Page 8 - THE_ISLAND_RULE_2012

1738 P. RAIA AND S. MEIRI

only variable retained in the model using stepwise backwards same populations (mean 95.6%, Wilcoxon matched pairs, T
regression ( 0.585, P 0.0001). 221, P 0.81).
Taken separately, the resource base is significantly and
positively correlated with size change (Spearman r 0.527, DISCUSSION
P 0.0001), and so is the presence of a smaller competitor
We suggest that different selective forces drive size evo-
(Spearman r 0.172, P 0.050), whereas other variables
are not (mass, r 0.097, P 0.27; diet, r 0.085, P lution in large insular herbivores and carnivores. Our results
0.33; predation, r 0.046, P 0.60; larger competitor, r lend strong support to two of our hypotheses: herbivore size
0.058, P 0.51, combined competition index, r 0.106, decreases in the absence of competitors and predators, and
P 0.23; area, r 0.05, P 0.59). decreases to a lesser extent when some competitors and pred-
ators are present (biological interaction effect hypothesis).
Thus, the resource base, even with the crude estimate we
Sizes of Crete deer were overdispersed, with size ratios be-
employ, explains a significant portion of the variance in car-
tween adjacent-sized species tending toward equality (ov-
nivore size, whereas sex, diet, island area and the presence
erdispersion hypothesis). Carnivore size patterns support our
of predators, and competitors small and large, seem to play
assumption that the nature of the resource base is an important
little or no role in affecting carnivore size evolution.
determinant of body size, but refute our prediction that com-
petition, predation, and diet will also be important (carnivore
Sexual Size Dimorphism Hypothesis
resource-competition hypothesis). Carnivores do not exhibit
Using only island mainland pairs for which we had data increased sexual size dimorphism on islands, refuting our
for both males and females of the same species, we tested SSD hypothesis.
for differences between insular and mainland SSD (calculated Size of large fossil ungulates was significantly affected by
as male CBL/female CBL), and whether such differences are the presence of smaller guild members, and by predators.
male or female driven. There are no differences between Competition appears to be more important than predation:
males and females in the pattern of size change on islands some dwarfing occurred even when large predators were pre-
(calculated as insular CBL/mainland CBL. Paired t-test, n sent. For instance, some four distinct faunal complexes set-
48, t 1.32, P 0.19). In fact, the patterns for males and tled in Sicily (Bonfiglio et al. 2002; Marra 2005). The oldest
females are highly correlated (product moment correlation, includes Elephas falconeri, a diminutive (some 23% the size
r 0.798, P 0.0001). Insular SSD is not different from of its mainland ancestor in linear dimensions), dimorphic and
mainland SSD (paired t-test, n 48, t 1.27, P 0.21). paedomorphic elephant (Palombo 2001; Raia et al. 2003),
The difference between mainland and insular SSD is nega- and no carnivores. Later faunal complexes were rich in both
tively correlated with female body mass (r 0.38, P carnivores and ungulates, and dwarfism occurs in six of eight
0.007), mainly because large carnivores are less dimorphic large ungulates, but to a lesser degree (SR 0.64–1.00). The
on islands. The difference between mainland and insular SSD four Sardinian complexes always feature at least one carni-
is positively correlated with predation level (Spearman r vore, and dwarfism is limited (online Appendix 1). Three
0.29, r 0.045). Island SSD is significantly lower than main- observations are particularly relevant, as they are at odds with
land SSD for predation category 2 (predation pressure low- models suggesting that island area, island isolation, and an-
er on islands), than they are for categories 1 (predation cestral size are the major determinants of body size evolution
pressure possibly lower on islands) and 0 (predation pressure on islands (Van Valen 1973; Heaney 1978; Lomolino 1985,
similar on islands and mainlands) (Fisher least significant 2005).
difference post-hoc test, differences between categories 2 First, in cases of repeated colonization of the same species
and 0, P 0.024, between 2 and 1, P 0.043, between in different islands, size change follows the competition/pre-
categories 1 and 0, P 0.715). Other variables are not dation regime regardless of phylogeny. Second, island iden-
correlated with the difference between insular and mainland tity and hence its area, isolation, and, arguably, total re-
SSD (diet, Spearman r 0.01, P 0.94; resource base, sources are not significant predictors of ungulate size evo-
Spearman r 0.16, P 0.29; presence of smaller com- lution. Admittedly, we did not test explicitly for the effect
petitors, Spearman r 0.18, P 0.21; presence of larger of island area and isolation in fossil ungulates because these
competitors, Spearman r 0.07, P 0.62; combined com- data are not available in most cases. Yet, it is worth noting
2
2
petition index, Spearman r 0.11, P 0.46). Because our that elephants from Crete (8300 km ), Cyprus (9200 km ),
2
2
2
hypothesis predicts enhanced insular SSD in dwarf popula- Rhodos (1400 km ), Sicily (25,700 km ), and Tylos (61 km )
tions only, we compared the difference between SSD of in- were of comparable size although these islands are (and cer-
sular dwarves (mean SR for males and females 1) with that tainly were) different in area by more than two orders of
of insular giants. Insular dwarves are as dimorphic as their magnitude. In addition, the two youngest Crete deer, living
near-mainland conspecifics (mean SSD 0.2% higher on is- along with at least four other deer species, were larger than
lands, n 30 population pairs), whereas insular giants are their mainland counterparts, although inhabiting a relatively
slightly less dimorphic than their near-mainland conspecifics small island.
(mean SSD 2.1% higher on the mainland, n 18 population Third, the body sizes of mainland ancestors/conspecifics
pairs). The difference between these groups is marginally do not affect the magnitude of size evolution after controlling
nonsignificant (Mann-Whitney U-test, U 186, P 0.074). for the effect of competition (in herbivores). An intriguing
SR of females in the 30 population of insular dwarves (mean: observation concerns Crete deer. In Crete, the smallest spe-
95.4%) is not significantly different than that of males in the cies were the oldest (Reese et al. 1996; Vos 2000). Yet, they