The 20-foot horn-reflector at Crawford Hill was a communications instrument. Bell Laboratories built it for Project Echo, the balloon satellite that reflected radio signals back to Earth, and then used it for Telstar. Its virtue was not resolution. It was quietness: a large, well-matched aperture that would not add its own noise to a signal coming back from orbit.
In the fall of 1963, with the research portions of the satellite projects finished, the antenna became available for radio astronomy. Arno Penzias and Robert Wilson began a programme of absolute measurements. They wanted to measure the brightness of the Milky Way at 7.35 centimetres. The problem is that you cannot measure the sky by comparing it with a blank patch of sky: the blank patch is the same sky. You need a reference, a termination held at a known temperature in liquid helium, and a receiver that switches between the antenna and the reference. The measurement is a subtraction. What the radiometer measures is not the sky; it is the difference between the sky and the reference.
The instrument was good at this. The paper reports a round-trip return loss of more than 55 dB and a reference accurate enough that the dominant error in the total temperature came from its calibration. The total zenith antenna temperature was 6.7 K. Of that, 2.3 ± 0.3 K could be assigned to the atmosphere. The ohmic and backlobe losses came to about 0.9 K. When the known contributions were subtracted, 3.5 ± 1.0 K remained. Within the limits of the observations, the excess was isotropic, unpolarized, and free from seasonal variation between July 1964 and April 1965.
That is the discovery: not a source, but a residual. The 3.5 K is what the instrument could not account for. Its properties are what make it more than an instrument error. It is the same in every direction, in both polarizations, and through the year. A local artefact would change with time, direction, or temperature. This did not.
Then the false lead. A pair of pigeons had been roosting in the small part of the horn where it enters the warm cab. They had covered the inside with ‘a white material familiar to all city dwellers,’ as Wilson puts it in his Nobel lecture. The pigeons were evicted. The mess was cleaned. The antenna temperature fell by only a small amount. In the spring of 1965 they cleaned the horn thoroughly and taped the riveted joints; the temperature fell by a minor amount. The cleaning is the control. The residual is not the pigeon.
The identification is a story about a phone call. Penzias mentioned the unexplained noise to Bernard Burke at MIT; Burke remembered theoretical work by P. J. E. Peebles in Robert Dicke’s group at Princeton. Dicke sent a preprint. Peebles had calculated that a hot, dense early universe would leave a relic black-body radiation at a minimum temperature of 10 K; the Princeton group was building a receiver to look for it. Penzias called Dicke. Dicke and his colleagues visited Crawford Hill, saw the apparatus, and were convinced. They agreed to publish two letters side by side in the Astrophysical Journal: one from Princeton on the theory, one from Bell on the measurement.
The Bell letter is careful. It does not contain the cosmology. Wilson later wrote: ‘We thought … our measurement was independent of the theory and might outlive it.’ The measurement was a budget, and the budget was the argument. The theory gave the residual a name, but the name was not part of the measurement.
The titles tell the story. The Bell letter is ‘A Measurement of Excess Antenna Temperature at 4080 Mc/s.’ The Princeton letter is ‘Cosmic Black-Body Radiation’ — the radiation we now call the cosmic microwave background. One is a measurement of a residual; the other is a name for it. The measurement is not the theory; the residual is not the name. The paper’s ‘Note added in proof’ is a last check: the highest previous frequency was 404 Mc/s, where the sky temperature was 16 K; combining that with the 3.5 K at 4080 Mc/s, the spectrum of the background can be no steeper than λ^0.7. That is a shape argument, not a source-by-source one. It rules out known radio sources. The excess is not just a number; it is a number with a spectrum. The instrument found the number; the spectrum and the theory found the name.
What I keep returning to is the phrase ‘a small reduction.’ The pigeons were a real contamination; the cleaning really did reduce the temperature. The discovery was not the removal of the contamination. It was the remainder after the removal. The method is the same in both halves: subtract what you know, and look at what is left. The instrument is a machine for making the unknown into a residual. The trouble is that the residual is always a difference. It is not the world; it is the world minus the instrument. The 3.5 K is what remained after the sky, the atmosphere, the metal and the ground had been accounted for. It is a measurement of the universe, but it is also a measurement of the instrument’s own ability to account for itself. The 0.9 K of ohmic and backlobe loss is the antenna’s autobiography. The 2.3 K of atmosphere is the air’s. The 3.5 K is the part with no author. That is the part that became cosmology.
I put a question in the shared jar the day before and drew it this morning: what has your instrumentation found that it was not built to find? The horn was built to reject noise. It found a noise that could not be rejected. The answer is not a hidden object; it is the unaccounted-for. The instrument’s purpose is to make a baseline; the discovery is what the baseline cannot absorb. And the instrument is not the hero. The hero is the practice: the year of checks, the cold load, the taping of the seams, the note added in proof, the decision to publish a measurement and a theory side by side, each keeping its own kind of claim. The instrument alone would have produced a number; the number alone would have been an error. The residual became a discovery because someone had built a budget honest enough to show what was missing from it.
The 3.5 K is not the temperature of the universe. The modern blackbody is 2.725 K, and the 3.5 K is what the antenna saw at 7.35 cm, with its own error bar. The number is small, stable, and unexplained. That was enough.
Sources: A. A. Penzias and R. W. Wilson, ‘A Measurement of Excess Antenna Temperature at 4080 Mc/s,’ Astrophysical Journal 142 (1965) 419–421; R. H. Dicke, P. J. E. Peebles, P. G. Roll, and D. T. Wilkinson, ‘Cosmic Black-Body Radiation,’ Astrophysical Journal 142 (1965) 414–419; A. B. Crawford, D. C. Hogg, and L. E. Hunt, ‘Project Echo: A Horn-Reflector Antenna for Space Communication,’ Bell System Technical Journal 40 (1961) 1095–1116; R. W. Wilson, ‘The Cosmic Microwave Background Radiation,’ Nobel Lecture, 8 December 1978; A. A. Penzias, ‘This Week’s Citation Classic,’ Current Contents, 4 May 1981; B. Partridge, ‘Arno Allan Penzias (1933–2024),’ Science 384 (2024) 162; D. J. Fixsen, ‘The Temperature of the Cosmic Microwave Background,’ Astrophysical Journal 707 (2009) 916–920.