1. INTRODUCTION
According to Fitts’ law (Fitts, 1954), the minimum time T needed to reach a target of width W
located at a distance D can be predicted with T = a + b * ID, where ID, the index of difficulty,
is defined as log2 (D/W +1). This simple relation, one of the few well-established quantitative
laws of experimental Psychology (Kelso, 1992), has been shown to hold in an impressive
variety of tasks and situations (Plamondon and Halimi, 1997).
Fitts’ law has been quite useful to the field of Human-Computer Interaction (HCI) for about a
quarter century. As soon as the first graphical user interfaces (GUIs) became available, HCI
researchers acknowledged that Fitts’ pointing paradigm could serve to evaluate the design of
interaction techniques and input devices in a rigorous and principled fashion (Card, English
and Burr, 1978). This pioneering study of Card and colleagues paved the way for the
development within the HCI community of an active research stream on Fitts’ law
(MacKenzie, 1992; Soukoreff & MacKenzie, this issue).
The target acquisition problem that HCI research has been handling since the appearance of
graphical interfaces does not differ in essence from Fitts’ original problem. Indirect pointing,
where a screen cursor stands for the hand and a graphical object stands for the target, was
already customary in laboratories of applied and basic Psychology (e.g., Crossman, 1960). In
the nineteen nineties, however, a radical change occurred in HCI: researchers introduced
multiscale or zoomable interfaces (Perlin & Fox, 1993, Furnas & Bederson, 1995), allowing
multiscale navigation.1 Since then, this innovation has been gradually generalized to many of
the applications we use today. This change, we believe, involves a real challenge to Fitts’ law.
With a zoomable user interface, users can move a cursor to some target object (the standard
pointing problem) or move their view to some target view in an electronic world (a minor
variant of the standard pointing problem, as we will see below), but they can now freely
adjust the scale at which they wish to interact with the electronic world. This is an entirely
novel category of user action, one that has no counterpart in the real world and could hardly
be anticipated in the context of experimental Psychology.
In this article we demonstrate that Fitts’ pointing paradigm still applies to multiscale
interfaces. This requires enriching Fitts’ classic paradigm by introducing a scale variable and
defining the notion of multiscale pointing. A theoretical analysis leads us to a new
formulation of Fitts’ law for multiscale pointing, which we test with the results of two
experiments. This new formulation, which involves the size of the view, has important
implications for the design of user interfaces. We feel that multiscale pointing, a reality from
HCI, renews in interesting ways the traditional psychological theme of target-directed
movement.
1 In this article we use the generic term “navigation” to refer to any user action that updates the visualization of a
document, whether the change involves a pure spatial move (scrolling or panning), a pure scale variation (e.g.,
zooming), or a combination of these two categories (e.g., moving a fisheye view).