Virtual reality assessment of a high-calorie food bias: Replication and food-specificity in healthy participants

Background: Theoretical models and behavioural studies indicate faster approach behaviour for high-calorie food (approach bias) among healthy participants. A previous study with Virtual Reality (VR) and online motion-capture quantified this approach bias towards food and non-food cues in a controlled VR environment with hand movements. The aim of this study was to test the specificity of a manual approach bias for high-calorie food in grasp movements compared to low-calorie food and neutral objects of different complexity, namely, simple balls and geometrically more complex office tools. Methods: In a VR setting, healthy participants (N = 27) repeatedly grasped or pushed high-calorie food, low-calorie food, balls and office tools in randomized order with 30 item repetitions. All objects were rated for valence and arousal. Results: High-calorie food was less attractive and more arousing in subjective ratings than low-calorie food and neutral objects. Movement onset was faster for high-calorie food in push-trials, but overall push responses were comparable. In contrast, responses to high-calorie food relative to low-calorie food and to control objects were faster in grasp trials for later stages of interaction (grasp and collect). Non-parametric tests confirmed an approach bias for high-calorie food. Conclusion: A behavioural bias for food was specific to high-calorie food objects. The results confirm the presence of bottom-up advantages in motor-cognitive behaviour for high-calorie food in a non-clinical population. More systematic variations of object fidelity and in clinical populations are outstanding. The utility of VR in assessing approach behaviour is confirmed in this study by exploring manual interactions in a controlled environment.


Introduction
Eating behaviour results from complex interactions between various bottom-up mechanisms of motivational drive such as hunger, reward sensitivity, or attention to food, and top-down cognitive control processes [7,12].An imbalance of bottom-up and top-down processes is widely assumed in pathological and disordered eating patterns [1,24].Accordingly, food intake is determined by subjective homeostatic and hedonic mechanisms, but also by automatic cognitive processing of environmental cues [35,44].In fact, in the modern Western environment, many automatic cognitive processes seem to be strongly captured by highly palatable and energy-dense high-calorie foods that can trigger incentive attribution processes.
The presence of cognitive and behavioural approach biases towards high-calorie food stimuli has been documented by several studies [5,16,66].Correspondingly, approach tendencies have been also explored as therapeutic target in different eating-related pathologies [20].However, although several studies support the clinical relevance of biased attentional and approach mechanisms towards food, the way in which these processes unfold in explicit behaviours is not yet clear.
It is largely established by eye-tracking research and paradigms such as the dot-probe task that an early attentional bias is involved in foodrelated mental disorders [5,58,64].Regarding biased approach mechanisms towards food, existing paradigms assessed cognitive and motor behavioral responses to food cues with simple movements derived from a joystick or computer tablet [18,70,73].In the approach-avoidance task (AAT), participants are required to push or pull a joystick or another response device in response to a stimulus on a computer screen [4,20,31,54,66,71].The AAT can detect approach biases towards food or towards appealing stimuli in faster response times for push-vs.pull-trials for the respective target category.For example, pull responses to high-calorie food were faster than push responses in a relatively large sample of normal-weight and overweight participants [21].Compared to more cognitive measures of approach-avoidance associations, early studies showed that the AAT was more sensitive to show impaired avoidance of sweet snacks with increased BMI [37,66].
Several studies have compared the extent of the approach bias to high-calorie food in the AAT across clinical groups with disordered eating [19].Excess weight has been associated with heightened approach in some studies [21,41,60], however, individuals with binge-eating disorder showed avoidance of low-calorie food [48].Another study during weight loss further observed comparable approach biases for healthy-weight and normal-weight children [57].In a cognitive bias modification trial, patients with bulimia nervosa and binge eating disorder did not differ in the AAT [3].Although it is assumed that automatic approach-avoidance tendencies would predict consumption (at least in individuals with poor inhibitory control), patients with anorexia nervosa, bulimia nervosa, and healthy controls did not differ in high-calorie nor low-calorie AAT [27].Finally, in another study testing patients with the restrictive type of anorexia nervosa on the affective manikin task, less approach tendencies were demonstrated compared to healthy controls [67].
As a possible intervention in obesity, AAT modification protocols have been introduced since they are hypothesized to overcome the low efficacy of conventional treatments by targeting early cognitive processing of food stimuli [4,49].However, AAT modification studies did not always yield consistent results [2,30].Moreover, transfer to actual eating-related situations from the lab is a rarely successfully tested assumption [21], but transfer could possibly be improved if trainings incorporated different relevant situations and overt behaviours, such as reaching out and grasping high-calorie food with more extensive movements.In contrast to promising effects of AAT training on abstinence rates in alcohol use disorder [38], different mechanisms may pertain to pathologies in the eating-related domain and restrict transfer to real-life settings [17].
Virtual Reality (VR) and motion tracking can be useful to overcome some of these limitations: In comparison with computer-based tasks, VR provides high sensorimotor and visuospatial integration and allows more complex movements than those needed to move a mouse or a joystick.Furthermore, VR has already been proposed as an effective technique for behavioural training programs that are difficult to replicate in the clinic [34,52], since the ability to create highly immersive virtual worlds is likely to allow for high adherence rates and good generalizability [10].Next to replicating behavioural tendencies, which can be successfully simulated also in less complex tasks [22,45], VR especially allows for tailored training in different virtual environments pertaining to a variety of psychopathological conditions [11].Thus, VR can open a new series of treatment approaches for eating-related pathologies [55], ranging from experiential bodily awareness [56] to experimental modulation of body-related personal cues [51].To summarize, VR has the potential to improve ecological validity of the approach bias paradigms, using a highly controlled environments in which the clinical stimuli, i.e., subtypes of food, can be modified, and improve the motivation and treatment adherence of participants.
In the present study, we focus on the presentation of subtypes of food, i.e., low-calorie and high-calorie food, and recordings of grasp and push hand movements in a non-clinical population.The aim of this study was to replicate and extend findings from a previous VR experiment, which assessed the relationship between motivational systems and overt approach bias towards high-calorie food [60].Comparable to the AAT, a small group of healthy participants performed grasp or push movements with their own hands when exposed to simulated 3D high-calorie food vs. neutral objects, which was enabled by concurrent real-time motion tracking with the LeapMotion® near-infrared hand-tracking sensor (cf.[32]).By recording the timing of the behavioural responses, results showed that food items were collected faster than neutral objects (balls).An important novelty of this study was represented by the use of a VR setting, which provided a high sensorimotor integration and allowed to systematically investigate more complex actions tendencies (like the manipulation of virtual objects) in a stereoscopic immersive environment.
With the present study, we aimed to extend on these findings by investigating the interplay of motor action execution and food-related approach biases when interacting with high-calorie food, low-calorie food, and neutral objects of different complexity, namely, balls and geometrically more complex tools.We investigated behavioural trajectories for 3D models of these objects by measuring time needed to collect them manually and time needed to push them away.Specifically, the results from our previous pilot study [60] only contrasted high-calorie food with balls and showed that hand movements towards high-calorie food were earlier initiated (movement onset), earlier grasped (object contact) and earlier collected.In contrast, similar studies with healthy participants but less extensive motor movement produced mixed results and low-calorie food was hardly studied in healthy populations [19,36,67].In the joystick AAT study by Maas et al. [37], for instance, low-fat and high-fat pictures alike were faster approached, but snacking-bothered participants were slower to avoid food pictures of all kinds compared to a control group and to scrambled neutral pictures.Furthermore, the joystick AAT in a comparison of anorexia nervosa patients with healthy controls showed slower general reactions to high-calorie vs. low-calorie food pictures [27].However, pull-and push-reactions were not statistically different for high-calorie vs. low-calorie food across groups [27].In immersive VR scenarios, the high degree of (simulated) interaction comes at the cost of perceptual matching of stimuli: Because stimuli can be rotated and the viewing perspective is dynamic, basic visual features such as shape and colour distributions are hard to control for, at least when compared to 2D pictures.Instead, in the current experiment, two different additional sets of stimulus objects that varied in their basic visual features were introduced to all participants in a repeated measures design to address this shortcoming of our previous results.All stimuli were rated regarding valence and arousal, to enable a differentiation of subjective evaluations from measures of behaviour.
In the present repeated measures study design, healthy participants were immersed in the VR and confronted with approach-and avoidance trials to four categories of stimuli in randomized sequence.We hypothesized the presence of a behavioural bias for collecting high-calorie food compared to low-calorie food, balls, and tools, as well as specificity for grasping as opposed to pushing interactions.More precisely, we expected faster grasp movements for high-calorie food, but not lowcalorie food, relative to the control stimuli and relative to push movements.

Participants
Healthy participants (N = 27; 9 males; mean age = 23.7 years, SD = 3.48 years, range = 18-32 years) within a healthy range of BMI (18.9 -24.9 kg/m 2 ) were recruited through mailing lists and flyers from the Tübingen student population.All participants were >18 years old, had normal or corrected-to-normal vision and were right-handed according P.A. Schroeder et al. to Oldfield's handedness inventory [47].Under-or overweight (i.e., a BMI exceeding the 18.5-24.9kg/m 2 range) was considered exclusion criterion; furthermore, left-handedness, epilepsy, or any other history of severe mental or physical disease led to exclusion.Participants reported moderate levels of hunger in the beginning of the experiment [mean = 6.00 (SD = 0.96); range = 5-8 on a scale from 1 to 10] and there was no eating restriction requirement before study participation.There was no indication of eating disorders and the mean global score in the eating attitudes test (EAT-26D) was below the cut-off score of 20 (M = 4.59, SD = 5.09).
The sample size was justified by available resources and acceptable power to reproduce previously reported behavioural differences between food and non-food 3D objects [60] in a within-subjects design.In particular, this sample size was sufficient to replicate a behavioural bias for food in VR-mediated motor responses (f = 0.48 [60]) in a factorial within-subjects design with a significance level of α =.05 and a power of 1-β = 0.95.In our previous study with a similar non-clinical sample, this large effect size was reported for the difference between high-calorie food cues and neutral cues in both movement onset and contact times; a direct replication of these two contrasts in a paired t-test required a sample of 17 participants.
Participants provided informed consent and received either course credit or a small monetary compensation for their participation.Ethical approval for the study was obtained from the University Hospital Tübingen Ethical Commission (No. of approval: 207/2015BO2).All data for this study were collected in 2015 before the Coronavirus pandemic.

Apparatus
To immerse participants in the VR, they were equipped with an Oculus Rift© DK2 stereoscopic head-mounted display (HMD; Oculus VR LLC, Menlo Park, California).Motion tracking of hand movements was realized with a LeapMotion© near-infrared sensor (LeapMotion Inc., San Francisco, California, USA, SDK version 3.1.3).The LeapMotion© sensor provides positional information regarding the palm, wrist, and phalanges.This data can be used to render a hand model in VR, as in previous studies [32,33,60].The whole experiment was implemented within the Unity® game engine 4.5 using the C# interface provided by the application programming interface.Instructions and feedback were presented on different text-fields, aligned at eye-height.

Procedure
Prior to the actual experimental testing, participants received a verbal instruction regarding proper handling of the VR equipment.Then they were equipped with the HMD and the experiment started with practice trials.During the practice trials, participants were asked to grasp and push objects according to the instructions (see Fig. 2) and each object was presented to them once.They were allowed to repeat the practice trials if needed.The experiment was conducted in a seated position.A researcher was present throughout the entire time of the experiment.All participants completed the same virtual experience and all independent variables were within-subject manipulations.
The supplementary video 1 shows examples of the trial sequence and Fig. 1 displays screenshots from the participants' view.The VR procedure was identical to our pilot study [60] except for the additional stimulus categories.Each trial consisted of two parts: (1) Preconditions had to be met for an interaction to begin: Participants had to move their right hand into a predefined and fixed starting position, indicated by red, semi-transparent spheres, at a comfortable height close to the participant.If the palms were within the positions and the hand was open, the semi-transparent spheres turned green.Furthermore, participants had to center their field of view on a fixation cross located at the outer bound of the task space.(2) Once the center of the visual field had been directed towards the fixation cross for at least 500 ms, the spheres and the fixation cross disappeared (stimulus-onset asynchrony: 200 ms), and a colored cue indicated the upcoming position of a target and indicated the requested action (e.g., blue cue for pushing and purple cue for grasping movements, 400 ms).This cue was then replaced with the target object, which appeared with slight motion directed towards the participant.Objects always appeared approximately 20 cm in front of the participants close to the position of the (removed) fixation cross, but exact location, rotation, and speed were jittered to make the task more challenging (please see supplementary video 1, for demonstration of these subtle variations).In the case of grasping, participants were requested to close their hand surrounding the virtual object, move it Fig. 1.Screenshots from trials from the perspective of participants.The virtual environment was a schematic square with a number of depth cues displayed on the floor and in the distance.Participants were positioned close to a box to which objects had to be collected.Up to three of previously objects were still shown in the box.To initiate a trial, preconditions had to be met: (A) Participants had to place their right-hand in a standardized starting position that was indicated by half-transparent red / green balls and (B) the head-mounted display had to be rotated towards a fixation cross.After that (C) a purple or blue sphere indicated the position of the upcoming object and the instruction (pull or push), respectively.(D) High-calorie food, low-calorie food, balls, or office tools were next shown at the position with random rotation and an intial movement vector directed towards the participant.With a simulated grasping movement (in-air pinching of fingers), objects could be collected and (E) moved to the box.Alternatively, in push-trials the objects appeared identically to (D) but any collision with the hand (without simulated grasping) could be utilized to push target stimuli.The mapping of instructions to purple and blue spheres was continuously shown in the upper field of the participants' view, together with a progress bar and block count.For illustration purposes, pictures display screenshots from different trials of the participants' left-eye and stimuli in an exemplary timeline.towards themselves, and place it in a box in front of them.In the case of pushing, participants were requested to hit the target object with their hand, which would then fly away after the collision.The three most recently collected objects remained in the box adjunct to the participants' feet, whereas pushed objects were cleared always before the next trial.
Trials were cancelled if the movement initiation took longer than two seconds.In case of such time-outs, early movements (earlier than 250 ms), or wrong responses, participants received according feedback in the form of a semi-transparent text-field.The whole experiment was self-paced, since trials only started when participants took the initial position and fixated the fixation cross.Hence, participants could (and they were encouraged to) take breaks between trials at any time.After half of all trials (i.e., approximately 30 minutes of VR), participants were asked to take off the HMD for a slightly longer break of some minutes.
In extension to ball and high calorie virtual food objects that were already studied before in a different sample of participants [60], two further categories of low calorie food and complex office-tools were selected All stimuli were created with 3D objects from the Unity® asset store and were modified to be equally sized Since no low-calorie and office-tools 3D stimuli were available from previous studies, the authors tried to maintain a consistent graphical style (low-poly, cartoon-like stimuli) with low rendering load to compliment fast refresh cycles and continuous motion capture.Screenshots of the object categories are shown in Fig. 2. Grasp and push interactions were requested in randomized order across the whole experiment for each of the 12 objects.Each object was presented 30 times, resulting in 360 trials in total.The experiment was subdivided into a total of 10 blocks, each comprising 36 trials, for a total duration of approximately 45 minutes.After the first part of 5 blocks, participants were encouraged to take a break before the second part of 5 blocks started.The entire experiment lasted approximately 2 hours.

Questionnaires and ratings 2.4.1. Hunger, BMI, presence, simulator sickness
Before entering VR, participants had to indicate their hunger on a 1-10 hunger scale with descriptive anchors (10 = painstakingly full, uncomfortable; 1 = need to eat now).After the experiment, all participants were asked to rate the VR exposure regarding experienced presence [62] and possible symptoms of simulator sickness [25].For presence, we used the established Igroup Presence Questionnaire [62] with subscales for spatial presence, involvement, realism, and a general presence item.Items were rated on a scale from − 3 (e.g., "not at all", "about as real as an imagined world") to +3 (e.g., "very much", "indistinguishable from the real world").Internal consistencies were weak in this study (Cronbach's α =.60 (spatial presence), α =.60 (involvement), α =.50 (realism), For the simulator sickness questionnaire [25], symptoms were rated from 1 ("not at all") to 4 ("strong") and internal consistency of the total score was acceptable (α =.79).
All participants self-reported their weight and height for computation of BMI (weight in kg / (height im cm) 2 ).The hunger and BMI scores were explored regarding their possible contribution to an approach food bias [5,43,46] in correlation analyses.

Rating of stimuli: valence and arousal
To evaluate the composition of object categories and differentiate subjective from behavioural responses to high-calorie food, participants also had to rate all objects for valence (negativepositive) and arousal (not at allvery arousing) on visual analog scales (VAS).The rating was performed after the experiment.Participants were shown a screenshot of each stimulus on a separate page in randomized order and were given two VAS items with the anchors "negative -positive" and "not at allvery arousing".For analysis, VAS ratings were saved on scales from 0 to 100.Internal consistency for high-calorie food valence was questionable (α =.53), consistencies for low-calorie food, ball, and office-tool valence ratings were acceptable (α =.63 -α =.74).Internal consistencies for arousal ratings were acceptable to excellent (α =.63 -α =.90).

Dependent measures and data treatment
Mean subjective ratings from VAS were calculated for each stimulus category.Response times (RTs) were recorded within the VR in milliseconds at three different stages of an experimental trial.Movement onset was defined as the time-point when the grasping hand left the starting position, object contact was triggered by the collision of the virtual hand with the target object, either due to grasping or pushing the object away.Finally, in grasp trials only, collection time was recorded once the object had been placed in the box next to the participant.All of these events were automatically triggered by the physics engine in Unity3D.Please see Fig. 1 for a visualization of these experimental stages within a sample trial.
This study was a repeated measures factorial design with the factors interaction (grasping vs. pushing) and stimulus (high-calorie food, lowcalorie food, balls, office tools).The primary outcome of the study was an anticipated two-way interaction, which we expected to be driven by faster grasp responses to high-calorie food objects.At later stages of the interaction (see Fig. 2), we expected this food bias to become more pronounced, and we also hypothesized an approach bias for collection times.Data from correct trials were individually aggregated to the four Fig. 2. Examples of 3D objects used in the study with non-food items (top row) and food items (bottom row) of different complexities and caloric content.All participants performed the VR grasping and pushing tasks with all types of objects in randomized order with 30 individual item repetitions.
categories for each type of interaction (grasping vs. pushing). 2Incorrect trials were excluded because of erroneous responding (14.2 %) or detection of early hand movements prior to onset of the stimulus (3.3 % of all trials).Further, an outlier correction of 2.5 SD was applied for each RT (1.83 % movement onset, 2.53 % object contact, 2.57 % object collection) before aggregation to the different condition means for each individual.Mean RTs and arousal VAS ratings were not normally distributed according to Kolmogorov Smirnov tests for normality (W = 0.95-0.97,ps <.002).Accordingly, we employed non-parametric statistical tests.We used the aligned rank transformed nonparametric factorial analysis (ART) [72] to investigate the two-way interaction of interaction and stimulus in movement onset and object contact RTs and the main effect of stimulus in collection times, for which only the interaction-level "grasp" was available.The ART procedure aligns data for each main effect and interaction, assigns ranks, and subsequently submits aligned ranked data to an F-test.Simple effects were implemented using paired nonparametric Wilcoxon tests with Bonferroni correction for multiple comparisons.The same analytic strategy was employed for VAS ratings for the main effect of stimulus.Finally, we explored Spearman's rank-order correlations between behaviour (i.e., collection times) and subjective variables such as BMI, hunger, stimulus ratings and questionnaire scores.We limited our correlation analyses to collection times as the eventual outcome of the full motor sequence, as we expected that possible associations are most pronounced with the highest amount of variance; moreover, grasp contact and collection times were positively correlated in the task (r =.45, p <.001).Analyses were performed using R 4.1.2(R [53]), the ARTool package [72] and rstatix [23] for nonparametric testing, the outlier()-function from schoRsch [50] and the tidyverse package [69].

Subjective ratings of 3D objects
All subjective ratings of valence and arousal are reported in Table 1 and in Fig. 3. Assumptions for parametric testing were met for valence, W = 0.99, p =.845, but arousal ratings were not normally distributed, W = 0.97, p =.008.

Movement onset: early food bias
RTs were not normally distributed; accordingly, we submitted values to aligned rank transform before testing the repeated-measures interaction between stimulus and direction.
Mean RTs and SDs are reported in Table 2.

Contact times: interaction intention and food bias
Contact times were not normally distributed; accordingly, we submitted values to the aligned rank transform procedure before testing the repeated-measures interaction between stimulus and direction.
In particular, for push movements, high-calorie food was neither faster reached when compared to ball objects, W = 191, p >.999, nor when compared to low-calorie food, W = 167, p >.999.Mean RTs and SDs are reported in Table 2.

Collection Times: Type of Control Object and Food Bias
Collection times were recorded in grasp trials only.The mean collection times as a function of object type are reported in  2 We also checked in a sensitivity analysis whether duration of the experiment had an influence.Specifically, we included another repeated-measures factor part (first part, second part), because the entire procedure was intermitted by a break after half of the trials.As expected, RTs in all conditions decreased for the second part of the experiment.For movement onset and object contact, all reported effects were comparable at large to the reported main analyses and there was no interaction between part × stimulus or part × stimulus × interaction.For object collection, a behavioural bias for high-calorie relative to low-calorie food was larger during the first part (241 ms) compared to the second part of the experiment (9 ms; F(3,75) = 3.52, p =.019, ηp 2 = 0.12).

Presence, simulation sickness, tolerability
Summary statistics for presence ratings obtained from the Igroup Presence Questionnaire [62] and simulation sickness questionnaire [25] are reported in Tables 2 and 3. Presence scores were within the range of normative data [62] and of previous studies with comparable setups [32,33].Despite the relatively long VR exposure, all participants tolerated the procedure.However, one additional session was cancelled due to the participant's frustration with placing his hand in the starting position.

Complementary analyses: correlations between BMI, hunger, subjective ratings, questionnaires, and movement dynamics
We investigated a set of bivariate Spearman's rank-order correlations between subjective and behavioural responses to the different stimuli, as well as individual variables.The results are reported in Tables 5 and 6.Analyses were limited to collection times as the final outcome of the grasping procedure.Interestingly, only responses to balls were positively associated with subjective ratings.Regarding presence, reaction times were positively associated with the degree of realism and general presence for all objects.In the present sample, there were no statistically significant associations with participants' BMI, eating disorder symptomatology as assessed with the EAT-26, and with hunger (see Table 5, Table 6, and Supplementary Table 1).

Discussion
Using motion tracking in VR, we evaluated the presence of behavioural biases towards virtual high-and low-calorie food in relation to non-food control objects of different complexity (balls and office objects).Consistent with a pilot study [60], results demonstrate a behavioural advantage for food grasping to reverberate in explicit hand-tracking movements in contact and collection times in a VR scenario.In extension, this was evident for high-calorie food compared to simple non-food cues, but also compared to low-calorie objects.As we  Note.M and SD represent mean and standard deviation, respectively. 3Please note that a sensitivity analyses revealed that this effect was significantly larger during the first part or the experiment (241 ms) compared to the second part of the experiment (9 ms).
outline in the following, the conciliation of results from previous AAT studies with the exact behavioural patterns obtained in the VR scenario complements and augments a scientific understanding of eating behaviour in the sense of motivational interactions.
Regarding the difference between low-calorie and high-calorie foods, behavioural patterns in different paradigms are theoretically consistent at first.In the VR scenario, a faster collection of high-calorie food was observed compared to most conditions, including low-calorie food  (tomato, pear, apple) and non-food cues.Results from the AAT partially corroborate this interplay; for instance, obese patients with binge eating disorder even revealed an avoidance bias for low-calorie food [48].In behavioural measures, energy-dense food stimuli are considered more attractive than non-food stimuli in clinical and non-clinical populations [15,68].Particularly the exposure to energy-dense food seems to determine a stronger activation of salience mechanisms and of automatic approach tendencies.However, further AAT results for low-calorie food diverge from our results.In healthy individuals and snacking-bothered participants, neither general approach or avoidance tendencies were different for low-calorie vs. high-calorie food pictures in the AAT [36].The same was observed in an AAT with eating disordered patients and healthy controls, although responses in general were slower to high-calorie food pictures [27].There are further behavioural differences observed in other setups, regarding both approach and avoidance tendencies.For instance, contrary to our task, food biases were not detected in classifications of task-irrelevant features such as cue color in an AAT for bias measurement, but only if the food status was task-relevant [31].Moreover, it is important to discuss the role of the control category since we only observed significant differences for ball objects and low-calorie food.More complex non-food office objects were somewhat faster grasped and collected than non-food balls and low-calorie food, although comparisons were not statistically significant.Object shape and affordances may be still considered critical variables that affect simulated interaction, and their roles for food-related biases should be further explored in future studies.Finally, regarding avoidance, we also observed faster movement onset for high-calorie food in push movements, possibly reflecting early attentional processes.However, this evened-out later at object contact in push-trials.
Automatic approach behaviours towards high-calorie food are likely to be an important contributor to unhealthy consumption behaviours to the pathogenesis of different eating-related pathologies [20].Nevertheless, also normal-weight participants are typically biased towards food dependent on their current hunger state [6], as reflected in the overt behavioural measures of the VR task.Their motivational bias towards food also differs from patients with the restrictive type of anorexia nervosa across traditional and VR paradigms [59,67].In the affective manikin task, patients made less errors in trials that requested the avoidance of high-calorie food, compared to healthy controls [67].This pattern was attributed to reduced motivational saliency of food.Furthermore, continuous movement recordings in VR revealed constant avoidance of restricting patients with anorexia nervosa across go-trials and stop-trials in a study on inhibitory control, possibly indicating habit-like food avoidance [59].
Interestingly, subjective evaluations of high-calorie food objects followed a different trajectory than behavioural responses in the movement dynamics.Participants rated high-calorie food and office tools as more negative, but more arousing; in contrast, only responses to non-food cues were correlated with collection times.This result might suggest the independence of both subjective evaluations from approach movements in relation to food.Moreover, the observation is in line with dual-process theories that indicate independent bottom-up and topdown processes, a pattern that was very recently also observed in a similar task with deviating explicit ratings and movement trajectories towards high-calorie food [59].
We did not observe strong associations between the subjective experience of presence and behavioural performance in this study.This finding is interesting with regard to the general emphasis of VR in generating presence, often defined as a feeling of being there (place illusion) and the belief that events in VR are actually happening (plausibility) [28,63].Previous theoretical work suggests that presence is necessary to induce realistic behaviour [42].In contrast, our findings question the relevance of presence in the context of behavioural assessment and seem to indicate that automatic food-related behaviour may even emerge in lower conditions of presence.However, it will be important to delineate whether this observation generalizes to more complex forms of behaviour, to clinical populations, and to understand the limiting boundaries for realistic behaviour.Furthermore, a certain   association in clinical populations between emotional responses (e.g., anxiety) and presence was reported [9].Thus, future work can provide more guidance regarding the minimal requirements of VR assessments regarding presence.Food-related biases had been observed in several previous paradigms with better established psychometric properties and retest-reliability, such as the emotional Stroop task, dot-probe paradigm, implicit association tests and the approach-avoidance task [5,64].Early attentional processes are critically involved in eating disorders and demonstrated food-related avoidance and vigilance; however, only with the more recent advent of eye-tracking VR headsets, it is possible to record attentional processes concurrently to VR tasks [61].With the present study, irrespective of a possible attention bias for food in healthy individuals, we could demonstrate that food-directed hand movement display a reaction time advantage when coupled with the intention of grasping, but not when pushing.Thus, trajectory recordings enable insights into the late aspects of food-related behaviour and their dynamics.
Even if not yet validated in clinical populations, our results seem to confirm the utility of VR in assessing approach bias mechanisms towards food by exploring movement trajectories in a controlled high-fidelity environment.Eating disorders seem to be characterized by the presence of an imbalance between bottom-up and top-down processes.In particular, an impairment in the upstream regulation processes is likely to play a crucial role in determining behavioural biases [5,8].Accordingly, a recent fNIRS study showed food-related interactions in VR to be related to prefrontal hemodynamic changes [39].Given the recent surge in technological development and availability as consumer mass media, VR also presents a feasible and increasingly accessible technology for possible future trainings.However, before trainings or modification paradigms can be implemented, the exact mechanisms underlying stimulus fidelity should be clarified.Furthermore, first evidence hints at more sensitive assessments of food-associated behaviour with VR parameters compared to two-dimensional setups [39]; further systematic comparisons between different technical setups and their validity are outstanding.

Limitations
This research investigated push and grasp movements with a relatively extensive motion capture task and with multiple sets of food and control objects.In the present sample of normal-weight individuals, no effect of BMI was observed, which highlights the need for future studies in clinical groups with a better control or manipulation of craving and hunger.
We reported food-related effects on behavioural trajectories; however, the present work did not include a psychometric evaluation of the measures collected in VR, nor their retest-reliability.A previous comparison of a computerized task with a VR-based inhibitory control task with the same stimuli indicated a relatively low construct validity [61], suggesting that VR might capture unique aspects of behaviour.In a separate study on food-related decision processes, again, behavioural responses in a tablet and a VR implementation of the same task with identical stimuli revealed slightly different results [39].However, the present study was not designed to compare different measures of approach biases and the psychometric properties of VR assessment are currently subject to further research.A further limitation is that we did not assess the presence or history of eating disorders in a clinical interview.Furthermore, since realism was associated with collection times, future studies can more specifically assess perceived realism of the 3D objects.The current results are limited by the relatively unrealistic, cartoon-like stimuli that were employed.Stimulus matching is challenging in VR, but all employed stimuli were kept in a similar visual style, despite variability in their geometries and low-level features.More realistic stimuli, virtual environments, and higher-resolution headsets should be used in future studies [61].A further limitation is the degree of craving induced by virtual food; for instance, previous studies found comparable subjective craving for virtual compared to real food [29,65], but physiological responses were limited to real chocolate exposure [65].We can speculate that the multi-modal induction of craving might be particularly effective using additional olfactory displays.

Conclusion
The present study revealed faster collection of high-calorie food objects, which emphasize a previously described food-related bias specifically for grasping as opposed to pushing interactions.Grasping and collecting virtual high-calorie food was faster compared to low-calorie food and simple non-food cues.These results confirm the presence of bottom-up advantages in motor-cognitive behaviour for high-calorie food in a non-clinical population.Future research can investigate effects of varying the degree of fidelity in stimulus materials and systematically compare technical setups.

What is already known on this subject?
Previous results demonstrated an approach bias to high-calorie food in different experimental paradigms.Among those, two recent VR studies showed that food items were collected faster than neutral balls and that high-calorie food selection was associated with prefrontal cortex activity.

What this study adds
This study replicates the faster collection of high-calorie food objects in virtual reality and suggests its specificity irrespective of subjective evaluations.Systematic comparisons confirmed the relevance of calorie content and of the control category.
P.A.Schroeder et al.

Fig. 3 .
Fig. 3. Mean ratings and standard error of the mean of arousal and valence for all stimulus categories.

Fig. 4 .
Fig. 4. Movement trajectories for push-instruction (left column) and grasp-instruction (right column) as a function of time.Top row shows approach movement (panels A and B) and bottom row shows height movements (panels C and D).Visualizations are generated with functions from the mousetrap-package in R [26].Continuous trajectories were recorded at a sampling rate of 100 Hz.
P.A.Schroeder et al.

Table 2 ,
full trajectories are shown in Fig. 4. Data were not normally distributed and

Table 1
Ratings of valence and arousal.Ratings were collected after the experiment on a visual analogue scale (VAS) with two anchors in a computerized questionnaire.

Table 2
Means and standard deviations for Object Contact as a function of a Target Category and Instruction.

Table 3
Presence ratings from the IPQ, ranging from 1 (not at all) to 7 (very high).

Table 5
Rank-order correlations between subjective evaluations (valence, arousal), BMI, and collection times (RT) for the four object categories.

Table 6
Rank-order correlations between scores in the eating attitudes test (EAT-26), collection times (RT), and subjective ratings for the four object categories.
P.A.Schroeder et al.