A discussion forum for history enthusiasts everywhere
HomeHome  Recent ActivityRecent Activity  FAQFAQ  RegisterRegister  Log inLog in  

Share | 

 Bayesian chronological modelling: Guesses disguised as statements of fact?

View previous topic View next topic Go down 
Nobiles Barbariæ

Posts : 5631
Join date : 2011-12-25

PostSubject: Bayesian chronological modelling: Guesses disguised as statements of fact?   Sun 13 May 2012, 09:57

A recent article in The Observer newspaper, with the rather misleading title of "How Stone Age Man Invented the Art of Raving", is actually about how two British researchers - Professor Alisdair Whittle and Dr Alex Bayliss - have re-examined the extant data regarding neolithic sites in Britain and have applied Bayesian modelling in order to arrive at a more precise estimate of age. The article calls this modelling "a revolutionary technique for dating ancient remains" and then, accepting the researchers' findings as established fact, proceeds to present these findings as a "corrected" version of the history of the period. The most important "corrections" cover a drastic reduction in the estimation of the length of time taken to construct certain neolithic sites, as well as a very specific chronological dating of the different stages in the spread of farming northwards throughout the island. On the strength of this research the individuals involved have now secured a £2M grant to extend the application of Bayesian modelling to similar sites throughout Europe.

The researchers quite understandably claim that their findings, if correct, could shed much new light on the development and systemic functionality of the societies under examination and could serve to increase our understanding of the people and the time in which they lived. The article in fact goes one better and says that it will do just that.

But will it? One is not too encouraged when one reads in the same article that Bayliss, despite championing his revised data as grounds for radical reassessment of how we view this period, then proceeds to trot out some very old assumptions which - even before this "revolutionary technique" came on the scene - had already been long questioned as reliable. His causewayed enclosures are "built by men" (anthropological modelling of the same data does not exclude women from that role). His communities are led by powerful "chieftains" who commission and direct these projects. They meet up at "festivals" (hence the silly "rave" claim in the article's headline) for which the main motivation is provided by "religious belief".

But that's my point anyway. How "revolutionary" is Bayesian modelling if it allows even its most enthusiastic adherents to persevere with assumptions which have not been empirically deduced? And, if the data has not therefore contributed to empirical deduction, are we to assume that its value is rather less than that which its adherents claim?

The answer to both these questions rests in what exactly Bayesian chronological modelling amounts to. The article cited above does not address this in the slightest whereas from my point of view this should have been its central theme, especially since data deduced according to Bayesian principles is being presented as factual. At its simplest, Bayesian theory states that probability of anything increases the more prior data supports it. Where it departs from more traditional empirical examination is when that probability is calculated as being so high that it can thus be treated as a certainty. Bayesian modelling then allows that "certainty" to be considered as factual data which, if necessary, can then be introduced as prior data in the establishment of another probability as a "certainty", and so on. Repeated application therefore produces quite specific deductions which could never have been arrived at using purely empirical assessment of the data.

Despite the inference that Whittle and Bayliss are "revolutionary" archaeologists in the article this application of Bayesian modelling in the field is not actually that new. A few years ago there was much controversy when similar research in South-East Asia (using carbon dating from blue shells found in one very specific site in Thailand) was used to drastically revise not only the timing of neolithic migrations in the region but even the identity of those migrating. Its authors, Charles Higham, Thomas Higham and Amphan Kijngam, were probably not seeking either controversy or profit from their claims, but their findings triggered a race to publication on the part of several authors with varying expertise in anthropology and archaeology whose own "findings" based on theirs were sensational enough to hopefully produce profitable sales. The academic voice of reason - which strove to point out that the Highams had produced what was essentially a guess - murmured away in the background but did not enjoy the same publicity. At this point in time it looks like reason will prevail however. Further Bayesian modelling elsewhere in the region has, of course, produced quite contradictory data presented as no less specific.

Ronald Fisher - a man of many talents and no mean statistician in his time - dismissed Bayesian method as so subjective in its acceptance of what constituted "prior" data as to be useless. I would go further and say that this aspect to it is what makes it the last thing one should ever allow to be applied to archaeological data. Where subjectivism is accepted it is not long before agenda-driven research introduces itself into the field. I can think of no academic discipline where one would actually want this, but in the case of archaeology I would suggest that this is one development the discipline most definitely does not need - it is polluted with enough agenda-driven practitioners as it is!

The article can presently be found here:The Observer/Guardian - How stone age man invented the art of raving
Back to top Go down

Posts : 2496
Join date : 2011-12-27

PostSubject: Re: Bayesian chronological modelling: Guesses disguised as statements of fact?   Sun 13 May 2012, 11:58

I'm not sure it's entirely fair to blame Whittle and Bayliss for the spin put on their work by a journalist but, not having read the book, it's not easy to judge. I’m assuming that ‘men’ is lazy language, rather than an expression of understanding.
Here's some information on the method
And a review by Alison Sheridan.
I see that it's in our library so, if time permits, I might have a look at it.
You're certainly right about Bayesian modelling having been used for rather a long time now and it does entirely depend on the quality of the data. I'm afraid I tend to switch off when statistical methodology starts to be discussed, and I know I should be more critically enquiring, but who can forget Refrew's Thiessien Polygons which managed to make the prehistory of Arran mindblowingly tedious?
Back to top Go down

Bayesian chronological modelling: Guesses disguised as statements of fact?

View previous topic View next topic Back to top 
Page 1 of 1

Permissions in this forum:You cannot reply to topics in this forum
Res Historica History Forum :: The history of ideas ... :: The history of history-