How a Date Gets Fixed

Everything in this chapter turns on one distinction: what a dating method dates, as opposed to what you want to know.

Start with the most famous. Radiocarbon dating works because living things exchange carbon with the atmosphere and carry the same proportion of the unstable isotope carbon-14 as their surroundings. At death that exchange stops and the carbon-14 begins to decay at a known rate. Measure what is left and you can say how long ago the exchange stopped.

Note what that sentence says. The method dates the death of an organism. It does not date an object, a building or an event, and the gap between those things has misdated more than one site. Wood was seasoned for years, stockpiled for decades, and salvaged from older structures routinely; a beam felled in 1290 can sit perfectly happily in a wall raised in 1400. Archaeologists call this the old-wood problem, and it is why dendrochronology aims for the sapwood or the bark edge, which gives the felling year, rather than a sample from somewhere comfortable in the middle of the trunk.

Radiocarbon has a second complication that is more interesting than it sounds. The method assumes the atmospheric level of carbon-14 was constant, and it was not: it shifted with solar activity and the carbon cycle, and then violently with fossil fuel burning and nuclear testing. So raw results have to be calibrated against sequences of tree rings whose ages are known by counting. This is why a published radiocarbon date is a range with a confidence level attached and never a year, and why anybody who quotes one as a single number has dropped the part that tells you how much to trust it.

Dendrochronology is the precision instrument. Tree rings vary in width with each year's weather, so a run of rings is a barcode that can be matched against master sequences assembled from overlapping timbers stretching back thousands of years. Where a match exists, the result is extraordinary: the year, and sometimes the season. Where the master sequence does not exist (the wrong region, the wrong species, too far back) the method has nothing to say at all.

Underneath all of it sits stratigraphy, which gives no absolute numbers and is the backbone of the whole enterprise. In an undisturbed sequence, the deeper layer is the older one. That yields relative dates (this before that) which is often exactly what an argument needs, and it works where nothing datable survives at all. Its weak point is the word undisturbed. Burrowing animals, later digging, ploughing and building all move material between layers, and reading a disturbed sequence as an intact one is the classic way to invent a chronology that never existed.

Which leaves the coin, and the most elegant piece of reasoning in the set. A coin of a known emperor turns up in a layer. What does it fix? It fixes the earliest possible date for that deposit, because a coin cannot be in the ground before it was struck. That is a terminus post quem: the date after which. It fixes nothing at all about the latest, because coins circulate, get hoarded, and are lost decades or centuries after issue; worn Roman coins turn up in genuinely medieval contexts and cause genuine confusion.

The asymmetry is the whole point. One direction is locked and the other is wide open, and treating the coin's date as the layer's date is the standard error: made, over the years, by people who knew better and wanted a number.