A novel mating system in a solitary carnivore: the fossa

Authors

  • C. E. Hawkins,

    1. School of Biological Sciences, University of Aberdeen, Aberdeen, UK
    2. School of Zoology, University of Tasmania, Hobart, Tas., Australia
    Search for more papers by this author
    • *Current address: School of Zoology, University of Tasmania, Hobart, Tas. 7001, Australia.

  • P. A. Racey

    1. School of Biological Sciences, University of Aberdeen, Aberdeen, UK
    Search for more papers by this author

  • Editor: Prof. Virginia Hayssen

Correspondence
Clare E. Hawkins, School of Biological Sciences, Zoology Building, University of Aberdeen, Aberdeen AB24 2TZ, UK.
Email: clare.hawkins@dpiw.tas.gov.au

Abstract

The mating strategies of male mammals have long been treated as broadly predictable on the basis of just two factors: the dispersion of females and the benefit of paternal care to male reproductive success. Female strategies and finer scale variations in mating systems remain poorly understood. In the fossa Cryptoprocta ferox, we had the rare opportunity to examine the mating system of a wild solitary carnivore directly, and identified features not classified or predicted by mating system theory. Males competed for mating opportunities at a traditional site monopolized by a female, high in a tree. The female mated with multiple males, repeatedly mated with some individuals and appeared to express mate choice. We observed three females thus, one replacing another on the site after each was seen to mate with four to five males over a period of 1–6 days. Copulations were prolonged (up to 3 h 8 min), involving a weak copulatory tie, and males appeared to guard females briefly after mating. Fossas are at low population density and do not use a den regularly; we suggest that both these factors impede individuals from locating a mate. We hypothesize that the observed mating system reduces this problem for both sexes, and increases the number of mates available to a female while ensuring a low risk of sexual harassment.

Introduction

The majority of Carnivora species are solitary, yet the majority of data on this order comes from social species (Gittleman, 1989). Solitary species are defined as those in which individuals never, except when mating, cooperate with conspecifics (Sandell, 1989). Very little is known of the mating systems of these elusive, typically nocturnal species, and much is inferred from spatial organization or denning behaviour (e.g. Gehrt & Fritzell, 1999).

The strategies by which individuals of different species locate a second parent for their offspring were linked by Emlen & Oring (1977) to a small group of ecological factors; this approach to explaining mating systems has since been developed and applied to a wide range of taxa. Yet predicting the mating system of a little known species remains difficult (Reynolds, 1996), and female strategy in particular remains poorly understood (Zeh & Zeh, 2003). Clutton-Brock (1989) explained mammalian male mating strategy in terms of two factors: (1) the significance of paternal care to the reproductive fitness of fathers and (2) female dispersion. Paternal care is absent in solitary species, so the mating system of solitary carnivores is predicted solely on the basis of female dispersion. Where solitary females occupy defensible ranges, the male is predicted to mate with the female(s) resident on the range he defends (facultative monogamy/polygamy); where female ranges are not defensible, the male is predicted to rove widely during the mating season, defending individual females during their receptive period. Sandell (1989), specifically considering the male mating systems of solitary carnivores, also considered the same two options of ‘staying’ and ‘roaming’; however, he proposed that population density, not female range defensibility, is the critical factor in determining the more beneficial strategy. In this scheme, solitary carnivores at a population density lower than average, relative to body size, are ‘roamers’, while those with higher than average densities relative to body size are ‘stayers’.

The limited data on solitary carnivore mating systems provide evidence of both ‘staying’ (civet Nandinia binotata, Charles-Dominique, 1978) and ‘roaming’ strategies (stoat Mustela erminea, Sandell, 1986; polar bear, Ramsay & Stirling, 1986), but other systems are also apparent. Male sea otters (Garshelis, Johnson & Garshelis, 1984) and some male cheetahs (Caro & Collins, 1987) appear to defend an area that females regularly pass through.

Evidence is increasing for female multiple mating in these species (e.g. Schenk & Kovacs, 1995). Sandell (1989) did not discuss female strategy, and Clutton-Brock (1989) remained non-committal regarding the factors that render female multiple mating a beneficial behaviour. Multiple mating is costly, for example in terms of energy, time and disease risk. It is difficult to explain the benefits to a female of mating multiply during a single period of receptivity, when a single brief copulation is potentially sufficient to fertilize all her eggs. Hypotheses concerning the benefits of female multiple mating are numerous but their tests remain inconclusive (Zeh & Zeh, 2003). Males benefit from participating in this behaviour through sperm competition, but again there are few data on the mechanism of this in mammals.

Further field data on both novel and half-understood mating systems are likely to help identify the circumstances in which female multiple mating is beneficial. The fossa Cryptoprocta ferox provided an opportunity to observe the mating system of a wild solitary carnivore in much closer detail than is commonly possible for such species; solitary carnivores are especially prone to exhibit different behaviours in captivity (e.g. Soderquist & Ealey, 1994). This large (head–body length c. 70 cm) euplerid (Hawkins, 1998), endemic to the forests of Madagascar, was reported to exhibit an unusual, relatively easily observed mating behaviour. Reports from captive (Vosseler, 1929) and wild observations (Albignac, 1970; C. Rakotondrasoa, pers. comm.; A. F. A. Hawkins, pers. comm.; Q. M. C. Bloxam, pers. comm.; J. Powzyk, pers. comm.) described a single female mating in a tree over a period of several days. Several males were commonly observed in and nearby the tree, and active competition was observed between them. Females appeared likely to have mated multiply, but lack of observation time or of distinguishing features among males precluded confirmation of this. C. Rakotondrasoa (pers. comm.) reported a traditional site, having observed fossas mating there regularly, in the same tree, over several years. Mating is clearly seasonal, having been observed over a period of a few weeks near the end of the dry season, between September and October (Albignac, 1973) or in November (C. Rakotondrasoa, pers. comm.).

Predictions of a mating system for the fossa are inconsistent. Clutton-Brock's (1989) scheme predicts male fossas to exhibit facultative monogamy/polygyny, or a ‘staying’ strategy, on the basis of the defensibility of female ranges by males. While defensibility of range is more or less impossible to measure, a recent study (Hawkins, 1998) found the ranges of female fossas to be exclusive, and therefore presumably defensible by males. As further evidence for this, males appear capable of territoriality, scent-marking frequently throughout the year in captivity (Albignac, 1973; Klockenkämper, 1988), and occupy ranges more than twice the size of those of females (Hawkins, 1998). Sandell's (1989) argument, on the other hand, predicts fossa males to exhibit ‘roaming’, because fossa population density relative to body size (Hawkins & Raccy, 2005) falls well below the regression line for solitary carnivores (Sandell, 1989) as well as for tropical carnivorous mammals (Peters, 1983).

We set up a study of marked individuals at the reported traditional site, to establish the mating system of the fossa. We sought to identify male and female mating strategies, and to investigate possible benefits of this system with a view to improving the predictive capacity of mammalian mating system theory.

Methods

The study was carried out in Kirindy Forest (44°41′E, 20°04′S), a 100 km2 concession of dry deciduous forest, 23 km from the west coast of Madagascar, where fossas had been observed mating in the same tree over several years (C. Rakotondrasoa, pers. comm.; A. F. A. Hawkins, pers. comm.; Q. M. C. Bloxam, pers. comm.).

Marking

Eighteen box traps (collapsible, single door ‘Bobcat traps’, 107 × 38 × 51 cm, by Tomahawk, WI) were placed on a grid, 500 m apart, and left set continuously for a 3-day period. After this period the traps were moved to new locations for another 3 days, and so on, such that an area of 16.75 km2 was trapped regularly over the dry seasons (May–October) of 1994 and 1995, and intermittently over the wet season (November–April) of 1994. Total trap effort was c. 1400 trap days.

Trapped fossas were anaesthetized with 5 mg ketamine (Vetalar V, Pharmacia & Upjohn Ltd, Corby, UK) and 1 mg xylazine (Rompun, Bayer, Bury St Edmunds, UK) per kg estimated body weight, and marked with a coloured and numbered ear tag (Rototag by Dalton, Henley-on-Thames, UK) in each ear, allowing identification at a distance. Measurements were taken from males of body mass, hind foot length, testis length and toothwear (scored 0–5; an indicator of age) as indicators of male attributes likely to affect mating success. In addition, five trapped individuals were radio-collared (Mod-335, Telonics, Mesa, AZ, USA). Collars were removed at the end of the radio-tracking study, at the start of the wet season in 1996.

Observations

Each radio-collared animal was located a minimum of five times per week, from 15 October to 8 December 1995, except during periods of direct observation. Mean radio-tracking error was 207 m (Hawkins, 1998). Mating activity was identified through radio-tracking as: (1) a grouping of individuals or (2) the location of a female in the same area for longer than a 24-h period. Fossas have no discernible daily activity rhythm, and do not regularly rest in a den or other location, except for females in January and February when they have recently given birth (Hawkins, 1998). Mating was then observed directly. In 1994, the traditional site was checked regularly in October and November, but mating was not observed. Out of concern that this regular checking might have disturbed the initiation of mating, it was not used in 1995, and radio location was used instead.

One of us (C. E. H.) made direct observations 30–50 m from the mating couple, during all daylight hours logistically possible, for the entire period that mating in the study area was seen to take place. Darkness precluded night-time observations, but continued vocalizations and the location of radio-collared individuals indicated similar behaviour to that observed during the day. All figures given below refer to daytime observations only, and are therefore likely to be underestimates of absolute figures.

We identified all individuals within view of the observer where possible, and recorded the nature of, and participants in, all visible interactions. Unmarked animals of the same sex were not reliably distinguishable, so we excluded them from analyses unless otherwise stated. When including them in analyses, we counted unmarked animals conservatively, so that an unmarked animal seen on more than one occasion within 50 m of its previous location was treated as a single animal unless seen simultaneously with another. We did this on the basis of the behaviour of marked animals and the low frequency with which unmarked animals were seen.

Because multiple mating by females had been reported, physical competition between males, sperm competition and/or cryptic female choice appeared likely. In order to identify any differences between mates, all intromissions were timed. Any discrete event or behaviour coinciding with termination of copulation, such as the approach of a new individual, was recorded. All approaches made by a male to a female (defined as when a male approached a lone female directly, making characteristic vocalizations) were also recorded, together with the response of the female. To identify guarding behaviour, the period for which a male stayed with the female in the absence of any other male was recorded, together with any interaction such a male had with any approaching male.

All behaviours indicating any potential conflict between the sexes were recorded in detail (in terms of timing and individuals involved), such conflict being a potential pointer as to which sex benefits from what behaviour (e.g. Stockley, 1997).

Analysis

To investigate the basis of any pre-copulation male–male competition, male attributes (body mass, head–body length, testis length, toothwear) were related, using Spearman's rank correlation, to: (1) mating success, in terms of the number of approaches each male made to each female; (2) the proportion of approaches by a male that led to copulation with the female concerned; (3) the median period of any resulting copulation, to investigate the basis of any female mate choice.

Results

Fourteen adult males and eight adult females were marked; of these, two adult males (M1 and M2) and three adult females (F1, F2 and F4) were radio-collared. Mating and related behaviour was observed across a 17-day period in November 1995 (Fig. 1). No other mating behaviour was detected during the study period.

Figure 1.

 Map and calendar of mating behaviour observed in November 1995, on two sites, each consisting of a 10–20 m2 area of trees, in Kirindy, a dry deciduous forest in western Madagascar. The first female observed, F1, mated on Site 1. The second female, F2, visited Site 1 but then moved to Site 2, where she mated. A third female, F3, subsequently mated on Site 2, after which no further animals were observed on either site. All observations were direct except where radio-tracking (r-t) specified.

Three females were observed mating. Where three values are given together, these are for F1, F2 and F3, respectively, in this order. Variation about medians is given as interquartile range (IQR). Mating took place on two different locations, c. 3 km apart (Sites 1 and 2; Fig. 1). Site 1 was the site where mating had been observed in previous years. Each site consisted of a few trees within a 10–20 m2 area. Mating was seen only in the branches of trees, predominantly in the largest, most open tree on the site (Quivisianthe papinae on Site 1; Adansonia sp. on Site 2), but also in other adjacent trees. Both sites lay within 50 m of the banks of the same dry river. In this, by Site 1, was the only known permanent source of water in Kirindy, and by Site 2, were two temporary pools which may have formed after the first heavy rain for several months, which fell on 15 November.

No mating activity was detected until 12 November when F1, M1 and M2 were located in the vicinity of Site 1. M1, but no other radio-collared individual, was also at Site 1 the previous day. F1 subsequently remained in the vicinity of the site, apparently almost continuously, for 8 days; 11 radiofixes during this period indicated her absence from the site only once, and she was seen in a tree on the site for 69% of the period (61 h 29 min) during which she was intensively observed (Table 1). During this period, which ran across 6 days, she mated with five males, after which she was not seen or radio-tracked on the site.

Table 1.   Mating behaviour of consecutive females
 F1F2F3Trend
  1. All data refer to the period when the female in the column was monopolizing the site; F3 mated with a fifth male before monopolizing the site. Except where stated, the table includes data for unmarked animals, which were counted as a single animal unless seen simultaneously. F1 mated once with an unmarked male; F2 mated five consecutive times with an unmarked male; F3 mated once with one unmarked male, and twice with another seen simultaneously.

Number of days monopolized site (observation period)631
(61 h 29 min)(19 h 32 min)(6 h 25 min) 
Total males mated554None
Total males observed on site598+
% time observed in tree spent mating (time observed in tree)57%52%56%None
(42 h 26 min)(13 h 6 min)(5 h 33 min) 
Median duration of intromission (min) (n, interquartile range)552120
(19, 41–119)(14, 12–45)(8, 9–28) 
% repeated matings with marked males (n)100%50%33%
(4)(4)(3) 
% marked male approaches refused (n)5%30%38%+
(20)(20)(13) 

F2 was observed through radio-tracking to approach the site intermittently during this period, but was not viewed there until the last day F1 mated. She was seen alone on Site 1 the morning after F1 had left it, and was located by radio-tracking at both Sites 1 and 2 for this and the following day, taking a minimum of 2 h to travel between the sites. She then remained at Site 2 and 2 days later was first observed mating there. Over a 3-day period, during which she was observed for 19 h 32 min, she mated with five males. After this F2 was not seen on the site again.

An unmarked female, F3, entered a tree on the site on the second of these 3 days, and briefly mated with one male. She remained on the site but did not mate again until F2 had left the site. Then, during a 6 h 25 min observation period, F3 mated with four further males. The following day, only males were present on the site, and no more females were observed or located by radio-tracking at either of the two sites for the rest of the study period.

No mating behaviour was observed in F4, and radio-tracking indicated that she did not approach either site.

The radio-collared males, M1 and M2, were present at both sites. After mating activity had ceased at these, neither animal could be located through radio-tracking during the subsequent week, indicating that they had left the area within which a radiosignal could be received (Fig. 1).

Male behaviour

Four of the nine marked males observed on the two sites mated multiply, with two females. Of the rest, three mated with a single female and the other two did not mate (Table 2). No male was seen for more than 4 consecutive days on a site, but may have been resting nearby; male presence out of the trees could not be easily monitored. The two radio-collared males were precisely located infrequently during the periods of direct observation, but remained within receiving range (Fig. 1) until the end of observed mating activity.

Table 2.   Traits of fossa Cryptoprocta ferox males which had previously been trapped, and various measures of their respective mating successes with each of three females
Male
identity
Body
mass
(kg)
Hind foot
length
(mm)
Tooth wear
score (0–5)
Testis
length
(mm)
Number of approachesMedian intromission
period (min) (n, IQR)
Days
present
TotalRejected
  1. Toothwear score gives a relative indication of age. An ‘approach’ occurred when a male approached a lone female in order to copulate with her (indicated by his direct movement towards her and vocalizations). Values in parentheses for male 15 with F3 refer to a copulation before her monopolization of the mating site. IQR, interquartile range.

F1
 M28.8127.3231.87099 (7, 49–149)4
 M15.9124.7232.95040 (5, 32–118)4
 149.9136.3135.220122.5 (2, 89.75–155.25)2
 309.2130.4433.83155 (2, 55–55)4
F2
 155.5128.3033.44035.5 (4, 19–49.75)1
 177.9125.6138.11048.5 (2, 29.75–67.25)2
 309.2130.4433.83163 (1)2
 106.9124.9433.01015 (1)1
 M28.8127.3231.80001
 M15.9124.7232.90002
 166.9129.9032.71102
 295.7125.2029.52201
F3
 M15.9124.7232.94014 (4, 6.25–24.5)1
 106.9124.9433.01065 (1)1
 155.5128.3033.41 (2)1(0 h 4 min)1
 309.2130.4433.80001
 166.9129.9032.70001
 295.7125.2029.50001
 177.9125.6138.11101

Males overtly competed with another, apparently for access to the female

Twelve males, in total, were seen on the two sites (Fig. 2): five were seen on Site 1, and 10 on Site 2. A median of 3 (n=10, IQR 2–6.5) males per day visited the active site (Fig. 2). Males fought and chased one another frequently: on 27 independent occasions (i.e. separated by at least 15 min of absence of aggression) males chased one another, emitting characteristic vocalizations. Three of these 27 occasions occurred when a mating male terminated mating and attacked a male ascending the tree; the others took place on the ground. The trees and undergrowth prevented us from observing further details of fighting on the ground, other than that two males exhibited numerous recent scratches, on the shoulders, thighs, face and rump. Termination of intromission coincided with the arrival of a second male on 11 out of the 12 occasions when a second male approached a mating couple, although on five of these, the original couple recommenced mating.

Figure 2.

 Number of male fossas Cryptoprocta ferox seen on a mating site per day, and cumulative total of new males seen over the whole observation period (14–28 November 1995). Two sites, in Kirindy Forest, western Madagascar, were observed. One site was monopolized for 6 days by a female (F1), after which a second, 3 km away, was monopolized first by female F2 and then by F3.

Mate guarding was observed

Up to 100% of the marked males mating with each female did so more than once (Table 1). Intromission was prolonged (Table 1), with the longest period of uninterrupted intromission lasting more than 3 h 8 min (having started before observation). Male pelvic thrusts were observed only during the initial penetration. A weak tie was apparent: if copulation ended suddenly, the couple were, unlike copulating domestic dogs, able to pull apart, but this evidently required much effort. Males also appeared to guard females, remaining in the tree, inactive, for a median period of 11 min (IQR 5–33.5) after 23 out of 40 intromissions where termination was observed. Out of 19 independent periods of guarding, males successfully reinitiated copulation on five occasions (two males reinitiated copulation more than once during a continuous period with the female). The male remained in the tree with a female (guarding or mating with her) for a median continuous period of 65.5 min (IQR 49.5–120, n=32).

No link was apparent between male attributes and mating success

Exploratory analyses using Spearman's rank correlations found no significant association between male attributes and the numbers of approaches a male made to each female, the proportion of these approaches that she accepted, nor the median period a male spent mating with any one female (Table 2). Rankings of success between males were not consistent across females.

Female behaviour

All females mated multiply and repeatedly

Each was observed to mate with five individual males, repeatedly mating with any one individual a maximum of seven (F1), four (F2) and four (F3) times.

Sites were monopolized by one female at a time, for a minimum period of 1 day

A maximum of two females was observed simultaneously on a site, with a second female on a site for only 3 of the 10 days mating was observed. The female which was observed at a site first was consistently preferred as a mate by all males: males were never observed to approach the female that had arrived second. On two independent occasions, females that had arrived second approached males in trees; after repeated approaches, the male briefly mounted the female, in one case mating with her for 4 min (F3, 26 November) and in the other, not mating at all (F2, 19 November). Only when the first female had permanently left the site did the males approach the second female. On one occasion, when the second female (F3) initially arrived on the site, the first female (F2) briefly chased her. Other than this, no aggression was observed between females even during the two periods when the second female (F2 and F3, respectively) approached a mating couple. F3 spent a 32-min period close enough to touch F2 and a male while they were mating, intermittently presenting her genitalia to them, but neither of the couple showed any reaction to her.

Females appeared to exert some choice over which male they mated with, and for how long

On 23% of 53 approaches observed, females did not mate (Table 1), chasing the male away if he persisted. In all other cases, the female either went into lordosis immediately or ascended into higher, thinner, horizontal branches, sometimes as little as c. 4 cm in diameter, before allowing the male to mount. During intromission, the male was in a relatively precarious position, because his forefeet were balanced on her shoulders, and the branch on which the couple were positioned was often thin and flexible. The female was thus able to make intromission difficult by sitting up, standing up or walking around. This behaviour was seen at some point in almost every intromission observed, and was typically followed by the male licking the female's neck, scratching her back with his forepaws and making high-pitched vocalizations, which seemed to encourage her to settle down. On occasions where this was not effective, females became more and more restless and copulation would eventually terminate. Of 40 intromissions where termination was observed, the termination of 18 were associated solely with this behaviour. On one of these, the female attacked the male, who immediately left the tree. The female also attacked the male when he was guarding on two occasions, after which he immediately left the tree. Females (F2 and F3) also controlled the identity of their mates by chasing off a second male, on four of the 11 occasions that intromission ended after arrival of a second male, and on five other occasions when another male was already present.

Each female mated with five males

Various behaviours differed across the three periods during which a female was observed to monopolize a site (Table 1). Median period of intromission, period of monopoly of the site, total number of males observed during any one period of monopoly, proportion of recognizable (i.e. marked) males with which a female mated repeatedly, and the proportion of male approaches which a female accepted all varied such that F2 and F3 each spent less time mating than their predecessor did. Nonetheless, each female was observed to mate with a total of five males.

Discussion

The mating system identified is somewhat different from either of those predicted for the fossa, and includes features not previously classified by mating system theory or reported in any other species. This study confirms the anecdotal reports described in the ‘Introduction’, and, being based on marked individuals, provides the first observations of female fossas sequentially monopolizing a site and mating both multiply and repeatedly.

Male behaviour

The prediction using Clutton-Brock's (1989) scheme, of facultative monogamy/polygyny by males, is not supported by the observations reported here. No continuing bond was apparent between any couple, and up to nine males were seen on a single site (Table 1). It seems unlikely that all nine defended territories that included the same site. Year-round radio-tracking data indicated a much lower level of male home-range overlap than this (Hawkins, 1998). Thus the defensibility of female ranges was not a good predictive factor of male mating strategy in this case.

The ‘roaming’ strategy, predicted through Sandell's (1989) scheme, is a more appropriate, but imperfect interpretation of the results. Males clearly visited individual females during their period of receptivity, although no male monopolized a female for more than a few hours. The male mating strategy differed from the typical ‘roaming’ behaviour in that female location was highly predictable. Site 1 was confirmed as a traditional site, and fossas (including F1, who retained her eartags because we were unable to recapture her) have since been observed and filmed mating there on three subsequent years (Oxford & Bish, 2000; F. Jantschke, pers. comm.; J. Keeling, pers. comm.). A male was also observed making the vocalizations associated with the mating period around Site 2 in late November 2000 (C. E. Hawkins, pers. obs.). Mating was not seen on Site 1 in November 2000, when a third site was identified; this was also by the banks of the river bed, c. 500 m west of Site 1 (M. Eberle, pers. comm.). Males did not stay constantly at the active site observed, and F2 took as little as 2 hours to travel the 3 km between Sites 1 and 2, so a male could visit several sites within one female's receptive period.

Males can thus easily search for females by patrolling river beds, checking traditional sites. Water is often a scarce resource in Kirindy in November, which is the end of the dry season, and Site 1 was beside the only known permanent source of water in the study area. The mating strategy could have developed as fossas regularly encountered each other when coming to drink. A traditional site may be especially important to this species with regard to mate location; not only does the fossa exist at an unusually low population density for a tropical carnivore of its body mass (Hawkins & Racey, 2005), but also it does not maintain any dens or other central places where it regularly rests (Hawkins, 1998).

Male–male competition was evident in the fighting observed. Since several males might be present together on the ground but only one male approached the female at a time, competition appeared to be for access to the female rather than defence of the site. It is possible that other males did not approach the couple more frequently because the female would have rejected them. As in this study, Albignac (1970) and Oxford & Bish (2000) both observed the female to repel a second male that approached while she was mating.

Sperm competition was also indicated by the behaviour of the males. Intromission period was the longest recorded in a carnivore (Ewer, 1973), and maintenance of intromission was assisted by a weak copulatory tie. Prolonged intromission, and the observed subsequent guarding of the female, would be expected to improve a male's chance of fertilizing a female.

J. Keeling (pers. comm.) saw the mating male biting the female's neck on occasion, sometimes inflicting deep wounds. The tie presumably arises from the structure of the penis, the anterior two-thirds of which is thickly covered with backwardly pointing spines that stick out when the penis is erect (Hawkins, 1998; Hawkins et al., 2002). The penile spines would discourage a female from terminating intromission before the penis became flaccid.

Sperm competition could also explain the observed repeated mating between specific pairs, if males gained from maximizing numbers of sperm transferred. This strategy would also be expected to favour a significant increase in testis size during the mating season, which was not observed, but a larger sample might give different results (mean testis volume September–February ±sd: 14.5±1.9 cm3, n=10; March–August: 11.4±3.4 cm3). An alternative explanation for repeated mating may be that the males could not discern when the females were ovulating, such that repeated mating throughout the receptive period was a strategy to increase a male's probability of fertilization.

Female behaviour

Females, rather than males, appeared to defend the mating site, because one female clearly monopolized a site at a time. This was a key difference between the observed system and a lek: in both, multiple males assemble to locate mates, but in the latter, each male defends a display area. Although almost no aggression was seen between females during this study, Oxford & Bish (2000) observed aggression between F1 and another female on 1 day in 1999 at Site 1, followed by the monopoly of the site for 4 days by F1. Additionally in 2001, J. Keeling (pers. comm.) observed two females present on Site 1 for several days, which initially fought intensively before one monopolized all matings. Aggression might be expected if two females' receptive periods coincide. In this study, males showed no inclination towards mating with a second female while the first female was present on the site. This could be explained if the probability of ovulation increased with the period of female receptivity: if so, the first female to appear would be more likely to have ovulated, so that mating with the second would be less beneficial than competing to mate with the first.

We have found no other report of a single female monopolizing a traditional mating site, with a new female replacing the departing one. What feature of the fossa would encourage such an unusual strategy is not clear. In species using leks or similar mating systems, males defend small territories where females predictably visit during the mating season (Höglund & Alatalo, 1995). In these cases, the females are typically numerous and competition to monopolize the area would be prohibitively intense. The traditional sites identified during this study, however, are sufficiently numerous for only a few females to mate on each one, so that monopoly of one site by a single female during her receptive period is possible. While this argument fails to explain why only fossas have been seen to follow this strategy, perhaps too few solitary species have been studied adequately.

The fossa mating system enables females to mate with multiple males, and to mate repeatedly with some or all of these males. Because females spent up to 6 days on a mating site, mating for a large proportion of this period (Table 1), a substantial benefit from multiple mating that outweighs such costs is expected. Hypotheses explaining female multiple and repeated mating are numerous (Reynolds, 1996; Jennions & Petrie, 2000) and our observations are insufficient to reject any, other than those concerning maintenance of a pair bond.

The female strategy appeared to maximize the number of mates. Firstly, the relatively long periods of behavioural oestrus seen in the first two females were associated with an accumulation of males that regularly visited the mating site (Fig. 2). Each female was seen to mate with a similar number of males, but achieved this in decreasing periods of time. The females that mated later thus appeared to benefit from having waited.

Male–female conflict was apparent over the ending of intromission. The female often sought to end intromission before males, assisted by her precarious position in the tree, while the male sought to continue intromission through a copulatory tie, guarding and biting (Albignac 1970, J. Keeling, pers. comm.). This conflict implies that the period of intromission had a significant effect on the probability of the mating male fathering the female's young. In controlling the intromission period, the female may have been choosing between males, a behaviour which is little understood in mammals. While no correlation was found between intromission period and measured male attributes, this is inconclusive: the former was often not fully measured, because the start of mating was not observed. Furthermore, where fewer than 10 males are being sampled by a female, correlations may be low even if females are expressing a choice (Benton & Evans, 1998).

Conflict was less marked over the initiation of intromission. The female's position in the tree may reduce harassment. Common in many species (Clutton-Brock & Parker, 1995), harassment might represent an especially high risk where females are at low density and receptive only once a year. In the tree, the female remained isolated from the several competing males, which limited one another's approaches to her, and she had some control over the termination of intromission. Captive females have been observed to climb as high as they can when approached by a male for mating (M. Klockenkämper, pers. comm.; T. Hornsey, pers. comm.). Harassment has also been proposed as an explanation for the masculinization of juvenile fossa females (Hawkins et al., 2002).

Conclusions

The mating system of the fossa, whereby females monopolize a site and maximize the available number of mates, appears to be unique. The present study adds to the wealth of evidence (Zeh & Zeh, 2001) that, contrary to the assumptions behind currently available schemes to explain mammalian systems, females also benefit from multiple mating. There is evidence of females mating multiply among other solitary carnivores (e.g. Schenk & Kovacs, 1995; Gehrt & Fritzell, 1999), but little is known of how mates locate one another. A lack of den use, and patchy dispersion of a limited resource, may favour the fossa's system. Where water is plentiful, fossas may either use sites related to a different resource, or exhibit a different mating system. We are hopeful that the continued investigation of the benefits to females of multiple mating will better explain mammalian systems. In the case of the fossa, genetic studies to establish the consequences of different intromission periods and mate combinations on the paternity of the offspring will likely reveal benefits to both males and females of this unusual behaviour.

Acknowledgements

We thank the Leverhulme Trust for funds; the Commission Tripartite of the Malagasy Government, the Université d'Antananarivo and the Département des Eaux et Forêts for our research permit; field assistants, particularly Rabeantoandro Zoelisoa and Rasamison Aimé; Victoria Braithwaite, Michael Jennions and an anonymous reviewer for helpful discussion and comments on earlier drafts. The research described above complied with the current laws of Madagascar.

Ancillary