Characterizing computer input with Fitts' law parameters: the information and non-information aspects of pointing

  • Authors:
  • Shumin Zhai

  • Affiliations:
  • IBM, Almaden Research Center, 650 Harry Road, NWE-B2, San Jose, CA

  • Venue:
  • International Journal of Human-Computer Studies - Special issue: Fitts law 50 years later: Applications and contributions from human-computer interaction
  • Year:
  • 2004

Quantified Score

Hi-index 0.01

Visualization

Abstract

Throughput (TP), also known as index of performance or bandwidth in Fitts' law tasks, has been a fundamental metric in quantifying input system performance. The operational definition of TP is varied in the literature. In part thanks to the common interpretations of International Standard ISO 9241-9, the "Ergonomic requirements for office work with visual display terminals-- Part 9: Requirements for non-keyboard input devices", the measurements of throughput have increasingly converged onto the average ratio of index of difficulty (ID) and trial completion time (MT), i.e. TP=ID/MT. In lieu of the complete Fitts' law regression results that can only be represented by both slope (b) and intercept (a) (or MT=a+b ID), TP has been used as the sole performance characteristic of input devices, which is problematic. We show that TP defined as ID/ MT is an ill-defined concept that may change its value with the set of ID values used for the same input device and cannot be generalized beyond specific experimental target distances and sizes. The greater the absolute value of a is, the more variable TP (=ID/MT) is. ID/MT only equals a constant 1/b when a = 0. We suggest that future studies should use the complete Fitts' law regression characterized by (a, b) parameters to characterize an input system, a reflects the noninformational aspect and b the informational aspect of input performance. For convenience, 1/b can be named as throughput which, unlike ID/MT, is conceptually a true constant.