Berry & Broadbent (1984, 1987, 1988) and others have argued that in dynamic tasks subjects can learn to control computer systems without knowing how the systems work. Two sets of computer simulations are examined to test the hypothesis that subjects do not learn how the system works because the signal to noise ratio in the data stream is too low. The first set of simulations show that the goal of obtaining a particular target position is in direct conflict with the goal of learning the rule. The second set of simulations demonstrates that, if subjects perform these tasks by remembering previous instances or by using a simple strategy based on abstracting over previous instances, they are even less likely to learn how the system works because the signal to noise ratio decreases even more. Finally, it is possible to explain, by examining the signal to noise ratio and how it can be varied, when subjects do learn the rule that controls the system's behaviour.
To receive an off print, please email Irene Jackson.