Re: Neanderthals' Noses Blow Scientists Away

Nick Maclaren (
8 Oct 1996 11:57:46 GMT

In article <>, Jane Andrews <> writes:
|> On 3 Oct 1996, Nick Maclaren wrote:
|> > In article <>,
|> > Don Staples <> wrote:
|> > >
|> > >Species identification needs more precision than the limited factor of
|> > >singular observations. I beleave that there are enough racial
|> > >differences in modern man to allow destinction by race from skeletal
|> > >structure, but all HS.
|> >
|> > Yes, there are. It is trivial to separate most populations and often
|> > easy to distinguish even individual skeletons. In fact, I am 90%
|> > certain that you can separate mainly celtic groups from mainly
|> > germanic ones on the basis of the skulls alone! The same is true for
|> > many geographical variants of other mammals.
|> >
|> I'm sitting at my computer just a few hundred yards from where Nick wrote
|> his message, but despite this geographical proximity, my understanding of
|> the use of craniometrics is rather distant from his.

I think that you misunderstood what I said - I apologise if I was
being unclear. My first sentence was referring to the ability to
separate populations on the basis of whole skeletons. As I pointed
out in my second sentence, I was and am not entirely certain that
my belief on that point is correct - I was relying upon a few things
that I had read.

|> It is possible using large ammounts of data, reasonable sample sizes
|> and sophisticated mulivariate statistical analysis to compare
|> populations. However it is not possible to say anything about an
|> individual. This is because there is a degree of variation in size
|> and shape of skull within any population and there is lots of overlap
|> between groups. ...

You have quoted several common myths here.

The first is that such analyses are complex and need large amounts
of data. The theory of such analyses dates from the 1930s, and the
calaculations were right pigs before modern computers. But they
are actually very simple, and work even on small samples (which
does NOT mean that any particular pair of samples is distinguishable
by these means).

The second is that you can say nothing about individuals because
there is variation within populations and there is a lot of overlap.
This is just plain wrong, I am afraid. It is simple to get results
about individuals, but you have to remember that they will be in
terms of probabilities or likelihoods.

You may be correct that the statistics are such that skull shape is
not enough to be of much use here - I don't know, though it was
certainly used in the heyday of craniometrics. But you certainly
CAN use such techniques on small populations and individuals, at the
risk of getting the answer "insufficient evidence".

|> Before you all fall asleep, bored by my methodological lecture let me
|> point out a major flaw of craniometric methods. When trying to determine
|> relationships from cranial measurements, we assume that size and shape of
|> the skull is determined by genetic factors. Of course in reallity size
|> and shape are infulenced by both genetic and environmetal factors.

Well, yes, obviously. But I don't see that it is a problem.
Statisticians are accustomed to working with both relevant and
irrelevant factors, and have been since the beginning of modern

Nick Maclaren,
University of Cambridge Computer Laboratory,
New Museums Site, Pembroke Street, Cambridge CB2 3QG, England.
Tel.: +44 1223 334761 Fax: +44 1223 334679