Traits and/or states
rjensen at SAINTMARYS.EDU
rjensen at SAINTMARYS.EDU
Sat Nov 1 09:24:37 CST 2003
One thing is obvious - this discussion is hindered by semantic, philosophical,
and linguistic differences. The reason responders keep placing words in
quotes is that there is no single definition of these words that we all agree
on. As Tom Lammers noted, what he calls a character, and its subsequent
division into (I presume; knowing Tom, I think I'm right on this) mutually
exclusive states is a formulation that we all recognize.
That said, let me ask what should be a simple question: suppose my
observations of my favorite taxon lead to the singular conclusion that
structure X exists in two, and only two, mutually exclusive forms (call them
X1 and X2). The homology of the two forms is not in question. In fact, I know
from careful developmental and genetic studies that structure X is solely
under genetic control and that the two genotypes are completely reproductively
isolated. Given that the two genomes share a multitude of other identical
structural features that I know are (by virtue of my detailed genetic studies)
inherited independently of structure X, the observed appearance of structure X
provides an umambiguous means by which I can unfailingly assign an organism to
one of the two groups: X1 and X2.
I my mind, structure X is a character that exists in two states. When I code
my matrix for all of the organisms in my collection, I have two choices.
First, treat the mutually exclusive appearances of structure X as one
character with two states; the cells of one column (or row, depending on your
preferred orientation) are coded 1 or 2, depending on the observed appearance;
second, treat the two mutually exclusive appearances of structure X as two
separate characters; for one column, I will enter 1 or "not 1", the second
column will have 2 or "not 2".
My question is (I know my answer, but I would like to know how others
respond), what is the difference in these two coding schemes and why would one
be considered more (biologically, philosophically, phylogenetically, etc.)
correct than the other?
Quoting Gianluca Polgar <polgar at ALFANET.IT>:
> I would like to express some ideas on the character/states concepts.
> 1) When we speak of phylogenetic inference, we should speak of inheritance,
> i.e. of vertical transmission of genes.
> This is a first hypothesis.
> We know that widespread and iterative events of horizontal transmission
> (i.e. via retrotransposons or transposable elements in eukariotes) lead to
> impressive evolutionary changes. Maybe the same explosive radiation of
> eukaryotes and their extraordinary evolutionary versatility with respect to
> the prokaryotes, could take place only due to continuous and iterated
> genetic duplicative mechanisms mediated by transposition or
> 2) But even assuming that horizontal transmission could be treated as a
> background noise at the scale of microevolution, speaking of inheritance
> when observing a phenotype (i.e. a morphological structure), implies a
> biunivocal correlation between shared phenotypic similarities and shared
> genetic traits.
> This is a second, strong hypothesis.
> We know that the morphological structure and related function of an
> organism is the result of the continuous interaction between the organism
> itself and its "environment", during embriogenesis and ontogenesis
> (reaction norms, ecotypes...). Shared similarities could only show a
> similar "interactive history" of the organism.
> 3) But what is a "character" and its "states" from the point of view of
> The only reasonable explanation for these concepts that I find, taking into
> consideration points 1) and 2), is that the "character" is an abstraction:
> it is the hypothetical "type of stable structural plan" which is the
> ecological and genetic stable interaction that controls the differentiation
> and development of a complex morphological structure through generations.
> (Note that intended so, the "character" is then linked not only to genetic
> flux, but also to the environmental conditions in which this flux have
> occurred through generations... but let's follow the second hypothesis).
> Thus the structural genetic plan could be the "high-level construct"...
> From this point of view then, states would be the expression of a
> particular structural plan, or the interaction between particular alleles
> (i.e. genetic variants).
> (Note that this is an oversimplification: for each allele, SEVERAL reaction
> norms will be possible in different environmental conditions...)
> But how could we infer that two particular morphological features of two
> different species (here more problems arise, but we'd like to arrest the
> haemorrhage...) which apparently share similarities... ARE actually the
> outcome of a common (in the sense of a recent common ancestry) "structural
> genetic plan"?
> We may study their anatomy, physiology, autoecology, eco-ethology (which
> most often we don't), correlate their observed characteristics with their
> observed functions inside and outside of the organism, and deduce their
> homology, i.e. propose a parsimonious model of their evolution. This is how
> for instance, comparate anatomy seems to work.
> We thus recognise in that observed, complex, reasoned cluster of
> morphological details a "structure" and a "function": an identity of its
> (In an ideal condition, we would analyse the genetic control of the
> expression and differentiation of that particular phenotypic feature, and
> then compare the genetic distance on the basis of targeted markers...)
> Thus we may describe a "morphological trait", assuming that its
> characteristics correspond to an inheritable "structural plan", and we may
> call it a "character".
> This is a third, very strong hypothesis.
> Up to this point, seeing that in most cases taxonomists never have the
> possibility to go through the long and complex experimental and theoretical
> processes needed to support the mentioned hypotheses, isn't it more honest
> and scientifically correct to limit the phylogenetic inference to the
> subject/predicate constructs directly derived from observations? Nothing
> more, nothing less of that: a phylogenetic inference based on shared
> OBSERVED similarities... The subjective (and psychological) arbitrariness
> of this process can be actually limited by a number of considerations, and
> the taxonomist will with no doubt reasonably explain his choice of the
> characters and of the states... that's all.
> Gianluca Polgar
> c/o Prof L. Bullini
> Dip. di Gen. e Biol. Mol. "Charles Darwin"
> Univ. of Roma La Sapienza
> Rome Italy
> >At 10:42 AM 10/31/03 +0100, you wrote:
> >>The character/state distinction seems to be (to me, at least) reasonable
> >>not only from the purely practical point of view.
> >>What we call 'character' and 'state' in the phylogenetic analysis are
> >>indeed both high-level constructs, not observations (not speaking about
> >>all the problems with the process of "observation"). The character/state
> >>distinction is a way how to hierarchize such constructs; we can treat this
> >>hierarchic relation as a subject/predicate, variable/value or whatever but
> >>the language we are using can hardly change the merit. I have personally
> >>found nothing in your message (nor in your presentation referenced below)
> >>what would contradict this generally accepted opinion. Perhaps you could
> >>explain your position in more detail?
> >My thanks to Drs. Skala and Deleporte for their thoughtful comments. I will
> >try to respond here to both.
> >I'm not entirely sure I understand what is meant by "high-level
> >constructs." If descriptions of organisms, or the cells of a data matrix,
> >do not represent our observations, then it would be difficult to claim that
> >a phylogenetic hypothesis is an explanation of shared similarities.
> >The concern I have had with the "character" and "state" distinction is that
> >it is not consistent with the nature of observation. What we observe are
> >objects by way of their properties (= character, attribute, trait), and
> >these are communicated by subject-predicate relations. For instance, I
> >cannot observe "number of digits." But I can observe that the distal ends
> >of some set of appendages (subject) have five digits (predicate). The
> >"character," "number of digits," and the "state," "five," cannot represent
> >subject-predicate relations denoting observations since "number of digits"
> >does not represent the subject observed. What are observed are subjects
> >called appendages with the property (= character, attribute, trait) of five
> >digits. The cell of a data matrix for species X must code the observation
> >that individuals in this class have appendages with five digits. The cell
> >is not simply a "state;" the cell summarizes observations.
> >I don't find any hierarchic relation here, and such a relation would not be
> >expected for objects and their properties. On the other hand, an object can
> >be composed of subsidiary objects, in which case hierarchic relations can
> >be stated. But, this hierarchic relation of a part composed of parts is not
> >equivalent to the observations which allow us to speak of the relation
> >I hope this helps clarify what I said earlier.
> >J. Kirk Fitzhugh, Ph.D.
> >Associate Curator of Polychaetes
> >Invertebrate Zoology Section
> >Research & Collections Branch
> >Los Angeles County Museum of Natural History
> >900 Exposition Blvd
> >Los Angeles CA 90007
> >Phone: 213-763-3233
> >FAX: 213-746-2999
> >e-mail: kfitzhug at nhm.org
More information about the Taxacom