other methods, in particular when a molecular clock ex-
ists (shown in this case by the LINTRE test, see below)
[9]. Both the UPGMA and the linearized NJ trees were
obtained using F. coelebs as an outgroup. Analysis of the
genera
Loxia, Carpodacus, Uragus, Haematospiza,
Pyrrhula, Eophona, and Mycerobas suggested an evolu-
tionary rate of 0.44 ¥ 10–9 nonsynonymous substitutions
per nonsynonymous site per year and 1.43 ¥ 10–8 syn-
onymous substitutions per synonymous site per year, so
that the overall rate is 3.6 ± 0.14 ¥ 10–9 [10]. This results
in a substitution rate of 0.4% per million years, which
leads to a rough approximation of 8% nucleotide substi-
tutions per 10 million years. Taking into account the
mtDNA differences found between the most distant
species in the scaled constructed linearized tree shown in
figure 2, which also approximates to 8%, the radiation
seems to have started about 13 MYA, before that of the
Serinus and Carduelis genera (fig. 2) [refs 4, 5 and un-
published data]. F. coelebs was found to have diverged
16.5 MYA; this value was used in figure 2. The topology
of the UPGMA and linearized NJ trees (fig. 2) was very
similar when using chicken and pheasant as additional
outgroups (not shown). This is important for discussing
the divergence times hypotheses put forward in the pre-
sent work. However, the substitution rate found by us
(0.4% per million years) differs from the standard 2% per
million years. Furthermore, cranes show a slower rate of
nucleotide substitution (about 1% [24]), closer to our re-
sults for Carduelis, Serinus and Passer.
Estimates using chicken/pheasant as an outgroup were
similar to those obtained when Fringilla was the outgroup.
Nevertheless, the calculation of times of species diver-
gence needs to be confirmed by other methodologies and
with additional species. Passerines [8] are found to be
older than paleognathous birds, and the Carduelini species
(Carduelis and Serinus) which are relatively close to Old
World sparrows (genus Passer) are considered nearly as
old as Passer [4, 5, 25, 26]. The use of either chaffinch or
chicken and pheasant as outgroups for inferring phyloge-
nies in our sample would seem to be correct, because only
third position transitions appear to be saturated (fig. 1),
but we chose the more closely related Fringilla to root the
trees; the LINTRE computer program [9] also showed that
there the evolutionary rates are constant among the bird
lineages used. Thus, for cyt b, a molecular clock exists
among the studied species (table 1). Finally, a caveat
should be added because an early timing for passerine di-
vergence [4, 5, 8, 27] is not accepted by others who pro-
pose speciation during the Pleistocene [28].
Phylogeography of crossbills, bullfinches, grosbeaks,
and rosefinches
In the present work, the complete geographical range of
crossbills, bullfinches, grosbeaks, and rosefinches was
covered with the studied species.
In preliminary analyses (data not shown), canaries (genus
Serinus) were shown not to be related to either gold-
finches (genus Carduelis) or any of the genera newly
studied in the present paper.
1) Crossbills (genus Loxia, 4 species)
Crossbills are closely related to redpolls (Carduelis flam-
mea and C. hornemanni) in all analyses and all dendro-
grams (figs. 2, 3). Phylogenetic trees which included all
Carduelis [4] and Serinus [5] species showed the same
strong close redpoll/crossbill relatedness (data not
shown). Like redpolls, crossbills have a northern hemi-
sphere distribution, and a characteristic crossed mandible
for specialized extraction of conifer seeds. They seem to
have a more ancient origin than redpolls (see NJ lin-
earized tree, fig. 2). They probably originated from a Car-
duelis-like ancestor when conifers were very common on
Earth. Pine cones undergo irregular cycles of appearance
and redpolls may have evolved at a time when pine cones
were scarce and the hypothetical ancestor of the redpoll
was forced to emerge from the conifer woods to find food
in neighboring small mesothermal plants. The time when
this occurred is uncertain; if we take the time scale hy-
pothesized in figure 2, it would be about 9 MYA. There
was clearly a decline in the number of conifers after the
cold periods of Pleistocene glaciations, when redpolls
could have appeared (fig. 2). However, the time scale is
still debated [4, 5, 8, 27, 28]. Beak shape may change
very rapidly according to feeding needs (i.e., lack of
conifers [29]).
The Arctic redpoll (C. hornemanni, size 14 cm) is very
similar to Loxia curvirostra japonica, a very small Loxia
subspecies (14–15 cm). Red color varying in distribution
placements and intensity according to subspecies, is con-
served in both crossbill and redpoll males [30]. Typical
seasonal north-south migrating patterns occur in both
crossbills and redpolls, but the characteristic southern ir-
ruptive behavior of the former is unique, since the avail-
ability of pine cones is unpredictable in the pine woods
[30, 31].
Consideration should also be given to the possibility of
classifying redpolls apart from the genus Carduelis and
together with Loxia, since redpolls are genetically distant
from the twite/linnet couple [4]. Previously, twite, linnet,
and redpolls were considered as a subgroup (or even an-
other genus, Acanthis) [30, 31] within the genus Car-
duelis.
2) Bullfinches (genus Pyrrhula, 6 species)
These finches have a palearctic distribution, including
the Azores Islands and Japan. In the present work, we col-
lected species and subspecies representing the entire geo-
graphic range. Their beak has evolved to eat buds [30].
They cluster together with pine grosbeak, Pinicola enu-
cleator (presently placed in the genus Pinicola, 2 species)
CMLS, Cell. Mol. Life Sci.
Vol. 58, 2001
Research Article
5
