1.1 Background and literature review

Affordances, the possibilities for action that the environment offers to a given organism or agent (Gibson, 1979), play a crucial role in guiding an agent's behaviour. In the field of affordance perception, there has been debate regarding whether it is an automated process or not. Some researchers argue that affordance perception is an automatic process because of its fast and effortless nature (Bonner & Epstein, 2017; Goslin et al., 2012; Harel et al., 2022; Tucker & Ellis, 1998), whereas others suggest that it is not automated but rather highly contextualised and can be influenced by biases and expectations (Girardi et al., 2010; Kalénine et al., 2016; Mustile et al., 2021; Pellicano et al., 2010; Tipper et al., 2006; Wokke et al., 2016). However, a synthesis perspective proposes that affordance automaticity should be understood as a dynamic process that changes over time, whereby affordance perception may initially occur automatically but is later modulated by higher-level cognitive processes (Borghi & Riggio, 2015; Djebbara et al., 2022; Gastelum, 2020; Kourtis et al., 2018). The question of whether affordance perception is an automated process or not may thus depend on the specific context and temporal scale being considered. The present study used brain dynamic features from electroencephalography (EEG) to provide evidence for the hypothesis that affordance perception is a time-varying process.

The question of affordance automaticity is related to the top-down/bottom-up debate regarding affordance processing (Pezzulo & Cisek, 2016) with several findings being broadly consistent with an automatic bottom-up view indicating effortless processing of affordances. Using a stimulus–response compatibility (SRC) paradigm, Tucker and Ellis (1998) required participants to decide as fast as possible whether objects displayed on a computer monitor were upright or inverted. In their experiment, objects were presented in two horizontal orientations (compatible with either right-hand grasp or left-hand grasp) and two vertical orientations (upright or reverse). The authors found that participants' responses towards upright or inverted objects were faster when the object's handle was on the same side of the display as the response hand than when the handle was on the opposite side, although the handle of the object was irrelevant to the judgement about the objects' vertical orientation. The results indicated that perceived affordances to grasp an object are activated automatically, even in the absence of explicit intentions to act (Tucker & Ellis, 1998). Using functional magnetic resonance imaging (fMRI), Bonner and Epstein (2017) aimed to identify representations of navigational affordances in the human visual system. They found that affordance information was automatically elicited during scene perception, because affordance coding in scene-selective brain regions was observed although participants performed a perceptual-semantic recognition task in which they were not explicitly asked about the navigational affordances of the scene (Bonner & Epstein, 2017). Using Bonner and Epstein's (2017) paradigm combined with EEG, Harel et al. (2022) investigated how rapid visual information is transformed into navigationally relevant information by measuring visually evoked event-related potentials (ERPs) with a focus on the P2 component, a positive component peaking around 220 ms after stimulus onset. The authors presented participants with computer-generated room scenes featuring varying numbers of doors or paintings to manipulate the navigational affordance of the local environment. These scenes, which served as the background stimuli, were presented on the screen for 500 ms. The participants were asked to report whether the horizontal or vertical bar of the central fixation cross lengthened on each trial by a factor of 25% minimising participants' attention to the background scenes. Although the participants in the study were not required to physically move or imagine themselves moving, the authors observed that changes in navigational affordance modulated early ERPs, specifically the P2 amplitude. Harel et al. (2022) concluded that information about the potential for navigation in the scene is extracted rapidly and automatically even when participants pay no attention to the scene (Harel et al., 2022).

Other studies have challenged the view that affordances are automatically activated, based on findings that perception of object affordances is modulated by task and context (Mazzuca et al., 2021). Context, in this case, refers to the scenario in which the object is perceived, including the presence of other objects or situations in which objects are embedded (Girardi et al., 2010; Mazzuca et al., 2021; Mustile et al., 2021). For instance, studies by Pellicano et al. (2010) and Tipper et al. (2006) demonstrated that the affordance effects to grasp objects as proposed by Tucker and Ellis (1998) only emerge when the task requires deeper processing of the object's characteristics, such as object categorisation and shape recognition, but not in simpler perceptual tasks like colour discrimination (Pellicano et al., 2010; Tipper et al., 2006). In a behavioural study, Girardi et al. (2010) aimed to study object affordance effects in natural reach-to-grasp actions while focussing on the role of the action context in the processing of object affordances. They presented the participants with a priming picture consisting of a hand and an object, and required them to make a power or precision grasp to classify the object as manmade or animate. Results showed that when the size of the object was consistent with the required grasping action (e.g. precision grasp to classify a needle as manmade), the response time was faster. Notably, the presence of this priming effect depended on the spatial arrangement of the hand and object in the picture. Specifically, the effect disappeared when the hand and object were presented as separate entities rather than an interacting pattern. Based on their findings, Girardi et al. (2010) argued that the observed affordance effect strongly depends on the action context and is not automatically activated (Girardi et al., 2010). In a recent study, Mustile and colleagues (Mustile et al., 2021) investigated the sensorimotor brain dynamics of dangerous object perception and whether the object's location and task goal modulated the encoding of its motor properties. The participants were recorded with high-density EEG while passively perceiving dangerous and neutral objects located at different distances from the observer and performing either a reachability judgement task or a discrimination-categorisation task. The authors found that a frontal N2 potential, a component of the ERP associated with motor inhibition, was larger for dangerous objects only during the reachability judgement task, indicating task-dependent and context-dependent aversive affordances. Taken together, these studies provided evidence that affordances might not be activated automatically, because the way we look at and operate objects around us might differ depending on the meaning of that object and the context they are placed in.

Recently, a synthesis of the automated and activity-dependent views on affordance processing has emerged. According to this view, affordance perception is explained with respect to different timescales or different processing types. According to Djebbara et al. (2022), the sensorimotor coupling process (or sensorimotor dynamics) in enacting perception concerning affordances can be understood as changing from a macroscale to a microscale, referred to as sensorimotor contingencies (SMCs; the non-specific and ongoing resonances with the environment) and sensorimotor responses (SMRs; the particular and preferential resonances with the environment occurring in the early sensorimotor processes), respectively. In line with different stages of affordance processing, Kourtis et al. (2018) found that affordances can be separated into two categories based on the temporal mechanisms of the agent-object interaction, immediate (i.e. grasping showing size-related affordances), and long term (i.e. skillful use showing function-related affordances) with different neural representations of these two kinds of affordances (Kourtis et al., 2018).

1.2 Research question and hypothesis

The present study analysed human brain dynamics to investigate whether affordance processing differs in automaticity over time and depending on the context. More precisely, we analysed early and later components of evoked brain dynamics reflecting affordance processing of identical virtual spaces comparing two different movement conditions, physically walking versus joystick-controlled movements. We investigated whether different sensorimotor phases can dissociate different levels of automaticity in affordance processing dependent on the movement context. In accordance with existing research on affordance context (Girardi et al., 2010; Mustile et al., 2021), we operationally defined context as a property of the perceived scene that primes different sensorimotor coupling states. To this end, we varied the way participants could move through a virtual environment. Varying levels of movement were realised through different displays as a between-subject design, with a previous study Djebbara et al. (2019) using head mounted VR display with full-body movements through the virtual environments and the present study using a 2D display with a keyboard for movement control. While moving through the identical virtual environment, the two different movement contexts should prime distinct sensorimotor coupling states. We used a virtual environment from the study by Djebbara et al. (2019) in which participants actively walked through an immersive virtual environment with the task to pass differently wide doors (passable and impassable doors) to go from one room to an adjacent room. In contrast to the physical movements through 3D virtual rooms in the Djebbara study, the participants in our study used keyboard controls to move through the same rooms displayed on a 2D display. We statistically compared the ERP results from the Djebbara study with the results from the present study. Comparable early and late ERP components in both studies would support the assumption of automated affordance processing, irrespective of the movement (full body vs. keyboard). Differences in both, early and late ERP components, in contrast, would support the assumption of context-dependent affordance processing while comparable early but differences in late components would support a stage-dependent processing of affordances. According to Djebbara et al. (2019), the P1-N1 complex in their study reflects early visual evoked activity. Moreover, in order to interpret the temporal mechanism underlying early ERPs with different latencies, some literature regarding different ERP components of involuntary attention (Escera et al., 1998; Escera et al., 2001; Näätänen, 1990; Wang et al., 2010) and voluntary attention (Luck & Hillyard, 1994; Patel & Azzam, 2005; Schneider et al., 2012; Wang et al., 2010) will also be referred to in the discussion section. And following early ERPs, we referred to the research of Elbert et al. (1982) and Djebbara et al. (2019) to interpret the psychological meaning of motor-related ERPs.

2.1 Participants

Eighteen participants (seven females) without history of neurological pathologies were recruited from social media and a participant pool of the Technical University of Berlin, Germany. All participants signed the consent form, which was approved by the Ethics Committee of the Technical University of Berlin. After each data collection, the participants received either monetary compensation (10 €/hour) or accredited course hours. Their mean age was 24.0 years (σ = 3.8). All participants were right-handed, had normal or corrected-to-normal vision, and none had a specific background in architecture (i.e. no architects or architectural students). We excluded one male participant from the study because of experienced dizziness.

2.2 Experimental paradigm and procedure

We used the same experimental paradigm and stimuli as Djebbara et al. (2019, 2021) and repeated their experiment in a stationary desktop setup instead of physically walking through a virtual space with head mounted VR. The experiment was conducted in a dimly lit experimental room at the Berlin Mobile Brain/Body Imaging Laboratories (BeMoBIL) using a 27″, LED monitor (PROLITE XB2780HSU-B1 [EOL], full HD 1920 × 1080 resolution, horizontal sync 24–80 kHz and vertical sync 56–70 Hz) to display the virtual environment. The size of the virtual space was 9 m × 5 m (virtual meters), with room sizes of 4.5 m × 5 m for two adjacent rooms connected through a door (see Figure 1). The participants were seated in front of the monitor in a comfortable position, and their task was to move through the virtual environment using four arrow keys on the keyboard. The doors connecting the two rooms differed in width from wide (1.5 m, easily passable) to mid (1 m, passable) to narrow (.2 m, unpassable) in a pseudorandomised order. The door width varied to manipulate the transition affordance between rooms. The participants performed a forewarned (S1-S2) paradigm with two different movement instructions (pseudorandomised 50/50) indicating to walk through the door (‘Go’ trial) or not to walk through the door (‘NoGo’ trial). In Go trials, the participants were required to walk from one room to a second room and to pick up a token in half of the trials, whereas in the other half of the trials (NoGo), they simply judged the emotional experience of the first room they were located in without moving to the second room. To be clear, however, despite these labels, our experiment does not follow the traditional Go–NoGo task paradigm in which subjects should respond fast to frequently presented ‘Go’ targets while ignoring rare ‘NoGo’ stimuli (Aron & Poldrack, 2005). Instead, we aimed to investigate whether affordance-related neural activity depended on the condition of knowing whether he or she would need to pass through the door (Go signals) or not (NoGo signals). We thus use the term Go trials or NoGo trials to indicate the imperative stimulus in the respective trials, following the protocol from Djebbara et al. (2019).

The experiment was thus a 2 × 3 repeated-measures design including the factors imperative stimulus (Go, NoGo) and door width (narrow, mid, wide). For each participant, there were 240 trials in total, with 40 trials for each condition. In each trial, the participants started in a dark environment on a predefined starting square (Figure 1). After a random interval (mean = 3 s, SD = 1 s), the ‘lights’ were turned on, and the participants viewed a room with a closed door. After waiting for a random duration (mean = 6 s, σ = 1 s), the door colour changed into either green indicating a Go trial or red indicating a NoGo trial. If the door colour was green, the participants were instructed to move toward the door by pressing the arrow keys. Once they reached the door, the door slid open aside immediately. The participants were instructed to find and virtually touch a red rotating ring in the second room. Touching each ring earned participants 1 cent as a bonus to the hourly rate, which was also our incentive strategy. After the participants went back to the first room, they were instructed to return to the starting square and to provide an emotional rating using the self-assessment manikin (SAM) questionnaire. However, in the case of a red door, the participants did not walk through the transition but directly provided an emotional rating for the room they were located in. After the emotional rating, the participants pressed the SPACE bar to start the next trial.

In each Go trial, the participants were instructed to move through the door into the second room irrespective of whether the door was passable or not passable. This setting kept motor processes comparable for all Go trials irrespective of the transition affordance (passable or not passable). When the participants touched the surrounding walls, the walls turned red, indicating that they failed to pass and the red rotating ring located at the second room would disappear indicating that the participants did not get a bonus for this trial. The participants first finished a training session to become accustomed to the experimental environment and the different conditions. The experimenter communicated with the participants from a control room, separated from the experimental space, using a camera and a microphone–speaker system. Moreover, there were five breaks in total, including two mandatory breaks (2 min at least) and three optional breaks.

In order to investigate the potential impact of movement on affordance perception, we compared the ERP results with the findings of Djebbara et al. (2019). This is allowed for a direct comparison of the same visually presented transitional affordances in two different movements (physical walking vs. keyboard movement). In addition, we administered the Igroup Presence Questionnaire (IPQ) before and after the experiment to assess the participants' sense of presence (Slater & Sanchez-Vives, 2016). The IPQ consists of four dimensions: general presence (PRES), spatial presence (SP), involvement (INV) and experienced realism (REAL) (Grassini & Laumann, 2020; Schubert et al., 2001). The data were analysed using paired-samples t tests with the factor time with two levels (pre- and post-test) as a within-subjects factor.

2.3 Self-assessment manikin

To investigate the subjective experience of the task, we asked the participants to finish a virtual SAM questionnaire after each trial. The SAM is a pictorial assessment of pleasure, arousal and dominance on a 5-point Likert scale (Bradley & Lang, 1994). The participants were asked to self-assess their current state after each trial by pressing the respective keyboard number on the 5-point scale. A first key press selected the participants' answers but still allowed changes to the selection. By pressing the same key again, the response was confirmed. The SAM data were analysed using a 2 × 3 factorial repeated-measures analysis of variance (ANOVA) with the factors imperative stimulus (Go and NoGo) and door width (narrow, mid and wide). Post hoc analyses were computed using the least significant difference (LSD) test to correct for pairwise comparisons.

2.4 EEG recording and data analysis

Experimental stimuli based on Unity software were displayed on a standard monitor. All data streams, including experimental events such as participants' movement and location in the Unity environment, stimulus markers and key presses, SAM answers and EEG data, were recorded and synchronised using Lab Streaming Layer (LSL; Kothe et al., 2014).

EEG data were recorded with a 64-channel EEG system (ANT Neuro), sampled at 500 Hz and referenced to CPz. Impedances were kept below 10 kΩ. Data analysis was conducted using MATLAB (MathWorks), the EEGLAB toolbox (Delorme & Makeig, 2004) and custom scripts. Before data preprocessing, all training and resting data sessions were excluded. All preprocessing steps were based on the BeMoBIL pipeline (Klug et al., 2022), first down-sampling the data to 250 Hz and then applying the ZapLine-plus function to remove spectral peaks at 50 Hz, corresponding to the power line frequency (Klug & Kloosterman, 2022). Next, bad channels were detected and interpolated in a repeated process using the clean raw data plugin of EEGLAB. To ensure a reproducible bad channel detection, we ran the clean_artifacts function, with a recommended minimum of 10 iterations. For the automatic channel cleaning, the data were split into time windows of 5 s and robust interpolations of each channel were computed. Channels with a correlation to a random sample of other channels were marked for removal in case the correlation was below .8 for more than 50% of the processed data and subsequently interpolated using a spherical spline interpolation. As a final step, the data were re-referenced to the average of all scalp channels, excluding the EOG channel. Subsequently, on each cleaned dataset, we computed an independent component analysis (ICA) using an adaptive mixture independent component analysis (AMICA) algorithm (Palmer et al., 2012) with the recommended parameter values from Klug and Gramann (2021). Before computing ICA, we performed high-pass filtering with a .75 Hz cutoff frequency. Then the AMICA decomposition was computed using one model and 10 iterations of data rejection with a rejection threshold of 2.9 standard deviations from the log-likelihood of the explained data during the decomposition. Afterwards, an equivalent current dipole model was computed using the DIPFIT toolbox of EEGLAB. Independent components (ICs) were automatically classified using ICLabel (Pion-Tonachini et al., 2019). Subsequently, we removed ICs reflecting eye-movement activity. The computed AMICA information including rejections and dipole fitting was copied back to the initial preprocessed, but unfiltered, dataset.

The continuous datasets were band-pass filtered between .2 and 40 Hz, and two sets of epochs were extracted with the first set for the onset of the first room (lights on) and the second set with onset of the imperative stimulus (Go/NoGo). All epochs were extracted from −1000 ms before to 1000 ms after stimulus onset for narrow, mid and wide door trials separately. Then, all epochs were baseline corrected using a −200 and 0 ms pre-stimulus interval and cleaned using the bemobil_reject_epochs function to rank epochs with respect to their noise level and removing a fixed percentage of the worst ranking epochs (Klug et al., 2022). Each epoch was evaluated using four measures that were normalised by their median across epochs: (i) mean of channel means (weight = 1), (ii) mean of channel SDs (weight = 1), (iii) SDs of channel means (weight = 1) and (iv) the SD of channel SDs (weight = 1). Each epoch then received a final summed score, and the epochs were sorted according to that score. Finally, up to 10% of epochs were removed (mean = 23.62, SD = .697) accordingly.

Early perceptual components of the ERPs time-locked to the onset of the virtual room (lights on) were analysed at midline posterior leads (Pz, POz and Oz) comparing ERPs from Djebbara et al. (2019) with the current results. ERPs time-locked to the subsequent onset of the imperative stimulus (Go/NoGo) were analysed at the same leads with the inclusion of one additional anterior electrodes (Fz, FCz and Cz) to further allow for a direct statistical comparison of the motor-related ERPs of Djebbara et al. (2019) and the corresponding data of the current study. The results of the anterior leads focus on FCz as described in Djebbara et al. (2019) (additional results for the other two anterior electrodes Fz and Cz are provided in the supporting information, as seen in Figures S1–S8). These electrodes locations cover brain regions involved in processing visual information and regulating motor actions as used in previous studies (Bozzacchi et al., 2012; Bozzacchi et al., 2015; Djebbara et al., 2019; Harel et al., 2022; Mustile et al., 2021). We identified three peaks (N120, P164 and N260) after the ‘lights on’ stimulus onset through visual inspection at the grand average level. Individual peak latencies were determined considering both EEG polarity and a time range of −50 to +50 ms around the group-level peak latencies. Therefore, the search windows for individual peaks ranged from 70 to 170, 114 to 214 and 210 to 310 ms for the N1, P1 and N2 peaks, respectively. To ensure robust statistical analysis, each negative or positive peak was exported, computing the mean from the data covering the peak plus/minus 3 sample points before and after the peak. The data of P1 and N2 were analysed using a 2 × 3 mixed-measures ANOVA with a between-subjects factor movement (physical walking vs. keyboard movement) and a within-subjects factor door width (narrow, mid and wide). Post hoc analyses were computed using the LSD test to correct for pairwise comparisons. Because the N1 component was an additional negative peak detected in the present study that was absent in the study of Djebbara et al. (2019), we conducted an additional one-way ANOVA with the door width (narrow, mid and wide) as a repeated-measures factor and the amplitude of the N1 as a dependent measure. Post hoc analyses were computed using the LSD test to correct for pairwise comparisons.

For the later motor-related cortical potentials (MRCPs), we focused on the early post-imperative component (EPIC) and post-imperative negative variation (PINV) as reported in Djebbara et al. (2019). After the stimulus onset of the imperative (Go/NoGo) stimulus, a negative peak at posterior leads was observed around 236 ms and a positive peak at FCz electrode was observed around 280 ms, referred to as the EPIC (Djebbara et al., 2019). Search windows for individual peaks were ranging from 186 to 286 ms and 230 to 330 ms, respectively. For the sake of robust statistics, peak amplitudes were calculated including the peak plus/minus 3 sample points before and after this detected peak. The PINV can be considered a postimperative negative variation reflecting the immediate motor execution related to onset of an imperative stimulus (Casement et al., 2008; Djebbara et al., 2019; Elbert et al., 1982). As in the previous study of Djebbara et al. (2019), the PINV component was calculated as the mean amplitude in the range of 600–800 ms after stimulus onset (imperative stimulus: Go/NoGo). The data of EPIC and PINV were both analysed using a 2 × 2 × 3 mixed-measures ANOVA with a between-subjects factor movement (physical walking and keyboard movement), two within-subjects factors imperative stimulus (Go and NoGo) and door width (narrow, mid and wide). Post hoc analyses were computed using the LSD test to correct for pairwise comparisons. In the present paper, we described the p value lower than .05 as a significant result and the p value between .05 and .1 as a tendency or marginally significant result because of a relatively smaller sample.

3.1 IPQ data

In this study, the IPQ questionnaire was used before and after the main experiment in order to investigate the extent of participants' sense of presence. A paired sample t test for the pre- and post-test data for each dimension of the IPQ revealed significant differences in general presence (PRES) (T₁₆ = −5.191, P < .001, Cohen's D = −1.259), spatial presence (SP) (T₁₆ = −6.014, P < .001, Cohen's D = −1.459), involvement (INV) (T₁₆ = −2.507, P = .023, Cohen's D = −.608) and experienced realism (REAL) (T₁₆ = −3.228, P = .005, Cohen's D = −.783) (see Figure 2).

3.2 Subjective data: SAM ratings

The SAM questionnaire was completed for both Go and NoGo trials in all door width conditions. A 2 × 3 factorial repeated-measures ANOVA was used with the factor imperative stimulus (Go and NoGo) and the factor door width (narrow, mid and wide) for each emotional dimension of the SAM questionnaire (see Figure 3). For arousal, there was a significant main effect of the factor door width (F_2,32 = 9.628, P < .001, partial η² = .376) as well as a significant main effect of the factor imperative stimulus (F_1,16 = 53.061, P < .001, partial η² = .768). The interaction of both factors was significant (F_2,32 = 19.106, P < .001, partial η² = .544). Post hoc comparisons revealed that only the arousal score of medium-wide doors was significantly different from the arousal score of wide doors (P = .043) in the NoGo condition. However, in the Go condition, significant differences were identified for the comparison of narrow versus wide doors (P < .001), narrow versus mid doors (P < .001) and mid versus wide doors (P = .021). For dominance, there was a significant main effect of the factor door width (F_2,32 = 13.944, P < .001, partial η² = .466) as well as a significant main effect of the factor imperative stimulus (F_1,16 = 21.198, P < .001, partial η² = .570). Both factors interacted significantly (F_2,32 = 19.878, P < .001, partial η² = .554). Post hoc comparisons revealed no differences in dominance ratings between differently wide doors in the NoGo condition. However, in the Go condition, significantly lower ratings were observed for the narrow as compared to wide doors (P < .001), narrow as compared to mid doors (P < .001) and a tendency towards a difference for the comparison of mid versus wide doors (P = .061). These results were also found for the valence dimension of the SAM including significant main effects of the factor door width (F_2,32 = 33.737, P < .001, partial η² = .678) and imperative stimulus (F_1,16 = 46.657, P < .001, partial η² = .745) as well as their interaction (F_2,32 = 48.585, P < .001, partial η² = .752). The three door widths levels did not differ significantly under the NoGo condition. However, under the Go condition, significant differences were identified in narrow versus wide (P < .001), narrow versus mid (P < .001) and mid versus wide doors (P = .027).

3.3 EEG data: early ERPs

We hypothesised that ERP components in both studies should be comparable during early sensorimotor stages if affordance processing was an automated process, irrespective of the movement condition. To test this assumption, we analysed the early ERP components reflecting perceptual aspects of transition affordances from the present study and compared the results with the results reported by Djebbara et al. (2019). For ERPs with onset of the rooms including different door widths, three prominent peaks were observed over posterior leads. Over the occipital midline electrode, a first negative component peaked around 120 ms, followed by a positive peak around 164 ms and a negative peak around 260 ms (Figure 4). The first negative component was not observed in the study by Djebbara et al. (2019), and thus, no comparison between the movement conditions was possible. A statistical test for this N1 component was conducted only for the desktop condition, while the comparison of visual evoked potentials from Djebbara et al. and the present study was computed for the latter two components using a mixed-measures ANOVA with the between-subjects factor movement (physical walking vs. keyboard movement) and the within-subjects factor door width (narrow, mid and wide) (Figure 5). Before we conducted the ANOVA on separate electrodes, we included the electrode location as a factor in the ANOVA. The results revealed the factor electrode location to modulate statistical effects of the factor door width. For example, early ERPs revealed interactions between door width and electrode location (F_4,136 = 2.630, P = .037, partial η² = .072) for the P1 and N2 components (F_4,136 = 3.789, P = .006, partial η² = .100) at posterior electrodes. Therefore, our interpretation is that topographical differences point to the involvement of distinct processes that are implemented in spatially distributed neural structures which are recorded with the different electrodes as volume conducted signals. Based on the interaction effects involving the factor electrode location, we followed up by ANOVAs for single electrodes to explain in detail the differences for the ERPs.

3.4 Motor-related ERPs

Besides assuming comparable early ERPs to reflect automatic affordance perception, we further hypothesised that in the subsequent motor-related stage of sensorimotor coupling, the movement context might moderate late ERP components reflecting later movement context-dependent affordances. In order to test this assumption, we directly compared the motor-related potentials of transition affordances in the present study with the corresponding results of Djebbara et al. (2019). The EPIC and PINV were observed at central midline leads (FCz, Pz, POz and Oz) after the onset of the imperative stimulus, which instructed the participants to go or not to go through the door to the next room (Figure 6a, b). Both the posterior EPIC components and anterior EPIC (at electrode FCz) are displayed in Figure 7. Before we conducted the ANOVA on separate electrodes, we included the electrode location as a factor in the ANOVA. The results revealed the factor electrode location to modulate statistical effects of other factors. For example, first, for the EPIC component, there were an interaction between door width and electrode location (Pz, POz, Oz and FCz) (F_6,204 = 6.102, P < .001, partial η² = .152), and a marginally significant interaction between imperative and electrode location (Pz, POz, Oz and FCz) (F_3,102 = 2.501, P = .064, partial η² = .069). Second, besides the relevant interaction of the door widths and electrode location of the EPIC component, several two-way and three-way interactions were also observed when including the factor electrode location for the analyses of the PINV component. Therefore, with combining the relevant results of early ERPs and MRCPs, we conclude that topographical differences point to the involvement of distinct processes that are implemented in spatially distributed neural structures as reflected in different activation patterns in visual sensory and motor-related brain regions. Based on the interaction effects involving the factor electrode location, we followed up by ANOVAs for single electrodes to explain in detail the differences for the ERPs.

3.5 EPIC

3.6 PINV

Following EPIC, we noticed a long-lasting negative waveform over central midline electrodes after the onset of the imperative stimulus. As described in previous studies, this waveform represents the PINV, which reflected the immediate motor execution and action readiness related to onset of an imperative stimulus (Casement et al., 2008; Diener et al., 2009; Elbert et al., 1982; Klein et al., 1996). In the study of Djebbara et al. (2019), the PINV was regarded as reflecting architectural affordances (door width) or, as the authors phrased it, as the sensorimotor dynamics resonating to architectural affordances (Djebbara et al., 2019). To directly compare the PINV observed in the study of Djebbara et al. (2019) with the present study, the PINV component was calculated as the mean amplitude in the range of 600–800 ms after stimulus onset of the imperative stimulus. Both the anterior and posterior PINV components were reported here (Figure 8).

4.1 IPQ and SAM data

The analysis of the four dimensions of the IPQ (general presence, spatial presence, involvement and experienced realism) revealed a significant increase in the participants' sense of presence throughout the experiment, even though their movement through the virtual environment was displayed on a monitor and using keyboard controls. The results indicate that the virtual environment was sufficiently immersive to create a sense of movement through the space allowing for a comparison of the brain dynamics in the present setup with the corresponding data of Djebbara et al. (2019).

The analysis of the SAM ratings for arousal, dominance and valence consistently revealed significant differences in door width among Go trials but no differences across NoGo trials for all ratings, except the arousal dimension. Varying door sizes in Go trials consistently yielded differences among all three SAM items, with increasing values observed with increasing door widths. This result is consistent with the finding of Djebbara et al. (2019), except for the dominance dimension. The difference in the latter SAM scale might be the result of additional semantic instructions in the present study to clarify the participants' understanding for the dimension of dominance. In the present study, and in line with the original scales from Bradley and Lang (1994), we used the small SAM representing ‘being dominated’ whereas the large manikin represented ‘being dominating.’ Overall, these results reflected that our Go/NoGo setting differentially influenced the participants' emotional experience regarding the affordances in the Go and the NoGo conditions. They provided initial evidence from subjective data prior to comparing our motor-related ERPs results with the findings of Djebbara et al. (2019).

4.2 Cortical measures