Page 7 - Renaud_Michaux_2007

Mandibles and molars of the wood mouse, Apodemus sylvaticus (L.)

Mandible Molar M1
(a) 35 40 45 50 (c) 35 40 45 50
5 5
4 3 FR-PTCR 4 3 FR-IEV
Md shape (EFT-CA2) 2 1 0 SP-MUR SIC-M IBIZ IT-GROS FR-S FR-OL FR-CH FR-NM FR-IEV M1 shape (RFT-CA1) 2 1 0 SIC-FIC SP-MUR IBIZ SARD TARQ ELBA FR-S FR-CH FR-RE FR-LA

FR-PORQ
SP-N
PYR
SP-ALC
SP-N
FR-Y
FR-MTP
GROS
FR-LA
FR-NM
PYR
SIC-GRAT
FR-MTP
CO
IT-CAL
ELBA
–1
SIC-FIC
–2
SIC-GRAT SP-ALC SARD CO IT-TARQ FR-RE FR-Y –1 SIC-M IT-CAL FR-PTCR FR-OL
–2 –3 FR-PORQ
r = 0.788, P = 0.002 r = 0.814, P < 0.001
–3 –4
(b) Latitude (d) Latitude
6.8 7
ELBA
6.6 6.8
IBIZ ELBA 6.6 SIC-GRAT SIC-M IBIZ FR-PORQ FR-Y
Md size (EFT-Area H1) 6.2 SIC-FIC SP-ALC IT-TARQ PYR SP-N FR-S FR-OL FR-CH FR-IEV M1 size (RFT-A0) 6.2 6 SP-MUR IT-CAL SP-ALC SARD IT-TARQ FR-PTCR FR-S FR-OL FR-CH FR-LA FR-IEV
6.4
SIC-M
6.4
SP-N
IT-GROS
FR-MTP
CO
SP-MUR
SIC-FIC
IT-GROS
6.0
FR-RE
IT-CAL
FR-Y
CO
FR-MTP
FR-PORQ
5.8
PYR
FR-PTCR
5.8
FR-RE
SIC-GRAT
5.6
5.6
5.4
5.2 SARD FR-NM FR-LA 5.4 FR-NM
5.2
r = 0.621, P = 0.031 r = 0.711, P = 0.010
5.0 5
35 40 45 50 35 40 45 50
Figure 3 Shape and size variations of the mandible (a, b) and the M1 (c, d) with latitude. Shape is estimated by the second (mandible) and
first (M1) canonical axes, size is estimated by the area of the first ellipse (EFT) for the mandible and the zero harmonic (RFT) for the M1.
Symbols correspond to average values per geographical groups ± standard error (95%). Dotted line: regression of the mainland samples vs.
latitude. R and P values are given on the graphs (regression of the group means on latitude, N ¼ 12). The outlines visualize the corres-
ponding shape change.
axis CA1 for molars (linear correlation of group means with
Multivariate comparisons between shape, size,
latitude, N ¼ 12, CA1: R ¼ 0.449, P < 0.001; CA2: R ¼ 0.020,
genetics and latitude
P ¼ 0.801).
Most insular populations match the latitudinal gradient The canonical analyses of the Fourier coefficients showed that
for the mandible (Fig. 3a) and for the molar shape (Fig. 3c). the pattern of morphological differentiation was the result of a
However, in both cases, the few populations that depart complex interplay between the influences of latitude, insularity
from the overall gradient originate from islands. Sardinia, and genetic differences among clades. Furthermore, as size
Porquerolles and Port Cros are the most obvious cases of shows a latitudinal gradient, the shape gradient might be
departure of the mandible from the general gradient along linked to an allometric effect, which would imply different
axis CA2. causal effects. To disentangle these various factors, multivariate
The latitudinal gradient is also expressed on molar shape multiple regressions were undertaken using the Fourier
(Fig. 3c). As with mandibles, some insular populations depart coefficients as dependent variables.
from the general trend. The most obvious cases are the Firstly, using mainland only and mainland vs. island
Mediterranean French islands Porquerolles and Port-Cros, populations, the existence of the latitudinal shape gradient
followed by the Atlantic French islands Ole ´ron, Re ´ and was confirmed using multivariate multiple regressions of the
Noirmoutier. Fourier coefficients on latitude (Table 2). In order to separate
The size of the characters also varies with latitude and a possible allometric effect causing this shape gradient, FCs
insularity (Fig. 3b,d). Mandible size on the mainland decreases were regressed on size and the residuals were compared with
towards higher latitudes (N ¼ 12, R ¼ 0.621, P ¼ 0.031). latitude (Table 2). A significant latitudinal effect persisted after
Most of the island populations follow this trend as well. The removing the size effect for both mandibles and molars.
population from Ibiza provides the only obvious case of Secondly, multivariate analyses of variance were applied to
insular gigantism based on mandible size. test for a systematic effect of insularity and genetic clades on
The size of the molars for mainland populations is mandible and molar shape (Table 3). The clade effect proved
correlated in a similar way with latitude (N ¼ 12, R ¼ 0.711, important on raw FCs, but weakened significantly after
P ¼ 0.010). Teeth from Yeu and Elba, unfortunately both removing the latitudinal effect for the mandible. However,
sampled by a single specimen, constitute departures from the the clade effect remained highly significant for molars
overall trend. (Table 3). This result corroborates the pattern emerging from
Journal of Biogeography 34, 339–355 345
ª 2006 The Authors. Journal compilation ª 2006 Blackwell Publishing Ltd