Page 6 - MICHAUX2002

432 J.R. Michaux et al.

Mallorca Menorca Ibiza Porquerolles Port-Cros Marettimo (11) (12) yy transformed in order to stabilize variance (Harvey, 1982). All
regressions between contrasts were forced through the origin
(7) (8) (9) (10) y yy (Garland et al., 1992). In order to verify that contrasts were
standardized properly we performed a regression of the
yy y yy y absolute values of standardized contrasts vs. their standard
deviations (Garland et al., 1992) using CAIC.
Corsica Sardinia Elbe Sicily (5) (6) yy yy y yy yy yy yy y yy y
RESULTS
(3) (4) yy yy yy yy yy yy yy yy
We found a significant positive relationship between geo-
Spain y y y y y y y y y y y graphical distances and genetic distances using the permuta-
(2) tion test, showing that geographical isolation of A. sylvaticus
populations was related to a decrease in genetic exchanges
France Italy yy yy yy Hieraaetus pennatus y yy Larus melanocephalus y y yy yy yy yy yy yy yy yy Garrulus glandarius y y (r = 0.57, P = 0.001). The position of the population of Port-
(13) (1) Cros can be noted between the Italian group (Italy, Sicilia,
Corsica, Elba) and the Iberian (Spain, Balearics) and French
Table 2 continued. Locations (c) Accidental predator species Lynx pardinus Canis lupus Circus aeruginosus Circus pygargus Larus cachinnans Coracias garrulus Ciconia alba Egretta garzetta Ardea purpurea Ardea cinerea Bulbulcus ibis Corvus corone Pica pica (France, Porquerolles) groups.
(see map, Fig. 1a)
Using the whole dataset, we found that A. sylvaticus body
size is not related to area of source region /island (r = −0.098,
P = 0.75; Fig. 2a). However, this lack of correlation was due
to one point (Port-Cros). The regression was statistically
significant when this outlier was removed (r = −0.632, P =
0.028; Fig. 2a). Significant correlations were found between
A. sylvaticus body size and area (r = −0.71, P = 0.0098),
number of regular predator species and area (r = 0.94,
P < 0.0001), number of occasional predator species and area
(r = 0.98, P < 0.0001) and number of competitor species and
area size (r = 0.95, P < 0.0001) using independent contrasts.

We controlled for body size, total number of predator
species (occasional and regular) and the number of com-
petitor species in relation to area by using the residuals of the
preceding regressions using independent contrasts. We found
a negative correlation between the residuals of body size
(controlled for area) and the residuals of the total number of
predator species (controlled for area) using independent
contrasts (r = −0.76, P = 0.0045, Fig. 2b). A similar negative
relationship was found using the number of regular predator
species (r = −0.74, P = 0.0063). These results suggest that
the changes in body size of A. sylvaticus depend on predator
pressures (irrespective of area of source region/island).

No correlation was found between the residuals of body
size (controlled for area) and the residuals of the number of
competitor species (controlled for area) using independent
contrasts (r = −0.237, P = 0.46).

When using the raw data (i.e. not controlled for genetic
distances), a negative trend (also not statistically significant)
was found between the residuals of body size (controlled for
area) and the residuals of the number of competitor species
(controlled for area) (r = −0.54, P = 0.07, Fig. 2c).

We confirmed these findings by performing a multi-
regression using the permutation method (Legendre &
Legendre, 1998). Using a backward procedure, this method
allows the selection of a subset of explanatory (i.e. independent)

© 2002 Blackwell Science Ltd, Global Ecology & Biogeography, 11, 427– 436