Avoidance learning encompasses the acquisition of behaviours that enable individuals to evade or withdraw from potentially harmful stimuli, prior to their occurrence. Maladaptive avoidance is a crucial feature of anxiety and trauma-related disorders. In biological and clinical settings, avoidance behaviours usually involve uninstructed, idiosyncratic and complex motor actions. However, there is a lack of laboratory paradigms that allow investigating how such actions are acquired. To fill this gap, we developed a wireless virtual reality platform to investigate avoidance learning in naturalistic settings, with an uncomfortable sound as unconditioned stimulus (US), a physically plausible avoidance action, and allowing for unconstrained movements. This platform, the CogLearn Toolkit for Unity, is publicly available and allows conducting various types of learning experiments with simple text files as input. We validated this platform in an exploration-confirmation approach with five independent experiments. Overall, participants showed successful acquisition of avoidance behaviour in all experiments. In three exploration experiments, we refined the paradigm and identified mean distance from US location during conditioned stimulus (CS) presentation (before US occurs) as a sensitive measure of avoidance. Two confirmation experiments revealed stronger avoidance for CS+ than CS- during avoidance learning, whether or not this phase was preceded by Pavlovian acquisition. Furthermore, we demonstrated reduced avoidance during extinction with instruction to approach CS, but persistent residual avoidance during this phase. We found evidence of reinstatement in one of two confirmation experiments. Overall, our study provides robust evidence supporting the efficacy of our paradigm in studying avoidance learning in conditions of high ecological relevance.

Human sensorimotor decision making has a tendency to get ‘stuck in a rut’, being biased towards selecting a previously implemented action structure (hysteresis). Existing explanations propose this is the consequence of an agent efficiently modifying an existing plan, rather than creating a new plan from scratch. Instead, we propose that hysteresis is an emergent property of a system learning from the consequences of its actions. To examine this, 152 participants moved a cursor to a target on a tablet device while avoiding an obstacle. Hysteresis was observed when the obstacle moved sequentially across the screen between trials, whereby the participant continued moving around the same side of the obstacle despite it now requiring a larger movement than the alternative. Two further experiments (n = 20) showed an attenuation when time and resource constraints were eased. We created a simple computational model capturing probabilistic estimate updating that showed the same patterns of results. This provides, to our knowledge, the first computational demonstration of how sensorimotor decision making can get ‘stuck in a rut’ through the updating of the probability estimates associated with actions.

Humans exhibit complex and diverse movements when encountering threats. In this study, we examined spatio-temporal parameters, trunk segment angle, and movement intensity during escape behavior in virtual reality (VR) scenarios. What biomechanical constraints influence human escape decisions during VR threat scenarios? We analyzed the data from our previous study by Sporrer et al. [1], where a fully-immersive VR environment was designed with a 5 ×10 m physical space in which participants could move freely. Participants (n=29) were instructed to forage for fruit on a bush and to avoid contact with various threats over 68 independent episodes. Participants were given a safe shelter to avoid the approaching threat (Fig. 1a.). Threat distances and speeds were such that at assumed escape speed, participants had at least 1.5 s time to start escaping. Participants were equipped with a VR headset (HTC Vive Pro Eye HMD), hand controllers, and three position-tracking sensors on the waist and each foot. The waist tracker was used to derive velocity and acceleration; head and waist trackers were used to estimate the trunk segment angle; and all of the trackers were used to calculate movement intensity [2] (Fig. 1b.). Fig. 1. (a) Example scenario of the VR environment featuring bushes for foraging, a safehouse for escaping, and a panther as a predatory threat. (b) Location of the position-tracking sensor and the equations used to calculate relevant variables. (c) Results from the first 1.5 s after threat appeared (black plot), and the previous 1.5 s before the end state (blue plot).Download : Download high-res image (215KB)Download : Download full-size image Participants who were caught by the threat had a lower maximum velocity within the first 1.5 seconds of the threat appearing (0.19±0.02 m/s) compared to those who escaped (0.63±0.02 m/s). Participants who were caught by the threat also leaned forward more (p<0.001) within the first 1.5 seconds after the threat appeared. Movement intensity was higher in all trackers during the first 1.5 seconds after the threat appeared in successfully escaped trials (Fig. 1c.). However, in the last 1.5 seconds before the end state, the participants who were caught by the threat exhibited higher velocity (p<0.001), acceleration (p<0.001), and overall movement intensity than the escaped participants, which may indicate that they were still trying to escape. Our findings suggest that although participants had at least 1.5 s decision time in all scenarios, moving less quickly during this period would make later virtual death more likely. The forward trunk inclination during this period may also have imposed a biomechanical constraint that affected their ability to move, walk, or run in another direction. Overall movement intensity across all body parts may also predict if the participants will successfully escape from the threat. Further studies are required to elucidate biomechanical constraints on human escape.

Abstract Background There is a bidirectional link between sleep and migraine, however causality is difficult to determine. This study aimed to investigate this relationship using data collected from a smartphone application. Methods Self-reported data from 11,166 global users (aged 18–81 years, mean: 41.21, standard deviation: 11.49) were collected from the Migraine Buddy application (Healint Pte. Ltd.). Measures included: start and end times of sleep and migraine attacks, and pain intensity. Bayesian regression models were used to predict occurrence of a migraine attack the next day based on users’ deviations from average sleep, number of sleep interruptions, and hours slept the night before in those reporting ≥ 8 and < 25 migraine attacks on average per month. Conversely, we modelled whether attack occurrence and pain intensity predicted hours slept that night. Results There were 724 users (129 males, 412 females, 183 unknown, mean age = 41.88 years, SD = 11.63), with a mean monthly attack frequency of 9.94. More sleep interruptions (95% Highest Density Interval (95%HDI [0.11 – 0.21]) and deviation from a user’s mean sleep (95%HDI [0.04 – 0.08]) were significant predictors of a next day attack. Total hours slept was not a significant predictor (95%HDI [-0.04 – 0.04]). Pain intensity, but not attack occurrence was a positive predictor of hours slept. Conclusions Sleep fragmentation and deviation from typical sleep are the main drivers of the relationship between sleep and migraine. Having a migraine attack does not predict sleep duration, yet the pain associated with it does. This study highlights sleep as crucial in migraine management.

All animals have to respond to immediate threats in order to survive. In non-human animals, a diversity of sophisticated behaviours has been observed, but research in humans is hampered by ethical considerations. Here, we present a novel immersive VR toolkit for the Unity engine that allows assessing threat-related behaviour in single, semi-interactive, and semi-realistic threat encounters. The toolkit contains a suite of fully modelled naturalistic environments, interactive objects, animated threats, and scripted systems. These are arranged together by the researcher as a means of creating an experimental manipulation, to form a series of independent "episodes" in immersive VR. Several specifically designed tools aid the design of these episodes, including a system to allow for pre-sequencing the movement plans of animal threats. Episodes can be built with the assets included in the toolkit, but also easily extended with custom scripts, threats, and environments if required. During the experiments, the software stores behavioural, movement, and eye tracking data. With this software, we aim to facilitate the use of immersive VR in human threat avoidance research and thus to close a gap in the understanding of human behaviour under threat.

Animals including humans must cope with immediate threat and make rapid decisions to survive. Without much leeway for cognitive or motor errors, this poses a formidable computational problem. Utilizing fully-immersive virtual reality with 13 natural threats, we examined escape decisions in N = 59 humans. We show that escape goals are dynamically updated according to environmental changes. The decision whether and when to escape depends on time-to-impact, threat identity and predicted trajectory, and stable personal characteristics. Its implementation appears to integrate secondary goals such as behavioral affordances. Perturbance experiments show that the underlying decision algorithm exhibits planning properties and can integrate novel actions. In contrast, rapid information-seeking and foraging-suppression are only partly devaluation-sensitive. Instead of being instinctive or hardwired stimulus-response patterns, human escape decisions integrate multiple variables in a flexible computational architecture. Taken together, we provide steps toward a computational model of how the human brain rapidly solves survival challenges.

Understanding and characterising human movements is complex due to the diversity of human actions and their inherent inter, intra, and secular variability. Traditional marker-based, and more recently, some marker-less motion capture (MoCap) systems have demonstrated to be reliable tools for movement analysis. However, in complex experimental set ups involving virtual reality (VR) and free movements (as in [1]), accuracy and reliability tend to decrease due to occlusion, sensor blind spots, marker detachment, and other artifacts. Furthermore, when actions are less distinct, e.g., fast walk and slow run, current classification methods tend to fail when actions overlap, which is expected as even researchers struggle to manually label such actions. Can current marker-less MoCap systems, pose estimation (PE) algorithms, and advanced action classification (AC) methods: (1) accurately track participant movements in VR; (2) cluster participant actions. The experiment consisted of avoiding threats (Fig. 1A) whilst collecting fruit in VR environments (n=29 participants, 5x10m area), see [1]. The Unity® software [2], based on the Unity Experiment Framework [3], was used to create the VR experiment, which was streamed through an HTC vive pro (HTC Corporation) VR headset. Movements were recorded using 5 ELP cameras (1280×720 @120 Hz) synchronised with the Open Broadcaster Software® (OBS) [4]. Openpose [5] was employed for PE (Fig. 1B). Euclidean distances, and angular positions, velocities, and accelerations were derived from cartesian positions. Finally, Uniform Manifold Approximation and Projection (UMAP) was used to embed high-dimensional features into a low-dimensional space, and Hierarchical Density Based Spatial Clustering of Applications (HDBSCAN) was used for classification (see Fig. 1E), similar to B-SOiD [6]. Participants were virtually killed by the threat in 223 episodes, for which the participants' last poses were estimated. After applying UMAP and HDBSCAN, 5 pose clusters were found (see Fig. 1C-D), which depict: (a) stand up, picking fruit with slow escape; (b) stand up, arms extended and slow escape; (c) long retreat at fast speed; (d) short retreat at medium speed; (e) crouching and picking fruit; (x) 4% unlabelled. Fig. 1. (A) VR-threat, (B) Participant estimated 3D-pose, (C) Pose clusters, (D) Cluster examples, (E) Methodology.Download : Download high-res image (176KB)Download : Download full-size image Marker-less MoCap and PE methods were mostly successful for participants' last poses. However, in some cases, and during exploration, tracking was lost due to occlusion and sensor blind spots. The results from the AC methods are an indication of the potential use of unsupervised methods to find participant actions under threat in VR. Nevertheless, such clustering is rather general, and had some AC errors, which could not be quantified as further work is needed to understand and define where the threshold of overlapping actions occurs. The results are exciting and promising; however, further investigation is needed to validate the findings, and to improve the AC methods.

Background There is a bidirectional link between sleep and migraine, however causality is difficult to determine. Previous studies rely on retrospective questionnaires, and small samples to support their findings. This study aimed to overcome this by teasing apart this relationship using sleep assessment and concomitant migraine data collected from a smartphone application using Bayesian modelling.Methods Anonymized self-reported data on sleep and migraine from 11,166 global users (aged 18-81 years, mean: 41.21, standard deviation: 11.49) were collected from the Migraine Buddy application (Healint Pte. Ltd.) between 30th June and 31st December 2021. Measures included: demographics, start and end times of each sleep episode and migraine attack, and the pain intensity for each migraine attack (visual analogue scale 0-10). Bayesian regression models were used to predict occurrence of a migraine attack the next day based on users’ deviations from mean monthly sleep, number of sleep interruptions, and hours slept the night before in those reporting ≥ 4 and <25 migraine attacks on average per month. Conversely, we modelled whether attack occurrence and pain intensity could predict hours slept that same night.Results Once exclusion criteria were applied, there were 724 users (129 males, 412 females, 183 unknown) with an average age of 41.88 years ( SD = 11.63), with a mean monthly number of attacks of 9.94. A greater number of sleep interruptions (95% Highest Density Interval (95% HDI [0.112 – 0.205]) and deviation from a user’s mean sleep the night before (95% HDI [0.040 – 0.080]) were significant predictors of a next day migraine attack. Total hours slept was not a significant predictor (95% HDI [-0.04 – 0.04]). Pain intensity, but not attack occurrence was a positive predictor of hours slept.Discussion Sleep fragmentation and deviation from typical sleep are the main drivers of the relationship between sleep and migraine, whereas overall sleep duration is not. Conversely, simply having a migraine attack does not predict sleep duration, it is the pain associated with an attack which alters sleep. This study has shed light on the causal mechanisms of sleep and migraine and highlights sleep hygiene as crucial in migraine management.### Competing Interest StatementHealint provided in-kind dataset, prior data-collection, data preparation, and intellectual support. J. Hoffmann reports honoraria for consulting activities and/or serving on advisory boards from Allergan, Autonomic Technologies Inc., Cannovex BV, Chordate Medical AB, Eli Lilly, Hormosan Pharma, Lundbeck, Novartis, Sanofi and Teva. He received personal fees for Medico-Legal work as well as from Oxford University Press, Quintessence Publishing, Sage Publishing and Springer Healthcare. He also reports a research grant from Bristol Myers Squibb. J. Hoffmann serves as Associate Editor for Cephalalgia, Cephalalgia Reports, Journal of Oral & Facial Pain and Headache as well as for Frontiers in Pain Research. He is an elected member of the Board of Trustees of the International Headache Society (IHS) and serves as a Council Member and Treasurer of the British Association for the Study of Headache (BASH). All these activities are unrelated to the submitted work. P.R.H. reports, unrelated to the current project, honoraria for educational and advisory purposes from Allergan, Novartis and Teva as well as research funding from Eli Lilly, Amgen, Kallyope and Cellgene/Bristol Myers Squibb. P.R.H serves as an associate editor for Cephalalgia and the British Journal of Pharmacology. JR.P and A.U. are employees of Healint Pte Ltd ### Funding StatementThis work has been supported by a Medical Research Council PhD studentship (ECS; MR/N013700/1).### Author DeclarationsI confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.YesThe details of the IRB/oversight body that provided approval or exemption for the research described are given below:Ethics committee of King's College London waived ethical approval for this work due to the study involving anonymised, de-identified / commercially available data.I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.YesI understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).YesI have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable.YesAll data produced in the present study are available upon reasonable request to the authors

The integration of visual and tactile cues can enhance perception. However, the nature of this integration, and the subsequent benefits on perception and action execution, are context-dependent. Here, we examined how visual-tactile integration can influence performance on a complex motor task using virtual reality. We asked participants to wear a VR head-mounted display while using a tracked physical putter to make golf putts on a VR golf course in two conditions. In the ‘tactile’ condition, putter contact with the virtual golf ball coincided with physical contact with a physical ball. In a second ‘no tactile’ condition, no physical ball was present, such that only the virtual ball contacted the putter. In contrast to our pre-registered prediction that performance would benefit from the integration of visual and tactile cues, we found golf putting accuracy was higher in the no tactile condition compared to the tactile condition. Participants exhibited higher lateral error variance and over/undershooting when the physical ball was present. These differences in performance between the conditions suggest that tactile cues, when available, were integrated with visual cues. Second, this integration is not necessarily beneficial to performance. We suggest that the decreased performance caused by the addition of a physical ball may have been due to minor incongruencies between the virtual visual cues and the physical tactile cues. We discuss the implications of these results on the use of VR sports training and highlight that the absence of matched tactile cues in VR can result in sub-optimal learning and performance.

In light of recent advances in technology, there has been growing interest in virtual reality (VR) simulations for training purposes in a range of high-performance environments, from sport to nuclear decommissioning. For a VR simulation to elicit effective transfer of training to the real-world, it must provide a sufficient level of validity, that is, it must be representative of the real-world skill. In order to develop the most effective simulations, assessments of validity should be carried out prior to implementing simulations in training. The aim of this work was to test elements of the physical fidelity, psychological fidelity and construct validity of a VR golf putting simulation. Self-report measures of task load and presence in the simulation were taken following real and simulated golf putting to assess psychological and physical fidelity. The performance of novice and expert golfers in the simulation was also compared as an initial test of construct validity. Participants reported a high degree of presence in the simulation, and there was little difference between real and virtual putting in terms of task demands. Experts performed significantly better in the simulation than novices ( p = .001, d = 1.23), and there was a significant relationship between performance on the real and virtual tasks ( r = .46, p = .004). The results indicated that the simulation exhibited an acceptable degree of construct validity and psychological fidelity. However, some differences between the real and virtual tasks emerged, suggesting further validation work is required.

Disturbance forces facilitate motor learning, but theoretical explanations for this counterintuitive phenomenon are lacking. Smooth arm movements require predictions (inference) about the force-field associated with a workspace. The Free Energy Principle (FEP) suggests that such 'active inference' is driven by 'surprise'. We used these insights to create a formal model that explains why disturbance might help learning. In two experiments, participants undertook a continuous tracking task where they learned how to move their arm in different directions through a novel 3D force field. We compared baseline performance before and after exposure to the novel field to quantify learning. In Experiment 1, the exposure phases (but not the baseline measures) were delivered under three different conditions: (i) robot haptic assistance; (ii) no guidance; (iii) robot haptic disturbance. The disturbance group showed the best learning as our model predicted. Experiment 2 further tested our FEP inspired model. Assistive and/or disturbance forces were applied as a function of performance (low surprise), and compared to a random error manipulation (high surprise). The random group showed the most improvement as predicted by the model. Thus, motor learning can be conceptualised as a process of entropy reduction. Short term motor strategies (e.g. global impedance) can mitigate unexpected perturbations, but continuous movements require active inference about external force-fields in order to create accurate internal models of the external world (motor learning). Our findings reconcile research on the relationship between noise, variability, and motor learning, and show that information is the currency of motor learning.

Virtual reality (VR) systems hold significant potential for both training and experimentation purposes as they provide precise control over the environment and the possibility to untether tasks from their normal physical constraints. However, the artificial creation of depth in stereoscopic displays, and reduced availability of haptic information, may affect how visually-guided motor tasks are performed in the virtual world. If so, tasks learned in VR may be unrepresentative of real skills, and therefore unlikely to elicit positive transfer to the real-world. In Experiment 1 we tested whether learning a visually-guided motor skill (golf putting) in virtual reality could transfer to real-world improvements. Despite the perceptual limitations imposed by the virtual environment, training novice golfers in VR led to improvements in real putting that were comparable to real-world practice. Experiment 2 explored these effects in more skilled golfers, and examined changes in gaze behaviour (quiet eye) that resulted from the more immediate use of VR (i.e. as a tool for ‘warming up’). VR use was found to cause impairments to gaze control (quiet eye) and putting accuracy, when used immediately prior to real world putting. Overall, these findings demonstrate the potential for VR training, but also highlight that fundamental questions remain about how the altered perceptual environment of VR affects visually-guided skills.

Losing a point in tennis could result from poor shot selection or faulty stroke execution. To explore how the brain responds to these different types of errors, we examined feedback-locked EEG activity while participants completed a modified version of a standard three-armed bandit probabilistic reward task. Our task framed unrewarded outcomes as either the result of errors of selection or errors of execution. We examined whether amplitude of a medial frontal negativity (the Feedback-Related Negativity; FRN) was sensitive to the different forms of error attribution. Consistent with previous reports, selection errors elicited a large FRN relative to rewards and amplitude of this signal correlated behavioral adjustment following these errors. A different pattern was observed in response to execution errors. These outcomes produced a larger FRN, a frontocentral attenuation in activity preceding this component, and a subsequent enhanced error positivity in parietal sites. Notably, the only correlations with behavioral adjustment were with the early frontocentral attenuation and amplitude of the parietal signal; FRN differences between execution errors and rewarded trials did not correlate with subsequent changes in behavior. Our findings highlight distinct neural correlates of selection and execution error processing, providing insight into how the brain responds to the different classes of error that determine future action.### Competing Interest StatementThe authors have declared no competing interest.

Virtual reality (VR) systems offer a powerful tool for human behavior research. The ability to create three-dimensional visual scenes and to measure responses to the visual stimuli enables the behavioral researcher to test hypotheses in a manner and scale that were previously unfeasible. For example, a researcher wanting to understand interceptive timing behavior might wish to violate Newtonian mechanics so that objects can move in novel 3-D trajectories. The same researcher might wish to collect such data with hundreds of participants outside the laboratory, and the use of a VR headset makes this a realistic proposition. The difficulty facing the researcher is that sophisticated 3-D graphics engines (e.g., Unity) have been created for game designers rather than behavioral scientists. To overcome this barrier, we have created a set of tools and programming syntaxes that allow logical encoding of the common experimental features required by the behavioral scientist. The Unity Experiment Framework (UXF) allows researchers to readily implement several forms of data collection and provides them with the ability to easily modify independent variables. UXF does not offer any stimulus presentation features, so the full power of the Unity game engine can be exploited. We use a case study experiment, measuring postural sway in response to an oscillating virtual room, to show that UXF can replicate and advance upon behavioral research paradigms. We show that UXF can simplify and speed up the development of VR experiments created in commercial gaming software and facilitate the efficient acquisition of large quantities of behavioral research data.

Robotics is increasing in popularity as a method of providing rich, personalized and cost-effective physiotherapy to individuals with some degree of upper limb paralysis, such as those who have suffered a stroke. These robotic rehabilitation systems are often high powered, and exoskeletal systems can attach to the person in a restrictive manner. Therefore, ensuring the mechanical safety of these devices before they come in contact with individuals is a priority. Additionally, rehabilitation systems may use novel sensor systems to measure current arm position. Used to capture and assess patient movements, these first need to be verified for accuracy by an external system. We present the ALAN-Arm, a humanoid robotic arm designed to be used for both accuracy benchmarking and safety testing of robotic rehabilitation systems. The system can be attached to a rehabilitation device and then replay generated or human movement trajectories, as well as autonomously play rehabilitation games or activities. Tests of the ALAN-Arm indicated it could recreate the path of a generated slow movement path with a maximum error of 14.2mm (mean = 5.8mm) and perform cyclic movements up to 0.6Hz with low gain (<;1.5dB). Replaying human data trajectories showed the ability to largely preserve human movement characteristics with slightly higher path length and lower normalised jerk.