3TA (was Human and ape phylogeny)
kfitzhug at NHM.ORG
Wed Apr 9 14:19:58 CDT 2003
Thanks Dick. I'm not saying that hypothesis (B,C) cannot be tested. What
I'm saying is that those effects which were used to infer the hypothesis in
the first place cannot also be tests of the hypothesis. As well, a
hypothesis inferred from premises with 25 characters, plus some theory,
stands as the best hypothesis for accounting for just those characters. At
the point that the premises are changed to 30 characters, plus some theory,
one has an entirely new inference. The old inference only has relevance to
the 25 characters. The new inference only has relevance to the 30 characters.
The test of a hypothesis is to judge the truth of the hypothesis - that
some set of causal events did occur in the past to produce the effects we
know perceive. Proceeding from premises with 25 characters to premises with
30 characters has done nothing to impugn the truth of the first hypothesis.
A valid test of an explanatory hypothesis seeks evidence of the specific
causal events which led to the originally observed effects. That evidence
stands wholly apart from characters in need of explanation. For instance,
one must seek the physical evidence for some event(s) which led to
speciation; one must seek the physical evidence for some event(s) which led
to the transformation from a plesiomorphic to an apomorphic character in an
ancestral species prior to speciation. These are classes of effects
deductively implied by a cladistic hypothesis, plus some causal theory. The
objective of testing is to find those effects. As was aptly put by Cleland
("Historical science, experimental science, and the scientific method."
2001 Geology 29: 987-990), historical scientists search for "a smoking gun"
type of evidence when it comes to testing.
At 02:18 PM 4/9/03 -0500, you wrote:
>I'll play along. What you are saying is that, if I have data that allow me to
>hypothesize that (B,C) is a natural subset of (A,B,C), then there are no
>data that can be used to test this hypothesis. Suppose my initial
>based on 25 characters. After a period of rigorous research, I have 5 new
>characters and when I include them in the analysis, I now find (total
>that (A,B) is recognized as a natural subset. My initial hypothesis was put
>forward, based on the information then at hand, and was tested with new
>that indicates my initial hypothesis is inconsistent with the data now at
>Why isn't this a test of a hypothesis?
>Kirk Fitzhugh wrote:
> > Actually, no "test" has been performed. The first inference, leading to
> > (A(B,C)), is entirely distinct from the inference leading to (A(B,C)).
> > These are different inferences derived from different sets of premises. One
> > has nothing to do with the other except to show that one has modified a
> > previous inference to take into consideration additional observations in
> > need of explanation in the same context as the original observations - this
> > is nothing more than abiding by the requirement of total evidence. One
> > might claim that they deduced, as consequences of (A(B,C)), additional
> > synapomorphies under the guise of a Popperian test. Unfortunately, such a
> > deduction is not a valid test since (A(B,C)) will account for those new
> > data no matter what. What are of relevance as tests are consequences of the
> > causal events claimed by (A(B,C)). Such consequences must be of a class of
> > effects independent from the class of effects, i.e., characters, the
> > hypothesis is intended to address.
> > Kirk
> > At 01:19 PM 4/9/03 -0500, you wrote:
> > >A phenetic classification is a hypothesis about natural
> > >relationships. What are
> > >the relationships among A ,B and C? Our initial data set yields
> (A(B,C)). We
> > >"discover" a new set of characters that we believe will help clarify the
> > >relationships. We include these with our initial data and the answer is
> > >(A(B,C)).
> > >Thus, our initial hypothesis has been tested against the new information
> > >and has
> > >been found to be consistent with the new information (or, at the least,
> > >the new set
> > >of characters did not have a strong enough signal to alter the
> > >result). Of course,
> > >one could argue that the 250 characters in our first data set completely
> > >swamped
> > >the signal in the 10 newly discovered characters, but that's another
> > >entirely. The second result is based on total evidence and we have no
> > >reason to
> > >question our hypothesis that (B,C) is a natural group with respect to A.
> > >
> > >Cheers,
> > >
> > >Dick
>Richard J. Jensen | tel: 574-284-4674
>Department of Biology | fax: 574-284-4716
>Saint Mary's College | e-mail: rjensen at saintmarys.edu
>Notre Dame, IN 46556 | http://www.saintmarys.edu/~rjensen
More information about the Taxacom