Psychology Expert Needed

EBSCO Publishing Citation Format: APA (American Psychological Assoc.):

NOTE: Review the instructions at http://support.ebsco.com.library.capella.edu/help/?int=ehost&lang=&feature_id=APA and make any

necessary corrections before using. Pay special attention to personal names, capitalization, and dates. Always consult your

library resources for the exact formatting and punctuation guidelines.

References

Day, B., Ebrahimi, E., Hartman, L. S., Pagano, C. C., Robb, A. C., & Babu, S. V. (2018). Examining the effects of altered

avatars on perception-action in virtual reality. Journal of Experimental Psychology: Applied. https://doi-

org.library.capella.edu/10.1037/xap0000192

<!--Additional Information:

Persistent link to this record (Permalink): http://library.capella.edu/login?url=http://search.ebscohost.com

/login.aspx?direct=true&db=pdh&AN=2018-52919-001&site=ehost-live&scope=site

End of citation-->

Examining the Effects of Altered Avatars on Perception-Action in Virtual Reality

By: Brian Day

Department of Psychology, Butler University;

Elham Ebrahimi

Department of Computer Science, University of North Carolina–Wilmington

Leah S. Hartman

Department of Psychology, Clemson University

Christopher C. Pagano

Department of Psychology, Clemson University

Andrew C. Robb

School of Computing, Clemson University

Sabarish V. Babu

School of Computing, Clemson University

Acknowledgement:

Virtual reality (VR) systems often represent the user as an avatar. Avatars are animated graphical representations of

people embedded within virtual environments (Lin, Rieser, & Bodenheimer, 2015). Past work has emphasized the

importance of providing an avatar for the accurate perception of virtual environments (Creem-Regehr, Stefanucci, &

Thompson, 2015). In some cases, an avatar may faithfully represent the anthropometric dimensions of the user (Lin,

Rieser, & Bodenheimer, 2012, 2013, 2015; McManus et al., 2011; Mohler, Creem-Regehr, Thompson, & Bülthoff,

2010). In other cases, the avatar may not be a direct reproduction of the user. This can occur mistakenly, or the

avatar may be purposively different (Jun, Stefanucci, Creem-Regehr, Geuss, & Thompson, 2015; Leyrer,

Linkenauger, Bülthoff, Kloos, & Mohler, 2011). For example, activities that are impossible to perform in the real world,

such as altering the length of limbs or size of one’s body in real time, can be done in VR, and simulated objects can

be incorporated into the body representation and treated as part of the participant (Slater, Pérez-Marcos, Ehrsson, &

Sanchez-Vives, 2008). Until recently, self-avatars were rarely used because of the technological difficulties in

rendering them realistically (McManus et al., 2011), but this has begun to change. As a result of the increasing use of

avatars, recent studies have investigated the effect that systematically altered avatars have on actors’ perception

and motor performance.

The perception of action capabilities in the virtual environment can be altered based on the presence or absence of

an avatar (Mohler et al., 2010; Renner, Velichkovsky, & Helmert, 2013; Ries, Interrante, Kaeding, & Phillips, 2009).

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

1 of 36 2/23/2019, 4:19 PM

Lin et al. (2013) had participants judge their ability to safely step off of a virtual ledge without falling. Participants with

an avatar that faithfully represented their own body dimensions estimated they could safely step off ledges of

appropriate height, whereas participants without an avatar judged they could safely step off ledges which were, in

fact, too tall to step off safely. Similarly, Lin et al. (2015) found that providing a self-avatar in VR generates action

judgments that are not significantly different from action judgments made in the real world. The authors concluded

that having participants perceive their action capabilities in VR with the presence of an avatar allows for accurate

judgment within a virtual environment. The results of Lin et al. (2013, 2015) suggest that providing faithful avatars

improves peoples’ ability to perceive critical information when deciding how to act in VR.

The present work investigates the effect of presenting participants with an avatar that extends their reach capabilities

by depicting an avatar arm that is longer than their own arm. It has been suggested that acting in VR with a

lengthened arm may be similar to acting in the real word with a hand-held tool that extends ones’ reach (Day,

Ebrahimi, Hartman, Pagano, & Babu, 2017). Humans frequently extend or augment their action capabilities through

tool use, which can be regarded as short-term changes to the body. Previous research has supported the idea that

objects attached to the body, such as tools, are perceived as functional extensions of the body (Bongers, Smitsman,

& Michaels, 2003; Day et al., 2017; Gibson, 1979; Wagman & Chemero, 2014). The extension of the body through

tools aids actors in both perceptual and behavioral tasks. The phenomenon of perceiving aspects of a distal surface

by means of a handheld tool is also known as extended haptic perception (Burton, 1993; Carello, Fitzpatrick, &

Turvey, 1992). Research has shown that the use of a tool that functionally increases reaching ability causes an

extension of the space that is perceived as reachable (Day et al., 2017; Maravita & Iriki, 2004). Interestingly, the

extension of perceived space persists after the actor has discontinued using the tool. The dispositions of

attachments to the body are perceived in the same manner as the appendages of the body itself, and can thus be

considered a type of proprioception (Pagano & Turvey, 1998), and numerous studies have supported this claim

(Bongers, Michaels, & Smitsman, 2004; Fitzpatrick, Carello, & Turvey, 1994; Wagman, Caputo, & Stoffregen, 2016;

Wagman & Taylor, 2005; Witt, Proffitt, & Epstein, 2005). Given that perception via a tool is not fundamentally different

than proprioception, then it might be expected that proprioception in VR with an avatar that lengthens ones’ arm, and

thus extends ones’ reach, may result in similar performance to reaching with a tool in the real world.

For successful action, one must be able to perceive what is or is not possible in the environment (Lessard,

Linkenauger, & Proffitt, 2009). As an example, it is frustrating to be unable to reach an object on a high shelf that is

just beyond arm’s reach. Fortunately, there are many things that can be done to obtain the object in question, such

as standing on a step stool or using a tool that extends one’s reach. This highlights the idea that successful action

requires that we be in tune with our action capabilities. Luckily, human beings are quite good at perceiving what we

can or cannot do in the environment, meaning we are quite good at perceiving affordances (Gibson, 1979). Findings

from the last 30 years of laboratory investigations have established that all types of organisms perceive their

surroundings in terms of the opportunities for action that are afforded (Cabrera, Sanabria, Jiménez, & Covarrubias,

2013; Heft, 1993; Mark, 1987; Wagman, Langley, & Farmer-Dougan, 2017; Wagman, Thomas, McBride, & Day,

2013; Warren, 1984; Warren & Whang, 1987). Additionally, the results of Lin et al. (2015) suggest that providing

faithful avatars allows people to more accurately perceive affordances in VR. Just as perceiving what is reachable is

an example of affordance perception (Carello, Grosofsky, Reichel, Solomon, & Turvey, 1989), so is perceiving what

is reachable with a tool (Day et al., 2017; Wagman & Morgan, 2010), as well as perceiving what is reachable in VR

using an avatar. However, the relationship between capabilities of an actor and environmental features continually

changes over short and long time scales.

Because the relationship constituted by an actor and their environment is constantly changing, affordances are

dynamic (Fajen, Riley, & Turvey, 2009; Wagman, Taheny, & Higuchi, 2014). Fortunately, our perception-action

system is flexible enough to adapt to changes in action capabilities (for recent reviews see Brand & de Oliveira,

2017; van Andel, Cole, & Pepping, 2017). According to Welch (1986)calibration is the semipermanent perceptual-

motor change that minimizes or eliminates a discrepancy between sensory modalities, within a sensory modality, or

the errors in behavior attributable to perceptual-motor discrepancies. Put differently, calibration is the process by

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

2 of 36 2/23/2019, 4:19 PM

which the execution of actions in response to the perception of affordances becomes scaled to the (changing)

relationship between environmental features and the actor’s action capabilities (Bingham & Pagano, 1998; Bingham,

Pan, & Mon-Williams, 2014; Coats, Pan, & Bingham, 2014; Day et al., 2017; Fajen, 2005; Withagen & Michaels,

2004, 2007). Generally, research into the process of calibration has highlighted the plasticity of the human

perception-action system in response to perceptual-motor discrepancies.

A hallmark study in the research on calibration was carried out by Mark (1987). The results of this investigation

showed that participants were able to judge whether chairs were sit-on-able or not accurately. Further, after a

manipulation of participants’ sitting ability, accomplished by attaching 10-cm-tall wooden blocks to their feet,

participants quickly calibrated to their new capabilities by altering their judgments of what was sit-on-able. Afterward,

Mark asked participants to judge how tall the block attachments were, and these estimates tended to be inaccurate.

The findings suggest that the human perception-action system has the ability to calibrate to altered capabilities

without knowledge of the specific alterations.

Research has shown that the visual perception of affordances for a particular behavior can become calibrated after

explicit practice performing that behavior (Bingham & Pagano, 1998; Day et al., 2017; Franchak, van der Zalm, &

Adolph, 2010; Wagman, 2012). Determining what is within reach of the body as humans interact with hand held tools

and graspable objects of various dimensions is a task that requires constant calibration. Once a tool or object is

grasped, the capabilities of an actor change. Specifically, what is now within reachable space changes. For example,

when using a tool, action possibilities in the environment generally increase and humans calibrate to this change in

capabilities quite readily (Day et al., 2017; Witt & Proffitt, 2008). Ebrahimi, Babu, Pagano, and Jörg (2016; Day et al.,

2017) urged future researchers to investigate what may be a similar effect of manipulating an immersive self-avatar

in VR, and examine its effects on human reach actions.

Recent research has investigated whether calibration occurs in the same manner in VR as it does in the real world.

Altenhoff et al. (2012) studied the effect of visual and haptic feedback on depth estimations in VR. Participants who

received visual and haptic feedback made more accurate distance estimates after the calibration phase, suggesting

that calibration of depth estimates can occur in VR. Ebrahimi et al. (2014; Ebrahimi, Altenhoff, Pagano, & Babu,

2015) have shown that participants calibrate to perturbed visual distances when their visually presented end effector

is shown to be nearer. It appeared to participants that they were underestimating their reaches and thus after

feedback they began to overestimate their reaches. Similarly, Ebrahimi et al. (2016) found that depth judgments in

VR are more accurate when scaled to visual and haptic feedback during closed-loop reaches than depth judgments

made in an open-loop manner in the real world. This finding is important because it suggests that visual feedback is

necessary for the calibration of actions in VR, and that congruent visuo-haptic feedback is most effective for

calibration. Collectively, past work has revealed that the presence of accurate visual feedback alone is sufficient for

the calibration of reaching actions in VR (Altenhoff et al., 2012; Ebrahimi et al., 2014, 2015, 2016). Interestingly, none

of these studies incorporated a fully rendered avatar for the user in VR.

Jun et al. (2015) investigated the perception of action capabilities in VR with avatars that were altered in some way.

Their participants were tasked with judging the width of a gap in virtual reality and were also asked to judge whether

they could safely cross that gap. All participants were shown only the disembodied feet of an avatar. In the small foot

condition, the presented feet were only 50% of the standard American male foot, and in the large foot condition the

presented feet were 200% of the standard foot. The researchers found that when participants viewed the

environment with large feet they judged distances as being relatively shorter and indicated that they could step over

relatively wider gaps in comparison to the small feet condition. In addition to showing an effect on judged distances

in virtual reality, the authors demonstrated that the perception of action capabilities can be altered via manipulations

in the size of portions of an avatar (Jun et al., 2015; Leyrer et al., 2011; Lin et al., 2012; Linkenauger, Leyrer, Bülthoff,

& Mohler, 2013). Thus, it seems that by manipulating an avatar, perception of affordances in a virtual environment

can be altered, further evidencing that calibration occurs in VR.

The present experiments investigated this possibility by asking participants to make reaches in VR with a self-avatar

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

3 of 36 2/23/2019, 4:19 PM

possessing arms that were longer than their own arms, and comparing their performance with participants who

reached with a faithful avatar. In the first experiment the participants performed three phases in VR; a pretest

baseline phase with no feedback, a calibration phase with feedback about the results of their reaches, and a posttest

phase without feedback. The participants in the second experiment performed the same three phases, but with both

the pretest and posttest occurring in the real world and only the calibration phase occurring in VR.

In both experiments it was expected that the calibration effect of the lengthened avatar arm would carry over to the

posttest phase, and performance in the posttest would differ from the pretest for the altered avatar condition. It was

also expected that in the absence of visual feedback during the posttest, the calibration effect would disappear as

the participants revert to performing in accordance with the actual length of their arms. This will be because in the

absence of vision the limbs’ dispositions are perceived via kinesthesis alone (Pagano & Turvey, 1995, 1998; van de

Langenberg, Kingma, & Beek, 2008). This effect of kinesthesis (also referred to as dynamic touch or the muscular

sense) has been shown to be independent of visual perturbations via prism goggle adaptation (Riley & Turvey,

2001).

The findings of these studies have significant ramifications for understanding the malleability of the body schema in

VR and whether such malleability carries over to the real world. The present research also has implications outside

of virtual environments. For instance, the present work is similar to the idea of accepting a limb that is bigger (or

smaller) than your own limb, such as when amputees receive artificial limbs. These artificial limbs may or may not be

the exact same size as the limb that they are intended to replace. Further, the present work has ramifications for

robot teleoperation, in that if humans can readily use avatar limbs that are longer than their own, then the same

phenomena might hold true for the teleoperation of robotic limbs that are very different from the more familiar human

limbs (e.g., Moore et al., 2006). In addition, the findings of these experiments further our basic understanding of the

malleability of the body schema in general.

Experiment 1

Experiment 1 addressed the following questions: If an avatar possesses different anthropometric dimensions than

your body, can your perception-action system quickly calibrate to the dimensions of the altered avatar when

attempting a simple action? If participants are able to calibrate to the altered dimensions and action capabilities of

the lengthened avatar, how long does this effect persist after the removal of feedback concerning the presence of a

lengthened avatar? Experiment 1 contained two primary avatar types (altered avatar vs. normal avatar) and utilized

three blocks of experimental trials (pretest, calibration, and posttest). The tasks that constituted each block involved

participants performing reaches to virtual targets at various distances.

Hypotheses

The current study had four primary hypotheses. First, based on previous findings involving the use of a tool that

similarly extended ones’ reach (e.g., Day et al., 2017), it was predicted that calibration to an altered avatar would

occur but it would not be instantaneous. Rather, we expected a linear improvement to a critical point in error over

trials in the calibration phase. Second, based on the findings in the calibration literature that have revealed

malleability of the body schema over periods of brief exposure to altered action capabilities, it was predicted that

calibration to the altered avatar with feedback would occur more quickly than reversion from the altered avatar back

to the arm’s actual length without feedback. This would be evidenced by steeper improvement in error across trials in

the calibration phase as compared to the posttest within the altered avatar condition. Third, it was expected that

reversion back to the user’s normal body dimensions would occur in the posttest. Reversion was defined as reaching

errors that match those of the normal avatar group. The fourth hypothesis pertains to effects of calibration carrying

over to posttest performance before this reversion takes place. Participants in the altered avatar group were

expected to calibrate to reaching with a lengthened arm that allowed them to bring the end of the avatar arm to the

virtual target by physically reaching to shorter distances. Thus, it was predicted that in the posttest they would exhibit

greater underreaches and attempt reach to more unreachable targets compared to the normal avatar group.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

4 of 36 2/23/2019, 4:19 PM

Method

Participants

Twenty-eight undergraduate students (22 females and six males, age M = 18.68, SD = 0.72) from Clemson

University participated in this experiment. The study was performed with approval of the Institutional Review Board

of Clemson University. Data from two participants were discarded because of a malfunction of the tracking system.

Participants were required to be right-handed as all equipment used was for right-handed participants. All

participants received credit in their psychology courses in exchange for participation. As participants entered the

testing area they were given a brief overview of the purpose of the experiment and informed consent was obtained.

Participants with a history of stroke or epilepsy were ineligible to participate. If participants needed glasses or

corrective lenses they were asked to wear those while participating. Participants were administered tests for visual

pathologies (such as refractive error or stereo blindness) before completing any trials. If participants failed these

tests they were unable to participate. Participants were randomly assigned to either the altered avatar condition or

normal avatar condition.

Design

Experiment 1 utilized a 2 (Avatar Type: Altered avatar vs. Normal avatar) by 3 (Phase: Pretest, Calibration, Posttest)

mixed groups design. Avatar type was a between-subjects variable, and phase was a within-subject variable. The

normal avatar condition involved use of an avatar’s arm that was directly proportional to the dimensions of the user’s

own arm. The altered avatar condition involved use of an avatar whose arm length, and thus reaching capabilities,

were increased by 30 cm.

Materials and apparatus

Figure 1 depicts the apparatus that was used to present the virtual environment. Participants were seated in a

wooden chair, which was situated approximately 20 cm from the edge of the wooden table. The tabletop was 50 cm

wide by 130 cm long, and was 76.2 cm tall (which is a standard table height). The center of the table was aligned

with the midpoint between the center of the participants’ head and their right shoulder. Participants were outfitted

with five Pohlemus sensors on the forehead, neck, right shoulder, right elbow, and on the hand-held tool. Aside from

the sensor on the forehead and on the tool, the other three sensors were all placed on the bony protrusions at those

points on the body. The base for the Pohlemus system was located underneath the table and out of view of the

participants. The virtual environment, which was a recreation of the actual room that the experiment was conducted

in, was displayed using a HTC VIVE head mounted display (HMD). The HTC VIVE is a binocular display system that

displays stereo information by presenting different information to each eye, with a combined resolution of 2160 ×

1200 pixels, a 90 Hz refresh rate, and a 110-degree field of view. Further details regarding this system can be found

in Ebrahimi et al. (2016). A virtual table and chair, whose dimensions and positions were the same as the real table

and chair, were placed centrally in the virtual room (see Figures 1, 2, and 3).

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

5 of 36 2/23/2019, 4:19 PM

Figure 1. Top: A view of the VIVE controllers, wrist worn mount, and both tools (normal and long). Bottom: The table

apparatus. This configuration was also rendered in virtual reality. Participants were asked to reach to targets

presented at the horizontal midpoint of the table.

Figure 2. The rendering of the virtual environment and the avatar as seen by participants. Each picture corresponds

to the virtual scene the participant would see for each of the four images in Figure 1, respectively. Starting on the top

left and going clockwise: (A) Both hands extended; (B) Altered avatar reaching for a target; (C) Resting position; (D)

Normal avatar reaching for a target.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

6 of 36 2/23/2019, 4:19 PM

Figure 3. The rendering of the virtual environment and the avatar as seen by experimenter. Each picture

corresponds to the virtual scene the experimenter would see for (A) resting position, (B) normal avatar reach, and

(C) altered avatar reach.

Participants were given a VIVE controller to hold and reach with. The VIVE controller was 26.5 cm long from base to

tip, 3 cm wide at the base of the handle, 5 cm wide at the top of the handle, 3 cm deep at the handle, and is 12 cm

wide at its widest point. The VIVE controller was mounted on a plastic mold affixed to the top of both of a

participant’s wrists, as seen in Figure 1. The wrist brace allowed for the wrist to remain in a consistent orientation

across all trials and across all participants. Mounting the VIVE controller on top of the wrist brace allowed the

experimenters to accurately model participants’ wrist position and hand position in VR. In this way, the participants

were presented with an avatar that accurately represented the orientation of their arms in the real world. Participants

were unable to see their shoulders or upper arm segment while reaching in VR. The plastic mold designed to hold

the controller also held a plastic rod with a rubber tip. When participants were reaching with the normal avatar in VR,

a 10 cm plastic rod was inserted into the mold. When participants were reaching with the altered avatar in VR, a 40

cm plastic rod was inserted into the mold while the 10 cm rod was depicted in VR, with the lengthened avatar arm

equaling the length of the normal arm plus the longer rod.

Participant’s head and hand movements in the real world were tracked and this information was used to update the

image displayed in the HMD so that the head and hand movements of the avatar were consonant with participants’

movements in the real world. Inverse kinematic (IK) algorithms were used to update the position of the forearm and

upper arm segments based on the position of the head, shoulder, and hand. Generally, IK can accurately predict the

position of the arm segments, yet there was a chance for error in the positioning of the virtual arm (however, this did

not commonly occur in the study). By outfitting participants with the wrist brace, the orientation of their wrist and

fingers was consistent for the entire experiment, and this position was maintained in the appearance of the virtual

avatars wrist and fingers as well.

In this study, in the normal avatar condition, participants observed the self-avatar holding the short tool. For

participants in the altered avatar condition, the IK algorithms elongated the upper arm and forearm segments by a

cumulative 30 cm. No other dimensions of the arm or hand were altered.

Participants were asked to reach for a visual target in virtual reality. For any trial, the target consisted of a virtual

representation of three luminous LEDs. The middle LED corresponded to the target distance. With the other two

LEDs luminous the length of the target area was 3 cm. In the pre- and posttest, targets were presented at 13

different distances, ranging from 20.5 cm to 121.5 cm. The difference between each target was approximately eight

cm. In the calibration phase, targets were presented at nine new distances, ranging from 17.4 cm to 114 cm, and the

distance between each target was approximately 10.5 cm. Every target was presented randomly, and each target

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

7 of 36 2/23/2019, 4:19 PM

distance was presented five times for a total of 65 reaches in the pre- and posttest, and 45 reaches in the calibration

phase.

Procedure

As participants entered the testing area, they were given a brief overview of the purpose of the experiment and

informed consent was obtained. Participants were administered the Stereo Fly Test (Stereo Optical, Chicago, IL),

which tested gross stereopsis and fine depth perception. Participants were then administered a test to determine

interpupillary distance (IPD) to help ensure the VIVE VR headset was properly adjusted to each participant. As

detailed by Willemsen, Gooch, Thompson, and Creem-Regehr (2008), the IPD test called for participants to look into

a mirror from a set distance and mark the location of each pupil in the mirror. The experimenter then measured the

distance between the two marks. The measured IPD was used to set the interocular distance on the VR headset

accordingly. By ensuring that the IPD of the VR headset was adjusted correctly for each participant, retinal disparity

and vergence would remain intact when participants were viewing the virtual environment.

All participants were asked to sit on the wooden chair at one end of the wooden table. Various motion sensors were

placed on the participant through the use of a long sleeve shirt. The sensors were attached to the shirt with Velcro,

and the cords for the sensors were strapped to the arm of the participant. The straps helped keep the shirt tight to

the arm of the participant so as to not interfere with their reach, and the straps helped to keep the wires of the

tracking system from pulling on the system. The physical location of each sensor was measured before and after

data collection to ensure that the sensors did not move over the course of the experiment.

Before putting on the HMD, the experimenter demonstrated the types of reaches that were appropriate in the

experiment. Then, after putting on the headset, but before any trials occurred, the participant engaged in three tasks

to familiarize themselves with being in VR. Participants were able to see their self-avatar in a mirror in the VR

simulation. The purpose of completing these tasks was also to induce a feeling of body ownership with the self-

avatar. Leyrer et al. (2011) have found that changes in presented eye height only affected judgments of distance

when ownership of the avatar was felt, suggesting that participants who did not feel as if the avatar was their own

were unwilling to calibrate to the altered dimensions. The tasks were based upon those frequently used by Slater in

his research on presence in VR (Banakou et al., 2013; Kilteni, Groten, & Slater, 2012; Maselli & Slater, 2013). The

first familiarization task required participants to bring their arms up to their side and move them around so they could

see how the movements of their body caused the avatar to move simultaneously. The second task required

participants to stretch their arms out straight in front of them and rotate their wrists. Lastly, participants were asked to

stretch their arms up over their head and move their arms around.

During data collection, participants were instructed to reach as quickly and as accurately as possible on each trial.

The major restriction placed on participants was they needed to remain seated (i.e., keep their weight on the seat

pan) and keep their feet flat on the floor during each reach. During the course of the actual reach participants could

engage their arm only, or they could engage their entire upper body (i.e., bending at the waist to reach farther).

Regardless of phase, each trial began with the participant resting their right forearm on the armrest of the chair and

their back against the back of the chair. Participants were instructed that this was the starting point for each trial. To

ensure uniformity in starting positions across participants, it was emphasized to participants that this starting posture

is critical for the study. Across all phases, participants were instructed to reach out as quickly and as accurately as

possible, and place the tip of the stylus as close to the center of the target at possible.

Pretest

In the pretest, participants were instructed to reach to the target that appeared on the table at various distances from

them. As part of each trial the participant was asked to make a judgment if they could reach the target or not. If the

participants answered in the affirmative (by saying “yes”), they were then instructed to initiate a reach. To ensure that

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

8 of 36 2/23/2019, 4:19 PM

participants could not see the target while reaching or receive informative feedback about their reach, at the initiation

of their reach, participants were shown a gray screen to simulate closing their eyes. After attempting to reach the

target, participants were instructed to return their hand and arm to the starting point to begin the next trial. If

participants did not believe they could reach the target they were instructed to say “no,” and the next target distance

was presented. Regardless of condition, all participants performed the pretest with a normal avatar (i.e., accurately

customized for their arm length). In this phase, participants only received haptic feedback when the controller they

are wielding in the real world contacts the surface of the table, but this feedback did not inform them about how close

their reach was to the target.

Thirteen different target distances were presented to each participant in the pretest. The distances ranged from

approximately 20 cm to 120 cm away from the participant. Each distance was separated by approximately 8 cm.

Targets were presented randomly, and each target distance was presented five times, for a total of 65 trials.

Calibration phase

After the pretest, participants completed the calibration phase. In the calibration phase, participants performed fewer

reaches to fewer targets than in either the pretest or posttest phases. Nine new distances that had not been

presented in the pretest (and were not presented in the posttest) were presented in the calibration phase. Each of

the nine targets was presented five times for a total of 45 trials, and all targets were presented randomly.

The task in the calibration phase was very similar to the pretest, except in this phase participants could see the result

of their reaches. After being shown a target, participants still gave a judgment if they could reach to the target or not.

Regardless of their response, participants were asked to reach to the target when the screen went blank. Once the

initial reach was made and participants kept the tip of the tool on the table, the virtual scene was restored to the

headset so participants could see the result of their reach. At this point, if the target was within reach, participants

were asked to adjust their reach to the center of the target area and hold there for one second before returning their

hand to the starting position. If the target was clearly out of reach participants returned to the starting position.

The primary manipulation of the experiment occurred in the calibration phase. Participants in the normal avatar

condition continued reaching with a normal avatar. However, participants in the altered avatar condition reached with

an avatar arm that was 30 cm longer than their normal avatar. For participants in the altered avatar condition, a

plastic rod that increased reach by 30 cm was substituted for the plastic rod that was used in the pretest. Participants

were not told of this functional increase in reaching ability. In the calibration phase, participants received haptic

feedback from when the (unseen) physical controller brace they were wielding in the real world contacted the surface

of the (physical) table. As stated above, once contact was made with the table, participants were shown the virtual

scene again and told to adjust their reach so the end of the virtually presented hand was in the center of the target,

thus receiving visual feedback as well.

Posttest

The posttest was identical to the pretest. Importantly, the experimenters ensured there was minimal delay (i.e., no

longer than 45 seconds for any participant) between the calibration phase and the posttest. By doing so, we hoped

to preserve the just modified action capabilities of the avatar for the posttest, as a long delay between these two

phases might cause the calibration to disappear.

Post data collection

After the conclusion of data collection, the experimenter again measured various aspects of the participant’s arm to

ensure that the positions of the sensors did not move over the course of the experiment. If a sensor was found to

have moved more than three cm, the data for that participant was deemed unusable and not included in the

statistical analysis (as mentioned in the participant section above). In addition, the participant was asked to perform

three reaches with their arm only (i.e., reaching their arm straight out as far as they could without engaging their

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

9 of 36 2/23/2019, 4:19 PM

shoulder or back) and three maximum reaches with their entire upper body (i.e., reaching as far as they possible

could and touching the table with no restrictions other than remaining seated in the chair with their feet flat on the

floor). Participants were given a brief questionnaire designed to measure the degree of body ownership they felt over

the avatar in VR (see the Appendix). The questionnaire contained items similar to those used in previous research

(see Maselli & Slater, 2013; Slater, Pérez-Marcos, Ehrsson, & Sanchez-Vives, 2009; Slater, Spanlang, Sanchez-

Vives, & Blanke, 2010). A manipulation check was also administered to participants. Participants were asked

whether they noticed anything odd that occurred during the course of the experiment. Lastly, participants were asked

about their previous use of VR simulations.

Results

Body ownership

There were no significant differences in feelings of body ownership between participants in the normal and altered

avatar groups. In response to the postdata collection manipulation check, 18 participants (64% of total participants)

indicated that they noticed something odd during the course of the experiment while 10 participants (36% of total

participants) indicated that nothing seemed odd. Of those 18 participants who responded that they noticed

something odd during the experiment, only seven participants (39% of yes responders) mentioned something about

the arm of the avatar being extended, manipulated, or larger. Common responses were “color of the eyes and skin

was off,” “it looked weird but not sure what,” or “arms were longer.” Broken down by avatar type, of the 13

participants in the normal avatar group, six participants (46%) indicated they noticed something odd and seven

participants (54%) said they did not notice anything. Of the 15 participants in the altered avatar group, 12 participants

(80%) indicated they noticed something odd and three participants (20%) said they did not notice anything. Of those

12, six participants (50%) specifically mentioned something about the arm of the avatar being extended,

manipulated, or larger.

Transformation variables

Figure 4 demonstrates the raw data in terms of reached distance (the distance to which participants reached with the

tip of the tool) and presented target distance. The overall data are shown, as well as the data for each phase.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

10 of 36 2/23/2019, 4:19 PM

Figure 4. Estimated distance as a function of presented distance (clockwise from top left) (A) overall, (B) pretest, (C)

calibration phase, and (D) posttest. The solid black line in each graph represents perfect performance (y = 1x + 0).

Absolute error was calculated by taking the distance between the presented target distance and the estimated

distance for each trial (error = reached distance − presented target distance), where negative values indicate

underreaching and positive values indicate overreaching. Then, the absolute value of the error term was computed.

Categorical variables included condition (normal avatar group used as reference category), phase (pretest used as

reference category), and error direction (overreach used as reference category). Three additional (binary) variables

were created using information contained in this error term. First, negative error values were coded as 0

(underreach) and positive error values were coded as 1 (overreach). Second, a correct judgment term was computed

that evaluates whether participants correctly judged if the presented target distance was within their reach envelope.

The correct judgment variable takes into account whether the target was within reach or not on a given trial and the

participants’ response on that trial. Attempting to reach to targets outside of the reach envelope or not reaching to

targets that were within reach were coded as incorrect judgments (0). Reaching to distances that were within reach

and not reaching to targets that were out of the reach envelope were coded as correct judgments (1). Lastly,

regardless of correct judgment, if participants performed a reach, that trial was classified as “action taken” (1). If they

did not reach, that trial was classified as “no action taken” (0). This variable is referred to as action taken. For

example, if on a given trial the participant overreached the target by reaching to a distance further than the target

distance, that trial would be coded as 1 for overreaching, 1 for correct judgment, and 1 for action taken. Conversely,

if on a given trial the participant undershot the actual target distance by reaching too short to a target that was out of

reach, this trial would be coded as 0 for underreaching, 0 for correct judgment, and 1 for action taken.

Outlier analysis

For each analysis, individual outlier analyses for full models were conducted. Residuals were obtained,

standardized, and examined for any potential outliers that were outside of the normal distribution (Cohen, Cohen,

West, & Aiken, 2003). Outlier analysis was based on data visualization as well. Data points that were likely a result of

malfunctions in the tracking equipment and were not physically possible were removed for each specific analysis. In

all of the analyses less than 2% of the trials were removed due to outliers.

Hierarchical linear modeling

The intraclass correlation (ICC) of the intercept only model (null model) was used to assess the overall nesting within

participants for each of the main dependent variables (correct judgment and absolute error). Because of the

repeated-measures design of the experiment, variables had significant nesting within participants. For example, the

ICC of absolute error was approximately 15%. An ICC greater than 2–3% indicates nesting that demands a

multilevel modeling approach (Bliese, 1998; Heck, Thomas, & Tabata, 2010). Multilevel modeling offers a more

flexible approach to accurately modeling data produced in repeated-measures designs over traditional analyses

such as a repeated-measures ANOVA (Cohen et al., 2003).

As previously stated, predictor variables for Level 1 were collected at each trial occasion (e.g., presented distance,

presented distance quadratic, and action taken), and person-level predictors (Level 2) were collected for variables

such as condition. Interactions terms were also created which could be interlevel interactions (e.g., Level 1 by Level

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

11 of 36 2/23/2019, 4:19 PM

1 or Level 2 by Level 2) or cross-level interactions (e.g., Level 1 by Level 2).

In multilevel modeling, effect sizes, also known as pseudo-R , are indexed by a measure of percent reduction in

error variance. Level 1 error variance is indexed by a reduction in residual variance for Level 1 predictors. Level 2

error variance is indexed by a reduction in intercept variance for Level 2 predictors. Reduction in error variances for

cross-level interactions (Level 1 by Level 2) is indexed by the percent reduction in the Level 1 slope variance. The R

change is only calculated for significant effects, and the unique effects controlling for all other variables in the model.

Multilevel modeling relies on both general linear models and generalized linear models. Thus, multilevel modeling

can be applied to both normally and non-normally distributed outcome variables. Unless otherwise specified, all

analyses presented in the following paragraphs pertain to data collected during each phase (pretest, calibration

phase, and posttest).

The current study had four primary hypotheses. The first three hypotheses are contingent upon a significant three-

way interaction involving trial number moderated by phase and avatar type. An analysis of the three-way interaction

will precede an analysis of simple effects testing the form of that three-way interaction. Please refer to the hypothesis

section in the Introduction for information regarding how evidence for each hypothesis is obtained in MLM. First,

based on previous findings, in terms of absolute error as the primary dependent variable (estimated distance—target

distance), it was predicted that calibration to an altered avatar would occur but it would not be instantaneous.

Second, calibration to an altered avatar would occur more quickly than reversion back to the normal arm length. And

third, reversion back to the normal arm length would still occur in the posttest. The fourth hypothesis was not

contingent upon the significant three-way interaction, and thus it will be addressed separately.

Interaction testing Hypotheses 1–3

A multilevel model with absolute error as the outcome was conducted to evaluate the first three hypotheses. Avatar

type, phase, and trial number were entered into the model as predictors, as well as all appropriate interactions. The

L1 variables of phase and trial number both had significant random effects, but their interaction did not. The

presence of significant random effects indicates that there were individual differences for the effect of phase and trial

number when predicting absolute error. Phase, F(2, 13.57) = 6.14, p = .013 and avatar type, F(1, 22.42) = 7.84, p =

.010, had a significant main effects. The two-way interactions of phase by trial number, F(1, 2830.66) = 18.06, p <

.001, avatar type by phase, F(2, 26.80) = 4.53, p = .02, and avatar type by trial number, F(1, 25.03) = 6.14, p = .02

were statistically significant. The three-way interaction between avatar type, phase, and trial number was statistically

significant as well, F(2, 2830.66) = 14.85, p < .001 (see Table 1). The significant three-way interaction of trial number

moderated by phase and avatar type means that across the three phases, the two avatar types demonstrated

different absolute error trends across trial number.

F values, Significance Tests, and R2Δ for Absolute Error in Experiment 1

To further investigate the significant three-way interaction of avatar type, phase, and trial number, the data file was

2

2

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

12 of 36 2/23/2019, 4:19 PM

split by phase, and three two-way interactions between avatar type and trial number were analyzed separately for

each phase. The two-way interaction between avatar type and trial number was significant in the calibration phase,

F(1, 22.41) = 4.59, p = .043, and the posttest, F(1, 21.28) = 8.06, p = .01, but was not significant in the pretest.

Thus, participants in the altered avatar condition exhibited greater amounts of error in their reaches across the

course of trials within the calibration phase and the posttest than participants in the normal avatar group. This finding

indicates that participants in the altered avatar group in the calibration phase and posttest demonstrated a greater

disparity between the target distance and their reach distance across trial number, suggesting that the process of

calibration to an altered avatar and the process of reversion back to the normal arm length each occurred over the

course of dozens of trials (see Table 2).

Predicted Means and Standard Errors for the Avatar Type × Phase × Trial Number Interaction

Examination of Hypothesis 1

The first hypothesis was that calibration to an altered avatar would occur but it would not be instantaneous. This was

supported by the effect of trial number being moderated by avatar type in the calibration phase resulting in a steeper

negative slope in the altered avatar condition than the normal avatar condition. When investigating the significant

three-way interaction of phase by avatar type by trial number, the two-way interaction of avatar type by trial number

in the calibration phase was significant. As hypothesized, participants in the normal avatar group exhibited a less

steep slope of trial number predicting absolute error. Participants in the altered avatar group exhibited a negative

linear slope predicting absolute error across trial number in the calibration phase as predicted. Per each unit

increase in trial in the calibration phase, participants in the altered avatar group exhibited a slope of −0.08, which

indicates the hypothesized direction for the simple slope. This slope was significantly different than zero, t(19) =

−5.02, p ≤ 0.001, which was greater than the critical t value of −2.09 (see Figure 5).

Figure 5. Simple slopes for each avatar type of trial number predicting absolute error in the calibration phase.

Examination of Hypothesis 2

The second hypothesis stated that calibration to an altered avatar would occur more quickly than reversion back to

the normal arm length as evidenced by trial number being moderated by phase and a steeper slope in the calibration

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

13 of 36 2/23/2019, 4:19 PM

phase than the posttest for the altered avatar condition. Support for this hypothesis was obtained in the form of a

significant interaction between avatar type, phase, and trial number. First the data file was split by avatar type, and

then again by phase, to highlight the effect of trial number for each avatar type in each phase. This analysis revealed

a steeper negative slope for participants in the altered avatar group in the calibration phase (coefficient = −0.08) than

in the posttest phase (coefficient = −0.03). This finding indicates that calibration to an altered avatar and reversion

back to the normal arm length both occurred, but calibration to an altered avatar occurred more quickly than

reversion back to the normal arm length because of the steeper slope in the calibration phase.

Examination of Hypothesis 3

The third hypothesis, that reversion back to the normal arm length would occur in the posttest, was partially

supported as participants in the altered avatar condition in the posttest exhibited a negative slope predicting absolute

error across trials (coefficient of −0.03), but it was not significantly different than zero, t(19) = −1.80, p = .09, which

was not greater than the critical t of −2.09. This means that in the posttest, participants in the altered avatar group

exhibited decreasing amounts of absolute error over the course of the phase, suggesting that they were slowly

reverting back to acting based off of their stored body schema instead of the body schema of the altered avatar they

had calibrated to in the previous phase. Participants in the normal avatar condition exhibited an increase in absolute

error across trials (coefficient of 0.04), which was significantly different than zero, t(19) = 2.19, p = .04, which was

greater than the critical t = 2.09 (see Figure 6). This means that in the posttest, participants in the normal avatar

group exhibited increasing amounts of absolute error over the course of the phase, suggesting a decrement in

performance perhaps attributable to fatigue from completing the same task more than 150 times or a decrease in

accuracy attributable to the lack of feedback in that phase (Bingham & Pagano, 1998). However, it should be noted

that across all phases, participants in the normal avatar condition exhibited minimal amounts of error and the

posttest was the only phase where those participants exhibited an increase in absolute error over trial number that

significantly differed from zero. Evidence for reversion can be seen in the altered avatar group where participants

exhibited absolute error that was similar to that demonstrated by participants in the normal avatar condition at the

end of the block of trials. Importantly, this effect was not immediate, as reversion only occurred after close to 40 trials

had occurred in the posttest.

Figure 6. Simple slopes for each avatar type of trial number predicting absolute error in the posttest phase.

Hypothesis 4

The fourth hypothesis pertained to effects of calibration carrying over to posttest performance before reversion took

place. Participants in the altered avatar group were expected to calibrate to reaching with a lengthened arm that

allowed them to bring the end of the avatar arm to the virtual target by physically reaching to shorter distances. It

was predicted that in the posttest they would exhibit greater underreaches and attempt reach to more unreachable

targets compared to the normal avatar group. To investigate this hypothesis two multilevel models were run. The first

investigated absolute error as the dependent variable, and included phase, avatar type, and error direction as

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

14 of 36 2/23/2019, 4:19 PM

predictors. Phase, F(2, 2934.73) = 41.31, p < .001, error direction, F(1, 2939.67) = 9.37, p = .002, and avatar type,

F(1, 23.37) = 7.26, p = .013 all had significant main effects. The two-way interactions of phase by error direction, F(2,

2922.11) = 8.48, p < .001 and avatar type by error direction, F(2, 2926.35) = 5.59, p = .018 were significant as well.

The three-way interaction of phase by avatar type by error direction was also significant, F(2, 2922.11) = 4.25, p =

.014 (see Table 3).

F Values, Significance Tests, and R2Δ for Absolute Error

To further investigate the significant three-way interaction, the data file was split by phase and three two-way

interactions between avatar type and error direction were conducted (one for each phase). The interactions between

avatar type and error direction were significant for the pretest, F(1, 911.51) = 10.35, p = .001 and the posttest, F(1,

1089.49) = 19.08, p < .001. The interaction between avatar type and error direction in the calibration phase was not

significant. Means for this interaction can be seen in Table 4. Overall, the difference between avatar type for

underreaching and for overreaching differed in the pretest and posttest.

Predicted Means and Standard Errors for the Avatar Type × Phase × Over/Under-Reach Interaction

To further examine the simple effects for the significant two-way interaction between error direction and avatar type

in the pretest and posttest, the effect of avatar type was examined within the pretest and posttest separately. The

data file was further broken down by type of error (either an under- or overreach). Condition was not a significant

predictor for absolute error when participants underreached in the pretest, F(1, 15.31) = 2.46, p = .14, or

overreached in the pretest, F(1, 20.62) = 1.61, p = .219. Similarly, condition was not a significant predictor of

absolute error when participants overreached in the posttest, F(1, 21.66) = 0.018, p = .90. Together, these findings

mean that there was no significant difference between the conditions in predicting the amount of absolute error when

participants over or underreached in the pretest, and when they overreached in the posttest. However, condition was

a significant predictor for absolute error when participants underreached in the posttest, F(1, 28.97) = 9.60, p = .004.

According to the predicted means generated by the model, participants in the altered avatar condition (M = 5.77, SE

= 0.60) underreached significantly more than those in the normal avatar condition (M = 2.26, SE = 0.96). This finding

provides further evidence that calibration to an altered avatar in the calibration phase carried over to the posttest in

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

15 of 36 2/23/2019, 4:19 PM

that when participants in the altered avatar group underreached their reach to a target they did so by a margin

significantly greater than participants in the normal avatar group. This finding suggests that calibrating to the

lengthened avatar caused participants to underreach target distances to a greater extent in the posttest than the

group who used the normal avatar during each phase.

Analysis of judgment as the dependent variable

The second multilevel model was a binary logistic model that investigated correct judgment as the dependent

variable, with phase, avatar type, and trial number as predictors. In terms of predicting whether participants made a

correct judgment, there was a significant two-way interaction of phase moderated by condition (see Table 5).

Fixed Coefficients for the Binary Logistic Regression on Correct Judgment

Overall, participants in the altered avatar condition were more likely to make incorrect judgments in the posttest (a

probability of 0.11) as compared with participants in the normal avatar condition (probability of 0.04). That is, the

participants in the altered avatar condition were more likely to either reach to targets that were unreachable or fail to

reach to targets that were within reach. This finding also suggests that calibration to an altered avatar in the

intervening calibration phase carried over to the posttest, in that participants in the altered avatar condition continued

to reach to target distances that would have been reachable in the calibration phase with the altered avatar but were

no longer reachable in the posttest with the normal avatar (see Table 6).

Predicted Probability of Making an Incorrect Reach Judgment

Summary of results

Experiment 1 yielded several interesting and novel findings. As expected, calibration to an altered avatar occurred in

the presence of explicit feedback. Calibration was evidenced by the direction of the slope for trial number predicting

absolute error in the calibration phase for participants in the altered avatar group and by the carry over effects

demonstrated in the posttest by the altered avatar group (i.e., a large absolute error at the beginning of the posttest).

Then, in the absence of explicit informative feedback in the posttest, reversion back to the normal arm length

occurred as well. This effect was not immediate, as reversion only occurred after approximately 40 posttest trials.

Interestingly, calibration to an altered avatar in the calibration phase occurred more quickly than reversion back to

the normal arm length in the posttest phase. Lastly, participants in the altered avatar group exhibited greater

underreaches than participants in the normal avatar group in the posttest and made more incorrect judgments in the

posttest, further indicating that they had calibrated to the extended length of the avatar arm. Taken together, these

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

16 of 36 2/23/2019, 4:19 PM

results confirm that in immersive VR actors can calibrate to an avatar with different bodily dimensions than their own,

and that this calibration is not instantaneous. Then, once an actor has calibrated to the altered avatar, if explicit

feedback regarding their reaching behavior is removed they will revert back to their body schema with their original

arm length.

Experiment 2

VR simulations are ideally suited for training purposes because training programs can be implemented where real-

world training would be dangerous, expensive, or where scenarios would be too difficult to control (Rose et al.,

2000). Virtual training allows for complete control over the presentation of stimuli and the type of feedback that the

trainees receive. VR applications have been implemented in a variety of settings ranging from training airline pilots

(Lintern, Roscoe, Koonce, & Segal, 1990), firefighters (Bliss, Tidwell, & Guest, 1997), police officers (Bertram,

Moskaliuk, & Cress, 2015), and surgeons (Hyltander, Liljegren, Rhodin, & Lönroth, 2002). In Experiment 2, a transfer

of calibration paradigm was used to test whether the training effect observed in Experiment 1 would carry over from

VR to the real world.

It has been assumed that training in VR will transfer to real-world performance, but there is conflicting evidence to

support this claim. For example, the early findings from Kozak, Hancock, Arthur, and Chrysler (1993) suggested that

transfer from virtual training to the real world might not occur. When we act in the real world, we act as a unitary

system that integrates information from a variety of perceptual systems (kinesthetic, haptic, visual, etc.; Stoffregen &

Bardy, 2001). Yet in the virtual world, there is often an interruption between the information specifying kinesthetic and

visual invariants. Previous research has also shown differences in performance on tasks completed in the real world

compared to the same task completed in VR (Napieralski et al., 2011). Ebrahimi et al. (2016) showed that

participants were more accurate performing a reaching task in the real world compared with the virtual world, and

that the movement patterns differed in the two conditions. Bufton, Campbell, Howie, and Straker (2014) also showed

that when playing the same game (ping-pong) in the real world or in a virtual setting, participants exhibited different

movement patterns across the two modalities. These findings suggest that the difference in movement patterns may

interfere with using VR to learn a real-world motor skill.

Other studies have found evidence to suggest that training in VR does transfer to real world environments (Bertram

et al., 2015; Ganier, Hoareau, & Tisseau, 2014; Hyltander et al., 2002; Larrue et al., 2014; Regian, 1997; Rose et al.,

2000). Interestingly, some training programs do not represent the user with an avatar, whereas others use very low

fidelity avatars (Koritnik, Koenig, Bajd, Riener, & Munih, 2010), avatars that are not scaled to the dimensions of the

user (Bertram et al., 2015; Bufton et al., 2014), or disembodied avatar limbs (Ganier et al., 2014; Grabowski &

Jankowski, 2015). Thus, it is unknown whether the type of avatar used to represent the user has an impact on the

transfer of calibration of action capabilities to the real world. Must the avatar be scaled exactly to the dimensions of

the user’s biological body for the skills learned in VR to transfer to the real world?

As recommended by previous researchers (Bufton et al., 2014; Ebrahimi et al., 2014), this study will test whether

calibration in VR carries over to performance in the real world, and if the size of the avatar impacts the transfer of

calibration. Experiment 2 replicated the first experiment with one crucial change; participants performed the reaching

task in the real world in the pre- and posttests and in VR only during the calibration phase.

Hypotheses

The second experiment had three primary hypotheses. Based on previous findings, we first predicted that calibration

to a faithful avatar in the calibration phase will occur more quickly than calibration to an altered avatar. This means

that in the calibration phase participants in the normal avatar condition will have a smaller intercept and a much

flatter slope than participants in the altered avatar condition who will have a larger intercept and a steeper negative

slope. Second, we predicted that reversion back to ones’ normal body capabilities in the posttest will occur more

quickly in the normal avatar condition compared to the altered avatar condition. This means that in the posttest

participants in the normal condition will have a smaller intercept and flatter slope than participants in the altered

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

17 of 36 2/23/2019, 4:19 PM

condition, who will have a larger intercept and steeper slope. Third, we predict that reversion back to the user’s

normal body dimensions will occur in the posttest, and this reversion will be evidenced by participants in the altered

avatar condition demonstrating the same amount of absolute error in the posttest as participants in the normal avatar

condition.

Method

Participants

Twenty-three undergraduate students (15 females and 8 males, age M = 19.28, SD = 1.1) from Clemson University

participated in this experiment. The study was performed with approval of the Institutional Review Board of Clemson

University. Reaching data from one participant was discarded for failure to follow directions. The remaining details

regarding the Participants were identical to Experiment 1.

Design, apparatus, and procedure

All methods were identical to Experiment 1, except that participants completed their reaches in the real world in the

pretest and posttest while the calibration phase was completed in VR. The transitions from reaching in one modality

to another were made as brief as possible, and no transition lasted longer than three minutes. When viewing the

room in the real-world condition, the room was dark and only the target was illuminated, as compared with viewing

the room in VR where the entire room was illuminated. However, upon initiating any reach, targets both in the real-

world and in VR were removed, and the screen was blacked-out in VR until completion of the reach. This ensured

that participants engaged in reaches in identical conditions in the real-world and in VR. In the calibration phase, once

the initial reach was made and participants kept the tip of the tool on the table, the virtual scene was restored to the

headset so participants could see the result of their reach, and be provided with feedback in a manner identical to

the first experiment.

Results

Body ownership

There were no significant differences in feelings of body ownership between participants in the normal and altered

avatar groups. In response to the postdata collection manipulation check, 16 participants (73% of total participants)

indicated that they noticed something odd during the course of the experiment whereas six participants (27% of total

participants) indicated that nothing seemed odd. Of those 16 participants who responded that they noticed

something odd during the experiment, only six participants (37% of yes responders) mentioned something about the

arm of the avatar being extended, manipulated, or larger. Common responses were “the face looked weird,” “it

looked weird but not sure what,” “posture was odd,” or “arms were longer.” Broken down by avatar type, of the 11

participants in the normal avatar group, eight participants (73%) indicated they noticed something odd and three

participants (27%) said they did not notice anything. Of the 11 participants in the altered avatar group, eight

participants (73%) indicated they noticed something odd and three participants (27%) said they did not notice

anything. Of those 12, four participants (33%) specifically mentioned something about the arm of the avatar being

extended, manipulated, or larger.

Hierarchical linear modeling

Hierarchical linear modeling was employed after an outlier analyses was conducted as in Experiment 1, and less

than 2% of the trials were removed because of outliers. Figure 7 demonstrates the raw data in terms of estimated

distance (the distance to which participants reached with the tip of the tool) and presented target distance. The

overall data are shown, as well as the data for each phase.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

18 of 36 2/23/2019, 4:19 PM

Figure 7. Estimated distance as a function of presented distance (clockwise from top left) (A) overall, (B) pretest, (C)

calibration phase, and (D) posttest. The solid black line in each graph represents perfect performance (y = 1x + 0).

The ICC of the intercept only model (null model) was used to assess the overall nesting within participants for each

of the main dependent variables (correct judgment and absolute error). Because of the repeated-measures design of

the experiment, variables had significant nesting within participants. For example, the obtained ICC for absolute error

as the DV was approximately 7%.

The current study had three primary hypotheses, all of which are contingent upon an interaction of trial number

moderated by avatar type and phase. Please refer to the hypothesis section in the Introduction for information

regarding how evidence for each hypothesis is obtained in MLM. First, based on previous findings, we predicted that

calibration to a faithful avatar in the calibration phase would occur more quickly than calibration to an altered avatar.

Second, we predicted that reversion back to one’s normal body capabilities in the posttest would occur more quickly

in the normal avatar condition as compared to the altered avatar condition. Third, we predict that reversion back to

the user’s normal body dimensions would still occur in the posttest.

Interaction testing primary hypotheses

A multilevel model with absolute error as the outcome was conducted. Avatar type, phase, and trial number were

entered into the model as predictors, as well as all appropriate interactions. Phase, F(2, 2696.70) = 47.52, p < .001,

and trial number, F(1, 2692.90) = 5.49, p = .019, had significant main effects. The two-way interactions of phase by

trial number, F(2, 2685.72) = 3.80, p = .022, avatar type by phase, F(2, 2687.20) = 4.15, p = .016, and avatar type by

trial number, F(2, 2685.76) = 6.46, p = .011 were all statistically significant. The three-way interaction between

avatar, phase, and trial number was not statistically significant (see Table 7). Because of the nonsignificant three-

way interaction of trial number moderated by phase and condition, none of the three hypotheses were fully

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

19 of 36 2/23/2019, 4:19 PM

supported. In each phase, participants in the two conditions demonstrated similar absolute error across all trials,

suggesting that they calibrated at similar rates. However, the significant two-way interactions between phase and

trial number indicates there was a difference in the rate of calibration over each of the three phases. The significant

two-way interaction between phase and avatar type suggests that there was a difference in the mean absolute error

between the two conditions across each phase. The significant two-way interaction between condition and trial

number indicates that the altered avatar group and normal avatar group differed in the mean absolute error

demonstrated across trials in general.

F Values, Significance Tests, and R2Δ for Absolute Error in Experiment 2

Across all phases, participants in the normal avatar condition exhibited less absolute error (M = 7.68, SE = 0.66)

than participants in the altered avatar condition (M = 10.00, SE = 0.69). To further investigate the significant two-way

interaction between phase and avatar type, means were produced for each condition in each phase (see Table 8). In

general, there was no difference in absolute error between avatar types in the pretest and calibration phases.

However, participants in the altered avatar condition exhibited greater absolute error in the posttest (M = 10.26, SE =

0.80) than participants in the normal avatar condition (M = 6.77, SE = 0.77).

Mean Absolute Error for Each Condition Broken Down by Phase

A graph illustrating the significant two-way interactions of trial number moderated by condition can be seen in Figure

8. As can be seen in the graph, participants in the altered avatar condition showed decreasing amounts of absolute

error over trials. After splitting the file by avatar type, to investigate simple effects, it was revealed that the simple

slopes for the altered avatar and normal avatar groups were not significantly different from zero.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

20 of 36 2/23/2019, 4:19 PM

Figure 8. Two-way interactions of trial number moderated by condition.

Lastly, simple effects were identified for the phase by trial number interaction. The data file was split by phase to

examine the effect of trial number in each phase. Trial number was not a significant predictor of absolute error in the

pretest. However, trial number was a significant predictor of absolute error in the calibration phase, F(2, 731.42) =

23.06, p < .001, and the posttest, F(2, 975.93) = 6.43, p = .011. Across both avatar types in the calibration phase,

per each unit increase in trial number, participants exhibited a decrease in absolute error of −0.06 cm. Similarly,

across both avatar types in the calibration phase, per each unit increase in trial number, participants exhibited a

decrease in absolute error of −.04 cm.

Because of the nonsignificant three-way interaction of trial number moderated by phase and condition, none of the

three hypotheses was fully supported.

Additional analyses

To replicate the analysis performed on Experiment 1, additional multilevel models were run on Experiment 2 data.

The first investigated absolute error as the dependent variable, and included phase, avatar type, and error direction

as predictors. Phase, F(2, 2703.57) = 11.35, p < .001, and error direction, F(1, 2660.36) = 86.75, p < .001, had

significant main effects. The two-way interactions of phase by error direction, F(2, 2703.02) = 22.95, p < .001, phase

by avatar type, F(2, 2698.42) = 9.36, p < .001, and avatar type by error direction, F(2, 2661.24) = 9.01, p = .003,

were all significant as well. The three-way interaction of phase by avatar type by error direction was not significant,

F(2, 2703.02) = 2.29, p = .101 (see Table 9).

F Values, Significance Tests, and R2Δ for Absolute Error Regarding Under and Over-Reaches

Means for each of the two-way interactions can be seen in Tables 10–12. As indicated in the interaction of error

direction moderated by phase, participants underreached to a greater extent in the pretest and posttest as compared

to the calibration phase. Next, in the posttest only, participants in the altered avatar condition exhibited larger

amounts of absolute error than participants in the normal avatar condition. Lastly, regardless of phase, participants in

both avatar types exhibited similar amounts of error when underreaching, but participants in the altered avatar

condition exhibited greater overreaches than participants in the normal avatar condition.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

21 of 36 2/23/2019, 4:19 PM

Predicted Means and Standard Errors for the Phase by Error Direction Interaction

Predicted Means and Standard Errors for the Phase by Avatar Type (Middle) Interaction

Predicted Means and Standard Errors for the Avatar Type by Error Direction Interaction

Predicted Means and Standard Errors for the Phase by Error Direction Interaction

Predicted Means and Standard Errors for the Phase by Avatar Type (Middle) Interaction

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

22 of 36 2/23/2019, 4:19 PM

Predicted Means and Standard Errors for the Avatar Type by Error Direction Interaction

To further examine the simple effects for the significant two-way interactions, the effect of avatar type was examined

for each phase and for over/underreaches separately. The data file was broken down by phase and by type of error

(either an under- or overreach), respectively. Condition was not a significant predictor of absolute error for any of the

phases. Further, condition was not a significant predictor of absolute error when participants underreached.

However, condition was a significant predictor of absolute error when participants overreached, F(1, 16.72) = 5.31, p

= .034. According to the predicted means generated by the model (and not accounting for the effect of trial number),

participants in the altered avatar condition (M = 4.82, SE = 0.50) overreached significantly more than those in the

normal avatar condition (M = 3.21, SE = 0.49). Lastly, the data file was again split by phase to investigate the effect

of error direction. Error direction was a significant predictor in the pretest, F(1, 664.85) = 88.91, p < .001, and the

posttest, F(1, 994.49) = 34.51, p < .001, but not in the calibration phase. According to the predicted means

generated by this model, regardless of avatar type, participants underreached their reaches to targets to a greater

degree in the pretest (M = 9.14, SE = 0.53) than their overreaches (M = 3.22, SE = 0.64). The same pattern was

found in the posttest, where participants underreached their reaches to targets to a greater degree (M = 9.01, SE =

0.94) than their overreaches (M = 3.96, SE = 1.18). This finding suggests that when participants were acting and

receiving no feedback they tended to underreach their reaches to targets, and that in the presence of no feedback

participants reverted back to acting with their stored body schema. Further, these findings suggest that participants

had no difficulty calibrating to reaching with an avatar in the intervening calibration phase. This is likely to have

occurred as a result of the presence of explicit and informative visual feedback during the calibration phase.

Correct judgment

The second multilevel model was a binary logistic model that investigated correct judgment as the dependent

variable, with phase, avatar type, and trial number as predictors. In terms of predicting whether participants made a

correct judgment, there was a significant two-way interaction of phase moderated by condition (see Table 13).

Fixed Coefficients for the Binary Logistic Regression on Correct Judgment

Overall, participants in the altered avatar condition were more likely to make incorrect judgments in the posttest (a

probability of 0.22) as compared with participants in the normal avatar condition (probability of 0.15). That is, the

participants in the altered avatar condition were more likely to either reach to targets that were unreachable or fail to

reach to targets that were within reach. This finding also suggests that calibration to an altered avatar in the

intervening calibration phase carried over to the posttest, in that participants in the altered avatar condition continued

to reach to target distances that would have been reachable in the calibration phase with the altered avatar but were

no longer reachable in the posttest (see Table 14).

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

23 of 36 2/23/2019, 4:19 PM

Predicted Probability of Making an Incorrect Reach Judgment

Summary of results

The results of Experiment 2 suggest that calibration to an altered avatar in VR carries over to the real world when

participants once again act again with their normal body, that calibration to an altered avatar and to a normal avatar

in VR occur at the same rate, and that reversion back to the normal body when acting again in the real world also

occurs at the same rate for both avatar conditions. It was revealed that participants in the altered avatar group

exhibited more error in their behaviors, and made more incorrect judgments in the posttest than participants in the

normal avatar group. These results suggest that calibration to an avatar can occur in VR, whether the avatar is

faithful or altered, and that this calibration carries over to the real world for a period of time before the process of

reversion back to the normal arm length occurs.

General Discussion

The current work capitalized on virtual reality technology to directly manipulate participants’ arm lengths and to test

the effect of this alteration on reaching performance in both VR and the real world. Specifically, the experiments

investigated whether an actor can calibrate to the action capabilities of a virtual avatar that possessed different

anthropometric dimensions than themselves. In two experiments, feedback allowed participants to calibrate to an

avatar with lengthened arms, and this calibration carried over to a subsequent posttest phase where the feedback

was removed. Over the course of the posttest, and in the absence of the explicit feedback, the participants reverted

back to their original arm length. These results confirm that actors can calibrate to an avatar with different bodily

dimensions than their own, and that this calibration is neither instantaneous nor permanent. The reversion in the

posttest phase indicates the importance of continuous calibration to maintain accurate performance (Bingham &

Pagano, 1998).

Although the calibration to an altered avatar in the calibration phase required a number of trials, it occurred more

quickly than the reversion back to the normal arm length in the posttest. Thus, the change in performance occurred

more quickly with visual feedback than without it. These processes occurred in the same manner whether the

participants performed in VR or in the real world before and after the calibration to the altered avatar in VR, with the

calibration in VR carrying over to performance in the real world in a similar fashion to how it carries over to continued

performance in VR.

Calibration

The results from both experiments contribute to the existing literature regarding calibration. Studies explicitly

investigating the rate of calibration in various situations are lacking, and it is important to identify the amount of

experience that is required for effective calibration to occur (Bingham & Romack, 1999; Brand & de Oliveira, 2017;

Ebrahimi et al., 2015; van Andel et al., 2017). An important contribution of the current research is that we were able

to track the change in our primary DV (absolute error) over the course of each individual trial in the calibration and

posttest phases without aggregating the data. Of the 23 papers included in the systematic review of calibration

research by van Andel et al. (2017), fewer than half included any information regarding the rate of calibration. In this

way, we are able to make specific contributions to the existing literature regarding the rate of calibration.

Previous research has shown that calibration can occur very quickly in VR, such as one second for braking (Fajen,

2007) or a single manual reach for distance perception (Linkenauger, Bülthoff, & Mohler, 2015). In contrast,

participants in the altered avatar group of the present Experiment 1 needed upward of 45 trials in the calibration

phase before exhibiting an absolute error similar to that in the normal avatar condition. Thus, although calibration to

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

24 of 36 2/23/2019, 4:19 PM

an altered avatar can occur, the process is relatively slow compared with previously published results. However,

calibration in the real world often requires a similar extended period of time, such as 30 min for judging sitting and

stepping height (Mark, 1987; as described in Mark et al., 1990).

The results of the current research are mostly in accord with the main findings of van Andel et al. (2017) and Brand

and de Oliveira (2017). Their literature reviews revealed that the timeframe for calibration is variable, it is contingent

upon the aptness of the information available to the perception-action system, and when active perceptual

exploration is allowed calibration occurs relatively quickly. Perhaps occluding participants’ view of their avatar during

the reaches made during the calibration phase of the present experiments caused the process of calibration to occur

relatively slowly. However, after the initial reach was made the trial was not ended until the participants were able to

readjust their reach to be 100% accurate and this readjustment was made in full view. Thus, participants were

allowed to explore and perceive the results of their reaching movements in some way. Nonetheless, the differences

in the available visual information between the different phases of the present experiments may explain why

calibration to an altered avatar occurred more quickly than the reversion back to the normal arm length. Alternatively,

perhaps acting with a full avatar rendered in VR caused calibration to occur relatively slowly.

The current studies also highlight that there is a lack of necessary criteria in the literature to help define whether

calibration has occurred or not. In the given definitions of calibration there is no set criteria for what calibrated action

looks like other than to produce environmentally directed action that is informed by a scaling between action and

perception (Mon-Williams & Bingham, 2007). Generally, calibration is measured by the action judgments that are

produced in a posttest after calibration has supposedly occurred. For example, if given a tool that increases reaching

distance and participants calibrate to that tool, evidence for this is taken in the form of those participants perceiving

further distances to be within reach in a posttest even after the tool has been removed from the system. Yet this way

of measuring calibration does not investigate the actual process of calibration itself as it is occurring (see Bingham &

Romack, 1999). As revealed by the current Experiment 1, participants in the altered avatar condition consistently

exhibited greater absolute error in the calibration phase than participants whose action capabilities had not been

manipulated. This same pattern of results, where calibration is said to have occurred but participants still exhibit error

in their behavior, is typical of previous work as well (Kelly, Donaldson, Sjolund, & Freiberg, 2013; Kelly, Hammel,

Siegel, & Sjolund, 2014; Mon-Williams & Bingham, 2007; Scott & Gray, 2010). Even in studies where the authors

report that calibration occurred relatively quickly, errors in behavior compared with control groups are still evident.

Can we confidently claim that calibration has occurred when the action judgments between two groups are similar,

but there are differences between the groups in the error exhibited when carrying out the actual motor behaviors? At

the very least, the calibration seems to be less than complete. Moving forward, it is important to consider both action

judgments, defined as choosing to engage in a behavior or not (i.e., reaching to a target or not reaching to a target),

and movement control, defined as level of accuracy or error in a completed action, as criteria for determining

whether calibration has occurred and how successful the process of calibration was (Day et al., 2017). Many

previous investigations into the process of calibration have ignored data produced in the calibration phase, and only

relied on action judgments in the posttest to make claims that calibration has occurred. However, action judgments

are not necessarily correlated with the actual accuracy of the movement control when carrying out an action. It is

important to understand how actions performed in the calibration phase potentially differ from actions performed in

the pretest and posttest. When attempting to show that calibration has occurred, one method to indicate whether

there are any differences in movements when acting toward targets is through tracking exhibited error.

Researchers often differentiate between calibration and recalibration (Brand & de Oliveira, 2017). Recalibration

occurs after a system that has been previously calibrated is subjected to a perturbation that “renders the perception-

action link inaccurate, thereby initiating a rescaling of that link (p. 55).” However, evidence suggests that rather than

being in a “calibrated” state, with changes in perception or action capabilities requiring that the system adjust or

“recalibrate,” the perception-action system is in a constant state of continuous (re)calibration. The removal of

feedback regarding the outcomes of ones’ actions is itself a perturbation, and without continuous calibration the

system drifts, becoming progressively less accurate (Bingham & Pagano, 1998; Ebrahimi et al., 2016; Vindras &

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

25 of 36 2/23/2019, 4:19 PM

Viviani, 1998; Wann & Ibrahim, 1992; Wickelgren, McConnell, & Bingham, 2000). This is an important distinction,

because viewing the system as something that has been calibrated implies that the calibration mechanism is

something that might be switched off once it is determined that a specific perception-action system is performing

accurately and has thus been calibrated, and then the calibration mechanism must be switched back on when a

perturbation occurs and a persistent discrepancy has been detected. The evidence suggests, however, that there

may be no recalibration, just a constant state of continuously ongoing calibration. This would result in a more flexible

system that reacts very quickly to perturbations. This is also a more parsimonious view of the mechanism underlying

calibration, because there is no assumption that the system must detect how calibrated it is, whether or not

calibration is complete, whether or not calibration must be turned on or off, and perturbations need not even be

perceived as perturbations per se. Rather, the calibration mechanism may simply remain functioning as an integral

part of the perception-action system, continuously adjusting for any detected error in performance due to fatigue,

changes in environmental conditions, the incorporation of a tool, and so forth.

Human actors are quite adept at assimilating tools designed to extend reach into their body schema, and this alters

their perception of distance in reachable space (Day et al., 2017; Maravita & Iriki, 2004; Proffitt & Linkenauger, 2013;

Witt et al., 2005). One question is whether or not using an avatar with an extended arm in VR is akin to using a tool

in the real world. There are similarities between acting with a tool and an avatar because participants are able to

calibrate to the extension of their reaching capabilities and this calibration is somewhat enduring. Conversely, the

results of the present work suggest that calibration to an altered avatar in VR occurs more slowly than calibration to

a handheld tool in the real world, as the latter was identified previously (Day et al., 2017). An interesting idea for

future research would be to compare calibration to a lengthened avatar arm with calibration to a virtual tool held in a

faithful avatar arm, such that both result in the same functional reach length. The condition with the virtual tool and

faithful avatar arm in VR can then be compared with reaching in the real world with an actual tool. The present

findings, along with those of Day et al. (2017), suggest that all three conditions should result in successful

calibration, but with calibration in VR taking longer than calibration in the real world. The specific cause of this effect,

and how it might possibly be eliminated, would also be an interesting topic for further study.

The Embodied Action Schema

The present results have implications for conceptualizations of the body schema, namely that the body schema is

both malleable and stable at the same time. Generally, research into the process of calibration has highlighted the

plasticity of the human perception-action system in responding to discrepancies (Bingham et al., 2014; Bingham &

Pagano, 1998; Bourgeois & Coello, 2012; Day et al., 2017; Fajen, 2005; Mon-Williams & Bingham, 2007; Welch,

1986). Just as the process of calibration works to keep behavior constant across perturbations, such as sensory or

action based perturbations, the body schema is similar. It would not be functionally efficient for the body schema to

be permanent, as it is a fact that our action capabilities change on a regular basis, such as when we use tools or

across the life span due to changes like increases in strength and acquiring new skills.

The concept of a body schema has existed in the literature for over 100 years (Head, 1920; Head & Holmes,

1911–1912). A body schema is a stable representation of the body and its potential for action. It is typically believed

that the body schema is learned early in life and is based on information provided by the proprioceptive, vestibular,

and kinesthetic senses (Iodice, Scuderi, Saggini, & Pezzulo, 2015). Originally, Head (1920) postulated that any

changes to the body and its action capabilities are compared to a fixed body schema stored in memory. More

recently, it has been hypothesized that the body schema is neither innate nor learned. Rather, the body schema is

perceived (Pagano & Turvey, 1995, 1998). Accepting the hypothesis that the body schema is fluid and malleable

allows for a body schema that is continuously perceived as the body moves and is equipped with items such as

clothing, hand-held tools, and so forth (Pagano & Turvey, 1998). A body schema is malleable in that it can be

adjusted due to permanent or temporary changes made to the body or the body’s abilities. Over short time scales,

people equip themselves with tools that require calibration to new action capabilities (Day et al., 2017; Maravita &

Iriki, 2004). Over longer time scales, the body grows and develops, which requires calibration as well. Iodice et al.

(2015) found that although changes to bodily dimensions can result in adopting a new body schema, this is a

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

26 of 36 2/23/2019, 4:19 PM

relatively long and slow process. The process of calibration to the new capabilities, however, occurs much more

quickly. Based on this finding, it seems that the malleability of the body schema is not durable when changes to the

body are not permanent or cemented in to the perception-action system of the actor.

Crucial aspects of the body schema may be perceived online. In this sense, the body schema is fluid, and perceived

continuously as the limbs and their attachments change (Maravita & Iriki, 2004; Pagano & Turvey, 1998). A key

finding is that both limbs and hand-held objects are perceived through the same mechanism. That is, the same

principles underlie both the perception of attached objects and the perception of the body itself (Pagano & Turvey,

1995, 1998). This finding links our understanding regarding the malleability of the body schema and our

understanding of how attached objects are perceived and then incorporated into the body schema, because our

perception-action system treats them like they are part of the body. The body schema does not seem to distinguish

between objects and limbs, but rather it represents the effects of a tool as the lengthening of the arm (Cardinali et al.,

2009; Day et al., 2017; Maravita & Iriki, 2004; Sposito, Bolognini, Vallar, & Maravita, 2012). Perhaps updating the

body schema via calibration is the means by which the perception-action system maintains accuracy across such

alterations.

One goal of the present work was to test these ideas by investigating the malleability of the online body schema in

the context of reaching in VR with an extended avatar arm. The findings are consistent with the observation that

calibration is action specific, and that it involves a mapping from embodied units of perception to embodied units of

action (Bingham & Pagano, 1998; Coats et al., 2014; Pan, Coats, & Bingham, 2014). That calibration is action

specific means that the calibration of one action will not generalize to another action that involves a different unit of

action, such as between reaching, throwing or walking (e.g., Proffitt, 2008; Rieser, Pick, Ashmead, & Garing, 1995;

Witt, 2011; Witt, Proffitt, & Epstein, 2010). That calibration involves embodied units of perception and embodied units

of action means that rather than using external metrics like inches or centimeters, the relevant units are intrinsic to

the scale of the body and its action capabilities (Bingham & Stassen, 1994; Cutting, 1986; Gibson, 1979; Mantel,

Stoffregen, Campbell, & Bardy, 2015; Pagano, Grutzmacher, & Jenkins, 2001; Proffitt & Linkenauger, 2013). It also

means that what is specifically being calibrated is the mapping between these intrinsic units of perception and the

intrinsic units of action (Bingham et al., 2014; Bingham & Pagano, 1998; Pan et al., 2014). Following Day et al.

(2017), we believe it is appropriate to move away from using the traditional term body schema that implies a stable

entity, and replacing it with embodied action schema, which is more in accord with recent findings regarding its

malleability and the embodied nature of calibration (e.g., Coats et al., 2014; Pan et al., 2014). The embodied action

schema represents the action capabilities of an actor, which is what has been shown to be calibrated in the present

and past work.

Future Research

Many important questions remain with regard to calibration to an avatar in VR. In the present study, the length of the

avatar arm was manipulated. Future work should investigate other types of manipulations. Leyrer et al. (2011)

manipulated the eye height of the avatar while asking participants to judge distances in virtual reality. They found that

participants with an avatar whose eye height was increased perceived distances as shorter in comparison with

participants who viewed distances through a shorter eye height. Interestingly, participants who had their eye height

decreased did not show an increase in distance perception, nor was there a difference in distance perception

between the shortened eye height group and the group who did not have their eye height manipulated. From the

standpoint of calibration, this asymmetrical finding is quite intriguing, as it suggests that calibration to altered

dimensions of an avatar does not occur in all cases. Ebrahimi et al. (2014, 2015) found a similar asymmetry in

reaching behavior. In their experiments, calibration to lengthened reaches was much greater than the calibration to

shortened reaches. It is possible that people are more used to calibrating to alterations that extend their abilities,

such as tools, and so forth, than manipulations that cause one’s limbs to be effectively shorter. In the case of altered

eye height, however, it may be that people are used to manipulations that shorten their eye height, such as sitting,

and they have learned that such an alteration should not rescale the one’s perception of distances in the world.

Future work should be directed at uncovering what types of manipulations can and cannot be calibrated to.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

27 of 36 2/23/2019, 4:19 PM

It was mentioned above that calibration is action specific, meaning that the calibration of one action typically does

not generalize to very different actions (e.g., Pan et al., 2014; Proffitt, 2008; Rieser et al., 1995; Witt et al., 2010). We

would not expect calibration to an altered avatar arm to generalize to the ability to judge broader distances in the

virtual space. Nonetheless, it remains important to test the extent to which calibration to an altered avatar transfers to

other tasks involving that avatar. For example, if manual reaches are used to calibrate to a lengthened avatar arm

then one would expect the calibration to carry over to other behaviors involving that arm, such as perceiving the

minimum doorway width that one can pass through with the arm stretched out to the side. This is because the

lengthened arm has been incorporated into the embodied action schema. Similarly, it is important to confirm that the

alterations are restricted to actions performed with the arm, and do not involve a perceived enlargement of the entire

body. No changes were made to information pertaining to the size of the body as a whole, such as the participants’

eye height in VR, and the full avatar with alterations to just the arms was seen by the subject. Thus, calibration to a

lengthened avatar arm would not be expected to carry over to behaviors involving the scale of the whole body, such

as perceiving the minimum doorway height that one can pass under while walking.

Of note, within the current studies, participants performed their reaching behaviors in VR or the real world over the

course of about an hour. Thus, the current findings can only generalize to applications where VR is used for training

during this time frame. Future research should investigate if repeated exposures (i.e., multiple hours of training over

multiple days) to VR training with an altered avatar results in similar findings. This topic deserves further attention.

Practical Implications and Conclusion

The present work has several practical implications. For example, this work has implications for understanding how

people may accept a virtual limb larger than their own limb and how amputees accept artificial limbs (Imaizumi, Asai,

& Koyama, 2016). Artificial limbs may or may not be the exact same size as their lost limb, and they typically do not

possess the same weight properties. The results of the present work suggest that calibration to these altered limbs is

possible and that the artificial limbs can be incorporated into the embodied action schema through the process of

calibration, but only after experience using them.

VR technology has numerous current applications, and the number of meaningful applications is growing. Currently,

VR is being used to aid rehabilitation, as a training exercise in many fields such as the medical field and combat,

education, behavioral research, and entertainment. Most technology companies offer some version of virtual reality

gaming product. For any of these uses of VR technology to be maximally useful, especially when used as a training

aid, a user’s perception-action processes in VR must match their perception-action processes in the real world. The

underlying assumption behind such applications of VR is that such VR rehabilitation or training will carry over to later

behavior in the real world. Although carry-over effects are desired in such applications, undesired carry-over effects

may be equally likely. Alterations made to ones reaching or driving behavior in VR may carry over to subsequent

reaching or driving behavior in the real world, with undesirable consequences. After using VR as part of

rehabilitation, training, gaming, entertainment, employment, scientific experimentation, work settings, and so forth,

people may need to perform tasks outside of VR in order to calibrate once again to the real world before exposing

themselves to real world consequences.

Previous research has highlighted the importance of providing a user with an avatar when acting in VR (Mohler et

al., 2010). However, as demonstrated in the current work and other related work (Linkenauger et al., 2015), merely

providing an avatar does not resolve all potential issues. There are major implications for how a user will perceive

and act in the immersive virtual environment based off the size and dimensions of their avatar. The present results

support the idea that calibration to an avatar that possesses different anthropometric dimensions than ones’ own

body is possible, but this process is relatively lengthy in comparison to other reported rates of calibration in that it

takes approximately 45 trials over the span of 10 to 15 min. Further, participants who were given an avatar with

extended arms produced behaviors that involved more discrepancies between the presented target distance and

their estimated reach distance than those who reached with an avatar scaled to the size of their body. These findings

suggest that for training in VR to be most effective, the avatar given to a user must either faithfully represent the

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

28 of 36 2/23/2019, 4:19 PM

dimensions of their body or an adequate calibration period with feedback about the outcomes of ones’ actions must

be provided.

Overall, the findings from the studies presented herein have direct implications for how avatars are designed and

presented to the user in immersive virtual environments. Presenting users with avatars that do not represent their

normal body causes a change in how the virtual environment is perceived, and altered avatars cannot be used to act

in a manner that is representative of how that user would act with an avatar that is designed to match their bodily

dimensions immediately. The results support the conclusion that participants are able to calibrate effectively to an

avatar that possesses longer arms than their own body. However, if the virtual environment is to be used for training

that must translate back to the real world, then the avatar must either match the actual dimensions of the users’ body

or an ample period of calibration must be provided during the transition back to the real world.

Footnotes

The meaning of the terms adaptation and calibration are not generally agreed upon in the literature, and they are

often treated as synonyms and used interchangeably (e.g., Mon-Williams & Bingham, 2007; van Andel et al., 2017).

We offer the following disambiguation: Adaptation is a more general process, with calibration being a special type of

adaptation. Adaptation includes processes that produce changes in sensations alone, even in the absence of an

accompanying action or in the absence of the perception of an affordance. Examples include decreases in neural

sensitivity due to the constant application of a stimulus, such as with color and motion adaptation and their

associated afterimages, as well as increases in sensitivity such as dark adaptation. Such adaptation can occur in the

absence of an action or other types of feedback. Calibration specifically refers to a type of adaptation that involves a

rescaling of the output of a perception-action system, and thus the two terms are not synonymous.

Just as in power analyses for traditional statistical techniques, estimating power in a multilevel study is a function of

Type I error rate, sample size, and effect size. Two other considerations for estimating power in a multilevel study are

the sample size of Level 2 units compared with the sample size of Level 1 units, and the intraclass correlation (ICC).

Power estimation in MLM is a complex procedure because it requires additional assumptions because of nesting and

the Level 1 and Level 2 estimates. Simulations have been run manipulating the n at Level 1 and N at Level 2 to

determine the standard error in various scenarios. Using absolute error as the dependent variable of interest, based

off previous research (Day et al., 2017; Ebrahimi et al., 2016) our estimated ICC was 0.15. To be conservative we

followed guidelines presented in Hox, Moerbeek, and van de Schoot (2010) by estimating the design effect based on

both 20 and 24 participants. Using these participant estimates and 130 L1 units, the design effect was 20.35. The

effective sample size was 128 for 20 participants, and 153 for 24 participants. For Cohen’s medium effect size (f ) of

.15, with an alpha level of 0.05 and seven IVs, power would be between 0.88 and 0.94, for 20 and 24 participants,

respectively. To have power of at least 0.80, we would need an effective sample size of at least 105. We chose an

effective sample size that falls comfortably between these two estimated effective sample sizes. Assuming the ICC

of our dependent variable does not exceed 0.25 (which is likely), 26 participants (an N at L2 of 26), each of whom

will complete 130 trials (an n at L1 of 130), we would far exceed an estimation of power at 0.80.

References

Altenhoff, B. M., Napieralski, P. E., Long, L. O., Bertrand, J. W., Pagano, C. C., Babu, S. V., & Davis, T. A. (2012).

Effects of visual and haptic feedback on near-field depth perception in an immersive virtual environment. In

Proceedings of the ACM symposium on applied perception (pp. 71–78). New York, NY: ACM.

Banakou, D., Groten, R., & Slater, M. (2013). Illusory ownership of a virtual child body causes overestimation of

object sizes and implicit attitude changes. Proceedings of the National Academy of Sciences of the United States of

America, 110, 12846–12851. 10.1073/pnas.1306779110

Bertram, J., Moskaliuk, J., & Cress, U. (2015). Virtual training: Making reality work?Computers in Human Behavior,

43, 284–292. 10.1016/j.chb.2014.10.032

1

2

2

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

29 of 36 2/23/2019, 4:19 PM

Bingham, G. P., & Pagano, C. C. (1998). The necessity of a perception-action approach to definite distance

perception: Monocular distance perception to guide reaching. Journal of Experimental Psychology: Human

Perception and Performance, 24, 145–168. 10.1037/0096-1523.24.1.145

Bingham, G. P., Pan, J. S., & Mon-Williams, M. A. (2014). Calibration is both functional and anatomical. Journal of

Experimental Psychology: Human Perception and Performance, 40, 61–70. 10.1037/a0033458

Bingham, G. P., & Romack, J. L. (1999). The rate of adaptation to displacement prisms remains constant despite

acquisition of rapid calibration. Journal of Experimental Psychology: Human Perception and Performance, 25,

1331–1346. 10.1037/0096-1523.25.5.1331

Bingham, G. P., & Stassen, M. G. (1994). Monocular egocentric distance information generated by head movement.

Ecological Psychology, 6, 219–238. 10.1207/s15326969eco0603_4

Bliese, P. D. (1998). Group size, ICC values, and group-level correlations: A simulation. Organizational Research

Methods, 1, 355–373. 10.1177/109442819814001

Bliss, J. P., Tidwell, P. D., & Guest, M. A. (1997). The effectiveness of virtual reality for administering spatial

navigation training to firefighters. Presence, 6, 73–86. 10.1162/pres.1997.6.1.73

Bongers, R. M., Michaels, C. F., & Smitsman, A. W. (2004). Variations of tool and task characteristics reveal that tool-

use postures are anticipated. Journal of Motor Behavior, 36, 305–315. 10.3200/JMBR.36.3.305-315

Bongers, R. M., Smitsman, A. W., & Michaels, C. F. (2003). Geometrics and dynamics of a rod determine how it is

used for reaching. Journal of Motor Behavior, 35, 4–22. 10.1080/00222890309602117

Bourgeois, J., & Coello, Y. (2012). Effect of visuomotor calibration and uncertainty on the perception of peripersonal

space. Attention, Perception, & Psychophysics, 74, 1268–1283. 10.3758/s13414-012-0316-x

Brand, M. T., & de Oliveira, R. F. (2017). Recalibration in functional perceptual-motor tasks: A systematic review.

Human Movement Science, 56, 54–70. 10.1016/j.humov.2017.10.020

Bufton, A., Campbell, A., Howie, E., & Straker, L. (2014). A comparison of the upper limb movement kinematics

utilized by children playing virtual and real table tennis. Human Movement Science, 38, 84–93.

10.1016/j.humov.2014.08.004

Burton, G. (1993). Non-neural extensions of haptic sensitivity. Ecological Psychology, 5, 105–124.

10.1207/s15326969eco0502_1

Cabrera, F., Sanabria, F., Jiménez, Á. A., & Covarrubias, P. (2013). An affordance analysis of unconditioned lever

pressing in rats and hamsters. Behavioural Processes, 92, 36–46. 10.1016/j.beproc.2012.10.003

Cardinali, L., Frassinetti, F., Brozzoli, C., Urquizar, C., Roy, A. C., & Farnè, A. (2009). Tool-use induces

morphological updating of the body schema. Current Biology, 19, R478–R479. 10.1016/j.cub.2009.05.009

Carello, C., Fitzpatrick, P., & Turvey, M. T. (1992). Haptic probing: Perceiving the length of a probe and the distance

of a surface probed. Perception & Psychophysics, 51, 580–598. 10.3758/BF03211655

Carello, C., Grosofsky, A., Reichel, F. D., Solomon, H. Y., & Turvey, M. T. (1989). Visually perceiving what is

reachable. Ecological Psychology, 1, 27–54. 10.1207/s15326969eco0101_3

Coats, R. O., Pan, J. S., & Bingham, G. P. (2014). Perturbation of perceptual units reveals dominance hierarchy in

cross calibration. Journal of Experimental Psychology: Human Perception and Performance, 40, 328–341.

10.1037/a0033802

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

30 of 36 2/23/2019, 4:19 PM

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple correlation/regression analysis for the

behavioral sciences. London, UK: Taylor & Francis.

Creem-Regehr, S. H., Stefanucci, J. K., & Thompson, W. B. (2015). Perceiving absolute scale in virtual

environments: How theory and application have mutually informed the role of body-based perception. In B. H.Ross

(Ed.), Psychology of learning and motivation (pp. 195–224). San Diego, CA: Academic Press.

10.1016/bs.plm.2014.09.006

Cutting, J. E. (1986). Perception with an eye for motion. Cambridge, MA: MIT Press.

Day, B., Ebrahimi, E., Hartman, L. S., Pagano, C. C., & Babu, S. V. (2017). Calibration to tool use during visually-

guided reaching. Acta Psychologica, 181, 27–39. 10.1016/j.actpsy.2017.09.014

Ebrahimi, E., Altenhoff, B., Hartman, L., Jones, J. A., Babu, S. V., Pagano, C. C., & Davis, T. A. (2014). Effects of

visual and proprioceptive information in visuo-motor calibration during a closed-loop physical reach task in immersive

virtual environments. In Proceedings of the ACM symposium on applied perception (ACM SAP), 103–110,

Vancouver, BC.

Ebrahimi, E., Altenhoff, B. M., Pagano, C. C., & Babu, S. V. (2015). Carryover effects of calibration to visual and

proprioceptive information on near field distance judgments in 3D user interaction. In 2015 IEEE Symposium on 3D

user interfaces (3DUI, pp. 97–104), Arles, France: IEEE.

Ebrahimi, E., Babu, S. V., Pagano, C. C., & Jörg, S. (2016). An empirical evaluation of visuo-haptic feedback on

physical reaching behaviors during 3D interaction in real and immersive virtual environments. ACM Transactions on

Applied Perception, 13, 19:1–19:21.

Fajen, B. R. (2005). Perceiving possibilities for action: On the necessity of calibration and perceptual learning for the

visual guidance of action. Perception, 34, 717–740. 10.1068/p5405

Fajen, B. R. (2007). Rapid recalibration based on optic flow in visually guided action. Experimental Brain Research,

183, 61–74. 10.1007/s00221-007-1021-1

Fajen, B. R., Riley, M. A., & Turvey, M. T. (2009). Information, affordances, and the control of action in sport.

International Journal of Sport Psychology, 40, 79–107.

Fitzpatrick, P., Carello, C., & Turvey, M. T. (1994). Eigenvalues of the inertia tensor and exteroception by the

“muscular sense.”Neuroscience, 60, 551–568. 10.1016/0306-4522(94)90264-X

Franchak, J. M., van der Zalm, D. J., & Adolph, K. E. (2010). Learning by doing: Action performance facilitates

affordance perception. Vision Research, 50, 2758–2765. 10.1016/j.visres.2010.09.019

Ganier, F., Hoareau, C., & Tisseau, J. (2014). Evaluation of procedural learning transfer from a virtual environment to

a real situation: A case study on tank maintenance training. Ergonomics, 57, 828–843.

10.1080/00140139.2014.899628

Gibson, J. J. (1979). The ecological approach to visual perception. Boston, MA: Houghton Mifflin.

Grabowski, A., & Jankowski, J. (2015). Virtual Reality-based pilot training for underground coal miners. Safety

Science, 72, 310–314. 10.1016/j.ssci.2014.09.017

Head, H. (1920). Studies in neurology (Vol. 2). London, UK: Oxford University Press.

Head, H., & Holmes, G. (1911–1912). Sensory disturbances from cerebral lesions. Brain: A Journal of Neurology, 34,

102–254. 10.1093/brain/34.2-3.102

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

31 of 36 2/23/2019, 4:19 PM

Heck, R., Thomas, S., & Tabata, L. (2010). Multilevel and longitudinal modeling with iBM sPSS (Quantitative

methodology series). New York, NY: Routledge.

Heft, H. (1993). A methodological note on overestimates of reaching distance: Distinguishing between perceptual

and analytical judgments. Ecological Psychology, 5, 255–271. 10.1207/s15326969eco0503_3

Hox, J. J., Moerbeek, M., & van de Schoot, R. (2010). Multilevel analysis: Techniques and applications. London, UK:

Routledge.

Hyltander, A., Liljegren, E., Rhodin, P. H., & Lönroth, H. (2002). The transfer of basic skills learned in a laparoscopic

simulator to the operating room. Surgical Endoscopy and Other Interventional Techniques, 16, 1324–1328.

10.1007/s00464-001-9184-5

Imaizumi, S., Asai, T., & Koyama, S. (2016). Embodied prosthetic arm stabilizes body posture, while unembodied

one perturbs it. Consciousness and Cognition, 45, 75–88. 10.1016/j.concog.2016.08.019

Iodice, P., Scuderi, N., Saggini, R., & Pezzulo, G. (2015). Multiple timescales of body schema reorganization due to

plastic surgery. Human Movement Science, 42, 54–70. 10.1016/j.humov.2015.04.009

Jun, E., Stefanucci, J. K., Creem-Regehr, S. H., Geuss, M. N., & Thompson, W. B. (2015). Big foot: Using the size of

a virtual foot to scale gap width. ACM Transactions on Applied Perception, 12, 16. 10.1145/2811266

Kelly, J. W., Donaldson, L. S., Sjolund, L. A., & Freiberg, J. B. (2013). More than just perception-action recalibration:

Walking through a virtual environment causes rescaling of perceived space. Attention, Perception, & Psychophysics,

75, 1473–1485. 10.3758/s13414-013-0503-4

Kelly, J. W., Hammel, W. W., Siegel, Z. D., & Sjolund, L. A. (2014). Recalibration of perceived distance in virtual

environments occurs rapidly and transfers asymmetrically across scale. IEEE Transactions on Visualization and

Computer Graphics, 20, 588–595. 10.1109/TVCG.2014.36

Kilteni, K., Groten, R., & Slater, M. (2012). The sense of embodiment in virtual reality. Presence, 21, 373–387.

10.1162/PRES_a_00124

Koritnik, T., Koenig, A., Bajd, T., Riener, R., & Munih, M. (2010). Comparison of visual and haptic feedback during

training of lower extremities. Gait & Posture, 32, 540–546. 10.1016/j.gaitpost.2010.07.017

Kozak, J. J., Hancock, P. A., Arthur, E. J., & Chrysler, S. T. (1993). Transfer of training from virtual reality.

Ergonomics, 36, 777–784. 10.1080/00140139308967941

Larrue, F., Sauzeon, H., Wallet, G., Foloppe, D., Cazalets, J. R., Gross, C., & N′Kaoua, B. (2014). Influence of body-

centered information on the transfer of spatial learning from a virtual to a real environment. Journal of Cognitive

Psychology, 26, 906–918. 10.1080/20445911.2014.965714

Lessard, D. A., Linkenauger, S. A., & Proffitt, D. R. (2009). Look before you leap: Jumping ability affects distance

perception. Perception, 38, 1863–1866. 10.1068/p6509

Leyrer, M., Linkenauger, S. A., Bülthoff, H. H., Kloos, U., & Mohler, B. (2011). The influence of eye height and

avatars on egocentric distance estimates in immersive virtual environments. In Proceedings of the ACM SIGGRAPH

symposium on applied perception in graphics and visualization (pp. 67–74). New York, NY: ACM.

10.1145/2077451.2077464

Lin, Q., Rieser, J., & Bodenheimer, B. (2012). Stepping over and ducking under: The influence of an avatar on

locomotion in an HMD-based immersive virtual environment. In Proceedings of the ACM symposium on applied

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

32 of 36 2/23/2019, 4:19 PM

perception (pp. 7–10). New York, NY: ACM.

Lin, Q., Rieser, J. J., & Bodenheimer, B. (2013). Stepping off a ledge in an HMD-based immersive virtual

environment. In Proceedings of the ACM symposium on applied perception (pp. 107–110). New York, NY: ACM.

10.1145/2492494.2492511

Lin, Q., Rieser, J., & Bodenheimer, B. (2015). Affordance judgments in HMD-based virtual environments: Stepping

over a pole and stepping off a ledge. ACM Transactions on Applied Perception, 12, 6. 10.1145/2720020

Linkenauger, S. A., Bülthoff, H. H., & Mohler, B. J. (2015). Virtual arm’s reach influences perceived distances but

only after experience reaching. Neuropsychologia, 70, 393–401. 10.1016/j.neuropsychologia.2014.10.034

Linkenauger, S. A., Leyrer, M., Bülthoff, H. H., & Mohler, B. J. (2013). Welcome to wonderland: The influence of the

size and shape of a virtual hand on the perceived size and shape of virtual objects. PLoS ONE, 8, e68594.

10.1371/journal.pone.0068594

Lintern, G., Roscoe, S. N., Koonce, J. M., & Segal, L. D. (1990). Transfer of landing skills in beginning flight training.

Human Factors, 32, 319–327. 10.1177/001872089003200305

Mantel, B., Stoffregen, T. A., Campbell, A., & Bardy, B. G. (2015). Exploratory movement generates higher-order

information that is sufficient for accurate perception of scaled egocentric distance. PLoS ONE, 10, e0120025.

10.1371/journal.pone.0120025

Maravita, A., & Iriki, A. (2004). Tools for the body (schema). Trends in Cognitive Sciences, 8, 79–86.

10.1016/j.tics.2003.12.008

Mark, L. S. (1987). Eyeheight-scaled information about affordances: A study of sitting and stair climbing. Journal of

Experimental Psychology: Human Perception and Performance, 13, 361–370. 10.1037/0096-1523.13.3.361

Mark, L. S., Balliett, J. A., Craver, K. D., Douglas, S. D., & Fox, T. (1990). What an actor must do in order to perceive

the affordance for sitting. Ecological Psychology, 2, 325–366. 10.1207/s15326969eco0204_2

Maselli, A., & Slater, M. (2013). The building blocks of the full body ownership illusion. Frontiers in Human

Neuroscience, 7, 83. 10.3389/fnhum.2013.00083

McManus, E. A., Bodenheimer, B., Streuber, S., De La Rosa, S., Bülthoff, H. H., & Mohler, B. J. (2011). The influence

of avatar (self and character) animations on distance estimation, object interaction and locomotion in immersive

virtual environments. In Proceedings of the ACM SIGGRAPH symposium on applied perception in graphics and

visualization (pp. 37–44). New York, NY: ACM.

Mohler, B. J., Creem-Regehr, S. H., Thompson, W. B., & Bülthoff, H. H. (2010). The effect of viewing a self-avatar on

distance judgments in an HMD-based virtual environment. Presence, 19, 230–242. 10.1162/pres.19.3.230

Mon-Williams, M., & Bingham, G. P. (2007). Calibrating reach distance to visual targets. Journal of Experimental

Psychology: Human Perception and Performance, 33, 645–656. 10.1037/0096-1523.33.3.645

Moore, K., Rodes, W., Csencsits, M., Kwoka, M., Gomer, J., & Pagano, C. (2006). Interface evaluation for soft

robotic manipulators. In G. R.Gerhart, C. M.Shoemaker, & D. W.Gage (Eds.) Unmanned systems technology VIII:

Proceedings of SPIE the international society for optical engineering defense and security symposium (Vol. 6230,

pp. 62301C1–62301C11). Kissimmee, Florida: International Society for Optics and Photonics.

Napieralski, P. E., Altenhoff, B. M., Bertrand, J. W., Long, L. O., Babu, S. V., Pagano, C. C., . . .Davis, T. A. (2011).

Comparing near field distance perception in real and virtual environments using both verbal and action responses.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

33 of 36 2/23/2019, 4:19 PM

ACM Transactions on Applied Perception, 8, 18:1–18:19.

Pagano, C. C., Grutzmacher, R. P., & Jenkins, J. C. (2001). Comparing verbal and reaching responses to visually

perceived egocentric distances. Ecological Psychology, 13, 197–226. 10.1207/S15326969ECO1303_2

Pagano, C. C., & Turvey, M. T. (1995). The inertia tensor as a basis for the perception of limb orientation. Journal of

Experimental Psychology: Human Perception and Performance, 21, 1070–1087. 10.1037/0096-1523.21.5.1070

Pagano, C. C., & Turvey, M. T. (1998). Eigenvectors of the inertia tensor and perceiving the orientation of limbs and

objects. Journal of Applied Biomechanics, 14, 331–359. 10.1123/jab.14.4.331

Pan, J. S., Coats, R. O., & Bingham, G. P. (2014). Calibration is action specific but perturbation of perceptual units is

not. Journal of Experimental Psychology: Human Perception and Performance, 40, 404–415. 10.1037/a0033795

Proffitt, D. R. (2008). An action-specific approach to spatial perception. In R.Klatzky, B.MacWhinney, & M.Berhmann

(Eds.), Embodiment, ego-space, and action (pp. 179–202). Mahwah, NJ: Erlbaum.

Proffitt, D. R., & Linkenauger, S. A. (2013). Perception viewed as a phenotypic expression. Action Science:

Foundations of an Emerging Discipline, 171–197.

Regian, J. W. (1997). Virtual reality for training: Evaluating transfer. In J.Seidel & P. R.Chatelier (Eds.), Virtual reality:

Training’s future? (pp. 31–40). New York, NY: Springer.

Renner, R. S., Velichkovsky, B. M., & Helmert, J. R. (2013). The perception of egocentric distances in virtual

environments-a review. [CSUR]. ACM Computing Surveys, 46, 23:1–23:40. 10.1145/2543581.2543590

Ries, B., Interrante, V., Kaeding, M., & Phillips, L. (2009). Analyzing the effect of a virtual avatar’s geometric and

motion fidelity on ego-centric spatial perception in immersive virtual environments. In Proceedings of the 16th ACM

symposium on virtual reality software and technology (pp. 59–66). Kyoto, Japan: ACM. 10.1145/1643928.1643943

Rieser, J. J., Pick, H. L., Jr., Ashmead, D. H., & Garing, A. E. (1995). Calibration of human locomotion and models of

perceptual-motor organization. Journal of Experimental Psychology: Human Perception and Performance, 21,

480–497. 10.1037/0096-1523.21.3.480

Riley, M. A., & Turvey, M. T. (2001). Inertial constraints on limb proprioception are independent of visual calibration.

Journal of Experimental Psychology: Human Perception and Performance, 27, 438–455.

10.1037/0096-1523.27.2.438

Rose, F. D., Attree, E. A., Brooks, B. M., Parslow, D. M., & Penn, P. R. (2000). Training in virtual environments:

Transfer to real world tasks and equivalence to real task training. Ergonomics, 43, 494–511.

10.1080/001401300184378

Scott, S., & Gray, R. (2010). Switching tools: Perceptual-motor recalibration to weight changes. Experimental Brain

Research, 201, 177–189. 10.1007/s00221-009-2022-z

Slater, M., Pérez-Marcos, D., Ehrsson, H. H., & Sanchez-Vives, M. V. (2008). Towards a digital body: The virtual arm

illusion. Frontiers in Human Neuroscience, 2, 6. 10.3389/neuro.09.006.2008

Slater, M., Pérez-Marcos, D., Ehrsson, H. H., & Sanchez-Vives, M. V. (2009). Inducing illusory ownership of a virtual

body. Frontiers in Neuroscience, 3, 214–220. 10.3389/neuro.01.029.2009

Slater, M., Spanlang, B., Sanchez-Vives, M. V., & Blanke, O. (2010). First person experience of body transfer in

virtual reality. PLoS ONE, 5, e10564. 10.1371/journal.pone.0010564

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

34 of 36 2/23/2019, 4:19 PM

Sposito, A., Bolognini, N., Vallar, G., & Maravita, A. (2012). Extension of perceived arm length following tool-use:

Clues to plasticity of body metrics. Neuropsychologia, 50, 2187–2194. 10.1016/j.neuropsychologia.2012.05.022

Stoffregen, T. A., & Bardy, B. G. (2001). On specification and the senses. Behavioral and Brain Sciences, 24,

195–213. 10.1017/S0140525X01003946

van Andel, S., Cole, M. H., & Pepping, G. J. (2017). A systematic review on perceptual-motor calibration to changes

in action capabilities. Human Movement Science, 51, 59–71. 10.1016/j.humov.2016.11.004

van de Langenberg, R., Kingma, I., & Beek, P. J. (2008). The perception of limb orientation depends on the center of

mass. Journal of Experimental Psychology: Human Perception and Performance, 34, 624–639.

10.1037/0096-1523.34.3.624

Vindras, P., & Viviani, P. (1998). Frames of reference and control parameters in visuomanual pointing. Journal of

Experimental Psychology: Human Perception and Performance, 24, 569–591. 10.1037/0096-1523.24.2.569

Wagman, J. B. (2012). Perception of maximum reaching height reflects impending changes in reaching ability and

improvements transfer to unpracticed reaching tasks. Experimental Brain Research, 219, 467–476.

10.1007/s00221-012-3104-x

Wagman, J. B., Caputo, S. E., & Stoffregen, T. A. (2016). Sensitivity to hierarchical relations among affordances in

the assembly of asymmetric tools. Experimental Brain Research, 234, 2923–2933. 10.1007/s00221-016-4695-4

Wagman, J. B., & Chemero, A. (2014). The end of the debate over extended cognition (pp. 105–124). London, UK:

Palgrave Macmillan.

Wagman, J. B., Langley, M. D., & Farmer-Dougan, V. (2017). Doggone affordances: Canine perception of

affordances for reaching. Psychonomic Bulletin & Review, 24, 1097–1103. 10.3758/s13423-016-1183-6

Wagman, J. B., & Morgan, L. L. (2010). Nested prospectivity in perception: Perceived maximum reaching height

reflects anticipated changes in reaching ability. Psychonomic Bulletin & Review, 17, 905–909. 10.3758/PBR.17.6.905

Wagman, J. B., Taheny, C. A., & Higuchi, T. (2014). Improvements in perception of maximum reaching height

transfer to increases or decreases in reaching ability. The American Journal of Psychology, 127, 269–279.

10.5406/amerjpsyc.127.3.0269

Wagman, J. B., & Taylor, K. R. (2005). Perceiving affordances for aperture crossing for the person-plus-object

system. Ecological Psychology, 17, 105–130. 10.1207/s15326969eco1702_3

Wagman, J. B., Thomas, B. J., McBride, D. M., & Day, B. M. (2013). Perception of maximum reaching height when

the means of reaching are no longer in view. Ecological Psychology, 25, 63–80. 10.1080/10407413.2013.753810

Wann, J. P., & Ibrahim, S. F. (1992). Does limb proprioception drift?Experimental Brain Research, 91, 162–166.

10.1007/BF00230024

Warren, W. (1984). Perceiving affordances: The visual guidance of stair stepping. Journal of Experimental

Psychology: Human Perception and Performance, 10, 683–703.

Warren, W. H., Jr., & Whang, S. (1987). Visual guidance of walking through apertures: Body-scaled information for

affordances. Journal of Experimental Psychology: Human Perception and Performance, 13, 371–383.

10.1037/0096-1523.13.3.371

Welch, R. (1986). Adaptation of space perception. In K. B.Boff, L.Kaufman, & J. P.Thomas (Eds.), Handbook of

perception and human performance (pp. 1–45). Hoboken, NJ: Wiley.

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

35 of 36 2/23/2019, 4:19 PM

Wickelgren, E. A., McConnell, D. S., & Bingham, G. P. (2000). Reaching measures of monocular distance

perception: Forward versus side-to-side head movements and haptic feedback. Perception & Psychophysics, 62,

1051–1059. 10.3758/BF03212088

Willemsen, P., Gooch, A. A., Thompson, W. B., & Creem-Regehr, S. H. (2008). Effects of stereo viewing conditions

on distance perception in virtual environments. Presence, 17, 91–101. 10.1162/pres.17.1.91

Withagen, R., & Michaels, C. F. (2004). Transfer of calibration in length perception by dynamic touch. Perception &

Psychophysics, 66, 1282–1292. 10.3758/BF03194998

Withagen, R., & Michaels, C. F. (2007). Transfer of calibration between length and sweet-spot perception by dynamic

touch. Ecological Psychology, 19, 1–19. 10.1080/10407410709336948

Witt, J. K. (2011). Action’s effect on perception. Current Directions in Psychological Science, 20, 201–206.

10.1177/0963721411408770

Witt, J. K., & Proffitt, D. R. (2008). Action-specific influences on distance perception: A role for motor simulation.

Journal of Experimental Psychology: Human Perception and Performance, 34, 1479–1492. 10.1037/a0010781

Witt, J. K., Proffitt, D. R., & Epstein, W. (2005). Tool use affects perceived distance, but only when you intend to use

it. Journal of Experimental Psychology: Human Perception and Performance, 31, 880–888.

10.1037/0096-1523.31.5.880

Witt, J. K., Proffitt, D. R., & Epstein, W. (2010). When and how are spatial perceptions scaled?Journal of

Experimental Psychology: Human Perception and Performance, 36, 1153–1160. 10.1037/a0019947

APPENDIX

APPENDIX A: Body Ownership Questionnaire

1. When you were looking down from above how much did you feel a strong connection with the avatar as if you

were looking down at yourself?

NOT AT ALL 0 1 2 3 4 5 6 7 8 9 10 VERY MUCH

2. How much did you feel that the seated avatar’s body was your body?

NOT AT ALL 0 1 2 3 4 5 6 7 8 9 10 VERY MUCH

3. How strong was the feeling that the movements of the avatar were caused by your own movements?

NOT AT ALL 0 1 2 3 4 5 6 7 8 9 10 VERY MUCH

4. How much did you feel that the virtual body was another person?

NOT AT ALL 0 1 2 3 4 5 6 7 8 9 10 VERY MUCH

5. How much was this experience more like watching a scene from the outside compared to being part of the scene?

NOT AT ALL 0 1 2 3 4 5 6 7 8 9 10 VERY MUCH

6. How strong was the feeling that the body of the person in the mirror was your body?

NOT AT ALL 0 1 2 3 4 5 6 7 8 9 10 VERY MUCH

Submitted: February 19, 2018 Revised: July 24, 2018 Accepted: July 28, 2018

This publication is protected by US and international copyright laws and its content may not be copied without the

copyright holders express written permission except for the print or download capabilities of the retrieval software

used for access. This content is intended solely for the use of the individual user.

Source: Journal of Experimental Psychology: Applied. Oct 22, 2018

Accession Number: 2018-52919-001

Digital Object Identifier: 10.1037/xap0000192

EBSCOhost http://web.b.ebscohost.com.library.capella.edu/ehost/delivery?sid=090f3...

36 of 36 2/23/2019, 4:19 PM