Page 6 - Asylv_molars_Pmax_revised_2014_01

6

Sex
of
the
specimens
was
known
in
three
well-­‐sampled
populations
(Mimizan,
Lantabat,
Tourch)
in

which
the
occurrence
of
sexual
dimorphism
in
molar
size
and/or
shape
was
investigated.
The

Lantabat
sample
was
trapped
in
four
successive
seasons
(two
in
spring
and
two
in
autumn)
allowing

the
investigation
of
the
influence
of
the
period
of
trapping.
The
Mimizan
sample
was
composed
of
a

set
of
four
places
distant
of
few
kilometers,
and
this
allowed
the
investigation
of
possible
local

variations
in
a
metapopulation.

All
specimens
had
all
their
third
molars
erupted.
Because
the
third
molar
erupts
at
the
period
of

weaning,
this
is
a
classical
criterion
for
excluding
juvenile
specimens.
All
animals
were
thus

considered
as
sub-­‐adults
and
adults.

Methods

Outline
analysis
of
the
first
upper
molar

Molar
shape
was
estimated
by
the
two
dimensional
outline
of
the
tooth
in
occlusal
view,
with
focus

towards
the
basis
of
the
crown,
which
is
affected
late
by
progressive
wear
during
the
animal’s

life.This
molar
shape
quantification
has
been
shown
to
be
relatively
insensitive
to
age
classes
in

wood
mice,
except
for
very
young
or
very
old
animals
that
are
anyway
rare
in
most
wild
populations

(Renaud
2005).
Such
an
approach
has
been
shown
efficient
to
evidence
patterns
of
biogeographic

variation
in
the
wood
mouse
(Renaud
2005;
Renaud
and
Michaux
2007).

Each
outline
was
described
by
a
set
of
64
point
at
equal
distance
along
the
outline.
The
starting
point

was
tentatively
located
at
the
anteriormost
part
of
the
tooth,
and
adjusted
by
aligning
the
starting

point
with
the
long
axis
of
the
tooth.
Outlines
were
then
analyzed
using
a
radial
Fourier
transform

(RFT)
(Renaud
et
al.
2006;
Renaud
and
Auffray
2013).
Using
this
method,
a
set
of
radii
(i.e.
distance

from
each
point
to
the
center
of
gravity
of
the
outline)
was
calculated
from
the
original
coordinates,

and
this
set
was
decomposed
into
Fourier
functions,
i.e.
trigonometric
functions
of
decreasing

wavelength
termed
‘harmonics’.
Each
harmonic
is
weighed
by
two
Fourier
coefficients
(FCs)
that

constitute
the
shape
variables.
The
zero
harmonic
(A0)
corresponds
to
a
best-­‐fit
circle
to
the
outline.

It
was
used
as
size
estimator,
and
to
standardize
all
other
FCs
for
isometric
size
differences.
This
was

performed
by
dividing
all
raw
FCs
by
A0.
Consideration
of
the
first
seven
harmonics
was
considered
a

satisfactory
compromise
between
information
content
and
number
of
variables
for
both
characters

(e.g.
Renaud
and
Michaux
2007).
The
14
resulting
FCs
(two
FCs
per
seven
harmonics)
were
used
as

shape
variables.