July 1948

A Mathematical Theory of Communication

First installment of Shannon's 'A Mathematical Theory of Communication' published

Shannon began by deliberately setting aside what messages mean — so telegrams, speech, pictures and machine data could all be treated by the same mathematics.

July 1948: half of a foundational paper

The July 1948 issue of the Bell System Technical Journal carried pages 379–423 of Claude E. Shannon's 'A Mathematical Theory of Communication' — the first installment of a paper serialized in two parts, with 34 more pages following in October. The issues are dated by month, so 1948-07 is the strongest supported precision.

Shannon framed communication as selecting a message at one point and reproducing it exactly or approximately at another — treating meaning as irrelevant to the engineering problem, not unimportant in every human sense.

Five boxes and an intruder

The paper presented a general system of five components — information source, transmitter, channel, receiver, destination — with a sixth intruder feeding corruption into the channel: a noise source.

Within that frame, Shannon defined information logarithmically, developed entropy and redundancy, and stated fundamental coding results for noiseless and noisy discrete channels — showing that noise did not merely require 'shouting louder': coding could define how closely reliable communication could approach a channel's limit.

A word called 'bit'

The paper used 'bit' as an abbreviation for binary digit — explicitly crediting the word to John W. Tukey; Shannon's contribution was its influential published use and formal informational meaning. He also identified Harry Nyquist and Ralph Hartley's earlier Bell System work as foundations he was extending.

The 1949 University of Illinois Press book was principally a republication of the paper with Warren Weaver's explanatory essay — not a newly expanded technical work. Later titles like 'Magna Carta of the Information Age' belong to the paper's reception, not to what changed in July 1948.

Sources

Researched 23 Aug 2026 5 sources 3 audit passes

How this was checked

Researched from the web into a fact sheet, rewritten from that sheet, then audited against it by a different model. How the pipeline works →

What the sources leave uncertain

The listed title says October while the supplied date and slug say July. The paper was published in two installments: July 1948 and October 1948. July is therefore correct for the first publication event; October is correct only for completion of the serialization.
The journal issues are dated by month rather than a verified day, so the strongest supported ISO precision is 1948-07.