Page 8 - Asylv_molars_Pmax_revised_2014_01

8

In
order
to
estimate
the
robustness
of
Pmax
and
of
the
evolutionary
directions
regarding
sampling,

each
considered
geographic
group
was
bootstrapped
100
times.
Pmax,
allometric
directions,
and

directions
of
changes
between
groups
were
calculated
on
these
bootstrapped
samples,
and

compared
with
the
original
direction.
The
mean
correlation
value
provided
an
assessment
of
the

stability
of
the
vector
regarding
sampling.

Results

Test
of
sexual
dimorphism
and
local
variation

Sexual
dimorphism
was
evidenced
in
none
of
the
cases
(Table
2),
a
result
congruent
with
other

studies
on
wild
populations
of
wood
mice
(Renaud
2005;
Ledevin
et
al.
2012)
and
domestic
mice

(Valenzuela-­‐Lamas
et
al.
2011).
Males
and
females
were
therefore
pooled
together
in
subsequent

analyses.

Season
of
trapping
in
Lantabat
was
not
related
to
any
differences,
supporting
the
idea
that
age

structure
of
a
population
had
no
impact
on
estimates
of
molar
shape.
Differences
between
sub-­‐
localities
in
Mimizan
were
significant
for
molar
size
and
close
to
the
significant
threshold
in
molar

shape,
indicating
that
variations
at
a
small
spatial
scale
might
interfere
with
variations
at
larger
scale.

Geographic
patterns
in
molar
morphology

Differences
in
molar
shape
were
significant
among
groups
(MANOVA:
P
<
0.0001).
Mainland

populations
tend
to
scatter
in
two
groups
along
the
first
multivariate
axis
PC1,
with
most
localities

being
close
to
the
origin
but
all
Western
French
populations
clustering
towards
positive
PC1
values

(Fig.
2A;
for
a
visualization
on
a
geographic
map,
see
Fig.
3A).
These
populations
share
slender
molar

shape.
Almost
all
insular
populations
scatter
around
this
core
mainland
variation.
Several
islands

(Hyères
Islands
of
Porquerolles
and
Port-­‐Cros,
French
Atlantic
islands
of
Oléron
and
Ré,
Southern

Corsica,
Marettimo
islet
off
Sicily)
display
extreme
negative
values
along
PC1,
corresponding
to
very

broad
molars
(Fig.
2B).
The
Gambarie
sample
(southernmost
extremity
of
mainland
Italy)
constitutes

the
only
mainland
sample
clustering
with
these
insular
populations.
Sicilian
populations
tend
to

isolate
along
PC2.
This
axis
tends
to
opposes
molars
wide
at
their
posterior
part
(positive
PC2
values)

to
molars
with
a
short
and
broad
forepart.

Molar
size
also
varied
significantly
across
groups
(ANOVA
and
KW:
P
<
0.001).
According
to
previous

results
(Renaud
and
Michaux
2007),
the
main
pattern
corresponds
to
a
trend
of
larger
molars