Fitts' law as a performance model in human-computer interaction
暂无分享,去创建一个
The theoretical basis for Fitts' law is developed and extended. Several modifications are claimed to improve the model's prediction power in general, and to extend its applicability to movement tasks common on today's interactive computer systems. Original contributions include (a) a correction to the model based on a direct analogy with Shannon's Theorem 17, (b) two methods for accommodating approach angle in the calculation of task difficulty, and (c) an extension of the model to dragging tasks and compound pointing-plus-dragging tasks.
In a re-analysis of published data from experiments by Fitts' and others, the Shannon formulation is shown to provide higher correlations than the Fitts' or Welford formulation, particularly when task conditions extend under 3 bits. It is demonstrated that the Shannon formulation cannot yield a negative rating for task difficulty, unlike the Fitts or Welford formulation.
Three experiments were conducted. In the first, the mouse, tablet-with-stylus, and trackball were tested on serial pointing and dragging tasks. Model comparisons provided moderate additional evidence of an improved fit using the Shannon formulation. Using the index of performance (in bits/s) as the criterion, the tablet-with-stylus and mouse performed similarly during the pointing tasks; but the tablet-with-stylus was slightly better than the mouse during dragging. The trackball was a poor performer for both tasks.
In the second experiment, the mouse was used in a two-dimensional point-select task with rectangular targets. Two models performed significantly better than the status quo model--one substituting for target width the smaller of the target's width or height (the "smaller-of" model), the other substituting the span of the target along an approach vector through the target's centre (the W$\sp\prime$ model).
In the third experiment, the mouse was tested in a point-drag-select task. Observations support the claim that Fitts' law can model both pointing and dragging operations, even when they occur as a point-select (button-down) action followed by a drag-select (button-up) action.