On May 22, 1960, the sea floor off southern Chile broke open along a fault that ran for hundreds of kilometers beneath the ocean, close to the city of Valdivia. Within a day, the tsunami it set off had crossed the Pacific and killed sixty-one people in Hilo, Hawaii, and more along the coast of Japan. Seismologists measured the earthquake the way they measured any other: by reading the amplitude of the surface waves it sent around the planet. The number that came back was magnitude 8.3.
An 8.3 placed the Chilean earthquake among the largest of the century, but not uniquely so. Other earthquakes going back to 1900 had registered in roughly the same range. What that single number could not show was that the scale itself had stopped working. The surface-wave magnitude scale was built around waves with a period of about twenty seconds, and it tracked the earthquake well only up to a rupture of about a hundred kilometers. Once a fault broke longer than that, the waves it sent out stopped growing in step with the earthquake, even as the earthquake kept growing. Seismologists call the result saturation: past a certain size, every earthquake reads about the same, no matter how much bigger the next one actually is. By the mid-1970s, several seismologists, Kanamori among them, had already identified this saturation as a known defect in the scale; what was still missing was a replacement that did not share it.
In 1977 Kanamori, a seismologist at Caltech, supplied that replacement. He worked back from a more basic physical quantity, the seismic moment, which multiplies a fault's rupture area, how far the two sides slipped past each other, and the rigidity of the rock. From that he estimated the strain energy a rupture had released, a minimum figure where the fault's stress drop was only partial, and built a new scale, Mw, from that estimate rather than from a wave reading that had already maxed out. Applied to the 1960 Chilean earthquake, Mw came out at 9.5; by Kanamori's own formula relating magnitude to estimated energy, that is more than sixty times the energy an 8.3 would imply. The same correction moved the 1964 Alaska earthquake from 8.4 to 9.2, and the 1952 Kamchatka earthquake from the low 8s to 9.0.
Kanamori called the scale's failure plainly what it was: an inconvenient and sometimes confusing flaw in an instrument seismology had trusted for decades. What had been missing, until 1977, was a formula built to keep reading past the point where the old one gave up.
A magnitude scale is only as honest as the range it was built for. The surface-wave scale had been roughly accurate for most earthquakes small enough to fit inside its own assumptions, and wrong by an order of magnitude for the handful of earthquakes that mattered most to get right, because those were the ones that moved coastlines and sent water across oceans.
The 1960 earthquake never changed size. The number describing it did not catch up until 1977.