Publishing on human origins

[David Orlovich]

> If this were the case for molecular data then I would not have an
> argument
> about molecular characters not being cladistic. Please provide
> reference to
> a published example as this is not an approach used in hominoid
> molecular
> systematics so far as I can see.
>

Dear John and others.

The above comment really surprised me. I haven't read a paper about
hominoid systematics, so I can't cite published examples from that
field. However, the a priori designation of polarity (and homology) are
what I have come to understand is what outgroup rooting and cladistics
are all about, so when I read that this approach is not used in
hominoid systematics, I am at a loss to understand what these authors
must mean if they use the word cladistics.

 From my reading, I understand that in the 'old days', people used to
look at a bunch of character states and say "this is a primitive
character state, and this other one is derived". They could then
construct a freehand tree with taxa having primitive character states
branching near the bottom and those with derived character states
diverging later. This is quite problematic as it is at the whim of what
the worker took to be primitive and derived. Anyway, let's construct an
unrooted tree by any method (lets say by finding the tree on to which
the character state changes fit with the least number of steps - this
is what I call parsimony, be they molecular or morphological
characters). So far, I haven't had to make any decision about polarity
of characters, I have simply fit character state changes on to a tree.
These changes can go in either direction. Now, I want to root the tree,
say using the outgroup method. Once the tree is rooted, it gets an
'evolutionary direction'. I am then polarising all the character states
on the tree - and I haven't had to make any explicit decisions about
polarity of characters, molecular or morphological. Occasionally this
is problematic when the outgroup is too distantly related to the
ingroup and the position of attachment to the tree is equivocal, but
this simply requires working around with other outgroups etc. What I
have described is what I call a cladistic analysis, and at no time did
I force polarity on any characters until I actually rooted the tree,
and rooting the tree didn't require me to make explicit decisions about
polarity, they happen 'automagically' when the tree is rooted.

In a sense what I have just described applies to homology as well. Let
me describe a less-than-perfect scenario. We aim to compare like with
like in constructing a character-state matrix, but with morphological
characters at least, we have to rely on criteria (Remanes criteria I
think they're called) to help us decide which character states are
actually states of the same character. We hope that we might be
choosing homologous character states in our matrix, and in case we make
a few mistakes, we choose lots and lots of characters so we can
minimise any misinterpretations (i.e. homoplasy). Once we have
constructed our character-state matrix, we can try to fit those
characters on to a tree. If we again use the criterion of parsominy, we
try to fit those character state changes on to a tree in the shortest
number of steps. After we root the tree, we have done again what I call
a cladistic analysis. From this tree, we can see that some character
states evolved once (these are interpreted to be homologous) and other
character states evolved more than once (these are interpreted to be
homoplasy). However, this might sound problematic as it is a circular
argument. On the one hand we tried to chose homologous character states
before the analysis, and on the other hand we are using the results of
our cladistic analysis to determine which characters are homologous. I
prefer to think of the character state matrix at the start of an
analysis as a hypothesis of homology, and a test of that hypothesis of
homology is where those character states came out on the tree. This can
be done with molecular or morphological characters. I see one clear
advantage with DNA sequence data - that the prior designation of
homology is done by alignment algorithms and (when done by hand) by the
conservative minimisation of character state changes. This removes some
of the subjectivity that arises when using Remane's criteria for
morphological characters. It's not perfect of course, but it can be
less problematic because of fewer ambiguities with interpreting
character states.

OK, so I've just outlined what I think cladistics is. Thinking about
groups of taxa as monophyletic, paraphyletic etc is also part of
cladistics, but that doesn't seem to be at issue at this point. I've
admitted that I think molecular data works better with cladistic
analyses than morphological data, but I see both have value. In view of
the difficulty with prior designation of homology, I prefer to use
molecular data for analyses, and then to map morphological character
state changes on to my cladistic tree a priori.

I've outlined this above because when I read John's posts about whether
or not molecular characters are cladistic or not, i think to myself
"What is he talking about?" and I see lots of responses to echoing
similar sentiments. What I have outlined above is what plant
systematists (and the odd animal systematist I know) do when they 'do
cladistics' - at least that's how I understand it. Whether the
character is molecular or morphological isn't relevant to the method
(even though it might be relevant to the outcome). I wonder sometimes
if there a still lots of hangovers in the minds of some from the 'old
days' when people were more concerned with prior designation of
homology and polarity. Computer programs like MacClade and PAUP still
allow us to do these things prior to the analysis, but whenever this
sort of thing is done, I feel it should be assessed on a case-by-case
basis and doesn't constitute an essential part of what I see as the
cladistic method.

Please let me know if i've made any glaring errors here - I'm thinking
as I go.

Cheers, David Orlovich.

Dr David Orlovich,
Department of Botany,
University of Otago,
P.O. Box 56,
Dunedin,
New Zealand.

Phone: +643 479 9060
Fax: +643 4769 7583