New Zealand Rats: Repeatability and When Can We Toss The Dates?

Kambiz has an interesting post on the rat paper, which, just by serendipity, provides an excellent jumping point for the second post I had planned on this subject. Kambiz points out that:

One hypothesis, suggested in this 1996 Nature article, “Arrival of rats in New Zealand,” indicates people arrived with rats roughly 2,800 years ago. This was established using carbon dating of rat bones.

Which raises a problem. If the current dates conflict with the previous dates, what reason do we have for preferring one set to the other? After all, creationists frequently accuse archaeologists of picking and choosing dates based on their agreement with evolutionary theory. What about repeatability? Shouldn’t we get roughly the same dates?
Back in 1996 some of the first AMS radiocarbon dates on rats in New Zealand were published in the Nature article above. The dates were about a thousand years older than the earliest dated archaeological sites. These dates, it was argued, supported the early colonization of New Zealand. Further analysis of Holdaway’s (and other’s) dates indicated an interesting phenomena. Older dates were on bones processed earlier (1995-96) while the younger dates were on bones processed later (after 1996). There were a number of proposed explanations ranging from inadequate pretreatment of the bones, dietary uptake of old carbon, or simply more experience in the AMS technique. The possibility of inadequate pretreatment uptake of old carbon are addressed in the Wilmshurst et al paper. How?
By repeating the experiment. First, they re-excavated two of the caves that provided the original rat bones and acquired new material to date. The results of the redating were mentioned in my previous post. Second, they tracked down some of the original rat bones and redated them. In all, they redated thirteen of the original bones originating from seven sites in New Zealand. To address the issue of dietary uptake, stable isotope analysis was performed – which indicated that the rats were feeding on terrestrial vegetation (marine diets, for example, could cause a bias). Consequently, this was ruled out as a factor. Wilmshurst et al conclude that the reason for the earlier dates in previous studies was the result of inadequate pretreatment. Personally, I don’t think that experience can be ruled out either as Anderson’s paper (;inked to above) is pretty thorough.
What all this goes to show, is that radiocarbon dates are not tossed out lightly. Usually, a lot of research goes into showing that the date in question is biased in some form or another. One has to demonstrate, experimentally, that the date is in fact biased and show the cause of the bias. In this case, the fact that the older dates were inconsistent with current theories on the colonization of New Zealand was an insufficient reason to dismiss them.