Abstract-An experiment was carried out to investigate whether it is useful to add system information to procedural information in instructional text. It was assumed that readers of instructions construct bothDoes someone who operates a device need to know how that device works? Many tend to answer in the negative: "Just tell me what I should do and don't bother me with the technical details." Researchers of user instructions seem to agree; they advocate an action-directed or procedural approach with some additional information only in small introductory or concluding notes (e.g., [1]). Nevertheless, it seems plausible that at least some understanding of the working of a technical device or a software program is useful for operating it adequately and that this understanding might be enhanced by explanatory information in user instructions. In this article, we make a distinction between two types of relevant knowledge for task performance. PROCEDURAL KNOWLEDGE includes knowledge about goals, actions, and effects of actions, also called a mental plan [2]. SYSTEM KNOWLEDGE includes knowledge about the different parts of the device, their functions, the processes that occur within the device, etc. (for the distinction, see also [3] and [4]). Parallel to these two forms of knowledge, we also distinguish two types of information in instructional documents: PROCEDURAL INFORMATION and SYSTEM INFORMATION.
LITERATURE REVIEWPrevious research is not conclusive to the question of whether system information improves task performance. Smith and Goodman showed that instructions with descriptive information enhanced task performance [5]. Kieras and Bovair found that a schematic description of the internal mechanism of a device led to reduction of errors in task performance [6]. However, their study also suggested that this effect occurs only if the information describes the internal functioning of the product, not when it merely stresses motivational aspects or general principles. Experiments by Dixon et al. showed no effect of system information on how accurately subjects chose the actions needed to complete a task [7]. However, they did show an effect of system information on the recall of actions that needed to be performed.Ummelen investigated the use and effects of "declarative" information (including system information) in an experiment with an unknown and rather complex spreadsheet program [8]. Subjects were offered a manual on screen containing two types of text blocks: procedural (stepwise instructions) and declarative (task-related information about the system). The text blocks were blurred and thus illegible, except for the headings that suggested what kind of information the text block contained. By clicking with the mouse on a block, it became legible for the subjects. This "click and read technique" enabled Ummelen to see when, how often, and how long procedural and declarative information were read.The experiment showed that subjects read procedural text blocks more often and for longer periods of time than declara...