Acarina, pycnogonids, Permian ticks?

[Kirk Fitzhugh]

At 12:23 PM 4/20/01 -0400, you wrote:
>I think this seriously misrepresents Popper. Even if it were true (and I
>dont see it) that Popper wished to "relegate all of the arena of hypothesis
>formation to the level of psychologism", this would not lead to the
>exclusion of cladistics (your view of cladistics) from the realm of science.
>For Popper was quite explicit that the issue of hypothesis formation was not
>the relevant issue when deciding if something was to be considered
>scientific or not. To repeat the point I made yesterday, the demarcation
>between science and non-science is the testability of hypotheses, not the
>process used in their formation.

I have not misrepresented Popper's position. He considered the inference of
hypotheses/theories to be by whatever means one wishes. What concerned him
was that those hypotheses/theories be available for potential
falsification. My "view" of cladistics is quite consistent with Popper's
view of science in that the testing of a hypothesis is by way of deducing
consequences. The problem is that neither the initial inference of a
cladogram nor the further inference of cladograms based on additional data
follow the rules for valid deductive inference, hence cannot follow
Popper's requirement for testing. My point then still holds that the
majority of work within cladistics is non-deductive, therefore stands
outside Popper's conception of science. As of yet, I've not seen anyone
provide a correct deductive syllogism for inferring cladograms. Indeed, it
has been known for several hundred years now that one cannot deductively
infer cause (= cladogram) from effects. This direction of inference is
always non-deductive.

>I agree. All possibile phylogenetic hypotheses, including the "true" one,
>are already known - they are the set of alternative trees of life.
>Creativity is a wonderful thing, but not necessary. It does characterize
>many hypotheses, but, once again, it is only testability that makes a
>hypothesis scientific, and it is only success at passing tests that enhances
>its value.

Actually, one does not start from the observations of effects and then
"knows" all possible causes. Inference from effects to cause does not
operate in this way, except maybe in the case of computer algorithms that
are sorting through all possible topologies per some causal theory that
functions as an optimality criterion. Algorithms merely mimic the more
limited mental processes we use to conjoin a causal theory with observed
effects to (non-deductively) infer causal hypotheses. When one sees a pool
of clear liquid on the floor next to an umbrella, do they automatically
infer every conceivable cause for the occurrence of the pool, much less the
composition of the liquid? Of course not. It is also worth noting that any
argument along the lines that we employ deductive arguments in which the
premises consist of all possible cladograms, from which one "deduces" a
subset of cladograms, is also in violation of the rules of valid deduction.
For instance, one must assume the truth of the premises in any inference;
it stretches the credulity of inference to think that all possible causal
situations exist. Similarly, valid deductions would never allow for
mutually exclusive conclusions. One of the hallmarks of valid deduction is
that the conclusion *must* be true given the truth of the premises. There
are no aspects of cladistic inference that abide by those rules.

>No doubt that testing historical hypotheses is difficult - we seem to be
>restricted to mining the same evidentiary base from which we made our
>initial selection - i.e. to discover new characters and to test their
>congruence with our preferred hypothesis.

This is not correct as far as evidence goes. The test of any historical
hypothesis is not by way of those effects the hypothesis is intended to
explain. At the point one introduces additional effects, one has altered
the premises of an earlier inference, producing an entirely new causal
hypothesis. The new and hypotheses are entirely irrelevant to one another
because they are based on different premises. There has been no testing
performed. Congruence is irrelevant here just as it is in the comparison of
cladograms based on different data sets. There have been no deductive
consequences provided, because the direction of the inference is from the
effects observed in the present to a conclusion of events in the past.
There is no valid deductive framework that can characterize this direction.
The standard, and accepted procedures for testing historical hypotheses is
to deduce consequences of other, entirely independent effects that must be
found in the present that directly provide empirical evidence of the
specified causal mechanisms originally hypothesized. As I pointed out in my
last message, Popper makes this immanently clear.