Positivism vs Realism

Tom DiBenedetto tdib at UMICH.EDU
Sun Dec 14 13:02:33 CST 1997

 James Francis Lyons-Weiler wrote:

>       Hypotheses of homology that have failed the test, according
>       to you, have risen from the ashes as hypotheses of convergence.

Yes. The original hypothesis, growing out of the sum of our
biological work, is that we have a set of homology/monophyly
statements encoded in our data. These speak, by their very nature, to
the relationships of the same taxa. With a perspective which
recognizes the coherence of the organism, and posits a single history
of life, we combine the monophyly statements with an expectation of
congruence. Those hypotheses that fail the test must be explained as
convergence (or reversal etc.). Since these explanations are
themselves hypothetical, we can refer to them as hypotheses of
convergence. I truly am at a loss to understand your seeming surprise
at this.

>       If you are saying that hypotheses of convergence that
>       are retained as result of a parsimony analysis are not
>       corroborated as hypotheses of convergence by the parsimony "test",

huh? We do not begin with any hypotheses of convergence. They are a
result we are forced to accept when we encounter incongruence.

>       then you must also be saying that hypotheses of homology that
>       are retained are also not corroborated.

huh? Since your premise is wrong, the conclusion is meaningless.

>       If parsimony is not
>       test in the sense that convergences are corrboraborated,

There are no convergences to be corroborated before the test is run.

>       then non-convergences (informative synapomorphies) can't
>       be corroborated either.


>       Popper and Miller's proofs in the early '80s show clearly
>       that when a test only serves to restate the obvious, (1)
>       no corroboration results,

and you think that a specific phylogenetic hypothesis is ever

>        (2) the test was not a critical test,

given the number that are falsified, how can you claim the test isnt

>       (3) time is better spent either thinking of new and
>       bold hypotheses or devising more critical tests that
>       surpass the no longer salient ones in their revelations.

so let me hear your bold hypothesis for the data of character

>> As I said above,,,the "amount of surprise" that Popper was referring
>> to was not a personal psychological state, it was a casual way of
>> making refernce to the relationship bvetween the hypothesis and
>> background knowledge.

>       This straw man is obvious - no psychologism is
>       needed.  When I said "surprise" and background
>       knowledge, I meant it in the same meaning as in
>       Popper's model of epistemology.  Interestingly,
>       he was charged with requiring a subjective knower,
>       which he vehemently denied.

C'mon James, thats no answer. I meant the strawman as a bit of a prod
to get you to reflect on what "amount of surprise" was referring to
(although I wasnt completely convinced that you knew it wasnt a
psycological standard). To simply say that you mean it the way Popper
did isnt very helpful. So do I. And we seem to understand it

>         We need only use tests that are set up to tell
>       us if and when a significant level of congruence exists.
>       Parsimony doesn't address this.

But that is another issue Your significance test is not going to
yield a set of relationships. That is what we are after in
systematics, y'know...

>       I won't go on and on about it, but the tests I have
>       devised require no modelling or character weighting.
>       They are designed to to what parsimony should be doing,
>       but doesn't.

It seems to me they are doing something rather different.
>> When something is labeled homoplasy and we are forced to accept
>> an ad hoc explanation for its presence in the data, we are
>> acknowledging that it has failed to be corroborated by out test - it
>> is a falsification. How can this be seen as dismissing Popperian
>> corroboration?????

>       Because if it to be dismissed as a hypothesis of homology,
>       it must be a hypothesis of convergence.

???? If data cannot be explained by the tested hypothesis, then we
need a new (smaller scale) explanation to account for it. This is not
in any way non-Popperian.

>         If your pefectly
>       objective system is really Popperian, then each hypothesis
>       to which the data pertain receive the same dose of
>       corroboration, whether we are interested in them (the
>       hypotheses we might fail to recognize) or not.

???? All the hypotheses that are tested are hypotheses of monophyly.
Some are corroborated, some are not (because it really is a critical

JAMES said this:
>> >        i.e., it is applied to the
>> >        hypotheses of congruence homologies, but it is not allocated
>> >        to the emergent hypotheses of discordant homologies, and
>> >        yet they represent the archetype of bold hypotheses because
>> >        they were unexpected.

Tom said this:
>> An ad hoc hypothesis is the archtype of bold hypotheis......buddy you
>> are getting lost in space now!

>       I don't think so.  Popper defines what he means as an
>       ad-hoc hypoothesis in his famous paper on the problem of
>       demarcation:
>       "I call a conjecture 'ad hoc' is it is introduced... to
>       explain a particular difficulty, but if... it cannot
>       be tested independently".

Perfectly consistent with what Tom has been saying!

>       Popper is consistent in his use of the term, and
>       never equates a bold hypothesis with an ad-hoc one;

Yes James, that is what I have been saying. It is you who said that
homoplasy (ad hoc) is bold because it is unexpected. That is why I
accused you of being lost in space. See above.

>         At the same, bold hypotheses need not be
>       testable - no definition of boldness is tied to
>       falsifiability.

Well this is completely wrong, and it might explain your problems
understanding Popper. Boldness is directly and fundamentally tied to
testability and falsifiability in Popper's system - it is in fact
part of his main point! I havent the time to track down a reference
to the complete explanation, but here are a few quotes you can piece
together to see this:
        "a high probability is a dubious reward for saying very
little, or nothing. In other words, the rule 'Obtain high
probabilities' puts a premium on ad hoc hypotheses. All this presents
a most uninspiring picture of science- a picture, moreover, that does
not in the least resemble the original. Indeed, what makes the
original so inspiring is its boldness; its boldly conceived
hypotheses, boldly submitted to every kind of criticism - to every
refutaton we can think of, including the most severe tests....."
        "Since testability in its turn can be measured by the content
of the theory, and since content, in its turn, can be measured by the
absolute logical improbability of the theory, content and
improbability stand in the same close relation to degree of
corroboration as does testability itself."

These two quotes show the link between logically probable
statements,ad hoc statements, and a lack of empirical content on the
one hand, and boldness, content, logical improbability, and
testability on the other. A few more snippets:

        "..in order that [the hypothesis] should be maximally
corroborable, its content or degree of testability must be maximal"
        "...corroborability equals testability and empirical
        "Why [do] scientists invaribly prefer a highly testable
theory whose content goes far beyond all observed evidence to an ad
hoc hypothesis, designed to explain just this evidence, and little
beyond it, even though the latter must be more probable than the
former on any given evidence"

Boldness clearly relates to hypotheses which are improbable on the
background knowledge alone; they are the opposite of ad hoc
hypotheses which say very little. And this is directly tied to
testability. A careful consideration of the formal definition of the
degree of corroboration, C(h,e,b)= p(e,hb)-p(e,b) should enable you
to see this. (The denominator is merely a normalizing factor)
This can be related to parsimony as follows:

        "It can be shown that what is usually called the simplicity
of a theory is associated with its logical improbability.....This
indeed, allows us top deduce,....why it is always advantageous to try
the simplest theories first. They are those which offer us the best
chance to submit them to severe tests; the simpler theory has always
a higher degree of testability than the more complicated one."

Cladistics entails a bold conjecture, a simple theory, maximally
testable (all of our homology hypotheses are valid statements of
evolutionary relationships). Parsimony is the implementation of the
test of congruence; a test directly deducable from the notions of
"organism" and descent. Alternative explanations for parts of the
data are taken up as necessary in light of the results of the test. I
dont see how this research program can be ignored unless you are
willing to deny the notion of descent as a factor which can order
homologies in a hierarchial manner.

>       Some bold hypotheses can be ad-hoc,

No. Ad hoc hypotheses are inherintly unbold. They account for
specific phenomena and say nothing beyond that. Boldness refers to
the scope of explanation that a hypothesis attempts.

>        and vice
>       versa, but a hypotheses need not be ad-hoc to be bold,

huh? I have consistently said the opposite.
Bold hypotheses are bold because they have high information
content. Ad hocs have low information content.

>         It is clear that the only useful
>       bold hypotheses are the testable ones, but that's
>       only because they are hypotheses, and this is true of
>       any hypothesis...

No, boldness and testability are inherintly linked.

re, cladograms
>       It is far better and more accurate to say that they are
>       hypotheses - to be tested in the future by independent
>       means - than to say that they are phylogenies, which
>       is downright misleading.

Misleading to who? They are phylogenetic hypotheses. OK?

re. PAUP*
 >The *, by the way, means (* and other methods).

Yes I know, I always thought he should call it PAUAAA - Phylogenetic
Analysis Using Anything At All.

>        the "test" you rely on
>       will pass any data matrix with variable states -
>       with or without a remarkable degree of congruence
>       (remarkable in the Popperian sense).

it will report the degree of corroboration of the hypotheses. You are
intent on making a judgement as to whether that degree is
significant. That is a different question. Whether it is or isnt
significant by your criterea, the results are not to be ignored in
favor of a less corroborated hypothesis.

>> I dont understand this. You think you can detect (e.g) convergence
>> without a tree?

>       YES.

>> If you can,,,,,and you can do it convincingly, then I wont propose it
>> as a homology. Such concerns are addressed in the character defintion
>> phase. Whats the problem?

>       Great!  I'll send reprints and other materials.

Fine, but I hope you realize that systematists (especially
morphologists) have been doing  this all along. That is why we study
organisms so hard.

Tom DiBenedetto                 http://www-personal.umich.edu/~tdib/
Fish Division                                   tdib at umich.edu
University of Michigan Museum of Zoology

More information about the Taxacom mailing list