Compston's machine allowed such rocks to be dated with unparalleled precision. The prototype SHRIMP was built and machined in the Earth Science department's own workshops, and looked like something that had been built from spare parts on a budget, but it worked great. On its first formal test, in 1982, it dated the oldest thing ever found—a 4.3-billion-year-old rock from Western Australia.

She took me down the hall to see the current model, SHRIMP II. It was a big heavy piece of stainless-steel apparatus, perhaps twelve feet long and five feet high, and as solidly built as a deep-sea probe. At a console in front of it, keeping an eye on ever-changing strings of figures on a screen, was a man named Bob from Canterbury University in New Zealand. He had been there since 4 A.M., he told me. SHRIMP II runs twenty-four hours a day; there's that many rocks to date. It was just after 9 A.M. and Bob had the machine till noon. Ask a pair of geochemists how something like this works, and they will start talking about isotopic abundances and ionization levels with an enthusiasm that is more endearing than fathomable. The upshot of it, however, was that the machine, by bombarding a sample of rock with streams of charged atoms, is able to detect subtle differences in the amounts of lead and uranium in the zircon samples, by which means the age of rocks can be accurately adduced.