Every email address contains a tiny piece of computing history. Between a person's name and the domain sits a character that most of us type without thinking: @.
It seems inevitable now. We read name@example.com almost instinctively as “name at example.com.” But in 1971 there was no universal rule saying an electronic address should look that way. The convention began with a programmer at Bolt Beranek and Newman in Cambridge, Massachusetts, who needed one character that would solve a very practical problem.
Ray Tomlinson looked at the keyboard in front of him and chose @.
It was a small design decision made during an experiment on ARPANET, the research network that helped lead to today's Internet. More than half a century later, the same separator remains embedded in billions of email addresses.
Email existed before the @ sign
Calling Tomlinson simply “the inventor of email” hides an important part of the story. Electronic messages already existed on large time-sharing computers in the 1960s. Multiple people could use the same machine through terminals, and systems allowed one user to leave a message for another user on that computer.
The limitation was geographical in a computational sense: sender and recipient belonged to the same host.
Tomlinson worked at BBN, one of the companies central to the development of ARPANET. BBN built the network's Interface Message Processors — early packet-switching machines that performed a role broadly comparable to network routers — and its researchers worked with computers running the TENEX operating system.
One existing program, SNDMSG, allowed users to append messages to another user's mailbox on the same computer. Tomlinson had also worked on CPYNET, an experimental program capable of transferring files between computers over ARPANET.
In 1971, he combined the ideas. If CPYNET could move a file across the network, why not make SNDMSG deliver a message to a mailbox on another machine?
That created a new addressing problem. A local username was no longer enough. The program needed to know two things: which user should receive the message and which computer held that user's mailbox.
Tomlinson therefore needed a separator.
Why @ was almost perfect
The separator had to satisfy a mundane but crucial requirement: it should not normally appear inside people's usernames. Otherwise the software could not easily tell where the person's identifier ended and the computer name began.
Tomlinson later recalled scanning the available symbols on his keyboard. Commas, slashes and brackets were already used in names or computer syntax. The @ sign was comparatively uncommon.
Then there was its meaning.
In English, @ had long been read as “at,” especially in commercial notation such as quantities sold “at” a particular price. In an electronic address, that meaning was remarkably natural. A user called alice whose account lived on a machine called host could be written as alice@host: Alice at host.
Tomlinson later joked that @ was effectively the only preposition on the keyboard. The address simultaneously satisfied the computer's need for an unambiguous delimiter and the human reader's desire for something intelligible.
No committee appears to have spent months debating the symbol. It was an engineering choice made because it worked.
Yes, the Model 33 keyboard mattered — but not quite as we remember it
The keyboard Tomlinson was using was associated with a Teletype Model 33 terminal, a ubiquitous machine in early computing. Modern retellings often show someone reaching for Shift+2 because that is where @ appears on many present-day US keyboards.
Tomlinson himself pointed out that this is historically wrong.
In an oral-history interview preserved by the Computer History Museum, he explained that the @ sign on the Model 33 was not on the number 2 key. It sat beside the letter P on the letter row. The arrangement reflected the relationship between the terminal's physical keyboard and ASCII character codes.
This detail is minor, but it captures how quickly familiar technology rewrites our memory of older machines. The @ sign survived Tomlinson's experiment; the keyboard layout around it did not.
The symbol was also not completely unused in computing. Tomlinson recalled that on MIT's Multics system, for example, @ had a special terminal function that made it awkward to type literally. What mattered for his own addressing scheme was that it was suitable as a separator and unlikely to be confused with an ordinary username.
The first network email crossed a very short distance
Tomlinson tested his modified software between two TENEX computers at BBN that were connected to ARPANET. Physically, the machines were close together. Logically, however, the message traveled from one host to another through the network.
That distinction is what made the experiment important.
The first message itself has been lost to history. Tomlinson repeatedly said he did not remember its contents. The early transmissions were simply tests, probably strings of arbitrary characters typed to see whether the system worked.
Stories sometimes assign a neat phrase to the world's first network email, but there is no reliable basis for doing so. Unlike the famous “LO” sent during ARPANET's first host-to-host connection in 1969, Tomlinson's first message was not preserved as a ceremonial moment. He was testing software, not staging a historical event.
Once the program worked, however, he sent colleagues instructions explaining how to use it. The useful message was not the forgotten first test but the realization that people on different computers could now address one another directly.
The basic pattern was already recognizable: user@host.
How a convenience became an Internet standard
Tomlinson's network mail software spread to other ARPANET sites as TENEX and related tools were distributed. Email adoption accelerated rapidly. The Computer History Museum notes that by 1973, email accounted for well over half of the traffic on the research-oriented ARPANET.
Other networks did not immediately agree on a single addressing system. Different conventions competed during the 1970s and 1980s, and @ was not instantly universal. But Tomlinson's notation survived the evolution from experimental ARPANET mail toward standardized Internet email.
Tomlinson also participated in developing early mail standards. By the 1970s, fields familiar to anyone opening an email client today — including From, Subject and Date — were being formalized. Later, the Simple Mail Transfer Protocol, SMTP, became the foundation for transferring Internet mail.
The machinery underneath email changed enormously. Hostnames became Internet domains. Personal computers replaced terminals. Graphical inboxes replaced command-line tools. Networks expanded from a small research community to billions of connected devices.
But @ remained.
The symbol is much older than email
Tomlinson did not invent the @ sign. Its pre-digital history stretches back centuries, although the precise origin of its shape is debated.
Merchants and accountants used versions of @ as a commercial abbreviation meaning “at the rate of.” A notation such as “10 items @ $2” meant ten items at two dollars each. Its commercial usefulness helped the symbol find a place on typewriters and later teletypes.
That is why it was waiting on Tomlinson's keyboard in 1971.
There is an appealing chain of technological inheritance here. A mark associated with trade entered mechanical writing machines; ASCII assigned it a computer code; a Teletype keyboard made it available to a programmer; and a network experiment gave it a completely new global identity.
In many languages, the symbol later acquired colorful nicknames based on its shape. Italians call it the chiocciola, or snail. In Dutch it has been called an “ape tail,” while other languages have compared it with a monkey, an elephant's trunk, a strudel or various curled animals and foods. The English reading “at,” however, is exactly what made Tomlinson's choice so semantically efficient.
A one-character lesson in good design
The success of @ illustrates a principle that runs through the history of computing: enduring standards are not always born from grand inventions. Sometimes they begin as small choices that fit both machines and people unusually well.
Tomlinson needed to tell software that a username belonged to an account on another computer. He could have chosen a more arbitrary punctuation mark. Instead, he selected one that was uncommon enough for the machine and meaningful enough for the human.
That combination made the notation almost self-explanatory.
Today an email address can point to a server farm on another continent rather than a TENEX machine a few meters away. The network underneath it bears little resemblance to ARPANET. Yet when we type @, we are still expressing the same relationship Tomlinson needed in 1971: this person, at this destination.
One character solved the problem so cleanly that the Internet never found a compelling reason to replace it.
