Darwin (was: Phylogenetic evidence)

[pierre deleporte]

A 10:15 25/07/2002 +0200, Zdenek Skala wrote :

>From: pierre deleporte:
> >Monophyletic taxa in strict cladistic classifications are effectively
> >classes, if viewed as sets of objects (individuals) sharing one or several
> >properties. These properties are the synapomorphies shared by the sets of
> >individual organisms composing the monophyletic taxa.
>
>I must disagree (despite I like most points that Pierre made). The clade
>definition does not exclude character-state reversals, so there can exist
>clades (and do exist - usually the large ones) whose members do not share
>any single character state in common.

But they must share characters in common, unless they would not be
identified as clades.

>Consequently, the clades are not classes but objects defined by a
>distance/similarity measure like clusters in cluster analysis.

Clades are classes if you consider reversals as apomorphies, what cladists
do as a rule. They classify organisms AFTER phylogenetic analysis, thus
'reversed' absence of characters, as well as "reversed" states similar to
plesiomorphic ones, are identified as apomorphies through optimizing
character evolutionary scenarios on the cladogram (which is the cladistic
logic for infreing the best fitting cladogram, and reading the results
"back to character scenarios").
Classes based on these nested sets of synapomorphies, interpreted in the
light of the cladogram, including possible apomorphic absence - reversals,
are thus classes of equivalence, not similarity clusters.
In my view, you can't reason for phylogenetic systematics by considering
"face value" character states in the data matrix. This is effectively what
cluster analysis does classically. But phylogenetic systematics consists in
considering a cladogram, taken as a reliable phylogeny, and accepting all
the logical consequences (including reversals to absence). Reversals to
absence are already identified as such apomorphic character states BEFORE
the phylogenetic systematic system is settled (= naming taxa), with taxa =
clades = fully consistent classes defined by equivalent nested synapomorphies.

>More importantly and regarding our months old discussion with Pierre:
>I am not sure if our agreement on the additional criterion in the eclectic
>classifications was really complete.

Not yet. Just wait a bit  :-)

>Most of the discussion boiled down to a rather simple point:
>having a cladogram with: taxa A,B,C,D,E

I presume, a pectinate cladogram ((((A,B)C)D)E)

>and "significant gaps" B|C and D|E, a cladist will probably name the taxa
>(((AB)CD)E).

That is (ABCDE), (ABCD), and AB.

>Now, what will do an eclecticist? Will (s)he name the taxa (((AB)(CD))(E))
>or ((AB)((CD)(E)))? The point is that there should be a criterion of
>quantity of "gap significance" (be it number of character-state changes or
>whatever) that will decide which taxon is more homogeneous: (ABCD) or
>(CDE). This is viewed (by Pierre, at least) as that "additional criterion"
>deserved by eclectic classifications. However (and here I still feel a
>disagreement), I would argue that even cladistic taxa splitting will
>deserve some measure of gap significance (say "S") - at least to decide at
>which level will be the clades named

Yes, but to decide only that.
Now, the eclecticist will not only decide when monophyletic clades are
worth naming (on the same basis of "S"), but also when paraphyletic clades
are worth naming rather than monophyletic ones. The eclecticist has to
decide when a gap is "significant" for paraphyly delineation, and when it
is not, while this is not a problem for a strict cladist. For the strict
cladist, "all significant gaps" will be worth naming the monophyletic group
they support, while the eclecticist will need a supplementary criterion:
among all significant gaps, what are the "more significant ones" worth
delineating a paraphyletic group "downward" (and not only supporting the
corresponding monophyletic group "upward"). Which gaps act as a support for
a group, and which ones act as both a support and a "stop" to a
paraphyletic group.

>(e.g. at 5 character-state changes (a wild example)). Moreover, it is
>highly improbable that it will be possible to apply the same significance
>level ("5 changes") over all the cladogram. Hence, the process will be
>quite similar to the ecleclectic one: split the cladogram at the level of
>5 changes, then split the pieces at the level of 3 changes etc.

Yes, but only for naming monophyletic groups "upward" all the time, thus
splitting the cladogram at the "5-to-3 gap level" throughout. The
difference between a 5-gap and a 3-gap has strictly no consequence for
naming in cladistics. They simply are, or are not, significant ... (in the
real world, qualitative notions of usage, in fact "changes in characters of
interest" are involved in the decision of naming, and not such a systematic
notion of "quantitative gaps", but I'm playing the game of simple
quantitative gap-coding for the clarity of the discussion).

>Anyway, just the concept of having some measure of gap significance (S)
>enables to make an ordering of gaps in the lines of
>S(B|C)>S(D|E)>S(A|B)>...etc. This is sufficient and necessary for both,
>cladists and eclecticists.

I simply don't think the sequence (hierarchy of gaps) is necessary for
cladists. It's only "gap significant or not", and thus "monophyletic group
named or not named". As for hierarchy (for using a hierarchy of formal
"ranks", for instance : species in genera in families...), the historical
hierarchy (nested clades) seems pretty sufficient for cladists. Am I
missing something here ?

>  Hence, I must repeat, I see no difference between cladistic and eclectic
> classification in this respect.

Let's think it over again... What does a cladist care of a hierarchy of
gaps, beyond "significant or not", and thus "clade worth naming or not" ?

My intuitive basis for reasoning is that only the eclecticist needs a
"stop" for a paraphyletic group, and thus a criterion for this. What is
this criterion ? If it is "significant gap larger than other significant
gaps", OK, but what does the cladist cares for this ?
If it is a question of naming "families" in sister clades, and then
"genera" inside these families, I can see the point, this may be a common
preoccupation based on common criteria (some way: sister clades will not
alway provide the cladist with symmetric gaps of same amplitude, and sister
clades will be named at the same rank, even if one of the gaps is "not
significant"...), but this has still nothing to do with "putting a stop or
not" to a paraphyletic group, which requires a criterion of its own.
Without such an additional criterion, eclecticists would logically name
monophyletic clades only, which would clearly be detrimental to their
reputation.
It seems that Cladists don't care the "hierarchy of gaps" in this respect,
much less phylocoders (no ranks). Gap / no gap seems quite sufficient for them.

(Just a hint: when considering the Kinman system, it seems to me that some
specific criteria involved in the decision of naming a paraphyletic group
include tradition (hence nomenclatural stability), and possibly length of a
pectinate series (A(B(C(D.......(X(Y(....omega)))))))) which would suggest
to splitting it into more handable smaller pieces, and limit the hierarchy
of ranks, the more for frequently studied groups with a tradition of
paraphyletic naming. These are examples of possible extra criteria
relatively to simple "signifiance of gap", which could make no difference
between the two methods by itself).

Best and cheers,
Pierre

>pierre:
> >As for the arbitrariness involved in naming not all clades in a cladistic
> >classification, and the arbitrariness of naming not all monophyletic and
> not all >paraphyletic groups in an eclectic classification, the two approaches
> >share the arbitrariness of deciding which gaps are relevant for naming
> >clades, but the second involves the supplementary arbitrariness of deciding
> >when gaps also vail for delineating paraphyly rather than monophyly only
> >(when to "look down the tree" and not only up the tree).
> >Note that I think I could reach "off list" a common agreement with Zdenek
> >Skala on the last point...
>
>Best!
>Zdenek Skala