Methodology I: Task Analysis - Foundations for Designing User-Centered Systems - page 317

not all tasks will have a unit task substructure. The time to do the whole task is
calculated by simply adding up the calculated times for all the component unit
tasks. It should be noted that you also need to include the time it takes to acquire
the task, i.e., the time it takes for the user to work out what to do. The KLM does
not provide a way of calculating this but Card et al. ( 1980 ) suggest that the time
needed for each unit task when manipulating a manuscript is 2-3 s if the
instructions are written down, 5-30 s if it is a routine design situation, and even
longer for creative composition situations.
To calculate the time for a unit action involves analyzing the operations into
their elemental perceptual, motor, and cognitive actions (e.g., keystrokes). Then
by adding together the times for these individual actions it can be possible to make
time predictions for expert, error-free performance of that task.
11.6.1 Description of KLM Components
The execution of a unit-task requires operators of (basically) four types:
1. Keystrokes (K): unit time based on typing speeds (0.08-1.2 s/keystroke, mouse
click, or button press)
2. Pointing (P): moving mouse to target (clicking is a keystroke) (approximately
1.1 s, but Fitts' law can also be used)
3. Homing (H(mouse) or H(keyboard)): moving hand to/from mouse and key-
board (approximately 0.4 s)
4. Drawing (D): dragging mouse in straight-line segments (0.9 n + 0.16 l where n
= number of segments and l = length of segments).
To these should be added some number of mental operators (M or Mop, 1.35 s).
Well practiced tasks with a simpler structure will require less mental operators.
Getting the number of mental operators correct is a difficulty in using this method,
but if it is done consistently across applications it is less of a problem. Finally, the
system's response time (Sys), if it limits the user's task performance, also has to be
included. This can be estimated, or, again, if the estimate can be assumed to be
consistent across interfaces, is less important than you might fear.
The number of mental operators is computed using a set of rules—basically
between all operators, except those linked through knowledge or skill (e.g., the
keystrokes in a single word, or a set such as point and click a mouse).
Where there are selection rules governing the choice of methods then it is up to
the analyst to decide whether to go for best or worst case time predictions.
The time to perform these operators as noted above can be done in an approxi-
mate style. There are several ways to improve the predictions of KLM models.
(1) Keystroke times vary. Card et al. ( 1983 ) include tables that provide different
times for different keystrokes. These tables do not differentiate different typist
speeds, but do show that different keys take different times. This suggests that