Bat-eared foxes ‘insectivory’ and luck: lessons from an extreme canid
Abstract and Keywords
This chapter presents a case study of bat-eared foxes in the Serengeti in order to shed light on the questions: how does their behaviour differ from that of other canids, and why? Topics discussed include group composition, dispersal, and philopatry; litter size, sex ratio, and reproductive success; energetics of female reproduction; rabies and mortality; and resources and sociality.
What underlies interspecific variation in the behaviour of the wild Canidae? Answers focus on two factors, the impacts first of phylogeny and second of ecological adaptations (Macdonald and Sillero-Zubiri, Chapter 1, this volume). In both respects, the bat-eared fox, Otocyon megalotis, is a revealing element in the pattern of canid variation. First, phylogenetically, although bat-eared foxes probably arrived in Africa as recently as the Pliocene, the clade of which they are the sole survivors split off from other modern canids long ago (Petter 1964; van Valen 1964; Wang et al. Chapter 2, this volume). Views differ as to exactly when they diverged— molecular evidence puts them at the base of the fox clade or even lower (Geffen et al. 1992e; Wayne et al. 1997), whereas morphologists place them closest to the gray fox, Urocyon, clade (Tedford et al. 1997); either way, the last time they shared a common ancestor with any other modern canid was more than 6 million years ago (m.y.a.), and perhaps closer to 10–12 m.y.a. A comparably long, distinct evolutionary (p.228) history is illustrated by the raccoon dog, Nyctereutes procyonoides (the subject of Chapter 13). Second, just as their ancestry is distinct, so too amongst modern canids the extent of their specialization on insectivory is extreme. Their dentition is unlike any other heterodont placental mammal because they possess between 1 and 4 pairs of extra molars (Coetze 1971; Clutton-Brock et al. 1976). Their teeth are small compared to those of other canids and, uniquely in the family, they have no carnassial shear. A sub-angular lobe at the insertion of the digastric muscle allows them to take 4–5 bites per second. Feeding on dung beetles, Scarabidae, links their fortunes to large ungulates (Malcolm 2001), but harvester termites, Hodotermes mossambicus or other termiter of the gonera Macrotermes or Odentotermes, are their most important food throughout their range (Nel 1978 1990; Nel and Mackie 1990; Wright 2004; Pacew 2000). Scarabidae, in particular, are conspicuously spatio-temporally heterogeneous in abundance within and between years—characteristics that are particularly relevant to one set of ideas seeking to explain inter- and intraspecific variation in sociality (Johnson et al. 2002; Macdonald et al. Chapter 4, this volume).
Against this background of phylogenetic and trophic extremism, we present a case study of bat-eared foxes in the Serengeti in order to shed light on the question, how does their behaviour differ from that of other canids, and why?
Study area and general methods
Between 1986 and 1990, one of us observed bat-eared foxes for a total of 2500 h in the Serengeti National Park, Tanzania (Maas 1993a). In the Serengeti, bat-eared foxes are common in open grassland and woodland boundaries but not short-grass plains (Lamprecht 1979; Malcolm 1986). Hendrichs (1972) recorded a density of 0.3–1.0 foxes/km2 in the Serengeti, but as many as 9.2 foxes/km2 in the breeding season, and 2.3 foxes/km2 at other times have been recorded in Botswana (Berry 1978). Groups forage as a unit and have home ranges from less than 1 km2 to more than 3 km2 that are sometimes overlapping (Nel 1978; Lamprecht 1979; Malcolm 1986; Mackie and Nel 1989). In Laikipia, Kenya, neigh-bouring pairs occupied ranges which overlapped widely (c. 20%) and averaged 3.3 km2 (Wright 2004). In the Seregenti's woodland boundary, and the open grasslands of southern and East Africa, insects are the primary food sources, with harvester termites (H. mossambicus) and beetles predominating, supplemented by smaller numbers of orthopterans, beetle larvae, and ants (Slater 1900; Shortridge 1934; Nel 1978; Lamprecht 1979; Berry 1981; Waser 1980; Stuart 1981; Malcolm 1986; Mackie 1988; Skinner and Smithers 1990).
The study area consisted of irregularly spaced Acacia trees in open bush country, forming part of the transitional boundary between Acacia tortillas woodland to the north and open grassland to the south. Rainfall is strongly seasonal with two peaks occurring during a rainy season from November to May (Sinclair 1979). Insect abundance is linked to rainfall (Waser 1980), and the onset of the rainy season in November is characterized by an explosion of insect activity. Harvester termite holes, vegetation, and ungulate droppings were surveyed during the dry season of 1987. A total of 2989 m2 samples were collected in 25 m intervals along 72.5 km of parallel transects, spaced 100 m apart. Transects extended across bat-eared fox territories as well as similarly sized adjacent areas.
General demographic data (group size and composition, litter size, and mortality) were collected from 16 groups in 1986, 19 in 1987, 18 in 1988, and 13 in 1989 (of these, 12 were studied continuously). However, sample sizes are smaller for some measures, depending on when they were taken, because of the formation and disappearance of some groups over the course of each year, particularly in 1987 and 1988, when rabies broke out. In 1987, rabies struck when the young foxes were nutritionally independent and affected 95% of family groups, but in 1988, cubs were less than 2 months old when rabies infected 68% of groups.
These bat-eared foxes were individually recognizable, and were observed at close quarters from a vehicle by day through binoculars, and by night using an image intensifier. Following Altman (1974), scan samples of all group members in view were recorded in 1-min intervals during daylight and 2-min intervals at night, together with proximity data for all individuals in view. Focal groups were watched weekly during at least one morning (05:00–08:00 h) and one evening (17:00 to 19:30–20:30 h). Beginning on the fifth day before each full moon, 8–10 6-h night watches were carried out between 18:00 and 07:30 h, to provide data equivalent to one composite night each month for each group.
(p.229) Group composition, dispersal, and philopatry
Bat-eared foxes in the Serengeti lived in family groups of one male and up to three females (Table 14.1), all of which invariably bred. Average breeding group size over 4 years was 2.44 (SE ± 0.09, n = 18), with 1.44 (± 0.09) females per group. However, in 1986, each of the 16 territories was occupied by just one pair, whereas during the following 3 years (1987, 1988, and 1989) 44%, 67%, and 54% of groups, respectively, contained more than one breeding female. This inter-annual variation in group size was statistically significant (x 2 9.24, df = 3, p < 0.002). Of 65 breeding events recorded, only one took place in a group that contained an apparently non-breeding adult, a yearling male (see below).
At about 9 months old, both males and females dispersed, whereas a proportion of females displayed natal philopatry, and this was apparently the only mechanism to join a group with a surviving female. There was one case of permanent male philopatry when a non-dispersing male remained following his putative father's death. Of 136 cubs that survived to dispersal age, 105 (77%) emigrated, almost all (95%) of them at the start of the dry season in June. The number of cubs that dispersed from a group was positively correlated with the number surviving to dispersal age (Kendall correlation coefficient, 1986: T = 0.7, z = 3.02, p < 0.01, n = 14; 1987: T = 0.74, z = 4.13, p < 0.0001, n = 18; 1988: T = 0.83, z = 4.46, p = 0.0001, n = 13). At the time of dispersal, group size averaged 6 (± 0.41, n = 18). Although variation between the 4 years 1986–89 did not reach statistical significance (x 2 = 1.17, df = 3, p < 0.07), group size at dispersal was smallest in 1989 (averaging 5.69 ± 0.38) when rabies struck while the cubs were small, and largest in 1987(6.21 ± 0.43) When there
Table 14.1 Average ± SE number of breeding females per group
Mean ± SE
2.5 ± 0.15
2.7 ± 0.18
2.5 ± 0.14
Most (93%) dispersers left the boundaries of their natal ranges, but seven known dispersers and one unidentified male formed four pairs that attempted neighbourhood settlement within one or both of their parents' territories. Of these, only one pair (both of whom could be identified) bred successfully for at least two subsequent seasons, displacing the female's natal group after a year. The newly formed pair had carved off a small part from the female's parental home range (SRI3, which had an extremely high termite density), and a large part of a neighbouring area (SR13B which, while not a breeding territory, had been occupied by a pair that had recently succumbed to disease). This sequence ended in one of five documented cases of adult dispersal. In this case, the occupants of SR13B— by then comprising the male, female and three almost fully grown cubs—finally took over SR13 after rabies had reduced the original group's membership from ten to just three. This followed a confrontation with the larger group in SR13B, whereupon SR13 settled in an unoccupied area about 1 km away; there, the surviving female was killed by a python, whereupon the male and their cub also disappeared. SR13B then took over almost exactly the original boundaries of SR13, and abandoned the peripheral area that had previously been their sanctuary.
Previously, one approximately 16-month-old subordinate female member of SRI3 disappeared after a period of escalating aggressive behaviour towards all other group members, especially her putative father, until the alpha-pair eventually drove her out.
In another instance, a male whose mate was killed by a vehicle apparently left the area and was not seen again. His territory was taken over by a newly formed pair the following year.
The only case of male natal philopatry involved a juvenile inheriting the territory at the end of the mating season in July, following the death then of his putative father; the male subsequently held his dead father's territory for three years. In another instance, a yearling male, after having left his family in October prior to the birth of the family's next (p.230) litter, rejoined his group for 6 weeks after an absence of two. He then groomed and guarded the new cubs and defended them against predators (he, and one cub, disappeared during the same night).
Case histories suggested that the female recruits to groups were invariably born there and one-third were dominant amongst their siblings. Of 30 philopatric female recruits, 18 (60%) became the dominant or sole female, whereas 12 (40%) became subordinate members of a group. Ten (55.6%) of these eighteen heiresses became dominant over their siblings, while eight (44.4%) lived in pairs, cohabiting with their father and in one case, their brother during the next mating season. All eight of the latter were seen to be mounted by the male kin with which they cohabited. Territory inheritance by a daughter was never accompanied by her father's emigration. Although full copulation was never seen, males attempted to mount all the females in oestrus, regardless of whether or not they were their daughters or sisters. During 1988 and 1989, a total of eight pairs consisted of an adult male and one of his adult daughters. The average number of cubs raised from these apparently incestuous matings was 2.88 (± 0.398), and was not significantly different to the number of cubs raised from potentially non-incestuous matings over the same period (2.5 ± 0.5; Mann–Whitney U-test: p = 0.5, z = 0.67, U = 19.5).
The dominant female's status was clearly defined by her close relationship with the male, rather than hostility to subordinates, and she was also the only female to urine-mark, either alone or in tandem with the male. The dominant female appeared to set the route when the group set out to forage in the evening. Of 30 females recruited into their natal groups, 18 inherited their mother's alpha status during the study, invariably following her death. The age at which these females acceded to dominance was 12.6 months (range 4–18, n = 10) in 1988 and 11.1 months (range 3–15, N = 8) in 1989; six females attained alpha status before they were 7 months old and before their siblings dispersed.
There was no relationship between recruitment and either numbers of cubs at emergence per family (Kendall correlation coefficient, 1986: T = 0.2, z = 0.98, p = 0.3, n = 14; 1987: T = 0.28, z = 1.45, p = 0.15, n = 17; 1988: T = 0.2, z = 0.96, p = 0.3, n = 13) or cub survival to dispersal age (1986: T = 0.12, z = 0.61, p = 0.5, n = 14; 1987: T = 0.22, z = 1.16, p = 0.2, n = 17; 1988: T = 0.19, z = 0.91, p = 0.4, n = 13).
Litter size, sex ratio, and reproductive success
Bat-eared foxes become sexually mature at 8–9 months of age and mate for life. Pair-bonding and mating take place from July to August with up to 10 copulations per day for several days (see also Rosenberg 1971), and with a copulatory tie lasting c. 4 min, followed by peculiar post-copulatory play (Le Clus 1971). Bat-eared foxes have one litter, of up to six cubs, per year, with births occurring from October to December (Nel et al. 1984), following a gestation period of 60–75 days.
To judge by the male's behaviour, in the Serengeti, all adult females in each group came into oestrus within a few days of each other, and they all bore and raised cubs, and lactated. Breeding dens contained different size classes of cubs, and the maximum number of distinct sizes always corresponded to the number of adult females.
Average litter size at emergence 8–12 days after parturition was 2.56 ± 0.13 (n = 90), but this average disguises substantial inter-annual variation. In 1986, when all territories were occupied solely by pairs, the single female per territory bore an average of 4.43 cubs ( 0.36, n ± 14 pairs), significantly more than the average litter size for 1987–89 (Fig. 14.1; Wilcoxon
Assuming that the single adult male associated with each group fathered all the cubs, then male reproductive success was highest in 1986 (averaging 4.43 cubs). During the years of lower productivity, from 1987 to 1989, males co-habiting with groups of females enjoyed a higher reproductive success (averaging 3.9 ± 0.4 cubs, n = 15) than did males in pairs (2.7 ± 0.2 cubs, n = 17; Mann–Whitney U-test: p < 0.01, z = −2.61, U = 194.5, n = 31).
There was a general shift from an unbiased offspring sex ratio in 1986 to a female-biased sex ratio over the following 3 years. Of a total of 198 cubs from litters of which all members were sexed at emergence, 58% were female (Binomial test: n = 198, z = −2.06, p < 0.05). However, this overview disguises a marked shift in sex ratio following emergence which swung from parity in 1986 (52% of 58 cubs were female), to a strong female bias between 1987 and 1989 (67% of 140).
Parental care and parent–offspring proximity
While the females gave birth inside the den, males were usually resting at one of the den entrances. Newborn cubs spent their first 10 days or so inside the natal den accompanied by their mother. Consequently, we know nothing of their number or sex ratio at birth. After 8–12 days, cubs appeared at the den entrance and began to explore the den area and, later, its immediate vicinity. Males were never seen to regurgitate to their mates or the cubs. However, following cub emergence, mothers spent increasing time away from the cubs whereas males took over all parental duties, with the obvious exception of nursing.
Males guarded, groomed, carried the cubs, and played with them. By day, males rested at one of the den entrances while females typically slept slightly further away, often hidden from the cubs' view by a small bush or tuft of grass. Thus, when the cubs emerged they would always encounter the male first. When sleeping outside the den, and until they were more than 3 months old, cubs invariably slept next to the male. Males played with and groomed the cubs assiduously, and from the age of 4 weeks cubs reciprocated; in contrast, until 3 months old, amicable socializing such as grooming and playing were rare between females and cubs. Females almost always rejected approaches by the cubs, walking away from them, sometimes snarling, growling, and even snapping at their young.
(p.232) By night and while the cubs were young, males would remain at the den while females foraged. Occasionally males would leave the den together with the female and forage for 15–45 min. Thereafter, the male would remain at the den, or forage in its immediate vicinity, until the female returned 2.5–6 h later. Then, while the male foraged, the female would stay with the cubs for 15 min to 2 h until the male returned. Female guarding shifts were always shorter than those of males.
Once the cubs were 4 weeks old, males led them on their first foraging trips, during which they kept close to the male, which frequently indicated or passed items of food to them. Usually males would indicate food, be it carpets of foraging Hodotermes, or patches of active dung beetles. Occasionally, there were large dung beetles, which the male crushed in his mouth for the cubs. Both males and females accompanied the cubs on foraging trips once they were approximately 3–4.5 months old.
Sex differences in parent–offspring proximity
Spatial geometry can reveal social structures (e.g. Macdonald et al. 2000), and this was so for the tendency of cubs to spend time (recorded during 2-min scan samples), at given proximities to male and female adults. Between weeks 3 and 12 (at weaning age), males spent a significantly greater proportion of time in all proximity ranges than did females (t-test, < 1 m: t = 9.66, p < 0.0001; < 5 m: t = 6.26, p > 0.0001; < 20 m: t 5.33, p < 0.001; < 50 m: 6.61, p < 0.0001; Fig. 14.3), spending 90% of their time, on average, within 50 m of the cubs, compared to the females' average of 56%. This difference was comparably significant (t > 5.03, p < 0.001 in all cases) whether the foxes were members of a pair or a group, and whether, within groups, the females were dominant or subordinate. Males cohabiting with a group of females spent more time with the cubs (t > 2.31, p < 0.05 for all proximity ranges) than did those living as a pair, but there was no such distinction between females from groups and pairs. Within groups, subordinate females spent a significantly greater proportion of time in the 5 m (t = − 2.84, p < 0.05) and 20 m (t = − 3.42, p < 0.01) range of cubs than did dominant females. Time spent within 1 m of the cubs appeared to be exceptionally demanding,
Energetics of female reproduction
Female bat-eared foxes suckled their young for up to 14 weeks. In groups with more than one female present, only one female ever nursed cubs at a time, and when she did so all the cubs nursed. During periods of the day when all group members including females were present in the vicinity of the den together, while one female suckled the others would remain at a distance and lying down, rendering their teats unavailable. Suckling bouts were infrequent, typically occurring in the early evening prior to the female's departure from the den, and in the morning on her return, as well as during nightly guard shifts, and occasionally in the early afternoon. Suckling was usually initiated when the female summoned the cubs from the den or its vicinity with a soft whimpering cry. The average suckling bout lasted 3 min (191.11±8.06 s, n = 18 females and 61 bouts). There were no significant differences between years in suckling bout length, for paired females (1986 versus 1987: t = 0.22, df = 5, p > 0.05), dominants (1987 versus 1988: t = − 0.32, df = 8, p > 0.05), or subordinates (1987 versus 1988: t = 0.19, df = 11, p > 0.05). However, as revealed in Table 14.2, there were significant differences in suckling bout length between females categorized by group size and status (p.233) (ANOVA: df = 5, p < 0.001, n = 29). Thus, within groups, subordinates suckled for longer than did dominants (t = −6.48, df= 18, p = 0.0001), and female members of a pair suckled for longer than did the dominant female within a group (t = − 4.03, df = 11, p = 0.002).
Sucking time may be proportional to milk intake. The product of sucking time and litter size gives a measure of the drain on the nursing female (referred to as ‘cub time’ on Table 14.2), and by this measure, too, subordinates invested more than did dominant females overall (Mann–Whitney U-test: p = 0.03, z = − 2.19, U = 79, n = 20), and more than did paired females in 1987 (Mann–Whitney U-test: p = 0.05, z= − 2, U = 12, n = 8) but not in 1986.
Both sucking bout length and cub time were significantly positively related to H. mossambicus foraging hole density in bat-eared fox territories (Kendall Rank correlation coefficient; suckling bout length: p = 0.05, z = 1.96, T = 0.8, N = 8; cub time: p = 0.05, z = 1.96, T = 0.8, N = 8).
What are the costs of reproduction to female bat-eared foxes? First, an estimate of the energetic cost of pregnancy can be obtained by relating litter size to maternal body weight (Gittleman 1986; Oftedal and Gittleman 1989). Second, an estimate of the energetic cost of nursing at peak lactation can be estimated from litter metabolic mass (LMM = litter size × cub weight0.83) by multiplying LMM by 227, and an index of maternal energy investment is provided by the ratio of LMM to maternal metabolic mass (W0.75) (Oftedal and Gittleman 1989). An LMM ratio indicates a high energy demand relative to maternal metabolism and, as argued by Oftedal (1984), indicates high expenditure of energy.
Based on the weights of five bat-eared foxes in the Serengeti, we use a maternal metabolic mass of 3.50.75 kg. Neonate weight was estimated at 120 g
Table 14.2 Average sucking bout and cub time durations for paired females in 1986 and 1987, and for dominant (mother) and subordinate (daughter) females in groups in 1987 and 1988 (no focal observations were made on pairs in 1988)
Mean sucking bout length
Mean cub time
Maternal weight (kg)
Litter metabolic mass (kg0.83)
Weight of young (kg) at peak lactation
Litter weight as % maternal weight
Daily milk energy output (kcal)
Metabolic mass ratio
Rabies and mortality
Mortality levels were significantly higher in 1987 and 1988 than in 1986 and 1989 (G = 55.72, df = 3, p = 0.0001), due to rabies outbreaks in those years. Rabies killed 85 (90%) of the 94 individually recognizable animals that died over the 4-year study, primarily in 1987 and 1988. The first rabies cases in 1987 were in early February when cubs born in 1986 were being weaned, but in 1988, rabies broke out in November, while the females were still suckling. Female mortality from rabies was significantly higher in 1988 (71% of 31 females died) than in 1987 (when 27% of 26 females died; X 2 = 10.98, df = 1, p = 0.001), but cub mortality was slightly lower (44% of 70 cubs in 1987 versus 36% of 61 cubs in 1988), although this difference was not statistically significant (X 2 = 0.914, df = 1, p = 0.339). Mortality in 1986 and 1989 was considerably lower for both cubs (approximately 3% mortality in both years, out of 62 and 41 cubs, respectively) and females (no mortality in 16 females in 1986 and approximately 11% in 20 females in 1989). No male mortality was recorded in 18 and 13 individuals in 1987 and 1989, respectively, but in 1986, 1(6%) of the 16 males died (from rabies) and in 1988, 5 of 18 males died, 3(16.7%) of them from rabies. Male mortality from rabies was significantly less than female mortality (1987: X 2 = 8.69, df = 1, p = 0.003; 1988: X 2 = 13.47, df = 1, p < 0.0001).
Deaths were diagnosed as rabies on the basis of the presence of inclusion bodies in the brain and tested positive using Indirect Fluorescent Antibody Technique, by the Tanzania Livestock Research Organization; the remainder were inferred.
Other causes of mortality were predation (three individuals) and road accidents (three individuals). Predators, all of which ate at least some of the foxes they killed, included martial eagles (Polemaetus belicosus), pythons (Python sebae), and spotted hyaenas (Crocuta crocuta). When a bat-eared fox was attacked by a martial eagle, the fox uttered an alarm call, whereupon five other members of its group charged towards the eagle. Together the foxes snapped and lunged at the bird, leaping into the air as it laboured to take-off. The foxes mobbed black-backed jackals (n = 22), golden jackals (n = 12), and spotted hyenas (n = 45), and occasionally aardwolves, mongooses, goshawk, and various snakes (total n = 19). An average of 2.8 ± 0.32 (range 1–7) foxes was involved in 22 mobbings of black-backed jackals, and 4.62 ± 0.36 (range 1–11) mobbed hyaenas, on average. These figures include instances where more foxes were available to mob the smaller predators, and thus the number participating can be interpreted as a measure of their assessment of risk. Groups of bat-eared foxes mobbing hyaenas were significantly larger than those mobbing jackals (Mann–Whitney U-test: U = 1119, z =−3.56, N = 79, p < 0.001); there were no differences in the size of mob harassing the two species of jackals (U = 135, z = − 0.111, n = 34, p = 0.9). Occasionally, when the cubs were threatened by hyaenas, the dominant male fox attempted alone to distract and lead them away.
Resources and sociality
Bat-eared foxes were clearly territorial. Territories sketched onto maps averaged 0.62 km2 (range 0.4–0.87, n = 6), and appeared to tessellate neatly. Although movements were not studied in detail, territorial borders appeared to be stable between years. There was a negative relationship between Hodotermes foraging hole density and territory size (Fig. 14.4) (Kendall's tau = − 0.867, z = − 2.442, p = 0.01).
Areas occupied by bat-eared fox territories were clustered in dispersed pockets. Both H. mossambicus (p.235)
Rainfall measured monthly at a rain gauge within the study area varied inter-annually. Rainfall in the whelping season was strongly correlated with scarabid abundance. The highest rainfall in a whelping season fell in 1986, when 107 mm fell in October, followed by 3 years of lower rainfall in this month (13, 64, and 4 mm in 1987–89, respectively).
When feeding on termite patches, group members fed closely together, but when feeding on beetles, beetle larvae, or grasshoppers, they foraged up to 200 m apart. Group members called each other to rich food patches with a low whistle.
Group size, philopatry, and territory quality
There was no significant relationship between the number of adult females per group and Hodotermes hole density (Kendell correlation coefficient: T = 0.15, Z = 0.474, p = 0.6, n = 7). However, this result may be confounded by high rabies mortality in 1987 and 1988, and a small sample size. Female
Reproductive success and territory quality
Between 1986 and 1989, reproductive success (measured as average number of cubs per female at emergence) was positively correlated with Hodotermes foraging hole density measured in 1987, and this (p.236) relationship persisted during 1987–89, years of lower October rainfall (although total annual rainfall was not particularly low in these years), with female reproductive success positively correlated with Hodotermes foraging hole density in 1987 (Kendall rank correlation coefficient: p = 0.03, z = 2.16, T = 0.683, N = 7). During this period, reproductive success varied negatively with the number of females per group (Kendall rank correlation coefficient: p = 0.02, z =− 2.54, T =−0.51, N = 14). The cumulative consequences of these relationships are that male reproductive success was positively correlated with number of females per group (Fig. 14.6).
Social dynamics and territory quality
One obvious line of thought is that the females' reluctance to associate with cubs stemmed from the harrowing demands of foraging—which consumes their time and limits their energy. To shed light on this, one pair was selected—the female had spent less time close to her cubs than had any other in 1986—and in 1987, she was provided with commercial dry cat food, raisins, and water for 12 weeks after whelping. Despite the fact that this period of 1987 was, for most foxes, one of substantially scarcer resources than it had been in 1986, the provisioned female spent a significantly greater proportion of her time in close proximity to her cubs in 1987 (<1 m: 11% versus
These observations are complemented by the predicament of a female whose mate died before the birth of her cubs in 1986. As a single mother, she spent significantly more time within 5–30 m of her cubs than her peers in 1986 did with theirs (<5 m: t = 7.56, p < 0.0001; <20 m: t = 4.68, p < 0.01; < 30 m: t = 5.36, p < 0.001); in contrast, she did not spend significantly more time within 1 m of her cubs than did other mothers (< 1 m range: t = 2.05, p = 0.05). Indeed, the single mother spent a significantly lower proportion of time (24.2%) in the < 1 m range of her cubs than did pair-living males (39.7%; t = 2.99, p = 0.01), but otherwise her proximity scores did not differ from those of males that year. By 1987, that individual had recruited daughters to form a group along with her philopatric son and her behaviour towards her next litter of cubs was statistically indistinguishable to that of other females in her circumstances that year.
Five questions of general theoretical interest emerge from these observations of one population of bat-eared foxes. In logical order these are, first, how do the energetics of insectivory and the foxes' life-history parameters interact to fashion their lifestyle? Second, how is that lifestyle moderated by patterns in the dispersion of their prey? Third, within the society that emerges from these foregoing considerations, what dictates their systems of mating and parental care and, fourth, sex ratio allocation? Finally, what lessons have emerged from this study with regard to the influences of phylogeny, resources and luck (as exemplified by the uncontrollable vicissitudes of rainfall and disease) on the biology of bat-eared foxes?
Energetics and insectivory
Body size is associated with various allometries and morphological constraints. In the case of canids, it is also associated with variation in social systems, (p.237) running on a continuum from small species with female-biased groups and predominantly male dispersal to larger species with male-biased groups and predominantly female dispersal (e.g. Macdonald and Moehlman 1982; Creel and Macdonald 1995). As summarized by Macdonald and Sillero-Zubiri (Chapter 1, this volume), this continuum is explained in terms of the effect of body size on neonate development by Moehlman (1986, 1989) and Moehlman and Hofer (1997), whereas Geffen et al. (1996) favour instead resource-based explanations. Similarly, and with consequences reviewed by Macdonald et al. (Chapter 4, this volume) body size is linked via metabolic rate to energetic demands, leading Carbone et al. (1999) to argue that the relatively low availability of small prey limits the size of carnivores that they can sustain. In addition to the size-related allometry in metabolic demands imposed by surface-to-volume ratios, some species (e.g. African wild dog) have higher energy demands than predicted by their mass (Gorman et al. 199n).
In this context, no canid lives more exclusively on prey smaller than the insects to which bat-eared foxes are committed. While their gestation costs may be unexceptional, extrapolations from our observations of litter size and suckling bout lengths suggest that the energetic demands on lactating bat-eared foxes may be even higher than expected (high metabolic mass ratio and high estimated milk energy output at peak lactation). This situation may explain (1) their lower litter size following poor rainfall and (2) their parental behaviour.
We know of no other canid in which females are so determinedly stand-offish towards their emergent offspring, or in which males are so correspondingly paternally diligent, and we attribute both to energetic brinkmanship of the females. Because of the small and fiddly nature of insect prey, and bearing in mind that bat-eared foxes do not regurgitate, the only way to transport energy to cubs is in milk; the entire burden must be born by the female which, if Carbone et al. (1999) are correct, is already likely to face severe time constraints. Gittleman and Thompson (1989) have emphasized the necessity for rest and sleep for females during lactation (McNab 1987).
The extent to which male bat-eared foxes monopolize the task of parental care is at an extreme for canids (and indeed for mammals, Kleiman and Malcolm 1981). Komers and Brotherton (1997) suggest that paternal care is less relevant to the evolution of monogamy in mammals than in the dispersion of females. The effectiveness of paternal care in mammals is largely unquantified (but see captive studies by Wynne-Edwards and Lisha 1989; Cantoni and Brown 1997, and observation in the wild by Gubernick and Tejeri 2000). By spending most of their time with the cubs, males enable females to maximize their foraging time. Consequently, their own foraging time is curtailed, and studies elsewhere have revealed that they have smaller ranges than do lactating females (Mackie and Nel 1989) and spend less time actually feeding when accompanied by cubs (Nel 1990). In our study, males from multi-female groups spent a greater proportion of time close to cubs than did males from pairs, indicating advantages resulting from better food availability and/or labour-sharing between group members. The uniformity of maternal behaviour—irrespective of group size or litter size— suggests females have little room for manoeuvre, an interpretation supported by the atypical motherliness of the one female, given supplementary food and water. Although in a different ecosystem, Wright (2004) found that the proportion of bat-eared fox cubs nursing to 14 weeks increased with the proportion of time males spent at the den, which itself was related to measures of termite abundance with territory.
Communal care of the young may also allow females to conserve energy and share the burden of lactation (e.g. by increasing the interval between suckling) (Creel and Creel 1995). The costs to female canids of sharing a male (and therefore paternal care for the young) may often be primarily associated with the capture of prey and feeding of the young— two features not applicable to the natural history of these foxes, whereas the benefits of vigilance and guarding are non-depreciable.
Although they rarely eat mammalian prey (Lamprecht 1979; Skinner and Smithers 1990), bat-eared foxes are nonetheless dependent on mammals insofar as the distribution of their insect prey is (p.238) heavily dependent upon the dung of large ungulates (Malcolm 2001). In the Serengeti, the presence, and grazing activity, of ungulates (and thus of their dung) is heterogeneous, and therefore so too is the dispersion of insects. Seasonal migrations bring the herds to the Serengeti for only the wet season, commencing in October (McNaughton 1989), primarily in search of phosphorus (Murray 2001). Once there, fine scale variations in soil and vegetation type cause them to graze patchily (Snaydon 1962; McNaughton and Banyikwa 1995), and the potential for insect communities to develop, sustained by their dung, is correspondingly patchy. The extent to which that potential is realized is determined by rainfall. As described above, when, as in 1986, rainfall is high between October and December, scarabid availability soars, whereas when the rains during these months following whelping are poor, as in 1987–89, so too is the supply of beetles and harvester termites' surface activity. Waser (1980) confirmed a direct link between rainfall and the availability of insect prey (see also Nel 1978, 1990). Locally, at least in some areas, numbers of bat-eared foxes can fluctuate from abundant to rare depending on rainfall and thus food availability (Waser 1980; Nel et al. 1984), and climate change in southern Africa has been invoked to explain the species' range extension (e.g. Stuart 1981; Macdonald 1982; Marais and Griffin 1993).
At the broad scale of their settlement throughout the ecosystem, the distribution of bat-eared foxes in the Serengeti was highly clustered into island communities, whose location matched the patchiness in insect availability driven by ungulate behaviour and thus, ultimately, interactions of soils, minerals, and vegetation. Families of bat-eared foxes occur in clusters in areas where harvester termites are present and are absent from areas lacking termites. Insofar as the territories in such clusters are occupied (a situation altered by epizootic disease), the prospects for dispersal may be limited—a situation analogous to that of an Ethiopian wolf facing dispersal from a mountain plateau (Sillero-Zubiri et al. Chapter 20, this volume).
At the fine scale of social and spatial arrangements within and between territories, the spatio-temporal heterogeneity in the dispersion of insects available to bat-eared foxes in the Serengeti creates almost exactly the conditions proposed by the Resource Dispersion Hypothesis (RDH) to facilitate social group formation (Macdonald 1983; Carr and Macdonald 1986; Macdonald et al. Chapter 4, this volume). Specifically, Macdonald and Carr (1989) argue that two ecological pressures will affect fundamentally the balance of costs and benefits of group membership: (1) the probability of successful dispersal and (2) the probability of accessing adequate resource security as a group-member (see Johnson et al. 2002). Watching the foxes, it appeared that food availability was patchy in that the foraging party would clearly make beeline journeys to particular foraging areas. Where resources are spatio-temporally heterogeneous, additional members may share the smallest territory required by the basic social unit (in canids generally a pair) at minimal cost, and if the risks of dispersal are high it may be advantageous to do so. That advantage may be enhanced if group membership brings sociological benefit—for example, through cooperation—and therefore the question of whether each additional group member is desirable (from the different perspectives of the candidate and the existing members) will be affected by the marginal advantage arising from its admission. If that marginal advantage is great, it may even pay to bear the costs of expanding the minimum necessary territory to accommodate the resource requirements of the new member (Kruuk and Macdonald 1985). Do these ideas help interpret the behaviour of the bat-eared foxes described here?
Assuming each H. mossambicus foraging hole is an equivalent measure of resource availability, all else being equal, each pair of foxes will require the same number of holes. RDH predicts that where these holes are widely dispersed, territories will be larger than where the holes are densely packed. As predicted, territory sizes were inversely related to the densities of Hodotermes foraging holes. Group size was not correlated to H. mossambicus foraging hole density (although this result may be confounded by the impact of rabies), but both the frequency of philopatry and reproductive success were. RDH predicts a correlation between group size and patch richness, but in this case H. mossambicus availability (including their often high renewal rate) was only one component of this richness—the other was scarabid availability.
On the basis of these results, we suggest that the following model might be tested by future study. (p.239) First, territory size is dictated by the availability of H. mossambicus, and territories are configured to sustain a pair of foxes under the likely worst conditions. We expect the surface abundance of H. mossambicus to increase following the grass growth stimulated by good rainy seasons and to wane as their food-stores deplete through prolonged drought. Between 1986 and 1989, bat-eared fox territories appeared stable. Whether territory sizes adapt to a fresh bottleneck each year, or are adapted over a sequence of years, the question arises of when is the bottleneck? We suggest that it is the dry season. At this time alternative prey are rare and the foxes depend almost entirely on Hodotermes. The availability of termites to foxes (i.e. the time they spend foraging above ground) is likely to depend on the supply of dried grass to be harvested, and this in turn depends on the rainfall during the preceding rainy season and current dry season.
Second, RDH predicts that group size will be determined by patch richness (e.g. the aggregate availability of Hodotermes and scarabids). Of these, variation in the availability of scarabids is conspicuous, and determined by wet (whelping) season rainfall. The substantially higher mean litter size in 1986, when October rainfall was high, suggests that cub survival is determined by the availability of scarabids and, following the rapid growth of grass, of surface active harvester termites (the mean litter size for pairs in 1986 was significantly higher than the mean litter size for pairs for 1987–89).
Whether these recruits remain in the group to become breeders during the following year will depend on the number of them surviving predation and disease, and whether the suggested dry season bottleneck in surface-feeding termites is sufficiently lenient to sustain them. We could not test the latter prediction readily because rainfall during the dry seasons within our intensive study was not particularly variable. However, remembering that in October 1986 all the foxes were settled in pairs, we predicted that the previous dry season must have imposed a stringent bottleneck. Indeed, rainfall in June 1986 was only 8 mm (well below the 29-year average of 34 mm), and this savage drought persisted until the end of September. Sinclair (1975) states that 23 mm of rain monthly is required to sustain grass growth, and termite surface activity is related to grass growth. The foregoing model, although speculative, suggests other ideas for future field test. Since the density of termite holes is seemingly not related to fox group size, whereas group size is negatively related to reproductive success, it is likely that the smaller litters produced by group-living females (in comparison to contemporary pair-living females) has a sociological explanation. Indeed, communal denning in Ethiopian wolves (Canis simensis) (Chapter 20, this volume), Arctic foxes (Alopex lagopus) (Chapter 8, this volume), and perhaps bush dogs (Speothos venaticus) (Macdonald 1996) and red foxes (Vulpes vulpes) (Macdonald 1987) have all been associated with reduced litter sizes.
In 1986, associated with high October–December rainfall, food availability appeared to be high during the time of lactation, in marked contrast to subsequent years when, by the time of emergence, litters were half the size of those in 1986. This raises the question of whether this litter-size reduction occurred pre- or postpartum, and in either case by what mechanism. The model proposed above would rationalize litter reduction as a consequence of reduced rainfall leading to a catastrophic delay in the appearance of scarabids, plausibly translating into a lowered capacity to generate milk—suckling bout lengths were longer in territories with high Hodotermes scores. The relevance of water is further supported by observations of nursing females licking dew from blades of grass.
We do not know when or how the litter reduction took place, but in other species, females may reduce their litter sizes both pre- and post-natally under suboptimal conditions (Harvey et al. 1988; Lindström 1988; Clutton-Brock et al. 1989; Packer et al. 1992). Smithers (1983) reports that bat-eared foxes characteristically carry 4–6 embryos (they have six teats). We have no evidence that during pregnancy the females can forecast the likely rainfall that will influence her food supply during the crucial months of lactation. To that extent, we turn to postpartum litter reduction. By what mechanism might this occur? Insofar as all emergent cubs looked healthy, general starvation seems implausible, raising the possibility of infanticide. Since the father almost never entered the den prior to cub emergence, and since litters were small in 1987–89, even in territories with only one breeding female, as well as in groups, the mother is a strong candidate. Insofar as litter reduction occurs, (p.240) and if it is the means of sex ratio variation, it suggests selective infanticide. Two anecdotes are noteworthy: only six cubs can nurse simultaneously from one female, and even in groups the mothers suckled at different times; in only two cases did an aggregate litter in excess of six emerge from the den, and in both cases the ‘surplus’ cub disappeared within days. Intriguingly, Smither's (1983) notes that his pet bat-eared fox had only four teats (he was unaware that this was anomalous) and gave birth to litters of 4–6 cubs—on all four occasions when her litters were larger than four, but not otherwise, she ate any cubs in excess of four within days of their birth (and thus before investing substantially in nursing them).
Mating system and parental care
Amongst group-living canids, the generality is for reproduction to be the prerogative of only one, dominant female in each group, and for non-breeding subordinate females to act as alloparents (Macdonald et al. Chapter 4, this volume). Where more than one female does breed within a group, there are cases of communal denning and allosuckling, but again, the generality is for signs of hostility (sometimes fatal for the pups) between the mothers (Creel and Macdonald 1995). More broadly amongst carnivores, all or several females within groups of felids (e.g. Bertram 1979), hyaenids (e.g. Kruuk 1972), and some herpestids (e.g. Rood 1986) may reproduce communally—all families within the feliform branch of the Order. The bat-eared foxes of the Serengeti are thus a significant outlier in canid sociality in that all females bear cubs, den comunally, and allosuckle indiscriminately.
What selective pressures favour communal breeding by these foxes? Key observations are that (1) bat-eared foxes have six teats, (2) even in groups of three females, only one female at a time ever nursed the cubs, and (3) even the largest aggregate litter never exceeded six for more than 3 weeks of age. The critical question is why all three females in our groups produced about two cubs each, rather than the dominant producing six which were nursed by the other two. Are there reasons why (1) a dominant female should be unable to suppress the reproduction of her subordinates and (2) it would be advantageous for all females to breed?
Possibly, reproductive suppression and spontaneous lactation, may, phylogenetically, not be open to bat-eared foxes, as it seems that regurgitation is not. The only way for these foxes to transport nourishment to their cubs is as milk. A hypothetical dominant female might thus gain by allowing her female group-mates to bear cubs, which she then kills so that the bereaved mothers nurse her off-spring. This did not happen (although it does in other canids, see Chapter 4). One possibility is that the act of infanticide is simply too risky—within the confines of a shared den, the subordinates might resist with the result that the dominant loses control and her own cubs suffer, and infanticide could proliferate to a tragedy of the commons. Several factors might erode the advantage to the dominant of monopolizing reproduction. Most obviously, the closeness of her relatedness to the subordinates' cubs (see below, inclusive fitness). A further clue may lie in the highly structured pattern whereby, even though two or three females were at the den, only one at a time would suckle the cubs; this may indicate optimization of either or both of milk yield and consumption. Most interestingly, subordinate females nursed for longer, and guarded more frequently, than did dominant females—perhaps reflecting that the dominant's status puts other drains on her time and energy, and perhaps indicating that she benefits from the greater assiduousness of her subordinate daughters and grand-daughters—benefits that might translate into her own lifetime reproductive success.
Benefits of cooperation
Our account reveals that potential benefits of group-formation in this population of bat-eared foxes include the added production of related cubs, matrilineal territory inheritance, collective nursing, grooming and general care of the cubs, corporate defence (and vigilance) against predators and defence of the territory, and huddling for warmth. How do these behaviours affect the balance of costs and benefits of increased group membership. On the one hand, females in groups produced smaller litters than did those in pairs. On the other hand, those in larger groups can share vigilance, maternal antibodies, and other duties, and if one is killed another is at hand to nurse her orphans. A further speculation (p.241) is that communal nursing is a form of risk-sharing in the face of an uncertain prey supply and high mortality rate. The costs of tolerating additional group members—a factor emphasised by Macdonald and Carr (1989)—may be unusually low in bat-eared foxes. To a dominant female, the cost of sharing a male (who guards a den with equal assiduousness whether it contains one litter or two) or food (which is often highly renewable) may in some respects be low.
Territory inheritance and inbreeding
Territory inheritance can provide an advantage, from the perspectives of both parent and offspring, to natal philopatry. However, it is often supposed to involve a long wait, and to bring with it risks of inbreeding. In this case, the wait was short and the probability of inheritance high. Although the widespread evidence of multi-male mating in mammals (Wolff and Macdonald 2004) raises the expectation that extra-pair copulations will be found in bat-eared foxes, the frequent observations of fathers mounting their daughters in our study suggest—but do not prove—that father–daughter or sibling inbreeding may have been the norm in this population. In a different population, where bat-eared foxes invariably lived as pairs, Wright (2004) found confirming evidence of cubs fathered by extra-pair males in only 2 litters out of 14. He also noted that the close proximity maintained within each pair could have made philandering difficult. This may, of course, be different in circumstances where there is more than one female in a group.
Clearly, one factor that may contribute to cooperative behaviour of the bat-eared foxes is their genetic relatedness. In this context we report two systems. The most simple is the case where an unrelated males with a single female or with two sisters. In this case, the coefficient of relatedness between parents and cubs is 0.5, and between aunt and nieces/nephews is 0.25. However, in some years (1987–1989), some foxes lived in groups comprised of one adult male with two females, and a number of cubs. In these groups the male's relatedness to any observed cub can arise in two ways. Observations in those years suggested that for 80% of matings (12/15) the male was also the father of its mate, and therefore simultaneously the grandfather of the cub (in some groups this was true for all cubs, as both females were the daughters of the male). For these matings, the male/cub relatedness will exceed 0.50 (indeed, because the coefficient of relatedness between grandparent–grandoffspring is 0.25, the male/cub relatedness is estimated to be 0.75). The same applies to those cases where a female is related to her cub as its mother and through incestuous mating with her father. Furthermore, the coefficients of relatedness of communally nursing females to each others cubs may also be elevated—for example, that of a female to her sister's cub may be 0.5 (0.25 as their aunt and 0.25 through their shared father). In short, the relatedness between all adults in these groups, and the cubs, may be unusually high. A different point is that the evidence of the neighbourhood settlement suggests that the members of adjoining territories may also have kinship ties.
Whereas sex ratio of cubs at emergence was equal in 1986 (the breeding season of peak food availability), a preponderance of daughters emerged during the failed rainy seasons on 1987–89 when food was less abundant. Departures from equal sex ratios among offspring can be explained when the determinants of fitness vary between the sexes (Clutton-Brock 1991). Thus, if traits affected by parental investment influence the fitness of one sex more than the other, it may pay parents to invest more in the sex expected to produce the most grandchildren per unit investment. That investment can occur during pregnancy or in the den or thereafter. Although we have no measures of differential parental investment in the sexes at any of these stages, we can observe its outcome in the sex ratio of cubs at emergence. Although the preponderance of daughters emerging in each year from 1987 to 1989 might be explained by factors beyond the control of the parents, for example, a disease afflicting males, we turn to two hypotheses that seek to explain such variation in terms of parental investment. First, Trivers and Willard (1973) hypothesized that for species with significant sexual dimorphism biased towards males, such that the costs of producing sons exceed those of daughters, only qualitatively (p.242) superior mothers can afford to rear sons, while inferior mothers should produce daughters. Alternatively, the local resource competition hypothesis (Clarke 1978; Silk 1983; Julliard 2000) predicts that mothers in poor condition (in low quality territories) should produce offspring of the sex that is most likely to disperse from the natal area to reduce competition for resources in that locality. Alternatively, mothers in good condition (in high quality habitats) should produce the most philopatric sex. Both hypotheses make assumptions. The crucial ones for Trivers and Willard is that sons are more costly to produce, and that greater investment in them may reap rewards later. For local resource competition, the assumptions are that resources are spatially variable and that reproductive success better in some patches than in others; furthermore, dispersal rate should differ between males and females. All these conditions would seem to apply, although the modest sexual dimorphism characteristic of bat-eared foxes makes the extra cost (and value) of hefty sons the least clearly supported. Overall, and in the Serengeti, males (4.06 kg) are heavier than females (3.9 kg) (Gittleman 1983, 1989), although, in a sample from Botswana, females weighed marginally more than males (male: 4.03 kg, range 3.4–4.91, n = 22; female: 4.11 kg, 3.18–5.36, n = 29; Smithers 1971). The predictions of these two hypotheses are opposite. In breeding seasons when food was short, the foxes produced predominantly daughters—in accord with the offspring quality hypothesis and contrary to the local resource quality hypothesis. Amongst carnivores, this accords also with findings for badgers (Meles meles), but is opposite to those for Arctic foxes (Dugdale et al., 2003; Goltsman et al. submitted). However, the interpretation of such results is seldom as simple as the beguilingly straightforward predictions of these two hypotheses might suggest. For example, in the case of the bat-eared foxes we have already established that opportunities for dispersal may have been low, which would alter the predictions of Julliard's hypothesis. Indeed, a diversity of complicating factors is already known to affect offspring sex ratio in polytocous species, such as maternal parity (African wild dogs, Lycaon pictus; Creel et al. 1998), maternal age and condition (coypu, Myocastor coypus; Gosling 1986b), and stress (golden hamsters, Mesocricetus auratus; Pratt et al. 1989).
Whatever the reason for the skewed sex ratio, how is it achieved? We speculate above that infanticide is a strong candidate, and raise the question of which individuals do the killing. Maternity analysis of the surviving cubs may suggest an answer.
Phylogeny, resources, and luck
Bat-eared foxes are survivors of an ancient lineage (currently represented by two subspecies (Coetzee 1977): O. m. megalotis (southern Africa), O. m. virgatus (East Africa)). The extent to which their ancient separation has bequeathed upon these foxes different constraints to those faced by other canids is unknown. There is evidence that they cannot regurgitate food, nor it seems can vulpine canids. There is no evidence of social suppression of female reproduction (common in both lupine and vulpine canids). The inverted U-posture of the tail is seemingly unique (Nel and Bester 1983). Nonetheless, there have been sufficient studies to reveal that, like other wild canids, bat-eared foxes display some intraspecific variation (Nel and Maas 2004; Wright 2004). That variation is doubtless rooted in regional differences in ecological circumstances and perhaps to the behaviour of the various species of termite on which they feed and, even within our study area, fox sociology varied between years and between groups within years. There is a plausible case that this can be attributed to variations in the dispersion of invertebrate prey, and the RDH suggests some predictions to test this. Nonetheless, our study reveals the enormous impact that an essentially chance event—infectious disease—can have on these predictions. It also reveals as crucial the question of whether—in the context of litter size—they can count to six. While the social lives of bat-eared foxes take a direction determined by their ancestry and local ecology, the outcome would appear to owe much to luck.