Very few early tetrapod groups have survived the intense phylogenetic reshuffling of
recent analyses. Among those that have, diadectomorphs appear repeatedly as the nearest
relatives of crown-amniotes; likewise, the stem-tetrapod affinities of colosteids and most
Devonian forms have been retrieved consistently by different authors, despite differences
in taxon sample size and the use of contrasting character ordering, weighting, and coding
regimes (see also Ruta et al., in press). These data suggest (although not conclusively) that
some regions of the tetrapod tree are better corroborated and more stable than others
(Panchen and Smithson 1987, 1988; Sumida and Lombard 1991; Berman et al. 1992;
Lombard and Sumida 1992; Sumida et al. 1992; Laurin and Reisz 1997, 1999; Lee and
Spencer 1997; Sumida 1997; Berman et al. 1998; Laurin 1998a–c; Paton et al. 1999;
Berman 2000; Clack 2001).
Methodological noteThe strict consensus topologies deriving from the most widely discussed published
datasets including Caerorhachis are considered here (Figures 11.2–11.7). The strict
consensus trees resulting from our new analysis (Figures 11.8–11.9) and from
experiments of character removal (Figure 11.10) are also illustrated. As in Ruta et al.’s
(2001) paper, Lebedev and Coates’ (1995) and Clack’s (1998b,d) analyses have been
omitted, since they are superseded by Coates’ (1996) and Paton et al.’s (1999) works,
respectively. Strict consensus trees are plotted on a stratigraphical scale resolved down to
stage level (geological timescale based on Briggs and Crowther 2001, and references
therein). For simplicity, stages are drawn to the same length, and not proportional to
their actual duration, although dates in millions of years before present (Ma) are
appended, where possible, to stage names. In addition, the known ranges of major early
tetrapods groups are used (Benton 1993), instead of specific occurrences of individual
species. The use of whole ranges permits rapid and easy comparisons between tree
shapes, and circumvents the problem of comparing time trees built on different taxon
samples for each group. Internodes within monophyletic groups are represented by
vertical bars of fixed, arbitrary length (except where ghost ranges are present; Smith
1994). This length represents merely a graphical expedient and does not imply an equal
time for the origin of adjacent nodes. It has, however, the inconvenient effect of
generating chronologically ‘deep’ origin events for some groups, depending upon the
number of internodes and the placement of the stratigraphically oldest members of a
group. Since the actual time occurring between adjacent nodes is unknown, the age of a
node leading to two sister taxa is conservatively taken to coincide with the age of the
older taxon.
Where species or genera are used as Operational Taxonomic Units (OTUs), it is possible
to identify the point of divergence between sister groups, even if whole stratigraphical
ranges are employed. For example, in Anderson’s (2001) tree, the stem-caecilian
Eocaecilia micropoda is the sister taxon to brachystelechid microsaurs. Therefore, the
divergence of caecilians can be graphically plotted within the stratigraphical range of
microsaurs instead of at the base of such a range (Figure 11.7). Paraphyletic groups pose
problems when whole ranges are used. A possible way around this consists of splitting the
ranges of large groups into the smaller ranges in which their component subgroups
234 BONES, MOLECULES, AND CROWN-TETRAPOD ORIGINS