nal species that has been described as social (George and
Crowther 1981). Although no information is available on
group size, the population size at a six-hectare study site in
the Namib desert was sixteen individuals (Withers 1983).
Ctenodactylidae
The ctenodactylid family of rodents contains five species,
all of which are obligate rock-specialists (table 35.1). Cten-
odactylid rodents inhabit the semidesert and desert moun-
tainous regions of northern Africa (George 1974). All five
species live in social groups, although densities range from
0.3 gundis /hectare in the widely dispersed Mzab gundi
(Massoutiera mzabi) to 237 gundis /hectare in Speke’s gundi
(Pectinator spekei;George 1981). The common gundi has
been the most extensively studied of all five ctenodactylid
rodents. This species is thought to be a cooperative breeder
that lives in groups of up to twenty individuals (Gouat and
Gouat 1983; Nutt 2005). Each social group contains be-
tween one and three breeding males and females and their
offspring, the number of breeding individuals in a social
group increasing with group size (Nutt 2003). Preliminary
field observations of Val’s gundi (C. vali) and the Mzab
gundi suggest that these species live in paired social groups
(Gouat 1988a). The diversity of social structure exhibited
by ctenodactylid rodents provides an excellent opportunity
for comparative studies on the effect of density and patchi-
ness of rocky habitat on degree of sociality (Gouat 1988a;
Nutt 2005).
Hypotheses for Observed Levels of Social Behavior
in Rock-Dwelling Rodents
Of the twenty-one saxatile rodents listed in table 35.1, at
least fourteen (67%) have been classified as group living
(this includes species classified as social, as well as those that
have been described as living in family groups, extended
family groups, loose social groups, multifamily groups, or
groups that contain multiple females per male). Note that
the Namaqua rock mouse is not included within these four-
teen species, even though one study claimed that this spe-
cies lives in family groups. Furthermore, some “colonial”
species may also be social. Regardless, over half of all rock-
specialist rodents appear to live in groups. Lacey and Sher-
man (chapter 21 this volume) suggest that there are at least
seventy species of social rodents. According to this estimate,
the number of group-living petrophilic rodents represents
approximately 20% of all social rodents.
Several hypotheses have been put forward to account for
group living in mammals: (1) the phylogenetic constraints
hypothesis (Rowe and Honeycutt 2002), (2) the need for
extended parental care (Armitage 1999), (3) high predation
levels (Jarman 1974), and (4) the clumped distribution of
a limited resource (Emlen and Oring 1977). There is some
evidence to suggest that to some extent, each of these four
constraints may influence the degree of social behavior ob-
served in rock-dwelling rodents.The phylogenetic constraints hypothesis
It has been suggested that sociality within some rodent lin-
eages is constrained by phylogeny (Rowe and Honeycutt
2002, but see Trillmich et al. 2004 for rebuttal). Is it pos-
sible that shared ancestry leads to the observed levels of
social behavior in rock-dwelling rodents? Considering the
highly diverse taxonomic distribution of petrophilic ro-
dents (table 35.2), it is unlikely that similarities in social
behavior among taxa are governed by shared ancestry. Phy-
logeny may, however, influence social behavior indirectly
in one of two ways: (1) phylogeny may influence choice of
habitat, which may directly influence degree of sociality
(shared ancestry for habitat preference is exemplified by the
adaptive radiations of rock-specialist species in several ro-
dent genera; table 35.2), and (2) phylogeny may limit rather
than define the type of social behavior that is observed. For
example, the restricted availability of crevices in rock out-
crops forces bushy-tailed woodrat and Allegheny wood-
rat individuals into close proximity. However, unlike many
other petrophilic rodents, the rock-dwelling woodrats do
not form cohesive social groups. Instead, loose social groups
or colonies are formed, possibly because woodrats (Neo-
tomasp.) in general tend to be solitary (Nowak 1991) and
are not prone to exhibiting high degrees of social behavior.
Although phylogeny may minimally influence the degree of
social behavior observed in some petrophilic rodents, eco-
logical constraints are likely to have a greater effect on ob-
served levels of social affiliation.Need for extended parental care in extreme environments
For some rodents, group living seems to have evolved be-
cause of a need for extended parental care in harsh environ-
ments (Barash 1974a; Armitage 1999). Since many pet-
rophilic rodents also live in extreme environments, it is
possible that they too form social groups as a result of de-
layed dispersal and the need for extended parental care.
There are two lines of evidence to suggest that this may be
the case. First, many rock-specialist rodents are K-selected,
meaning that they have only a few young to which they give
a large amount of parental care. As examples, the rock cavy
has the longest gestation length, the smallest mean litter
size, and the lowest mean litter weight of all the caviinae
(cavies; Roberts et al. 1984), the punaré has the longest ges-
tation length by 50% and the lowest annual and lifetime
reproductive rates of any echimyid (spiny mouse; Roberts424 Chapter Thirty-Five