Institute for Computer-Based Learning

Heriot-Watt University

Edinburgh EH14 4AS

UK

email: terry@uk.ac.hw.icbl

ABSTRACT

Hypermedia and multimedia have been placed rather uncritically at the centre of current developments in learning technology. This paper seeks to ask some fundamental questions about how learning is best supported by hypermedia, and concludes that the most successful aspects are not those normally emphasised. A striking observation is that the best learning experience is enjoyed by hypermedia courseware authors rather than students. This is understandable from a constructivist view of learning, in which the key aim is to engage the learner in carrying out a task which leads to better comprehension. Deep learning is a by-product of comprehension. The paper discusses some approaches to designing software - cognitive tools for learning - which illustrate this constructivist approach.

Introduction

The overall purpose of this paper is to discuss the way in which computers are currently being used to support learning. In particular, some of the assumptions underlying the development of hypermedia for learning will be examined, and a rather different perspective offered - that which has become known as constructivism. This perspective will be justified by referring to some of my previous and current work in the area of learning from computers, and by presenting a view of learning grounded in cognitive psychology.

Answers will be sought to the following questions:

* What are hypermedia and multimedia good for?

* Why do the authors of hypermedia courseware learn more than the students who use it?

* What are the right conditions for effective learning?

* What are cognitive tools?

* Where should we direct our development effort in the future?

Hypermedia for learning: the background

Hypertext represented a very significant step forward in the development of the use of computers in education. Traditional CAL, with frames programmed in some awkward authoring language, was still seen as labouring under the dead-hand of programmed instruction. The teaching machine era of the sixties and early seventies was widely seen as having failed, although the principles of programmed instruction took some of the blame that should have been directed at the crude attempts to apply it to education. Meanwhile, in the psychology of learning, the emphasis moved decisively away from a study of the acquisition of knowledge or skill towards the nature of the competence and expertise that learning produces (Glaser, 1990). The study of competence in complex performance has produced significant advances in our understanding of the organisation of memory, in the nature of information processing for problem soving, and in the qualitative and quantitative changes that result from extended practice and in the development of expertise. Much of this was then reflected in the attempts to derive knowledge based approaches to tutoring. Throughout this period, though, a kind of Piagetian sub-plot inspired much educational theory and continued to place learning processes at the top of the agenda. This influence can be seen in the LOGO movement, and more generally in the focus on problem solving as the main paradigm for the development of CAL (see, e.g. O'Shea & Self, 1982 ).

Hypertext and its derivatives in multimedia seem to owe little to any of this background. Hypertext was not initially seen by many of its proponents as a particularly suitable vehicle for learning. It soon became apparent, however, that the power of information access provided by such systems offered opportunities for the design of reactive learning environments. Making a virtue out of the lack of student modelling provided in such systems, many saw exploratory or discovery learning not only as a welcome relief from the apparent difficulties of designing systems capable of genuine dialogue with learners, but more importantly as the paradigm offering most promise for active learning. In hypermedia, we now have systems offering complete learner control, with a high degree of interactivity through direct manipulation interfaces to large databases of multimedia teaching materials. It is worth noting in passing that it has sometimes been argued that the hypermedia paradigm is particularly suitable for learning because it somehow reflects the apparently associative nature of human thought. This argument, like some others promoted in support of learning from hypermedia, is probably spurious.

What is the case for designing hypermedia systems for learning? Essentially, there are two kinds of reasons: those which depend on the "hyper" features, and those which are due to the use of multimedia.

THE CASE FOR THE "HYPER" IDEA

The case for the "hyper" idea can be presented as follows:

* The hypertext idea itself, as Conklin (1987) has pointed out, conveyed the important notion that computers could help to augment human cognition. Hypertext uses the computer in the role of a tool for supporting various kinds of cognitive actions. Thus, at a general level, it helps to shift the focus away from the idea that a computer, because it seems to consist mainly of a screen, is simply a device for presenting information.

* The idea also conveys the concept of 'just-in-time' information access. It is the learner who is in the best position to judge what information he or she needs next. Hypermedia gives the learner control to access information that is most relevant to the particular learning need of the moment. It therefore contrasts with most previous forms of CAL, which have presented information in an "expository" mode.

* Hypermedia also provides a high degree of interactivity. Interactive learning is widely assumed to be effective because active learning produces more effective learning outcomes.

* Developing the idea of learner-control, it was soon appreciated that computers could also provide environments in which discovery learning could occur. Learning-by-browsing emphasised this paradigm, and browsing was seen as the natural mode of navigating through hypermedia.

Perhaps the system that best exemplifies CAL based on the provision of specific guidance tools built on top of a hypertext network is HitchHikers' Guide (Hammond & Allinson, 1988). The principle espoused here is that of extending and tailoring basic hypertext facilities, not only with the fairly standard aids to access, such as browsers, but also with tools that help the user to explore the material conceptually, such as guided tours, indexes, and quizzes. Allinson & Hammond (1989) have referred to such a system as a Learning Support Environment (LSE).

There are problems with each of the above arguments which raise questions about the validity of the case for hypermedia. First, the fact that hypertext provides a cognitive tool does not necessarily imply that it will be effective in support of the process of learning. Secondly, as Hammond (1992) has pointed out, the learner is not always going to choose what information to see next in a way that will lead to effective learning. Unguided choice may be as inefficient as no choice. Thirdly, there are problems with the simple idea that interactivity is a necessary attribute of effective learning from computers. Just what it is about interactivity that succeeds in promoting better learning has rarely been questioned. Indeed, some learning software is described as "interactive" simply because the learner has to press the space bar to proceed to the next screen of information. It is evident that a more critical analysis of interactivity is needed before we can be confident of identifying the essential feature(s) which lead to effective learning. Some authors have referred to "engagement" as the necessary factor. However, that shifts the locus of effect onto the cognitive level, and does not necessarily help us to design environments in which engagement is likely to occur. Finally, questions can be raised about discovery learning through browsing. Unless the browsing can be motivated by seeking answers to questions, or by some kind of problem solving, then it may only support a shallow learning experience. An important research question could be built around the observation of enjoyment that readers of magazines experience in browsing. A very major publishing industry is based on this almost universal feature of human reading. Readers seem to enjoy the experience of browsing in a domain in which they are already knowledgable. It is not clear that browsing in an unfamiliar domain occurs at all as a primary learning strategy. Anecdotal evidence would suggest that browsing becomes enjoyable only when a certain level of familiarity, or to put it in cognitive terms, when schemata have become sufficiently well developed for browsing to represent a process of "tuning" (Norman, 1982 ). An important, and still unanswered, research question concerns this aspect of what learning "feels like'.

There are also the disadvantages of the "hyper" idea described by Conklin (1987). That is, the feeling of disorientation engendered by trying to navigate in a non-linear information space; and the cognitive overhead of being required continually to make choices. Finally, a serious drawback to hypermedia is the complexity involved in its authoring.

THE CASE FOR THE MULTIMEDIA IDEA

Turning now to the characteristic that has led, above all, to the enthusiasm for using computers in support of learning we must consider the case for multimedia. Taking advantage of the graphic and sound capabilities of modern desktop machines, using processing power to generate simulated microworlds, and integrating some of the potential of digital video, now allows the creation and delivery of vivid interactive courseware. One well-known example of this is Palenque (Wilson, 1988). Palenque is a discovery learning system in which the learner is able to explore all the paths of an ancient Mayan site. As the user travels through the site a multimedia database in the form of a museum provides moving video, stills, audio and text about the rainforest, the Mayans, maps of the area, and glyph writing. What characterises Palenque is the variety of methods and media the user is offered for accessing the knowledge. The components are: video overviews; surrogate or virtual navigation; a multimedia database; characters as experts and guides; simulated tools; and games. The interface in Palenque employs visual menus and dynamic icons, spatial and thematic navigation, and a simple input device. The virtual travel around the Palenque site can be regarded as a main menu for an exploratory mode in which various options and subprograms are distributed spatially at meaningful locations. Icons represent such options as branch points in travel, available pans and information zooms, and narrations.

It seems self-evident that multimedia will support more effective learning, but as with our examination of the "hyper" idea, a closer consideration of the arguments and evidence begins to raise some doubts. The combination of media in displays, particularly the high resolution, colour, full-motion video with speech variety carries with it a vividness that cannot be questioned. However, Taylor & Thomson (1982) attempted to pin this down in a comprehensive review of work on the "vividness" effect. Their conclusions were surprising:

"Everyone knows that vividly presented information is impactful and persuasive.....There is one problem with this self-evident proposition. The available evidence suggests that it is not true".

The research failed to show that concrete descriptions have any greater impact than dull ones; that pictorially illustrated information is more effective than that which is not illustrated; or that videotaped information has more impact than oral or written. One can react to this negative conclusion by rejecting the validity of the research. The authors, however, make the point that vividness can never be simply a function of the presentation. The impact will always depend on an interaction with user characteristics. There is no evidence, for example, that a more "vivid" experience is a more memorable one, if by "vivid" we mean some combination of characteristics of the presentation. Vividness is entirely "in the eye of the beholder".

The fundamental point here is that as learners we are not easily enticed by surface aspects of information, and the attempt to use computers to somehow make the learning experience more attractive or more palatable is doomed to failure. Information that is poured into the learners head through the "Nurnberg Funnel" (Carroll, 1990) is only likely to be better learned as a consequence of being presented through multimedia if it is thereby better understood. Hypermedia or multimedia will therefore be successful to the extent that they promote better understanding.

The StrathTutor experience

From 1986, my colleagues and I worked with a hypermedia system that we called StrathTutor. This has been fully described elsewhere ( Mayes et al, 1988) but a short description will summarise its main features.

StrathTutor consists of frames of text and graphics on some topic which a learner explores. In StrathTutorlinks between frames are computed on the basis of attribute coding, from a set of up to 60 attributes predefined by the author for the particular domain. Each designated 'hotspot' of text and/or graphics is so coded. The system computes the 'relatedness' of all remaining unseen frames to the current frame ('frame' is arbitrarily set at the size of a single screen) or hotspot. Each frame can be represented as a profile of attributes, summed across all hotspots in that frame. The learner can choose to navigate by accepting the 'related' frames offered by the system, or can proceed to access named frames. Details of the way in which StrathTutor achieves this computation are given in Kibby & Mayes (1989). There is a traditional hypertext feature whereby some hotspots are explicitly linked to windows presenting explanatory material. Nevertheless, a much more important feature of this system is the opportunity it offers learners to try out hypotheses about the meaning of attributes and the relationships between them. A learner can 'interrogate' the system by designating a combination of attributes that may be beginning to seem meaningful and the system will respond by giving the learner a 'guided tour' of all frames that are coded with that particular subset of attributes.

The main instructional approach embodied in StrathTutor is one of learning by challenge. Despite its conventional frame-based appearance it can actually be seen as a problem generating system. The StrathTutor 'quiz' invites the learner to play a kind of game, in which he or she tries to identify the areas across the two frames which have maximum overlap in attributes. Here the learners are pitting themselves against the author who created the attribute tags on each hotspot. In each case the learners are expected to create for themselves a view of the underlying conceptual space.

Thus, as originally conceived, StrathTutor provided the following features:

* learning-by-browsing

* runtime links

* exploration through a variety of browsing techniques

* easy authoring

* games features for motivation

STRATHTUTOR AS UNDERSTOOD IN 1993

Now, our perspective on StrathTutor is rather different. The main shift in emphasis has occurred as a result of observing StrathTutor being used, in a variety of different contexts. This has led us to the view that the users of StrathTutor who benefit most, those who experience the deep learning experience, are not the students but the authors. Time and again this observation has been emphasised. Finally we accepted the obvious point. The requirement for an author to analyse the subject matter at the level of attributes provided a task that it was not possible to achieve satisfactorily without reaching a deep understanding of the material. We had created an effective learning tool, but it was most effectively used not by trying to double-guess the coding that someone else had put on the material, but by doing so directly. Simply by changing roles - by placing the learner in the role of author - we realised that we had created a cognitive tool for learning.

Cognitive Tools for Learning

A cognitive tool for learning is simply a device, or technique, for focusing the learner's analytical processes. In the context used here, a cognitive tool provides computer support for a task, the explicit purpose of which is to lead to active and durable learning of the information manipulated or organised in some way by the task. The primary task is not learning per se. Rather, learning is an inescapable by-product of comprehension. The point of the cognitive tool is to make it possible for the learner to acquire a deeper understanding of the material. Then learning takes care of itself.

Mayes (1992) gives an account of cognitive tools derived from cognitive psychology. The experimental underpinning of this is provided by the work on levels of processing (Craik & Lockhart, 1975) and the enactment effect (Cohen, 1981). But it is a long-established theme in the study of human cognition that learning flows from understanding, and understanding flows from action and problem-solving (Bartlett, 1932).

A recent book by Kommers, Jonassen & Mayes (1992) draws together many examples of cognitive tools for learning. The techniques range from requiring learners to engage in concept mapping based on constructing semantic nets (eg SemNet) through to simple attempts to represent the domain as a series of rules by asking students to use a simple expert-system shell ( Trollip et al, 1992). In each of these examples the students can be placed in some sense in the role of teacher, by being required to structure their developing understanding for someone else to view. This capitalises on the oft-quoted observation that the best way to learn something is to teach it. It strongly suggests that the most promising way in which computers can aid learning is for us to stop thinking of authoring as a task for subject matter experts and to start putting authoring tools into the hands of students. It is consistent with a view of learning that can be characterised as constructivist (Duffy et al, 1993). Put simply, this approach rests on the basic assumption that understanding has to be constructed by the learner. There is a strong tradition of constructivism in higher education. That is the basis for the student essay, or laboratory report. For a while computers-as-presentation-devices have distracted us from the constructivist approach. It is time to return to basics.

Learning from computers: a constructivist manifesto

In summary, the following points emphasise the constructivist argument underlying this paper:

* Hypermedia/multimedia learning systems will be effective in so far as they support the learner in the performance of knowledge construction tasks. The influence of the features usually emphasised - learner-control, interactivity, browsing, vividness of presentation - will be secondary, and probably will have only marginal benefit for learning.

* "Interactivity" by itself is not enough. The interaction must be at the level of meaning, whereby the learner seeks answers to new questions, arranges the material into new structures, or performs other manipulations which succeed in raising the level of comprehension. Deep learning will then follow naturally.

* We already have many computer-based tools which can serve to support learners in constructing knowledge. Some of these are authoring tools. Since our observations lead us to conclude that authors do the best learning, the obvious step is to shift our perspective and regard them as cognitive tools for learning.

* There are many other possible computer-based cognitive tools for learning. A kind that has not been discussed here is based on computer-mediated communication. Computers now provide powerful opportunities for learners to support each other. This is consistent with the constructivist approach of student-as-teacher. Learner-learner communication has become an important theme in our work on the ISLE (Intensely-Supportive Learning Environment) Project at Heriot-Watt.

* A key to the design of cognitive tools for learning lies in our understanding why some learning tasks are pleasurable, while others are aversive. This remains a fundamental challenge for future research.

REFERENCES

Allinson,L.J. & Hammond,N.V. (1989). A learning support environment: The hitch-hiker's guide. In R.McAleese (Ed.) Hypertext: theory into practice. Ablex: Norwood, NJ.

Bartlett, F.C. (1932) Remembering. Cambridge: Cambridge University Press.

Carroll, J.M. (1990) The Nurnberg Funnel: Designing minimalist instruction for practical computer skill. Cambridge, MA: MIT Press

Cohen, R.L. (1981) On the generality of some memory laws. Scandinavian Journal of Psychology, 22, 267-281

Conklin, J. (1987) Hypertext: an introduction and survey. IEEE Computer, 20, 9, 17-41.

Craik, F.I.M. & Lockhart, R.S. (1975) Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behaviour, 11, 671-684.

Duffy, T. Jonassen, D. & Lowyck, J. (Eds) (1993) Designing constructivist learning environments. Heidelberg, FRG: Springer-Verlag

Glaser, R. (1990) The reemergence of learning theory within instructional research. American Psychologist, 45, 1, 29-39

Hammond, N.V. (1992) Tailoring hypertext for the learner. In P. Kommers, D. Jonassen & J.T Mayes (Eds) Cognitive Tools for Learning, Heidelberg, FRG: Springer-Verlag.

Hammond, N.V. and Allinson, L.J. (1988) Development and evaluation of a CAL system for non-formal domains: the hitch-hiker's guide to cognition. Computers and Education, 12, 215-220.

Kibby M.R. and Mayes J.T. (1989) Towards intelligent hypertext. In R. McAleese ed., Hypertext: theory into practice. Ablex, Norwood, New Jersey.

Kommers, P., Jonassen, D. & Mayes J.T. (Eds) (1992) Cognitive Tools for Learning, Heidelberg, FRG: Springer-Verlag.

Mayes J.T., Kibby M.R. and Watson H. (1988) StrathTutor: the development and evaluation of a learning-by-browsing system on the Macintosh. Computers and Education, 12, 221-229.

Mayes, J.T. (1992) Cognitive Tools: a suitable case for learning. In P. Kommers, D. Jonassen & J.T Mayes (Eds) Cognitive Tools for Learning, Heidelberg, FRG: Springer-Verlag.

Norman, D.A. (1982) Learning and memory. San Fransisco: Freeman.

O'Shea, T. & Self, J.A. (1982) Learning and teaching with computers. Englewood Cliffs, NJ: Prentice-Hall

Taylor, S.E. & Thomson,S.C. (1982) Stalking the elusive `vividness' effect. Psychological Review, 89, 2, 155-181.

Trollip, S.H., Renate, C.L., Starfield, A.M. & Smith, K.A. (1992) Building Knowledge Bases: An environment for making cognitive connections. In P. Kommers, D. Jonassen & J.T Mayes (Eds) Cognitive Tools for Learning, Heidelberg, FRG: Springer-Verlag.

Wilson, K.S. (1988) Palenque: an interactive multimedia digital video interactive prototype for children. In Proceedings of CHI'88. ACM, NY., pp 275-279.