Embodied Mental Rotation – Does It Affect Postural Stability?

Abstract The effect of different human body part stimuli in mental rotation tasks (MRTs) on postural stability was investigated in two dual-task experiments. There were significant differences within egocentric MRTs (Experiment 1, N = 46): Hand and foot stimuli tended to cause more body sway than whole-body figures and showed increased body sway for higher rotation angles in the MRTs. In object-based MRTs (Experiment 2, N = 109) different stimuli did not evoke different levels of body sway, but higher rotation angles led to higher body sway. Both experiments showed a stabilizing effect of MRTs compared to the control condition. Exploratorily analyses identified reaction time in MRTs as a significant predictor of body sway. The results suggest a heterogeneous impact of mental rotation on postural stability.


Introduction
T he relationship of mental rotation, the ability to imagine rotated objects in the mind (Shepard & Metzler, 1971), and motor ability is widely accepted (Voyer & Jansen, 2017). Especially the relationship between mental rotation and postural stability has been demonstrated by Hofmann and Jansen (2021). Since spatial abilities (e.g., mental rotation) play an essential role in daily living, their interference with other daily processes, like basic motor processes, is of great importance. This study will focus on the role of the embodiment of mental rotation stimuli (images of human body parts) on postural stability in more detail.

Postural Stability
Postural stability, also often referred to as balance, is the ability to control the body's center of gravity in relation to the support surface (Shumway-Cook & Woollacott, 2007). To quantify postural stability, usually, the Center of Pressure (CoP) is calculated using a force plate (Rhea et al., 2014). During standing, the CoP is the point representing the weighted average of the sum of the vertical ground reaction forces exerted by both feet onto the force plate (Winter, 1995). Typically, lower CoP displacements are referred to as a more stable stance (Palmieri et al., 2002). However, other approaches exist that consider some variability in the CoP signal to be a functional component of stance (Haddad et al., 2013). So, an upright stance, with high CoP-displacements over time, evaluated solely based on the CoPcourse, does not have to be a less stable stance if it does not involve a fall or loss of balance. Nonetheless, the approach described before is common (Rhea et al., 2015).

Mental Rotation
Mental rotation refers to the ability to imagine an object rotated in the mind (Shepard & Metzler, 1971). There are two types of mental rotation tasks: objectbased and egocentric tasks (Zacks et al., 2000). An object-based mental rotation task is represented by presenting two stimuli on a screen, where the right stimulus is a rotated or mirrored version of the left stimulus. Participants must decide whether both items are rotated, or in other words, are "the same" or mirrored, which means that they are "different." In an egocentric task, a common variation is that a human figure, raising one arm, is rotated. The participant must decide whether the right or the left arm is raised. One of the main differences between these two types of mental rotation tasks is that in an object-based task, the subject's position about the environment remains unchanged, and only the stimulus is mentally rotated. In contrast, participants change their perspective and imagine rotating themselves to solve the egocentric task. Accordingly, egocentric transformations cause a simulative rotation process of one's body (Kessler & Rutherford, 2010).
A crucial role in mental rotation research is the influence of the type of stimuli. The use of human bodies or body-parts as stimuli are direct consequence of what is known of "embodied cognition" (e.g. Lakoff & Johnson, 1999). Embodied cognition means that cognition arises from the need to control body's function in the natural environment (Wilson, 2002). Saying this, embodied cognition is based on the assumption that mental processes and physical processes are related to each other (Glenberg, 2010). Several taxonomies exist within the embodied cognition or grounded cognition theories as it is for example the one of common coding, internal models, or simulation theory, that operate on different levels and interfaces (Jansen, 2022). However, stimuli that provoke a motor response should facilitate cognitive tasks. Using body parts or human stimuli in mental rotation tasks has been often used to introduce embodied cognition in mental rotation tasks (Amorim et al., 2006;Voyer & Jansen, 2016). The importance of the stimulus material in mental rotation tasks was also investigated by Kosslyn et al. (1998). They showed in a PET study, that hand stimuli are mentally rotated different compared to cube stimuli because only the mental rotation of hands resulted in activation of brain regions that are related to low-level motor processes. Other studies also provide reasons to assume that body-part stimuli, mental rotation, and the body are related to each other. For example, the mental rotation of pictures of body parts correlated with the time needed by the participants to imagine the corresponding process of the body part (Parsons, 1987(Parsons, , 1994. The author also showed that participants could more easily imagine a biomechanically comfortable hand rotation than an uncomfortable hand rotation. He concluded that the subjects' strategy was to imagine their own hand as a comparison for the rotated stimulus. Amorim et al. (2006) demonstrated that the embodiment of the stimulus material, for example, adding a human head in classical cube figures, leads to better mental rotation performance. The authors attributed this to the spatial and motor mapping of one's body axes onto the axes of the stimulus to be rotated.
Furthermore, according to Kessler and Thomson (2010), egocentric mental rotation tasks are embodied differently from object-based mental rotation tasks since interactions between the direction of mental rotation and one's body posture only occur in egocentric and not in object-based rotation tasks.

Relationship between Mental Rotation, Motor Tasks, and Embodiment
It has been assumed that motor and mental rotations share standard processes (Wohlschl€ ager & Wohlschl€ ager, 1998), and the mental rotation process can be considered a covert motor rotation (Wexler et al., 1998). Furthermore, more recent studies have also demonstrated a relationship between mental rotation ability and motor expertise Pietsch & Jansen, 2012;Steggemann et al., 2011;Voyer & Jansen, 2017). In balance ability, significant correlations were shown with rotational performance (Jansen & Heil, 2010;Jansen & Kaltner, 2014). Kawasaki and Higuchi (2013) linked mental rotation, different types of pictures of the body, and postural stability tasks. They found that performing an egocentric mental rotation task with feet as stimulus material but not with cars as stimuli led to less sway when body sway was measured immediately after the mental rotation task. This effect could only be shown for the one-legged stance, not for the two-legged one. For other investigated body parts (hand stimuli), no significant correlation with body sway could be demonstrated in the single-leg stance (Kawasaki et al., 2014). Also, an effect of a mental rotation intervention on postural stability lasting up to 60 minutes in a single-leg stance was more effective for foot stimuli than for hand stimuli (Kawasaki & Higuchi, 2016). As a possible explanation, they suggest that the foot/ankle is more related to the stance and therefore plays a more essential role in postural control than hands. Consequently, they may also interfere more with postural control as hands as stimulus material. The study of Kawasaki and Higuchi (2016) shows the different roles of the "embodied" stimulus material on postural stability.
The direct relationship between motor and mental rotation tasks can be investigated with the dual-task approach: Dual-task paradigms are an experimental procedure in which participants are asked to perform two tasks simultaneously. Suppose a performance decline in one of the tasks compared to the single execution of this task can be shown. In that case, it is assumed that both tasks compete for similar information processing resources. Although the quiet upright stance seems to be one of the simplest everyday motor tasks, it has already been shown that it can be influenced by cognitive tasks performed simultaneously. Whether this influence leads to postural stabilization or deterioration is not clear. While in some dual-task studies on this topic, the simultaneous accomplishment of different cognitive tasks leads to a deterioration of postural stability (Mujdeci et al., 2016;Pellecchia, 2003;A. Shumway-Cook et al., 1997;Simoneau et al., 1999), other studies show that stabilization of body sway occurs (Andersson et al., 2002;Dault et al., 2001;M. C. Hunter & Hoffman, 2001;Potvin-Desrochers et al., 2017;Vuillerme et al., 2000). The interference in dual-task designs can be explained in the framework of the "bottleneck," the "capacity," or "crosstalk"-theories (Pashler, 1994).
Studies on dual tasks of mental rotation and postural stability are rare. Dault et al. (2001) investigated the influence of different working memory tasks in various postural tasks and the effect of an egocentric mental rotation task with a stick figure on the variability of body sway in the anterior-posterior and medio-lateral directions. They showed a reduction in body sway compared to a control condition (looking at a fixation point) but not to other working memory tasks. The same pattern was shown by Hofmann and Jansen (2021). For a double-legged stance, decreasing body sway during egocentric and object-based mental rotation tasks was shown for embodied and non-embodied stimuli (Budde et al., 2021;Hofmann & Jansen, 2021). Hofmann and Jansen (2021) also showed a significantly higher body sway for higher rotation angles in object-based mental rotation tasks but not in egocentric mental rotation tasks. Furthermore, they found that embodied stimuli (whole body figures) in mental rotation tasks lead to less postural sway than mental rotation tasks with the classical cube figures.

Aim of the Study
The overall goal of the study is to investigate the effect of different stimuli of human body parts (whole body, foot, hand) in mental rotation tasks (egocentric and object-based tasks) and a control condition (fixation cross) on a postural stability task that is performed simultaneously. We intended to replicate the results from Hofmann and Jansen (2021) and the results from Kawasaki and Higuchi (2016) in a dual-task situation. Furthermore, the relevance of human body parts stimuli in mental rotation tasks on postural stability must be investigated in more depth (Budde et al., 2021;Hofmann & Jansen, 2021;Kawasaki & Higuchi, 2016). As Hofmann and Jansen (2021) couldn't find a common pattern for egocentric and object-based mental rotation tasks on postural stability, both tasks will be investigated separately.

Hypotheses
Hypothesis 1: There will be less body sway during egocentric mental rotation tasks compared to the cognitive control task (fixation cross) (Hofmann & Jansen, 2021). Hypothesis 2: Posture-related body parts as stimuli in egocentric mental rotation tasks induce a more substantial decrease of postural sway compared to body parts which are not an essential part of postural control. a. The postural sway will be reduced in the egocentric mental rotation tasks with foot stimuli compared to egocentric mental rotation tasks with hand stimuli (Kawasaki & Higuchi, 2016). b. Postural sway for egocentric mental rotation tasks with whole-body stimuli will be less compared to egocentric mental rotation tasks with hand stimuli but not compared to egocentric mental rotation tasks with feet stimuli (Hofmann & Jansen, 2021;Kawasaki & Higuchi, 2016). c. Exploratively, a possible difference between egocentric mental rotation tasks with whole-body stimuli and egocentric mental rotation tasks with feet stimuli will be examined. d. Furthermore, a possible influence of the factor "rotation angle" will be examined exploratively.

Participants
Based on the results of Hofmann and Jansen (2021) and a small effect, according to Kawasaki and Higuchi (2016), Cohen's f ¼ 0.422059 was assumed. At an alphalevel of p ¼ .05 and a desired power of 1-beta ¼ .95, a power analysis (G Ã Power (Faul et al., 2007)) resulted in N ¼ 46. Exclusion criteria were diseases or injuries affecting the balance. In total, there were 47 participants recorded. Due to technical problems, one participant had to be excluded, and another participant was recorded to achieve N ¼ 46 for the data analysis (30 females and 16 males). The resulting sample comprised students from the study subject "Applied Movement Sciences." The acquisition of the participants took place through a newsletter. The female participants had a mean age of 22.63 (SD ¼ 2.85) and a mean height of 167.83 (SD ¼ 5.92) cm. The male participants had a mean age of 23.75 (SD ¼ 3.00) years and a mean height of 182.69 (SD ¼ 8.62) cm. The Ethical Board of the University of Regensburg approved the study and the study has been preregistered (https://osf.io/mxyn9). Participants were informed about the study's goal and the privacy policy concerning the data. All participants gave their written informed consent to participate in this study.

Material
Cognitive Tasks. The cognitive tasks (three egocentric mental rotation tasks and one fixation cross task) were presented on a laptop (HP Probook 650 G4, 1366 Â 768), using the software OpenSesame (Mathôt et al., 2012). The laptop was placed on a height-adjustable tripod to present the stimuli at the participant's eye level. The distance to the laptop was allowed to be freely chosen by the participant. Three blue movable screens were placed around the laptop to minimize possible visual distractions. Task instructions were to solve all tasks as fast and as correctly as possible. All stimuli were presented on a white screen. To answer the mental rotation tasks, participants were given one Bluetooth mouse in each hand, of which only the mouse in the right hand was switched on and could be used for answering. During the main experiment, no feedback was given; for the practice trials, there was feedback after each test on whether it was solved correctly. The three mental rotation tasks consisted of 1) a whole-body task, 2) a task with a hand stimulus, and 3) a task with a foot stimulus. In the first task, participants were presented with the front view of a female figure with blonde hair (Hofmann & Jansen, 2021;Kaltner & Jansen, 2018). The woman in the picture raised the right or left arm to the side, and the participants must either click the left mouse button when the left arm is raised or the right mouse button when the woman raises the right arm. In the second task, a hand stimulus providing a hand picture in dorsal view was used, the same stimulus as Kawasaki and Higuchi (2016). The task procedure was the same as for the whole-body task. Participants had to decide whether it was the right or the left hand and press either the right or left mouse button. In the third task, a foot stimulus was used, which was also taken from Kawasaki and Higuchi (2016) and showed a picture of a foot in dorsal view. Again, task instructions were the same, and participants had to decide whether it was the right or left foot. After every presented stimulus, a fixation cross was shown for 1000 ms. Mental rotation stimuli were introduced in the following angular disparities: 0 , 60 /300 , 120 /240 and 180 . The control condition showed a fixation cross in the middle of the screen (see Figure 1). Task instructions stated to look at the cross while doing the postural stability task.  (Hufschmidt et al., 1980;Palmieri et al., 2002). The selection of parameters was based on the results of Prieto et al. (1996) and the methodology used by Hofmann and Jansen (2021). All participants stood in a double-legged close stance on the force plate, wearing ultra-thin socks to ensure better hygiene while maintaining the barefoot condition. Before each trial, participants were positioned on the force plate by the experimenter, using a taped "T" of 3 cm wide tape on the force plate, so that the foot position was standardized for each subject (Richer & Lajoie, 2019). The participants were instructed to stand as still and upright as possible during the task. The arms should hang relaxed by their sides and the palms of their hands were to face the body. Only one type of cognitive task was solved per trial. One trial lasted at least 70 sec (Carpenter et al., 2001). To prevent fatigue, there was a 90 sec break between each run where participants sat down while rating perceived cognitive and physical load (see section: Perceived Cognitive and Physical Load).
Perceived Cognitive and Physical Load. To control for possible confounding factors, as perceived task difficulty or physical fatigue, a modified version of the German scale ASS ("Effort Scale Sport") was used. The ASSscale is a ten-level scale with complete level labels with a semantic meaningful sentence, including the adjective "effortful" at each level (B€ usch et al., 2021). Additionally, the color-coding as presented in B€ usch et al. (2015) was used. Here, colors are used to represent the stages of very light effort in shades of green up to increasingly greater effort, which ends in a strong red. To obtain a meaningful sentence for the tenth level of the scale for cognitive effort, only this level was slightly modified (from "so anstrengend, dass ich abbrechen muss" to "so anstrengend, dass ich nicht l€ osen kann"). During each 90 seconds break between blocks, the participants assessed the perceived cognitive and physical effort of the last block by saying in each case a number corresponding to the perceived level.

Procedure
The study was conducted as within-subject repeated measures dual-task design. All participants completed eight blocks of cognitive tasks while standing as still as possible on a force plate. All tests took place in a laboratory at the University of Regensburg, lasted about 45 minutes and had the same experimenter. During the experiment only the participant and the experimenter were in the laboratory. The experiment consisted out of a practice session and a main session. The trials of the main session were block-randomized with two blocks, each containing the four tasks (three mental rotation tasks and control task) once (every task for 70 seconds). In the practice session, participants were presented all tasks for 30 seconds in random order in their selfselected speed. The exemplary testing procedure is shown in Figure 2. Participants were allowed to start the trials themselves, simultaneously the measurement of the body sway started automatically. The mental rotation trials could exceed 70 seconds because the participant was always allowed to solve the last mental rotation task, which started within the 70 seconds.

Data Analysis
The CoP course over time was characterized by the four sway parameters mentioned above. To avoid movements, where the proband may have anticipated the start or ending of a block (M. C. Hunter & Hoffman, 2001), only the CoP-data within the 5 s to 65 s interval was evaluated. All sway parameters were calculated during the actual execution of the cognitive task. The time between two mental rotation tasks (1000 ms fixation cross) was not analyzed. Only correct responses to nonmirrored stimuli were included in the analysis, as angular disparity is not clearly defined for mirror-reversed stimuli (Jolicoeur et al., 1985). All data was first processed in Matlab (R2020b) and then imported in SPSS (IBM SPSS Statistics 26) for statistical analysis. For the comparison between all four cognitive tasks in general (all mental rotation tasks and the neutral condition with the fixation cross) a one-way repeated measures ANOVA was calculated for each sway parameter. To account for the angular disparity of the three different mental rotation tasks, a two-factor repeated measures ANOVA using the factors "stimulus type" (human, foot, hand) and "rotation angle" (0 , 60 /300 , 120 /240 , 180 ) was calculated for each sway parameter. The Friedman test was calculated for the ordinal data of the ASS-scale. The alpha-level was set to .05 for all analyses and the respective post-hoc tests. For each hypothesis post-hoc tests were Bonferroni corrected for all comparisons and all parameters, when needed (depending on which main effects or interactions showed significant results). The Bonferroni correction for hypothesis 1 resulted in a corrected alpha value of 0.0028. For hypothesis 2 it resulted in a corrected alpha of 0.0019. For the ordinal data of the ASS-scale, Dunn-Bonferroni tests were used as posthoc tests. To correct for violations of sphericity the Greenhouse-Geisser adjustment was used.
Additional analyses: After the data analysis as preregistered and described above was finished, a visual inspection of the mental rotation graph (Supplementary Material) showed similar patterns to the CoP-data of the parameters mean amplitude, maximum range of CoP course in anterior-posterior direction and maximum range of CoP course in medio-lateral direction. To test the role of "reaction time", the following explorative analyses were done successively and only for the parameter mean amplitude, to avoid a flood of analyses.
1. Pearson correlations, independent from the stimulus, between mean amplitude per angle and reaction time per angle. 2. Linear mixed modeling based on an additive model, where "reaction time" was added as fixed effect.
Linear mixed models were conducted using lme4 package (version 1.1.26; (Bates et al., 2015)) in R (version 3.6.1, (R Core Team, 2019). Model parameters were estimated by maximum likelihood estimation using bobyqa algorithm wrapped by optimx package (version 2021.10.12; (Nash & Varadhan, 2011)) as optimizer. Model fit was calculated by using likelihood ratio tests to compare models with and without the fixed effect of interest to a significance level of .05. For model building, all models started with a full random effects structure (StimulusType Ã Angle Ã ReactionTime) and based on the research of Barr et al. (2013) and Matuschek et al. (2017), the model complexity was reduced by stepwise reducing non-converging models and dropping of nonsignificant variance components at the significance level of .2. Non-significant fixed effects were further stepwise removed from the model, such that effects, which least decreased model fit were removed first and a model containing only significant fixed effects remained. This model was considered as the final model.

Mean Sway Values for the Four Cognitive Tasks
Mean values for each sway parameter over all four angles (0 , 60 /300 , 120 /240 , 180 ) per stimulus type were calculated for each participant (see Table 1). Prior to analysis, to handle missing data, the respective column mean was imputed for each stimulus type per angle and the fixation cross. None of the single data columns had more than three missing values. In total there were 1.84% of the data missing.
Concerning the mean amplitude of the CoP-course over time a repeated measures ANOVA with the factor "stimulus type" revealed significant differences between the four stimuli (F(1.103, 49.616) ¼ 381.233, p < .001, partial g 2 ¼ .894). Bonferroni adjusted post-hoc tests showed that only in the fixation cross condition the mean amplitude was significantly higher than in the egocentric mental rotation tasks (all p < .001). The maximum range of CoP-course in anterior-posterior direction also showed significant differences between conditions (F(1.048, 47.176) ¼ 411.369, p < .001, partial g 2 ¼ .901). Post hoc tests revealed only a higher range in anterior-posterior direction for the fixation cross condition (p < .001). Regarding the maximum range of CoP-course in medio-lateral direction also significant differences were found (F(1.064, 47.863) ¼ 395.157, p < .001, partial g 2 ¼ .898). Post hoc analysis revealed those differences only for the fixation cross condition, which had significantly higher values than the mental rotation conditions (p < .001). For the parameter sway velocity no significant differences between conditions were found (F(2.047, 92.100) ¼ 0.502, p ¼ .611, partial g 2 ¼ .011).

Sway Values for Different Rotation Angles
Mean Amplitude. For the parameter mean amplitude (see Figure 3(A)), the repeated measures ANOVA revealed no significant main effect for the factor "stimulus type", (F(2, 90) ¼ 2.414, p ¼ .095, partial g 2 ¼ .051) and a significant main effect for the factor "rotation angle" (F(2.196, 98.813) ¼ 27.750, p < .001, partial g 2 ¼ .381). Additionally a significant interaction between both factors could be shown (F(4.564, 205.363) ¼ 4.900, p < .001, partial g 2 ¼ .098). To check for the differences of the rotation angles dependent on stimulus, simple main effects were analyzed. For whole-body stimuli, no significant differences between any rotation angles could be found. For foot stimuli, the mean amplitude at the rotation angle 180 was significantly higher than the other rotation angles (all p < .001). For hand stimuli, the mean amplitude was higher at the rotation angles 180 and 120 than at 60 and 0 (all p < .001). To check for differences of stimulus types at specific rotation angles, another simple main effect analysis was conducted. No differences between stimuli were significant for any rotation angle.

Maximum Range of CoP in Anterior-Posterior Direction.
Regarding the maximum range of CoP-course in anterior-posterior direction (see Figure 3(B)), the repeated measures ANOVA showed no significant main effect for the factor "stimulus type" (F(2, 90) ¼ 1.677, p ¼ .193, partial g 2 ¼ .036) but a significant main effect for the factor "rotation angle" (F(2.219, 99.855) ¼ 23.245, p < .001, partial g 2 ¼ .341). The interaction between both factors revealed a significant result (F(3.541, 159.328) ¼ 3.971, p ¼ .006, partial g 2 ¼ .081). To check for the differences of the rotation angles dependent on stimulus, simple main effect analyses were conducted. For wholebody stimuli, no significant differences between all rotation angles were found. For foot stimuli, the maximum range of CoP-course in anterior-posterior direction at the rotation angle 180 was significantly higher than for all other rotation angles (all p < .001). The maximum range of CoP-course in anterior-posterior direction at the rotation angle 120 was significantly higher than at 60 (p < .001) and at 0 (p ¼ .0014). For hand stimuli, the maximum range of CoP-course in anterior-posterior direction was higher at the rotation angle 180 than at 60 (p < .001). At the rotation angle 120 , it differed significantly to 60 (p ¼ .0014). To check for differences of stimulus types at specific rotation angles, another simple main effect analysis was conducted. At an angular disparity of 0 , no differences between stimuli were significant. At an angular disparity of 60 , foot stimuli had a significantly lower maximum range of CoP-course in anteriorposterior direction than whole-body stimuli (p < .001).
For 120 and 180 , no significant differences between stimuli were found.

Maximum Range of CoP in Medio-Lateral Direction.
Regarding the maximum range of CoP-course in mediolateral direction (see Figure 3(C)) the repeated measures ANOVA showed no significant main effect for the factor "stimulus type" (F(2, 90) ¼ 2.195, p ¼ .117, partial g 2 ¼ .047) but a significant main effect for the factor "rotation angle" (F(2.046, 92.054) ¼ 25.671, p < .001, partial g 2 ¼ .363). The interaction between both factors revealed a significant result (F(4.187, 188.411) ¼ 5.573, p < .001, partial g 2 ¼ .110). To check for the differences of the rotation angles dependent on stimulus, simple main effect analyses were conducted. For whole-body stimuli, no significant differences in the maximum range of CoP-course in medio-lateral direction between all rotation angles were found. For foot stimuli, the maximum range of CoP-course in medio-lateral direction at the rotation angle 180 was significantly higher than at all other rotation angles (p < .001). For hand stimuli, the maximum range of CoP-course in medio-lateral direction was higher at the rotation angles 180 than at 60 and 0 (all p < .001). At the rotation angle 120 , it was significantly higher than at 60 and 0 (all p < .001). To check for differences of stimulus types at specific rotation angles, further simple main effect analyses were conducted. No differences between stimuli were significant for any rotation angle.

Controlling for Reaction Time of MR Tasks in Mean Amplitude
As Pearson correlations (Step 1, see Section Data Analysis) between the mean amplitude values and the mental rotation task reaction times at the corresponding equal rotation angles (0 (r ¼ .465. p ¼ .001), 60 (r ¼ .520, p < .001), 120 (r ¼ .560, p < .001), 180 (r ¼ .556, p < .001)) showed significant results, the additive version (Step 2, see Section Data Analysis) of the linear mixed model was fitted with reaction time only as fixed effect (StimulusType Ã Angle þ reaction time). The resulting model identified only reaction time (p < .001) as significant predictor of the mean amplitude of the CoP course. To explore other relationships, the linear mixed model with the full three-way interaction structure (StimulusType Ã Angle Ã ReactionTime) as fixed effects was fitted (Step 3, see Section Data Analysis). The resulting model also revealed only the significant influence of reaction time (p < .001). The additional influence of the reaction time is displayed in Figure 4.

Subjective Cognitive and Physical Effort
As each subject rated each condition twice, the median of these two ratings was calculated per condition. For the perceived cognitive effort, there was a significant difference between conditions (v 2 (3) ¼ 32.476, p < .001). Post-hoc tests showed a lower rating for the condition "fixation cross" than for all other conditions (p < .05). The stimulus "hand" was rated higher than human or foot stimuli (p < .05). The perceived physical effort also showed significant differences (v 2 (3) ¼ 45.721, p < .001). Post-hoc tests revealed a significantly higher rating for the "fixation cross" task (p < .001).

Summary of Experiment 1
Experiment 1 aimed to further work out the effects of human body parts in egocentric mental rotation tasks. The first hypothesis that mental rotation tasks reduce the extent of postural sway and in dual-task designs was confirmed. This is in line with a former study by Hofmann and Jansen (2021). Neither without nor with the inclusion of rotation angle consistent effects for the influence of the stimulus material on body sway can be seen for hypotheses 2 a -c. Only descriptively, slight tendencies for a reduced body sway can be found for solving tasks with foot stimuli compared to solving tasks with hand stimuli. Furthermore, it can only be partially identified descriptively that the body sway during solving human figures is higher than the body sway during foot stimuli and below the body sway during hand stimuli. Higher body sway values for higher rotation angles in Embodied Mental Rotation -Does It Affect Postural Stability? egocentric mental rotation tasks occurred only for foot and hand stimuli (Hypothesis 2 d). The additional analysis of the role of reaction time in egocentric mental rotation tasks identified it as a significant predictor for postural stability.

Experiment 2: Object-Based Mental Rotation Tasks
Since different rotation processes seem to be ongoing in egocentric and object-based mental rotation (Zacks & Michelon, 2005) and Hofmann and Jansen (2021) also did not find a consistent pattern of effects on postural stability for both types of tasks, the task types were analyzed separately in this study. The presentation of the order of the experiments in this study does not represent a temporal order of the experiments.

Hypotheses
Hypothesis 1: There will be less body sway during the object-based mental rotation tasks compared to the cognitive control task (fixation cross) (Hofmann & Jansen, 2021). Hypothesis 2: Posture-related body parts, as stimuli in object-based mental rotation tasks, induce a more substantial decrease of postural sway, depending on the angle, compared to body parts which are not an essential part of postural control. a. Postural sway will be smaller in object-based mental rotation tasks with foot stimuli compared with object-based mental rotation tasks with hand stimuli. (Kawasaki & Higuchi, 2016) b. Postural sway for object-based mental rotation tasks with whole-body stimuli will be less than objectbased mental rotation tasks with hand stimuli but not than object-based mental rotation tasks with foot stimuli. (Hofmann & Jansen, 2021;Kawasaki & Higuchi, 2016) c. Exploratively, a possible difference between objectbased mental rotation tasks with whole-body stimuli and object-based mental rotation tasks with feet stimuli will be examined.

Participants
Based on Hofmann and Jansen's results (2021), Cohen's f ¼ 0.2695805 was assumed. At an alpha level of p ¼ .05 and the desired power of 1-beta ¼ .95, a power analysis with G Ã Power (Faul et al., 2007) resulted in N ¼ 109. Exclusion criteria were diseases or injuries affecting the balance. Due to technical problems, four participants had to be excluded. After a check of the demographic questionnaire, six more participants had to be excluded as some information about current injuries met the exclusion criteria. For these ten excluded participants, another ten participants were recorded to achieve N ¼ 109 for the data analysis (63 females and 46 males). The resulting sample comprised participants from the study subject "Applied Movement Sciences." The acquisition of the participants took place through a newsletter. None of the participants had participated in Experiment 1 (see Section Participants, Experiment 1). The female participants had a mean age of 21.97 (SD ¼ 1.89) and a mean height of 167.97 (SD ¼ 7.23) cm. The male participants had a mean age of 22.13 (SD ¼ 1.92) and a mean size of 182.54 (SD ¼ 7.02) cm. The Ethical Board of the University of Regensburg approved the study, and it has been preregistered (https://osf.io/mxyn9). Participants were informed about the study's goal and the privacy policy concerning the data. All participants gave their written informed consent to participate in this study.

Material
Cognitive Tasks. The stimulus material, hardware, and software were the same as in Experiment 1. In contrast to Experiment 1, only object-based mental rotation tasks were examined (see Figure 5). Therefore, two stimuli were presented at the same time. Participants had to either click the left mouse button when the two stimuli were the same, meaning the right stimulus was rotated compared with the left stimulus, or the right mouse button when the stimuli were different, representing the right stimulus was mirror-reversed to the left stimulus. The left stimulus always remained in the same position.
Postural Stability Measurement. The measurement was the same as in Experiment 1.
Perceived Cognitive and Physical Load. Perceived cognitive and physical load was measured the same way as in Experiment 1.

Procedure
The procedure was the same as in Experiment 1 (see Figure 2).

Data Analysis
The preregistered data analyses and the additional exploratory analyses were the same as in Experiment 1. The Bonferroni correction for both hypotheses 1 and 2 resulted in a corrected alpha value of 0.0021.

Mean Sway Values for the Four Cognitive Tasks
Mean values for each sway parameter over all four angles (0 , 60 /300 , 120 /240 , 180 ) per stimulus type were calculated for each participant (see Table 2). Before analysis, the respective column mean was imputed for each stimulus type per angle and the fixation cross to handle missing data. None of the single data columns had more than five missing values. In total there were 0.71% of the data was missing.
Concerning the mean amplitude of the CoP-course over time, a repeated measures ANOVA with the factor "stimulus type" revealed significant differences between the four stimuli (F(1.072, 115.799) ¼ 944.212, p < .001, partial g 2 ¼ .897). Bonferroni adjusted post hoc tests showed that the mean amplitude was higher in the fixation cross condition than in the object-based mental rotation tasks (p < .001). The maximum range of CoPcourse in anterior-posterior direction also showed significant differences between conditions (F(1.041, 112.411) ¼ 1335.345, p < .001, partial g 2 ¼ .925). Post hoc tests revealed only a higher range in the anterior-posterior direction for the fixation cross condition (p < .001). Regarding the maximum range of CoP-course in mediolateral direction also significant differences were found, (F(1.087, 117.346) ¼ 1404.104, p < .001, partial g 2 ¼ .929). Post hoc analysis revealed those differences only for the fixation cross condition, which had significantly higher values than the mental rotation conditions (p < .001). The parameter sway velocity revealed significant differences between conditions, (F(2.193, 236.871) ¼ 5.937, p ¼ .002, partial g 2 ¼ .052). Post hoc tests showed a lower sway velocity for hand stimuli than for the fixation cross condition (p < .001).

Maximum Range of CoP-Course in Anterior-Posterior
Direction. Regarding the maximum range of CoP-course in anterior-posterior direction (see Figure 6(B)), the repeated measures ANOVA showed a significant main effect for both the factor "stimulus type" (F(2,216) ¼ 3.051, p ¼ .049, partial g 2 ¼ .027) and the factor "rotation angle" (F(2.311, 249.592) ¼ 135.263, p < .001, partial g 2 ¼ .556). The interaction between both factors did not reveal a significant result (F(4.683, 505.799) ¼ 1.230, p ¼ .295, partial g 2 ¼ .011). Post hoc analysis revealed no differences between stimuli but  there were significant differences for the maximum range of CoP-course in the anterior-posterior direction between all angular disparities (all p < .001).

Maximum Range of CoP-Course in Medio-Lateral
Direction. Regarding the maximum range of CoP-course in mediolateral direction (see Figure 6(C)) the repeated measures ANOVA showed no significant main effect for the factor "stimulus type" (F(2,216) ¼ 1.854, p ¼ .159, partial g 2 ¼ .017) and a significant main effect for the factor "rotation angle" (F(2.558, 276.257) ¼ 113.240, p < .001, partial g 2 ¼ .512). The interaction between both factors did not reveal a significant result (F(4.887, 527.795) ¼ 0.605, p ¼ .692, partial g 2 ¼ .006). Post hoc analysis revealed significant differences in the parameter maximum range of CoP-course in the mediolateral direction between all angular disparities (all p < .001).

Controlling for Reaction Time of MR Tasks in Mean Amplitude
As Pearson correlations (Step 1, see Section Data Analysis) between the mean amplitude values and the mental rotation task reaction times at the corresponding equal rotation angles (0 (r ¼ .290, p ¼ .002), 60 (r ¼ .431, p < .001), 120 (r ¼ .483, p < .001), 180 (r ¼ .470, p < .001)) showed significant results, the additive version (Step 2, see Section Data Analysis) of the linear mixed model was fitted with reaction time only as fixed effect (StimulusType Ã Angle þ reaction time). The resulting model showed a significant influence of reaction time (p < .001) and rotation angle (p ¼ .014). To explore further relationships, the linear mixed model with the full three-way interaction structure (StimulusType Ã Angle Ã ReactionTime) as fixed effects was fitted (Step 3, see Section Data Analysis). The resulting model revealed a significt influence of the interaction of rotation angle and reaction time (p < .001). The additional influence of the reaction time is displayed in Figure 7.

Subjective Cognitive and Physical Effort
As each subject rated each condition twice, the median of these two ratings was calculated per condition. For the perceived cognitive effort, there was a significant difference between conditions (v 2 (3) ¼ 74.294, p < .001). Post-hoc tests showed a lower rating for the condition "fixation cross" than for all other conditions (p < .001). The foot stimulus was rated higher than human stimuli (p < .05). The perceived physical effort also showed significant differences (v 2 (3) ¼ 74.859, p < .001). Post-hoc tests revealed a significantly higher rating for the "fixation cross" task (p < .001).

Summary of Experiment 2
Experiment 2 aimed to investigate further the effect of human body parts in object-based mental rotation tasks on postural stability. As in experiment 1, the already known result that mental rotation tasks cause a reduction of postural sway (Hofmann & Jansen, 2021) was confirmed (Hypothesis 1). Contrary to hypothesis 2 a), it could not be shown that foot stimuli have lower body sway values than hand stimuli. Also, whole-body stimuli were not shown to have lower body sway values than hand stimuli (Hypothesis 2 b). Also, for hypothesis 2 c), no difference could be demonstrated in body sway during solving mental rotation tasks with whole-body stimuli and with foot stimuli. There were slight tendencies that foot stimuli had the highest body sway. In the mentioned exploratory analyses, reaction time in object-based mental rotation tasks was identified as a significant predictor.

General Discussion
This study examined the effect of human body part stimuli on postural stability in the two standard versions of mental rotation tests, egocentric and object-based mental rotation tests.

Mental Rotation Effects on Postural Stabilization
In both experiments, mental rotation tasks were shown to reduce body sway relative to a neutral condition in which a fixation cross was viewed. This confirms, on the one hand, the results of Dault et al. (2001), who showed postural sway reduction for egocentric mental rotation tasks with stick figures, and on the other hand, the results of Hofmann and Jansen (2021), who demonstrated a postural sway reduction effect for both egocentric and object-based mental rotation tasks against the same neutral condition (fixation cross). In the broad field of dual-task studies, several explanations exist for a more stable stance during the simultaneous execution of a cognitive task. One explanation could be an attentional shift toward simultaneously solving the cognitive task, leading to more automated processing of the postural task (Donker et al., 2007). However, postural control is generally a very automatic process (Massion, 1992). Simple standing with a focus on the quiet upright stance is thereby somewhat artificial (Wulf et al., 2001) and thus may hinder the automated process. An alternative explanatory approach could be a co-contraction control strategy of postural muscles of the central nervous system, which may lead to tighter control of body sway (Dault et al., 2001). However, the authors postulate that the co-contraction control strategy is independent of task difficulty. Since the mental rotation tasks in this study differ in their difficulty, we cannot confirm this postulation. The results of our study indicate that task difficulty does matter, as our study tends to find higher body sway values at higher angles. In mental rotation tasks, higher reaction times and higher error rates are usually found for larger rotation angles, suggesting that these are more difficult. More difficult cognitive tasks require more cognitive resources, increasing dual-task costs and worsening postural stability (Pellecchia, 2003). However, the fact that the ASS scale for the fixation cross condition shows the lowest cognitive effort but the largest body sway values contradicts this. We think it is difficult to compare this condition with the mental rotation tasks because, as we discussed, it may be that the simultaneous processing of a cognitive task (in this study: mental rotation tasks) creates a shift of attention to the cognitive task, meaning an external focus, which leads to a minor variability in CoP-course (Donker et al., 2007). This shift of attention might not take place in the fixation cross condition.
Moreover, the perceived physical effort in the fixation cross condition is significantly higher than the perceived effort in the mental rotation tasks in both experiments. This would also support the theory that the focus in the fixation cross condition is on the physical activity instead of the deflected attention for the cognitive tasks. It is conceivable that the higher perceived physical effort and focus on the physical task is accompanied by more muscle activity, which could mean that no automatized control occurred here, and thus more sway is measured. However, this is purely speculative since we did not measure focus or any muscle activity. Future studies should thus either employ such measurements or different control conditions, which are expected to produce comparable attentional shifts as the experimental conditions. Furthermore, this study exploratively examined the influence of rotation angle in egocentric mental rotation tasks, as Hofmann and Jansen (2021) did not observe an increase in postural sway values at higher angles for egocentric mental rotation tasks with a whole-body figure. This effect is replicated in this experiment for the same whole-body figure from Hofmann and Jansen (2021). However, it does not carry over to egocentric tasks in general, as in this experiment foot and hand stimuli also had occasionally higher body sway values at higher rotation angles in egocentric mental rotation tasks. This study showed evidence for a higher body sway at higher rotation angles, independent of the stimulus for objectbased mental rotation tasks. This also confirms the results of Hofmann and Jansen (2021). They found higher body sway at the angular rotation of 180 compared to the other angles in an object-based mental rotation task with the same whole-body stimulus as in this study. Budde et al. (2021), who also tested whole-body stimuli in egocentric and object-based mental rotation tasks, found, on the one hand, evidence for higher body sway for the angle 135 compared to 45 , but on the other hand, also for lower body sway for 180 compared to 135 . So, there is no consistent pattern for body sway for increasing angle.
It should be noted that their results were shown for both mental rotation tasks together, and in their study, the whole-body stimulus was shown from the back. Thus there was one axis of mental rotation less (the longitudinal axis) than in this study. Therefore, both studies are not entirely comparable.
In line with the results of Hofmann and Jansen (2021), this study also shows that mainly the parameters measuring the range (mean amplitude, the maximum range of CoP-course in the anterior-posterior direction, and maximum range of CoP-course in medio-lateral direction) are affected by the simultaneous solving of the mental rotation tasks. The sway velocity, however, seems to be unaffected. Hofmann and Jansen (2021) offered a speculative approach, that this might relate to the constant mental rotation speed (Shepard & Metzler, 1971). DiDomenico and Nussbaum (2005) showed that with increasing mental demands, the Root mean square-distance of CoP-course decreased, but the sway velocity remained the same. Compared to the results in this study, this is either in line or contrary to our results: When we consider mental rotation tasks as increasing mental demands, compared to the fixation cross condition, the results of this study show the same pattern. However, our results show the contrary pattern when considering higher rotation angles within the mental rotation tasks as increasing mental demands. But differences in sway velocity between rotation angles with simultaneous mental rotation are found in Budde et al. (2021).

Embodiment Effects on CoP-Course
Mainly, the hypotheses of this study were based on the results of a study by Kawasaki and Higuchi (2016). They showed up to 60 min lasting postural sway reduction for unipedal standing after the egocentric mental rotation of foot stimuli. Based on the embodiment effect (Wilson, 2002), it is assumed that the mental rotation of body parts causes cognitive processes, which are used for motor imagery and motor execution of the specific body part (Parsons, 1994;Schwoebel et al., 2001) and therefore might interfere with actual motor tasks. As the foot plays an essential role in maintaining postural stability (Gage et al., 2004), we hypothesized that the findings from Kawasaki and Higuchi (2016) might result from the fact that only those body parts interfere with postural stability, which are directly involved in postural control. So, the aim was to transfer the postural stabilizing effects, after the egocentric mental rotation of foot stimuli, to a dual-task paradigm to determine the effects of simultaneous mental rotation (egocentric and objectbased) of body parts on postural stability. The descriptive analysis of the CoP-courses of the parameters Mean Amplitude, Maximum Range of CoP-course in the anterior-posterior direction, and Maximum Range of CoPcourse in the medio-lateral direction per angle suggests that an effect might exist in egocentric mental rotation. Although the graphs show a similar curve for foot and hand stimuli, the values of the foot stimuli are mostly slightly lower. However, the results of the ANOVA don't show any differences.
Since the results of Kawasaki and Higuchi (2016) are only shown in the single-leg stance, which causes greater sway values than the double-leg stance, it is possible that the calculated power, according to the results of Kawasaki and Higuchi (2016), of our study was too small. It is therefore critical to note that the attempt to directly apply the results of Kawasaki and Higuchi (2016) to a double-leg dual-task design, may have been too ambitious. Thus, repeating our experiments in a single-leg stance would be interesting. However, the results suggest that the interaction Kawasaki and Higuchi (2016) found does not necessarily transfer in the same way to other balance tasks, even when the same mechanism of embodiment should apply.
Whole-body stimuli were also thought to be more involved in postural control, or at least to show more body parts involved, than hand stimuli. This assumption could not be confirmed: It is conceivable that this pattern would also be more evident in the single-leg stance, as this is a more difficult postural task (Remaud et al., 2012).
In the egocentric tasks, the maximum range of CoPcourse in the anterior-posterior direction at 60 shows a significant smaller postural sway during the foot stimuli compared to the whole-body stimuli. Because of the many non-significant results, we would not overestimate this result, but it may indicate what the descriptive results suggest.

Influence of Reaction Time
For both experiments, we found positive correlations between the reaction time in mental rotation tasks and the body sway values of the parameter mean amplitude, so reaction time was exploratively included in the analysis. In both experiments, both versions of the linear mixed models identified reaction time as a significant influencing predictor. It was even the only explanatory predictor in the egocentric mental rotation tasks. The graphical illustrations and the significant correlations show that, between-subject, with higher reaction times in the egocentric and object-based mental rotation tasks also higher values for the mean amplitude, meaning worse postural stability, are measured. One possibility could be that in trials, that take a longer time to be solved, is simply more time to sway. However, this idea should be discarded, since range and velocity parameters should be nearly time-independent measures. Nevertheless, to control for different reaction times future dual-task experiments (especially: reaction time tasks vs postural stability tasks) might think about analyzing only equal periods of time from the CoP-course. It would be task specific how large these sections should be and at which point they should be cut out. In the case of mental rotation tasks, the problem arises that the exact temporal structure of the mental rotation phases is not clear (Heil & Rolke, 2002).
Another possible explanation for the significant influence of reaction time is that a more difficult cognitive task, in the context of mental rotation: higher rotation angles, leads to more postural sway because of higher attentional demands (Pellecchia, 2003). However, Kawasaki et al. (2014) found lower body sway values, in single-leg stance, significantly correlated with faster reaction times in following mental rotation tasks with foot stimuli. Since the mental rotation tasks were conducted after the postural stability measurements, it is excluded that these two tasks influence each other like in a dual-task paradigm. These findings suggest that it is not exclusively the actual amount of time performing the postural stability task but rather a person's specific mental rotation ability. S. W. Hunter et al. (2020) identified executive functions, with the Trail-Making-Test, as mediator between visual acuity and postural stability. One aspect of executive functions is working memory, which plays a role in visual-spatial cognitive abilities (Miyake et al., 2001) and, therefore, also in mental rotation (Linn Embodied Mental Rotation -Does It Affect Postural Stability? & Petersen, 1985). The impact of executive functions on postural stability is not clear yet, but it would be interesting if mental rotation training could improve postural control. First evidence might have come from Kawasaki and Higuchi (2016), who have already shown evidence for positive short-term effects (up to 60 minutes) of mental rotation on postural stability but concluded this was due to embodiment effects. Future studies could investigate the role of mental rotation training, besides training of executive functions in general, for improving postural control. Further insights could make mental rotation an interesting rehabilitation tool for people with impaired postural control.

Limitations
A limitation of the study is that only healthy young sports students who presumably have higher motor expertise than the general population are examined. Since a relation is known between mental rotation ability and motor expertise (Voyer & Jansen, 2017), this might influence the results. Furthermore, the bipedal stance as a motor task might be too easy to detect a clearer interference with mental rotation tasks, especially for this sample. In general, more extended measurements are needed for reliable postural stability measurements (Carpenter et al., 2001;Le Clair & Riach, 1996;van der Kooij et al., 2011). Therefore, it is a limitation that we only analyze concise time periods. More fundamental structures in the body sway can probably not be detected in this way. Also, the influence of reaction time may overshadow possible embodiment effects, as these might be much smaller. Ideally, future studies would identify stimuli with the same angle-reaction time relation to control for reaction time and test them for differences.

Conclusion
This study shows the effects of two types of mental rotation tasks (egocentric vs. object-based) with human body part stimuli on postural stabilization. In general, human body part stimuli led to more postural stability than neutral conditions in both mental rotation tasks. We conclude that mental rotation tasks, in general, affect postural stability in dual tasks. However, the different types of human body part stimuli had no consistent effects on postural stability between the mental rotation tasks. Additionally, reaction time in mental rotation tasks was shown to be a significant predictor of postural stability. For a better understanding of the influence of embodiment in mental rotation tasks on postural stability, this requires further research and maybe more difficult motor tasks to enlarge dual-task costs. Once the interaction between mental rotation and postural stability is better understood, mental rotation may offer an interesting approach for rehabilitating patients with impaired postural control.

ACKNOWLEDGMENTS
We would like to express our gratitude to the entire research team, particularly to Franziska Schroter and Markus Siebertz for their advice and helpful discussions regarding statistics. We would also like to extend our thanks to Tsubasa Kawasaki, who provided us with the original stimulus material from the studies on which our hypotheses are mostly based on.

Disclosure Statement
No potential conflict of interest was reported by the authors.

Funding
The author(s) reported there is no funding associated with the work featured in this article.

Data Availability Statement
The data that supports the findings of this study is available at https://osf.io/m2u5e/files. The study was preregistered at https://osf.io/mxyn9.