To understand human interaction and locomotor navigation strategies, it is essential to explore the interrelationships that exist between the task, the environment, and the individual; this offers the ability to predict human motion in relation to the surroundings (Gibson, 1966). The Perception-Action (P-A) Theory explores the constraints between these three components for movement control, effective training, and skill acquisition. Past research in this area suggests that motor control strategies are optimized for individuals to locomote through changing environments, with vision playing a critical role in successful movements (Heft, 2001). Two constructs of the P-A theory that are thought to evolve together in mammals include visual perception, describing the interpretation of visual information by the brain, and action, which can be defined as something being done by the system (Gibson, 1966). These components, although independent constructs, are thought to be in a close interrelationship from birth, developing simultaneously as new visual experiences are acquired (e.g., Barbu-Roth et al., 2009; Barbu-Roth et al., 2014); this allows for adaptive motor strategies in changing environments (Warren, 2006).

A vital component of this interrelationship is optic flow, which can be defined as the pattern of motion of the visual environment caused by the movement between the observer and the visual scene (Warren et al., 2001). Individuals develop an understanding of self-motion while experiencing these flow patterns, as the visual system and associated brain areas capture information about the body position within the physical environment (Marigold, 2008). This visual information is subsequently decoded into flow patterns, which yield rotational and translational movements contributing to the perception of self-motion. Optic flow also provides critical information pertaining to motor control strategies such as speed or heading direction (e.g., Jörges et al., 2024; Warren & Hannon, 1988), depth perception (e.g., Liu et al., 2023), time to contact with obstacles (e.g., Lee & Aronson, 1974; Ramirez & López-Moliner, 2020), and environmental structure and layout (e.g., object placement and surface orientation, Le et al., 2018). Collectively, information captured in optical flow patterns can serve as a means for guiding locomotion. While other sensory systems (i.e., vestibular and somatosensory) also play a role in locomotor control, visual information and especially optic flow, is unique in providing feedforward control information to plan and guide movements within dynamic environments (Warren, 2006).

The effects of optic flow on movement in real-world environments have also been observed in virtually simulated scenes (e.g., Berthoz et al., 1979; Chou et al., 2009; Warren et al., 2001). Within these virtual environments, providing optical flow cues (i.e., textured motion, radial expansion and contraction, objects) can elicit self-motion perception in humans in the absence of movement within the physical environment. An additional similarity between real-world and virtual environments is the ability for optic flow to elicit central nervous system (CNS) compensatory responses in reaction to visual stimuli. For example, participants in Antley and Slater’s study (2010) saw an increase in preparatory postural muscle activation via electromyography (EMG) when exposed to scenes with increased visual perceptual demands. Throughout the literature, in both real-world and virtual environments, we see that changing visual environments with differing stimuli (i.e., visually rich [e.g., complex and detailed visual input, such as a forest trail] versus less visually rich environments [e.g., sparse or minimal visual stimuli, such as a white-walled room]) may pose varying roles in the perception of self-motion (e.g., Jörges & Harris, 2021). Neuroscientific evidence suggests that these differences may be underpinned by activity in visual-vestibular integration areas, such as the ventral intraparietal area and medial superior temporal area, which contribute to spatial orientation and process optic flow (Bremmer et al., 2002; Smith et al., 2012). Animal studies have further demonstrated that circuits essential for spatial navigation, including the hippocampal place and grid cells, are able to adapt dynamically to environmental salience and changes in complexity (Chen et al., 2019). Together, these neuroscientific findings support the notion that visual scene characteristics contribute to the modulation of neural processing related to self-motion, which may, in turn, reflect evolutionary adaptations made to help interpret environmental affordances. While important scientific findings have been reported in this area of study, the behavioural consequences of presenting visual cues with varying richness, particularly within postural control and locomotor patterns, remain underexplored.

Another important consideration of the influence of optic flow on movement control is the impact of the adopted base of support (BoS) while experiencing visual information. Previous research has shown that optic flow influences postural sway in both the mediolateral and anterior-posterior directions (Phu et al., 2023); it appears that larger adopted BoS (e.g., seated versus standing positions) results in improved postural stability (Nam et al., 2017). In situations that challenge our BoS (e.g., changing tasks or environmental conditions), visual information can provide crucial body position information that may enable the fine-tuning of postural control (Nam et al., 2017), i.e. a return to perceived ‘stability’. However, if visual information is presented in a way that is in conflict with other sensory information (i.e., proprioception), postural stability may become disturbed even in seemingly stable BoS positions (Wang et al., 2022). If an individual feels unstable or uncomfortable within their BoS, perhaps due to conflicting or reduced proprioceptive information, their postural sway may increase (Mochizuki et al., 2006); this further emphasizes the importance of visual flow in the generation of motor control strategies.

While P-A theory research has elucidated how visual perception and action exist in an interrelationship to control movement effectively, much of the support for this theory has been derived from studies assessing newborn (e.g., Barbu-Roth et al., 2009; Barbu-Roth et al., 2014; Forma et al., 2018) and infantile (e.g., Bertenthal et al., 1997; Lee & Aronson, 1974) stages of development. We have discussed some of the available literature surrounding the role of optic flow on movement into adulthood; however, the strength of this relationship has minimal support, especially on a neurological level. More specifically, very little is understood about the interplay between changing visual cue complexity and body position in space and whether BoS will influence how visual information is integrated to inform muscle activation patterns involved in maintaining postural control. Additionally, much of the literature explores the effects of optic flow on visuomotor control on aggregated male and female data (e.g., Chou et al., 2009; Wade et al., 1995; Yoo et al., 2014), overlooking potential sex-based differences in sensory processing and motor response. There is some evidence to suggest that males and females experience optical flow patterns differently, which could influence downstream motor outcomes (e.g., Raffi et al., 2014). Some differences between the sexes may stem from underlying variations in visual attention, spatial orientation strategies, or sensory weighting preferences. For example, female participants have been shown to rely more heavily on visual cues when executing visually guided movements (e.g., Gorbet et al., 2007; Barnett-Cowan et al., 2010) compared to their male equivalents. These findings underscore the importance of considering sex within optic flow studies, as visual scene manipulations may not exert uniform effects across the sexes.

This study aims to evaluate the changes in muscle activity and postural sway in the absence of overt movement when exposed to four complex video projected visuomotor environments (i.e., an indoor hallway, a boardwalk outdoors, outside on a hiking trail, and no video) in three different body postures (i.e., seated with feet planted, seated with feet dangling, and standing with feet shoulder-width apart) in both male and female participants. We have various hypotheses for the present study, dependent on the condition presented. First, we expected that exposure to all three visually complex environments, in any postural condition, would result in greater muscle activation and postural sway compared to the no video condition due to P-A coupling (i.e., more visual information to inform self-motion, Marigold, 2008). Additionally, we anticipated that the greatest muscle activity and postural sway would occur when standing and viewing the hiking trail video due to the small BoS paired with increased visual perception requirements in this environment (e.g., trees, branches and rocks), as well as the perceived need to prepare for navigation through a more complex pathway (e.g., Antley & Slater, 2010). Finally, we predict that the seated posture with feet dangling would result in greater muscle activation and increased postural sway compared to the seated posture with feet planted, due to reduced proprioceptive information from the lower limbs (Wang et al., 2022). Without reliable somatosensory input from stable ground contact, it is hypothesized that the postural response would be similar to that of an unstable surface (i.e., foam; Vuillerme & Pinsault, 2007), where the body increases neuromuscular effort to compensate, reflected in greater muscle activation and increased sway.

On the day of the in-person laboratory visit, and prior to the commencement of the protocol, participants were first provided verbal instructions of what they could anticipate throughout the duration of the study, and verbal informed consent was reconfirmed. Participants were then palpated for four muscle bellies of the right leg (i.e., gastrocnemius, tibialis anterior, vastus lateralis, and biceps femoris) using contraction and palpation procedures (Hermens et al., 1999). The skin above the muscle belly was then prepped by shaving and rubbing the area with isopropyl alcohol. Four EMG bipolar electrode sets (Ambu 32 mm Sensors, Kego, Ontario, CA) were then fitted to the identified areas (Fig. 1B). A ground electrode was placed over the right patella or right fibular head (dependent on the participant's physique) and taped to ensure the electrode was secure. EMG was measured via Bortec EMG System (1000 Hz, Bortec, Alberta, CA). Following EMG preparation, participants were then fitted with 27 retroreflective markers placed at anatomical landmarks following the OptiTrack Conventional Upper Body skeleton (Motive software; Version 2.2.0, Oregon, USA). An additional two markers were placed on the iliac crests (Fig. 1B). A surrounding 10-camera kinematic system (100 Hz; OptiTrack, Natural Point, Oregon, USA) tracked the movement of these markers. In addition to the above equipment, the experimental setup required a projection screen (spanned 3.30 m x 2.50 m, from ceiling to floor in the lab), an overhead projector (HY300 Pro, QRJ Eokeiy, China; connected via computer), over-ear headphones (Sony WH-CH720N, Tokyo, Japan), and a height-adjustable stool with no backrest.

The protocol started by asking participants to stand in front of and face the white projector screen. From this position, participants completed nine quiet trials of 30 seconds each; this included three trials in each of the three body postures (9 trials x 30 seconds, three trials per posture). For the standing posture, participants were told to stand with their feet shoulder-width apart and hands by their sides. For the seated with feet dangling posture, a stool was adjusted to a height where participants' feet could not reach the ground; they were told to sit upright, with hands resting on the side of the chair, and not to touch the ground with their feet during the trials. In the seated with feet planted position, participants were told to sit upright, with hands resting on the side of the stool and feet firmly planted on the ground. During quiet trials, participants were instructed to remain stationary and look straight ahead; this allowed for baseline muscle activity and postural sway to be captured. Once quiet trials were collected, maximal head range of motion (ROM) trials were collected in the standing posture, which involved participants completing yaw and pitch head movements within one trial (approximately one minute). This served as a baseline measure for head ROM.

To begin the experimental trials, participants were placed again in front of the screen (approximately 1.42 m from centre) and were surrounded by black curtains. All overhead lighting was off to minimize distractions from the surrounding laboratory environment. Participants had a total visual angle of approximately 90° in front of the screen. Headphones were placed over the ears and ambient sound (appropriate for the visual environment being displayed, e.g., ventilation system low hum in hallway; bird song on hiking trail) was played to reduce laboratory background noises and create a more immersive sensation. Each trial consisted of a one-minute video that would stream in front of the participant via the computer-based projection. The videos were pre-recorded on an iPhone (iPhone 15 Pro, Apple, California, USA) through first-person perspective in three environments; this included an indoor hallway of a university building, an outdoor boardwalk, and an outside hiking trail through a conservation area (Fig. 1C). This was meant to simulate real-world environments within the virtual lab-setting.

A total of 45 experimental trials were collected. In the first 15 trials, individuals watched each type of video five times in the seated with feet planted posture. For the next 15 trials, individuals watched each type of video five times in the seated with feet dangling posture. For the final 15 trials, individuals watched each type of video five times in the standing posture. The postural order was block randomized between participants while the videos consistently played in the same order. During trials, participants were instructed to adopt a comfortable hand position for the duration of the trials, remain facing forward, and watch the video in front of them. While the videos were one minute in duration, kinematic and EMG data were sampled for only 30 seconds, approximately at the midway point of the video, to capture a period where the participants were fully immersed in the virtual setting. Once the 45 trials were collected, the markers and electrodes were removed, and testing was concluded. The total in-person testing session took approximately 90 minutes.

The same COM_HTP model was used to calculate measures of head and trunk ROM for each trial within the V3D software. To evaluate head ROM, the joint angle was computed between the right head marker and the generated trunk segment for yaw (rotation about longitudinal axes) and pitch angles (rotation about medio-lateral axis). To evaluate trunk ROM, the joint angle was computed between the trunk and pelvis segments, also in the yaw and pitch directions. Once the joint angles were generated for each measure, the maximum and minimum angle per trial was identified. The range of ROM was then computed by calculating the difference between the maximum and minimum values and exported (MS Excel). Range values were then averaged across the experimental condition (vision x posture; 3 x 3) to calculate the independent variables of interest: head and trunk ROM, along the yaw and pitch directions. Once again, outlier trials were removed if they were greater or less than two standard deviations from the mean of the condition.

To calculate our EMG dependent variable of interest, the trapezoidal integrated root-mean square (RMS; e.g., Pourhashemi et al., 2025), the raw data was processed within Python (Python Software Foundation, Delaware, USA). Raw data was band-pass filtered from 10–499 Hz, and then further filtered using a 4th order Butterworth filter at 10 Hz. The resulting signal was then smoothed using a 200-ms RMS rolling window (i). Once complete, a trapezoidal integration was completed across the signal with unit spacing (all trials had the same sampling rate; see Eq. 1), and the total value was produced in millivolts (mV).

[Equation 1]

This yielded a single value for each trial, for each visual and postural condition, across all participants. The data were exported to Microsoft Excel, where, for each participant, an average of the no video trials was calculated. Each virtual cue video trial (hallway, boardwalk, and hiking trail) was then normalized to its corresponding no video condition by dividing the trial’s RMS value by that participant’s average corresponding no video RMS. Outlier trials were removed if they were greater than or less than one standard deviation from the mean of the condition. The remaining values were then averaged across each experimental condition (vision x posture; 3 x 3) for each participant, in preparation for statistical analyses.

As we were interested in understanding how visual input shapes sensorimotor and postural control, this protocol was designed to manipulate visual scene complexity (no video, hallway, boardwalk, and trail scenes) and BoS (seated with feet planted, seated with feet dangling, and standing) and assess how these modifications influence postural sway (ML and AP displacement and acceleration), segmental movement (head and trunk ROM), and muscle activity (integrated RMS) across the sexes. Based on previous research in P-A coupling, where perception is inherently linked to the control of action (Gibson, 1966; Warren et al., 2001), our approach aimed to isolate how visual scenes with varying complexity may serve as feedforward control sources in the absence of overt movement. Through this methodology, we were able to examine how visuomotor responsiveness changes with different conditions of postural demand. Our findings partially supported our hypotheses, whereby we predicted that visually complex environments (e.g., hallway, boardwalk, and trail) would elicit greater postural sway and sensorimotor engagement, especially under reduced BoS (e.g., seated with feet dangling and standing) due to the reliance on visual inputs. The results suggest that the complexity of a visual cue plays an important role in the modulation of postural control mechanisms, more critically in the standing posture, and that this response may not be uniform across the sexes.

In conclusion, visual cue complexity and posture interact to control balance-related motor behaviours across multiple physiological systems. Findings underscore the need to consider sex as a key variable in visuomotor control research and suggests that complex visual environments can differentially challenge or facilitate postural control depending on the sensorimotor context. Future studies should investigate these visuo-postural dynamics during more ecologically valid locomotor tasks and utilize neurophysiological tools (e.g. mobile electroencephalography (EEG) or functional near-infrared spectroscopy (fNIRS)) to map sensory weighting mechanisms in action further.