250, respectively, when G. gallus is used as an outgroup.
These numbers are appropriate to establish sound phylo-
genetic comparisons [21]. Data related to Gallus are in-
cluded throughout the text when discussing the species
time of divergence hypotheses.
The nucleotide distribution pattern of the cyt b gene of
the birds under study was similar to that found in a previ-
ous analysis of this gene in other birds and mammals
[13]. At the first codon positions, the four bases had sim-
ilar frequencies; at the second position, fewer G residues
and more T residues were seen. At the third codon posi-
tion, the bias against G and T was strong, as demonstrated
in previous studies [6, 13]. This bias in base composition
was similar in all species studied (results not shown).
Thus, the parsimony methodology seems to be adequate
for all species studied [22]. The variability within the cyt
b gene was theoretically sufficient to establish phyloge-
netic relationships according to the number of observed
phylogenetically informative sites [21]; most of the dif-
ferences were silent substitutions, as expected for a pro-
tein-coding gene, particularly for close relatives, such as
species within a single genus [23].
As expected for a gene evolving relatively rapidly under
strong functional constraints, 84.1% of the third codon
positions were variable among species. The correspond-
ing figures for the first and second positions were 18.5%
and 3.9%, respectively. The base composition according
to codon position was similar in all species; thus, a cor-
rect phylogeny may be inferred from the parsimony
analysis.
The relative tempo of crossbill, rosefinch, bullfinch and
grosbeak evolution was calculated. To estimate the rate
of evolution, we used the type of calculations previously
employed by ourselves and others [4, 5, 10]. After cali-
bration of evolutionary rates, the times of origin of Dar-
win’s finches, canaries, and goldfinches were estimated.
An UPGMA dendrogram (not shown) and a linearized NJ
dendrogram with Kimura biparametric distances (fig. 2)
were constructed because these types of phylogenetic tree
are more suitable for estimating divergence times than
4
A. Arnaiz-Villena et al.
Phylogeography of Carduelini birds
F
Figure 2. Linearized NJ tree calculated with Kimura biparametric distances. Divergence times are estimated for mt cyt b genes under a
mixed molecular and fossil hypothesis [10]. The confidence probability (CP) is shown in the interior part of the node. Bootstrap values are
shown above branches; branch lenghts (¥ 1000) are shown underlined below branches. Bird species used are detailed in table 1. Tree length
(¥ 1000) = 1082. Genetic distances are depicted under the time scale. [Birds are modified from ref. 30. Maps were drawn in our labora-
tory using refs 30 and 34.
