[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]


George Olshevsky wrote:
>When you start looking at these so-called "reversals" in detail [...]
>you see that these turn out not to be reversals at all: Each
>lineage loses its flying ability in a different way. No two such "reversals"
>are exactly alike,[...] In fact, these so-called "reversals" are actually
        For the purposes of the general public I feel it is important to
make known the assumptions and definitions under which George makes his
statements about the supposed rarity and unacceptability of reversals. Given
this knowledge, everyone can make their own judgments about the
appropriateness of his claims.
        We are apparently dealing with different definitions of "reversal".
The definition George uses is specifically when a character returns to the
*exact* primitive state. The definition cladists use is when a taxon evolves
such that it must be scored with the primitive state.
        As George has pointed out, the likelihood of returning to exactly
the same state as an ancestor by random morphological chance is extremely
low to impossible. I have noted in response that evolution is not entirely
random, old genetic code may be resucitated, increasing the chance of
returning to a primitive state. However, I did not understand that George
meant the *exact* same state. This appears to be impossible under the terms
George describes.
        George notes later in his post, using his random model of
morphologic change, that it is more likely that a character will return to a
morphology similar to its original morphology (i.e. cabaple of serving the
same function). This is especially true if we observe that similar selection
pressures will limit the limit the range of other possible morphologies
available to evoltion (ala convergence). So it may also be with reversals
(in the cladistic sense).
        However, as George has pointed out, characters are rarely
cut-and-dried binary features. If we are to specify that the character
returns to the ancestral state, we must first identify exactly what that
state is. For example, in the case of therizinosaur pedes, is it the pes of
_Eoraptor_? Or _Herrerasaurus_? or something we haven't found yet? In the
flight/flightlessness case, is it advanced arctomet curorality,
scansorality, rapid dashing cursorality, or something else? Character states
do change over time, and identifying which one is the "true" ancestral
condition seems problematic at best.
        What is the exact morphology of a character anyway? In the
morphospaces George discusses, a character CANNOT be a point. A character
state is a cloud consisting of plots of the morphologies of all of the
members of the species in question. If there is no identity of form within
individuals, how can there be identity between species?
        Even in the case of so-called "genetic reversals", where the genetic
code of the ancestral state has been partially or wholly reused to generate
the new form, we would hardly expect a complete return to the "original"
form of the character. The new taxon, the one featuring the reversal, is a
different animal, with a different starting position in overall morphospace,
and is under different selection pressures. "Genetic" reversals are
adaptations like any other, although they may (and I am not expert here)
have a different balance of advantages and disadvantages relative to de novo
character state transformations.
        Characters and character states are classifications (see below) of
morphology. As such, we design them to be meaningful. We must not, however,
try to invoke meaning which is not there. An animal does not have an
arctometatarsus. Sure you can say that we all recognize that the femur of
_Ophicodon_ and _Homo_ are different. But how much difference is there? In
an analysis of all life, both of these taxa have the same character state:
"Presence of humerus: yup." We try to break the difference down as much as
we can to test phylogeny and homology. And test we must, for that is the
only way we will know if character states are homologous or not. But lets
not try to ascribe too much meaning to our characters. A "four-toed pes" or
"flight" can encompass a wide wide range of morphology or behavior, and
realtive similarity may not necessarily indicate homology.
        George has further pointed out that determining whether a reveral
(sensu cladistica) is "genetic" or "convergent" may be impossible. I agree
strongly with this assertion. George says that some "reversals" are actually
convergences. This results from his acceptance of either the definition
given above, or the definition of a "genetic reversal" as the true defintion
of reversal. This is indeed why the "cladistic" definition of reversal is an
operational defintion, not an explicit invocation of an evolutionary
process. We cannot know always what processes were afoot (sorry) in the
re-evolution of the four-toed therizinosaur pes. Indeed, our catagorization
of these processes seems to be largely a construct of our imagination rather
than recognition of real phenomena. Evolution doesn't "care" whether you
call it a convergence or a reversal, all it cares is that the animal is
shaped to fit the environment.
        Given these considerations, it does indeed seem that reversals sensu
George, involving an exact return to the original state, are very unlikely
to impossible (assuming an exact state can be identified). However,
"cladistic" reversals (where a character must be scored with the ancestral
state, even though an ancestor had the derived state) do appear in analyses
with some frequency. Please do not allow the difference in definitions to
sway your judgement. When I say the pes of therizinosaurs has exhibits a
reversal, all I mean is that, for some reason, they have a pedal morphology
similar enough to their ancestors that we call it the same character state,
but they did not inherit this morphology from more immediate ancestors.
George might call this "convergence" or a "reversal" or whatever. What he
has said here and in other posts suggests that he should have no objection
to this possibility.

        Here's my oppinion, informed or otherwise, on this entire mess:
        Characters are classes, artificial groupings of morphology which we
invent through observation and use as discrete quanta of morphology in our
analysis. Nature does not accept uniformity of morphology in her creations,
as we all know too well. Characters are never simple, never wholly
straightforward, and do not always stand still for you to look at them. As a
professor of mine likes to say, "all discrete characters are continuous
characters in disguise." While I do not necessarily accept this statement in
the spirit it is intended, the point is clear. We cannot ascribe some
supernatural existance to characters. We cannot say they "exist", that there
is an ideal form of a character or a single point in morphospace.
        Similarly, we cannot inherently tell if characters are homologous or
not. There is no key to homology, no a means of rejecting it outright a
priori of phylogenetic analysis, although there can be independant lines of
evidence supporting or refuting it (e.g. embryology). The only way to *know*
that a character state is homologous is to observe the evolution of the
higher taxon within which it is expressed. The most straightforward means of
testing homology is using a phylogenetic hypothesis.
        While homology derives directly from the process of ancestry and
descent, homoplaisy does not (excepting, perhaps, the elusive "genetic
reversal"). As such, I am not sure that homoplasy is a "real" phenomenon.
Homoplasy may simply be a consequence of our classification of characters.
All homolasy is is the presence of a character which must be scored with the
same state in different taxa which did not inherit that state from a common
ancestor. Whether you call this a reversal, a convergence, parallelism, or a
hoo-hoo-dilly is a matter of convention. Homoplasy does not in and of itself
imply a process, nor is it clear to me that the "processes" of, say,
convergent evolution are anything more than a classification in the first
place. The cladistic convention on the use of the term reversal is an
acceptance of the ambiguity.


    Jonathan R. Wagner, Dept. of Geosciences, TTU, Lubbock, TX 79409-1053
                    "...To fight legends." - Kosh Naranek