Re: An alternative to ST and AAT

Paul Crowley (
Mon, 25 Nov 96 20:10:54 GMT

In article <> "Phillip Bigelow" writes:

PB> An extrapolated hypothesis taken from hard data........hmmmm.
PB> Whoa!...somebody should have woke me up! (Where's my ringer?)
PB> Ding! Ding! Ding! Ding! Ding! Ding!
PB> Gerrit wins this week's SUBSTANTIVE POST AWARD for
PB> not using frivolous speculation in his post.

PB> Somebody should have warned me that science was going
PB> to be posted! Shees...take a nap and something always happens.

When you get a "Phillip Bigelow award", ALARM bells should ring.
You likely to be in deep doo-dah. I'm afraid that IMO Gerrit's
post was a classic example of bad science: false maths based on
unstated assumptions; the unthinking application of some general
rule; the extrapolation from inappropriate data, little of which
is "hard", to other data which no one claims is "hard" to produce
figures to a completely spurious degree of accuracy.

> Gerrit Hanenburg wrote:
> > >the life expectancy of the australopithecines?
> : 42.2 years.
> : Calculated using the regression equation for primates from
> : Harvey and Clutton-Brock (1985),
> :
> : log (lifespan)=log (780.9)+0.29*log (female bodyweight)
> : (lifespan in days,female bodyweight in grams)
> :
> : Female bodyweight A.afarensis= 29.3kg. (McHenry,1992)

I don't have Gerrit's sources, but I doubt if either bodyweight
or lifespan figures for many primates have much reliability.
He quotes a correlation co-efficient of "r=0.78". Firstly there
is the question of: "Why only females?". The exclusion of males
may be justifiable, but it needs justification. It should not
be done just because it seems a good idea at the time.

Another problem, of course, is the exceedingly limited data on
australopithecine weight. Gerrit gives a spurious accuracy to
his calculation with a figure of 42.2. At best he should say
something like "about 40 with a probable error of plus/minus 10".

However, the main problem is this: If the australopithecines
were more or less standard leaf-eating monkeys living high in the
canopy, it might be reasonable to use an "r=0.78" to estimate
life expectancy from body weight for females (stating the margin
of error, of course). But they were not. We do not know how
they lived or have any idea of their social structure, diet, or
general ecology. We do know it was nothing like that of a
typical primate.

The weakness of Gerrit's method can be seen by applying it to
known hominoids: Does it work for female gibbons, chimps, orangs,
gorillas and humans? Gerrit can enlighten us as to the figures
but I reckon the answers are: gibbons=No, chimps=Yes, orangs=No,
gorillas=No and humans=No. IOW, when tested, we get an
approximately correct answer in 1 case out of 5. If this is so,
or if results are not close to 3 out of 5, then the method is
highly questionable, to say the least.

The application of any such rule to males would probably reveal a
"0 out of 5" score. This might be answerable; or it might not.

Even if we got a high score, the question should be posed:
"Should we _expect_ to get a high score?". I'd suggest that the
answer is "No". Too much depends on particular aspects of the
animal's lifestyle. The unthinking application of general rules
is *bad_science*.

It is far better and much more simple to ask "What are the life
expectancies for known relatives: orangs, gorillas, chimps and
humans?" and cautiously extrapolate from them in the light of
whatever other information we have. Of course that would not
look so "scientific" and would never win a Phillip Bigelow award.

It is so easy to fool yourself in science, especially when you
introduce a teeny bit of maths. I'm sure most of us were fooled.
I certainly was.