Systematic errors In performance are an important aspect of human behavior that have not received adequate explanation. One such systematic error is termed postcompletion error; a typical example is leaving one's card In the automatic teller after withdrawing cash. This type of error seems to occur when people have an extra step to perform in a procedure after the main goal has been satisfied. The fact that people frequently make this type of error, but do not make this error every time, may best be explained by considering the working memory load at the time the step is to be performed: The error is made when the load on working memory is high, but will not be made when the load is low. A model of performance In the task was constructed using Just and Carpenter's (1992) CAPS that predicted that high working memory load should be associated with postcompletion errors. Two experiments confirmed that such errors can be produced in a laboratory as well as a naturalistic setting, and that the conditions under which the CAPS model makes the error are consistent with the conditions under which the errors occur in the laboratory.

Kieras and Polson (1985) proposed an approach for making quantitative predictions on ease of learning and ease of use of a system, based on a production system version of the goals, operators, methods, and selection rules (GOMS) model of Card, Moran, and Newell (1983). This article describes the principles for constructing such models and obtaining predictions of learning and execution time. A production rule model for a simulated text editor is described in detail and is compared to experimental data on learning and performance. The model accounted well for both learning and execution time and for the details of the increase in speed with practice. The relationship between the performance model and the Keystroke-Level Model of Card et al. (1983) is discussed. The results provide strong support for the original proposal that production rule models can make quantitative predictions for both ease of learning and ease of use.

: The processes of acquiring procedures from text are important to understand for both practical and theoretical reasons. This paper outlines a theory of procedure acquisition that is based on empirical and theoretical work on the value of production-system representations of procedural knowledge. The key process in acquiring procedures from text is thus constructing an adequate set of production rules from the textual input. The existing empirical literature is interpreted and criticized within this framework. Two general conclusions are that the amount of research on text that is intended to convey procedures is much less than the topic deserves, and future research needs to more precisely distinguish the different processes involved in acquiring procedures from text. Keywords: Text processing, Instructions, Procedure acquisition.

This paper describes a successful test of a quantitative model that accounts for large positive transfer effects between similar screen editors, between different line editors and from line editors to a screen editor, and between text and graphic editors. The model is tested in an experiment using two very similar full-screen text-editors differing only in the structure of their editing commands, verb-noun vs noun-verb. Quantitative predictions for training time were derived from a production system model based on the Polson and Kieras (1985) model of text editing.

Learning a cognitive skill from written instructions can be viewed as consisting of converting the propositional content of the written material into a representation of procedural knowledge, such as production rules. In a transfer of training experiment, subjects learned from step-by-step instructions a series of related procedures, in different training orders, for operating a simple device. The strong between-procedure transfer effects were predicted by a simple model of transfer in which individual production rules can be transferred or re-used in the representation of a new procedure if they had been used in a previously learned procedure. Apparently, this transfer mechanism acts on declarative propositional representations of the production rules, suggesting that it is more similar to comprehension processes than to conventional practice mechanisms, or to Anderson's learning principles (1982, Psychological Review, 89, 369–406; 1983, The architecture of cognition, Cambridge, MA, Harvard Univ. Press).

: Three questions were addressed in an experiment in which subjects followed instructions to complete tasks involving several pieces of electronic equipment: (1) Two instruction formats were compared: a historical menu format containing natural chunks of instructions was not superior overall to a simple step-by-step instruction format. The menu format was superior only if the subject was familiar with the type of device, and was sometimes substantially inferior otherwise. (2) Experts were compared to nonexperts, and found to be faster overall, and able to operate equipment with fewer instructions in the menu condition. They were also faster when complex physical actions were involved. Thus, there were both specific and general effects of expertise. (3) Evidence was sought that knowledge of how to operate equipment was schematic. It was expected that when subjects in the menu format condition operated a device without selecting any instructions to read, their sequence of actions should correspond to stereotyped schema-like patterns. This occurred only weakly, suggesting that even experts operate everyday devices in a problem-solving mode, rather than by retrieved complete procedures.

This report presents three studies concerned with learning how to operate a simple control panel device, and how this learning is affected by understanding a device model that describes the internal mechanism of the device. The first experiment compared two groups, one of which learned a set of operating procedures for the device by rote, and the other learned the device model before receiving the identical procedure training. The model group learned the procedures faster, retained them more accurately, executed them faster, and simplified inefficient procedures far more often, than the rote group. The second study demonstrated that the model group is able to infer the procedures much more easily than the rote group, which would lead to more rapid learning and better recall performance. The third study showed that the important content of the device model was the specific configuration of components and controls, and not the motivational aspects, component descriptions, or general principles. This specific information is what is logically required to infer the procedures. Thus, the benefits of having a device model depend on whether it supports direct and simple inference of the exact steps required to operate the device.