The History of AI

Transmission of Information

Era 1 · The foundations (1763–1942) › Information and circuits

Lineage
Machines and computation
Kind
Publication

Who

Bell Labs

What happened

Ralph Hartley, an engineer at Bell Labs, publishes "Transmission of Information" in the Bell System Technical Journal, proposing to measure information with a logarithm.

Why it mattered

The first quantitative measure of information, independent of what messages mean. Shannon starts from it in 1948 and names it in the introduction to information theory.

The full story

In the 1920s the telephone companies wanted to know how many messages a line could carry. Harry Nyquist, at Bell Labs, had studied in 1924 the maximum speed at which telegraph signals can be sent over a channel. What was still missing was a way to measure what was being sent.

Ralph Hartley, a Bell Labs colleague of Nyquist's, presented a proposal in September 1927 at a telegraphy and telephony congress at Lake Como and published it in July 1928. The first step was to set meaning aside: to an engineer, a message is a choice among many possible ones, and what matters is how many possibilities there were. If each symbol is chosen from s possible ones and n symbols are sent, there are s to the power n different messages. Hartley took the logarithm of that number as the measure of information, so that information grows in proportion to the length of the message.

With this measure, Hartley related the amount of information a system can carry to the bandwidth and time available: to send more, you have to use more frequencies or more time.

The measure assumes that all symbols are equally likely. Twenty years later Claude Shannon generalized it to symbols with different probabilities, and in the introduction to "A Mathematical Theory of Communication" cited Nyquist and Hartley as forerunners and credited Hartley with the logarithmic measure.

Under the hood

If each symbol is chosen from s possible ones and the message has n symbols, there are sⁿ different messages. Hartley defines information as H = n · log s, the logarithm of the number of possible messages. The logarithm makes the information of two messages in a row the sum of the information of each.

Shannon keeps this measure for equally likely symbols and generalizes it: if each symbol i has probability pᵢ, the average information per symbol is H = −Σ pᵢ log pᵢ. With base-2 logarithms it is measured in bits.

Leads to

  • A Mathematical Theory of Communication (July 1948)

    Documented direct influence

    The introduction cites Nyquist and Hartley as forerunners and credits Hartley with the logarithmic measure of information

Sources

How to cite

“Transmission of Information.” The History of AI, revised November 13, 2026. https://thehistoryofai.com/en/milestone/hartley-measure-of-information

This entry was last reviewed on November 13, 2026.

Open in the interactive timeline