The mouse beat a knee lever in a 1967 paper, once users had practice

Your knee was once a serious candidate for moving the cursor. The mouse had competition, and some of it was underneath the desk.

William English, Douglas Engelbart and Melvyn Berman’s March 1967 paper compares screen-selection devices for SRI’s text-manipulation system. The experiment included a knee-operated control alongside a mouse, joystick, graphical tablet and light pen. This paper is the source for the comparison, rather than a modern recreation of its tests.

Experienced participants did best with the mouse. Beginners tended to fare better with the light pen and knee control. An interface that looks easy on first contact can behave differently once someone has learned it; that distinction was already causing trouble for tidy winner declarations before most people had ever touched a computer mouse.

The researchers also timed getting to the device from the keyboard, rather than treating pointing as the entire job. Moving your hand, finding the control and returning to other work consume time too. A beautifully quick isolated action can become less impressive when the rest of the task is invited back into the room.

Their tested joystick was slowest and least accurate. The paper then does something worth appreciating: it warns against extending those results to every possible joystick or other device. They were judging “the particular examples that were tested.”

That caution matters because controls have implementations, not just names. A stiff button, awkward mounting or unfortunate movement ratio can spoil a particular example without proving the whole idea worthless. Nobody should have to defend every keyboard ever made because one office bought a terrible keyboard (procurement would appreciate that arrangement).

The knee lever also makes the task’s constraints easy to see. If pointing does not require a hand, the hands can remain available for something else. That potential benefit still has to survive an actual comparison of control, fatigue and the work being done. An unusual input method is useful when the measured improvement survives the rest of the task, including its awkward parts.

For teams comparing new interfaces, the useful modern experiment would preserve that distinction. Someone learning a system for an occasional task has different needs from someone who will use it all day. Combining them into one average can make the final number easier to present while making the decision harder to trust.

More unusual controls will keep arriving, and each can be tested against the actual work instead of the assumption that unfamiliarity itself is progress. The old paper leaves room for better examples to change the outcome, which is a healthier future than announcing that one experiment settled input forever.

The knee had a chance.