In the late afternoon of Tuesday on October 17, 1989, a magnitude-7 earthquake struck the San Francisco Bay area of California.
The damage was enormous. Thousands were injured and 63 died, most of them when a two-kilometre length of double-decked freeway overpass collapsed, killing 42.
Apart from the pancaked freeway, the earthquake is best remembered now for interrupting that day's World Series baseball game between the Oakland Athletics and the San Francisco Giants.
Less well-remembered is that among the other events San Francisco had to cancel was a conference of the world's leading seismologists on the science of earthquake prediction.
I know, I know, it sounds like a bad joke, with shades of the fortune-tellers' meeting that was postponed due to unforeseen circumstances. But there is a serious point here. In 1989, seismologists knew they couldn't accurately forecast earthquakes, and the science has barely advanced since.
The reason is simple enough. Earthquakes happen because stresses build up in the rocks of the earth's crust. Sooner or later, those stresses become insupportable and a fracture occurs, usually - but not always - along a pre-existing fault zone.
The trouble is that there is no reliable way to measure stresses in rock. All we can measure is strain - the deformation produced by those stresses.
And that deformation happens when the rock fractures, in other words, when there is an earthquake.
This makes accurately forecasting seismic events impossible. The best scientists can do is look at past patterns of earthquakes, guess at the pent-up stresses and offer a probabilistic estimate of future seismic activity.
So, for example, today seismologists reckon there is a 67 per cent chance of a major earthquake hitting Los Angeles in the next 30 years.
In contrast, the probability of a major earthquake occurring within 100 kilometres of Hong Kong at any time in the next 50 years is put at just 1.5 per cent.
When in March 2009 Giulio Selvaggi, director of Italy's National Earthquake Centre, and six of his colleagues issued a statement saying there was little chance that a major earthquake was imminent in the central Apennine mountains, they were basing their reassurances on just such a probabilistic estimate.
Alas, just one week later, a magnitude-6 earthquake devastated the Apennine town of L'Aquila, killing 308 people.
Selvaggi and his colleagues were subsequently put on trial, with the prosecutors accusing them of negligence and alleging they "provided an approximate, generic and ineffective assessment of seismic activity risks".
In their defence, the scientists explained the limitations of earthquake forecasting, stressing that it was an inexact science based on assessments of probabilities.
They said a big earthquake was improbable in the near term. But, as anyone who has ever thrown three double-sixes in a row at backgammon (a one in 50,000 chance) can attest, improbable events do happen.
The judge was unmoved, and earlier this week convicted the seven of manslaughter.
His verdict was surely the most execrable display of wilful judicial ignorance since Galileo Galilei was found guilty of heresy in 1633 for daring to suggest that the earth might not be the centre of the universe but that it actually orbited the sun.
The L'Aquila verdict is ridiculous. By implication, if you are going to hold seismologists liable for earthquake deaths, you should sue meteorologists for the damage inflicted by typhoons, racing tipsters for your betting losses, and financial columnists for any money you lose playing the stock market.
The Italian seismologists issued the best forecast they could with the imperfect tools available. Treating them as scapegoats will only deter scientists from pursuing research on more accurate earthquake forecasting.
This ludicrous verdict must be reversed on appeal.