Positivism vs Realism

Richard Zander bryo at COMMTECH.NET
Mon Dec 15 10:13:58 CST 1997

Byron Adams wrote:
> Richard Zander wrote:
> >Byron Adams wrote:
> >>
> >> Re: the thread of realism vs positivism:
> >>
> >>         It seems reasonable to me that any scientific research program or
> >> methodology employing discovery operations subservient to positivism should
> >> be exposed as such.  I think this is a very clever and valid way to
> >> criticise a research program or methodology, and I would like to see it
> >> continue.  So far (from reading the threads) I am not convinced that this
> >> argument has been clearly made or defended.
> >
> >Two things about this, Byron. "Discovery operations" assume that there
> >is a reality out there, say, a pattern, that exists independently of
> >your theory and method of research. That's okay, you're a realist. I
> >think that this leads to overconfidence about what "converging to the
> >truth" means, especially using a non-statistical method like parsimony
> >analysis. Postitivism is on the outs now, since W. Quine published his
> >Two Dogmas paper (see his web page) that showed that all statements are
> >theory-laden, and the difference between analytic (idea-generated)
> >statements and synthetic (observational) is rather vague. On the other
> >hand, he said "there is no place for a priori philosophy." To the extent
> >that your assumption that metaphysics HAS a place in science, you are
> >handicapping yourself with a burdensom circularity.
> >
>         First of all, do we agree on a definition of metaphysics?  I'm not
> referring to that associated with crystals, pyramids, the occult,
> or windham hill samplers, although your post (and an earlier one from
> James, who, in a criticism of cladistics said,  "My comments are not
> intended to place cladistics firmly into metaphysics...") hints at this.
> I'm using the term metaphysics to mean the science that deals with
> ontological reality at its most fundamental level.

My dictionary gives a number of definitions of metaphysics,
corresponding to various philosophers' points of view: 1.Bacon: a quest
for formal and final causes;  Kant: the science of pure or a priori
reason; Hegel: logic, Schopenhauer: analysis of experience. Apparently
there is a distinction between epistomology as the theory of knowledge,
and metaphysics as ontology. Ontology is defined as the science of being
or reality. I'll agree with you with this last.

>         Are we in sync?  If so, I maintain that not only does metaphysics
> have a place in science, but that we cannot have the one without the other.
> I'm not familiar with the context of Quine's statement about the inadequacy
> of _a priori_ philosophy, but you seem to be implying that:
>         1.  There is no place for preconceived notions in science
>         2.  I have a preconceived notion (realism), thus the circularity of
> my scientific paradigm

Quine is both a pragmatist and a realist, which I think is an uneasy
pair. Apparently, from my reading, an antirealist nowadays is someone
who denies the reality of a preconceived notion except as a construct
(which may be real in itself) and uses ideas and observational
statements, whether metaphysical or not, as tools to construct theories,
especially for prediction. They need not be true or have some reality to
do this, but may be convenient schemes for summarizing, classifying,
ordering and predicting. See more on this at:
(slow server, may take a while).

>         As a scientist, I know of no other way to approach my subjects
> (dear little nematodes) than that they, and all of their evolutionary
> history exists independently of any of my notions, be they _a priori_ or _a
> posteriori_ of any phylogenetic analysis.

You could assume that your observations are perfectly real, but you do
not have to assume that the patterns and theories you develop based on
your observations have ontological reality "out there." You could
develop the same results without the burden of these assumptions, and
use the results to do predictions. The problem with assuming external
realities based on ideas is that it is so easy to think you have proved
the existence of an external reality (pattern, past relationships, other
non-observational inference) by relying in the consistency of your
method, that it "approaches" reality, that reality lies just beyond
reach waiting to be discovered.

Donald Davidson, a protege' of Quine, asserts that truth-knowledge is
the best interpretation made by a totally independent interpreter of an
alien sentence, based on how the alien uses the sentence in practice.
That's truth to a realist. Now there would be nothing wrong with this
bit of philosophy if cladists did not claim to approach truth based on
consistency of their method and the above angle on truth. Apropos of all
this, in my dictionary under metaphysics, J.S. Mill is quoted thus:
"Metaphysics, that fertile field of delusion propagated by language." Of
course, we all know Mill was wrong is thinking that induction can infer
truth with certainty; so what does he know? (irony :]).

>If "what really happened" during
> the course of nematode evolution can be different for you than it is for
> me, then the sum of all our evolutionary science is incoherent because what
> really happened could conflict -- you could have one tree depicting
> phylogenetic relationships, I another (quite different one), and we could
> both have recovered the *real* tree.

Conflict is normal. What is the test, then? After the usual arguments
about simplicity, consistency, theory, philosophy, etc., it comes down
to whose is most useful in prediction. I could agree with you that there
is a real tree, but that cannot affect its reality.

>As it regards evolutionary science, I
> am convinced that this is ontologically impossible.

And I think is is normal, ontologically irrelevant, and totally
scientific that we deal with conficting results. But then, going back to
your argument that we could have both recovered the *real* tree, I
recover nothing but instead infer a pattern on paper that can be tested
against observational phenomena. Actually the only conflict is in
different patterns on paper.

>I am familiar with
> some of the arguments of philosphical "conventionalism" or "relativism."
> Some of these I find appealing (I must admit to not completely
> understanding all of them).  But to borrow an analogy from Dennett
> (_Darwin's Dangerous Idea_), relativism in evolutionary biology is akin to
> playing tennis with the net down.  In the end, there is only confusion, and
> nobody has any fun.
>         I doubt you are advocating that as systematic biologists we
> approach our subject as relativists.  But you imply an achilles heel of
> realism.  Are you proposing another, more philosphically and operationally
> satisfying paradigm?  Or, do you suggest that there is no room for *any*
> paradigm?

Sure, antirealism/pragmatism/instrumentalism/skepticism/phenomenalism.

>         Also, you seem to imply that if I use statistical, as opposed to
> cladistic methods to recover evolutionary history, I am less likely to
> deceive myself into thinking I have recovered "what really happened" when
> in fact I have erroneously recovered something else (I call this a "type
> III error).  I can interpret this a couple of ways:  Do you mean to say
> that, "like a stick of dynamite, parsimony is a powerful tool when used
> properly, but dangerous when inappropriate"?  Or was your intent more to
> the tune of, "if you're going to eat snow, don't eat yellow snow"?

No. As my past postings state (see Taxacom archives since Sept.) I think
both cladistics and statistical phylogenetics can err in presenting a
single tree as a best approximation. They do not err when the tree is
totally reasonable in cladistics, and have posterior probability greater
than .5 in stat. phylogenetics, but they do err in cladistics when there
is a group of totally reasonable trees that can explain all data with
descent with modification plus some reasonable convergence, and in stat.
phylogenetics when the tree of maximum likelihood has a posterior
probability less than .5. That's the gist. All the remaining
philosophizing has to do with trying to figure out WHY poorly supported
trees are passed off as good approximations or probabilistic "results".

>         If your intentions were of the first scenario, then my
> interpretation is that "parsimony analysis" (especially in the hands of
> someone unfamiliar with its use) can be more dangerous than
> statistical-based phylogenetic inference.

A little knowledge... I agree. Note, however, that the regularity
assumptions of both cladistics and statistical phylogenetics, plus small
morphological data sets and gene data sets based on only 4 characters
means little knowledge for even the most sophisticated analyst. I think
both techniques claim more than they should.

> I can accept this without
> further discussion.  But if your intentions were of the latter persuasion,
> then I find this position a bit harder to swallow (pun intended!).
> byron
> Byron J. Adams
> Department of Plant Pathology
> 406 Plant Sciences Hall
> P.O. Box 830722
> Lincoln, NE 68583-0722
> lab (402) 472 5598
> fax (402) 472-2853
> http://ianrwww.unl.edu/ianr/plntpath/nematode/badams.htm


Richard H. Zander, Buffalo Museum of Science
1020 Humboldt Pkwy, Buffalo, NY 14211 USA bryo at commtech.net

More information about the Taxacom mailing list