Virtual environments enable a given information space to be traversed in di
fferent ways by different individuals, using different routes and navigatio
n tools. However, we urgently need robust user models to enable us to optim
ize the deployment of such facilities. Research into individual differences
suggests that the notion of cognitive style may be useful in this process.
Many such styles have been identified. However, it is argued that Pask's w
ork on holist and serialist strategies and associated styles of information
processing are particularly promising in terms of the development of adapt
ive information systems. These constructs are reviewed, and their potential
utility in "real-world" situations assessed. Suggestions are made for ways
in which they could be used in the development of virtual environments cap
able of optimizing the stylistic strengths and complementing the weaknesses
of individual users. The role of neural networks in handling the essential
ly fuzzy nature of user models is discussed. Neural networks may be useful
in dynamically mapping users' navigational behavior onto user models to ena
ble them to generate appropriate adaptive responses. However, their learnin
g capacity may also be particularly useful in the process of improving syst
em performance and in the cumulative development of more robust user models
.