Summary

  • Carnegie Mellon’s recovered 1982 discussion records a group trying several ways to distinguish jokes from serious posts after a sarcastic elevator warning was taken literally.
  • Scott Fahlman proposed :-) as a joke marker on 19 September; the surviving exchange shows a community convention forming, not a standards decision or proof that he invented every emoticon.

A joke became a safety warning

The thread begins before the smiley. In September 1982, computer-science staff and students at Carnegie Mellon were trading physics puzzles on a shared electronic bulletin board. One asked what would happen to a candle and a drop of mercury if an elevator cable snapped. A follow-up announced that the left elevator had been contaminated with mercury and slightly damaged by fire. It was meant as a joke. Readers treated it as an actual warning; another post had to say there had been no spill.

That confusion was more than an awkward punchline. A bulletin-board post could be read by people who were not present when the joke was made. The board carried ordinary questions, notices and arguments alongside humor. A sarcastic sentence could be mistaken for a serious statement, draw a long reply, and bury the discussion that came before it. Fahlman later described this as a recurring problem in a text-only community where tone and body language were absent. The recovered original thread lets readers see the event and the repair proposals in order.

The first fix was not the face

On 17 September, Neil Swartz suggested putting an asterisk in the subject line of any notice meant as a joke. The archive then shows Fahlman posting an absurd elevator story with (*) in its subject—followed by the sentence that, despite the asterisk, the story was not a joke and should be taken literally. The mark could not carry a stable meaning on its own.

Other participants offered %, & and #, sometimes with elaborate explanations of why a particular character looked funny. One objected that jokes should not need a mark, especially when they resembled real safety warnings. The thread is a small record of interface design by argument: what should the mark mean, where should it go, and could a reader recognize it without a separate legend?

Two days later Fahlman proposed the sideways sequence :-) for jokes. In the same short message, he suggested :-( for things that were not jokes, adding that it might be more economical to flag the exceptions. This reverse proposal did not remain stable. The frown soon came to signal displeasure or sadness, as Fahlman later acknowledged in “Smiley Lore”.

A convention that fit the terminal

The design worked with characters already available on ordinary keyboards. No new image format, editor, server feature or board software was required: a writer could type the sequence, and a reader could turn it sideways. In a 2021 retrospective, Fahlman describes the constraint as a single line of ASCII in the subject field. The choice was therefore easy to copy between machines and messages, but it remained an invitation rather than an enforced annotation. Nothing in the archive shows a parser checking whether a joke had been marked.

The thread also refuses a tidy invention story. The archive begins on 16 September and warns that some earlier posts are missing. It preserves the joke-marker debate, rather than every discussion that may have led to it. Fahlman himself later wrote that he was “or at least” one of the inventors of the sideways face, and said he had no hard evidence about earlier private uses. The claim supported by the record is narrower: this is a dated, recovered proposal of the recognizable ASCII pair in a CMU discussion, followed by visible adoption.

How it travelled

The archived posts show variants appearing in later CMU messages. A 10 November email from James Morris, then at Xerox PARC, explains the convention to colleagues and lists several offshoots. Fahlman’s later estimate is that use spread widely through CMU within a week and to other ARPAnet universities and laboratories within a few months. The November message is an early contemporary trace of the move beyond the original board; the wider timetable is a participant’s retrospective, not a measured adoption survey.

That route matters. The symbol did not become useful because an authority declared it a standard. It travelled because people could read the same plain characters across a working network, repeat the suggestion and adapt it. As the original joke marker acquired meanings such as happiness, the shared form stayed recognizable while its use broadened.

What the smiley could not decide

A marker can help readers infer intent; it cannot settle intent for them. One reply to Fahlman’s proposal argued that a joke symbol was not enough and sketched a more elaborate scheme for rating messages. Other posts kept testing, parodying and extending the idea. The archive records no rule requiring the marker and no evidence that every disagreement disappeared.

The useful historical distinction is between a technical feature and a social convention. The board did not need a new release for three printable characters to circulate. But adoption still depended on readers who recognized the cue, writers who chose to use it, and a network through which the practice could be observed. The face looks like a tiny interface. Its history is a record of people negotiating how much a mark can clarify—and what remains for the reader to judge.

Sources: CMU’s recovered “Joke” thread; Scott Fahlman, “Smiley Lore”; Fahlman’s 2021 retrospective; Carnegie Mellon’s institutional account; scholarly history of earlier face-like symbols; November 1982 board backup transcript.