Remembering details about the program

In two separate studies [19,18], Rohr investigated how people represent knowledge about the use of an application in their minds, and how this affects their performance in using the application. He ran an experiment in which subjects were asked to group a number of editing commands together according to their apparent relationship to one another. For one group, the commands were presented as icons, and for another the commands were presented as command words. The results showed that those using the icons tended to group commands according to the entity on which they operated (e.g. file operations), whereas those using command words tended to make groups consisting of collections of commands that frequently appear in sequence. Rohr then had the subjects use the command groups to perform editing operations.

Rohr found that the first group achieved a significant performance improvement over the second group. He attributed this to the icons enforcing a spatial representation of the concepts needed to understand the workings of the program. His model of the operation of the mind is one in which there is a small area of working memory, and a large area of long-term memory. When trying to understand a complex system, a user cannot keep the entire structure of the system in working memory at once. Therefore, some mechanism is needed to extract from long-term memory only the areas of the system that are required for the immediate task.

Rohr identified two mechanisms for this purpose. The first is to develop cues for sequential recall--that is, remembering a list of concepts recalled one after the other. The list can easily be recalled, but starting anywhere except at the beginning is difficult. The second mechanism is to integrate all the concepts under consideration into a picture, whose overall structure can be retrieved at will, and which then allows focus to be transferred to a specific element. With this method, it is much easier to recall just the parts of the image that are of most interest (see Figure 2.1). The first mechanism would tend to lend itself more to natural language, and hence to command words, whereas the second would be more naturally suited to icons, since they are already represented pictorially. The first group of subjects had used the pictorial representation, and were able to recall details of the software system more efficiently than the second group.

Figure 2.1: Two methods for storing information

The result of this is that the choice of command words or icons to represent information can determine what kind of mental model the user constructs in order to comprehend an unknown system. Furthermore, since the former leads to sequential recall of concepts, which tends to be less efficient, the use of icons to represent ideas can lead to improved efficiency. My program makes extensive use of iconic representations to exploit this mechanism.