RAPID COMMUNICATION

Equivalence of Real-World and Virtual-RealityRoute Learning: A Pilot Study

Joanne Lloyd, Ph.D.,1 Nathan V. Persaud, M.Sc.,1 and Theresa E. Powell, Ph.D.1,2

Abstract

There is good evidence for effective transfer of learning from virtual to real-world environments, and this holdstrue even for complex spatial tasks such as route learning. However, there is little research into the simpleequivalence of an individual’s performance across real and virtual environments, an important topic whichcould support the use of virtual reality as an assessment and research tool. This pilot study compared route-learning performance in a desktop virtual town with performance around a real-world route. Participantswere ‘‘driven’’ around a route through a virtual town and around a different (but equally complex) routethrough a real-world suburb, then asked to direct the driver back around each of the routes from memory. Theycompleted strategy checklists after learning each route. Results indicated good equivalence between the realand virtual environments, with comparable error rates and no differences in strategy preferences. This dem-onstrates that simple desktop virtual environments may be a useful tool for assessment of and research intoroute learning.

Introduction

While the value of virtual reality (VR) in risky sit-uations such as flight training1 and surgical training2 is

well recognized, a less obvious field in which it has recentlybeen embraced is that of wayfinding. It has been found, forexample, that spatial information learned in a simulated en-vironment can be transferred to the real world; this haspractical implications for scenarios such as military training,where soldiers can familiarize themselves with territory vir-tually to improve their navigation efficiency in subsequentreal-world maneuvers.3 In psychological research, VR can beused to generate completely novel, controlled, and consistentenvironments, where participants’ wayfinding can be studiedwithout prior familiarity with a location confounding results.It also circumvents mobility impairment and physical fatigue(particularly when studying walked routes), and because of‘‘time compression,’’3 it can allow more to be achieved withina given time period.

Good evidence for the ecological validity of virtual way-finding comes from studies showing transfer of learning fromreal to virtual environments. Researchers have demonstratedtransfer of knowledge about the layout of a school buildingfrom a desktop virtual simulation to the real-world building

itself in children with mobility impairments4–6 and have evenobserved generalized improvement on nontrained tasks.4,5 Inan outdoor orienteering task, Darken and Banker3 found thatfor some participants (i.e., ‘‘intermediate’’ standard naviga-tors), preexposure to a real-world environment in the form ofa virtual simulation resulted in better performance than pre-exposure to a map or even to the real world itself; and Farrellet al.7 demonstrated transfer of spatial knowledge from avirtual building to its real-world counterpart.

Other studies demonstrate the equivalence of wayfindingacross real and virtual environments. Ruddle et al.8 highlightsimilar patterns of biases when people learn layouts of realand virtual buildings. In a virtual office, they found learn-ing by navigation to promote most accurate estimations of‘‘path’’ distances, while learning from a map resulted in su-perior estimates of ‘‘as-the-crow-flies’’ distances, consistentwith biases seen in a similar real-world task.9 Studies ofparticipants with acquired brain injury (ABI) also support theequivalence of real and virtual wayfinding, showing real-world spatial learning deficits to be mirrored in both sim-plistic10 and complex11 virtual environments.

A relatively neglected question, however, concerns whe-ther individuals’ route-learning performance as measured ina virtual task is equivalent to their real-world performance. If

1School of Psychology, University of Birmingham, Edgbaston, Birmingham, United Kingdom.2Moor Green, South Birmingham PCT, Moseley Hall Hospital, Moseley, Birmingham, United Kingdom.

CYBERPSYCHOLOGY & BEHAVIOR

Volume 12, Number 4, 2009ª Mary Ann Liebert, Inc.DOI: 10.1089=cpb.2008.0326

423

real-world wayfinding skill can be correlated with perfor-mance on a virtual task, we can demonstrate the validity ofvirtual environments as an arena for psychological researchand assessment. The present study therefore directly com-pares individuals’ performance when learning a real-worldroute with their performance in a route through a virtualtown.

Route-learning strategies are an important considerationbecause they impact performance,12,13 and considerable in-dividual differences in preference are observed.14,15 Majorstrategies include the use of a cardinal reference system(i.e., compass points), turn-sequence memorization, and land-marks.13 Preferred landmarks may be proximal (set alongone’s route) or distal (visible from a distance).12,16 Somewayfinders continuously update their position relative to astart point in a technique known as ‘‘dead reckoning.’’17

Others use a ‘‘look-back’’ strategy,18 turning around period-ically to familiarize oneself with a route’s return perspective.A particularly effective wayfinding strategy for someeg,19 is‘‘cognitive mapping,’’ or the formation of a mental map of anenvironment.20 In order to enhance comparison of routelearning in the real and virtual environments, checklistsasking participants to rate the extent to which they used eachof these typical wayfinding strategies were administered afterboth the virtual and the real-world route-learning tasks.

Method

Design

A repeated-measures, within-participants design was em-ployed. All participants learned a route through a VR townand a route through a different real-world town (of similarlength and complexity). The order of conditions was coun-terbalanced to control for practice and fatigue effects. Theindependent variable was the route type (real or virtual), andthe main dependent variable was number of errors made onattempting to recall the route. An additional dependent var-iable was self-reported strategy use.

Participants

Participants were a convenience sample of 14 neurologi-cally healthy volunteers, 8 males and 6 females, ages 18 to 54years, with a mean of 23.1 years (SD¼ 9.21). Participantswere recruited through word-of-mouth and through theuniversity research participation scheme. All participantsreceived an information leaflet explaining the study and gaveinformed consent. Exclusion criteria were prior experiencewith the video game used in the VR condition and familiaritywith the area in which the real-world route was located.

Materials

The VR route was presented via a Sony PlayStation 2games console, linked via a SCART lead to a 21-inch colortelevision. The software used was a modified version of thecommercial title Driv3r (Reflections Interactive Ltd.), anATARI studio, in which extraneous game-related informationhad been removed from the screen in order to reduce dis-tractions. The virtual environment in which the route was setwas a simulation of the real-world town of Nice, France. Thereal-world route was a completely different route of similarlength and complexity, located in a suburb of Birmingham,

United Kingdom. The location was an area with whichpotential participants were unlikely to be familiar, beingreasonably distant from the university campus and studentarea.

The routes each contained a total of 15 turns, which brokedown into equivalent percentages of left, right, and straightahead choices and equivalent percentages of 2- and 3-choicepoint junctions. Pilot testing found that the average timetaken to complete the real-world route was 5 minutes, and theaverage time taken to complete the VR route was 4 minutes50 seconds. It also took a similar amount of time to return tothe start point of the real-world route (4 min.) and the VRroute (3 min. 30 sec.). The 2 routes are shown in Figure 1.

Participants rated the extent to which they relied on eachof the 9 route-learning strategies, summarized in Table 1, on a5-point Likert-type scale with the categories 0, not at all; 1,a little; 2, a moderate amount; 3, a lot; and 4, almost totally.

Procedure

In each condition, participants were taken to the start of theroute, shown once around the route, and then returned to thestart point. From here, they were instructed to call out direc-tions to the experimenter, with the goal of retracing the sameroute. Each direction had to be called out at least 5 secondsprior to actually reaching the turning point. This was in orderto allow the experimenter to drive safely, using indicators.This could be conceived as having added ecological validityto the task, as it mimics the task demands of directing a driverin a real-life scenario. Participants were asked to call outdirections even when the correct choice at a junction was tocontinue straight ahead. One point was awarded for eachcorrect instruction, with a maximum score of 15 points. Im-mediately after completing each test trial, participants wereasked to fill in a questionnaire rating the degree to which theyhad employed a variety of common wayfinding strategies.

In the VR condition, participants were seated at approxi-mately arm’s length from the television screen and asked towatch as the experimenter drove them along a route throughthe town, imagining that they were actually traveling throughthe virtual environment. After being taken around the routeonce, they were returned to the start. Then, as in the real-worldtask, they were instructed to direct the experimenter backaround the route, ensuring directions were called out 5 sec-onds in advance.

Results

The mean number of errors made in the virtual environ-ment was 2.57 (SD¼ 1.01), and the number made in the realworld was 2.43 (SD¼ 1.55). Paired-samples t test revealedthat there was no significant difference in errors made be-tween the two conditions (t¼�0.56, df¼ 13, p¼ 0.58); rather,there was a highly significant correlation between them(r¼ 0.81, n¼ 14, p< 0.0005).

Correlations between frequency of use of strategies in thereal and virtual environments are shown in Table 1. Spear-man’s correlations are used because ratings are on a Likert-type scale, for which nonparametric tests are deemed mostappropriate.21

The extent to which people rely on creation of a cognitivemap, use of landmarks, guessing, and instinct are all corre-lated across the real and virtual environments ( p’s< 0.05).

424 LLOYD ET AL.

Correlations between virtual and real-world use of cardinalreference points and tall distant buildings as landmarksnarrowly miss significance ( p’s # 0.07). There was no signif-icant correlation between real and virtual tasks in the relianceon learning sequences of left and right turns, use of dead

reckoning, and use of the look-back strategy ( p’s> 0.05).However, paired-samples t tests showed that there were nosignificant differences between the extents to which any of thestrategies were employed in the real versus virtual environ-ments ( p’s> 0.05).

FIG. 1. A: Real-World Route; B: Virtual Route.

Table 1. Correlations between Use of Strategies in Real and Virtual Route Learning Tasks*

Spearman’s correlations between frequency of use of strategiesin real and virtual environments (n¼ 14, 2-tailed)

Strategy Rho p

Creating a cognitive bird’s-eye map 0.88 <0.0005Using landmarks 0.58 0.03Guessing 0.53 0.05Using instinct 0.54 0.05Using cardinal reference points 0.53 0.05Using dead reckoning 0.45 0.11Using tall, distant buildings 0.50 0.07Memorizing turn sequences 0.37 0.19Using look-back strategy �0.08 0.80

*Shows nonparametric correlations of Likert-type scale frequency-of-use ratings for strategies across real and virtual route-learningconditions.

EQUIVALENCE OF REAL AND VIRTUAL ROUTE LEARNING 425

Discussion

The present study compared route-learning performancein a VR town with performance in a distinct real-world routeof similar length and configuration. Participants rated theextent to which they used 7 major wayfinding strategies,along with instinct and guesswork, when learning each route.Results showed a strong correlation between the number oferrors made in the real and virtual environments, supportingthe equivalence of real and VR route learning. Furthermore,use of the majority of strategies was significantly correlatedacross the real and virtual environments. The few strategiesnot showing a significant correlation did not differ signifi-cantly across the two environments.

The equivalence of performance between the virtual andreal worlds is consistent with previous findings both in gen-eral VR research22 and in route-learning studies.5,7,8 Whereasprevious studies have demonstrated equivalence of real andvirtual route learning through similar patterns of spatialknowledge acquisition22 and impairment10 in real and virtualtasks and in transfer of learning from a virtual environment toits real-world counterpart,3 the present study demonstratesequivalence using a real town and a virtual town that arecompletely distinct from one another. Therefore, we demon-strate that general route-learning ability can be assessed in aVR town and that performance on a generic virtual route isrepresentative of real-world route-learning aptitude.

The correlations seen between strategy use across the realand virtual conditions also support the ecological validity ofVR for studying route learning, demonstrating that peoplerely on similar techniques to memorize a real and a virtualroute. Although not every strategy was significantly corre-lated across real and virtual routes, the lack of any significantdifference in usage is encouraging. Possible explanations forthe lack of outright correlation include the prospect that smallsample size resulted in poor power of the statistical tests todetect correlations for the scarcely used techniques with smallamounts of variance, such as the look-back strategy.

This study is limited by its small sample size, and furthercomparisons of real and virtual wayfinding performance andtechniques are needed before concrete conclusions arereached. If the present results can be replicated, it seems thatsimple, desktop VR using commercially available softwarecould be an economical and ecologically valid tool for psy-chological research. This could be valuable as a research toolfor use in populations with route-learning deficits, such aspeople with acquired brain injury23 or Alzheimer dementia.24

Assessing or analyzing route learning virtually is timeefficient, minimizes fatigue, and circumvents issues of dif-ferential familiarity with real-world environments that canconfound between-participants studies.

Disclosure Statement

No competing financial interests exist.

References

1. Hays RT, Jacobs JW, Prince C, et al. Flight simulator trainingeffectiveness: a meta-analysis. Military Psychology 1992;4:63–74.

2. McCloy R, Stone R. Science, medicine, and the future: virtualreality in surgery. British Medical Journal 2001; 323:912–5.

3. Darken RP, Banker WP. (1998) Navigating in natural envi-ronments: a virtual environment training transfer study.Proceedings of the Virtual Reality Annual International Sympo-sium (VRAIS).Washington, DC: IEEE Computer SocietyPress, pp. 12–9.

4. Stanton D, Foreman N, Wilson PN. (1998) Uses of virtualreality in clinical training: developing the spatial skills ofchildren with mobility impairments. In Riva G, ed. Virtualenvironments in clinical psychology and neuroscience. Am-sterdam, Netherlands: IOS Press, pp. 219–32.

5. Stanton D, Wilson P, Foreman N, et al. (2000) Virtualenvironments as spatial training aids for children andadults with physical disabilities. In Sharkey P, Cesarani A,Pugnetti L, et al., eds. Proceedings of the 3rd InternationalConference on Disability, Virtual Reality and Associated Tech-nologies (ICDVRAT). Sardinia, Italy: University of Reading,pp. 123–8.

6. Wilson PN, Foreman N, Tlauka M. Transfer of spatial in-formation from a virtual to a real environment in physi-cally disabled children. Disability & Rehabilitation 1996;18:633–7.

7. Farrell M, Arnold P, Pettifer S, et al. Transfer of routelearning from virtual to real environments. Journal of Ex-perimental Psychology: Applied 2003; 9:219–27.

8. Ruddle RA, Payne SJ, Jones DM. Navigating buildings in‘‘desk-top’’ virtual environments: experimental investiga-tions using extended navigational experience. Journal ofExperimental Psychology: Applied 1997; 3:143–59.

9. Thorndyke PW, Hayes-Roth B. Differences in spatial knowl-edge acquired from maps and navigation. Cognitive Psy-chology 1982; 14:560–89.

10. Skelton RW, Bukach CM, Laurance HE, et al. Humans withtraumatic brain injuries show place-learning deficits incomputer-generated virtual space. Journal of Clinical & Ex-perimental Neuropsychology 2000; 22:157–75.

11. Spiers HJ, Burgess N, Maguire EA, et al. Unilateral temporallobectomy patients show lateralized topographical and epi-sodic memory deficits in a virtual town. Brain 2001; 124:2476–89.

12. Cornell EH, Soronsen A, Mio T. Human sense of directionand wayfinding. Annals of the Association of AmericanGeographers 2003; 93:399–415.

13. Kato Y, Takeuchi Y. Individual differences in wayfindingstrategies. Journal of Environmental Psychology 2003; 23:171–88.

14. Lawton CA. Gender differences in wayfinding strategies:relationships to spatial ability and spatial anxiety. Sex Roles1994; 30:765–79.

15. Lawton CA, Kallai J. Gender differences in wayfindingstrategies and anxiety about wayfinding: a cross-culturalcomparison. Sex Roles: A Journal of Research 2002; 47:389–401.

16. Jacobs WJ, Laurance HE, Thomas KGF. Place learning invirtual space 1: acquisition, overshadowing, and transfer.Learning & Motivation 1997; 28:521–41.

17. Cutmore TRH, Hine TJ, Maberly KJ, et al. Cognitive andgender factors influencing navigation in a virtual environ-ment. International Journal of Human-Computer Studies2000; 53:223–49.

18. Heth DC, Cornell EH, Flood TL. Self-ratings of sense of di-rection and route reversal performance. Applied CognitivePsychology 2002; 16:309–34.

19. Siegel AW, White SH. (1975) The development of spatialrepresentations of large-scale environments. In Reese HW,

426 LLOYD ET AL.

ed. Advances in child development and behavior. New York:Academic Press, pp. 9–55.

20. Prestopnik JL. Roskos-Ewoldsen B. The relations amongwayfinding strategy use, sense of direction, sex, familiarity,and wayfinding ability. Journal of Environmental Psychol-ogy 2000; 20:177–91.

21. Clason DL, Dormody TJ. Analyzing data measured by in-dividual Likert-type items. Journal of Agricultural Educa-tion 1994; 35:31–5.

22. Rose FD, Attree EA, Brooks BM, et al. Training in virtualenvironments: transfer to real world tasks and equivalenceto real task training. Ergonomics 2000; 43:494–511.

23. Barrash J, Damasio H, Adolphs R, et al. The neuroanatom-ical correlates of route learning impairment. Neuropsycho-logia 2000; 38:820–36.

24. Uc EY, Rizzo M, Anderson SW, et al. Driver route-followingand safety errors in early Alzheimer’s disease. Neurology2004; 63:832–7.

Address correspondence to:Dr. Theresa E. Powell

School of PsychologyUniversity of Birmingham

EdgbastonUnited Kingdom, B15 2TT

E-mail: T.Powell@Bham.ac.uk

EQUIVALENCE OF REAL AND VIRTUAL ROUTE LEARNING 427