Re: Popper and Palaeontology (was: Re: "running" elephants - locomotary analoges)

[John Conway <john_conway@mac.com>]

On Monday, April 21, 2003, at 01:17  AM, David Marjanovic wrote:

> 	Back to part of the topic: Cladobabble. The way I see it at least, 
> every cladogram represents the following 2 hypotheses:
> 1. The cladogram is the most parsimonious (or, for molecules, often 
> most likely) way to represent the data.
> 2. Evolution is most parsimonious (respectively, what we call "maximum 
> likelihood" is really that), or at least it's parsimonious enough that 
> the few extra reversals and convergences would not change the 
> cladogram, except if we had lots more data and would use those to make 
> the cladogram.

We don't know that evolution is parsimonious "enough". If we assumed it 
was, this would serve as an axiom of evolutionary biology (see below). 
But I am fighting a fight I didn't mean to get into here.

> The first can be tested very easily (and therefore rarely is): run the 
> matrix through the program again (exhaustive search), run it through 
> another 1, 2 or 3 programs, if you still don't believe it dig through 
> the source code (if Farris gives you that of Hennig86) and look for 
> bugs. Respectively look for typos and other mistakes in the matrix -- 
> but don't add information that wasn't genuinely overlooked or 
> dismissed by the authors, because then you're making another > cladogram.
> 	Which in turn is a good idea, because the fewer you take the more 
> likely you are to choose a subset that will not give you the signal 
> that there is in the totality of the data. In short, add characters 
> and (a bit more important, according to theoretical studies) add taxa. 
> That's what's being done all the time, and why ever-growing data 
> matrices are being used to examine the same old problems (the 
> phylogeny of certain groups) again and again.
>
> The second is much harder to test (though that shouldn't be completely 
> impossible... breed fruit flies for a few decades, so you know their 
> true phylogeny, examine their characters, make a cladogram with those 
> and compare...). Nevertheless, we must operate close to it, or 
> cladistics is completely futile. All I can say is that most 
> parsimonious and most likely trees tend to look similar, both tend to 
> look similar to the results of traditional, more intuitive approaches, 
> and both tend to look similar for different data sets (e. g. 
> morphological vs. molecular) and the same groups. Besides, I can't see 
> a reason to assume that evolution is extra munificent.

I don't disagree. Maybe I didn't make myself clear in my last post; I 
think cladistics works (to a degree) and it might be because of 
parsimony. My point was that Popper's ideas are not necessarily the 
most natural fit for cladistics, or biology in general.

> > Popper's ideas certainly do provide a robust philosophical framework, 
> > but they are notoriously difficult to apply to biological sciences.
>
> Why do you think so?

I think so, and so do plenty of others (refs if needed). It has to do 
with Popper's characterising scientific theories as positing laws that 
forbid certain things - which seems to be very difficult indeed to 
apply to biology, where it is questionable that there are any laws at 
all.

> > They are even harder to apply to biological sciences such as 
> > palaeontology, where testing is so limited.
>
> The famous time machine problem is not so horrible IMHO. Because we 
> can't prove anything anyway (and having a time machine would come very 
> close to proving that something was so and not different), we need 
> only find some prediction that our hypotheses make and that can be 
> falsified without a time machine. Old example: The K-T catastrophe 
> could have been produced by a supernova. This would leave 
> plutonium-244 respectively its decay products in K-T sediments. There 
> aren't any => hypothesis falsified.

My point doesn't go to proof - Popper thought proof was impossible 
anyway. And yes, of course you can construct falsifiable theories in 
biology and palaeontology, it would be absurd to suggest otherwise. 
However, it's not obvious how a lot of methods in biology (such as 
cladistics) conform to a falsificationist programme. One Popper's 
central theses was that there is an asymmetry between verification and 
falsification, in many biological methods this asymmetry doesn't seem 
to be present. See "Reconstructing the Past" (Sober, 1988) for detailed 
arguments against asymmetry of verification and falsification in 
cladistics.

> > (Maybe Popper would have said: "if those old astronomers and 
> > physicists had used my method, they wouldn't have taken so long to 
> > get where they did!" ;-))
>
> Very likely. Because usually (not my idea, some philosopher's) science 
> behaves more like a punctuated equilibrium: a certain view is in place 
> (maybe because certain professors are in their positions), and for a 
> long time nobody notices the contradicting evidence. When someone 
> discovers some, he (have been few women so far... related to the fact 
> that IIRC 58 % of Austrian students, but only 7 % of Austrian 
> "ordinary" professors are women) isn't taken serious. Decades (or, 
> today, maybe months :-) ) later, when some more evidence crops up, the 
> establishment begins to take it serious, but tries to explain it away. 
> Then sometime or other the smoking gun comes, everyone thinks "bingo, 
> that's how it is", hagiographies are written about the 3 people who 
> held the view before everyone else, and it becomes the new paradigma. 
> Until the next guy comes and suggests a modification.

So, is the evolution of science parsimonious? ;-)

As a potted history of science that ain't so bad I suppose - but the 
whole area is minefield of controversy. "Smoking guns" are difficult to 
find in the history of science - they tend to be played up later on.

> > Popper was quite clear that he thought the creation of a hypothesis 
> > was not methodological,
>
> What does this mean?

Popper argued that it was not important where a hypothesis came from - 
that it was outside the logical analysis of science. To quote the man 
himself (The Logic of Scientific Discovery, p. 7):

"[....] the initial stage, the act of conceiving or inventing theories, 
seems to me neither to call for logical analysis nor be susceptible of 
it. [....] it is irrelevant to the logical analysis of scientific 
knowledge."

I hope that's clearer now.

Of course, Popper also thought that the most falsifiable hypotheses 
(i.e. the ones with the largest empirical content - and so by his 
definition the most parsimonious) are the best, and should be first in 
line for testing. But their being the most falsifiable is not in itself 
the test. So, where does the testing come in cladistics? Phylogenetic 
hypothesis are rejected solely on the basis that they are not the most 
parsimonious. This method may not be wrong - that's not what I'm 
arguing - just that it is not obviously derivable from Popper's 
philosophy.

> there aren't many [axioms in evolutionary biology] , if any. :-) What 
> could be such axioms?

It depends on the level of analysis. Phylogenetics has at least one 
obvious axiom: organisms share a common ancestor. I won't go into this 
in great detail - it wasn't my point to begin with.




John Conway, Palaeoartist & Protophilosopher

"Do I contradict myself? Very well, then I contradict myself, I am 
large, I contain multitudes." - Walt Whitman

Palaeoart: http://homepage.mac.com/john_conway/_palaeoart/
Palaeontology & philosophy discussion forum: 
http://clouds.proboards16.com/