Genetic and spatial structure within a swift fox population



    Corresponding author
    1. Department of Forest, Range and Wildlife Sciences, Utah State University, Logan UT 84322–5230, USA;
      Ann M. Kitchen, Department of Forest, Range and Wildlife Sciences, Utah State University, Logan UT 84322–5230, USA. E-mail:
    Search for more papers by this author

    1. US Department of Agriculture, Wildlife Services, National Wildlife Research Center, Department of Forest, Range and Wildlife Sciences, Utah State University, Logan UT 84322–5230, USA; and
    Search for more papers by this author

    1. Department of Fish and Wildlife Resources, University of Idaho, Moscow ID 83844–1136, USA
    Search for more papers by this author

    1. Department of Forest, Range and Wildlife Sciences, Utah State University, Logan UT 84322–5230, USA;
    Search for more papers by this author

    1. Department of Forest, Range and Wildlife Sciences, Utah State University, Logan UT 84322–5230, USA;
    Search for more papers by this author

Ann M. Kitchen, Department of Forest, Range and Wildlife Sciences, Utah State University, Logan UT 84322–5230, USA. E-mail:


  • 1We incorporated spatial data on swift foxes (Vulpes velox) with genetic analysis to assess the influence of relatedness between individuals on their social and spatial ecology. We recorded the space use patterns of 188 radio-collared swift foxes in south-eastern Colorado from January 1997 to December 2000. One hundred and sixty-seven foxes were also genotyped at 11 microsatellite DNA loci and the degree of relatedness between individuals was estimated.
  • 2We described the genetic structure of the population by examining the relatedness of neighbours and the relationship between the spatial and genetic distance of all individuals. We found that close kin appeared to cluster within the population. Neighbours were significantly more related (mean R = 0·089 ± 0·01) than non-neighbours (mean R = 0·003 ± 0·01; randomization test, P < 0·0002). Female clusters were more extensive than male clusters.
  • 3The degree of genetic relatedness among foxes was useful in explaining why foxes tolerated encroachment of their home ranges by neighbours; the more closely related neighbours were, the more home-range overlap they tolerated (Mantel test, P = 0·0004). Foxes did not appear to orientate their home ranges to avoid neighbours and home ranges overlapped by as much as 54·77% (x = 14·13% ± 0·41). Neighbours also occasionally engaged in concurrent den sharing.
  • 4Relatedness influenced the likelihood that an individual would inherit a newly vacated home range, with a mean relatedness of range inheritors to previous owners of 0·333 ± 0·074. Thus, the genetic structure of the population and interactions between kin were interrelated to space-use patterns and social ecology of the swift fox.


Mammals interact with their neighbours and other conspecifics in a variety of ways, ranging from the use of non-exclusive home ranges to defence of exclusive territories (Geffen et al. 1999). Home-range defence occurs for a number of reasons, including monopolization of mates and defence of other resources such as food and dens (Davies 1978; Messier & Barrette 1982; Gese 2001). Territorial behaviour is costly, however. Time and energy are spent patrolling the territory, and if encounters with neighbours are aggressive, injury or even death can ensue. Territoriality should occur only when the benefits outweigh the costs (Davies & Houston 1984). The benefits of territoriality will be reduced in cases where resources are abundant and evenly distributed or intraspecific competition for resources is low.

Inclusive fitness may mediate intraspecific competition for resources when relatives are aggregated spatially and likely to interact cooperatively (Hamilton 1964). If kin are less aggressive towards each other than toward non-kin (Waldman 1988), and if inclusive fitness is increased as a result, cooperative behaviours among kin will evolve (e.g. Garza et al. 1997; Hoglund et al. 1999). Kinship has been used to explain affiliative and cooperative behaviour within and between social groups in a variety of birds and mammals (e.g. Packer et al. 1991; Hatchwell et al. 2001; Walls & Kenward 2001; Wimmer, Tautz & Kappeler 2002). Furthermore, variation in relatedness is hypothesized to correlate with variation in behavioural interactions (Gompper & Wayne 1996). For example, the benefits of cooperating with kin may also influence where individuals settle, thereby affecting population genetic structure. It has been suggested that the kit fox (Vulpes macrotis Merriam), a closely related species to the swift fox (Vulpes velox Say), has an ‘expanded social structure’ in which foxes often interact with foxes from neighbouring social groups (O’Neal, Flinders & Clary 1987). This spatial structure may have been influenced by kin effects as kit fox neighbours are often related (Ralls et al. 2001).

Here we describe the spatial and genetic structure of the swift fox. The swift fox is one of the smaller North American fox species, inhabits short and mid-grass prairies of North America, and ranged historically from Canada to northern Texas (Scott-Brown, Herrero & Reynolds 1987). They are highly fossorial, using dens year-round (Egoscue 1979). Little is known of the breeding system or intraspecific interactions of this fox, and the genetic structure of swift fox populations has not been investigated previously. We hypothesized that if swift foxes gain benefits by interacting amicably with kin, this cooperative behaviour would be evident in the degree of overlap between home ranges and in its concurrent use. The more closely related neighbouring swift foxes are, the more their home ranges will overlap and the more individual foxes should tolerate use of overlapping areas. We hypothesized further that if this pattern of cooperation exists, it will be apparent in the genetic structure of the population. Because kin receive benefits from being neighbours, dispersing individuals should settle preferentially near relatives creating kin clusters. Thus, an increased tolerance among neighbouring kin will influence both the spatial ecology and genetic structure of swift fox populations.


study site

The study area (Pinon Canyon Maneuver Site, PCMS) is located in Las Animas County, north-east of Trinidad, Colorado. The foxes sampled inhabited an area of 736 km2. The climate is semi-arid, with a mean annual precipitation ranging between 26 and 38 cm. Mean monthly temperatures range from −1 °C in January to 23 °C in July. Elevations range from 1310 to 1740 m. The two main vegetation types are shortgrass prairie and pinyon pine (Pinus edulis)–juniper (Juniperus monosperma) communities (Shaw et al. 1989). The study area was used primarily for cattle ranching prior to 1982, at which time the US Army acquired the PCMS.

genetic analyses

Genetic relatedness between individual swift foxes within the population on the PCMS was assessed using 11 microsatellite loci. Blood samples were obtained from foxes caught in box traps and were frozen or stored in lysis buffer (Longmire et al. 1991) at a 1 : 5 ratio of blood to buffer. Tissue samples were taken from radio-collared animals found dead and frozen until analysis. Fresh scat samples were collected during trapping allowing for positive correlation between the scat and the fox and then frozen.

DNA was extracted from blood and tissue samples using a blood or tissue Qiagen protocol (Qiagen Inc., Valencia, CA, USA), or a phenol/chloroform protocol (Vardenplas et al. 1984). DNA was extracted from scat samples in a room dedicated to processing low-quantity samples using standard protocols of a Qiagen stool kit, using multiple negative controls to test for contamination. Samples were amplified through a polymerase chain reaction (PCR) with microsatellite primers (Saiki, Scharf & Faloona 1985). PCR products were first run on a 1·5% agarose gel to test the quality of DNA extractions, and if necessary (i.e. no band of the appropriate size appearing on the gel) a second extraction of the alternate type was performed. When PCR amplification was faint or absent after the second extraction, DNA extracts were concentrated and purified to remove inhibitors using standard Geneclean protocols (Qbiogene, Inc., Carlsbad, CA, USA).

Microsatellite primers developed for the dog genome and used for the closely related kit fox (Östrander, Sprague & Rine 1993; Fredholm & Wintero 1995; Francisco et al. 1996; Ralls et al. 2001) were optimized for the swift fox samples (Table 1). The following primers were used successfully: CXX20, CXX30, CXX173, CXX263, CXX403, CXX250, CXX109, CXX2062, CXX377, FH2054 and CPH3. Other primers that were tested but not used were CXX123, CXX225, CXX2001 (rejected due to unsatisfactory amplification) and CXX172, CXX200 and FH2140 (rejected due to an insufficient number of alleles). For blood and tissue samples, the 20 µL PCR reactions were cycled 35 times, with denaturation at 94 °C for 30 s, annealing at 51 or 55 °C, depending on the primer pair (Table 1), for 30 s, and extension at 72 °C for 30 s. Each primer was labelled with a fluorochrome (FAM, HEX or TET). Multiplexes of primers were developed to allow multiple loci to be run simultaneously. CXX20/CPH3, CXX109/CXX30, CXX403/CXX173 and CXX2062/CXX250, were run as multiplexes and CXX263, CXX377 and FH2054 were run as singleplexes. The concentrations of reagents that did not vary between reactions were dNTPs (0·25 mm), bovine serum albumin (BSA) (1·01 mg/mL), regular Taq buffer (1×), and regular Taq (0·5 U). Reagents that varied were the primer concentrations and MgCl2 (Table 1). For scat samples, the PCR reactions were cycled 55 times, and 0·2 U of Gold Taq DNA polymerase replaced regular Taq. Microsatellite genotypes were obtained using an Applied Biosystems 377 sequencer (Applied Biosystems) with a genescan 500-Tamra size standard. The genotypes of the individual foxes were obtained using the software programs genescan‘ version 3·1 and genotyper‘ version 2·1 (Applied Biosystems). The blood and tissue samples from 20 individuals were re-amplified and the observed error rate per single locus genotype was estimated by calculating the number of errors/number of PCRs. Due to low quality and concentrations of DNA, scat samples were analysed repeatedly and the confidence of the genotypes was estimated using the software package reliotype (Miller, Joyce & Waits 2002). reliotype is a program for assessing the reliability of an observed multilocus genotype and for directing further replication if it is not sufficiently reliable. Genotypes were replicated until a confidence level of obtaining a correct multilocus genotype of 99% was obtained (an average of 4·11 PCRs per locus were used for scat samples). For some samples, unreliable loci were dropped from the analysis. Samples were removed from the analysis unless they were assessed at ≥ 9 loci. Due to possible effects on relatedness values, deviations from Hardy–Weinberg equilibrium and linkage disequilibrium were tested using the program genepop (Raymond & Rousset 1995) and corrected for multiple tests using Bonferroni adjustment.

Table 1.  Optimization information, polymorphism information, observed and expected heterozygosities, and reagent concentrations for the microsatellite loci
LocusnAnnealing temp (°C)No. of allelesSize rangeHOHEPrimer (uM)MgCl2 (mM)
CPH316755 6151–1610·6470·6830·10·25
CXX40316755 4273–2810·3050·3050·20·25
CXX26316651 4114–1220·4820·6110·20·19
CXX25016651 7132–1400·4400·6310·30·38
FH205416651 6175–1870·6270·6500·20·25
CXX2016755 9129–1450·7070·7190·30·25
CXX17316755 3124–1280·3170·2980·20·25
CXX10916755 3168–1720·4430·6610·30·25
CXX301225511141–1570·7870·828 0·10·25
CXX206216655 6137–1540·6510·6830·10·38
CXX37716455 8173–1910·5610·6420·20·25
162  6·09 0·5420·610  

Relatedness between individuals was assessed using the program kinshipTM 1·1·2 (Goodnight & Queller 1999). kinshipTM estimates Grafen's relatedness coefficient (Grafen 1985) between all possible pairs of individuals. This coefficient measures the degree to which two individuals share identical alleles, taking into account the allele frequencies in the population and each individual's genotype (Goodnight & Queller 1999). Loci exhibiting lower than expected heterozygosity levels contribute less to the calculation of R than loci with higher levels of heterozygosity. R-values range between −1 and 1. A positive R-value between two individuals indicates that they are more related (i.e. they share more alleles that are identical by descent) than expected by chance, and a negative R-value indicates that they are less related than expected by chance. These calculations are based on a reference set which was made up of all sampled foxes. It is important to note that these do not constitute an isolated population, however.

interaction analyses

The interaction of swift fox neighbours was documented using radiotelemetry. Foxes were captured using double-door box traps (80 × 25 × 25 cm) baited with chicken (Covell 1992). Traps were deployed in the evening and checked the following morning. Trapping ceased during periods when night-time temperatures dropped below −10 °C. A radio-collar and ear tag was attached to the fox and the weight, sex and age of the animal was recorded. To recapture certain individuals in order to change their radio-collar, a trap-enclosure system as described by Covell (1992) was used. All foxes were released at the site of capture. The foxes were caught in five areas on the study site. The boundaries of the five areas were > 1·5 home-range diameters apart, and although some foxes dispersed between areas, no fox inhabited more than one area at a time (Schauster, Gese & Kitchen 2002).

Telemetry procedures followed recommendations by White & Garrott (1990). Radio-tagged foxes were monitored to determine home-range estimates, overlap and range inheritance. Relocations on the foxes were attempted approximately every 1–3 days with locations obtained throughout the 24-h period to reduce bias in home-range estimates. Home-range estimates were obtained by a 95% fixed-kernel range estimator (Worton 1989) with least-squares cross-validation smoothing. A social group was defined as foxes that shared a range and concurrently shared dens (Kitchen, Gese & Schauster 1999). Foxes were considered neighbours if they were residents in adjacent ranges with a common boundary or some overlap, as determined by the 95% fixed-kernel home-range estimator.

We compared the relatedness between neighbours to that between non-neighbours. We evaluated the level of overlap between neighbours of different sex combinations (e.g. male–male, female–male and female–female), and whether the relatedness of neighbours correlated with the level of overlap between their ranges. Mean overlap was calculated using an index that varies between 0 and 100, with 100 indicating complete overlap (Minta 1992).


We assessed whether observed overlap among neighbouring foxes was different from an estimate of overlap expected by chance derived by rotating the ranges randomly about their centroids (Geffen & Macdonald 1992). We compared the actual overlap with the average overlap calculated from randomly rotating one of the ranges three times. If the overlap observed was greater than that expected, individual foxes were considered to be attracted to one another, whereas if the observed overlap was less than expected foxes were assumed to be avoiding one another (Madison 1980). The actual overlap was used instead of the Minta index (Minta 1992) for overlap, as the size of the neighbouring home range was inconsequential to the orientation of a fox's home range.

To evaluate whether foxes temporally avoided their neighbours, we assessed how much neighbouring foxes were using the overlapping areas of their ranges concurrently. To do this, we calculated the simultaneous use of the overlapping area of their range by calculating the percentage of the total points (simultaneous and non-simultaneous) in the overlapping area when both animals were found there within 1 h of each other (deemed simultaneous). This was compared to the percentage of the total points in the non-overlapping area when both animals were located within 1 h using a mixed factor analysis of variance model. The sex combination of the neighbours (female and female, male and female, male and male) was incorporated into the model. Five randomly chosen overlapping neighbours per season per year were used for these analyses. We assessed whether the level of temporal avoidance displayed by neighbours (i.e. the difference between the simultaneous use of overlapping areas vs. non-overlapping areas) was correlated to their relatedness using a Pearson product–moment correlation. We also assessed the frequency of occurrence of den-sharing between neighbours and the relatedness between neighbouring animals that engaged in concurrent den-sharing.

We evaluated the relatedness between swift foxes that inherited empty ranges and the previous owner and compared this to the average population relatedness (the mean relatedness of all dyads of sampled foxes). We assessed this in cases where an apparently single range-holder died, or a mated pair died within a short period. In the latter case, the relatedness of the last pair-member to die was used. If a mated pair began to utilize an empty range at the same time, the more closely related of the members to the former range-holder was used. This method was chosen as swift fox pairs are generally unrelated (Kitchen 2004); thus, only one individual would be related to the previous home-range owner. We believe this method to be valid and did not introduce bias into the analysis; the average R-value of single inheritors (mean R = 0·381 ± 0·093, n = 7) was higher than the more closely related of the members of a mated pair (mean R = 0·217 ± 0·108, n = 3).

If kin settled near each other, we hypothesized that distance between home-range centroids would be correlated with genetic relatedness. That is, the more closely related foxes were to one another, the closer their home ranges would be. Distance was grouped into < 3 km (approximately distance between neighbour centroids), 3–6 km (approximate distance between neighbours twice removed) and more than 6 km. Because of the issue of pseudoreplication in the multiple pairwise comparisons of fox relatedness, we used Mantel's randomization tests (Mantel 1967) and analysis of variance and two-sample tests by randomization performed in the software program rt version 2·1 (Manly 1997) instead of conventional statistical tests. These randomization tests were carried out with 5000 permutations to assess: (1) the relatedness of male individuals to other males vs. female individuals to other females within the population; (2) the relatedness of neighbours vs. non-neighbours; (3) the relationship between geographical distance and relatedness of individuals; and (4) whether there was a relationship between home-range overlap and relatedness. Probability values in multiple comparisons were adjusted using a Bonferroni adjustment.

For the purposes of analyses, we defined seasons on the basis of energetic demands (due to climatic changes and prey abundance) and behavioural characteristics (including breeding, gestation, pup-rearing and dispersal) which were applicable to swift foxes as follows: pup-rearing season (15 April−14 August), dispersal season (15 August−14 December) and breeding/gestation season (15 December−14 April). Spatial analyses were performed using arcview version 3·0 (Environmental Systems Research Institute, Inc., Redlands, CA, USA). Statistical analyses were performed using sas (2001). Data were examined for normality and homoscedasticity in all parametric tests. Data were log-transformed for assessment of spatial avoidance owing to deviations from normality.


We obtained 32 556 locations on 188 swift foxes, with continuous data collection from January 1997 to December 2000. We analysed genetic samples on 167 foxes. The mean proportion of individuals genotyped at each locus was 0·972, and 164 individuals were genotyped at ≥ 10 loci. The error rate of genotyping was estimated at 0·9%. Forty-four of 52 scat samples met the data reliability criteria discussed in the Methods; the remaining samples were omitted from analysis. Observed heterozygosities per locus varied from 0·305 to 0·787 with an average of 0·542 (Table 1). Significant deviations from Hardy–Weinberg equilibrium were observed at five of the 11 loci when testing the population as a whole after Bonferroni adjustment for multiple tests. When the Hardy–Weinberg equilibrium was tested within each of the five sampling areas individually, there was an average of 1·6 loci per area that deviated significantly (Table 2). A number of factors may influence Hardy–Weinberg equilibrium within a population, such as non-random mating. However, because relatives appear to be clustered within our population (see below), the deviations from Hardy–Weinberg equilibrium seen in our population may be due to the Wahlund effect (Wahlund 1928). The Wahlund effect occurs when populations with different allelic frequencies are combined in a single sample. The kin clusters were essentially subpopulations within the population. This is supported further by the fact that deviations were reduced when we tested each area separately, and that we had a low error rate reducing the possibility that deviations were due to null alleles. This effect has also been seen in kit fox populations, where similar kin clustering occurs (Ralls et al. 2001). Loci pairs that were in linkage disequilibrium at the 0·05 level in the population overall after adjustment for multiple tests using a Bonferroni adjustment were CXX30 and CXX377, CXX2062, CXX173; FH2054 and CXX20, CXX250; and CXX2062 and CXX377. We tested for linkage disequilibrium within each area, and found that there was an average of 1·8 loci pairs in disequilibrium per area (compared to the six pairs that were in disequilibrium in the population overall). Of the six loci out of equilibrium for the entire data set only one pair (CXX30 and CXX2062) was out in more than one area and this pair was out of equilibrium only in areas 1 and 4. Areas 2 and 5 show no disequilibrium. Thus we feel that the observed linkage disequilibrium in our data set is due to population substructure and sporadic effects rather than physical linkage of these loci in the genome.

Table 2.  Exact test P-values for Hardy–Weinberg (HW) equilibrium for the whole population and each area individually adjusted for multiple comparisons using Bonferroni adjustment. Values marked * indicate significant deviation from HW equilibrium at the level of 0·05
LocusWhole populationArea 1Area 2Area 3Area 4Area 5

Females and males had similar relatedness within the population (mean R = 0·009 ± 0·004 vs. 0·008 ± 0·004; randomization test, no. of dyads = 999, P = 0·362). Neighbours were significantly more related (mean R = 0·089 ± 0·01) than non-neighbours (mean R = 0·003 ± 0·01; randomization test, no. of dyads = 990, P < 0·0002). There was significantly more home-range overlap with increasing levels of relatedness (Mantel test, no. of dyads = 999, P = 0·0004, Fig. 1).

Figure 1.

Average Minta index for overlap (± SE) between neighbouring home ranges for four classes of relatedness between swift foxes in south-eastern Colorado, 1997–2000.

Overall, the mean overlap between neighbouring home ranges was 14·13% ± 0·41. The distribution of overlaps was negatively skewed, however; neighbours overlapped each other's ranges by as much as 54·77%. Male–male neighbours exhibited a slightly smaller overlap ( = 13·64% ± 0·789) than female–female neighbours ( = 14·239% ± 0·803) or female–male neighbours ( = 14·308% ± 0·589). This overlap was not significantly different (randomization test, no. of dyads = 999, P = 0·95).

When assessing spatial avoidance, we found that the actual overlap ( = 1·20 km2 ± 0·14) was not significantly smaller than the overlap seen from the rotated ranges ( = 1·28 km2 ± 0·13; F2,93 = 0·02, P = 0·89); thus, the foxes did not appear to be orientating their ranges in such a way as to reduce overlap with their neighbours.

There was significantly less simultaneous use of the overlapping area ( = 3·49% ± 0·89) than there was of the non-overlapping areas of neighbouring home ranges ( = 21·52% ± 1·40; F1,98 = 102·96, P < 0·0001). There was no difference in temporal avoidance for the different sex combinations (F2,98 = 0·43, P = 0·653) or the different seasons (F2,98 = 0·99, P = 0·374). In addition, there was a significant positive correlation (r58 = 0·31, P = 0·028) between the relatedness of neighbours and the level of simultaneous use of the portion of their home ranges that overlapped relative to the simultaneous use of non-overlapping areas.

Concurrent den-sharing of non-social group foxes occurred nine times, eight times between neighbours and once between a resident and a transient. In seven of the nine cases, the sharing occurred between members of the same sex. Four cases occurred in the breeding season and five in the dispersal season. Concurrent den-sharers from neighbouring ranges were generally not closely related (mean R = 0·105 ± 0·177).

We found that the 10 range inheritors were. on averag. highly related to the foxes from which they inherited the range ( = 0·333 ± 0·074). This was significantly higher than the average relatedness of all dyads in the sample ( = 0·009 ± 0·003; t = 4·37, d.f. = 9, P = 0·0018). There appeared to be no sex bias in inheritance, with two males and three females inheriting ranges from males, and two males and three females inheriting ranges from females. There were three range inheritors that were collared prior to the death of the range owner. All three range inheritors came from their natal dens; two from neighbouring ranges and one from two home ranges away.

The geographical distance between foxes’ home-range centroids was correlated to relatedness, with higher levels of relatedness among foxes with geographical proximity for all sex combinations (Table 3, Fig. 2). Females exhibited a more extensive kin clustered pattern with higher relatedness extending for larger distances than that seen between males (Table 3, Fig. 2).

Table 3.  Randomization test significance values and number of dyads assessed (restricted to 999) when comparing the average relatedness relative to the distance between their home-range centroids for female–female (FF), male–male (MM) and female–male (FM) sex combinations of swift foxes in south-eastern Colorado, 1997–2000. Values are adjusted for multiple comparisons using a Bonferroni adjustment
Distance (km)FFMMFM
PNo. of dyadsPNo. of dyadsPNo. of dyads
< 3 vs. 3–6  0·107402  0·017278  0·039962
3–6 vs. 6 +  0·001999  0·375999  0·002999
Overall< 0·001999< 0·001999< 0·001999
Figure 2.

Average relatedness for three classes of distance between home-range centroids for female–female (FF), male–male (MM) and female–male (FM) sex combinations of swift foxes in south-eastern Colorado, 1997–2000.


Our data suggest that swift fox populations are genetically structured at a fine-scale with kin clustering evident. Neighbours were related more closely than expected from the average population relatedness. In addition, foxes were increasingly related as the spatial proximity of their home ranges increased (Table 3, Fig. 2). We propose that the clustering among kin in the swift fox population has led to an increased level of tolerance among neighbours. We found a positive correlation between tolerance levels and the degree of kinship. Overlap of neighbours increased with increasing relatedness (Fig. 1), as did the use of overlapping areas. While the influence of relatedness on tolerance has not always been clear in field studies (e.g. Spong & Creel 2004), a positive correlation between cooperative behaviour and relatedness has been noted in carnivores, primates and birds (e.g. Morin et al. 1994; Hatchwell et al. 2001; Widdig et al. 2001; Creel & Creel 2002). Girman et al. (1997) found that African wild dog (Lycaon pictus) neighbours show higher relatedness than expected by chance, and hypothesized that dispersal to neighbouring packs may reduce the frequency and intensity of interpack encounters.

The tolerance between neighbours and its positive correlation with relatedness in swift foxes demonstrate that kin facilitation (Hamilton 1964) may play an important role in the social ecology and space-use patterns of the swift fox. The spatial organization of kin clusters probably facilitated the foxes by reducing the costs inherent in home-range defence, and also benefited foxes in that empty ranges were often inherited by related foxes. Inheritance of ranges by relatives has been seen in birds (e.g. Emlen 1991; Cockburn 1998) and other canids (Moehlman 1989; Schmidt & Mech 1997). Tolerance extended to den-sharing between neighbours. However, the exhibition of den-sharing in swift foxes with unrelated individuals was unexpected but may be due to the presence of a predator as dens appear to be important in predator avoidance for the swift fox.

The difference between swift foxes and other canids that do not exhibit both spatial–genetic structuring and tolerance of neighbours due to relatedness may be due to either intrinsic behavioural factors or extrinsic human-caused influences. Tolerance of neighbours will be viable only when breeding opportunities are not limited. Many canids that have been studied genetically live in packs (e.g. coyotes, Canis latrans, Lehman & Wayne 1991; Williams et al. 2003) or confined areas (e.g. island foxes, Urocyon littoralis, Roemer et al. 2001); thus, the stronger competition for mating opportunities may have led to decreased tolerance of kin. Kin competition would lead to higher dispersal rates and less spatial–genetic structuring such as that seen in coyotes (e.g. Lehman & Wayne 1991) or, where dispersal is not possible, to decreased tolerance of neighbours such as that seen in the island fox (Roemer et al. 2001). Alternatively, the lack of spatial–genetic structuring in many canid populations may be due to the high turnover of individuals resulting from high mortality due to factors such as human exploitation (e.g. Williams et al. 2003).

The kin-clustering evident in swift foxes indicates that settlement decisions may be influenced by the relatedness of neighbours, and indeed short-range dispersal was seen in the swift fox population (Schauster et al. 2002). Dispersal is influenced by kinship in African wild dogs (Girman et al. 1997), with dispersal events often coinciding with a change in dominance hierarchy and dispersers often moving to areas close to relatives. Long-range dispersal events were also documented (Schauster et al. 2002) and may be an important mechanism of inbreeding avoidance (Gandon 1999). However, the reduced risk and benefits of kin facilitation when living in clusters of relatives seems to have selected against obligatory long-range dispersal in the swift fox. Alternatively, the kin clustering may have been a result of the dispersal patterns displayed by swift foxes.

There was a difference between sexes in the structure of relatedness within the population with female kin clusters more extensive than male kin clusters. Many species exhibit a sex bias in the degree of philopatry to natal sites or social groups (Greenwood 1980). A bias towards female philopatry is most common among social mammals (Eisenberg 1997), and is typical of small-bodied canids (e.g. red foxes, V. vulpes, von Schantz 1981; bat-eared foxes, Otocyon megalotis, Nel, Mills & Van Aarde 1984; crab-eating foxes, Cerdocyon thous, Macdonald & Courtenay 1996). There was a slight bias towards male dispersal within the swift fox population with more males dispersing than females (Karki 2003).

The tolerance shown to neighbouring related foxes indicates that swift foxes can identify related individuals and maintain long-term relationships with them. Continuing social relationships between adults and dispersed offspring have also been documented in crab-eating foxes (Macdonald & Courtenay 1996), and is likely in kit foxes (Ralls et al. 2001). The ability to recognize relatives and adjust one's behaviour accordingly has important implications for the evolution of mammalian social systems.

In conclusion, our data suggests that a kin-clustered structure occurs in swift fox society. We found that tolerance of conspecifics was correlated positively with their degree of kinship. Thus, the indirect or kin-selected benefits accrued from living in kin clusters may be of fundamental importance to the evolution of swift fox breeding systems and space use patterns.


Funding and logistical support were provided by the Rob and Bessie Welder Wildlife Foundation, Sinton, TX, the US Army, Directorate of Environmental Compliance and Management, Fort Carson, CO, through the US Fish and Wildlife Service, Colorado Fish and Wildlife Assistance Office, Golden, CO and the Utah Cooperative Fish and Wildlife Research Unit at Utah State University. Additional support provided by the US Department of Agriculture, Wildlife Services, National Wildlife Research Center, Logan Field Station at Utah State University. We thank T. Warren, G. Belew, R. Bunn, and B. Rosenlund for support and project coordination. We thank M. Klavetter, E. Bergman, K. Bly, J. Bolis, C. Bromley, R. Cavallaro, L. Gorman, S. Hahn, C. Hamblin, K. Hansen, R. Hare, E. Joyce, S. Kiffe, J. King, S. Langeland, S. Lupis, S. McLellan, J. Milner, M. Pangraze, L. Schafer, L. Schleub, L. Schutte, J. Stamp, H. Tall, J. Weber, W. Weber, M. Wedermyer, B. Wirchansky and T. Young for assistance in the field, P. Terletzky for geographical help and J. Adams, M. Murphy, C. Miller, C. Anderson and other graduate students and employees at the Laboratory of Ecological and Conservation Genetics at University of Idaho for their help in the laboratory. K. Ralls provided optimization specifications of primers used for the kit fox. J. Bissonette, T. Edwards, K. Mock, D. Ramsey, K. Sullivan and two unidentified reviewers provided helpful comments on the manuscript. Research protocols were approved by Institutional Animal Care and Use Committees at Utah State University and the National Wildlife Research Center.