Keystroke-level Model

jackmyers.info

One way to model performance is to utilize the well-understood human-motor system. In 1980, Stuart Card, Thomas Moran and Allen Newell published the Keystroke-level model (KLM) to predict time of task execution. KLM is low-level GOMS model where the method is specified.

Evaluating User Performance

To truly evaluate how long it would take a user to complete a task, you must consider two phases of task completion:

the acquisition of the task — i.e., how the user cognitively builds a mental representation of the task
the execution of the task — i.e., how long it will take the user to physically complete the task once a task-completion strategy has been formulated.

Operator	Remarks¹	Time (sec)
K	Press key good typist (90 wpm) poor typist (40 wpm) nontypist More categories available in Card article	0.12 0.28 1.20
B	Mouse button press down or up click	0.10 0.20
P	Point with mouse Fitt's Law² average movement	0.10 log₂(D/S + 0.5) 0.20
H	Home hands to and from keyboard	0.40
D	Drawing³ — domain dependent	0.9n_D + 0.16l_D
M	Mentally prepare	1.35
R	Response time from system	—

The KLM Model

KLM, obviously, deals only with execution. This model decomposes execution into five physical motor operators, a mental operator, and a system response operator.

K: Keystroking— striking keys, including modifier keys like Shift
B: Pressing a mouse button
P: Pointing by moving the mouse, using the trackpad, etc. to place the cursor over an item
H: Homing — switching the hand between mouse and keyboard
D: Drawing lines using the mouse

Mental Operator

M: Mentally preparing for a physical action

System Response

R: System response (This may be ignored if the user does not have to wait for a response.)

References

^{1. Dix, Alan, Janet Finlay, Gregory D. Abowd, Russell Beale.
Human-Computer Interaction, 3rd Edition. 2003.
↩}

^{2. Fitt's Law has many forms. This is the original form
used with the original KLM. D is the distance to the target and S is the
target size.
↩}

^{3. n_D = total number of line segments;
l_D = total length of all line segments.
↩}