Geoscience Reference
In-Depth Information
carbonate, raises the possibility that at least some of
the still water bodies were brackish, even if the sea
was far away. Furthermore, Eberth & Berman
(1993, p. 46) seem to have interpreted the presence
of the dipnoan Sagenodus and the osteolepidid Loh-
sania as freshwater indicators. Sagenodus was prob-
ably euryhaline, however, and is thought to have
occurred in marine and freshwater environments
(Schultze & Chorn 1997). We have found no
detailed data on the presumed habitat of Lohsania
but, given the general reinterpretation of osteolepi-
dids from freshwater to marine and euryhaline
forms, the freshwater interpretation of the Cutler
Formation does not appear to be supported by fau-
nistic criteria. We have provisionally accepted the
conclusions of Eberth & Berman (1993) but it
would be interesting to study the mollusks, arthro-
pods and other metazoans from that formation.
uncertainties about the palaeoenvironmental
interpretation of many localities, adding more taxa
would not have changed the global pattern. Further-
more, the information presented in Table 2 enables
any interested palaeobiologist to expand the analy-
sis to additional taxa.
To determine if character optimization yields
reliable information on ancestral states, the presence
of a phylogenetic signal should be assessed (Laurin
2004). The high number of polytomies constrains
the choice of randomization procedure because the
number of steps required by trees which include
soft or hard polytomies, and of trees with randomly
resolved polytomies, differs. An appropriate ran-
domization procedure is to reshuffle terminal taxa
randomly on the tree in which topology and
branch lengths are kept constant, as was done by
Laurin (2004). In this case, all random trees
include the same number and type of polytomies.
Another solution would have been to randomly
resolve the polytomies several times (ten or more)
to investigate the phylogenetic signal in all of
these trees and to average the probabilities; that sol-
ution would be more time consuming and poten-
tially less accurate, however (unless a much
greater number of random resolutions were exam-
ined). In both cases, the probability that the distri-
bution of the character states is independent of the
phylogeny is given by the number of random trees
(produced by reshuffling) which implies the same
number (or fewer) transitions as the reference tree,
divided by the total number of random trees (here,
10 000). The three states (0: marine; 1: brackish;
and 2: freshwater) were ordered according to a
salinity gradient.
Given the controversies surrounding palaeoen-
vironmental interpretations of most Palaeozoic fos-
siliferous localities in which stegocephalians were
found, two optimizations of habitat are presented.
The first presents the most traditional interpretation:
many localities are interpreted as freshwater
environments or, when clearly marine or brackish
water, stegocephalian remains are interpreted as
allochtonous elements brought in by rivers
(Fig. 3). Since the phylogenetic signal is highly sig-
nificant ( p ¼ 0.0002), the character can be opti-
mized. This optimization suggests that the first
sarcopterygians and tetrapodomorphs lived in a
brackish or marine environment (which is not
new, of course) and that the move to freshwater
environments took place before the last common
ancestor of Tiktaalik and stegocephalians. The few
Palaeozoic stegocephalians which tolerated brack-
ish water represent returns to a marginal marine
environment.
The second optimization presents the alternative
interpretation of a more marine (or at least brackish)
environment of most fossiliferous localities. Under
Evolutionary analysis of habitat in early
stegocephalians
A time-calibrated supertree was compiled from the
literature (Figs 3 & 4). Given the large time-span
encompassed by the tree (Givetian to Roadian on
the figure, but it really extends to the Holocene),
all terminal taxa were placed within the proper geo-
logical stage. No attempt was made to achieve
greater stratigraphic precision, for various reasons.
First, the gained precision would not be visible on
the figure, unless a non-linear timescale was used.
Second, given the stratigraphic uncertainties on
the age of many fossils and the still greater uncer-
tainty about the actual (as opposed to observed) stra-
tigraphic range of most terminal taxa (Marshall
1997; Marjanovi´ & Laurin 2008a), the gains in pre-
cision would be more apparent than real. Among
many arbitrary branch length values that could
have been used, we set both terminal branches to a
minimal length of 1 Ma and internal branches to a
minimal length of 2 Ma. We placed the end of the
stratigraphic range of all terminal taxa at the top
of the geological stage to which they belong, as
was done by Laurin (2004) and Marjanovi´ &
Laurin (2007). Given the large number of terminal
taxa (86) and of polytomies included in the tree,
this procedure results in reasonable ages for the
hypothetical ancestors. Presumed terrestrial taxa
were excluded because the distinction between salt
and freshwater habitats does not apply to them.
Early amniotes are therefore represented by
Ophiacodon, which may have been amphibious to
aquatic (Romer 1958; Germain & Laurin 2005).
Hylonomus is included (but not coded for habitat)
to provide a temporal calibration of this part of the
tree only. The tree is not exhaustive but, during its
compilation,
it
became
clear
that
given
the
