Page 4 - THE_ISLAND_RULE_2012

1734 P. RAIA AND S. MEIRI

600 years BP; Reese et al. 1996). Middle Pleistocene fau- than 1/10 the size of the focal species present. The lower
nas include Crete hippos and the smallest elephant (Reese et limit for the inclusion of predators in our dataset was 2 kg.
al. 1996), Sicilian E. falconeri (550,000 25–30% years BP; Carnivores can subdue prey as much as five times larger than
Belluomini and Bada 1985) and Sardinia faunal complexes themselves, and even 10 times larger in the case of group-
with deer and mammoth; Abbazzi et al. 2004). Late Pleis- hunting lions (Radloff and Du Toit 2004). Our size limit of
tocene faunas also include the Sardinian larger mammoth 1/10 the size of the focal prey was decided accordingly. Cat-
(Palombo et al. 2004); younger Sicilian faunas with E. mnaid- egory 2 is meant to indicate occasional predation at best.
riensis plus carnivores, hippo, deer, auroch, and bison (whose Small predators may be important to dwarfing herbivores:
absolute dating ranges between 200,000 and 88,000 years wolves or hyenas were probably not a serious threat to main-
ago; Bada et al. 1991). land straight-tusked elephants, but might have preyed on ju-
Species with uncertain biostratigraphic position or of un- veniles and even on adults of the most dwarfed insular forms.
clear descent (e.g., Balearic Myotragus spp., Sardinian Ne- Thus, for example, the Sicilian elephant E. mnaidriensis
sogoral spp.) were excluded (online Appendix 1). We also (some 30% the mass of its mainland ancestor) was ascribed
excluded all island fossil assemblages either known or sus- to predation category 2 given the presence of cave lion, wolf,
pected to belong to a phase of connection between the island and cave hyena. Category 3 indicates more severe predation
and the mainland. pressure, being restricted to assemblages with carnivores
large enough to prey on the focal herbivores.
Modern Carnivore Faunas Guilds were defined by feeding habits (ascertained by com-
parison with living relatives, see Nowak 1999) and phylo-
We measured crania of medium to large (larger than mar-
genetic criteria as follows: deer were considered to be mixed
ten-sized, 2 kg) carnivores in museum collections (see Ac-
feeders because large deer can tolerate low-quality food
knowledgments), using condylo-basal length (CBL) as an
(Geist 1998), and most insular deer originated from mixed-
index of size. CBL was chosen because it is invariant in
feeding red and fallow deer. Elephants were also considered
adults, has low intrapopulation variability, low measurement
mixed feeders. Bos, Bison, and Equus species were considered
error (Dayan et al. 2002; Meiri et al. 2005c), and does not
to be grazers. Suids were considered to be omnivores, and
change with time since collection as do some skin measure-
hippos were classified in a unique dietary group because of
ments (Winker 1993). We used only specimens with complete
their peculiar niche.
closure of the dorsal sutures and treated males and females
We recognize three competition levels (CLs): (1) no com-
as separate morphospecies. We compared only population
petitor present; (2) competing species of a different guild
pairs from which we had at least three specimens of a given
present or a smaller competitor in the same guild and no
morphospecies on both the island and the mainland.
species in other guilds; (3) smaller competitors in the same
guild (more than half the size of the focal species) and at
Analyses least one other species, irrespective of guild, present. Cate-
gory 1 includes the species freed from competition at the
The biological interaction effect hypothesis
onset of size reduction. Category 2 includes species that faced
We tested this hypothesis by comparing sizes of extinct diffuse competition from species of different guilds exploit-
insular species with those of their likely mainland ancestors, ing overlapping resources (e.g., grasses for both mixed feed-
taken from the recent paleontological literature (online Ap- ers and grazers), or have just one species in the same guild
pendix 1). A size ratio (SR; size on island/size on mainland) but are free to exploit resources in other guilds. Category 3
was computed by taking the ratio of linear measurements, was intended to test our assumption that smaller species in
chosen to maximize sample size. We calculated SR using the the same guild decrease little in size when available free niche
same measurement for all descendants of a common ancestor. space in other guilds is scarce. For example, if the fossil
Size ratio implies scaling is isometric. However, many (but assemblage includes both an elephant and a hippopotamus,
by no means all) insular dwarfed ungulates have often been competition was set at level 1 for both. If there were an
noted for their comparatively large teeth and short metapo- elephant and deer, competition was set at level 2 for both.
dials (Gould 1975; Sondaar 1977). Thus, size ratios calcu- If there were two bovids and a deer intermediate in size
lated using teeth have higher SR values (i.e., a lesser degree between these bovids, competition was set at level 3 for the
of dwarfism) than values based on metapodials. Consequent- larger bovid, level 1 for the smaller, and level 2 for the deer.
ly, we performed separate tests for long bones and teeth. We We performed two separate analyses of covariance (AN-
tested for the influence of predation, competition, and an- COVAs) on SRs with competition and predation levels as
cestral body mass on SR. Most of the effects of predation factors, and log-transformed estimates of ancestral mass (on-
and competition depend on population densities, which is line Appendix 2) as a covariate to test whether mass, rather
unknown for fossils. Thus, we tried to model predation and than faunal composition, is a predictor of size change, as
competition independently of density effects, by categorizing predicted by the island rule (Lomolino 1985, 2005). Our SR
whether any degree of predation is likely to have prevailed. estimates are conservative because our samples may comprise
Predation was categorized by distinguishing three predation some individuals that were fossilized before the dwarfing
levels (PLs) of increasing intensity: (1) no predator known reached its final (and presumably greatest) extent. Further-
to be present, (2) mammalian predators not larger than more, Capasso Barbato and Gliozzi (1997) suggested there
1/10 the size of the species in question and/or large birds of was gene flow between mainland red deer and Capri Cervus
prey known to be present, (3) mammalian predators larger elaphus thyrrenicus, and Palombo et al. (2004) agued that