A lake that freezes each winter lays down its sediment in pairs of layers, a coarse one in the melt season and a fine one under the ice, and a core taken through the deposit can in principle be counted like tree rings. The count is what is called a floating chronology: it gives the number of years between one horizon and another without saying when either of them fell. To fix it in calendar time some second method is needed, and radiocarbon dating of plant fragments held within the same core is the one ordinarily used.
The two methods disagree, and in the older parts of long cores they disagree by centuries. The received treatment of this -- call it the calibration view -- is that the count is the party at fault. Layer counting is subject to errors that accumulate, since every layer missed is missing from every total thereafter, and the errors are asymmetric: a faint couplet is more easily overlooked than an ordinary one is counted twice. So the count is adjusted to the radiocarbon series, and the adjusted count is what appears in print.
Both halves of that reasoning should be accepted. Counting errors do accumulate, and they do run in one direction more readily than the other, so the adjustment is not arbitrary in its sign, and a reader who expected a count to be corrected downward would be right more often than wrong. What does not follow is the use the literature then makes of the result. Once the count has been brought to the radiocarbon series, the agreement between the two is no longer an observation about the world. It is a property of the procedure, and it is cited as though it were the former.
The concession is not free, and it should be said what it costs. Granting that the count is the less reliable of the two removes the strongest argument against making the adjustment at all, and leaves nothing to be said against performing it. The objection here is therefore not that the adjusted chronologies are wrong; several of them are very probably right. It is that a chronology tuned to a second series cannot afterwards be offered as independent support for that series, and a good deal of what is written about the mutual confirmation of the two methods does exactly that.
There is a remedy and it is not expensive. Publish the raw count alongside the adjusted one, with the number of layers whose status was uncertain and the horizons at which the adjustment was applied. Where that has been done, the raw and adjusted series have on occasion diverged in ways that pointed to a real feature of the basin -- a season of unusual inflow, a break in deposition -- rather than to a miscount. A discrepancy is informative only while it is still visible, and the present convention makes it invisible before anyone has looked at it.
