For efficient and effective pen-based sketching of user interface designs, what widget representations are best? To answer this question, the authors of this paper undertook two experiments. The first identified users’ preferences for representations. The second involved actual pen-based sketching of widgets using a sketching tool capable of generating runnable user interfaces. The hypothesis that user performance depends on the complexity of widget representation is strongly supported by the results of the second experiment.

Table 5 clearly shows how the widget recognition rate was found to worsen as the number of atomic components and constraints of a widget representation increased. Timing data below this table reveal that sketching complex widgets required an average of over seven seconds, and an average of less than two seconds for simple widgets.

Table 6 identifies the number of atomic components, the sequence constraint, and the complex inclusion constraint as the main governing factors for sketching time and the number of delete operations. The authors conclude that it is important to minimize the number of constraints to be satisfied if one is to sketch a widget efficiently and effectively.

Those who are familiar with goals, operators, methods, and selection (GOMS) keystroke-level modeling will find the results obvious, but they will agree that it is useful to have the empirical foundation and the treatment of complexity characterization that this study provides.

Criticisms of this paper include the failure to discuss widget palettes with drag-and-drop functionality, and the failure to discuss pattern-based user interface design. Despite these criticisms, I strongly recommend this paper to human-computer interaction specialists.