Abstract New digital video technologies provide a wide spectrum of multimedia interface capabilities for educational courses running on personal computers. A formal experiment was conducted using a digital video course on code inspection to determine the effects of such capabilities on recall performance and attitude. The findings suggest that the presentation of material as motion video rather than as a slide show within an interactive video course leads to better recall performance. In addition, the presence of motion video in the interfaces and the use of surrogate travel for navigation promote better student opinions toward the subject matter.

Animated demonstrations display the execution of interface procedures. They appear to be a natural and fast way for users to learn direct-manipulation interfaces by watching. To assess their effectiveness for users learning HyperCard™, we compared carefully matched animated demonstrations, procedural textual instructions, and demonstrations combined with spoken procedural text. During training, demonstration users were faster and more accurate than text-only users. Without the instructions, 7 days later, text-only users were faster and as accurate as demonstration users in recalling and performing identical and similar tasks without the instructions. Surprisingly, users of the combined demonstrations with spoken text closely mirrored the results of the demonstration-only users. The poor retention and transfer for the demonstration users appeared to be due to mimicry of the demonstrated procedures. Even with accompanying spoken text, the simplicity of using animated demonstrations may encourage superficial processing and disregard for the procedural text.

Over the past 8 years, The TextLab Research Group within the Department of Computer Science at the University of North Carolina has developed a collection of tools and techniques for recording users' interactions with graphics-based direct manipulation computer systems in machine-readable form and for automatically analyzing and displaying those data. This article describes these tools, discusses their methodological context, and considers their implications for software design and studies of human-computer interaction. Tools discussed include the following: tracking users' behaviors and producing a machine-recorded protocol at the level of users' actions, replaying users' sessions from the protocol data, modeling users' strategies using formal cognitive grammars, analyzing user sessions by parsing them with the grammars, and displaying results in visual form-both static and animated--to facilitate interpretation and understanding by researchers. These tools are placed in a methodological context by reviewing issues associated with concurrent think-aloud, keystroke, X-Windows, and video protocols; other support systems for working with these forms of protocol data are also reviewed. The discussion concludes with our reflections on the methodology and its application to computer systems and research objectives different from our own.

We describe marking menus, an extension of pie menus, which are well suited for stylus-based interfaces. Pie menus are circular menus subdivided into sectors, each of which might correspond to a different command. One moves the cursor from the center of the pie into the desired sector. Marking menus are invisible pie menus in which the movement of the cursor during a selection leaves an "ink trail" similar to a pen stroke on paper. The combination of a pie menu and a marking menu supports an efficient transition from novice to expert performance. Novices can "pop-up" a pie menu and make a selection, whereas experts can simply make the corresponding mark without waiting for the menu to appear. This article describes an experiment in which we explored both articulatory and cognitive aspects of marking menus for different numbers of items per menu and using different input devices (mouse, trackball, and stylus). The articulatory aspects are how well subjects could execute the physical actions necessary to select from pie marking menus. Articulatory aspects were investigated by presenting one group of subjects with the task of selecting from fully visible menus. Because one feature of marking menus is that users should be able to select from them without seeing the menus (by making a mark), we also ran two groups of subjects with invisible pie menus: one group with an ink trail and one without. These subjects were therefore faced with the task of either mentally representing the menu or associating marks with the commands they invoked through practice. These then are the cognitive aspects to which we refer. Our results indicate that subjects' performance degraded as the number of items increased. When menus were hidden, however, subjects performance did not degrade as rapidly when menus contained even numbers of items. We also found subjects performed better with the mouse and stylus than with the trackball.

The development of schemes to support group work, whether behavioral methods or new technologies like groupware, should be based on detailed knowledge about how groups work, what they do well, and what they have trouble with. Such data can be used to suggest what kinds of tools people might need as well as to provide a baseline for evaluating the effects of schemes for improvement. We present details of how real groups engage in a representative collaborative task - early software design meetings - to provide such knowledge. We studied 10 design meetings from four projects in two organizations. The meetings were videotaped, transcribed, and then analyzed using a coding scheme that looked at participants' problem solving and the activities they used to coordinate and manage themselves. We also analyzed the structure of their design arguments. We found, to our surprise, that although the meetings differed in how many issues were covered they were strikingly similar in both how people spent their time and in the sequential organization of that activity. Overall, only 40% of the time was spent in direct discussions of design, with many swift transitions between alternative ideas and their evaluation. The groups spent another 30% taking stock of their progress through walkthroughs and summaries. Pure coordination activities consumed about 20 %, and clarification of ideas-a crosscutting classification-took one third of the time, indicating how much time was spent in both orchestrating and sharing expertise among group members. The pattern of transitions revealed these activities were clustered into two general classes - design and management. Although most issues had more than one alternative offered and discussed, there was rarely a wide set discussed, and one third of them were never explicitly evaluated. The results have implications for both the characterization of collaboration itself and for the way in which it might be supported through technology. Finally, the coding schemes developed may be useful for a wide range of problem-solving meetings other than design.

This article describes a three-part empirical approach to understanding the development of expertise within the UNIX1 operating system. We studied UNIX users with varying levels of expertise. The first part of our research attempted to ascertain the nature of their conceptualizations of the UNIX system. The second part measured users' performance in tasks requiring them to comprehend and produce UNIX commands. The third part was a longitudinal rather than cross-sectional analysis of the emergence of expertise. The conceptualization data suggest important differences in the models of UNIX structure formed by each group. Experts best represent the higher levels of the UNIX system; novices more fully represent the lower, more concrete levels of the system, including specific commands. UNIX users also differ markedly in performance, according to their history of use with the operating system. Only experts could successfully produce composite commands that required use of the distinctive features of UNIX (e.g., pipes and other redirection symbols), even though the intermediates and novices evidenced the component knowledge required for composite commands. This finding is somewhat surprising, inasmuch as these are fundamental design features of UNIX, and these features are taught in elementary classes. These data suggest, however, that these features can be used reliably only after extensive experience. The longitudinal data suggest that most subjects increased in expertise. However, expertise can decrease as a function of time, depending on system use. Those subjects who increased in expertise acquired the ability to produce the simple commands and represent the basic modules before they acquired knowledge of complex commands and advanced utilities. The nature of expertise is considered with respect to both system design and user characteristics, including users' conceptual models of system structure.

Abstract This article reports two studies examining how the use of computer technology affects writers' processes and resulting written products. The first experiment, a partial replication of Gould's (1981) study, compared the efficiency and quality of 15 experienced writers' persuasive letters written in three counterbalanced conditions: pen and paper, standard personal computer, and advanced workstation. Analyses of time to compose and length of the resulting letters showed that when composing with the advanced work-stations, writers wrote for longer periods of time and composed longer letters than when composing with pen and paper. Rate of composition (words per min) was similar in all three conditions. Scores for content quality, mechanics quality, and total quality were also collected and analyzed. Letters composed with the personal computer were poorer in content quality and total quality. There were no differences in mechanics quality. In the second experiment, 8 of the original 15 writers revised two of their letters in the medium with which they were composed. Think-aloud protocols collected while the writers revised their letters were collected and analyzed. When revising with pen and paper, writers planned more than when revising on-line; they also planned more before beginning their revisions. However, when revising in the computer conditions, writers reread their texts more and paid more attention to the medium than when revising on paper.

Should the structure of external databases correspond to the structure of long term memory? This question is relevant to human factor specialists engaged in the design of user friendly information retrieval systems. The question presupposes agreement on the assumption that external cognitive acts are performed in the same way as analogous internal mental acts. This assumption is not new and might be considered as the opposite of the internalization hypothesis set forth by, for example, Vygotsky (1978) and Piaget (1953). The internalization hypothesis holds that intelligence as such is the product of perceptual-motor activities of the young child. Manipulations of concrete objects are internalized and integrated into the individual’s mental equipment to deal with the environment.

The aim of this article is to find the optimal combination of written instruction and on-line practice for learning a new computer application. Subjects in the experiment learned commands for an electronic spreadsheet by reading brief user-manual descriptions and working training problems on-line. The form of the training problems was varied in a within-subjects design to control how much independent problem solving subjects engaged in while learning any given command. There were three forms of practice: (a) pure guided practice, in which subjects were told exactly what keystrokes to type to solve the problems; (b) pure problem-solving practice, in which subjects solved problems without guidance; and (c) mixed practice, in which the first problem for a command was presented in guided practice form and two others in problem-solving form. The spacing of the training problems was also manipulated; the problems pertaining to a given command were either massed (i.e., presented consecutively) or distributed (i.e., separated by other instructional material). After a 2-day delay, subjects solved new problems on the computer without referring to the instructional materials. The results indicate that problem solving was a more difficult form of training than guided practice, but it produced the best performance at test. Distributing the spacing of training problems during training also improved performance at test. The results have clear pragmatic implications for the design of interactive tutorial manuals as well as implications for cognitive models of skill acquisition.