This experiment was designed to investigate the effects of incorporating multiple metaphors in interface design for complex systems. It compared the performance of subjects who were trained to use a complex information system either through a single interface or multiple interfaces.
Based on existing theories, we hypothesized that the trends of development for using single and multiple interfaces in learning to use systems are not parallel. There would be an interaction between experience with a system and the number of interface models used for training on performance. That is, single-interface learning environments for systems would be simpler for users to start with, but would result in less sophisticated mental models. Therefore, learning might occur faster but would stop at a lower performance level in the long run. In contrast, multiple-interface learning environments might be more difficult for the users at the initial learning stage, nevertheless, once the new representation is acquired, the representation would be more powerful. As a consequence, learning to use complex systems through multiple interfaces might progress at a slower rate, but as experience with the systems progresses, users would reach a higher performance level. The hypothesized interaction of experience with a complex system and number of interfaces used on user performance is illustrated in Figure 4.
Figure 4: The hypothesized interaction between experience with a hypertext system and number of interfaces used for learning on user performance.
According to this hypothesis, there is a point at which the two learning curves cross each other. If comparisons of user performance are made before the point is reached, the single-interface users would found to be performing better than the multiple-interface users. On the contrary, if the comparisons take place after the cross-over point is reached, the result would be the opposite.
The results of this experiment did not show any significant differences between the single-interface users and the multiple-interface users. For more details, see Lin (1989).
We hypothesized that the reason for this lack of difference was the short amount of training for the subjects in experiment 2, so that the "cross-over" point was not reached before the end of the experiment. This hypothesis was investigated in Experiment 3.