Age, musth and paternity success in wild male African elephants, Loxodonta africana

ANIMAL BEHAVIOUR, 2007, 74, 287e296
doi:10.1016/j.anbehav.2006.12.008
Age, musth and paternity success in wild male
African elephants, Loxodonta africana
JULIE A. HOLLISTER-SMIT H* , JOYCE H. POOLE†, ELI ZA B ETH A. AR CHIE* †, ERIC A. VA NC E ‡,
N IC HO L A S J . G E O R G IA DIS§, C YNTHIA J. MOS S† & SUSAN C. ALBERTS *
*Department of Biology, Duke University, Durham, North Carolina, U.S.A.
yAmboseli Elephant Research Project, Nairobi, Kenya
zDepartment of Statistical Science, Duke University, Durham, North Carolina, U.S.A.
xMpala Research Centre, Nanyuki, Kenya
(Received 21 September 2006; initial acceptance 3 November 2006;
?nal acceptance 1 December 2006; published online 13 July 2007; MS. number: A10565)
Male African elephants experience intense intrasexual selection in gaining access to oestrous females, who
represent a very scarce and highly mobile resource. An unusual combination of behavioural and physio-
logical traits in males probably re?ects this intense selection pressure. Males show prolonged growth, grow-
ing throughout much or perhaps all of their long life span (ca. 60e65 years), and they show musth,
a physiological and behavioural condition exclusive to elephants, which is manifested by bouts of elevated
testosterone and aggression and heightened sexual activity. Most observed matings are by males over 35
years of age and in musth, suggesting that age and musth are both important factors contributing to
male reproductive success. Here we report the results of a genetic paternity analysis of a well-studied pop-
ulation of wild African elephants. Patterns of paternity for 119 calves born over a 22-year period showed
signi?cant effects of both age and musth on paternity success. Among males in musth, paternity success
increased signi?cantly with age until the very oldest age classes, when it modestly declined. When not in
musth, males experienced relatively constant, low levels of paternity success at all ages. Thus, despite the
importance of both musth and age in determining male paternity success, adult males both in and out of
musth, and of all ages, produced calves. In general, however, older males had markedly elevated paternity
success compared with younger males, suggesting the possibility of sexual selection for longevity in this
species.
Ó 2007 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Keywords: African elephant; age-related behaviour; competition; dominance status; genetic paternity success; Loxodonta
africana; maleemale con?ict; mating; musth
Dominance status is a recurring theme in analyses of male
?ghting ability (Moczek & Emlen 2000; Reichard et al.
reproduction in many animal taxa. This re?ects the fact
2005; Carlini et al. 2006). In other species, other factors,
that in many (but not all) species, access to reproductive
such as physiological state, may contribute to a male’s
females is strongly in?uenced by maleemale competition
ability to dominate other males. For instance, in some co-
(Andersson 1994; Ellis 1995). However, males attain their
operatively breeding species, dominant males may be dis-
dominance status in different ways. In some species and
tinguished from other males by their higher androgen
across a wide range of taxa (e.g. invertebrates, mammals
pro?les and they may socially and/or physiologically sup-
and ?sh), size alone is a strong predictor of dominance
press reproduction in subordinates (Clarke & Faulkes
rank, presumably because it is a strong predictor of
1998; Peters et al. 2001; Bales et al. 2006; Bender et al.
2006). The common theme in all these examples is that
males in many taxa have experienced strong selection to
Correspondence and present address: J. Hollister-Smith, Oregon Health
dominate other males for reproductive opportunities.
& Science University, Oregon Clinical and Research Institute, 3303 SW
Bond Avenue, Portland, OR 97239, U.S.A. (email: juliehollister-smith@
This selection pressure has resulted in a diversity of means
alumni.duke.edu).
by which they do so.
287
0003e 3472/07/$30.00/0
Ó 2007 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

---

288
ANIMAL BEHAVIOUR, 74, 2
Elephants represent a particularly interesting and un-
of musth increases from 2 days for males aged 16e25
usual example of this diversity. Male elephants face very
years, to 13 days for males aged 26e35 years, to 52 days
intense
reproductive
competition
and
they
show
for males aged 36e40 years, to 69 days for males aged
a unique dominance pattern in which individual males
41e45, to 81 days for males aged 46e50 years, and then
repeatedly experience periods of both high and low
declines again to 54 days for males aged 51e60 years of
dominance status. Both the intense reproductive com-
age (Poole et al., in press). Furthermore, young males expe-
petition that male elephants experience and the distinc-
rience musth at irregular intervals, but as an individual
tive dominance patterns that they show arise from the
male ages, his musth periods generally stabilize to an ap-
particular behavioural ecology and life history of ele-
proximately annual occurrence (Poole 1987, 1989a, b;
phants. African elephants live in ?ssionefusion societies
Poole et al., in press). However, musth does not occur syn-
in which males and females do not co-reside in perma-
chronously among adult males. Musth may be observed in
nent social groups. Females live in ?uid social groupings
every month of the year, but relatively few males are in
generally with their female relatives and calves (Archie
musth concurrently within a population (Poole 1987,
et al. 2006), while males range independently of these
1989a). Being in musth temporarily raises a male’s domi-
female groups, joining them only occasionally for brief
nance status above males not in musth, including those
periods to travel and mate (Moss & Poole 1983; Poole
larger than himself that he would otherwise rank below
& Moss 1989). Female groups are highly mobile, vary
(Poole 1987, 1989a). Male elephants compete directly
over time in size and composition, and only rarely con-
and sometimes intensely for access to mates, and even
tain females in oestrus; each adult female is sexually re-
occasionally kill each other; escalated aggressive interac-
ceptive only for 3e6 days every 3e9 years (reviewed in
tions generally involve a male in musth (Hall-Martin
Moss 1983; Poole & Moss 1989). Thus, male elephants
1987; Poole 1989a). Musth appears to be an energetically
face the reproductive challenge of locating very scarce,
costly state to maintain; males lose weight and their phys-
highly mobile reproductive females, and of preventing
ical condition visibly deteriorates as musth progresses
other males from gaining access to them (Poole 1989b;
(Poole 1989a), presumably because of decreased foraging
Poole & Moss 1989).
(Poole 1982) coupled with increased distance travelled
Male elephants show an unusual combination of be-
(Poole 1982) and alterations in body homeostasis (Schulte
havioural, morphological and physiological traits that are
& Rasmussen 1999). The oldest males experience the
thought to re?ect the intense competition for access to
largest deterioration because they maintain musth for
females that they experience (Poole 1989a, b; Poole &
the longest periods (Poole 1989a).
Moss 1989; Poole et al., in press). First, they show pro-
Musth also in?uences the association patterns and
longed and possibly indeterminate growth: they continue
sexual behaviour of both male and female elephants.
to grow in stature, body mass and tusk weight throughout
Males in musth range more widely than nonmusth males,
much and possibly all of their life (Roth 1984; Haynes
spending more time with female groups or alone and less
1991; Lindeque & van Jaarsveld 1993; Lee & Moss
time with other males (Hall-Martin 1987; Poole 1989a;
1995). Second, they have a very long life span. Median
Rasmussen et al. 1996). Female elephants appear to prefer
life expectancy for male elephants in Amboseli is esti-
older musth males as mates, maintaining close physical
mated to be 26 years; about 25% of males survive beyond
proximity to older males who are in musth when they
the age of 40 (Poole et al., in press) and males in their 60s
are in oestrus (Moss 1983). Musth males have higher
are regularly sighted in the population (Lee & Moss 1995;
mate guarding and mating success than nonmusth males.
Moss 2001). Male elephants thus have a very long poten-
Nonmusth males do show breeding behaviour and mate
tial life span (ca. 65 years; Haynes 1991). Consequently,
successfully (Poole 1989b), however, most observed
the size disparity among reproductively mature males
matings are by musth males over 35 years of age (Moss
may be very large depending on population demograph-
1983; Poole 1989b).
ics. For instance, a 40-year-old male may be twice the
Here, we addressed two questions using a genetic
weight and 30% taller than a 20-year-old male (Poole
paternity analysis of the wild population of African
et al., in press). Because dominance rank is size dependent,
elephants living in and around Amboseli National Park,
and males grow throughout life, males gain in dominance
Kenya. First, we tested the hypothesis that the higher
rank as they age (Poole 1989a).
mating success experienced by older males in musth
In addition, healthy adult male elephants show musth,
(Poole 1989a, b; Poole et al., in press) translated into
a physiological and behavioural condition that is man-
higher paternity success. Genetic determination of pater-
ifested by bouts of elevated testosterone and aggression,
nity was important because studies in some species have
and heightened sexual activity. The importance of musth
shown that dominance rank, mate guarding and even
has been well documented by behavioural observations
mating itself are not always good predictors of actual
(Moss 1983; Hall-Martin & van der Walt 1984; Hall-Martin
paternity (Pemberton et al. 1992; Hughes 1998; Coltman
1987; Poole 1987, 1989a, b, 1999). Males begin to experi-
et al. 1999; Eady & Hardy 2001; Preston et al. 2001).
ence musth at a mean of 29 years of age (Poole 1987), and,
Second, we quanti?ed the impact of both age and musth
as a male ages, his musth episodes typically increase in
on male paternity success. Mating success increases mark-
duration throughout his life. Among young males, musth
edly with age (Poole 1989b; Poole et al., in press), as does
generally lasts for only a few days or weeks, whereas in
the duration of musth (Poole 1987, 1989a; Poole et al., in
older males, it typically lasts for several uninterrupted
press). We developed a model to test the contributions of
months (Poole 1987, 1989a). Speci?cally, median duration
each to male paternity success.

---

HOLLISTER-SMITH ET AL.: WILD AFRICAN ELEPHANT PATERNITY
289
METHODS
For individuals with multiple samples, we replicated
DNA ampli?cation from two samples, either a tissue and
Data Set
a faecal sample or, if tissue was not collected, from two
faecal samples collected on separate days. Genotypes for
The subjects of our study were wild African elephants
faecal and tissue samples matched for 57 of the 58 individ-
living in and around Amboseli National Park, Kenya. This
uals for whom we had both. In the one instance where
population has been continuously studied since 1972
tissue and faeces did not match (mismatched at 5 of 8
(Moss 2001) and all elephants are individually recogniz-
loci), we assumed that the mismatch was due to misiden-
able based on individual physical characteristics and are
ti?cation of a faecal sample in the ?eld, and we used only
habituated to researchers’ presence (Moss 2001). Sightings
the tissue sample (see Buchan et al. 2005). To be conserva-
of individuals are done on an opportunistic basis, by
tive, we assigned a given allele to a given individual only
a core of ?ve experienced elephant researchers (mean ¼
when it ampli?ed consistently either in a minimum of
19.8 years with the project, range 12e30 years). Life his-
two reactions with at most one other allele also amplify-
tory and behavioural data for the approximately 1300 ele-
ing (for heterozygous individuals) or in a minimum of
phants in the population (including births, deaths,
seven reactions with no other alleles amplifying (for ho-
musth, oestrus, mating and mate guarding records), are
mozygous individuals). Lastly, we conducted Mendelian
documented in the Amboseli Elephant Research Project
checks for all mothereoffspring pairs.
(AERP) databases (Moss 2001).
The population showed high genetic diversity, with
Every animal in the study population has been assigned
a mean number of alleles per locus of 13.25 (range 9e19);
an age as part of the ongoing long-term research. Ages for
all eight loci provided informative data. All loci were in
elephants born after 1975, with few exceptions, were
HardyeWeinberg
equilibrium
(CERVUS
chi-square:
known to within Æ2 weeks. Elephants born from 1972
P < 0.01) and mean expected heterozygosity was 0.834.
through 1975 were known to within Æ3 months. Ages of
The null allele frequency was near zero for all loci (range
individuals born before 1972 were estimated based on
validated techniques that have been used in multiple
À0.0141, þ0.0121), suggesting that no loci had to be
excluded from the analysis because of allele nonampli?-
elephant populations (Haynes 1991; Lee & Moss 1995;
cation
resulting
from
primer-binding-site
mutation
Moss 2001). Ages of individuals born in 1970e1971
(Marshall et al. 1998). The mean observed error rate across
were considered accurate to Æ6 months and those born
all loci, calculated by CERVUS, was 0.0082. The cumula-
in 1968e1969 were considered accurate to Æ1 year. The
tive power of exclusion for identifying the second parent
ages of individuals born in 1963e1967 were considered
when one parent was known was 0.9999.
accurate to Æ2.5 years and individuals born before 1963
were considered accurate to Æ5 years. These accuracies
are based on known patterns of variance in size measures
with increasing age (Haynes 1991; Lindeque & van
Paternity Analysis
Jaarsveld 1993; Lee & Moss 1995; Morrison et al. 2005).
Putative fathers were identi?ed using CERVUS 2.0
(Marshall et al. 1998). The offspring analysed were con-
Genetic Sampling and Genotyping
ceived over a period of 22 years (1977e1998). Because
CERVUS is very sensitive to the proportion of candidate
Samples from 89 adult male Amboseli elephants (age
parents sampled (Kru
¨tzen et al. 2004), and this proportion
17e59 years old) and 279 calves and their mothers were
varied over the 22-year period, we ran different simula-
used in the analysis described here. For nearly 85% of all
tions of CERVUS for periods with different proportions
elephants sampled, we obtained multiple faecal samples,
of candidate males sampled. We kept the following input
collected on separate days (mean ¼ 3.9 samples per indi-
parameters constant for all CERVUS simulations: 10 000
vidual, range 2e14 samples). For the remaining 15%, we
cycles, 90 candidate parents, 100% of loci typed, 1% of
collected a single faecal sample. In addition, a single tissue
loci mistyped and con?dence levels of 95% strict and
sample was collected from 58 individuals. DNA was ex-
80% relaxed. However, we varied the value for the propor-
tracted from faeces using a QIAmp DNA Stool Kit (Qiagen)
tion of candidate males sampled depending upon concep-
following the methods described in detail by Archie et al.
tion years: 33% (1977e1980), 45% (1981e1985), 55%
(2003). Tissue DNA was extracted using a DNeasy Tissue
(1986e1990), 61% (1991e1995) and 74% (1996e1998).
Kit (Qiagen, Valencia, California, U.S.A.) following the
This procedure resulted in paternity assignments with
manufacturer’s protocol. PCR products were separated
95%
con?dence for
119
calves. For
most
calves
using an ABI PRISM 3700 DNA analyser (Applied Biosys-
(N ¼ 114), the assigned male was the only male who had
tems, Foster City, California) and allele sizes were deter-
zero mismatches with the offspring (i.e. all other adult
mined
using
GENOTYPER
2.0
software
(Applied
males had one or more mismatches). In the other ?ve
Biosystems).
cases, the assigned male mismatched at a single homozy-
We genotyped individuals at eight of the microsatellite
gous locus (N ¼ 4) or at two homozygous loci (N ¼ 1).
loci described by Archie et al. (2003): LaT05, LaT07,
These 119 calves represented approximately 10% of
LaT08, LaT13, LaT16, LaT17, LaT18 and LaT24. Loci
recorded births in the entire population during the study
were ampli?ed as in Archie et al. (2003), and genotypes
years (Table 1). Most of the calves for whom fathers were
were assigned using a modi?ed multiple-tubes methods
not assigned were from earlier years of the study,
(Taberlet et al. 1996), as described in Archie et al. (2006).
for which fewer mothers were sampled and smaller

---

290
ANIMAL BEHAVIOUR, 74, 2
Table 1. Distribution of elephant offspring births recorded in the
produce an offspring. For example, we calculated age-
study population by year and number of assigned paternities
speci?c paternity success at age 45 by ?rst counting the
number of offspring matched to males of age 45 (N
Recorded
Number
Percentage
¼ 8;
number of
of con?rmed
of assigned
one calf conceived in 1985, three in 1990, two in 1995,
Birth year
offspring born
paternities
paternities
and one calf each in 1996 and 1998). Then we calculated
the number of opportunities to produce offspring by
males age 45 by summing, over all years, the number of
1978
5
1
20
1979
57
3
5
calves assigned paternity multiplied by the number of
1980
53
2
4
males that were age 45 that year. In 1989, for example,
1981
25
1
4
one male was age 45 and six calves were assigned pater-
1982
33
0
0
nity; therefore, 45-year-old males had six opportunities
1983
82
2
2
1984
40
0
0
to sire calves that year. In 1990, two males were age 45
1985
61
4
7
and nine calves were assigned paternity (2 Â 9). Thus,
1986
24
1
4
there were 18 opportunities for 45-year-old males to sire
1987
66
7
11
calves in 1990. We de?ned the paternity success for each
1988
40
3
8
age class as the number of offspring produced by males
1989
28
0
0
1990
49
6
12
of that age class, divided by the total number of opportu-
1991
74
10
14
nities for conceiving calves by males of that age class
1992
39
4
10
pooled across all years of the study.
1993
35
7
20
To highlight the major age-related trends and minimize
1994
64
5
8
1995
71
9
13
the effects of noise in the data, we generated a smoothed
1996
74
8
11
curve (with 95% con?dence intervals) by using a 10-year
1997
53
8
15
moving average, for yearly intervals beginning with age
1998
35
6
17
21.5 years. That is, the average number of offspring per
1999
116
13
11
opportunity for age 21.5 years was generated by taking the
2000
114
19
17
total number of calves assigned to males for ages 17
Total
1238
119
10
through 26 years divided by the opportunities for the
males aged 17 through 26 years; the average for age 22.5
years was generated by summing the values for ages 18
proportions of candidate fathers were sampled. However,
through 27 years, and so on through age 54.5 years (the
other than their year of birth, there was no obvious source
last age for which we could take a 10-year average).
of bias (e.g. based on maternal home range or natal family
We examined the effects of musth on age-speci?c
group) in the set of calves for which we were able to assign
paternity success by examining the musth records of
paternity, so that these 119 calves should represent a ran-
each assigned father and all genotyped nonfathers during
dom sample of calves drawn from the population.
the month that each offspring was presumed to have been
conceived (de?ned as 22 months before the recorded birth
Analysis of Age and Musth Effects
month; Moss 1983). Males were considered to be in musth
if they showed secretions from swollen temporal glands
To examine the effects of age on paternity success, we
with facial staining and concurrent urine dribbling or
designated male age classes using 1-year intervals (e.g. the
the evidence of recent dribbling (e.g. urine stains on the
40-year-old age class comprised all males in their 40th
inside of the legs; Poole 1987). We assigned fathers to
year, etc). We then calculated the number of offspring
the ‘in musth’ as opposed to the ‘not known to be in
produced per opportunity by adult males in each class age
musth’ category only if they were unambiguously seen
(17 years and older). First, we assigned each offspring to
in musth during the month that conception was pre-
a male age class depending upon the age of the father at
sumed to occur. This was a highly conservative approach
that offspring’s conception. Next, we counted the number
because males in musth may not be seen by observers in
of opportunities that males in each age class had to father
a given month because of their protracted movement pat-
offspring, by counting the number of adult males alive in
terns and the relatively large size of the population. Con-
the population in each age class at the time of each
sequently, this approach resulted in assigning a number of
opportunity for paternity (each male had one opportunity
fathers to the ‘not known to be in musth’ state that were
for each conception). Owing to maturation and death, the
probably in musth at the time of conception, but were not
number of adult males in each age class changed for each
seen by observers; for instance, if a male was seen in
year of the study. Then, to get the number of offspring
musth during the month before the presumed conception
produced per opportunity by males in each age class, we
date and during the month after the presumed conception
divided the total number of offspring produced by males
date, but was not seen during the month of the presumed
in each age class by the number of opportunities that
conception, we conservatively assigned him to the ‘not
males had in those age classes (again, counting one
known to be in musth’ category. We then used a chi-
opportunity for each genotyped male that was present
square test to assess statistically whether fathers were
in the population at the time of that conception).
more likely to be in musth than nonfathers.
Each calf matched to a genotyped male in our study
For a subset of conceptions (N ¼ 56), the mother was
represented an opportunity for the genotyped males to
observed during oestrus. We used these behavioural

---

HOLLISTER-SMITH ET AL.: WILD AFRICAN ELEPHANT PATERNITY
291
records, which included records of female oestrous behav-
89
iour and male mate guarding and mounting, to examine
In musth
the effects of musth during the week of conception for
Not known to be in musth
these 56 conceptions. Females were considered in oestrus
if, in the presence of males, they showed wariness,
‘oestrous walk’, chase, mount and consort behaviours, as
described in Moss (1983).
Modelling
We developed a model to describe the relationship
between paternity success and age and musth in Amboseli
elephants. Our model for paternity success models the
Males
probability of a male siring a calf given the male’s age and
36
musth status in the month that the calf was conceived.
Our model computes the probability of siring each calf for
all the adult genotyped Amboseli males alive at that
conception. Using the R statistical package v 2.2.1 (The
R Foundation, Vienna, Austria), we performed a logistic
regression for males i ¼ 1, ., 89, and calves j ¼ 1, ., 119.
 


logit pij ¼ log pij=1 À pij
1
¼ b0 þ b1 þ b1Ageij þ b2Age2ij þ b3Musthij
17
25
30
35
40
45
50
55
60
where Age
Age (years)
ij is the age of male i at the time of conception of
calf j, and Musthij ¼ 0 if male i was not seen in musth dur-
Figure 1. Life history data for the 89 genotyped male African ele-
ing the month of conception of calf j, but Musthij ¼ 1 if
phants in this study. Each line represents the known life history of
male i was seen in musth during the month of conception
one of the 89 males in the sample. Lines in the upper half of the
of calf j. These variables then determine p
plot represent genotyped males that fathered no offspring in our
ij, the probability
that male i sired calf j. We determined the P values of the
data set. The plot shows the distribution of age ranges that we
parameter estimates to identify whether the terms in the
were able to analyse for each male and the cross-sectional nature
model accounted for signi?cant variance in the data.
of the data set; for some males, we captured the early part of the
life history; for others, we captured the later part of the life history.
Like all statistical models, this model contained some
Each symbol represents a reproductive event (a calf conceived) while
false assumptions. In particular, the model implicitly
the male was in musth (6) or not known to be in musth (B).
assumed that a male’s paternity success was independent
from one year to the next, an assumption that was almost
certainly incorrect, as some males were clearly better
males were also among the oldest genotyped males in
reproducers overall than others. However, the effects of
the population in 1998, at 58, 53 and 48 years old. How-
this false assumption on our conclusions were probably
ever, males sired offspring beginning in young adulthood
very minor; no single male or small subset of males
(the youngest father in our data set was 26 years old at the
contributed disproportionately to the data set (see
time that he conceived the calf), indicating that paternity
Fig. 1), and both age and musth produced overwhelm-
success was not strictly age dependent.
ingly strong effects in the model, an effect that could
not be due to psuedoreplication in this data set (see
Results).
Age and Paternity Success
RESULTS
The youngest and oldest males for whom we docu-
mented paternity were 26 and 59 years old, respectively, at
Males varied greatly in the number of offspring they
the time of calf conception. Although males as young as
produced, and in the ages at which they produced them
their mid-20s sired offspring, this was not a common
(Fig. 1). Thirty-six males, somewhat less than half of the
event; six males sired a total of eight calves among them
genotyped Amboseli males over the age of 25, fathered
during their 20s (Hollister-Smith 2005). Age-speci?c pater-
the 119 genotyped offspring in our analysis. Just three
nity success, as measured by the number of calves fathered
males were responsible for 36 (30%) of the 119 assigned
per opportunity at a given age, increased steadily from the
paternities. These three males, each of which fathered 12
mid-20s until a peak between 45 and 53 years of age
calves in our analysis, also had the highest guarding and
(Fig. 2). Paternity success then declined to levels compara-
mating success in our behavioural records of oestrus and
ble to a male in his early 40s. Siring offspring in old age
mating, supporting our hypothesis that mating success
was not a rare occurrence. Four males sired 14 calves
would predict paternity success in this population. These
among them when they were in their 50s.

---

292
ANIMAL BEHAVIOUR, 74, 2
data for the subset of 56 calves with known conception
0.10
10-year average
dates. For these calves, behavioural records of mothers’
95% CI
oestrus were available (see Methods), increasing the prob-
ability that we observed the father at the time of concep-
0.08
tion. In this subset, the proportion of fathers in musth at
conception was 79%. Taken together, these results sug-
gested that musth had an enormous impact in determin-
0.06
ing male paternity success, and led us to investigate the
importance of age and musth jointly in a formal model.
0.04
Modelling Effects of Age and Musth
on Male Paternity Success
0.02
Our model illustrates the conspicuous effects of age and
musth on male paternity success (logistic regression:
Calves born per opportunity per age class
c23 ¼ 279:72, P ( 0.0001; Table 2, Fig. 3). Both linear and
0
second-order effects of age were statistically signi?cant
0
10
20
30
40
50
60
(Table 2). Still more striking was the manner in which
Age (years)
musth contributed to paternity success. In particular, in
Figure 2. The relationship between male age and paternity success
the absence of musth, age contributed relatively little to
in African elephants, using data from 89 genotyped Amboseli males
variance in paternity success, because males not in musth
and 119 calves for whom we assigned paternity. The vertical lines
experienced relatively constant, low levels of success at all
denote the number of offspring produced per siring opportunity
ages. However, age did strongly contribute to the paternity
by males of each age in Amboseli. To highlight the major age-related
success of musth males (Fig. 3). Thus, a 45-year-old male
trends, the solid line represents a smoothed curve (with 95% con?-
dence intervals) generated by using a 10-year moving average of
in musth is predicted to outcompete a 25-year-old male
age for yearly intervals (see Methods).
in musth. However, a male in musth in his mid-20s expe-
riences a measurable advantage over his agemates that are
not in musth; he is predicted to achieve approximately
Musth and Paternity Success
the same paternity success as a 50-year-old nonmusth
male.
Musth males had much higher paternity success than
expected, based on the frequency of musth males in the
population at the time of conceptions. Speci?cally, 74%
DISCUSSION
(88/119) of the calves were fathered by males known to be
in musth and 26% of the calves were fathered by males
Our analyses indicate that two factors profoundly affected
not known to be in musth. This was quite different from
the paternity success of male elephants. First, male
the relative frequencies of musth and nonmusth males
paternity success increased with age until late in life.
in the population; only 12.7% of all genotyped males were
Second, most calves were sired by males who were in
in musth during conception opportunities (1066 males in
musth at the time of calf conception. The ?rst result
musth versus 7312 males not known to be in musth
supports the observation that age predicts a male’s posi-
pooled over the 119 conceptions; each adult male in the
tion in the male dominance hierarchy, and that older
population at the time of each conception was counted
males are able to outcompete younger males for access to
once, in musth or not known to be in musth, for each
females (Poole 1989b; Poole et al., in press). In contrast to
conception). This difference between the paternity success
age, which represents a dominance queue, musth appears
of musth and nonmusth males was highly signi?cant
to allow a male to ‘jump the queue’ of the age/size domi-
(c2
nance hierarchy; musth males can outrank nonmusth
1 ¼ 401:9, P ( 0.0001).
These striking effects of musth appeared in our data
males (Poole 1989a; Poole et al., in press), and because
even with our highly conservative manner of assigning
musth occurs asynchronously, unlike a seasonal rutting
musth status, in which any males not known with
period, adult males of all ages reproductively contribute
certainty to be in musth were assigned as ‘not known to
be in musth’. Consequently, it is very likely that our
Table 2. Results of logistic regression model for the effects of age
estimate of the number of conceptions attributed to
and musth on paternity success in male African elephants
fathers in musth was an underestimate (see Methods).
This is especially true given that most conceptions to
Standard
fathers not know to be in musth involved younger fathers.
Predictor
error of
Younger males typically stay in musth for much shorter
variable
Estimate
estimate
Z
P
periods, increasing the probability that we would not
have detected them in musth. Hence, we conclude that
Intercept
À13.67
2.07
À6.61
(0.0001
the difference we have documented in the number of
Age
0.395
0.105
3.778
0.0002
Age2
À0.004
0.001
À3.259
0.001
calves fathered by musth versus nonmusth males repre-
Musth
2.306
0.227
10.154
(0.0001
sents a minimum difference. This is supported by our

---

HOLLISTER-SMITH ET AL.: WILD AFRICAN ELEPHANT PATERNITY
293
males are larger than most adult females (Lee & Moss
0.14
In musth
1995), so it may be physically dif?cult for a female to resist
Not known to be in musth
mating attempts. Females actively attempt to evade
0.12
advances by young males (Moss 1983), however, they
are frequently harassed by multiple males during oestrus
0.10
if they are not guarded by a larger, older musth male
(Moss 1983; Poole 1989b). Younger males may also have
0.08
the advantage of sneaking mating opportunities when
older males interact with each other (Poole 1989b).
Finally, although females in general appear to prefer large
0.06
musth males, female preferences may be idiosyncratic to
some extent, possibly re?ecting preferences for males
0.04
Probability of siring a calf
with whom they are particularly compatible regardless of
age or size (see review in Neff & Pitcher 2005).
0.02
This pattern of continued high male reproductive out-
put into old age is unlike that of most described mamma-
0
lian species (Table 3; e.g. baboons, Papio cynocephalus
0
10
20
30
40
50
60
(Alberts et al. 2003, 2006), rhesus macaques, Macaca
Age (years)
mulatta (Bercovitch et al. 2003), red deer, Cervus elaphus
(Clutton-Brock et al. 1988), northern elephant seals, Mir-
Figure 3. Model predictions for the relationship between paternity
ounga angustirostris (Le Boeuf & Reiter 1988), fallow deer,
success and age and musth in male African elephants. Curves were
generated using the parameter estimates described in Table 2. See
Dama dama (McElligott & Hayden 2000; McElligott et al.
Methods for details.
2002), greater kudu, Tragelaphus strepsiceros (Owen-Smith
1993), lions, Panthera leo (Packer et al. 1988), and chim-
to the population. Musth does not annul the hierarchy,
panzees, Pan troglodytes (Boesch et al. 2006)). Typically,
however.
the reproductive output of a male mammal peaks when
Older musth males had a marked advantage over
he reaches his full adult body size, which occurs soon after
younger musth males (Fig. 3). The effect of musth was
he achieves sexual maturity. At this point males are usu-
more dramatic for older males, in part, because older
ally in their peak physical condition and at their highest
males in musth are dominant over younger males in
dominance rank. Males often maintain high dominance
musth as well as over all males not in musth. In contrast,
rank for a relatively short period, perhaps only one or
younger musth males are lower ranking than older males
a few breeding seasons. As body condition deteriorates
in musth, even though they can outrank males not in
and they can no longer compete with younger males, re-
musth. This situation ultimately reinforces the reproduc-
productive performance of many male mammals steadily
tive dominance of older males, so that reproductive suc-
declines. For most male mammals, peak reproductive out-
cess increases dramatically with age for musth males
put is achieved within the ?rst half of the total adult life
until late in life. Older males are also able to remain in
span (Clutton-Brock et al. 1988; Le Boeuf & Reiter 1988;
musth longer than younger males (Poole 1987, 1989a;
Packer et al. 1988; Owen-Smith 1993; McElligott &
Poole et al., in press), and this effect also reinforces the
Hayden 2000; McElligott et al. 2002; Alberts et al. 2003,
reproductive dominance of older males. In particular,
2006; Bercovitch et al. 2003; Boesch et al. 2006). The con-
males between 45 and 50 years of age produced calves at
sequence is that male breeding life span is compressed
an average annual rate six times that of 30-year-old males
into a fraction of its full potential because of intense com-
(Fig. 2). This ?nding suggests that male elephants may
petition from other males (Clutton-Brock 1988).
experience sexual selection for longevity (Poole 1989a,
Male elephants appear to represent the extreme among
b; Poole et al. in press). However, relatively few males in
mammals in the extent to which they show high mating
the population lived to 50 years, and survival to 50 years
and paternity success until late in life. Two noteworthy
is estimated at less than 10% for males (Moss 2001; Poole
features of elephant physiology and behaviour are prob-
et al., in press).
ably responsible for this: (1) male elephants continue to
In spite of the reproductive superiority of older musth
grow throughout much or all of life (Roth 1984; Haynes
males, younger males’ reproductive contribution was not
1991; Lindeque & van Jaarsveld 1993; Lee & Moss
inconsequential. Indeed, males under 35 years of age