Architectural Affordance Impacts Human Sensorimotor Brain Dynamics

Action is a medium of collecting sensory information about the environment, which in turn is shaped by architectural affordances. Affordances characterize the fit between the physical structure of the body and capacities for movement and interaction with the environment, thus relying on sensorimotor processes associated with exploring the surroundings. Central to sensorimotor brain dynamics, the attentional mechanisms directing the gating function of sensory signals share neuronal resources with motor-related processes necessary to inferring the external causes of sensory signals. Such a predictive coding approach suggests that sensorimotor dynamics are sensitive to architectural affordances that support or suppress specific kinds of actions for an individual. However, how architectural affordances relate to the attentional mechanisms underlying the gating function for sensory signals remains unknown. Here we demonstrate that event-related desynchronization of alpha-band oscillations in parieto-occipital and medio-temporal regions covary with the architectural affordances. Source-level time-frequency analysis of data recorded in a motor-priming Mobile Brain/Body Imaging experiment revealed strong event-related desynchronization of the alpha band to originate from the posterior cingulate complex and bilateral parahippocampal areas. Our results firstly contribute to the understanding of how the brain resolves architectural affordances relevant to behaviour. Second, our results indicate that the alpha-band originating from the posterior cingulate complex covaries with the architectural affordances before participants interact with the environment. During the interaction, the bilateral parahippocampal areas dynamically reflect the affordable behaviour as perceived through the visual system. We conclude that the sensorimotor dynamics are developed for processing behaviour-relevant features in the designed environment.


Introduction 34
When we act, we are changing the perceived environment according to a set of expectations that depend on 35 our body and the environment. The potential ways we can act depends on the affordances of the environment 1 . 36 Affordances refer to the possibilities for use, intervention, and action which the physical world offers and are 37 determined by the fit between a body's structure, skills, and capacities for movement and the action-related 38 properties of the environment 2 . Affordances are thus perceptual and action-related expectations that are 39 systematically reflected in sensorimotor dynamics 3 . In this sense, cognitive functions that depend on sensory or 40 motor activity are not bound by the physical structure of the body alone, but also by the functional ways in 41 which we interact with the environment. 42 Preparation and selection of motor action have been studied extensively using frequency-specific oscillatory 43 activity. It is generally held that a vital function of the oscillations in the brain entails transferring information 44 across regions to sustain the binding processes related to various cognitive functions 4-6 . Particularly, decreases in 45 alpha power relative to a baseline, generally referred to as event-related desynchronization ERD, for review: ,7-9 , are 46 consistently reported after a preparatory stimulus in sensorimotor tasks [10][11][12] . ERD, as opposed to event-related 47 synchronization (ERS), reflects the release from inhibitory mechanisms and is therefore linked to active cortical 48 processing 9 . In a traditional motor-related priming task (S1-S2 paradigm), it is held that the processes necessary 49 to interact with the upcoming event occur between the preparatory stimulus (S1) and the imperative stimulus 50 (S2) 13 . These processes are typically pronounced through ERD between S1 and S2. 51 Traditionally, the selection of perceived information and the selection of appropriate motor responses has 52 been attributed to attentional mechanisms as separate processes occurring in sequences 14 . In this sense, 53 attentional mechanisms are responsible for the selection processes regarding sensory information, which then is 54 followed by central executive transformation and selection processes that are then transferred to motor-related 55 mechanisms that implement the selection and execute relevant motor responses. However, recent studies suggest 56 that the preparation and execution of actions and the selection of sensory information unfold in parallel rather 57 than serial-like processes 3,15-17 . These studies suggest that the selection of motor response and the selection of 58 sensory information may be reflected by the same neural processes 18 . Therefore, we approach motor attention 59 as sharing the same neuronal resources as visual attention 19 . The ecological account and predictive coding 60 account of the brain are consistent with such an approach. Accordingly, actions comprise motor predictions 61 based on the sensory consequences of the trajectory of the action 20,21 , where the shape of these actions depends 62 on the built environment. Therefore, action and perception both shape and are shaped by the environment in 63 ways that comply with a dynamic set of motor predictions, i.e. proprioceptive sensations, and perceptual 64 predictions, i.e. exteroceptive consequences of action 20,22 . 65 Cortical oscillations as measured with electrophysiological instruments have for long been associated with 66 such behavioural and cognitive states 6,7,9,23 . Numerous frequency bands can be dissociated within the power 67 spectrum of the electroencephalogram (EEG), for example, theta, alpha, beta, and gamma-however, their 68 functional significance has caused numerous debates. The emergent alpha rhythm in parietal and occipital areas 69 has been linked to anticipatory attention 24,25 , re-allocation of resources 26,27 , inhibition of visual areas to suppress 70 the processing of irrelevant visual information 28 , top-down inhibition of task-irrelevant brain regions 6 , and 71 additionally, to the timing of inhibitory processes 9 . Given the many discovered characteristics of the alpha 72 rhythm 29 , it may be misguiding to ascribe very general concepts to specific frequency bands 30 . Instead, our 73 approach lends itself to a biological account of oscillations, where the rhythms are not end-products of a specific 74 cognitive function, but rather components underlying information processing in the brain with the interaction 75 between rhythms revealing more about brain function 30,31 . Here, we focus on how alpha ERD fits within the 76 processes that involve the selection and control of interactions with the environment. In this sense, we do not 77 cast the cortical mechanisms as semantic processes, but instead as ecological and goal-directed. Thus, the 78 concept of 'attention' is here used only relative to the task at hand and cannot be generalized. 79 We set out to investigate the claim that specific frequencies are modulated in a way that reflects the potential 80 to act in a given environment, i.e. affordances. In a previous study 3 , we demonstrated this systematic covariation 81 of affordance and brain dynamics in the time domain with modulations of early event-related potentials (ERPs) 82 reflecting the potential to act. Extending on these results, we investigate here how sensorimotor activity is 83 induced relative to motor-selectional mechanisms both (i) upon perceiving the environment and (ii) during the 84 interaction with the environment. By 'sensorimotor dynamics' we refer to the brain dynamics when an observer 85 estimates the state of the environment while estimating its own state by integrating sensory and motor 86 information. These dynamics, therefore, reflect one aspect of behaviour. Sensorimotor dynamics comprise 87 changes in the visual, somatosensory, motor, and multisensory areas dependent on aspects of the environment. 88 In this regard, we analysed brain data from an earlier Mobile Brain/Body Imaging (MoBI) 32-34 experiment that 89 was designed as a motor priming paradigm 3 . Participants were located in one of two rooms in a Virtual Reality 90 (VR) environment and were instructed to pass through transitions of varying width connecting the two rooms. A 91 preparatory stimulus (S1) revealed perceptual information about the upcoming transition (whether the transition 92 was too narrow to pass or passable), which was then followed by an imperative stimulus (S2) that revealed 93 whether participants had to actively move through the transition or not. In the interval between S1 and S2, the 94 participants anticipated how to interact with the opening, while S2 instructed participants to pass the transition 95 or remain in the same room. The paradigm thus provides an excellent opportunity to answer the following 96 research question: how are attentional mechanisms relative to the selective processing of actions reflected in the 97 brain? With the aforementioned literature in mind, we expected to find desynchronization over parietal, visual 98 and motor cortices reflecting attention, visual processing and motor activity to covary with the affordances of the 99 environment. We found systematic variations in the parieto-occipital alpha band when participants approached 100 transitions with positive or negative affordance that mirrored the desynchronization patterns. Further 101 examination of the alpha-band activity revealed desynchronization over temporo-occipital areas during 102 perceiving poor affordances (when the transition was too narrow to pass), as compared to the perception of 103 positive affordances (when the transition was passable). These results suggest that the brain dynamics followed a 104 pattern of affordance where the possibility to act determined the type of sensory gating in sensorimotor areas. 105 106

Method 107
The following description of the experimental paradigm corresponds to the original experiment conducted by 108 Djebbara and colleagues 3 . 109 Participants. The experimental procedure included 20 participants, of which 9 were female (mean age 28.1 110 years, σ = 6.2 years). All participants were recruited from a participant pool of the Technical University of Berlin, 111 Germany and received either monetary compensation (€10/h) or accredited course hours. None of the 112 participants had a history of neurological pathologies and normal, or corrected to normal, vision with no specific 113 background in architecture. A written informed consent was given and the protocol was approved by the local 114 Ethics Committee and signed by all participants. A single participant was excluded due to technical issues. 115 Paradigm. Integrating MoBI with Virtual Reality (VR) allowed for recording and analysing brain activity 116 using EEG in freely behaving human participants moving in virtually designed spaces. The experimentation 117 unfolded in the experimental room of 160 m 2 in the Berlin Mobile Brain/Body Imaging Laboratories 118 (BeMoBIL). The virtual space was designed to occupy a total of 9 m  5 m for both virtual rooms, i.e. each room 119 was 4.5 m  5 m. To complete the task, participants had to transit from one room to a second room-however, 120 doors of different widths manipulated the transition affordances between rooms. The widths of the door in VR 121 varied from impassable (0.2 m, Narrow), to passable (1 m, Mid) to easily passible (1.5 m, Wide). The paradigm was 122 a forewarned Go/NoGo paradigm, also known as an S1-S2 paradigm, where S1 was a first stimulus serving as a 123 preparatory signal. S1 presented the participants with the environment including the transition width and S2 was 124 the imperative stimulus. S2 revealed whether participants were allowed to interact with the environment (Go), or 125 not (NoGo). S2 was pseudorandomized for 50% Go and 50% NoGo. Therefore, the experimental design was a 3  126 2 repeated measures design where the factors were the type of doors with three levels (Narrow, Mid, Wide) and 127 the imperative movement instructions with two levels (Go/NoGo). Each participant responded to 240 trials with 128 40 trials for each factor level. A training phase before starting the 240 trials ensured that participants were 129 comfortable with the protocol and got accustomed to the VR environment. All events in the experiment were 130 registered and collected using LabStreamingLayer 35 . The main investigator withdrew to the control room and 131 observed the participants through two cameras and a mirrored display of the head-mounted displays the 132 participant was wearing. This ensured minimal-to-none interaction with the participant once the experiment was 133 commenced. 134 A single trial comprised starting in the dark in the first room inside a predefined starting square directed 135 towards the door (see Figure 1). The participants had to wait for 3 s on average (ITI = 3 s ± 1 s) before the 136 "lights" would go on (S1), so they could perceive the environment including the type of door they had to transit.

137
Facing the closed door, they had to wait for 6 s (ITI = 6 s ± 1 s) before the door would turn green or red for Go 138 or NoGo, respectively. In the case of a Go trial, participants were instructed to walk towards the door, which 139 would slide open when within a distance of 0.3 m, fetch a floating red circle in the second space using their 140 controller, and subsequently return to their starting square. Touching the red circle would elicit a reward of €0.1 141 added to their reimbursement. If the door was too narrow to pass, they were instructed to try until the walls 142 turned red indicating a collision with the wall resulting in a text informing that they have failed to pass. In the 143 case of a NoGo trial, the participants were instructed to not transition into the second room, but to fill in the 144 emotional questionnaire (Self-Assessment Manikin; SAM) and move on to the next trial. After each trial, they 145 were instructed to fill in the SAM questionnaire irrespective of whether they transitioned through the door or 146 not before moving on to the next trial. The SAM was filled in using a laser pointer from the hand-controller, 147 which also controlled when to turn the "lights off" to move on to the next trial.  scripted in, and powered by, Unity. The participants were equipped with a hand-controller by Acer that was 164 linked to the VR-system (see Figure 2). All EEG data were recorded (DC) with a 0.3 Hz high-pass filter and 165 sampled at 500 Hz with impedances kept below 10 k. Computational delays were measured by parallel 166 processing a direct event-marker and an event-marker through Unity. The 20 ms ± 4 ms were corrected during 167 the analysis. EEGLAB Toolbox 36 . The data were band-pass filtered between 1 Hz and 100 Hz which is within a reasonable 177 range 37 , and further downsampled to 250 Hz before undergoing automatic cleaning, which consisted of 178 automatically detecting and excluding the most deviant data. Specifically, the data were segmented in epochs of 179 1000 ms in each dataset where the rejection rate of 18% was based on the mean of epochs and channel 180 heterogeneity. Hereafter, channels with more than five standard deviations from the joint probability of the 181 electrodes were removed and interpolated, then all datasets were re-referenced to an average reference. Adaptive 182 mixture independent component analysis 38 was computed on the remaining ranks. This resulted in matrices of 183 ICA spheres and weights. All ICs were associated with an equivalent dipole model as computed by DIPFIT 184 routines 39 using a boundary element head model based on the MNI brain (Montreal Neurological Institute,  185 MNI, Montreal, QC, Canada). 186 Epochs used to analyse the LightsOn data epochs were time-locked to the Lights-On event (S1) from -500 ms 187 to 4000 ms after the stimulus onset for each factor. Since the imperative stimulus (S2) was presented after 5000 188 ms, the selected epochs did not contain any brain activity of the S2-onset. Equivalently, epochs used to analyse 189 the Threshold (when participants passed the transition) were time-locked to the Threshold event (S3) from -4000 ms 190 to 500 ms and thereby describing the brain activity before the event of passing the transition. Approximately 191 17% of all epochs for the Lights-On event while approximately 21% for the Threshold event were automatically 192 rejected since they deviated more than five standard deviations from the joint probability and distribution of the 193 activity of all electrodes. 194 Event-Related Spectral Perturbations (ERSPs) were computed using the newtimef() function in EEGLAB. 195 Frequency range from 3 Hz to 100 Hz in log-scale, using wavelet transformation with 2.6 cycles for low 196 frequency and 0.5 cycles for higher frequency. The baseline for the cluster precomputations was defined as -200 197 ms to 0 ms. 198 The group-level analysis was computed using all ICs with less than 75% in residual variance of their 199 equivalent dipole model. These ICs, which reflect instantaneous independent time source information, were 200 clustered based on their equivalent dipole locations (weighted = 10), grand-average ERSPs (weight = 5), grand-201 average ERPs (weight = 1), mean log spectra (weight = 1), and scalp topography (weight = 3) and a region of 202 interest (ROI) located in the occipital cortex. The clustering was driven by a repetitive k-means clustering 203 approach (Gramann et al., 2018) with 5000 repetitions to ensure replicability. The number of clusters was 204 determined by ICs per participant so that the total number of ICs (301 ICs) and the total number of participants 205 (19 datasets) yielded 16 clusters. The approach was divided into three steps. First, given our prior results 3 , the 206 occipital area was defined as ROI with the Talairach coordinates (x = -20, y = -90, z = 7). Second, since each 207 clustering repetition yields a solution, the cluster of interest was selected based on (i) the number of participants 208 with an IC in the cluster, (ii) the ratio of ICs/participant, (iii) the spread (average squared distance) of the cluster 209 centroid, (iv) the mean residual variance of the fitted dipoles, (v) the distance of the x-y-z coordinate of the 210 cluster centroid from the ROI, and (vi) the Mahalanobis distance of the cluster of interest from the median of 211 the total 5000 solutions. These quality measures (i = 4, ii = -3, iii = -1, iv = -2, v = -3, vi = -1) allowed to 212 optimize the clustering solution close to the ROI. Third, the solutions were ranked based on the summed score, 213 where the highest-ranked solution was chosen as the final clustering solution. 214 Permutation tests (1,000 permutations) using EEGLAB stats were first computed to indicate differences 215 across factor levels. These are visualized in Figure 5  included the occipital, motor, and parietal areas, namely Cluster 3, 6, 9, and 11 (see Figure 3). Therefore, we 248 selected these clusters for further analysis. In locating the origin of the clusters, the mean of the included ICs was 249 calculated and projected onto the MNI head-model (see Table 1). The cluster representing the cingulate area 250 (Cluster 3) was estimated to originate from BA23, which corresponds to the posterior cingulate cortex (PCC). 251 Further, we were able to identify two clusters in each hemisphere in the temporo-occipital region (Cluster 6 and 252 11), which were located to originate from BA19, corresponding to a location near the intersection of the occipital 253 extrastriate areas and the bilateral parahippocampal regions (PHC). A single cluster within the supplementary 254 cortex (SMA) was identified (Cluster 9) (see Figure 3). Given the limited spatial resolution of EEG as a 255 neuroimaging method, we interpret the estimated location of the clusters with care. We take them as suggestive 256 rather than absolute. Using the Talairach-coordinates of the mean of each cluster, we labelled the clusters to the 257 nearest grey matter Talairach    in the PCC-cluster (Cluster 3) and the premotor cluster (Cluster 9) reflected the affordances of the environment. 292 The ERSPs of the clusters displayed clear alpha ERD (all p-values are reported in Figure 6). These results 293 suggest that during the mobile session of the experiment, the PCC-cluster and premotor-cluster dynamically 294 reflected the affordances of the environment via alpha ERD (see Supplementary Figure 2 for all ERSPs in this 295 event). For an overview of the relevant clusters in each phase, see Figure 4. 296  With an action-selection approach to attention, we questioned how sensorimotor attentional processes are 313 influenced by architectural affordances. Computing time-frequency analysis and clustering the ICs of EEG data 314 in participants moving through architectural spaces, we found that during the immobile LightsOn phase, the 315 bilateral PHC covaried with the affordances of the space-however, while approaching the Threshold between the 316 two rooms, we found that the premotor area and PCC reflected the affordances. The deep structures in parieto-317 and temporo-occipital regions exhibited a significant involvement, which was reflected in the attenuation of the 318 alpha rhythm in all regions. From our clustering solution, we can differentiate between alpha originating from 319 the PCC, bilateral PHC, and premotor cortex. Post-hoc statistical analysis revealed that exclusively the Narrow 320 transition was processed differently from the other conditions over the parieto-and temporo-occipital sites. The 321 action-selection process involved with architectural affordances, according to these results, are resolved via 322 parieto-and temporo-occipital alpha that involves a context-sensitive network, which encompasses the PCC, 323 possibly RSC, premotor area and bilateral PHC. We discuss these results in light of the anatomical structures 324 involved in alpha-band attenuation and conclude with a discussion of the behavioural corollary in terms of 325 embodied predictions. 326

Behavioural aspect 327
Attention as a balance? Conservative sensorimotor views cast action as re-actions to stimuli so that action is 328 guided by the stimulus akin behaviourism 42 . Such an approach favours physical causes above mental causes. 329 However, the ideomotor view stresses goal-directed actions that originate from internal (volatile) causes of 330 action. In this sense, goal-representations, i.e. functional anticipations of actions, play a central role in the 331 emergence of action. Goals and goal-representations are prioritized above stimuli and responses in the 332 generation of action in this view. Nonetheless, in the current study, we conceive action as both dependent on 333 internal and external causes because we take actions to be structures that link movement to goals, and goals to 334 movements 43 . Such an ecological approach stresses the interaction between the action capabilities of the 335 environment and the physical structure of the body 1 . It requires parallel internal and external attention as 336 affordances are neither an objective fact nor a subjective feature of our existence. Therefore, since no participant 337 chose to disobey the instructions, this experimental paradigm encompasses a window into the action-selection 338 process (i) upon perceiving the type of transitions (LightsOn) and (ii) during the unfolding of the selected action 339 that in turn causes a changing environment (Go-Threshold). 340 In line with the aforementioned critique of 'attention' 30,31,44 , the folk-psychological concept is disregarded, 341 and instead, a biological and phylogenetic understanding is employed. Despite the many studies referring to the 342 parieto-occipital area as the key-node in the attention-network 45-47 , we stress the role of the parieto-and 343 temporo-occipital area in embodied decision-making processes related to environmental (external) changes 44,48 . 344 Given an ecological and predictive coding approach, our view of 'attention' stems rather from the motor 345 planning process directing the sensory gating function. 346 Embodied predictions. An active inference account of action and perception complies with the ecological 347 foundation of affordances 49 . In the framework of the free energy principle, the corollary, active inference, cast 348 action and perception as serving the same purpose, namely to minimize our uncertainty about the environment 349 by minimizing free energy 20 . Central to sensorimotor dynamics, active inference conceives action as the 350 fulfilment of predictions based on inferred states of the ecological environment. The world is present in all its 351 details because the necessary set of actions to bring up a specific perception or information is known and 352 immediately obtainable 21,50,51 . This essentially means that behaviour can be framed as optimal predictions based 353 on incoming sensory evidence, where the 'optimal' is biased by prior preferences or goals. In this sense, 354 embodied predictions 52 rests on consequences of sensorimotor contingencies on proprioceptive, interoceptive, 355 and exteroceptive levels where the variable to be controlled resides in the body whereas the controllable states 356 are sensorimotor-related 53 . By conceiving architectural affordances as embodied predictions, i.e. predictions 357 based on our physical structure, motor signals can be cast under the same neural mechanisms as visual signals.

358
The necessary 'attention' in sensory-gating can thus be understood as motor-related attunement where the 359 'attention', i.e. feedback or prediction-errors, reflects a behaviourally relevant selection process from multi-level 360 sensorimotor dynamics. In our case, we observe a strong modulation of the alpha-band as originating from the 361 parieto-and temporo-occipital area. 362

Electrophysiological and anatomical aspect 363
Anatomy of thalamocortical alpha. Several studies have shown the visual thalamus to be crucial in the 364 generation and modulation of posterior thalamocortical alpha 54-61 . Neurophysiologically, the alpha rhythm is 365 viewed as an active inhibitor mechanism gating sensory information relative to perception 9,24,62 . The thalamus, 366 which holds a large collection of relay neurons, is the only source of information for the neocortex about the 367 body and the environment (safe olfaction). It is connected to the cingulum, which in addition to interconnecting 368 the major regions of the brain also serves as a tract interconnecting with the thalamus, e.g. PCC and retrosplenial 369 complex (RSC). 63 . It is worth noting that thalamic-cingulate projections posterior to the splenium divide to form 370 separate fascicles in RSC and PHC and further forms a principal route for RSC and PCC projections to the PHC 371 64 . However, feedback connections returning to the thalamus, i.e. cortico-thalamic paths, outnumber the outward 372 projections by 5 to 10 fold 65-67 . Since these inputs relayed through the thalamus arrive via axons with targets 373 often being subcortical motor centres, it raises the possibility that thalamocortical projections also serve as 374 efference copies 68 . Therefore, the embodied predictions may already be directing the gating function of the 375 thalamus as the sensory information is dynamically collected and processed. This could explain the observed 376 alpha behaviour in the PCC while the participants were approaching the door. 377 The PCC has been shown to have strong functional connections to many other regions in the brain 69 , and to 378 be one of the most active regions during rest and during task-related challenges 70,71 . The cytoarchitectural 379 structure reveals that it is organized to process perceptual input relative to the limbic and hypothalamus regions, 380 which suggests an important role in the internal and external regulations 72,73 . Accordingly, the PCC has been 381 ascribed to the capability of controlling the balance between the internal 'attention', e.g. recalling 382 autobiographical memories, and external 'attention', e.g. environmental changes. Particularly the PCC and RSC 383 have been suggested as key areas in this balance 74,75 . Given that studies using nonhuman primates have 384 demonstrated the importance and responsiveness of the PCC in environmental changes 48 and the alterations of 385 behaviour 76 , the balancing of internal and external attention may be a plausible explanation for the involvement 386 of the PCC in this study. This is consistent with the 'affordance competition hypothesis', which suggests that as 387 the environment changes, the affordances change along, which is then manifested in dynamic embodied 388 predictions that propel the body effortlessly through the environment 77,78 . 389 However, with the low spatial resolution of the EEG, we speculate whether the origin could be in an adjacent 390 and closely related area. For instance, Bonner and Epstein 79 , in a functional neuroimaging study, found that a 391 scene-selective region of the visual cortex, labelled the occipital place area (OPA), could be used to predict the 392 affordances of scenes. Studies in nonhuman primates propose that the nature of the connectivity of the cortex in 393 the middle to dorsal levels of the parieto-occipital sulcus, specifically the V6/V6A areas, to be involved in visual 394 guidance of movement 80,81 . The involvement of the visuomotor area V6 has further shown to receive its primary 395 afferent from thalamic nuclei providing visual and somatic inputs 82 and further suggested to be heavily engaged 396 in sensorimotor integration 80 . As the PCC-cluster is exclusively observed to covary with the affordances during 397 movement, the V6/OPA area cannot be excluded. Therefore, the V6/OPA also qualifies as a source for the 398 observed variation-nonetheless, both the V6/OPA and PCC are theorized to interact directly with the 399 thalamus.Modulation of PCC alpha. We suspect the observed posterior alpha during LightsOn in the PCC-400 cluster to be of cortico-thalamic feedback origin. Interestingly, the PCC-cluster showed significant differences 401 across affordances only while performing the task of acting in the environment (Threshold), which is arguably 402 ideomotor-dominant, as opposed to the anticipative interval (LightsOn) that is sensorimotor-dominant. We 403 speculate whether this difference in the type of task reflected in the PCC-alpha reveals the nature of affordances 404 as the balancing between internal and external 'attention'. Altogether, with the central role of the thalamus in 405 sensorimotor integration and the PCC in the balancing of internal and external 'attention' in mind, we propose 406 that the alpha-band attenuations can be interpreted as an indirect measure of sensorimotor demand, i.e. the 407 demand of neuronal processing for estimating the environment and own state by integrating sensory and motor 408 information. The motor information may then be adjusted by the cortico-thalamic embodied predictions in the 409 thalamus and thereby guiding the gating function 68 . This means that the subsequent cortical response is 410 modulated already at the entry to the cerebral cortex. Indeed, if the source of the alpha rhythm during movement 411 originates from the PCC, we would expect to find alpha ERD to covary with the sensorimotor demand. For that 412 reason, the posterior alpha serves as an excellent marker of dynamic action-selection processes relative to the 413 affordances as actions unfold. 414 Modulation of premotor alpha. Although we observe ERD in the alpha-band over premotor cortex after 415 the LightsOn event, the differences during this phase did not reach significance. The presence of clear alpha ERD 416 suggests activity over the premotor cortex, however, this activity is not modulated by the affordances of the 417 transition but suggests rather that one is in general preparation. Instead, during the Threshold phase, the activity 418 seems to reflect the affordances of the door suggesting that the premotor cortex is highly involved in the 419 continuous thalamic action-perception feedback loop suggested above. A previous study shows stronger alpha 420 desynchronization as emerging from BA6 during visually guided reaching, which is equally observed in our 421 recordings 10 . In summary, the activity of the premotor cortex is not modulated by the affordances during 422 preparation but rather involved during the active phase. Interestingly, we identified the same areas-however, our interpretation stresses the role of the thalamocortical-429 corticothalamic interactions, particularly due to the recorded alpha-band ERD. Crucially, we observe condition-430 specific sustained alpha ERD exclusively during the passive LightsOn event in the PHC areas. In line with the 431 existing theories, we interpret the activity as the processing of contextual information-however, we suggest that 432 since the available sensory information provided by the thalamus is affordance-sensitive, the PHC activity 433 reflects the associated behaviours as possible in the environment rather than place-or location-specific 434 information processing. Since the contextual information is continuously and correctly predicted as the 435 participants start acting, the PHC area shows no condition-specific differences as opposed to the PCC. 436 According to the thalamic projections described above, we interpret the observed alpha-band activity in PHC to 437 stem from thalamic interactions through the cingulum and the thalamus-PCC-PHC link. 438

Conclusion 439
In the current study, it was asked how architectural affordances relate to the attentional mechanisms 440 underlying the gating function for sensory signals both upon perceiving the environment and during the 441 interaction with the environment, are induced. Assuming that the magnitude of the alpha oscillations reflects the 442 impact on information processing in the brain, the results suggest the parieto-and temporo-occipital alpha ERD 443 serves as a marker for sensorimotor integration during interaction with the environment. For instance, while 444 approaching staircases, passing through doors, and turning corners are all examples of situations that require 445 dynamic processing of architectural affordances. Once the corner is turned, the processing of the new 446 environment, our results suggest, is reflected in the medio-temporal alpha ERD. We suggest the PHC area to be 447 involved in the immediate processing of affordances. Particularly the interaction between the PCC, possibly the 448 RSC, and the bilateral PHCs via cortico-thalamic alpha ERD suggests an action-perception mechanism sensitive 449 to the architectural affordances. Interestingly, it is the same pattern that emerges here as in the study by Djebbara 450 et al. 3 . Additionally, by conceiving the PCC as a transmodal thalamocortical hub, the PCC becomes crucial to 451 the balance between internal and external 'attention', i.e. the breadth of internal and external selection of actions. 452 Therefore, we interpret the visual sensory processing as inherently biased by an understanding of the body by 453 internal attention and the environment by external attention. 454 Indeed, moving in space is to continuously construct a prediction of a world that we perceive as dependent 455 on our action potentials that in turn is manifested in cortical oscillations. This suggests that users of space hold a 456 principle of anticipation that architects should keep in mind when designing the context for actions. In this 457 sense, by designing our environments, architects design cortical activity. However, we still have a long road 458 ahead to decently understand how architectural affordances impact various levels of brain dynamics. 459 Further research. Provided the thalamocortical origin in both the parieto-occipital and temporal regions, we 460 speculate how architectural design may implicitly, but constantly, influence subcortical structures that in turn 461 project to numerous other regions in the brain. A critical factor in the analysed data is the brain activity of 462 behaving human beings. Indeed, using architectural design as a medium for investigating action-selection and 463 sensorimotor integration is an exceptional approach, which calls upon more experimentation. Here, the MoBI-464 approach proves an excellent way forward. Assisting such experimentations, particularly in articulating the type 465 of involvement of the known subcortical networks, we propose the use of dynamic causal modelling 89 to allow 466 for multiple models to compete and infer which of the models best explain the acquired data. 467