Interaction with co-located haptic feedback in virtual reality

Virtual Reality (2006) 10: 24–30DOI 10.1007/s10055-006-0027-5ORIGINAL ARTICLEDavid Swapp Æ Vijay Pawar Æ Ce´line LoscosInteraction with co-located haptic feedback in virtual realityReceived: 22 December 2005 / Accepted: 31 March 2006 / Published online: 27 April 2006Ó Springer-Verlag London Limited 2006Abstract This paper outlines a study into the effects ofco-location (the term ‘co-location’ is used throughout torefer to the co-location of haptic and visual sensorymodes, except where otherwise specified) of haptic andvisual sensory modes in VR simulations. The studyhypothesis is that co-location of these sensory modeswill lead to improved task performance within a VRenvironment. Technical challenges and technologicallimitations are outlined prior to a description of theimplementation adopted for this study. Experimentswere conducted to evaluate the effect on user performanceof co-located haptics (force feedback) in a 3Dvirtual environment. Results show that co-location is animportant factor, and when coupled with haptic feedbackthe performance of the user is greatly improved.Keywords Haptics Æ Co-location1 IntroductionPresence is likely to be enhanced by multi-modal input:in a virtual reality (VR) environment, the simulation ofadditional sensory modes should consolidate our senseof presence although this is with the proviso that allsimulated sensory modes accurately complement oneanother (Dinh et al. 1999) Conflicting sensory cues areliable to degrade the sense of presence. Research in VRis currently dominated by simulation for the visual andaudio sensory modes. While this can be partly explainedby the dominance of these sensory modes in shaping ourexperience of the world, there are also technologicalreasons. The visual and audio sensory modes benefitD. Swapp (&) Æ V. Pawar Æ C. LoscosDepartment of Computer Science, University College London,Malet Place, WC1E 6BT London, UKE-mail: d.swapp@cs.ucl.ac.ukTel.: +44-20-76797211Fax: +44-20-73871397E-mail: vijaympawar@hotmail.comE-mail: c.loscos@cs.ucl.ac.ukfrom a high availability of good quality display equipmentthat, in combination with accurate and low-latencytracking, allows the development of compelling simulations,sufficient to induce a strong feeling of presence inusers. In many application areas it is likely that touchcan also be a compelling factor in presence (Durlachet al. 2005; Frisoli et al. 2005; Loscos et al. 2004)transforming a simulation from a world of ghosts to aworld of solid forms. Haptics does not refer to a singularsensory apparatus, and can be considered to be composedof a number of sensory and motor elementsthough there is a degree of overlap among the tactile,force-feedback and proprioceptive elements (Dinh et al.1999; Hoffman 1998). The proprioceptive element canconsolidate visual cues of depth and spatial arrangementin a simulated 3D environment (Hayward et al. 2004).Other studies show that the addition of haptics can leadto improved task performance (Hoffman 1998; Sallna¨set al. 2000).Ideally all simulated sensory cues should be co-located—forexample a sound-producing object should getlouder when it visually appears closer and sound shouldbe perceived to come from the same direction that we seethe object. Likewise, for haptics we should be able to feelthe edge of an object at precisely the location where wesee that edge. Precise co-location of haptics is, however,technically harder to achieve than co-location of audioand visual cues. This is primarily due to the greaterperceptual latitude when locating the source of a stimulusvia both audio and visual cues. The phenomenonknown as the ‘‘ventriloquism effect’’, whereby an audiostimulus that is spatially close to a visual stimulus isperceived to emanate from the location of the visualstimulus (Jack and Thurlow 1973; Wallace et al. 2004)has also been demonstrated for spatial dominance ofother sensory cues over audio cues (Caclin et al. 2002).In other words, while we are to some degree tolerant ofinaccuracies in visual-audio co-location, this is not thecase for visual-haptic co-location.A commonly-implemented compromise to co-locationis the use of visual markers to represent the haptic

25contact points. In the study described in this paper, asmall cone-shaped marker is used—this also provides anindication of the orientation of the haptic contact. Otherstudies have used cross-hairs to pinpoint the hapticcontact (e.g. Loscos et al. 2004) Because these markersare visually rendered by the same graphics system as thevirtual environment, spatial correspondence is guaranteed—inother words, it will never be the case that thevisual marker is slightly offset from an object’s edgewhen the haptic contact is felt (assuming that the hapticcollision detection algorithm is effective). In the currentstudy, such a setup is referred to as non-colocatedhaptics.For many applications it is possible for a non-colocatedinterface to work well, with users adjusting quiterapidly to the built-in spatial offset. However this is notalways the case. The proprioceptive element of thehaptic feedback is of diminished benefit in a non-colocatedsetup, since there is an offset (albeit constant)between the visual depth cues and those inferred fromhaptic feedback. In this paper we present the results of acomparative study of user performance when using colocatedhaptic interaction versus non-colocated hapticinteraction. The specific tasks involved are designed totest users’ spatial awareness of the simulated environment,and to discover if there is a benefit to visual-hapticco-location. The following section reviews work relatingto the current paper. Section 3 discusses the problems ofimplementing co-location in immersive VR systems.Section 4 explains the set up and the measurements ofco-location. Section 5 presents the results of the experiments.2 Related workImproving the efficiency, friendliness and ease of userinteraction with multimedia devices is one of the maingoals of research in human–computer interaction. Inearly studies, it was demonstrated that direct manipulationof objects improves user performance and systemusability for novice users (Shneiderman 1983). Furtherdevelopment allowed the use of direct interaction forexample through tactile screens or portable devices(Shneiderman 1997). The concept of direct manipulationand direct interaction has been tried in virtual reality(Cheng and Pulo 2003) and augmented reality (Hoffman1998). The addition of passive haptics has also beenshown to enhance presence in VR (Meehan et al. 2001).This confirms the importance of designing direct interactionin VR.While desktop haptics is common (Sensable Technologies,http://www.sensable.com/) adding hapticinteraction in virtual reality is more challenging. A fewhaptic systems have been developed to fit in immersiveenvironments (Loscos et al. 2004; Dettori et al. 2003;Sato 2002). Desktop haptic devices have been used fordirect interaction with stereo-displayed applications, andspecific display devices have been created for that purpose.However with these devices the hand and the deviceare masked by the display surface.Several research studies have investigated the effect ofhaptics and stereovision on systems similar to the Reachindevice (Reachin, http://www.reachin.se/) Arsenaultand Ware performed experiments demonstrating thebenefits of correct visual perspective and haptic feedbackon user performance in a spatial task (Arsenault andWare 2000) Also in the context of co-located haptics,Bouguila et al. (2000) investigated calibration issuesbetween the visual and haptic modes in virtual environments.This study also demonstrated an improvementin users’ perception of depth when the interactionis coupled with haptic feedback. Similarly, Wall et al.(2002) showed that haptic feedback coupled with stereovision assists user performance in achieving tasks. Ernstet al. (2000) demonstrated a benefit of haptic feedback inthe perception of surface orientation via texture. Wareand Rose (1999) investigated the task of object rotationin a virtual environment with reference to a variety offactors. Among their findings was that displacement ofthe visual representation of a real object (held in onehand) by 60 cm led to a 35% slowdown in task completiontime.The notion of co-location of haptic and visual sensorymodes can also be considered in the more generalterms of sensory displacement. Experiments involvingthe use of prisms to spatially offset visual informationcan be traced to the work of nineteenth century psychophysicistsvon Helmholtz (1867) and Stratton (1896)Both noted that adaptation to the ‘‘shifting’’ of objectsin the field of view was achieved over a short period oftime for the simple task of reaching and touching anobject. The precise mechanisms involved in such visuomotoradaptation are still a matter of debate.The experiments described in the current study do notallow for such adaptation effects—we are instead interestedin the ability of people to interact with a virtualenvironment without a training phase. However, it isnotable that in some contexts visuo-motor adaptationhas not been observed to be total: Rolland et al. (1995)describe a study using a head-mounted display thatsignificantly displaces the wearer’s viewpoint. While theyobserved some adaptation of hand–eye co-ordinationduring trials, performance did not reach baseline levels.The aim of the current study is to demonstrate thebenefit of co-located haptic feedback for interactivetasks in a virtual environment. While the studies describedin this section (Arsenault and Ware 2000; Bouguilaet al. 2000; Wall et al. 2002; Ernst et al. 2000; Wareand Rose 1999) point to benefits of co-location (andcosts of non-colocation) in specific contexts, the aim ofthis current study is to investigate the benefit of colocationacross different classes of interactive task, usingboth co-location and haptic feedback as control variables.Additionally, these studies all use an implementationthat relies upon a reflection of a video monitor toachieve co-location. Such a setup cannot be scaled up toan immersive cave-like implementation. Although the

26current study is performed on a desktop setup, the largergoal is to implement co-located haptics in an immersivecave-like setting—this goal is achievable via a scaling-upof the desktop setting implemented here.In common with the studies described, the hapticdevice feedback provided in the study described here is aPhantom desktop force-feedback device, providing forcefeedback and proprioceptive elements, but without anytactile stimulus.As discussed in the following section, co-location isdifficult to implement and may require specifically builtsystems, thus we believe that it is important to show thatco-location is a significant factor for user performance.3 Implementation issues for co-location in immersivevirtual realityVarious technical obstacles must be overcome in orderto successfully implement visual-haptic co-location inimmersive VR environments. Some issues associatedwith depth perception are presented in (Bouguila et al.2000). In this paper, we discuss the issues in a moregeneral context of using haptics with stereo displays inVR. We have identified three broad classes of implementationproblem for visual-haptic co-location: occlusion,accommodation and calibration; we discuss theseproblems in the following sections.3.1 OcclusionFor screen-projection systems (as opposed to HMDs),occlusion problems arise when we reach behind a displayedgraphical object: instead of our hand being occludedby the object, the reverse is the case.3.2 AccommodationAccommodation, or focus, of the user’s eyes is an issue inany stereoscopic display. While appropriate convergenceof the eyes can be elicited by adjusting the disparity of theleft and right stereo pair images according to the distanceof the displayed object from the eyes, accommodation isdetermined by the distance to the display surface (Wannet al. 1994). For HMDs this distance tends to be both verysmall and invariant, leading to a considerable decouplingof accommodation and convergence (by contrast, whenviewing the real 3D world around us, there is a precisecorrespondence between the accommodation and convergencesystems). This accommodation-vergence issue,as well as its associated symptoms of visual fatigue, is lesssignificant with larger-screen 3D display systems, wherethe eye-display separation is larger. However, if a realobject is introduced into the field of view (e.g. the contactpoint of a haptic device) and is spatially co-located with avirtual object, this gives rise to a perceptual dissonance—wecan feel the object at our fingertip via hapticfeedback, but we cannot visually focus on both virtualobject and fingertip simultaneously. While there is nosimple technical solution to this problem, it can be mitigatedby designing, where possible, haptic simulations tolocate haptic contact points close to the display surface.3.3 CalibrationFor visual-haptic co-location to be successfully implemented,it is necessary to align the co-ordinate systemsof both these sensory modes as closely as possible. Thehaptic device itself can normally be calibrated veryprecisely. Once the base of a haptic device is positionedand registered, tracking of the contact point can beachieved with very high (sub-millimeter) accuracy. Errorsin calibration of haptic devices should normally beinsignificant in relation to other sources of error(tracking and display).In large-scale immersive environments (e.g. CAVEs),the user must have their head tracked so that the viewingpoint for the display can be continuously updated. Inthis case, tracking errors will lead to incorrect visualdepth perception. For desktop environments, it is feasiblefor the user of a system to position their eyes at theappropriate viewing points for the left and right projectionsof the stereo display. In this case inadvertenthead movements will lead to incorrect visual perceptionof the 3D location of objects. In either case, the magnitudeof 3D location errors is dependent on severalfactors, all contributing to a decoupling of the visual andhaptic renderings:• The magnitude of the disparity between projectionviewing points and eye positions.• The position of the viewed object relative to the displayscreen.• The position of the viewed object relative to the eyeposition.CRT display systems are liable to non-linearity in bothaxes across the projection surface. Physical measurementstaken of a calibrated grid projected from such adisplay system indicated errors of up to 1.2% of screenwidth in some regions of the screen. This was in spite ofcareful calibration of the CRT projector prior to measurement.Such non-linearity can affect the perceivedposition of a displayed object in all three axes.4 Design of experimentsA series of experiments was designed to test thehypothesis that co-location of the visual and hapticsensory modes will lead to improved task performanceand enhanced sense of presence within a VR environment.In order to evaluate the effect of co-location on userperformance, we designed three experiments to test if

27there is a benefit, in terms of enhanced spatial awarenessof the user, to be derived from co-location. The threeexperiments each test a different aspect of spatialawareness:• Interaction accuracy: the ability to locate, move toand touch an object in 3D space.• Ease of manipulation: the ability to control the motionof an object in 3D space via physical contact.• Agility: the ability to intercept or catch an objectmoving through 3D space.4.1 Equipment and calibrationThe experiments were run on a PC with Pentium 4,1.7 GHz processor and NVidia Quadro FX1100graphics, displayed on a CRT monitor. The participantswore CrystalEyes shutter glasses for stereo viewing.Haptic interaction was provided with a PhantomDesktop from Sensable technology (Sensable Technologies.http://www.sensable.com/) The Phantom device isa force-reflecting interface and can thus affect the forcefeedback and proprioceptive elements of haptic feedback.The workspace area of the device is16 · 12 · 12 cm, appropriate for desktop interaction.The Phantom was positioned to allow co-location andthe full workspace of the device. The interaction workspacewas between the screen and the participant, thesupport being on the right hand side of the participant(see Fig. 1).For each participant there was a two-stage calibrationprocedure. In the first stage, the participant was seated atarms length from the monitor screen such that they werelooking directly at the centre of the screen. Eye-separationvalues used for the stereo rendering were then finelyadjusted to match those of the participant. This wasachieved by asking the participant to compare the depthsof rendered objects to real reference positions.Positioning of the Phantom haptic device to allow colocationformed the second stage of calibration. Thisinvolved manually aligning the Phantom device workspacewith the virtual scene. This manual alignment wasfinely tuned using the Phantom calibration program.The interaction tasks were programmed using GhostSDK and OpenGL.4.2 Task designFor each of the three tasks there are two independentvariables: co-location and haptic feedback. For colocation,the Phantom is carefully positioned such thatthe point of interaction on the Phantom coincidesvisually with the point of contact in the 3D scene. Fornon-colocation, visual markers indicate this point ofcontact. When haptic feedback is turned off, the Phantomis used as a 3D joystick. Thus there are four classesof interaction:• Co-located haptics• Non-colocated haptics• Co-location with no haptic feedback• Non-colocation, no haptic feedback.For all tasks, there are three levels of difficulty, withincreasing numbers of objects, more complex spatialarrangement, and decreasing object size. For each trial,the time taken to complete the task is measured.The first task tests spatial accuracy. The participant isrequired to touch, one by one in a given sequence, a setof objects distributed in 3D space. Three levels of difficultyare designed. On the first level objects are distributedon a plane parallel to the viewport. On the secondlevel, objects are distributed in 3D space, and theirnumber increased. On the third level, the number ofobjects increases and their size is reduced. A screenshotof the setup of this task for level 1 is shown in Fig. 2.For each level, the time spent by the participant tocomplete the task is measured.The second task tests spatial manipulation. It involvesmanipulating a ball through an environmentconsisting of a sequence of objects, akin to moving itthrough a maze. Again, three levels of difficulty weredesigned, similar to the ones used for the task on spatialaccuracy. A screenshot of level 2 is shown in Fig. 3. Foreach level, the time needed for the participant toaccomplish the task is measured.The third task tests spatial response. Gravity is simulatedand the participant must juggle objects in theenvironment. The task stops when an object is dropped.Fig. 1 Experiment set up Fig. 2 Spatial accuracy test, level 1

28Fig. 3 Spatial manipulation, level 2Three levels were implemented, with an increased numberof objects at each level. A screenshot of level 2 isshown in Fig. 4. Time is measured from the beginning ofthe task until the first dropped object.4.3 Experiment procedureA within-groups design was employed on a set of sixparticipants. All participants were male and technicallyliterate,although none had experience of using hapticinterfaces. Each participant was given a verbal andwritten description of the tasks. For the first task, thiswas to touch each object in the environment in a predefinedcolour-coded sequence, with the colour sequencechanging at each level. For the second task, theinstruction was to manipulate a ball such that it touchedeach object in the environment, again in a colour-codedsequence. For the third task, the participant was askedto intercept each object as it fell, effectively juggling withthe objects.The participant was then asked to sit at arm’s lengthfrom the monitor. The system was calibrated (see Fig. 5)for stereo adaptation and visual-haptic co-location. Theparticipant was asked to keep their head as still aspossible to maintain correct stereo and co-location. Atraining period of a few minutes followed.The tasks were then presented in the following order:spatial accuracy, spatial manipulation, then spatial response.The order of presentation of the differentinteraction classes was designed such that any learningeffects would bias against our initial hypothesis—weexpected performance to be best when both hapticFig. 4 Spatial response, level 2Fig. 5 Experiment calibrationfeedback and co-location were implemented, and worstwhen neither were implemented. Thus each task wasperformed using the four interaction classes in the followingorder: co-located haptics; non-colocated haptics;co-location with no haptic feedback; non-co-locationwith no haptic feedback.5 ResultsAll participants completed the set of tasks and timeswere recorded. For the spatial response task, timingstopped when the participant failed to intercept one ofthe falling target objects. For the spatial accuracy andspatial manipulation tasks, timing stopped when the lastobject was touched. In some cases participants touchedthe same object twice or an incorrect object (i.e. not thenext one in the sequence). In the former case this appearedto be due to participants not being certain thatthey had touched the target (thus they touched it again),and in the latter case appeared to be accidental. In allcases participants went on to successfully complete thetasks.The mean times for task completion are plotted inFigs. 6, 7 and 8 for the different tasks. Error bars indicatethe standard deviation across the times recorded forall participants. For Figs. 6 and 7 shorter time indicatesbetter performance (i.e. quicker completion of the spatialaccuracy and spatial manipulation tasks respectively).For Fig. 8, longer time indicates betterperformance (i.e. prolonged ability to respond quickly toobjects moving in 3D space).While the graphs indicate a trend for all tasks ofbetter performance when interaction is via co-locatedhaptics, the error bars indicate substantial variance inthe recorded times.For each of the tasks, results were analysed via twowayrelated-measures ANOVA for each level of difficulty.The levels of difficulty had been introduced to theexperiment to allow measurements at some appropriatelevel (i.e. such that the tasks were neither trivially easynor impossible). Interactions among results from dif-

29Colocated hapticsNo colocation, hapticsColocation, no hapticsNo colocation, No hapticsColocated hapticsNo colocation, hapticsColocation, no hapticsNo colocation, No hapticsSpatial accuracy taskSpatial response task50Time (s)403020100Level 1 Level 2 Level 3Task levelFig. 6 Results for spatial accuracyferent levels of difficulty were not of interest and in anycase it would be difficult to draw conclusions from this.However, adding these additional tasks to the designalso has the negative effect of reducing the power ofresults. This was accounted for in our analysis by theBonferroni method, thus each F-ratio computed by theANOVAs is then divided by 9 (the number of tasks).This correction method is conservative, and lowers therisk of Type I errors, although at the risk of introducingType II errors.For the spatial accuracy task, completion times weresignificantly improved (p < 0.05) by co-location at alllevels, while a marginal improvement (0.05 < p < 0.1)was noted for the effect of haptic force feedback at levelsTime (s)180160140120100806040200Spatial manipulation taskLevel 1 Level 2 Level 3Task levelFig. 7 Results for spatial manipulationColocated hapticsNo colocation, hapticsColocation, no hapticsNo colocation, No hapticsTime (s)4030201002 and 3 of the task. This suggests that for tasks requiringaccurate positioning, haptic feedback alone (i.e. whenunaccompanied by accurate proprioceptive feedback) isof less benefit than having the accurate proprioceptionprovided by co-location (with or without haptic feedback).The results for the spatial manipulation task (Fig. 7)indicate a less pronounced effect, and this is borne outby the lack of statistical significance in the data.The spatial response task was almost impossible toperform unless co-located haptic feedback is implemented.Without co-located haptic feedback, the averagetime that participants managed to maintain thejuggling of objects was less than five seconds. However,this observation is confirmed by statistically significantresults only for level 3 of the task, for which both colocationand haptic feedback elicit a significant(p < 0.05) improvement in task performance.6 ConclusionsLevel 1 Level 2 Level 3Fig. 8 Results for spatial responseTask levelPrevious studies have demonstrated the beneficial effectof haptic feedback on task performance (Hoffman 1998;Sallna¨s et al. 2000). This study additionally indicatesthat co-location is a significant factor in improvinginteraction performance in a 3D environment, for tasksrequiring accuracy and rapid motion in user interaction.The experiments described have been performed only ina desktop setting with limitations in terms of tracking,field of view, freedom of movement and the associatedloss of perceptual cues. However in spite of these limitationsand the low power of the experimental design(mainly due to the small number of participants), significantbenefits for co-location have been demonstrated.The spatial manipulation task produced no significanteffects—it is possibly the case that this type of ‘‘close

30control’’ is not benefited by co-location. It may havebeen useful to measure the relationship between thedegree of control (e.g. mean distance maintained betweenhaptic contact point and target) and task performance.The next step for this research is to extend it to a fullyimmersive VE system equipped with a larger haptic device(Frisoli et al. 2005; Dettori et al. 2003; Sato2002).Head-tracking and a larger haptic workspace will allowus to investigate more fully some of the implementationproblems described earlier. A more immersive systemwill also enable a broader investigation incorporatingthe impact of multi-sensory co-location on presence.Acknowledgements We would like to thank Deepti Narwani andEmanuele Ruffaldi for their help in the current study. The workpresented in this paper was partially funded by the collaborativeEuropean project PUREFROM (IST-2000-29580), a 3-year RTDproject funded by the 5th Framework Information Society Technologies(IST) Programme of the European Union.ReferencesArsenault R, Ware C (2000) Eye–hand co-ordination with forcefeedback. In: Proceedings of the SIGCHI conference on humanfactors in computing systems, The Hague, pp 408–414Bouguila L, Ishii M, Sato M (2000) Effect of coupling haptics andstereopsis on depth perception in virtual environment. Worldmulticonference on systemics, cybernetics and informatics (SCI2000), Orlando, pp 406–414Caclin A, Soto-Faraco S, Kingstone A, Spence C (2002) Tactile‘‘capture’’ of audition. Percept Psychophys 64(4):616–630Cheng K, Pulo K (2003) Direct interaction with large scale displaysystem using infrared laser tracking devices. In: Australasiansymposium on information visualisation, Adelaide, 2003.Conferences in research and practice in information technology,vol 24Dettori A, Avizzano CA, Marcheschi S, Angerilli M, BergamascoM, Loscos C, Guerraz A (2003) Art touch with CREATEhaptic Interface, ICAR 2003. In: 11th International conferenceon advanced robotics, University of Coimbra, Portugal, June30–July 3Dinh HG, Walker N, Hodges LF, Song C, Kobayashi A (1999)Evaluating the importance of multi-sensory input on memoryand the sense of presence in virtual environments. In: RosenblumL, Astheimer P, Teichmann D (eds) Proceedings of theIEEE virtual reality ‘99 conference. IEEE Computer SocietyPress, Los Alamitos, pp 222–228Durlach P, Fowlkes J, Metevier C (2005) Effect of variations insensory feedback on performance in a virtual reaching task.Presence 14(4):450–462Ernst MO, Banks MS, Bulthoff HH (2000) Touch can change visualslant perception. Nat Neurosci 3(1):69–73Frisoli A, Jansson G, Bergamasco M, Loscos C (2005) Evaluationof the pure-form haptic displays used for exploration of worksof art at museums, World haptics conference, Pisa, March 18–20Hayward V, Astley OR, Cruz-Hernandez M, Grant D, Robles-De-La-Torre G (2004) Haptic interfaces and devices. Sensor Rev24(1):16–29von Helmholtz H (1867) Treatise on physiological optics, vol III(English translation by Southall J, 1925)Hoffman HG (1998) Physically touching virtual objects using tactileaugmentation enhances the realism of virtual environments.In: Proceedings of the IEEE virtual reality annual internationalsymposium’98, Atlanta, pp 59–63Jack CE, Thurlow WR (1973) Effects of degree of visual associationand angle of displacement on the ‘‘ventriloquism’’ effect.Percept Mot Skills 37:967–979Loscos C, Tecchia F, Carrozino M, Frisoli A, Ritter Widenfeld H,Swapp D, Bergamasco M (2004) The museum of pure form:touching real statues in a virtual museum, VAST 2004. In: 5thInternational symposium on virtual reality, archaeology andcultural heritage, BrusselsMeehan M, Insko B, Whitton M, Brooks FP (2001) Physiologicalmeasures of presence in virtual environments. In: Proceedingsof 4th international workshop on presence, Philadelphia, pp21–23Rolland JP, Biocca FA, Barlow T, Kancherla A (1995) Quantificationof adaptation to virtual-eye location in see-thru headmounteddisplays. In: Proceedings of the VRAIS, pp 56–66Sallnäs E, Rassmus-Gro¨hn K, Sjo¨stro¨m C (2000) Supportingpresence in collaborative environments by haptic force feedback.ACM Trans Comput-Hum Interact 7(4):461–476Sato M (2002) Development of string-based force display: SPI-DAR. In: 8th International conference on virtual systems andmultimedia (VSMM2002), Gyeongju (alias Kyongju), KoreaShneiderman B (1983) Direct manipulation: a step beyond programminglanguages. IEEE Comput 16(8):57–69Shneiderman B (1997) Designing the user interface, Chapter 9,interaction devices (Sections 9.1–9.3), 3rd edn. Addison-Wesley,Reading, pp 306–327Stratton G (1896) Some preliminary experiments on vision withoutinversion of the retinal image. Psychol Rev 3:611–617Wall SA, Paynter K, Shillito AM, Wright M, Scali S (2002) Theeffect of haptic feedback and stereo graphics in a 3D targetacquisition task. In: Proceedings of eurohaptics 2002, Universityof Edinburgh, 8–10th July, pp 23–29Wallace M, Roberson G, Hairston W, Stein B, Vaughan J, SchirilloJ (2004) Unifying multisensory signals across time and space.Exp Brain Res 158(2):252–258Wann J, Rushton S, Mon-Williams M (1994) Natural problems forstereoscopic depth perception in virtual environments. Vis Res35(19):2731–2736Ware C, Rose J (1999) Rotating virtual objects with real handles.ACM Trans Comput-Hum Interact 6(2):162–180