The first of an intended four volume set, Contributions in Science in the Making: 1798-1850, opens with an abbreviated history and some classic papers. A Mr Cartwright describes a patent steam engine and Mr John Rodman a new instrument for trepanning. A broad sweep of papers follow on “Electricity discovered” – from the voltaic pile to the electric motor”, and includes correspondence between Alexander Volta and Joseph Banks, and notes from M. Ampère on electromagnetic experiments. Subsequent sections focus on some of the seminal work of Michael Faraday and Humphry Davy, including safety-lamps for coal miners, fluid chlorine and the condensation of gases.

Succeeding papers concentrate on the nature of light and matter, drawing on work by David Brewster, Faraday, George Airy and John Herschell. A final section covers work by James Prescott Joule, including his famous experiments on the equivalence of mechanical work and heat. Anyone with an interest in the history of science will want access to these four volumes.

Parkfield has suffered moderate earthquakes on average every 22 years
since at least 1850. Based on the periodicity with which the last five such
quakes have happened (1966, 1934, 1922, 1901, and 1881), the next quake
should have turned up in about 1988. It did not. Max Wyss, of the University
of Colorado in Boulder, says he can show why it will hit in 1991.

Wyss claimed success in predicting an earthquake in 1986 in the Stone
Canyon region of the San Andreas fault by seeking out abnormal ‘quiescence’
along the fault. He and his team claim that the current absence of small
earthquakes in the area is a ‘precursory anomaly’ signalling a change in
the physical process governing the production of small earthquakes in the
region (Nature, vol 345, p 426).

The theory of seismic quiescence says that seismicity will die down
and remain at a low level for several years before an earthquake. By charting
the reduced activity over time, it is possible, claims Wyss, to predict
when an earthquake will happen.

Records from the US Geological Survey, which monitors seismicity near
Parkfield using dense arrays of seismometers, show that the rates for earthquakes
of magnitude greater than 2.4 and greater than 2 have decreased in recent
years by 80 and 46 per cent respectively. For quakes greater than 2 in magnitude,
quiescence began in September 1986 and for those greater than 2.4 magnitude
in January the same year. So the quiet period has already lasted between
3.5 and 4 years.

To the north of Parkfield, the anomaly seems to end near latitude 36.05Degree
N. But the Colorado seismologists say that they cannot define the southern
limit. What seems certain, though, is that the anomaly stretches for some
40 kilometres along the fault. This, together with the long duration of
the quiescence, suggests strongly that the anomaly heralds an earthquake
of about 5.7 rather than a small mainshock, says Wyss.

Wyss and his group arrive at the likely date for the quake of 1991 (plus
or minus one year), 4.5 years after September 1986. They base their estimate
on comparison with six other previous periods of quiescence that preceded
a main shock, and using case histories of quakes in California, Hawaii and
in Alaska.

They back their case for a significant anomaly with measurements that
they and members of the US Geological Survey have made using laser-ranging
instruments. These devices can measure distances between benchmarks to an
accuracy of 0.5 millimetres over 5 kilometres.

The geologists measured the lengths of lines across the fault in the
Parkfield region on 700 days from mid-1984. In a second report in the same
issue of Nature (p 428), the American group reports noticing in August 1986
that the rate of shortening of two geodetic survey lines across the San
Andreas fault decreased by about 20 per cent. This happened within a few
months of the start of a period of quiescence.

In the same issue of Nature (p 383), William Stuart of the US Geological
Survey suggests that the work of Wyss and his colleagues provides a basis
for a mechanical model which connects crustal deformation, background seismicity
and a forthcoming mainshock. ‘A detailed and accurate model might therefore
be adapted to forecast earthquakes just as atmospheric models are used to
forecast the weather,’ he says.