Perceptual-motor control in human-computer interaction.

Abstract

Three experiments compared the speed and accuracy of item selection from two types of hierarchical pop-up menus (click-open & walking). These menus are perceptually equivalent, but require very different action sequences to make selections from them. Two ways in which the action sequences differ involve (1) the amount of spatial constraint in the movement path, and (2) the position of the mouse button during the movement. The present results show that selection from click-open menus is faster and less error-prone than selection from walking menus. This disagrees with the Keystroke-Level Model of Card, Moran, and Newell (1983). The difference in selection times for the two menu styles stems from the position of the mouse button during movement. In particular, moving with the mouse button depressed (walking menus) is slower than moving with the mouse button released (click-open menus). It appears that this difference is caused by the postural awkwardness of moving with the button depressed throughout the primary phase of movement. The difference is not caused by (1) greater working-memory load, (2) greater friction between the mouse and the mouse pad, or (3) greater difficulty in coordinating the end of the movement phase with the terminal button release in walking menus. With numerically-ordered menus, selection time is dominated by motor factors and is well modeled by Fitts' law for both click-open and walking menus. With randomly organized menus, selection time is dominated by a systematic, top-to-bottom search process. Visual search and mouse movement are parallel rather than serial processes. Present results indicate the need for a process model of human-computer interaction that accommodates both serial and parallel processes. A start is made toward specifying such a model through the critical-path scheduling used by Schweikert (1978) and John (1988). Implications for menu selection and other computer-based tasks are also discussed.