You are on page 1of 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/11807555

How should Fitts' Law be applied to human-computer


interaction?

Article  in  Interacting with Computers · January 1993


DOI: 10.1016/0953-5438(92)90018-B · Source: PubMed

CITATIONS READS

30 4,157

5 authors, including:

Douglas Gillan Kritina L. Holden


North Carolina State University Leidos
161 PUBLICATIONS   2,127 CITATIONS    39 PUBLICATIONS   218 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Graph reading models View project

Mental Models View project

All content following this page was uploaded by Douglas Gillan on 13 November 2017.

The user has requested enhancement of the downloaded file.


How should Fitts’ Law be applied to
human-computer interaction?

Douglas J. Gillan, Kritina Holden*, Susan Adam’, Marianne Rudisill’ and


Laura Magee*

The paper challenges the notion that any Fitts’ Law model can be
applied generally to human-computer interaction, and proposes in-
stead that applying Fit& Law requires knowledge of the users’
sequence of movements, direction of movement, and typical move-
ment amplitudes as well as target sizes. Two experiments examined a
text selection task with sequences of controlled movements (point-click
and point-drag). For the point-click sequence, a Fitt’s Law model that
used the diagonal across the text object in the direction of pointing
(rather than the horizontal extent of the text object) as the target size
provided the best fit for the pointing time data, whereas for the
point-drag sequence, a Fitts’ Law model that used the vertical size of
the text object as the target size gave the best fit. Dragging times were
fitted well by Fit& Law models that used either the vertical or
horizontal size of the terminal character in the text object. Additional
results of note were that pointing in the point-click sequence was
consistently faster than in the point-drag sequence, and that pointing
in either sequence was consistently faster than dragging. The discus-
sion centres around the need to define task characteristics before
applying Fitts’ Law to an interface design or analysis, analyses of
pointing and of dragging, and implications for interface design.

Keywords: human-computer interaction, Fitts’ Law, text editing,


interface design

Cognitive engineeering typically involves decomposing performance into the


component processes of a task. One approach to using task decomposition for
design is to derive time estimates for the task components, then to calculate an
estimate of the total task time. A prominent example of this approach can be
found in the Model Human Processor and its application in the keystroke level
Goals-Operators-Methods-Selection rules (GOMS) model (Card et al., 1983).

Department of Psychology, University of Idaho, Moscow, ID 83843, USA. Tel: (208) 885-5555.
E-mail: INTERNET: gillan@iduil.csrv.uidaho.edu.
*Lockheed Engineering & Sciences Company
‘NASA/Johnson Space Center

291
The Model Human Processor employs a set of three interconnected subsys-
tems - perceptual, cognitive, and motor - each consisting of memories and
processors (e.g., Card et al., 1983). Associated with each subsystem are four key
parameters: memory capacity, decay, representation and processing cycle time.
In addition, the Model Human Processor obeys a set of principles of operation.
Card et al. use three principles of operation that provide parameters for
estimating performance time - Fitts’ Law, the Power Law of Practice, and the
Uncertainty Principle (or the Hick-Hyman Law) (e.g., Card et nl., 1983). The
principle task to which Card et al. (1983) apply their version of cognitive
engineering is text editing with a computer. This paper focuses on how Fitts’
Law should be applied to that task.
Fitts’ Law states that the time to move to a target, MT, will be related to the
movement distance, D, and target size, S, as follows:

MT (in set) = u + b log2 (D/S + 0.5)

(Fitts, 1954; see Welford, 1968, for the above formulation of Fitts’ Law; see also
Crossman and Goodeve, 1983, and Meyer et al., 1988, for alternative derivations
of the law). Log2 (D/S + 0.5) is known as the Index of Difficulty (ID) and
typically has been expressed in terms of bits. (However, this unit of measure-
ment is arbitrary because both D and S are measures of distance, thus their ratio
has no units [see MacKenzie, 19911.) The u term in the Fitts’ Law equation is the
coefficient for the y-intercept and b is the slope coefficient for the linear
regression of MT against ID. The coefficient for the y-intercept will be affected
by constant factors and those that vary randomly with respect to ID, such as the
response time of the computer and the time for the user to press a mouse
button. The slope coefficient has been shown to be affected by the speed and
accuracy requirements of the task (Fitts and Radford, 1966) and by the type of
limb used in making the movement (Langolf et al., 1976). The reciprocal of the
slope coefficient is referred to as the Index of Performance and is expressed in
bits/msec.
A fundamental assumption of the Model Human Processor approach to
cognitive engineering is that the subsystem parameters and operating princi-
ples can be applied under a wide variety of situations. Fitts’ Law (Fitts, 1954)
was originally derived under a fairly limited set of conditions, such as, moving
a stylus, a disk, or a pin by hand from right to left. In those experiments, the
movement distance increased in the right-to-left dimension, and the target size
increased in the direction of movement. However, various researchers have
shown that the Fitts’ Law relation of movement time to the distance and target
size holds for a variety of tasks and types of movements: for examples, discrete
movements (e.g., Fitts and Peterson, 1954), moving a foot to a pedal (Drury,
1975), and microscopically short movements of the finger and hand (Langolf, et
al., 1976). All of these conditions resembled the original Fitts (1954) research in
that direction of the movement and the direction of the change in target size
were the same.
Card et al. (1978) demonstrated that Fitts’ Law could be applied to people
using a control device, such as a mouse, to move a cursor to a target on a

292 Interacfing with Computers vd 4 no 3 (1992)


computer display. This type of target acquisition on a computer is commonly
called pointing. In the Card et af., (1978) research, pointing was part of a
point-click sequence in which people moved to a text object by pointing and
selected the object by clicking on a button on the mouse. Card et af.‘s analysis
showed that, in the point-click sequence, mouse-controlled pointing was
reasonably well described by this model based on Fitts’ Law: Pointing Time (in
msec) = 1030 + 96 logz (D/S + 0.5), where D is the distance the pointing cursor
has to be moved to reach a target text object and S is the horizontal extent of the
target. Card, er al.‘s (1978) experiment differed from Fitts’ (1954) research in that
the direction of pointing and the direction of the change in size of the text object
were not always the same. Based on Card et nt.‘s (1978) research, Card et al.
(1983) proposed that the time to move the hand across a distance, D, to a target
of size S is

MT = a + b log2 (D/S -i- 0.5)

where b = 100 (70-120) msec/bit was a general principle of operation for the
Model Human Processor.

Three challenges to the general applicabili~ of Fitts’ Law


We propose three reasons to question whether the Fitts’ Law model proposed
by Card et al. (1983) should serve as an operating principle that can be applied
generally in the analysis and design of user interfaces to computer systems.
First, pointing in computer-based tasks, such as text editing, typically occurs in
a sequence of movements, and the movement sequence may affect the user’s
pointing target. The point-click sequence examined by Card et al. (1978), in
which clicking on the text object selects it, can be used only to select character
strings of a predetermined size (e.g., a word, line, or paragraph). To select a
character string of a size that the user determines, he or she uses a point-drag
sequence. In the point-drag sequence, the user moves the cursor from a starting
point to the beginning of the character string to be selected, then drags to the
end of the string to indicate the text to be selected. Dragging with a mouse
generally involves moving the mouse on a planar surface while holding down at
least one mouse button. In a point-click sequence, a user can click anywhere on
a character string to select it; thus, the entire block of text is the user’s target for
pointing. In contrast, in the point-drag sequence, a user moves from an initial
point to the leftmost side of the string of characters to be selected (assuming a
left-to-right drag movement); that is to say, only the leftmost side of the block of
text should function as the target for pointing. Consequently, in accordance
with Fitts’ Law, variations in the horizontal extent of the character string to be
selected should affect pointing time in the point-click but not the point-drag
sequence.
The second challenge to the generality of Card et aI.‘s (1983) Fitts’ Law
operating principle comes from the response requirements of dragging follow-
ing pointing. Unlike clicking, dragging requires a controlled movement of the
cursor that has direction, a constrained path, and a temporal duration. Accor-
dingly, our second hypothesis is that experienced mouse users will coordinate

Gillan et al. 293


a b C

Figure 1. Example of S’ for very small, medium and large character strings: (a) S’ for
a one-character string; S’ for a 24-character string; and (c) S’ for a 26-character
string

the pointing and subsequent dragging responses, rather than performing them
as two completely independent responses. For example, to prepare for drag-
ging, people might adjust their hand position as they are pointing or slow to a
stop at the transition between pointing and dragging. Thus, we hypothesize
that when the transition between pointing and dragging occurs with minimal
change in direction (e.g., left-to-right pointing followed by left-to-right drag-
ging), the pointing time should be faster than when the transition between the
two movements involves an abrupt change in direction (e.g., right-to-left
pointing followed by left-to-right dragging).
The third challenge involves the relation between the direction of pointing
and the direction in which the target (e.g., a text object) changes size. Fitts’ Law
has been derived and supported in research in which the movement and the
target size increase in the same dimension, e.g., movement in the horizontal
plane and an increase in the width of the target. In real-world human-computer
interaction, particularly in text-editing tasks, the user approaches a target, such
as a character string, from whatever angle is most direct. This raises the
question of how to determine the target size, S, in text editing: Should S be
based on the size of the target in the direction of movement or in the horizontal
dimension? In their research on pointing with a point-click sequence, Card et
al. (1978) varied the angle of approach, using diagonal, horizontal, and vertical
approaches, but determined target size based on the horizontal size of the
character string. Although the pointing movement was often off horizontal and
the target size increased horizontally, Card et al. (1978) found a reasonably good
fit between movement time and IDS based on the horizontal extent of the
character string. MacKenzie (1991) has recently proposed a new measure of
target size, S’, for off-angle approaches. S’ is the diagonal across the vertical
dimension of the character string in the direction of movement, beginning at
the midpoint of the horizontal dimension (see Figure 1).
In addition to examining the generality of the application of Card et al.‘s
(1983) formulation of Fitts’ Law to pointing, the present research was designed
to investigate dragging. Although dragging is a common movement in current
human-computer interaction, it has not been the focus of much research.
Recently, MacKenzie et al. (1991) studied dragging in a Fitts-type task (moving a
cursor horizontally between boxes) on a computer. However, no research has
294 Interucting with Computers ml 4 no 3 (1992)
investigated dragging under more realistic conditions of use, such as, a text
selection task. The benefit of studying a phenomenon in a more real-world
situation while maintaining rigorous experimental control is improved external
validity and generalizability of the results (see Chapanis, 1988).
To analyse controlled movements in human-computer interaction and how
Fitts’ Law applies, we had subjects perform a component of text editing - a text
selection task in which they pointed at and selected a character string. We chose
text selection as the task to investigate because it was the task studied by Card
et al. (1978) and because of the frequency with which it is required in text
editing. To simulate the real-world task of text selection, a large range of
pointing distances and target sizes were used, the direction of pointing was
varied, and subjects pointed at an identifiable text object embedded in a larger
segment of text. However, to maintain adequate experimental control and to be
able to examine the subtle effects of the experimental manipulations, we
simplified the subjects’ task by requiring them only to point and click or point
and drag.
Pointing and dragging times in the sequence were each recorded separately,
and all subjects’ movements were recorded in detail to identify errors at any
point in pointing or selection. The experiments varied features of the display,
such as the pointing distance, the dragging distance, the pointing target size,
and pointing direction. In addition, Experiment 1 varied the movement
sequence - point-drag and point-click.

Experiment 1
Experiment 1 compared text selection using two different response sequences
- point-click and point-drag. The hypothesis underlying the experiment was
that the target for pointing differs between the point-click and point-drag
sequences. This hypothesis led to two predictions:

(1) The pointing time in the point-click movement sequence should be


sensitive to both the horizontal extent of the text object and the pointing
distance (as has previously been observed by Card et al. [1978]).
(2) The pointing time in the point-drag sequence should only be sensitive to
the pointing distance, not to changes in the horizontal extent of the text
object,

Method

Subjects
The subjects were 7 male and 3 female employees of Lockheed Engineering and
Sciences Company in engineering and research positions. Potential subjects for
the experiment filled out a screening questionnaire about their experience with
a mouse and a Macintosh. The criterion levels of experience were 100 hours
using a mouse and 50 hours using a Macintosh. Accordingly, subjects in the
experiment reported that they had substantial previous experience with a
mouse: Subjects had a mean of 3.4 years of mouse experience (range = 1.5 to 6.0

Cillanet al. 295


‘I
‘I
‘I,, Polnffng
11Distance = 2.0, 7.5 cm, 13.75 cm
‘I

Text 0b)ect: Dragging Distance = .25, 1.0,3.5,6.0 cm

meant? She couldn’t read French.

INEXT

Figure 2. Elements of the basic display. Note that stippled lines and italicized words
are shown in the Figure to indicate the features of the display and were not part of the
display

yr) and performed tasks using a mouse a mean of 11.5 hours/week (range = 3 to
20 hlwk).

Apparatus, stimuli, and data recording software


The experiment was conducted using an Apple Macintosh SE with a 9 in
diagonal screen and a Macintosh mouse (Model A9M0331). The mouse control-
led the movement of the cursor with a control/display ratio of approximately
0.53 (i.e., the second fastest setting for mouse sensitivity on the control panel).
The stimuli displayed on the computer screen were presented using the
Apple Software product, HyperCard. Several basic screen elements - the NEXT
and SELECT buttons, and the two-line block of text with a portion underlined -
are shown in Figure 2. The underlined text, which the subject was to select, is
referred to as the text object. The NEXT button was positioned in the lower
right-hand comer of the screen on all trials, but the SELECT button was placed at
various locations as the pointing distance and direction varied. Additional
screen elements were two types of cursors, only one of which was displayed at a
time: a pictographic hand, which was displayed over the entire display in the
point-click condition and over all of the display except the text block in the
point-drag condition, and an I-bar which was displayed over the text block for
dragging in the point-drag condition. Figure 2 also graphically indicates several

296 Interacting with Computers ~014 no 3 (1992)


of the key features of the experiment (information indicated by dashed lines
and italics are intended to help the reader and were not presented to the
subjects): the pointing distance, the dragging distance, and the text object.
The data, recorded using a custom version of Automat III software provided
by Genesis Software, consisted of the time and cursor location at both the start
and end of the pointing movement and the time and cursor location for all
planar movements of the mouse made with the mouse button down. This data
record could be replayed for observing the precise location of the start and end
of the pointing movement and the detailed topography of dragging.

Procedure
Subjects underwent two sets of two sessions each. The sessions within a set
were given on consecutive days, but the two sets were not necessarily given on
consecutive days. Subjects received three blocks of 48 trials each in all four
sessions, with a session lasting approximately 30 minutes. The blocks were
separated by a 1-2 minute break. All of the trials within a set used the same
movement sequence - either point-click or point-drag. Five subjects received
the point-click sessions first, and the remaining five subjects received the
point-drag sessions first. Before each session, each subject was instructed to
perform tasks as quickly and accurately as possible.
Each 48-trial block within a session replicated a complete factorial design
with three factors:

l The distance between the SELECT button (used to start the timing of a
pointing movement) and the nearest point on the text object - 2.0,7.5, and
13.75 cm.
l The direction of movement to the text object - up and to the left, up and to
the right, down and to the left, and down and to the right.
l The horizontal extent of the text object - 0.25 cm long (1 character), 1.0 cm
long (a 5-character string), 3.5 cm long (a 14-character string), and 6.0 cm
long (a 26-character string).

Trials were randomized within each block.


The general procedure on each trial consisted of the following: At the start of
a trial, the subject clicked on the NEXT button (advancing them to the next
display). Then, the subject positioned the cursor over the SELECT button, clicked
on the button, and moved the cursor to the text object. In the point-drag
condition, after the subject moved the cursor to the left edge of the text object,
he or she pressed the button on the mouse, dragged over the underlined text,
and released the button. As the text was selected, it was displayed in reverse
video. In the point-click condition, the subject moved the cursor to any point on
the text object and clicked the mouse button. If the click was within the text
object, the text was displayed in reverse video.

Results and discussion


Due to a computer memory failure, four trials from a point-drag session were
lost for one subject and were not analysed.

Gillan et al. 297


Response topography
The analysis of the types of responses was based on the categorization of the
responses by an observer who viewed replays of each trial (See Tables 1 and 2
in the Appendix for the categories of responses and their defining features,)
Subjects were highly accurate in both the point-drag and point-click condi-
tions: responses that resulted in an incorrect selection (i.e., pointing errors or
drag errors) or procedural errors occurred on 5.2% of the trials for the
point-click sequence and 3.0% of the trials for the point-drag sequence. In both
types of trials, subjects most commonly made procedural errors, especially
Incorrect Reselect (on 2.4% and 0.4 9'0 of point-click and point-drag trials,
respectively) and drag from SELECX(on 2.1% of point-click and 1.0% and
point-drag trials).

Pointing time
The analyses of pointing times were restricted to trials that resulted in a correct
response. The pointing time data show that the subjects maintained a constant
level of performance, with no significant decrease between Session 1 and
Session 2 within a set, F&9) = 1.88, p>O.OS, no decrease in pointing time in the
three blocks within a session, F(2,lS) = 3.07, p>O.O5, and no interaction
between session and blocks, F(2,18) = 0.93, p>O.OS. Accordingly, the pointing
time data from both sessions within a set were combined for further analyses.
Figure 3 shows the pointing time as a function of the pointing distance and
horizontal extent of the text object for both the point-drag and point-click
movement sequences. The point-click sequence produced pointing times more
than 400 msec (31%) faster, on average, than the point-drag sequence, F(L9) =

d 13.75
- 7.5
---I4- 2.0
~
Point-Drag
Point-Click

0 1 2 3 4 5 6 7 o I 2 3 4 5 6 7

Horizontal Size of Text Object (In em) Horizontal Slza of Text Object (in cm)

Figure 3. Mean pointing time as a function of the pointing distance and horizontal size
of text object for both mouement sequences in Experiment 1

298 lnterucfing with Computers ~014 no 3 (29921


Table 1. The effects of direction of pointing movement on pointing time

Movement sequence
Direction Point-click Point-drag

Down-left 973 msec 1459 msec


Up-left 976 msec 1409 msec
Down-right 973 msec 1391 msec
Up-right 1015 msec 1435 msec

48.35, ~~0.05. In both sequences, the pointing times increased as pointing


distance increased, F(2,lS) = 596.10, ~~0.05, and this relation did not differ
between sequences, pointing distance X sequence F(2,18) = 2.46, p>O.O5.
Although the overall pointing time decreased as the horizontal extent of the text
object increased, F(3,27) = 11.52, ~~0.05, examination of Figure 2 and the
interaction between movement sequence and horizontal extent of the text
object, F(3,27) = 31.5, ~~0.05, show that the effect of text object depended on the
sequence.
An analysis of the data for the sessions with the point-click sequence shows
that pointing time in that movement sequence was influenced by both the
pointing distance and the horizontal extent of the text object, F(2,18) = 349.42,
p<CJ.O5, and F(3,27) = 39.7, ~~0.05, respectively. A Newman-Keuls test shows
that only the text object of 0.25 cm (1 character) produced significantiy slower
pointing times (p<O.OS). This finding suggests that the effect of the size of the
target text object reached its asymptote between 0.25 and 1.0 cm (i.e., between
one and five 12-point Geneva characters). An analysis of the point-drag
sequence indicates that pointing time increased as the pointing distance
increased, F(2,18) = 775.32, p<O.O5, but was unaffected by variation in the
horizontal size of the text object, F-(3,27) = 0.70, p>O.O5. Thus, these data
support the first chailenge to Card ef @L’s (1983) application of Fitts’ Law to
pointing - that pointing is affected by the movement sequence.
In addition to the effects of pointing distance, sequence, and target size, Table
1 shows that the direction in which subjects moved the pointer from the
starting point to the target influenced their pointing times. The significant
interaction between the direction of movement and the movement sequence,
F(3,27) = 3.14, ~~0.05 reflects the relatively greater effect of the direction of
movement on pointing time during point-drag trials than point-click trials. in
the point-drag sequence, moving the cursor down and to the left produced the
slowest pointing. Pointing in a leftward direction required a change in the
direction of movement as the subjects made the transition from pointing to
dragging. Thus, the data are consistent with the second challenge to Card et al.‘s
(1983) application to Fitts’ Law to pointing - that pointing time may be affected
by the direction of pointing in the point-drag sequence.

Fiffs’ Law applied to pointing time


The hypothesis underlying this experiment suggests two models that would

Gillan et al. 299


Table 2. Comparison of two models applying Fit-k Law to pointing times (PT)
for the two movement sequences in Experiment 1

Sequence Model 1 R2

Point-click PT = 795 + 83 logz([PD/TOH] + 0.5) 0.56


Point-drag PT = 1300 + 54 logt([PDITOH] + 0.5) 0.25

Model 2

Point-click PT = 497 + 180 logz([PD/‘0.5] + 0.5) 0.8%


Point-drag PT = 918 + 187 lO@([PD/O.5] + 0.5) 0.99

No&: I’D = pointing distance, TOH = horizontal size of text object

bear examination. Model 1, based on Card et al. (1978), assumes that the
horizontal extent of the text object is the pointing target: pointing time = II + b
log2 ([pointing distance/text object width] f 0.5). Model 2 assumes that the left
edge of the text object is the pointing target. The vertical and horizontal sizes of
the left edge of the text object remained constant in this experiment at
approximately 0.5 cm; accordingly, Model 2 is pointing time = a + b log2
([pointing distance/O.51 + 0.5).
Table 2 displays the results of applying the above two Fitts’ Law models to
pointing time data from Experiment 1. The amount of variance accounted for by
Model 1 was substantially greater for the point-click sequence than for the
point-drag sequence, 56% versus 25%. However, Model 2 accounted for 99% of
the variance in the point-drag sequence. Despite the significant effect of the
horizontal size of the text object on pointing times described above, Model 2
was also a better fit for the point-click sequence than was Model 1, accounting
for 88% of the variance. The regression coefficients in Model 1 for the
point-click sequence resemble those described by Card et al, (1978) in which a =
1030 and b = 96. Model 2 for both of the movement sequences has smaller
y-intercepts and somewhat larger slopes than does the Card er al. (1978) model.
The poor fit between Model 1 and pointing time may have been caused by the
use of the horizontal extent of the text object as the measure of target size, rather
than S’, the change in the size of the target in the direction of movement
(MacKenzie, 1991). Card et al. (1978) used the horizontal extent of the text as
their target size, and Model 1 followed that approach. However, Card et al.
(1978) found a closer fit between pointing times and the logz ([D/S] + 0.05) in
their point-click sequence (R2 = 0.83) than in the present experiment. Accor-
dingly, we reanalysed the fit between Fitts’ Law and the pointing time data
from the point-click condition in Experiment 1 using S’. In addition to changing
the target size used in the Fitts’ Law model, the use of S’ also resulted in slight
changes in the pointing distances because the target location was assumed to be
the midpoint of the text object, rather than the point on the text object nearest
the START button. The S’ values for the text objects and the adjusted pointing
distances are shown in Table 3.

300 Interacting with Computers vol 4 no 3 (1992)


Table 3. S’ and adjusted pointing distances (cm) for all (original) pointing
distances and horizontal size of the text objects in Experiment 1

S’
Horizontal size of the text object (cm)
0.25 1.0 3.5 6.0

Original 2.0 0.2 0.5 0.5 0.4


pointing 7.5 0.3 0.5 0.5 0.5
distance (cm) 13.75 0.2 0.4 0.4 0.4

Adjusted pointing distance


Horizontal size of the text object (cm)
0.25 1.0 3.5 6.0

Original 2.0 2.5 2.2 1.6 1.5


pointing 7.5 8.2 8.0 7.2 6.5
distance (cm) 13.75 14.2 13.9 12.8 12.4

Replacing S with S’ in Model 1 resulted in the following: Pointing Time = 376


+ 153 log* ([D/S’] + 0.5, R* = 0.95, for the point-click sequence. (Substituting S’
for S in the Fitts’ Law Model 1 for the point-drag sequence resulted in an R* =
0.79, higher than that observed with S in Ihe model, but still substantially lower
than the fit with Model 2, R2 = 0.99). The analysis of the point-click data using
S’ in place of S resulted in a very good fit, even exceeding that observed by Card
et at. (1978). The success using S’ supports the third challenge to Card et LIP.
(1983) and suggests that the diagonal across the text object in the direction of
movement, not the horizontal extent of the text object, may be the functional
target for pointing in the point-click sequence when subjects approach the text
from an angle. Perhaps the reason that the effect of the horizontal size of the text
object on pointing time in the point-click sequence reached an asymptote by
the 5-character condition was because S’ asymptotes within a few characters in
12-point text.
Why have previous studies applying Fit&’ Law to computer controls found
that it accounted for a sizable proportion of the pointing time variance, even
though they used S rather than S’? Epps (1986) and Kantowitz and Elvers (1988)
both found that pointing times for moving a cursor to a geometric (nontext)
target were fit reasonably well by Fitts’ Law in certain conditions, accounting
for as much as 93% of the variance (although in many conditions accounting for
substantially less). However, subjects in both studies approached the target in
the horizontal dimension and the target size increased in the same dimension,
thus negating the usefulness of S’. Only Card et al. (1978) used an angular
approach to simulate real-world computer use for text editing. The high
proportion of variance accounted for by Fit& Law in Card ef at. (1978) may have
been because they examined relatively small text objects widths - 1, 2,4, and
10 characters (0.246 to 2.46 cm) where the change in S and S’ would be highly
correlated.
Gillan et d. 301
DragDistance (in cm)

Figure 4. Mean dragging time ~1sa function of the drag distance and pointing disfance
in Experiment I

Drugging time
The data on dragging time are from the trials that resulted in a correct selection
in the point-drag sequence. As with the pointing time results, the dragging
times were unaffected by the session within a set or the block within a session
and, accordingly, the analyses used the data from all blocks.
Figure 4 shows the dragging time as a function of both the dragging distance
and the pointing distance. The further the subject had to drag the cursor, the
longer the dragging time, F(3,27) = 315.59, ~~0.05. The relation between
dragging distance and dragging time was independent of the preceding
pointing distance as can be seen in Figure 4.

Fit&’ Law applied to dragging time


Only one model was developed to relate dragging time to the display features.
That model assumed that the drag target was the rightmost side of the text
object, for which the horizontal and vertical sizes remained constant (approxi-
mately 0.5 cm) throughout the experiment. Accordingly, the Fitts’ model for
dragging was dragging time = 684 + 328 log2 ([dragging distance/O.51 + 0.5).
This model accounts for 99% of the variance in dragging time.

Experiment 2
The purpose of Experiment 2 was to further examine the hypothesis that the
target of the pointing movement in a point-drag sequence is the leftmost edge
of the text object by varying the size of the left edge of the text object to
determine if such variance influences pointing time.

302 Interacting with Computers ml 4 no 3 (1992)


Method
The basic method for the five male and four female subjects in Experiment 2
was the same as that for Experiment 1, except as noted. The display differed
slightly from Experiment 1, with a smaller SELECT button and different words in
the text block. Subjects received three sessions of 108 trials. The design of the
experiment varied the pointing distance (2.5,7.0, and 11.75 cm), the direction of
pointing (up and to the left, and up and to the right, down and to the left, and
down and to the right), the font size of the text object (9 pt, 12 pt, and 26 pt
Geneva), and the drag distance (i.e., the horizontal extent of the text object, 1.0,
3.5, and 6.0 cm).
The entire text block read as follows:

“and fallible and that sometimes are reversed by posterity. Augustine Brannigan
has reviewed several instances of scientific discoveries in which retrospective
reinterpretation was at least as important as the original contribution had been.
For example, he makes an interesting case to the effect that our view”

At the 9-point font size, the entire text block was displayed. For the 12-point
font size, the text block began at “by posterity” and ended at “he makes”. The
26-point font size block began at “instances of” and ended at “at least as”. For all
three text blocks, “retrospective” was at the same location relative to the start of
the text - approximately one-third of the way into the second of three lines.
The first ‘r’ in “retrospective” served as the starting point for the text object on
all trials.
In order to maintain the same height-width ratios for all of the font sizes and
to keep the text readable, the vertical and horizontal sizes of the leftmost
character (and all other characters) in the text object co-varied as the font size
increased from 9- to 26-point Geneva. The mean vertical and horizontal
dimensions of a character in these font sizes were as follows: 9 pt = 0.4 x 0.17
cm; 12 pt = 0.5 x 0.24 cm; and 26 pt = 1.2 x 0.57 cm. Although these parameters
co-varied, the character width had no effect on the horizontal extent of the text
object, which was independent of font size. Consequently, for the various text
object size-by-drag distance combinations, different numbers of letters were in
the text object. For example, for a drag distance of 1.0 cm, the text object of 9 pt
characters consisted of the characters ‘retros’, whereas for that same drag
distance, the text object of 26 pt characters consisted of the characters ‘re’.

Results and discussion


Due to computer memory failures, all of the data from Session 1 for one subject
and 54 trials from that session for a second subject were lost and could not be
included in the analysis.

Response topography
The text selection responses were perfect on 69% of the trials. Of the remaining
trials, many resulted in a correct response with overshoots (on 8.5% of the total
trials), corrected drops (14.7%), corrected raises (0.8%) and correct reselects
(3.6%), for correct responses on a total of 96.6% of the trials. Pointing errors (on

Gillan et at. 303


F

@
=
,E c

g 800-r 1 B 800 I I
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5

a Vertical Size of Text Object (in cm) b Vertical Size of Text Object(incm)

lea0 -

ii:

E
F
B
2
g 0007 I

0.0 0.5 1.0 1.5

C VerticalSizeof Text Object (in cm)


Figure 5. Mean pointing time as a function of pointing distance, vertical size of text
object, and horizontal size of text object in Experiment 2: (a) horizontal size of text
object = 1.0 cm; (b) horizontal size of text object = 3.5 cm; and Cc) horizontal size of
text object = 6.0 cm

0.1% of the total trials), dragging errors (0.9%), and procedural errors (2.4%)
were the only types of errors.

Pointing time
As can be seen in Figure 5, pointing time increased as the pointing distance
increased, F(2,16) = 1190.9, p<O.O5, and decreased as the vertical size of the text
object increased, F(2,16) = 75.5, ~~0.05. In contrast to Experiment 1, the
direction that subjects moved the cursor had no affect on pointing time, F(3,24)
< 1. The effect of the vertical size of the text object further supports the
hypothesis that the pointing target in a point-drag sequence is the leftmost side
of the text to be selected and that variation in the size of that target would affect
pointing times.

Fitts’ Law applied to pointing


Table 4 shows the fit of three Fitts’ Law models relating pointing time to the
vertical size of the text object (Model 1), the horizontal size of the text object
(Model 2), and the size of S’, the diagonal across the text object in the direction
of movement (Model 3). Both Models 1 and 3 account for a substantial amount of
the variance, with Model 1 providing a somewhat closer fit to the data. Model 2
accounts for a low proportion of the variance. The fit between Model 1 and

304 Interacting with Computers vol4 no 3 (1992)


Table 4. Comparison of three models applying Fitts’ Law to pointing times
(PT) in Experiment 2

Model R2

Model 1: I’T = 729 + 182 log2([PD/TOV] + 0.5) 0.97


Model 2: P-l- = 1216 + 80 logz([PD/TOH] + 0.5) 0.21
Model 3: I-‘-I = 882 + 153 log2([PD/S’] + 0.5) 0.89

Note: I’D = pointing distance, TO” = vertical size of text object, TOH =
horizontal size of text object

pointing time provides additional support for the hypothesis that pointing
times for the point-drag sequence are under the control of two features of a
computer display - the pointing distance and the vertical size of the text object.

Dragging times
Figure 6 shows the dragging times as a function of the dragging distance and
the font size of the text object. As the figure shows, the dragging times
increased across the dragging distances, but decreased as the font size of the
text object increased, F’s(2,16) = 118.2 and 43.9, respectively, both p’s~O.05.
Increasing the font size increased the right side of the text object; thus the effect
of font size on dragging time is consistent with the hypothesis that the target for
the drag movement is the right side of the text object.

Fitts' Law applied to dragging


Because the variation in font size resulted in increased horizontal and vertical

Draaaina

- 6.0

- 3.5

- 1.0

r I I

12 24 36

Text Object Font Sire (in pt)

Figure 6. Mean dragging time as a function of dragging distance and text object font
size in Experiment 2

Gihn et al. 305


sizes of the characters, two models were fit to the dragging time data. In Model
1, Dragging Time = 312 + 355 log2 ([dragging distance/horizontal size of the
text object] + OS), RZ = 0.98. Model 2 takes the dragging distance and the height
of the text object into account and also fits the data very closely, dragging time
= 611 + 383 log2 ([dragging distance/vertical size of the text object] + 0.5), R2 =
0.97.

General discussion
Six basic findings emerged from these experiments:

pointing time for the point-click movement sequence was sensitive to


variation in the horizontal size of the text object, but reached asymptote
between 0.25 and 1.0 cm,
Pointing time for the point-drag movement sequence was sensitive to
variation in the vertical size of the text object and the pointing distance, but
not to variation in the horizontal size of the text object,
Pointing times for the point-click sequence were generally faster than that
for the point-drag sequence,
A modified Fitts’ Law model using S’ in place S best fit the pointing time
data in the point-click sequence (R2 = 0.95), but a Fitts’ Law model using
the vertical size of the left edge of the text object best fit the pointing time
data in the point-drag sequence (R2 = 0.97),
Dragging time was affected by both dragging distance and the font size of
the text object,
Fitts’ Law models using either the horizontal or vertical size of the
rightmost character as the target size provided good fits to the dragging
time data (X2 = 0.97 and 0.98, respectively).

Using Fitts’ Law as a design metric


Within the human-computer interface design community, there have been
suggestions that greater computational rigour in human performance models
will lead to improved human engineering of computers and other technology
(for examples, see Newell and Card, 1985; Elkind, et al., 1989). Newell and Card
(1985) have proposed that the use of design metrics, like those used in other
engineering fields, would be one way to increase rigour. Following the lead of
Card ef at. (1983), Newell and Card (1985) suggested that the Fitts’ Law model
and particularly the slope coefficient estimated by Card et al. (1978) couId be
used in human-computer interface design and evaluation.
Although rigour in the design process is desirable, the results of these
experiments suggest that rigour will not be increased by the Card et al. (1983)
application of Fitts’ Law to HCI design. The present experiments provide
support for each of the three challenges to the Fitts’ Law operating principle
proposed by Card et al. (1983). First, both experiments, taken together, indicate
that the pointing target is a function of the movement sequence; consequently,
a single target size parameter could not be used across these different condi-
tions. Second, Experiment 1 showed an effect of the direction of pointing only
306 Interacting with Computers sol 4 no 3 f1992J
in the point-drag condition, suggesting that, under some conditions, the
physical features of the pointing movement also need to be taken into account.
Third, the best fit to the pointing data from the point-click sequence in
Experiment 1 used a measure of target size, S’, that follows from the direction of
movement.
In addition to the challenges supported by the above research, a fourth
challenge to the general applicability of a Fitts’ Law operating principle is
suggested from research by Gan and Hoffman (1988). They demonstrated that
Fit& Law does not fit data well if the ID is 3 or less. Rather, data from those
conditions are better fit by the square root of the distance, with no effect of
target size. Card et al. (1978) used a reasonably large range of pointing distances,
from 1 to 16 cm, but a restricted range of target sizes, from 0.246 to 2.46 cm (l-10
characters), resulting in IDS generally larger than 3. Likewise, Epps (1986) and
Kantowitz and Elvers (1988) used large pointing distances and small targets,
resulting in large IDS. However, computer users often point over short distances
at large targets. For example, a person editing text will often point across
relatively short distances as they correct one error after another in sequence on
the same line. In addition, many of the changes that people make during text
editing involve changing relatively long character strings, as they change a
word or phrase. Thus, in real-world text editing, short pointing distances and
large target sizes will be encountered frequently, leading to IDS well under 3.
Thus, Gan and Hoffman’s findings as applied to text editing further circums-
cribe the usefulness of the Card et al. (1983) application of Fitts’ Law to user
interface design.
Taken as a whole, these findings indicate that the notion of a simple, general
application of Fitts’ Law as an operating prinicple is not tenable. Before Fitts’
Law could be applied successfully to an interface design, a detailed examination
of such task conditions as movement sequence, direction of movement, and the
ratio of pointing amplitude and target size would be needed. Only then could
designers know whether Fitts’ Law should be applied and, if so, what
y-intercept and slope coefficient estimates would be most reasonable. This
more complex, specific approach to cognitive engineering would be more
time-consuming, but ultimately, it would lead to more successful designs than
would the simpler notion of a general-purpose operating principle that could be
applied under all task conditions.

Poinfing
One of the interesting results in Experiment 1 was not anticipated: subjects
consistently pointed faster in the point-click sequence than in the point-drag
sequence. For example, on point-click trials with the one-character text object,
S, the target size with that movement sequence, was 0.2 cm. On point-drag
trials with the one character text object, the vertical extent of the text object, the
target size with that movement sequence, was 0.5 cm. Despite the greater target
size for the point-drag sequence, pointing was faster on point-click trials by
300-350 msec: 821 versus 1173 msec at a pointing distance of 2 cm, 1160 versus
1453 msec at a pointing distance of 7.5 cm, and 1331 versus 1651 msec at a
pointing distance of 13.75 cm.

Giffan et al. 307


One hypothesis to account for the difference between pointing times in the
point-click and point-drag sequences centres around the preparation for the
drag response during the pointing movement in the point-drag sequence. For
example, before beginning the drag movement, subjects had to change the
direction of movement from that used during pointing; on half of the trials, the
directional change was abrupt, from moving towards the left to moving towards
the right. If subjects began preparatory motor activities in anticipation of
dragging while they performed the pointing movement, they may have slowed
pointing (see Rumelhart and Norman (1982) for a discussion of preparatory
movements during another cognitive-motor task, typing). The greater effect of
movement direction on pointing time for the point-drag sequence than for the
point-click sequence and the greater effect of abrupt changes in direction from
leftward pointing to rightward dragging in Experiment 1 provide support for
this hypothesis. Future research might use videotape to examine preparatory
movements during pointing in both movement sequences.

Dragging
The present research shows that Fitts’ Law provides a good fit for dragging
time. The results of both experiments show a powerful effect of dragging
distance on dragging time, and the results of Experiment 2 suggest that the
target for dragging is the last character in the string to be selected. However, as
we proposed for pointing, the precise target for dragging and, consequently, the
display feature that controls the dragging movement, may vary as a function of
the details of the user’s task. For example, selecting a block of text by dragging
the cursor vertically down the text block would likely have a different target
than dragging horizontally through a text object. Thus, the present research
should not be used to propose a single Fitts’ Law model that is universally
applicable to dragging.
In both experiments, subjects dragged much more slowly than they pointed.
Averaged across the range of movement distances and targets, subjects dragged
the cursor at a mean rate of 2.0 cmlsec and 2.3 cm/set, whereas they pointed (in
the point-drag sequence) at a mean rate of 5.4 cm/set and 5.3 cm/set in
Experiments 1 and 2, respectively. The performance differences between
dragging and pointing in the present experiments might be due to a number of
differences between the movements. First, small physical differences between
pointing and dragging, such as having to press the button throughout the drag
movement, may have slowed dragging. MacKenzie et at. (1991) compared
pointing and dragging using identical horizontal directions of movement and
box targets, so that the only difference between the movements was pressing
the mouse button. They observed a small, but significant, speed advantage for
pointing with the mouse. However, the size of the difference in speed was
much greater in the present experiments than in MacKenzie et al. suggesting
that other factors may play a role in making pointing faster than dragging in a
text selection task. One possibility is that differences in the cognitive proces-
sing requirements between the two movement conditions may have slowed
dragging. For example, although not instructed to read the text, our subjects
may have read as they dragged across it, thereby slowing their movement. A

308 Interacting with Computers ~014 no 3 09921


Table 5. Examples of common computer-based tasks that use the point-click and point-drag
movement sequences

Task Use of point-click Use of point-drag

Icon/window l Expanding an icon into a window l Icon movement


management l Closing a window l Window movement
. Paging contents of a window l Paging contents of a window
Menu/button . Selecting a menu item l Selecting a menu item
interaction (e.g., full-screen menu) (e.g., pull-down menu)
. Activating a button *Moving a manu
(e.g., tear-off menu)
Word processing l Placing cursor l Selecting a character string of any size
l Setecting a word
. Selecting a line
Graphics . Selecting a graphical object l Moving a graphical object
processing l Selecting a graphics palette icon l Resizing a graphical object
l Drawing a straight line
l Drawing a box
Spreadsheets l Opening a cell l Selecting the contents of a cell

third possible explanation focuses on the relative cost of inaccuracy in the two
movements. In real-world use of word processing systems, recovery from small
errors in pointing takes little time and effort, simply repositioning the cursor a
few millimeters from the site at which the movement erroneously ended. In
contrast, a small error in dragging costs more in effort and time, typically
requiring the user to perform the entire drag movement again. Because of the
relative difference in the consequences of an error, users may assign different
weights to speed and accuracy in pointing and dragging (e.g., Fitts and
Radford, 1966).

Implications for human-computer interface design


Although the best fitting Fitts’ Law models in Experiments 1 and 2 account for a
substantial amount of the variance in pointing and dragging times, the absolute
sizes of the effects are small - typically 100-300 msec. Such small amounts of
time might be considered to be trivial in their consequences for interface
design. However, a small effect taken over many instances, can sum to a large
effect. For example, how many companies would be unwilling to pay sizable
sums of money to improve all their workers’ typing speeds by a few msec/
keystroke? Like typing, the point-click and point-drag sequences are common
movement sequences in human-computer interaction. Table 5 shows examples
of the uses of these movement sequences in current interfaces. In addition, as
direct manipulation interfaces proliferate and as interfaces for the production of
graphics increase, pointing, clicking and dragging (as well as other movements,
such as boxing and circling (Jackson and Roske-Hofstrand, 1989) will increase in
frequency. Thus, improvements of 200 msec per movement could add to savings
of person-years taken across all workers and all tasks within a corporation. In
addition, in certain environments in which either computer time or labour is
very expensive, small savings of time could result in notable savings in cost. For
example, reducing computer use by 1 minute per day per astronaut on the Space
Station, Freedom, (i.e., 300 pointing and/or dragging movements reduced, on

Gitlan et al. 309


average, by 200 msec) would result in an annual saving in labour cost of nearly
one million US dollars (using a projected cost of US $20,00O/person-hour).
The present results suggest two principal ways in which interface designs
could be improved to reduce pointing and dragging times. First, the finding
that pointing is faster in a point-click sequence than in a point-drag sequence
suggests that, when selectable objects can be predefined, users should have the
ability to select the objects via the point-click sequence. In addition, overall
selection times will be even faster using point-click than point-drag because
clicking is much faster than dragging. For example, making word processing
functions directly selectable by point and clicking instead of forcing users to
select them only via pointing and dragging through a pull-down menu should
result in both faster pointing and faster selection (for example, see Microsoft
Word - Macintosh Series, Version 5.0, or Windows Series, Version 1.1 - for
the use of buttons on ribbons and rulers). Likewise, text objects (or graphical
objects) that users are most likely to select in order to edit should be selectable
by pointing and clicking. However, because dragging provides the users with
the ability to determine the size of the object being selected if it cannot be
determined before the task, an interface design that combines both the
point-click and point-drag sequences will maximize both speed and flexibility.
The second result that points towards an improvement in interface design is
the finding that the sizes of various aspects of the text object affect pointing and
dragging. For example, creating text displays with greater selectable space
above and below the letters would increase the target sizes for pointing in both
the point-click and point-drag sequence by increasing the size of S’ and the
vertical size of the text object, respectively. One potential use of this design
feature would be in hypertext systems: selectable words (i.e., words that serve
as hypertext links) should be conceived of as targets for pointing, and
consequently, should have a large selectable area around them. In addition, in
word processing systems, the selectable area for the final letters in words and
after punctuation (likely places to end dragging) should be extended both
horizontally and vertically, thereby increasing the target for dragging. Typical-
ly, users want to select to the end of a word of sentence, and do not want to
select the space following or the letters above or below the word or sentence.
Thus, extending the selectable area both horizontally and vertically around the
terminal letter in a word or around a period would reduce the selectable areas in
the surrounding spaces (possibly reducing overshoot errors) and would make
dragging to the end of a word or sentence faster.

Acknowledgements
This research was funded by the National Aeronautics and Space Administra-
tion, Office of Aeronautics and Space Technology, and was performed at the
Johnson Space Center Human-Computer Interaction Laboratory. The authors
thank Drs Richard Pew, Tim McKay, and Mary Czwerwinski for their helpful
comments on earlier versions of this paper.
310 Interacting with Computers ~014 no 3 (1992)
Reprints may be obtained from Douglas Gillan at the University of Idaho,
Department of Psychology, Moscow, ID 83843, USA.

References
Card, S.K., English, W.K. and Burr, B.J. (1978) ‘Evaluation of the mouse, rate-controlled
isometric joystick, step keys, and text keys for text selection on a CRT’ Ergonomics 21,
601-613
Card, S.K., Moran, T-P. and Newell, A. (1983) 7%~ Ps~c~~o~ogy of ~~~rn~n-corn~u~~~r
~~fer~cf~on Lawrence Erlbaum Associates
Chapanis, A. (1988) ‘Some generalizations about generalization’ Humnlz Fncfors 30,
253-267
Crossman, E.R.F.W. and Goodeve, P.J. (1983) ‘Feedback control of hand-movement and
Fitts’ Law’ Quarterly 1. Experimental Psychol. 35A, 251-278 (Original work presented at
the meeting of the Experimental Psychology Society, Oxford, England, July 1963)
Drury, C.G. (1975) ‘Application of Fitts’ Law to foot-pedal design’ Human Fnctors X7,
368-373
Elkind, J.I., Card, S.K., Hochberg, J. and Huey, B.M. (eds (1980) Human PerJormance
Motels and Computer-Aided Engi~leeri~g National Academy Press
Epps, B-W. (1986) ‘Comparison of six cursor control devices based on Fitts’ Law models’
Proc. Human Facfors Society 30th Annual Meeting Human Factors Society, 327-329
Fitts, P.M. (1954) ‘The information capacity of the human motor system in controlling
the amplitude of movement’ I. Experimental Psychol. 47, 381-391
Fit&, P.M., and Peterson, J.R. (1964) ‘Information capacity of discrete motor responses’
1. Experimental Psychol. 67, 103-112
Fitts, P.M. and Radford, B.K. (1966) ‘Information capacity of discrete motor responses
under different cognitive sets’ f. Experimetttal Psychol. 71, 475-482
Gan, K.-C. and Hoffman, E.R. (1988) ‘Geometrical conditions for ballistic and visually
controlled movements’ ergonomics 31, 829-839
Jackson, J.C. and Roske-Hofstrand, R.J. (1989) ‘Circling: a method of mouse-based
selection without button presses’ in Bite, K. and Lewis, C. (eds) Proc. CHI ‘89: Human
Factors In Computing Systems (Austin, TX, USA: April 3O-May 4,1989) ACM, 161-166
Kantowitz, B.H. and Elvers, G.C. (1988) ‘Fitts’ Law with an isometric controller: Effects
of order of control and control-display gain’ 1. Motor Behav. 20, 53-66
Langolf, G.D., Chaffin, D.B. and Foulke, J.A. (1976) ‘An investigation of Fitts’ Law
using a wide range of movement amplitudes’ 1. Motor Behnu. 8, 113-128
MacKenzie, I-S. (1992) ‘Fit& Law as a research and design tool in human-computer
interaction’ H~rna~-Comp~iter ~n#erac~io~~7, 91-139
MacKenzie, I.S., Sellen, A- and Buxton, W. (1991) ‘A comparison of input devices in
elemental pointing and dragging tasks’ in Robertson, S.P., Olson, G.M. and Olson,
J.S. (eds) Proc. CHI ‘97: Human Factors In Computing Systems (New Orleans, LA, USA:
April 27-May 2, 1991) ACM, 161-166
Meyer, D.E., Abrams, R.A., Kornblum, S., Wright, C.E. and Smith, J.E.K. (1988)
‘Optimality in human motor performance: ideal control of rapid aimed movements’
lJsycho1. Review 95, 340-370
Newell, A. and Card, SK. (1985) ‘The prospects for psychoiogical science in human-
computer interaction’ Humor-Complier ~n~er~cfjo}z 1, 209-242

Gillan et al. 311


Rumelhart, D.E. and Norman, D.A. (1982) ‘Simulating a skilled typist: A study of skilled
cognitive-motor performance’ Cognitiue Sci. 6, 1-36
Salmoni, A.W. (1983) ‘A descriptive analysis of children performing Fitts’ reciprocal
tapping task’ J. Human Movement Studies 9, 81-95
Welford, A.T. (1968) Fundamentals Of Skill Methuen

Appendix 1
Table Al. Description of types of responses resulting in a correct selection

Response type Definition

Perfect selection Subject moved the cursor from the leftmost


character of the underlined area (+0.5 character)
directly to the end of the underlined area (+ 0.5
character).
Overshoot Subject moved the cursor from the leftmost
character of the underlined area (+0.5 character)
beyond the end of the underlined area, but then
moved back to the end of the underlined area (20.5
character) before releasing the mouse button.
Corrected drop Subject dragged the cursor vertically down to the
line below the underlined text, but raised it again
to the correct line before releasing the mouse
botton at the end of the underlined area.
Corrected raise Subject dragged the cursor vertically up to the line
above the underlined text, but lowered it again to
the correct line before releasing the mouse button
at the end of the underlined area.
Overshoot/corrected drop Subject dragged the cursor both more than one
character beyond the leftmost character of the
underlined area and down to the line below the
underlined area, but moved it back to the end of
the underlined area before releasing the mouse
button.
Overshoot/corrected raise Subject dragged the cursor both more than one
character beyond the leftmost character of the
underlined area and up to the line above the
underlined area, but moved it back to the end of
the underlined area before releasing the mouse
button.
Correct reselect Subject made an error in selection, but then clicked
on the SELECTbutton again and reselected the
target text as in a ‘perfect’ selection.

312 lnferucfing with Computers uot 4 no 3 fZ992)


Table AZ. Description of types of responses resulting in incorrect selections

Response type Definition

Pointing errors
Miss left Subject ended pointing and began dragging more
than 0.5 character to the left of the initial
underlined letter of the target text.
Miss right Subject ended pointing and began dragging more
than 0.5 character to the right of the initial
underlined letter of the target text.
Miss text Subject‘s click missed the target text. (Applies only
to point-click movement sequence of Experiment
1.1

Dragging errors
End left Subject ended the selection more than 0.5
characters to the left of the initial underlined letter
of the target text.
End right Subject ended the selection more than 0.5
characters to the right of the initial underlined
letter of the target text.
Uncorrected drop Subject ended the selection on the line below the
underlined area.
Uncorrected raise Subject ended the selection on the line above the
underlined area.

Procedural errors
Incorrect reselect Subject made an error in selection, then reselected
the text without first returning to the SELECT button
and clicking on it.
Miss select Subject missed the SELECTbutton.
Drag from select Subject dragged cursor off the SELECT button,
instead of simply clicking on it.
Intermediate click Subject clicked the mouse button while on the way
from the SELECTbutton to the target text.
Double-click Subject selected the text by double-clicking
anywhere on the first word in the underlined area
and dragging the cursor to any position on the last
word in the underlined area.
Backwards Subject began the selection with the rightmost
letter in the underlined text and dragged the cursor
to the left.
Text, then select Subject selected the target text without first
clicking on the Select button, then clicked on the
SELECT button.
Skip Subject skipped a trial.

et al.
Giflan 313

View publication stats

You might also like