Hopefully some resolution (was RE: Feathers are not magical things...)

["Thomas R. Holtz, Jr." <tholtz@geol.umd.edu>]

> From: philidor11 [mailto:philidor11@snet.net]
>
> One reason for the discussion to clarify things
> for me is
> a difference in approach to argument.  You tend to state principles, then
> add modifiers.  I tend to try to combine principle and modifiers into a
> single,
> comprehensive statement.  Assembling all that you say, following
> your way of
> explaining, my questions do get resolved.

Fair enough.  However, none of our time is unlimited here, so although long
on-going open-ended discussions are indeed very effective ways of learning,
they might not serve our interests as well as a few direct questions and
answers.

>
> So, if someone like me, say, were to produce a cladogram, using
> the software
> correctly and making my own choices about how many and which characters
> (given my definitions) and species to include, you would be able
> to evaluate
> that cladogram using information I don't have.  I think you'd be far more
> likely to be able to send my cladogram back to the drawing board than I
> would yours.

Probably (this is assuming, of course, that the taxon in question is one I
am more familiar with than you are).  Experience does certainly help, if
only in so far as being aware of the discovered diversity of anatomy among
the taxa in question.  This is why it is generally a good idea for people
who have direct experience with the fossils (or living organisms) in
question to take up their time conducting the analyses and taxonomies.

> You're saying that the evaluation portion is not necessary to
> regarding the
> hypothesis as 'scientific'.
> If this is correct, then I think I get it.

Hmmmm...  That is stretching what I said a bit, but in principle I would
agree.  HOWEVER (and this is a big HOWEVER), implicit in this is that the
characters you code accurately reflect the data.  (That is to say, that you
haven't entered data that is known at present to be incorrect).

If this cladogram you constructed were based on wholly hypothetical
organisms, and their hypothetical characters, then I would agree with you
completely (although it might not be a useful excerise, except as a learning
experience).

Which actually gets me to a related point, and one which might provide some
resolution: one learns better by doing than by just talking about it!!!  How
about trying to do your own phylogenetic analysis yourself?  This is a
useful thing, something I have students (in at least some classes) do.  Get
used to HOW the technique works, and then come back with (hopefully
specific) questions.

> In response to my formulation that:
> <So, a scientific argument about relationships includes:
> - a number of shared characters judged sufficient to indicate that certain
> animals are related and few enough shared characters in animals
> outside the
> group that these animals can be excluded,>
> you responded:
> <Not exactly.  It is the concordance (or consilience, or whatever similar
> term you'd like) of the characters, not the number.  Certainly a higher
> number of uniquely shared characters helps, but is not necessary (nor, in
> fact, sufficient).>
> Sorry, I don't understand this.

Then I can't help you, unless you go and read some more about this topic
elsewhere.  Anyone else want to join in and see if you can make this
clearer?

(Not one to like to give up, one more try:)
It is not some magic number which is judged sufficient, but rather the
pattern or distribution of the characters which is important.  However, a
greater number of characters consistent with a particular pattern does
suggest better support, just as a greater number of  points along a line on
an X-Y plot might lend better statistical support for that line.

> I've seen many summaries here which claim
> to prove relationship by a long list of shared characters, with no further
> analysis of why the characters are shared.

1) FORGET the "why" question.  Pattern first, scenario second.
2) Remember the postings here are, as you say, "summaries": they are not the
analyses themselves.  Thus, the posters are highlighting one aspect of the
research, but are not presenting the whole of the research project.

As for your (snipped for space by me) comments about character selection: we
have covered that ground many times here on the list, so please go back and
review.


> <(Something distresses me in your above point: it sounds to me
> that you are
> confusing phenetics (which clusters taxa based on some metrics of
> similarities) with cladistics (which clusters based on the simplest
> distribution of derived character states)).>
> I'd assumed that cladistics brings together animals based on similarities,
> then creates a time line (basal/derived) based on differences.  In
> descriptions of the use of software I've seen, there is a prescreening of
> which animals to include.  That prescreening and the later grouping into
> lineages (with 'sister' groups distinguished) both sound like your
> description of phenetics.  These two concepts seem to be conflated.
> How are they distinguished?

Again, some outside reading here would be VERY VERY helpful.  However:
In both modes of analysis, the worker chooses a batch of organisms
(operational taxonomic units) which are the subject of study.  The worker
also selects a number of features which to examine.  Every worker recognizes
that both the above aspects invokes some form of subjectivity on the part of
the scientist; however, unlike traditional methods of systematics these
steps must be made explicitly and are thus subject to counter-evaluation by
other workers.  (In previous methods, it was done "behind the scenes").

Additionally, as you will find in the many previous discussions of character
choice, or in textbooks, or course websites, or whatever, there are some
ways of eliminiting characters from cladistic studies a priori, in order to
shorten search times:
        *Characters unique to a single OTU will reflect the evolutionary 
history of
that taxon after it diverged from all the others, and thus will not help
resolve alternative topologies over the tree as a whole
        *Characters found in every single OTU reflect the common heritage of the
whole group rather than the different possible divergences AMONG the groups,
so they too can be ignored.

> My second hallmark of a scientific argument about relationships was:
> < - expectation that unknown characters will be consistent with others in
> the group, as in feathered velociraptors,>
> and you responded:
> <Yes. I also assume, without additional direct evidence to
> correct me, that
> all dinosaur species had eyes...>
> Ok, I'm learning.
> Still, what made Jura's assertion (that to be 'scientific'ally valid a
> character had to be observed) interesting to me was the fact that the
> 'scientific' method starts with observations and attempts to find a
> principle uniting these observations.  You're extending the use
> of the term
> 'scientific' not just into the hypothesis, but into the inferences drawn
> from the hypothesis.  There is not only a 'scientific' method; there are
> also 'scientific' inferences.

Aha!  Yes!

> The problem for me is that when you apply 'scientific' to unconfirmed
> concepts, it seems you're encouraging the view that science does not
> discover Truth (describe reality), but only generates useful, refutable
> ideas.

Bing!!  (Well, not entirely "bing").  Science does NOT discover Truth;
however, it DOES describe reality in attempts to more closely approximate
what can be observed.  Given that total knowledge is not possible, we can
never know everything.  However, we can use scientific methodology to get
more comprehensive or detailed approximations of reality.

Hence, while none of the following statements are absolutely true, they do
form a series of closer approximations to the reality of the shape of the
Earth:
*The Earth is a flat plane
*The Earth is a cube
*The Earth is a sphere
*The Earth is an oblate spheroid
*The Earth is an oblate spheroid with a number of buldges

> As pointed out in the Disparaging Popper discussion, this
> could turn
> science into a social event.

No, but scientific conferences can be social events... :-)  Note that most
of the referents in the "Disparaging Popper" discussion were from
philosophers of science, and not scientists themselves.  As most of you know
from your own life experiences, DOING something gives you a lot more
information about how something works that just talking about it.

>  That is, the ideas considered and in use may
> be in part the result of the preconceptions of the people holding them.

Yep.  True.  Scientific theories and hypotheses are "in part" the result of
preconceptions.  Anyone who denies that has but to look back at the record.
BUT (and this is the proverbial "Big But") there is a second component to
Science that is lacking in some other disciplines and ways of knowing: that
is, scientific hypotheses and theories *must* be tested against external
realties (i.e., the known phyiscal universe) in ways that other philosophies
and "ways of knowing" are not.  By recognizing and playing off of the
limitations of knowledge, science has been able to uncover many things.

> My third hallmark of a scientific argument about relationships was:
> <- expectation that evolutionary change in the group will be gradual and
> consistent (after all, if a large number of brand new characters suddenly
> appeared in the lineage,  the character similarity screen might fail to
> identify the
> animals as part of the group),>
> To which you responded with principle:
> <NO!  This assumption does not exist within phylogenetic
> methodology (or at
> least morphological phylogenetic methodology).  The acquisition of large
> amounts of new characters within a lineage are not a problem per se: they
> will only add a long branch length to that particular (apomorphic) clade.>
> and modifier:
> <That being said, large scale transformations which "overwrite" a large
> portion of the anatomy may obscure ancestral relationships, which is one
> reason for trying to find new fossil taxa.>
>
> So, if I'm reading you correctly, my statement would be correct
> if and only
> if I make the assumption that our knowledge of the fossil record of a
> particular set of animals is so incomplete that we (or, better, you) could
> be misled by a major change in anatomy.

No, you are not reading me correct.  You said that a gradual and consistent
tempo of evolution was an assumption in phylogenetic methodology.  It is, in
fact, not an assumption.

Punctuated equilibrium states only that most morphological change occurs in
geologically short speciation events, not that these changes involve vast
transformations of anatomy.  (It is common to confuse PE with "hopeful
monster" macromutations, but these are separate models of evolution.
Eldredge and Gould and Stanley and Vrba and company only proposed that most
morphological changes were restricted to short segments of the stratigraphic
range of a species, not that these changes were tremendous profound
alterations of anatomy).  Thus, lineages evolving in PE and in phyletic
gradualistic tempi and modes should be equally resolvable by cladistic
analysis, especially as in all cases you can only sample short time slices
of the record anyway!

Even if there were some major anatomical leaps in a geologically short
period of time, these would not necessarily confuse the analysis, unless:
        *Possibility 1: the anatomical transformations are so profound that they
"overwrite" the anatomical evidence of ancestral characteristics.  Parasites
are notorious for this.
        *Possibility 2: the anatomical transformations are not only profound, 
but
are also convergent onto the morphology of another, distantly related group.
This does lead to the possiblity of false recovery of a clade based on
homoplasies rather than synapomorphies.  If the evolutionary transition from
ancestral to descendant species were so rapid that recovery of the ancestral
or transitional species was unlikely, this could pose a problem.

> Finally,

Whew!

> I asked:
> <Suppose you could establish that PAUP gave a lower weight to punctuated
> equilibrium scenarios than to more gradualist models.  Would that
> invalidate
> the algorithm?  I realize that this must have occurred to other
> people; I'm
> asking because I can't solve it myself.>
> and you replied:
> <Interesting thought.  In the particular case in question, I
> think the whole
> situation is flip-flopped.  I think the failure of resolution of the PE vs
> PG question (which dominated a LOT of my classes as an undergraduate and
> graduate, far more so than systematics) is that to effectively test models
> of these two questions you need well-supported phylogenetic trees.  Even
> better, to test some of the particular models you would need to
> account for
> *all* the taxa within these trees.  Unfortunately, relatively few
> clades of
> fossil invertebrates have been subjected to explicit phylogenetic analysis
> at present, although some groups are getting addressed.  Within ten to
> twenty years, I expect that enough of the relevant speciose and
> individually-abundant taxa will be sorted out so that we can go back and
> address many of the questions raised by Eldredge and Gould and Stanley and
> company.  (Okay, some sooner than others).>

> The first question that comes to mind is how reliable the current
> phylogenetic trees should be considered while waiting for these
> investigations and analyses to be conducted.  I get the
> impression that the
> software as currently constituted may be more amenable to use
> with one model
> rather than another, though I may well be inferring too much.

I believe you are inferring too much.  My thoughts on the subject: use the
trees currently available, then test again later if you think new trees
might change your results.  You get more publications that way... :-)

> I also note the use of invertebrate results as a check on vertebrate
> analyses.

Yep, for a number of reasons:
1) Go back to the literature of punctuated equilibrium.  What are the test
cases?  Invertebrates for the most part.  If you are going to use
phylogenetic analyses to reciprocally illuminate past studies of PE, then
one will have to use the organisms used in those studies.
2) There was a very good reason for #1: sample size!!!  PE vs Phyletic
Gradualism requires some knowledge of the change in population distributions
through time, which requries statistically useful sample sizes at each
interval, over a statistically useful number of time intervals.  Almost no
vertebrate lineage has sufficient number of specimens known which satisfies
these requirements.
3) Similar to #2, there are a heckuva lot more species of invertebrates than
vertebrates!!  If one is attempting to find some general pattern of the
evolution of organisms through time, using examples from one really weird
bunch of giant highly energetic critters with an internal skeleton and a
number of unique tissue types might not be very appropriate.

> Is there a way of confirming the appropriateness of this
> assumption?  Do changes in snail shells over time parallel
> changes in bone?

Damn good question (and one I used to ask in class a lot).  How to answer
it?  Well, you have to do the PE tests on some invertebrates, and on some
vertebrates, and compare the results.  BUT, in order to actually test what
PE suggests, we'll kneed to know the phylogenetic relationships among the
test speces.  And in order to do that, we need people to do those tests...

                Thomas R. Holtz, Jr.
                Vertebrate Paleontologist
Department of Geology           Director, Earth, Life & Time Program
University of Maryland          College Park Scholars
                College Park, MD  20742
http://www.geol.umd.edu/~tholtz/tholtz.htm
http://www.geol.umd.edu/~jmerck/eltsite
Phone:  301-405-4084    Email:  tholtz@geol.umd.edu
Fax (Geol):  301-314-9661       Fax (CPS-ELT): 301-405-0796