1995 VR Conference Proceedings

Go to previous article 
Go to next article 
Return to 1995 VR Table of Contents 


Standen, P.J. & Cromby, J.J.

Department of Learning Disabilities, University of Nottingham
Queens Medical Centre, Clifton Boulevard
Nottingham, England NG7 2UH


** Overview

This paper describes a successful attempt to show generalisation to the real world of a skill practised in a desktop virtual environment. The people who took part in the study were a group of students with severe learning difficulties, who got better at a real supermarket shopping task after practising in a virtual supermarket.

The paper describes the special features of the people who took part, then draws upon developmental psychology to explain why virtual environments may be a particularly valuable tool in their education. The rationale for using desktop virtual environments in education is discussed, and the typical problems of producing generalisation and maintenance of learned skills amongst students with severe learning disabilities are outlined. Other attempts to show generalisation to the real world of skills acquired in virtual environments are reviewed, and the method and findings of our own work are presented and discussed.

** People with severe learning disabilities

Since terminology in this area is potentially confusing it is necessary to begin with some definitions. In this paper, the term "people with learning disabilities" is used to refer to adults, whilst "students with learning difficulties" is used to refer to younger people. The participants in this study were all students with severe learning disabilities attending a special school in Nottingham, England. In other countries, these people might be described as severely mentally retarded, mentally handicapped or developmentally disabled.

In the United Kingdom about three or four people in every thousand are described as having severe learning disabilities, defined as having an IQ of less than 50. They are an extremely heterogenous group. About 80% have some additional physical impairment, in addition to being learning disabled. Mobility problems, sensory impairments and heart defects are all common. Some have no speech, and must use signing systems to communicate. About 20% have recognised medical/genetic conditions such as Down Syndrome or Fragile X, and others are know to have been subject to pre- or perinatal injury or abuse, but the majority have no more specific medical diagnosis than "learning disabled".

Students with severe learning difficulties pose many problems for teachers. A class of such students typically encompasses a much broader spread of ability than a class in a mainstream school. They will also have a wide range of sensory and physical impairments, which the teacher has to take into consideration. In the UK long stay residential facilities for people with learning disabilities are being closed down and their residents cared for in the community. With this there has been a recognition of the importance of education for independent living and social competence (Siperstein, 1992). The growing expectation that many of the students currently in special education will one day look after themselves with only minimal support means that the need to devise more effective ways of educating them is more urgent than ever before.

Unlike other new technologies which have first of all been used in mainstream education and only later made their way across to special education, the potential of computer based virtual environments is already as apparent in special education as it is in mainstream schools. We now draw upon developmental psychology to explain why we think virtual environments may be especially valuable in the education of people with learning disabilities.

** Virtual Environments and special education

We have identified four reasons why virtual environments may be especially useful to teach life skills to people with learning disabilities.

First, virtual environments encourage active involvement in learning and give the user the experience of control over the learning process (Pantelidis, 1993). The crucial role of self-directed activity in learning has high priority in the developmental theories of Bruner (1968), Vygotsky (1978) and Piaget (see Wood, 1988). Research on the development of perception (Held and Hein, 1963) also emphasised the essential nature of self-directed activity and for individuals who are limited in the extent to which they can do this in the real world virtual environments provide an opportunity to initiate actions for themselves. A tendency to passivity has been noted in students with learning difficulties (Sims, 1994), which virtual environments may help to overcome.

Second, by reflecting the real world, virtual environments create the opportunity for people with learning disabilities to learn by making mistakes but without suffering the real consequences of their errors. This echoes the role accorded to play by both Vygotsky and Bruner. Play is implicated in the acquisition of knowledge about social roles, and also in the development of an adequate and accurate sense of self (Fein, 1984). In the real world, students with learning difficulties are often prevented from having playful experiences which might promote their further development by a combination of prejudice and discrimination, mobility or other impairments, and because their carers are scared of the consequences of allowing them to do things on their own (Shakespeare, 1975).

Third, in virtual environments rules and abstract concepts can be conveyed without the use of language or other symbol systems. Virtual environments have their own "natural semantics" (Bricken, 1991): the qualities of objects can be discovered by direct interaction with them. They can thus be used to facilitate concept attainment through practical activity, by-passing the need for disembedded thinking (Donaldson, 1978) which people with learning disabilities find especially difficult to acquire and use.

Fourth, provided they are accessed through a carefully designed interface and control medium, virtual environments can minimise the effects of many physical disabilities. Whether or not a student can easily move herself around in real life, in a virtual environment she can have complete control over her actions, be free to wander and explore anywhere she likes.

These features of virtual environments combine to make them a uniquely empowering tool in special education. Power is an important consideration, since there is much evidence that people with learning disabilities are a uniquely disadvantaged group whose disempowerment extends to all areas of everyday life. By giving control to the learner, facilitating playful activity, avoiding abstract language and negating or minimising the effects of physical impairments, virtual environments might be an empowering medium for people with learning disabilities to practise and acquire new skills. This empowering effect means that virtual environments may help to stimulate curiosity and overcome apathy and dependence.

** The Technology

Although virtual environments are typically associated with the use of head mounted displays, earphones and gloves which surround the user with a computer generated environment, presence in a virtual environment is not dependent upon either full immersion or any particular technological component. Steuer (1992) argues that since presence is an experience, attempts to define virtual environments in terms of technology will always be flawed, and that a conceptual definition should therefore be adopted. Helsel (1992) argues that in education a conceptual approach is especially valuable since it leads to a focus upon the cognitive, emotional and social experiences of the user, with the result that the technology becomes no more than a tool for replicating, creating or enhancing these educational experiences.

It is sometimes assumed that the maximum benefits can be gained from virtual environments by experiencing them in the most immersive way possible, and as a result there may be a tendency to view desktop virtual environments as poor second cousins to the real thing. Travis, Watson & Atyeo (1994) question this assumption. They note that even with the best systems currently available image quality is poor, there are significant time lags when the user's head moves, and that interacting with the computer images "is like feeling for an object in the dark - while wearing boxing gloves". Since the development of virtual environments is closely allied to the drive to make computers more intuitive and friendly to use, they argue that we should not forget that different computer users have particular tasks to perform, for which immersion may or may not be useful or helpful.

There are a number of reasons why in work with people with learning disabilities it is preferable to use desktop virtual environments. First, there are still unresolved health and safety issues associated with the use of head mounted displays which would make their use unethical. Second, because desktop systems display the virtual environment publicly they facilitate peer and tutor interaction with the learner. The educational value of instruction and peer interaction with others was highlighted by Vygotsky (1978), and the need to give appropriate and timely instruction is emphasised by Wood (1988). Unlike desktop systems, head mounted displays make the user look and feel isolated. This could be a particular problem for students with severe learning difficulties whose social isolation is often already great. Third, desktop virtual environments can be run on an entry-level IBM compatible computer, which brings them realistically within the budget of most schools: by contrast, immersive systems are currently much more expensive, and few schools could afford to purchase them until prices fall substantially.

** Generalisation

For virtual environments to be of any use in education, learning obviously has to generalise to similar experiences in physical reality. However there have been doubts about the ability of people with learning disabilities to transfer skills they have learnt in one environment to another. Clinicians and educators have frequently noted that individuals with severe, moderate or mild learning disabilities experience extreme difficulties in the generalisation of initial learning to novel situations (Stokes and Baer, 1977; Ward and Gow, 1982). Work on concept formation (Blount, 1968; Furth and Milgram, 1965; Landau and Hagan, 1974)) suggests that this result arises because competence in these tasks is measured with verbal responses or by the reliance on active retrieval strategies both of which provide difficulties for those with learning disabilities. Zigler and Balla (1982) also suggest that these results suggesting rigidity and inability to generalise arose from monotonous laboratory tasks and could be explained by the participants' attempts to gain social approval from the experimenter. Studies of the acquisition of vocational and independent living skills which try to avoid these contaminating factors have demonstrated a higher level of transfer to novel situations (Horner and McDonald, 1982; Day and Horner, 1986; Baty, Michie and Lindsay, 1989).

There is also evidence that people with learning disabilities fail to maintain performance gains for any significant period after training or practice ceases (Robertson, Richardson and Youngson, 1984; Davies and Rogers, 1985). In any case, maintenance is linked to generalisation: if skills fail to transfer out of the training situation, then clearly they will be seen as irrelevant once training ceases, and so are unlikely to be maintained.

So far, there have been few published studies which examine generalisation to the real world of competencies practised in virtual environments, even though there is already evidence that virtual environments can be used successfully in education and skill acquisition. Theasby (1992) points out that simulators of various kinds have been successfully used for many years to train complex skills where the errors made by novices could have dangerous or costly consequences, such as piloting an aircraft or managing a nuclear power plant. Wilson (1993) describes a successful attempt to teach children with physical disabilities the location of fire exits and appliances using a virtual environment which then transferred successfully to the environment that had been modelled. However, Kozak et al (1993) describe a failed attempt at generalisation, and attribute their lack of success to both the low veridacity of the virtual environment and the simplistic nature of the task, which involved moving cans to target locations.

The study to be reported set out to investigate whether students with severe learning difficulties could transfer to the real world skills they had acquired in a virtual environment. The skill we chose to investigate was shopping in a supermarket. This was chosen because it had high ecological validity, it was reasonably straightforward to test performance in the real world, and performance could be easily quantified both on the virtual and the real task.


** Selection of participants

The research took place at a special school for students with severe learning difficulties aged between 3 and 19 years. This school is described in another presentation. School staff nominated students to take part in the research if they met the following criteria:

  1. they had sufficient motor skills and visual ability to be able to use the computer terminal and joystick.
  2. they were sufficiently able to carry out a real shopping trip with minimal staff support
  3. they had used virtual environments on at least three previous occasions
  4. the parents or carers gave consent for their child to take part in the research
This resulted in a sample of 23 students whose teachers then completed the classroom version of the Vineland Adaptive Behaviour Scale (VABS). The students were then assigned to either the experimental or control group so that the two groups were matched on age, gender and Vineland score. Their characteristics are shown on Table I. Their parents were asked to complete a short questionnaire detailing how much their child accompanies them and helps them with shopping. There was no difference between the two groups on these measures.
Mean Age Gender Mean VABS
Experimental group 15.29 5M / 6F 51.8
Control group 15.67 6M / 4F 52.54

Table 1: characteristics of participants

** Design

Students' baseline performance on a shopping task in a real supermarket was compared with their performance after an eleven week interval during which the experimental group had twice-weekly sessions using a virtual supermarket. During this time the control group had access to other virtual environments, but were not permitted to use the virtual supermarket.

** Procedure

1) The real shopping task
Students were taken to the local supermarket and individually given a shopping list of four items which they were to find, put in their trolley and take to the checkout. The list was one of five made from both miniaturised colour reproductions of packets of commonly available consumer goods such as washing powder and breakfast cereal and actual labels cut from other items such as tinned goods, which were then mounted under clear film onto pieces of A4-sized card.

The students' performance was monitored by the accompanying adult who maintained a distance from the student and recorded the following measures:

  1. Total time:
    time taken from passing through the turnstile into the shopping area until stopping at the checkout with their shopping
  2. Total number of items:
    all items placed in the trolley even if they were later put back onto the shelves before reaching the checkout
  3. Total items at end:
    number of items in the trolley on arrival at checkout
  4. Total items correct:
    number of items from the shopping list in the trolley on arrival at the checkout

2) The virtual supermarket
All the virtual environments used were built using the Superscape world- building kit produced by Dimension International. They are displayed on an ordinary 14" colour SVGA monitor, driven by an IBM-compatible 486/DX2-66 computer with 8mb of RAM.

The supermarket consisted of a two-aisle store with movement through the store achieved by use of a joystick and selection of items by a mouse. Shelves were filled with a representative selection of goods found in the local supermarket.

There were five different lay-outs of goods on the shelves which the students were presented with at random each time they started a session. Students were given a list as described above and faced with the same task. Each session was initiated by entering the student's identifying number and the number of the shopping list used. The same set of measures that the staff collected on the real shopping trip were then obtained from the computer's record or by a researcher. In addition the researcher recorded the amount of time the students spent actually looking at the screen as a proportion of the total time they took to complete the task, to provide a quantitative measure of each students' level of engagement with the task.

3) Other virtual environments
While the experimental group used the virtual supermarket the control group were free to use any of the other virtual worlds which included a virtual city, a virtual house and a ski-slope.

In the last two days before the return visit to the real supermarket, students in the control group played a game with the shopping lists to ensure that they were not disadvantaged in familiarity with the shopping lists when they returned for the second real shopping trip.


Examination of both the mean times taken to complete the virtual shopping task and the measures of distraction gathered by the researcher showed that the performance of two students in the experimental group had not noticeably improved. If no learning had taken place there would be no skill to generalise; these two students were therefore excluded from the analysis

** There was a significant difference between the experimental group and the control group in the time taken to complete the shopping task on their return to the real supermarket:


Source of   Sum of            Mean            Sig
Variation   Squares     DF   Square      F   of F
TIME 1       56.139      1   56.139    2.774  .115
Main effect:
GROUP       145.177      1  145.177    7.173  .016
Explained   190.764      2   95.382    4.713  .025
Residual    323.819     16   20.239
Total       514.583     18   28.588

Examination of group means shows that the experimental group were nearly five minutes faster than the control group on their return visit to the real supermarket:

1st Visit 2nd Visit
Experimental group 11.97 11.59
Control group 11.21 16.92

Table 2: Mean times taken to complete the real-world shopping task

** On average, the experimental group also got significantly more items correct on their return visit to the supermarket than the control group: Mean number of items correct (target value = 4) S.D.

Experimental group 3.11 .78
Control group 2.20 1.32

Mann-Whitney U test: U=26.0, P= .05 (one-tailed)

Table 3: Mean number of items correct on the real-world shopping task

This result cannot be explained by the experimental group simply rushing around the supermarket and throwing as many items in their trolleys as possible, since the experimental group averaged fewer items in their trolleys on the return visit than the control group:

1st Visit 2nd Visit
Experimental group 4.11 4.0
Control group 4.2 6.4

Table 4: Mean number of items picked up on the real world shopping task

** The experimental group got significantly closer to the target value of 4 items in their trolleys on arrival at the checkout:

Mean error (from (target value of 4) S.D.

Experimental group .222 .441
Control group 3.7 7.59

Mann-Whitney U test: U=18.5, P=.01 (one-tailed)

Table 5: Mean error in number of items on arrival at the checkout

** There was a significant effect upon shopping times on the second visit to the real supermarket, according to whether or not students regularly went shopping with their parents/carers. However, the interaction between the effects of group and shopping with parents was not significant:


Source of Variation          SS      DF        MS         F  Sig of F

WITHIN+RESIDUAL          233.20      14     16.66
REGRESSION                 1.26       1      1.26       .08      .787
GROUP                    226.86       1    226.86     13.62      .002
WITH PARENTS              78.27       1     78.27      4.70      .048
GROUP by WITH PARENTS     51.62       1     51.62      3.10      .100
(Model)                  281.39       4     70.35      4.22      .019
(Total)                  514.58      18     28.59


These results support the conclusion that performance on a real shopping task benefits from experience of the virtual supermarket. When compared with their own baseline times neither group actually got faster, but on their return to the real supermarket the students who used the virtual supermarket were both faster and more accurate than the control group.

This failure to show any improvement over baseline times for both groups on the second visit occurred because the task was substantially harder this time. The second visit to the real supermarket was in mid-December, when stock had been increased by nearly 50%, from around 5000 products to around 7,500 for the Christmas period. The aisles were much more cluttered, and there were many attractive displays of goods carefully placed to draw the attention of shoppers. Additionally, many of the extra items in the shop were gifts and decorations that were likely to prove especially appealing to the students. So although neither group showed an improvement from baseline, when faced with the more difficult shopping task on the second occasion the experimental group took less time on average than did the control group.

Students in the experimental group also selected significantly more correct items than did the control group on their second visit. This cannot be explained in terms of the students improving their chances of a correct choice by just picking up more items at random, since the experimental group picked up fewer items than did the control group on their return visit to the supermarket.

This improvement over the control group was achieved using a very simple virtual supermarket bearing a limited resemblance to the real supermarket. Shopping in a supermarket involves a variety of competencies: memorising items from the list, visually searching for them, remembering the route taken so far and the locations of items already passed. The task is also made easier if the shopper has some understanding of how goods are categorised, such that finding coffee on the shelves makes it more likely that tea bags will be located nearby.

It is unlikely that the students in the present study improved on all of these component competencies, and future studies will attempt to isolate and evaluate the impact of virtual shopping on each. Doing so should give us more insight into why the students got better, and this knowledge should enable us to design more effective educational virtual environments in the future.

It is also possible that the advantage that the experimental group had over the control group was mediated by a more general increase in familiarity with the task and ensuing confidence gained from experience with a virtual version, or the fact that their sessions with the virtual supermarket were more structured than those on the virtual house or city. Non-specific outcome measures, looking for generalised increases in autonomy and the ability to take decisions, need to be developed in order to explore these possibilities. The development of measures to assess the possible non-specific outcomes of using educational virtual environments is a current priority of our work.

A further unresolved question concerns maintenance of the performance gains that were evident in this study. Before the efficacy of virtual environments as an educational medium is wholly accepted, research into the maintenance of skills acquired there needs to be carried out, to parallel this work on generalisation.

More generally, future research also needs to isolate the features of virtual environments which make them most conducive to learning. Here, we have identified two broad questions to guide future research.

The first concerns the appropriate degree of freedom which students should have within the virtual environment in order to give them the optimal learning experience. On the one hand, one of the strengths of virtual environments is their potential to empower students by giving them immediate and direct control over their own learning experience. On the other hand, some degree of structure or control must be programmed into the virtual environment if the intended lessons are to be learned. Research is needed to explore this continuum between empowerment and control, in order to identify the tasks most suited to relatively structured or unstructured educational virtual environments.

The second question concerns the appropriate level of detail to build into an educational virtual environment. By striving to replicate in a virtual environment the booming, buzzing confusion of the real world we might create an engaging and entertaining experience - but one with little educational value, since any lessons learned would be submerged within the mass of detailed information flowing from the screen. Yet at the same time, if the virtual environment is not sufficiently similar to the real world then skills acquired there will fail to generalise. Research is needed to investigate which aspects of the real world can most usefully be incorporated into virtual environments for students with severe learning disabilities: sounds, textures, the use of live- action video, photo-realistic images, programming autonomous agents into the virtual world are all possible options for investigation.


Virtual environments have great potentials in education in general, and special education in particular. Their ability to bypass or negate many of the obstacles to learning that are inherent in the more traditional teaching methods used with students with severe learning disabilities makes them especially valuable. At the same time, there is already a pressing need to teach basic life skills to students with learning disabilities in an efficient and effective way. For these reasons, further research into the possibilities and limitations of virtual environments in special education is now urgently needed.


Baty, F.J., Michie, A.M. & Lindsay, W.R. (1989) "Teaching mentally handicapped adults how to use a cafeteria" Journal of Mental Deficiency Research 33, 137-148

Blount, W.R. (1968) "Concept usage research with the mentally retarded" Psychological Bulletin 69, 81-94

Bricken, W. (1991) "Virtual reality learning environments: potentials and challenges" Computer Graphics 25, 3.

Bruner, J.S. (1968) "Process of cognitive growth: Infancy" Clark University Press, USA

Day, M.H. & Horner, R.H. (1986) "Response variation and the generalisation of a dressing skill: comparison of single instance and general case instruction" Applied Research in Mental Retardation 7, 189-202

Davies, R.R. & Rogers, E.S. (1985) "Social skills training with persons who are mentally retarded" Mental Retardation 23, 186-196

Donaldson, M. (1978) "Children's Minds" London, Fontana

Fein, G. (1984) "The self-building potential of pretend play, or 'I got a fish, all by myself'" in Yawkey, T.D. & Pellegrini, A.D. (eds) "Child's Play: developmental and applied" Erlbaum, Hillsdale, NJ

Furth, H.G. & Milgram, N.A. (1965) "The influence of language on classification: a theoretical model applied to normal, retarded and deaf children" Genetic Psychology Monographs 72, 317-351

Held, R. & Hein, A.V. (1963) "Movement produced stimulation in the development of visually guided behaviour" Journal of Comparative and Physiological Psychology 56, 872-976

Helsel, S. (1992) "Virtual reality in the classroom" Educational Technology, May 1992, 38-42

Horner, R.H., & McDonald, R.S. (1982) "A comparison of single instance and general case instruction in teaching a generalised vocational skill" Journal of the Association for the Severely Handicapped 7, 7-20

Kozak, J.J., Hancock, P.A., Arthur, E.J. & Chrysler, S.T. (1993) "Transfer of training from virtual reality" Ergonomics 36,7 777-784

Landau, B.L. & Hagan, J.W. (1974) "The effect of verbal cues on concept acquisition and retention in normal and educable mentally retarded children" Child Development 45, 643-650

Pantelidis, V.S. (1993) "Virtual reality in the classroom" Educational Technology, April, 23-27

Robertson, I., Richardson, A.M. & Youngson, S.C. (1984) "Social skills training with mentally handicapped people: a review" British Journal of Clinical Psychology 23, 241-264

Shakespeare, R. (1975) "The Psychology of Handicap" London, Methuen

Sims, D. (1994) "Multimedia camp empowers disabled kids" IEEE Computer Graphics and Applications January 13-14

Siperstein, G.R. (1992) "Social competence: An important construct in mental retardation" American Journal on Mental Retardation 96(4) iii-vi

Steuer, J. (1992) "Defining virtual reality: dimensions determining telepresence" Journal of Communication 42,4 73-93

Stokes, T.F. & Baer, D. (1977) "An implicit technology of generalisation" Journal of Applied Behaviour Analysis 10, 349-367

Theasby, P.J. (1992) "The virtues of virtual reality" GEC Review 7,3 131-145

Travis, D., Watson, T., & Atyeo, M. (1994) "Human psychology in virtual environments" in "Interacting with Virtual Environments" edited by MacDonald, L. & Vince, J. London, John Wiley

Vygotsky, L.S. (1978) "Mind in Society: the Development of Higher Mental Processes" Cole, M., John-Steiner, V., Scribner, S., & Souberman, E. (eds) Cambridge, Mass., Harvard University Press

Ward, J., & Gow, L. (1982) "Programming generalisation: a central problem area in clinical psychology" Educational Psychology 2, 231-248

Wilson, P. (1993) "Nearly there" Special Children 28-30

Wood, D.J. (1988) "How Children Think and Learn" Oxford, Blackwells

Zigler, E. & Balla, D.(1982) "Mental Retardation: the developmental- difference controversy" Erlbaum, Hillsdale, NJ

Go to previous article 
Go to next article 
Return to 1995 VR Table of Contents 
Return to Table of Proceedings 

Reprinted with author(s) permission. Author(s) retain copyright.