Richter said his scale was deliberately not defined by energy. His 1935 paper advertised it as an energy rating.

Corrected 2026-08-13, within the hour, after a fact-check. The original title was “Richter was asked to define magnitude in terms of energy, and refused on purpose”. It was wrong three ways: his 1978 sentence says “there have been suggestions” and “this was purposely not done” — nobody is named as asking and nobody is named as refusing — and, worst, his own 1935 paper proposes the scale “for rating shocks in terms of their original energy”, and he and Gutenberg published on magnitude and energy in 1936 and again in 1956. Also corrected below: ML numbers did not stay put between 1935 and 1975; the 0.13-unit offset was never carried through catalogues; the 1906 magnitude does have a seismogram-based redetermination; and the USGS lists fifteen magnitude types, not ten.

Every large earthquake produces a number, and almost everyone reporting it calls that number the Richter scale. The United States Geological Survey, which computes it, says this in plain language on its own FAQ page:

The Richter scale is an outdated method for measuring magnitude that is no longer used by the USGS for large, teleseismic earthquakes… The USGS currently reports earthquake magnitudes using the Moment Magnitude scale.

Note the qualifier “for large, teleseismic earthquakes”. Local magnitude has not been retired; the USGS still lists ML as the authoritative type below about magnitude 4.0, and most earthquakes are small ones. What Richter’s scale lost was the headlines, not the job.

The usual explanation for that runs like this: Richter’s scale was a crude first attempt, it turned out not to track the actual energy of an earthquake, and so it was replaced by something that does.

The last part of that is true. The middle part is not so much backwards as under-specified: the encyclopaedias do say “saturation”, they just tend to explain it in energy-underestimate language, which names the symptom rather than the cause. But there is a stranger document in the file — Charles Richter, on tape in 1978, saying an energy definition had been avoided on purpose.

The remark

Caltech’s oral history archive holds a long interview with Richter. In it he is asked about the scale, and he says this:

Frequently there have been suggestions that the magnitude scale should be defined in terms of energy and this was purposely not done, because to do that would have then involved continuous revisions, both numerical and theoretical.

Read that carefully, because it says less than it looks like it says. Nobody is named as asking, and “this was purposely not done” is a passive with no actor in it. What it does establish is a stated engineering reason for the choice: an energy definition would have to be revised every time the theory of how earthquakes radiate energy improved, and every revision would move numbers that had already been published.

So he defined it the other way. Here is the 1935 sentence, from the Bulletin of the Seismological Society of America:

The magnitude of a shock is defined as the logarithm of the calculated maximum trace amplitude, expressed in microns, with which the standard short-period torsion seismometer (T0 = 0.8, V = 2,800, h = 0.8) would register that shock at an epicentral distance of 100 kilometers.

Not the energy released. Not the size of the fault. The height of an ink line on a named machine at a named distance. Nothing in it depends on knowing what an earthquake is, which is the property he says he wanted.

He chose a number that would stay put over a number that would be true. That sentence was mine, not his, and a fact-check took most of it away from me. Two problems, and they are the interesting part of this essay.

First, his own 1935 abstract proposes the scale “for rating shocks in terms of their original energy”, and Gutenberg and Richter published Magnitude and Energy of Earthquakes in 1936 and again in 1956, deriving energy from magnitude. He did tie magnitude to energy in print, repeatedly, and then revised it — which is exactly the “continuous revisions, both numerical and theoretical” his 1978 remark gives as the reason for not doing it. The tape is a retrospective about how the scale was defined, not a policy he kept.

Second, the numbers did not stay put either. The distance-correction table that converts a real station’s reading into the hypothetical one at 100 km was revised by Gutenberg and Richter in 1942, revised again in Richter’s 1958 textbook, and recalibrated regionally by Hutton and Boore in 1987. The same ink line yields different magnitudes depending on which table you are holding.

Two things hiding in that definition

The first is that the standard machine is usually not there. Richter’s sentence says the amplitude the instrument would register at 100 kilometres. Stations sit where geology and money put them, not on a circle of fixed radius, so from the very first sentence the magnitude is a modelled quantity read off a distance-correction table — not a reading anyone took. The popular contrast between the old direct measurement and the new computed one does not survive contact with the original definition.

The second is the constant V = 2,800, the magnification of the Wood–Anderson torsion seismometer: how much bigger the ink line is than the ground motion. It is the anchor. Everything else in the definition is a period, a distance and a logarithm; the magnification is what converts the world into the number.

In 1990, Robert Uhrhammer and Eric Collins took actual Wood–Anderson instruments and measured it:

The static magnification of standard (Ts = 0.8 sec) Wood-Anderson torsion seismographs, determined from measurement of the free period and tilt sensitivity, is 2080 ± 60, and not 2800 as often reported.

Their reading of where 2800 came from is that it was the manufacturer’s theoretical figure — geometry, not calibration. Nobody had checked it against a real instrument. They found the damping was wrong too: 0.7, not the 0.8 in the definition. Two of the three instrument constants in the sentence everyone quotes are wrong.

The magnification discrepancy works out to about 0.13 magnitude units. I first wrote that this was “carried in catalogues for fifty-five years”. That was my invention and it is false: when the IASPEI standard adopted 2080 it also retuned the additive constant so that reported values would stay continuous with the historical scale. The offset was deliberately absorbed, not propagated. Nobody’s old magnitudes moved.

Which is a better irony than the one I reached for. The stability I wanted to credit to Richter’s 1935 construction was retrofitted by a standards committee seventy years later — on purpose, by picking a constant that would keep the numbers where they already were.

I want to be careful here, because this is the kind of finding I like too much: I found one paper making this measurement and no replication of it. It is a single-source claim from a peer-reviewed journal, which is a good sort of single source, and it is still one.

The number people quote for 1906

San Francisco is the test case, because the 1906 earthquake happened twenty-nine years before the scale existed.

The USGS says the traditional figure of 8¼ or 8.3 “comes from Richter (1958)” — his textbook, published fifty-two years after the event. The number the USGS publishes today is Mw 7.9, and its preferred derivation is not seismological: land-survey triangulation done before and after the earthquake, surveyors’ angles converted to fault slip, converted to seismic moment. (The same USGS page carries a seismogram-based redetermination too — Wald and others, 1993, giving about 7.7. The geodetic value is the preferred one.)

The most famous earthquake in American history has a magnitude that was never measured on the scale it is attributed to, and whose preferred modern value was produced by people with theodolites.

What actually expired

The 1978 remark says an energy definition would require continuous revision. Whatever its status as history, it was right as a prediction.

Moment magnitude — Kanamori in 1977, and the working formula from Hanks and Kanamori in 1979 — is defined from seismic moment, a physical property of the fault. And it did the thing the remark warns about: the numbers moved. The 1964 Alaska earthquake was reported at the time as Ms 8.4–8.5 and is now Mw 9.2. Chile 1960 became 9.5.

The reason his scale had to be replaced is one he understood better than the story gives him credit for. Local magnitude saturates: past a certain size, the number stops growing while the earthquake keeps growing. Kanamori’s 1977 paper puts the mechanism as saturation setting in once the rupture dimension exceeds the wavelength of the seismic waves the scale is measured from. (I have that as a close paraphrase from the abstract; my checker could not confirm it as a verbatim sentence, so I am not quoting it.) A machine tuned to a 0.8-second period is listening to a band; a rupture hundreds of kilometres long does most of its work below that band, and the needle simply has nothing left to tell you. (Joining that to Richter’s specific instrument choice is my inference — it is the standard textbook telling, but I did not find a single primary sentence saying it.)

Note what that is not. It is not the instrument maxing out. Per Boore’s 1989 review, Wood–Andersons could write onscale records of great earthquakes at teleseismic distances where other instruments clipped. The hardware was fine. The question had a ceiling.

Today the USGS lists fifteen magnitude types in operational use, each authoritative in a different distance and size regime, with an explicit precedence chain saying which one wins when several are computed. There is no such thing as “the magnitude” of an earthquake; there is a fallback chain, and which link fires depends on how far away the recording stations happened to be.

The name

One last thing, since it is on the same tape. Richter did not call it the Richter scale:

I called it the magnitude scale, and I refrained from attaching my personal name to it for a number of years. And I think it was Professor Byerly who started referring to it as the Richter Scale in public.

Perry Byerly was at Berkeley — a different institution. The eponym was not a credit grab; it was a convenience adopted by somebody else, out loud, and it stuck. Richter also says the name “somewhat underrates Gutenberg’s part,” Beno Gutenberg being the colleague who suggested plotting the amplitudes logarithmically in the first place.

There are two popular versions of this story: that Richter took the credit, and that he insisted for life it should be called Gutenberg–Richter. Susan Hough’s biography, which had his papers, describes something closer to a man with a man who never denied Gutenberg’s and Wood’s parts. I could not get to Hough directly, only reviews of her, so I am not going to characterise her further — but the tape is his own voice, and it does not support either of the tidy versions.

Sourcing note, updated after the fact-check: the deepest problem below is provenance, and it is mine. Neither agent could open the 1935 paper (403 from three hosts) or the Caltech oral history transcript (a 301 to a landing page). So the two quotations this essay turns on — the energy remark and the Byerly remark — rest on secondary quotation, and the 1935-abstract fact that undercuts my original title came to me the same way. Read accordingly. The 1935 definitional sentence is quoted from restatements — GeoScienceWorld, CaltechAUTHORS and one PDF host all refused the full text, so I have the sentence verbatim and identically from several places but have not read the paper. Boore 1989 and Hanks & Kanamori 1979 were read as abstracts and agency summaries, not in full. The 2080 ± 60 figure and the 0.13-unit offset come from Uhrhammer & Collins 1990 alone. The 1964 Alaska original magnitude is the one number here I could only get from secondary summaries, which is why it says “the mid-eights” and not a figure.