Re: Ancestors [was: Re: And while on the theory of phylogenetic reconstruction...]

[David Marjanovic <david.marjanovic@gmx.at>]

> > there is a paper on the phylogenetic position of turtles, [...]
> > and then says that mammals are the basalmost amniotes [...]
>
>     I NEED THIS REFERENCE NOW! I have a moldy, half-finished manuscript
> titled "Mammals are basal amniotes," and I need this reference!

Really? :-)
Rafael Zardoya & Axel Meyer: The evolutionary position of turtles revised,
Naturwissenschaften 88, 193 -- 200 (?May 2001)

"_Fig. 2A -- F_ Alternative hypotheses explaining the phylogenetic position
of turtles within living amniotes. _A_ Mammals are the most basal living
amniotes. Turtles are the only living representatives of anapsid reptiles.
Living diapsid reptiles include Lepidosauria (the tuatara, snakes, and
lizards) and Archosauria (crocodiles and birds) (e.g. Laurin and Reisz 1995;
Lee 1997). [...] _F_ The Haematothermia hypothesis: birds are the sister
group of mammals (e.g. Gardiner 1993)"

Fig. 2A, labels from bottom to top: Mammals, Turtles, Tuatara, Lizards,
Snakes, Crocodiles, Birds.

>     Some people use "cladist" to signify those phylogenetic systematists
of
> the mindset that parsimony is the only appropriate optimality criterion
for
> phylogenetic inference. This group is exemplified by the AMNH systematics
> community, and many of the members of the Willi Hennig Society. There
seems
> to be a connotation that many "cladists" in this sense means pattern
> cladist, although this is certainly not true (e.g., Arnold Kluge). I tend
to
> call such folks Kladists (with a K), a practice I got from George
Olshevsky.

Which folks, those who think only pattern cladists are "cladists", or those
who think only parsimony must be used?

> Of course, in biological circles, there are few people
> who really do not fall under that definition anymore (HINT HINT).

...?
I'm tired this rainy morning...

>     For the record, what David describes here is a SPECIES polytomy. It is
> possible that, when sufficiently sensitive molecular markers are
recovered,
> it will be found that organisms within each species share more recent
common
> ancestors with members of some of the other species than with members of
> others.

Possible -- but, judging from today's mainland *Podarcis* species (which is
quite diverse in its mt genes), it's quite likely that each such molecular
marker would produce a different tree.

> > Don't hold your breath,
>
> I'm not, but I'm banking my dissertation on it, which is arguably worse.

Oh!
Peter Forster & Alfred Toth: Toward a phylogenetic chronology of ancient
Gaulish, Celtic, and Indo-European, PNAS 100(15), 9079 -- 9084 (22 July
2003)

Abstract:
"Indo-European is the largest [gone crazy?] and best-documented language
family in the world, yet the reconstruction of the Indo-European tree, first
proposed in 1863, has remained controversial. Complications may include
ascertainment bias when choosing the linguistic data, and disregard of the
wave model[*] of 1872 when attempting to reconstruct the tree. Essentially
analogous problems were solved in evolutionary genetics by DNA sequencing
and phylogenetic network models, respectively. We now adapt these tools to
linguistics, and analyze Indo-European language data, focusing on Celtic and
in particular the ancient Celtic language of Gaul (modern France), by using
bilingual Gaulish-Latin inscriptions. Our phylogenetic network reveals an
early split of Celtic within Indo-European. [Apart from Celtic, only
English, Latin + Romance languages and Greek were used. Outgroup: Basque.
Boo.] Interestingly, the next branching event separates Gaulish (Continental
Celtic) from the British (Insular Celtic) languages, with Insular Celtic
subsequently splitting into Brythonic (Welsh, Breton) and Goidelic (Irish
and Scottish Gaelic). Taken together, the network thus suggests  that the
Celtic language arrived in the British Isles as a single wave (and then
differentiated locally), rather than in the traditional two-wave scenario
("P-Celtic" to Britain and "Q-Celtic" to Ireland). The phylogenetic network
furthermore permits the estimation of time in analogy to genetics, and we
obtain tentative dates for Indo-European at 8100 BC +- 1900 years, and for
the arrival of Celtic in Britain at 3200 BC +- 1500 years. The phylogenetic
method is easily executed by hand and promises to be an informative approach
for many problems in historical linguistics."

* Was the idea that languages don't have phylogenetic relationships, and
that instead linguistic novelties just spread over geographical areas. While
vastly exaggerated, the phenomenon does exist. For example, Ukrainian,
Belorussian, Slovak, Czech, and Upper Sorbian have experienced a G --> H
change, while it's obvious that Upper and Lower Sorbian are sistergroups,
and that Ukrainian and Belorussian are East Slavic languages, while the
others (and Polish) are West Slavic ones.

"Methods
_Construction of the Indo-European network._ We used the linguistic network
approach of Forster et al. (4). A phylogenetic network displays differences
of items between language lists (or DNA molecules) with links and branches
like a tree does, except that a network may contain reticulations when
convergence (i.e. through historical loan events or chance parallel changes,
or even through data misassignments by the researcher) has obscured the
evolutionary tree. The linguistic network approach is therefore expressly
intended to search for treelike structure in potentially 'messy' data. A
technical problem that needs to be overcome is the multitude of irrelevant
trees that a network may contain, in the form of reticulations, with even
modestly noisy data. The first phase of the linguistic network approach is
to remove characters (items) with more than a certain number of states
(e.g., lexemes) across the translations according to an empirically
determined threshold, because high variability is a sign either of inherent
instability of that item or of unreconstructable ancient changes. The second
phase is to process the binary characters (binary items) in the data to keep
initial complexity to a minimum, with the rationale that a less variable,
geographically widespread character state (e.g., a lexeme or phoneme) is
more likely to reflect genetic relationships between languages. Characters
that include states that are suffix losses are initially disregarded. In the
third phase, the multistate characters are processed by splitting them into
binary characters dictated by the current network: for each multistate
character, that binary split which partitions the largest group of closely
linked nodes or taxa (i.e., languages) is introduced first, following the
same rationale as in phase 2. The other states of the multistate character
are then split off the enlarged network. The fourth phase is to process the
suffix losses. We found these to be least reliable for tree construction,
because independent losses of a suffix frequently occur, causing convergence
and thus reticulation. In all four phases, the processing of a certain
character may contribute disproportionately to an increase in reticulations.
If this occurs, the step should be reverted and other characters should be
chosen iteratively for their ability to enlarge the network at low
complexity. The temporarily excluded item is reintroduced in the next round
of the phase. In the final phase, the uninformative binary items (i.e.,
those that differ in only one language) are added to the branch tips."

35 characters, 14 languages. Very little missing data (none of those in
Gaulish). Unfortunately I can't find a way to write the resulting network
(Fig. 3). Basque does not appear in the network at all.

"Outlook
The present analysis excludes a number of interesting ancient Indo-European
languages such as Hittite, Tocharian, etc. These omissions are an inevitable
side effect of including the fragmentary corpus of ancient Gaulish: other
ancient and fragmentary corpora would have little or no overlap with the
Gaulish items, thus preventing any comprehensive phylogenetic analysis. To
circumnavigate this difficulty and to arrive at a complete tree of ancient
and modern Indo-European languages, future analyses may focus on the
phylogenetic placement of a specific fragmentary language, as we have
performed here for Gaulish, and may then piece together the resulting
partial phylogenetic networks into a unified Indo-European language
network."

Hittite and Tocharian have left impressive amounts of texts, so I don't
think they're all that fragmentary...

Ref. (4) is
P. Forster, H. Toth & H.-J. Bandelt, J. Quant. Linguist. 5, 174 -- 187
(1998).