geldera* 3 a Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht 4 University, Maastricht, Limburg 6200 MD, The Netherlands. 5 * Correspondence: Beatrice de Gelder, Room 3.009, Oxfordlaan 55, 6229 EV Maastricht, The 6 Netherlands. 7 [email protected] 8 9 10 ABBREVIATED TITLE: Temporo-occipital nodes for action perception 11 12 Number of pages: 26 13 Number of figures: 5 14 Number of tables: 1 15 Number of words for abstract: 194 16 Number of words for introduction: 519 17 Number of words for discussion: 1215 18 19 Competing interests: The authors declare no competing interests. 20 21 Acknowledgments: 22 This work was supported by the European Research Council ERC Synergy grant (Grant 23 agreement 856495, Relevance), by the European Union’s Horizon 2020 research and innovation 24 programme (Grant agreement 101017884, GuestXR), the European Union’s Horizon Europe 25 research and innovation programme (Grant agreement 101070278, ReSilence), and by the Future 26 and Emerging Technologies (FET) Proactive Program H2020-EU.1.2.2 (Grant agreement 824160, 27 EnTimeMent). 28 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 2 Abstract 29 The temporo-occipital cortex (TOC) plays a key role in body and action perception, but current 30 understanding of its functions is still limited. TOC body regions are heterogeneous and their role 31 in action perception is poorly understood. This study adopted data-driven approaches to region 32 selectivity and investigated the connectivity of TOC nodes and the functional network sensitivity 33 for different whole body action videos. In two human 7T fMRI experiments using independent 34 component analysis, four adjacent body selective nodes were detected within the TOC network 35 with distinct connectivity profiles and functional roles. Action type independent connectivity was 36 observed for the posterior-ventral node to the visual cortex, the posterior-dorsal node to the 37 precuneus and the anterior nodes to the frontal cortex. Action specific connectivity modulations 38 were found in middle frontal gyrus for the aggressive condition with increased connectivity to 39 the anterior node and decreased connectivity to the posterior-dorsal node. But for the defensive 40 condition, node-nonspecific enhancement was found for the TOC-cingulate connectivity. By 41 addressing the issue of multiple nodes in the temporo-occipital network we show a functional 42 dissociation of different body selective centres related to the action type and a potential hierarchy 43 within the TOC body network. 44 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 3 1. Introduction 45 During social interactions, intentions, actions, and emotions are routinely read from nonverbal 46 communication signals provided by faces, body postures and whole-body movements as well47 documented in studies of humans (Argyle, 1976; de Gelder et al., 2010) and non-human primates 48 (Vogels, 2022). Despite its importance, the neural basis of whole-body perception is still much 49 less understood than that of faces (Deen et al., 2023). 50 Recent studies have revealed several brain areas in human and non-human primates. In humans, 51 the extrastriate body area (EBA) (Downing & Kanwisher, 2001) was the first one to be reported. 52 It is a region located in the extrastriate cortex that overlaps with other category-selective areas, 53 such as those dedicated to processing motion (Weiner & Grill-Spector, 2011), tools, and even 54 action related words (Lingnau & Downing, 2015). Subsequent studies have identified at least 55 three different body-selective clusters within the extrastriate cortex (Weiner & Grill-Spector, 56 2011). This anatomical diversity is complemented by functional diversity, as these clusters also 57 display varying patterns of connectivity (Zimmermann et al., 2018). 58 The actual contribution of these body selective areas is not well understood (de Gelder & Poyo 59 Solanas, 2021; de Gelder et al., 2010; Vogels, 2022). A number of studies found that these body 60 selective areas play a role in processing emotional expressions (de Gelder et al., 2004; Grèzes et 61 al., 2007; Peelen et al., 2007; Pichon et al., 2008), biological movement (Jastorff & Orban, 2009), 62 specific postural and kinematic features of the body (Marrazzo et al., 2023; Marrazzo et al., 2021; 63 Poyo Solanas et al., 2020), action recognition (Goldberg et al., 2014; Shmuelof & Zohary, 2005), 64 motor planning (Zimmermann et al., 2012), as well as social perception (Kret et al., 2011a; 65 Moreau et al., 2023). Taken together, these findings suggest that body selectivity may be 66 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 4 understood not simply as a matter of category selectivity but as resulting from the activity of a 67 sparsely distributed ensemble of brain areas (de Gelder & Poyo Solanas, 2021; Weiner & Grill68 Spector, 2011). Mapping the broader network’s activity may be an important step in uncovering 69 how these different body-specific nodes collectively contribute to the perception of whole-body 70 movements at the network level. Several network models have been proposed including the 71 Action Observation Network (AON) (Caspers et al., 2010), the Default Mode Network (DMN) 72 (Zhan et al., 2018) or a pathway for social perception (Haak & Beckmann, 2018). However, none 73 of these directly addresses the specific case of perceiving emotional expressions, intentions, and 74 actions of conspecifics routinely conveyed by body movements. 75 The goal of this study was to identify a dynamic whole-body network and to investigate how its 76 network activity and its connectivity with other brain areas supports specific actions such as 77 defensive or aggressive behaviour. Rather than following the traditional approach of contrast78 based selection of body regions, we approach the question at the network level with data-driven 79 methods. We used independent component analysis (ICA), which is widely applied in resting80 state and task-based fMRI studies (Du et al., 2017; Jarrahi et al., 2015; Jung et al., 2020), to 81 identify the body sensitive nodes within the TOC network and further tracked their whole-brain 82 communications during whole body action processing. 83 2. Method 84 The study consisted of two experiments: a localizer experiment and the main experiment. First, 85 we used the localizer experiment to identify the temporo-occipital network associated with body 86 action perception. This was accomplished through a data-driven strategy based on our previous 87 study (Li et al., 2023). Next, the data of the main experiment was employed to extract node 88 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 5 regions within this network and investigate their connectivity profiles as well as their modulation 89 by affective body conditions. Nineteen participants took part in the experiment. 90 2.1. Participants 91 Nineteen healthy participants (mean age = 24.58 years; age range = 19-30 years; 6 males, all 92 right-handed) took part in the experiment. All participants had a normal or correct-ed-to-normal 93 vision and no medical history of any psychiatric or neurological disorders. All participants 94 provided informed written consent before the start of the experiment and received a monetary 95 reward (vouchers) or course credits for their participation. The experiment was approved by the 96 Ethical Committee at Maastricht University and was performed in accordance with the 97 Declaration of Helsinki. 98 2.2. Experiment Design 99 2.2.1. Network localizer 100 The functional localizer followed a blocked design with twelve categories of videos com-posed 101 from three factors: (body / face / object) * (human / monkey) * (normal / scramble). Each 102 category consisted of ten different 1000-ms videos which were presented block-wise in a random 103 order. Within each block, the ten videos were interleaved by a fixed 500-ms inter-trial interval, 104 while two consecutive blocks were interleaved by an inter-block interval jittered around 11 105 seconds. The order of block conditions was randomized for each participant, and each condition 106 was repeated six times within three runs. Each run contained a catch block where, in one of the 107 trials, the fixation point changed its shape from a “+” to a “o”. Participants were instructed to 108 make a button-press response when detecting the fixation shape change. The total length of each 109 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 6 run was 735 seconds on average. A detailed description of the localizer stimuli and design can be 110 found in Li et al. (2023). 111 2.2.2. Main experiment 112 For the main experiment, we applied a mixed block/event-related design (Visscher et al., 2003) 113 consisting of five conditions of videos: three human body conditions (aggressive, defensive, and 114 neutral), one neutral human face condition and one neutral object condition (Figure 1). Each 115 condition consisted of ten different 1000-ms videos. The body and face videos were chosen from 116 the same stimulus set as described in Kret et al. (2011b). The object videos were selected from 117 the same set as used in the localizer experiment but differed from the ones used in that 118 experiment. 119 During the experiment, each condition was presented block-wise with a jittered inter-trial 120 interval of around 3 seconds and an inter-block interval of 12 seconds. For each trial, the video 121 was centered and presented on a uniform gray background. The sizes of stimuli were 3.5*7.5 122 degrees of visual angle for bodies and objects, and 3.5*3.5 degrees of visual angle for faces. The 123 order of block conditions was randomized for each participant, and each condition was repeated 124 ten times within five runs. Two extra blocks with a catch trial were inserted in each run where 125 participants were instructed to detect the fixation shape changes as described in the localizer 126 experiment. The total length of each run was 480 seconds. 127 Both the main experiment and the localizer experiment were programmed using the 128 Psychtoolbox (https://www.psychtoolbox.net) implemented in Matlab 2018b 129 (https://www.mathworks.com). Stimuli were projected onto a screen at the end of the scanner 130 bore with a Panasonic PT-EZ57OEL projector (screen size = 30 * 18 cm, resolution = 1920 * 131 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 7 1200 pixel). Participants viewed the stimuli through a mirror attached to the head coil (screen-to132 eye distance = 99 cm, visual angle = 17.23 * 10.38 degrees). 133 2.2.3. fMRI data acquisition 134 All images were acquired with a 7T MAGNETOM scanner at the Maastricht Brain Imaging 135 Centre (MBIC) of Maastricht University, the Netherlands. Functional images were collected 136 using the T2*-weighted multi-band accelerated EPI 2D BOLD sequence (TR/TE = 1000/20 ms, 137 multiband acceleration factor = 3, in-plane isotropic resolution = 1.6 mm, number of slices per 138 volume = 68, matrix size = 128 * 128, volume number = 735 for the network localizer and 480 139 for the main experiment). T1-weighted anatomical images were obtained using the 3D140 MP2RAGE sequence (TR/TE = 5000/2.47 ms, Inverse time TI1/I2 = 900/2750 ms, flip angle 141 FA1/FA2 = 5/3°, in-plane isotropic resolution = 0.7 mm, matrix size = 320 * 320, slice number = 142 240). Physiological parameters were recorded via pulse oximetry on the index finger of the left 143 hand and with a respiratory belt. 144 2.2.4. fMRI image preprocessing 145 Anatomical and functional images were preprocessed using the Brainvoyager 22 (Goebel, 2012), 146 and the Neuroelf toolbox in Matlab (https://neuroelf.net/). For anatomical images, brain 147 extraction was conducted with INV2 images to correct for MP2RAGE background noise. The 148 resolution was then downsampled to 0.8 mm for better alignment to the 1.6 mm resolution of 149 functional images. For functional images, the preprocessing steps included EPI distortion 150 correction (Breman et al., 2020), slice scan time correction, 3D head-motion correction, and 151 high-pass temporal filtering (GLM with Fourier basis set of 3 cycles, including linear trend). 152 Coregistration was first conducted between the anatomical image and its most adjacent 153 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 8 functional run using a boundary-based registration (BBR) algorithm (Greve & Fischl, 2009), and 154 all the other functional runs were coregistered to the aligned run. Individual images were 155 normalized to Talairach space (Collins, Neelin, Peters, & Evans, 1994) with 3 mm Gaussian 156 spatial smoothing. Trilinear/sinc interpolation was used in the motion correction step, and sinc 157 interpolation was used in all the other steps. 158 Physiological parameters were collected as confound factors for the functional imaging data. The 159 physiological data were preprocessed using the RETROspective Image CORrection 160 (RETROICOR; Glover et al., 2000; Harvey et al., 2008) pipeline, which uses Fourier expansions 161 of different orders for the phase of cardiac pulsation (3rd order), respiration (4th order) and 162 cardio-respiratory interaction (1st order). Eighteen physiological confound factors were finally 163 created for each participant. 164 The anatomical labeling of the brain areas reported in this study was performed according to the 165 Talairach Daemon (http://www.talairach.org/daemon.html) in combination with the Multilevel 166 Human Brain Atlas (https://ebrains.eu/service/human-brain-atlas). 167 2.3. Data analysis 168 2.3.1. Body network extraction 169 The Infomax algorithm implemented in the Group ICA of fMRI Toolbox (GIFT; Calhoun et al., 170 2001) was used to identify body selective networks within the localizer experiment. This resulted 171 in 75 spatial independent components. Individual ICs were back-reconstructed using the GIG172 ICA algorithm from the aggregated group ICs (Du & Fan, 2013). The stability of group ICA was 173 assessed by the ICASSO module implemented in the GIFT, which repeated the Infomax 174 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 9 decomposition 20 times and resulted in an index of stability (Iq) for each IC (Himberg et al., 175 2004). Prior to the group-ICA, physiological and motion confounds were regressed out from the 176 preprocessed functional images. The resulting time courses were then transformed into 177 percentages of signal change to enhance the ICA stability (Allen et al., 2011). Components 178 showing large white matter / cerebrospinal fluid coverage were excluded from further analysis. 179 To identify body selective networks, we conducted a GLM on each reconstructed subject-level 180 IC time course, which estimated the IC response for each condition. In the design matrix, each 181 condition predictor was modeled as a boxcar function with the same duration of the block and 182 convolved with the canonical HRF. The estimated betas were first averaged across all runs for 183 each participant and were then used to calculate the contrast of [2 * human body (normal - 184 scramble) – (human face (normal - scramble) + human object (normal - scramble))]. Right-tailed 185 t-tests and Benjamini-Hochberg multiple comparison corrections were conducted at the group 186 level to find significant body sensitivity. 187 To define the group-level coverage of the IC networks, the individual IC maps were normalized 188 to z-scores and averaged across all runs for each participant. A group t-test against zero was 189 computed using the z-scored maps of each subject and corrected using a cluster-threshold 190 statistical procedure based on Monte-Carlo simulation (initial p < 0.005, alpha level = 0.05, 191 iteration = 5000). The group-level coverage of the network was then used as the initial mask for 192 the body node extraction in the main experiment. 193 2.3.2. Connectivity of the TOC body nodes 194 The analysis for the main experiment is illustrated in Figure 1. First, a fixed-effects GLM was 195 conducted on each participant’s functional images with each different video as a separate 196 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 16 for TOPJ3 compared to PHT/TPOJ2, whereas PHT/TPOJ2 has been associated with stronger 326 body form selectivity (Glasser et al., 2016). Thus, the pMTG nodes and the LOCs node may be 327 sensitive to different kinds of visual features associated with specific features of social 328 information. 329 4.2. Different connectivity profiles 330 An ANOVA with seed nodes and affective conditions as factors was used to compare the nodes’ 331 connectivity profiles. Distinct global connectivity profiles were revealed by the significant main 332 effect of the seed node. The only cluster showing stronger connectivity to the LOCi node was 333 EVA/V2. Regarding to the LOCs node, the strongest connectivity was found with default mode 334 network and dorsal attention network nodes including PCC, precuneus, and superior frontal 335 gyrus (SFG). This is consistent with other studies reporting ventral stream connectivity to the 336 LOC (overlap with the LOCi) and dorsal stream connectivity to the EBA (overlap with the LOCs) 337 (Zimmermann et al., 2018). Compared to the LOCi and LOCs, more widespread connectivity 338 was found for the two anterior nodes, lpMTG and rpMTG, which suggested the pMTG nodes 339 may serve as network hubs connecting the TOC network to the global-level computation. Both 340 pMTG nodes had the highest connectivity to SMG and insula. However, in the case of the 341 rpMTG, this connectivity extended further to encompass the MFG, angular gyrus and SFG. Also, 342 different from the LOCs, the anterior nodes were linked to nodes of ventral attention / salience 343 networks (VAN & SN). The asymmetric results are consistent with the right-literalized 344 distribution of the VAN (Vossel et al., 2014). Thus, these two nodes could contribute to a 345 processing pathway for evaluating both valence as well as relevance in the observer. 346 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 17 4.3. Affective modulation for defensive vs. aggressive body movements 347 The most significant and novel finding derived from our analysis is that for defensive/fearful 348 body images we observed enhanced connectivity between the TOC and the PCC / precuneus as 349 well as the caudate. The former are part of Posteromedial cortex (PMC), known for involvement 350 in episodic memory, including autobiographical memory (Bubb et al., 2017; Leech & Sharp, 351 2014), and are also part of the default mode network. The PCC has also been shown to exhibit 352 increased activity when attention is directed towards a target of high motivational value (Leech 353 & Sharp, 2014). Furthermore, recent evidence indicates that Anterior Precuneus is causally 354 linked to the bodily self, or the brain resources involved with the observers’ body schema (Lyu et 355 al., 2023). Seen against this background, witnessing fear/defensive actions may engage brain 356 activity intricately associated with the bodily self, in the sense of increasing the involvement of 357 the observer. These results suggest that observing fearful expressions is associated with 358 concomitant neural activity preparing the brain for defensive actions (de Gelder et al., 2004). In 359 support, several brain areas identified in studies of body schema (Berlucchi & Aglioti, 2010), 360 sense of self and its deficits in pathological conditions (Dary et al., 2023) overlap with the 361 network described here. 362 In contrast, the connectivity modulation for aggressive actions showed a different pattern and 363 varied across different nodes. As revealed by the ANOVA analysis, a significant interaction effect 364 was observed for FEF and cerebellum regions. When viewing the aggressive videos, the two 365 regions showed increased connectivity to the lpMTG node, and decreased connectivity to the 366 LOCs node. As mentioned above, both lpMTG and LOCs were composed of voxels from social 367 perception regions such as TPOJ and PHT and may be related to the processing of different types 368 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 18 of social features. And compared to LOCs, the lpMTG node exhibited a higher level of global 369 connectivity. Moreover, the FEF region is known for its involvement in shaping attention 370 (Veniero et al., 2021) and as such the FEF may be part of decision making (Krajbich et al., 2021). 371 Thus, the opposite FEF connectivity pattern to the lpMTG and LOCs may suggest the 372 reorientation of salient features and the distributed computation during the presence of 373 aggressive stimuli. 374 5. Conclusions 375 We showed that body representation is implemented in the brain in several different body 376 sensitive hubs that each have their connectivity network. These hubs and connectivity networks 377 have a relatively specific functional role in body and action perception. This supports the notion 378 that these are body sensitive hubs that are not so much defined by abstract body category 379 selectivity than by their different functional roles in action understanding. 380 381 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 19 References 382 Allen, E. A., Erhardt, E. B., Damaraju, E., Gruner, W., Segall, J. M., Silva, R. F., . . . Calhoun, V. D. 383 (2011). A baseline for the multivariate comparison of resting-state networks. Frontiers in Systems 384 Neuroscience, 5, 2. doi:10.3389/fnsys.2011.00002 385 Argyle, M. (1976). Non-verbal Communication and Language. Royal Institute of Philosophy Supplements, 386 10, 63-78. 387 Berlucchi, G., & Aglioti, S. M. (2010). The body in the brain revisited. Experimental brain research, 200, 388 25-35. 389 Breman, H., Mulders, J., Fritz, L., Peters, J., Pyles, J., Eck, J., . . . Goebel, R. (2020). An image 390 registration-based method for epi distortion correction based on opposite phase encoding (COPE). 391 Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence 392 and Lecture Notes in Bioinformatics), 12120 LNCS, 122–130. doi:10.1007/978-3-030-50120393 4_12 394 Bubb, E. J., Kinnavane, L., & Aggleton, J. P. (2017). Hippocampal–diencephalic–cingulate networks for 395 memory and emotion: An anatomical guide. Brain and neuroscience advances, 1, 396 2398212817723443. 397 Calhoun, V. D., Adali, T., Pearlson, G. D., & Pekar, J. J. (2001). A method for making group inferences 398 from functional MRI data using independent component analysis. Human Brain Mapping, 14(3), 399 140–151. doi:10.1002/hbm.1048 400 Caspers, S., Zilles, K., Laird, A. R., & Eickhoff, S. B. (2010). ALE meta-analysis of action observation 401 and imitation in the human brain. NeuroImage, 50, 1148–1167. 402 doi:10.1016/j.neuroimage.2009.12.112 403 Dary, Z., Lenggenhager, B., Lagarde, S., Medina Villalon, S., Bartolomei, F., & Lopez, C. (2023). Neural 404 bases of the bodily self as revealed by electrical brain stimulation: A systematic review. Human 405 Brain Mapping, 44(7), 2936-2959. 406 de Gelder, B., & Poyo Solanas, M. (2021). A computational neuroethology perspective on body and 407 expression perception. Trends in Cognitive Sciences, 25, 744–756. doi:10.1016/j.tics.2021.05.010 408 de Gelder, B., Snyder, J., Greve, D., Gerard, G., & Hadjikhani, N. (2004). Fear fosters flight: A 409 mechanism for fear contagion when perceiving emotion expressed by a whole body. Proceedings 410 of the National Academy of Sciences of the United States of America, 101, 16701–16706. 411 doi:10.1073/pnas.0407042101 412 de Gelder, B., Van den Stock, J., Meeren, H. K. M., Sinke, C. B. A., Kret, M. E., & Tamietto, M. (2010). 413 Standing up for the body. Recent progress in uncovering the networks involved in the perception 414 of bodies and bodily expressions. Neuroscience and Biobehavioral Reviews, 34, 513–527. 415 doi:10.1016/j.neubiorev.2009.10.008 416 Deen, B., Schwiedrzik, C. M., Sliwa, J., & Freiwald, W. A. (2023). Specialized Networks for Social 417 Cognition in the Primate Brain. Annual Review of Neuroscience, 46, 381-401. 418 Downing, P., & Kanwisher, N. (2001). A cortical area specialized for visual processing of the human body. 419 Journal of Vision, 1, 2470–2473. doi:10.1167/1.3.341 420 Du, Y., & Fan, Y. (2013). Group information guided ICA for fMRI data analysis. NeuroImage, 69, 157– 421 197. 422 Du, Y., Pearlson, G. D., Lin, D., Sui, J., Chen, J., Salman, M., . . . Calhoun, V. D. (2017). Identifying 423 dynamic functional connectivity biomarkers using GIG-ICA: Application to schizophrenia, 424 schizoaffective disorder, and psychotic bipolar disorder. Human Brain Mapping, 38, 2683–2708. 425 doi:10.1002/hbm.23553 426 Glasser, M. F., Coalson, T. S., Robinson, E. C., Hacker, C. D., Harwell, J., Yacoub, E., . . . Jenkinson, M. 427 (2016). A multi-modal parcellation of human cerebral cortex. Nature, 536(7615), 171-178. 428 Glover, G. H., Li, T. Q., & Ress, D. (2000). Image  based method for retrospective correction of 429 physiological motion effects in fMRI: RETROICOR. Magnetic Resonance in Medicine, 44, 162– 430 167. doi:10.1002/1522-2594(200007)44:1<162::AID-MRM23>3.3.CO;2-5 431 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 20 Goebel, R. (2012). BrainVoyager—past, present, future. NeuroImage, 62(2), 748-756. 432 Goldberg, H., Preminger, S., & Malach, R. (2014). The emotion-action link? Naturalistic emotional 433 stimuli preferentially activate the human dorsal visual stream. NeuroImage, 84, 254–264. 434 doi:10.1016/j.neuroimage.2013.08.032 435 Grèzes, J., Pichon, S., & de Gelder, B. (2007). Perceiving fear in dynamic body expressions. NeuroImage, 436 35, 959–967. doi:10.1016/j.neuroimage.2006.11.030 437 Grill-Spector, K., & Malach, R. (2004). The human visual cortex. Annu. Rev. Neurosci., 27, 649-677. 438 Haak, K. V., & Beckmann, C. F. (2018). Objective analysis of the topological organization of the human 439 cortical visual connectome suggests three visual pathways. Cortex, 98, 73–83. 440 doi:10.1016/j.cortex.2017.03.020 441 Harvey, A. K., Pattinson, K. T. S., Brooks, J. C. W., Mayhew, S. D., Jenkinson, M., & Wise, R. G. (2008). 442 Brainstem functional magnetic resonance imaging: Disentangling signal from physiological noise. 443 Journal of Magnetic Resonance Imaging, 28, 1337–1344. doi:10.1002/jmri.21623 444 Himberg, J., Hyvärinen, A., & Esposito, F. (2004). Validating the independent components of 445 neuroimaging time series via clustering and visualization. NeuroImage, 22, 1214–1222. 446 doi:10.1016/j.neuroimage.2004.03.027 447 Jarrahi, B., Mantini, D., Balsters, J. H., Michels, L., Kessler, T. M., Mehnert, U., & Kollias, S. S. (2015). 448 Differential functional brain network connectivity during visceral interoception as revealed by 449 independent component analysis of fMRI time-series. Human Brain Mapping, 36, 4438–4468. 450 doi:10.1002/hbm.22929 451 Jastorff, J., & Orban, G. A. (2009). Human functional magnetic resonance imaging reveals separation and 452 integration of shape and motion cues in biological motion processing. Journal of Neuroscience, 453 29(22), 7315-7329. 454 Jung, J. Y., Bungert, A., Bowtell, R., & Jackson, S. R. (2020). Modulating Brain Networks With 455 Transcranial Magnetic Stimulation Over the Primary Motor Cortex: A Concurrent TMS/fMRI 456 Study. Frontiers in Human Neuroscience, 14, 31. doi:10.3389/fnhum.2020.00031 457 Kolster, H., Peeters, R., & Orban, G. A. (2010). The retinotopic organization of the human middle 458 temporal area MT/V5 and its cortical neighbors. Journal of Neuroscience, 30(29), 9801-9820. 459 Krajbich, I., Mitsumasu, A., Polania, R., Ruff, C. C., & Fehr, E. (2021). A causal role for the right frontal 460 eye fields in value comparison. eLife, 10, e67477. 461 Kret, M. E., Denollet, J., Grèzes, J., & de Gelder, B. (2011a). The role of negative affectivity and social 462 inhibition in perceiving social threat: an fMRI study. Neuropsychologia, 49(5), 1187-1193. 463 Kret, M. E., Pichon, S., Grèzes, J., & De Gelder, B. (2011b). Similarities and differences in perceiving 464 threat from dynamic faces and bodies. An fMRI study. NeuroImage, 54, 1755–1762. 465 doi:10.1016/j.neuroimage.2010.08.012 466 Larsson, J., & Heeger, D. J. (2006). Two retinotopic visual areas in human lateral occipital cortex. Journal 467 of Neuroscience, 26(51), 13128-13142. 468 Leech, R., & Sharp, D. J. (2014). The role of the posterior cingulate cortex in cognition and disease. Brain, 469 137(1), 12-32. 470 Li, B., Solanas, M. P., Marrazzo, G., Raman, R., Taubert, N., Giese, M., . . . de Gelder, B. (2023). A large471 scale brain network of species-specific dynamic human body perception. Progress in 472 Neurobiology, 221, 102398. 473 Lingnau, A., & Downing, P. E. (2015). The lateral occipitotemporal cortex in action. Trends in Cognitive 474 Sciences, 19(5), 268-277. 475 Lyu, D., Stieger, J. R., Xin, C., Ma, E., Lusk, Z., Aparicio, M. K., . . . Buch, V. (2023). Causal evidence 476 for the processing of bodily self in the anterior precuneus. Neuron. 477 Marrazzo, G., De Martino, F., Lage-Castellanos, A., Vaessen, M. J., & de Gelder, B. (2023). Voxelwise 478 encoding models of body stimuli reveal a representational gradient from low-level visual features 479 to postural features in occipitotemporal cortex. NeuroImage, 120240. 480 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 21 Marrazzo, G., Vaessen, M. J., & de Gelder, B. (2021). Decoding the difference between explicit and 481 implicit body expression representation in high level visual, prefrontal and inferior parietal cortex. 482 NeuroImage, 243, 118545. 483 Moreau, Q., Parrotta, E., Pesci, U. G., Era, V., & Candidi, M. (2023). Early categorization of social 484 affordances during the visual encoding of bodily stimuli. NeuroImage, 120151. 485 Peelen, M. V., Atkinson, A. P., Andersson, F., & Vuilleumier, P. (2007). Emotional modulation of body486 selective visual areas. Social cognitive and affective neuroscience, 2(4), 274–283. 487 Pichon, S., de Gelder, B., & Grèzes, J. (2008). Emotional modulation of visual and motor areas by 488 dynamic body expressions of anger. Social neuroscience, 3, 199–212. 489 doi:10.1080/17470910701394368 490 Poyo Solanas, M., Vaessen, M., & de Gelder, B. (2020). Computation-based feature representation of 491 body expressions in the human brain. Cerebral Cortex, 30(12), 6376–6390. 492 Robert, S., Ungerleider, L. G., & Vaziri-Pashkam, M. (2023). Disentangling object category 493 representations driven by dynamic and static visual input. Journal of Neuroscience, 43(4), 621494 634. 495 Sayres, R., & Grill-Spector, K. (2008). Relating retinotopic and object-selective responses in human 496 lateral occipital cortex. Journal of Neurophysiology, 100(1), 249-267. 497 Shmuelof, L., & Zohary, E. (2005). Dissociation between ventral and dorsal fMRI activation during 498 object and action recognition. Neuron, 47(3), 457-470. 499 Veniero, D., Gross, J., Morand, S., Duecker, F., Sack, A. T., & Thut, G. (2021). Top-down control of 500 visual cortex by the frontal eye fields through oscillatory realignment. Nature Communications, 501 12(1), 1757. 502 Visscher, K. M., Miezin, F. M., Kelly, J. E., Buckner, R. L., Donaldson, D. I., McAvoy, M. P., . . . Petersen, 503 S. E. (2003). Mixed blocked/event-related designs separate transient and sustained activity in 504 fMRI. NeuroImage, 19(4), 1694-1708. 505 Vogels, R. (2022). More Than the Face: Representations of Bodies in the Inferior Temporal Cortex. 506 Annual Review of Vision Science, 8. doi:10.1146/annurev-vision-100720-113429 507 Vossel, S., Geng, J. J., & Fink, G. R. (2014). Dorsal and ventral attention systems: distinct neural circuits 508 but collaborative roles. The Neuroscientist, 20(2), 150-159. 509 Weiner, K. S., & Grill-Spector, K. (2010). Sparsely-distributed organization of face and limb activations 510 in human ventral temporal cortex. NeuroImage, 52(4), 1559-1573. 511 Weiner, K. S., & Grill-Spector, K. (2011). Not one extrastriate body area: using anatomical landmarks, 512 hMT+, and visual field maps to parcellate limb-selective activations in human lateral 513 occipitotemporal cortex. NeuroImage, 56(4), 2183–2199. 514 Zhan, M., Goebel, R., & de Gelder, B. (2018). Ventral and dorsal pathways relate differently to visual 515 awareness of body postures under continuous flash suppression. Eneuro, 5(1). 516 Zimmermann, M., Mars, R. B., De Lange, F. P., Toni, I., & Verhagen, L. (2018). Is the extrastriate body 517 area part of the dorsal visuomotor stream? Brain Structure and Function, 223(1), 31–46. 518 Zimmermann, M., Meulenbroek, R. G., & de Lange, F. P. (2012). Motor planning is facilitated by 519 adopting an action's goal posture: an fMRI study. Cerebral Cortex, 22(1), 122-131. 520 521 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 2 2 Figures & tables 522 523 Figure 1. Illustration of the main experiment design and analysis. Five video conditions 524 (aggressive / defensive / neutral face blurred body, neutral face, and object ; face region overed 525 here for privacy) were presented in a mixed block/event-related design, in whi ch the stimuli were 526 blocked for each condition while with jittered inter-trial-interval around 3s. For each condition, 527 ten different videos were included and were repeated ten times across five runs. GLM was 528 conducted to estimate the response for each different video, resulting in 50 betas extracted for 529 each participant. The videowise betas were then entered to an ICA procedure within the body 530 sensitive TOC network identified by the localizer experiment. Body selective network nodes 531 were defined by higher component responses for body videos than for non-body videos. To track 532 the whole-brain connectivity of each selected node, the video-wise betas were zscored and 533 convolved with the hemodynamic response function within each condition, resulting in five 534 reconstructed time-courses for each component. The reconstructed timecourses for all selected 535 components were then added to a whole-brain GLM design matrix as the predictors for seed - 536 based connectivity. Finally, two-factor ANOVA was conducted with connectivity betas across all 537 participants to test their modulations from the body conditions or the node components. 538 2 ns ed re n, as for dy es ck nd ve ed - all (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 23 539 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 2 4 540 Figure 2. (a). The coverage of T OC network as defined in the network localizer experiment. (b). 541 Beta plots of the four body-nodes from the main experiment. Zeropoint indicates the mean beta 542 value across all masked voxels. Colors indicate the component indexes. (c). The spatial 543 distribution of the four body nodes. (d). The relative position of the centers of body nodes and 544 the hMT (black) on the left hemisphere. 545 546 4 b). eta ial nd (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint 2 5 547 Figure 3. (a). Map of voxels with significantly higher contribution from each of the nodes. The 548 voxelwise IC weight from each node was compared to the other three nodes and entered a 549 group-level t-test against zero (two-tailed). The resulting map was corrected by a cluster - 550 threshold statistical procedure based on MonteCarlo simulation (initial p < 0.005, alpha level = 551 0.05, iteration = 5000). Slice numbers indicate the Z coordinates in the Talairach space. (b). The 552 voxel composition for each cluster in (a.), labeled by the HCP-MMP atlas (Glasser et al., 2016). 553 554 5 he a - l = he (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 15, 2024. ; https://doi.org/10.1101/2024.01.12.574860doi: bioRxiv preprint