Apple replaced Lisa’s “Do It” button after a tester read it as “Dolt”

The user was trying to save his work, and the computer appeared to be calling him an idiot.

Andy Hertzfeld’s firsthand account describes Lisa tests organized by Larry Tesler. Some participants clicked Cancel when the intended action required Do It. One frustrated tester finally explained: “I’m not a dolt”.

In the system font, the capital I resembled a lowercase l. Lose the space while reading Do It and you get Dolt. Apple changed the affirmative button to OK, a word the team had initially considered too casual. Hertzfeld says users then had fewer problems.

The test setup Hertzfeld describes was designed to catch precisely the things a designer standing over somebody’s shoulder can accidentally conceal. Participants sat in a special room with a one-way mirror. A moderator introduced the mouse, then asked them to do tasks such as editing and saving text without further help. The observers had to watch people interpret the interface for themselves.

Tesler began organizing these sessions in the summer of 1981, recruiting friends and family. Hertzfeld describes him as Apple’s leading advocate for putting software in front of real users. That history gives the button episode a useful setting: an early graphical interface was being tested with people who hadn’t already learned its conventions.

The moderator encouraged people to say what they were thinking as they worked, and the sessions were recorded. That made a mistaken click more than a mark on a completion chart. The team could hear the interpretation that produced it. Clicking Cancel might otherwise have suggested a fear of saving, an unclear question or a misunderstood task. This user was objecting to what he thought he had been called.

The offending word was produced by several ordinary details interacting. The space didn’t register, the capital I looked like a lowercase l, and the user read the result as a single word. Spell-checking either label would have accomplished exactly nothing. Testing a different font in a document would also miss the arrangement of letters inside this particular control.

Apple’s preference had initially gone the other way: Hertzfeld says the team thought OK sounded too colloquial. The observed confusion supplied a reason to reconsider that judgment. His account reports fewer problems after the change, without offering a count or a controlled comparison. That is enough to explain the decision; it doesn’t supply a universal ranking of button labels for every modern confirmation screen.

A modern confirmation screen can still pass every spelling check while being misread. The next useful test begins with what someone believes the button will do.