Page 8 - Renaud_Michaux_2007

S. Renaud and J. R. Michaux

Table 2 Multivariate regressions of the shape parameters with latitude and size
Character Dependent variables Independent variable Samples N P

Mandible shape EFT7–FC Latitude Mainland + islands 216 < 0.0001
Molar shape RFT9–FC Latitude Mainland + islands 239 < 0.0001
Mandible shape EFT7–FC Latitude Mainland 143 0.0001
Molar shape RFT9–FC Latitude Mainland 158 < 0.0001
Mandible shape EFT7–FC Mandible size (area H1) Mainland + islands 216 < 0.0001
Residuals EFT7–FC/size Latitude Mainland + islands 216 0.0003
Molar shape RFT9–FC Molar size (A 0 ) Mainland + islands 239 < 0.0001
Residuals RFT9–FC/size Latitude Mainland + islands 239 0.0020

EFT, elliptic Fourier transform; RFT, radial Fourier transform; FC, Fourier coefficients; N, number of specimens considered; P, probability.

Table 3 manovas on shape coefficients

Character Dependent variables Grouping variable Samples N P
Mandible shape EFT7–FC Clade Mainland 143 0.0001
EFT7–FC/lat. Residuals Clade Mainland 143 0.0016
Molar shape RFT9–FC Clade Mainland 158 < 0.0001
Residuals RFT9–FC/lat. Clade Mainland 158 < 0.0001
Mandible shape Residuals EFT7–FC/lat. Insularity Mainland + islands 216 0.1136
Molar shape Residuals EFT7–FC/lat. Insularity Mainland + islands 239 < 0.0001
EFT, elliptic Fourier transform; RFT, radial Fourier transform; FC, Fourier coefficients; lat., latitude; N, number of specimens considered; P,
probability.

the canonical analysis, suggesting a stronger genetic effect on molars with both islands and the mainland. The correlation is
molar shape rather than on mandible shape. weaker when considering insular samples alone (Fig. 4),
Finally, the existence of a typical insular signature was especially with respect to shape differentiation. Overall,
investigated, once the latitude effect was removed, by consid- differentiation in size of mandibles and molars seems to be
ering residuals of a multiple regression of the FCs on latitude positively correlated (Fig. 4a; linear regression, N ¼ 12,
(Table 3). No systematic morphological shift corresponding to R ¼ 0.579, P ¼ 0.048). Wood mice from Ibiza constitute a
an insular trend was detected for mandibles. This may be due departure from the common trend. However, shape differen-
to a random scatter of the insular morphologies around the tiation of mandibles and molars seems to be uncoupled since
mainland ones as shown on the canonical plane (Fig. 2a). the amount of insular divergence of both characters is not
However, such an ‘insular trend’ existed for the molars. Most correlated (N ¼ 12, R ¼ 0.184, P ¼ 0.567), even tending
island samples are clustered in the canonical plane (Fig. 2b) towards a negative regression if the outlying sample from
with the exception of Sicilian molars. Yeu is excluded (N ¼ 11, R ¼ )0.357, P ¼ 0.282).
Identification of the causal factors of these patterns of
differentiation is somewhat intractable as they are likely to
Comparing patterns of differentiation between
mix latitude and various possible factors associated with
mandibles and molars
insular conditions (Table 5). Physical parameters such as
Table 4 shows the correlation of average scores for the area, elevation and distance from the mainland can be
different size and shape parameters of the mandible and considered as proxies for overall environmental conditions.

Table 4 Comparison between mandible and molar size and shape. Euclidean distances were computed for each variable (univariate
estimates of size) or set of variables (FCs). Matrices of distances were then compared using a Mantel test. The coefficient of correlation R is
given, as well as P ¼ 1 (the probability than a random R is larger than the observed R). Group means are considered
Variable 1 Variable 2 Samples N R P
M1 shape RFT9–FCs Md shape EFT7–FCs Mainland + islands 25 0.32735 0.0112
M1 size (A 0 ) M1 shape RFT9–FCs Mainland + islands 25 0.29266 0.0129
Md size (area H1) Md shape EFT7–FCs Mainland + islands 25 0.28993 0.0097
M1 size (A 0 ) (area H1) Md size Mainland + islands 25 0.22066 0.0268
N, initial number of samples involved in the calculation of the distances.

346 Journal of Biogeography 34, 339–355
ª 2006 The Authors. Journal compilation ª 2006 Blackwell Publishing Ltd