C14 Dating of Parchment: Testing the Test in ’72

I am, and have been, accepting of the radiocarbon dating results for the Voynich Manuscript, as released by the University of Arizona. Well, at least as far as I believe it is the best possible current method of testing samples of parchment and vellum for age. But I am a skeptic at heart, and a pragmatist by nature, and to not automatically assume the infallibility of science, or of scientists or their methods of experiment. Not being able to test their methods myself, in many cases, I would have to rely on the hope that their methods are correct and accurate. Better yet, in some cases the scientists test themselves, and their own methods and conclusions… and we would hope that when they do, they can be, well, “scientific” about it. I mean, we must even trust them, in this self-regulation.

In the case of the accuracy and value of the radiocarbon dating of vellum, there is at least one, seminal example. The paper is entitled, “Radiocarbon Dating of Parchment” (Nature, volume 235, January 21, 1972). It is a 1972 paper outlining an experiment meant to apply the current radiocarbon testing methods to parchment and vellum, both to determine if it would be an accurate method of determining the age of manuscripts, and also, as a cross check to the dendrochronology of tree rings… itself used as a check of radiocarbon dating methods. I wanted to see the article to learn more about C14, as I was already pretty much in awe of the ability to date vellum, and wanted to learn more about it. But… I have to say I was somewhat shocked at what I found.

In the test, several samples of vellum, with known dates, were radiocarbon dated. The point was to compare the results to the known dates, to see if the radiocarbon results were accurate, and could be used in the future with any accuracy. Here is the list of the results:

But I think there is a problem with these results. Below I’ve broken them down, and commented on them:

1820 ± 40, 1720 ± 30 or 1650 ± 15 for a 1788 document.

Ok, so they had three wildly varying results… peaks I suppose, or whatever they call them, and chose the closest ones, and ignored the 1650 ± 15 result? Why? Because they knew it was from 1788, so 1650 ± 15 must have been wrong… and, by a minimum of 123 years!

1750 ± 20 or 1680 ± 15 for a 1752 document.

Once again, they used the known date to come to a conclusion: One result was dead on, so they rejected the other… which was a minimum of 57, and a maximum of 72, years off.

1650 ± 15 for a 1666 document.

Very good, they got one right. Well, they got one result, which it happens, was correct. If they had no check, though, they would not have known, of course.

1600 ± 30 or 1500 ± 25 for all three of these: 1495, 1579, 1578.

OK now… this is interesting… they got two results, both the same, for three documents… of two very different eras. That alone is somehow disturbing to me. Why would a 1495 sample give the same results as a 1578 sample? That alone is almost 100 years in discrepancy. And then, they chose to feel it was accurate, but seemingly only by applying the 1600 ± 30 to the latter two documents, and 1500 ± 25 to the 1495 one. How do they know their results reflected the assigned ones? Because they knew the date written on them, that is how. But if they did not have the dates, they would, first of all, had a maximum age difference in the results of 155 years! That is, as old as 1475, to as new as 1630. The 1600 ± 30 result, if applied to the 1495 document, would be an error minumum of 75 years, and a maximum of 135 years.

The procedure they used, discarding results based on known dates, actually showed that the tests alone could not be relied on. They would, in effect, have been clueless with undated samples. In fact, although the testers seem to present the results as though “C14 works for vellum”, they actually have this very telling passage:

It is interesting that the radiocarbon dates after correction and calibration for secular variations correspond to thier known historical ages. But the nature of the calibration curve first developed by Suess sometimes permits age ranges or alternative dates rather than unique dates. Consequently, for samples of unknown age it may be necessary to use independent criteria to narrow the choice.

Italics are mine. But the point is, this whole test of the test could be summed up as follows: “If you don’t know the date of a vellum document, C14 will not give it to you. It could be well over a hundred years off.”

That is bad enough, but there are other problems. In the same article it is stated that vellum “…was used for writing within very short periods of time from manufacture”. But how do they know this? For who is to say that in the case of the 1788 document, the vellum was not made in 1650, as one result showed, and that the 1820 ± 40 result was not the one in error? Maybe the vellum was over a hundred years old. I mean, since vellum had not been accurately dated before, then how do they know how long it sat? There is some serious circular logic happening here… first, they assume that vellum was used soon after manufacture, to validate their use of the known date of the sample, which they then compare to several wildly varying results, then pick the one which closely matches the date written on the document, and conclude the test is accurate! It is an assumption used to chose a result, and then that result is used to back up the assumption. The ouroboros of scientific testing…. the snake eating it’s own tail… creating, the snake again.

Well of course the testing in the case of the Voynich may still be very accurate. For the time being, we really have to assume that. Unfortunately for us, though, the official test results have never been released. We do not know if other results came up during the testing, but were discarded, as they were in the 1972 test. And if there were other results, we do not know why they were discarded. In the above test, we can see what criteria they used… they knew the dates, and threw out the results which did not match their expectations. But in the case of the Voynich, for “expectations”, they would have to use the opinion of scholars. That is, whichever Voynich scholar they relied upon, to make that judgment call.