Role of forage, habitat and predation in the behavioural plasticity of a small African antelope


  • Justin S. Brashares,

    Corresponding author
    1. Centre for Applied Conservation Biology, Department of Forest Sciences, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
    Search for more papers by this author
  • Peter Arcese

    1. Centre for Applied Conservation Biology, Department of Forest Sciences, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
    Search for more papers by this author

Justin S. Brashares, Centre for Biodiversity Research, University of British Columbia, 6270 University Boulevard, Vancouver, BC, Canada V6T 1Z4. Tel: (604) 822–0862, Fax: (604) 822–0653. E-mail:


  • 1Ecological influences on social organization can best be measured by comparing populations of a socially variable species living in different environments. We examined causes of variability in social behaviour in a socially plastic small antelope, the oribi, Ourebia ourebi Zimmermann, in Ghana, West Africa.
  • 2In total, 161 individually identified oribi were observed at five sites along an ecological gradient through dry and rainy seasons. Behavioural observations and ecological data were used to test the relative influence of forage abundance and quality, habitat structure and predation pressure on female dispersion and male social behaviour.
  • 3Of nine ecological variables that were quantified, forage abundance and quality accounted best for variation in female dispersion among and within study populations. Specifically, female oribi formed larger groups and had smaller ranges where dry season forage was relatively abundant and low in fibre.
  • 4Male territorial behaviour differed among sites and was related to female home range size. Males were most active in territory maintenance where females had small home ranges, and males defended a female rather than a territory where females ranged widely.
  • 5These results show that variation in social organization among oribi subpopulations in Ghana reflects female responses to the availability and quality of dry season food resources and male responses to the variable distribution and ranging behaviour of females.


Resource distribution, habitat structure and predation pressure are all thought to drive the evolution of variation in social organization among species (Crook 1965; Jarman 1974; Bradbury & Vehrencamp 1977). More recent studies of mammals, birds, herptiles, fish and insects suggest that the extent to which these ecological factors influence social organization can best be measured by examining a single variable species in a range of environments (reviews in Lott 1991; Foster & Endler 1999). Intraspecific comparisons of populations minimize phylogenetic influences relative to those of ecology on social behaviour (Harvey & Pagel 1991; Brashares, Garland & Arcese 2000). However, few species vary sufficiently in social behaviour to permit detailed intraspecific comparisons, especially in the large mammals (but see Thirgood, Langbein & Putman 1999; Maher & Lott 2000). A notable exception occurs in a small African antelope, the oribi, Ourebia ourebi. Here, we describe how variation in habitat structure, predation pressure, and food quality and abundance influence female movement and group size, and male social behaviour of oribi in Ghana, West Africa. Several studies of large mammals have tested separately for a link between forage abundance and animal demography (e.g. Owen-Smith 1990), predation pressure and social organization (e.g. van Schaik & van Hoof 1983), or female dispersion and territoriality in males (e.g. Carranza, Garcia-Munoz & Dios Vargas 1995). In this study, we examined simultaneously and quantitatively these and other predicted links between ecology and behaviour.

Oribi occur in a mosaic of contiguous subpopulations throughout Ghana in habitats that vary from moist, open savannah on the southern coast to mixed woodland and subdesert scrub along the country's northern border (Ghana Wildlife Department, unpublished data; J. S. Brashares, unpublished data). Thus, in Ghana, oribi can be studied in a region that exhibits substantial ecological variation within a relatively small geographical area. In 1999, we studied 60 groups of oribi on territories at five sites in Ghana to address longstanding hypotheses about links between ecology and social organization.


The links between ecology and social organization can be summarized within a conceptual framework that we call the ‘resource, habitat and predation’ model (Fig. 1; reviewed below). Three constituent hypotheses of this model relate to the dispersion of female oribi in response to ecological conditions; the fourth relates to the behaviour of males in response to female dispersion. The first three hypotheses are non-exclusive, but may vary in effect. The first hypothesis posits that female group size is related positively to food abundance and quality. This may occur where competition for high-quality food is minimal, food occurs in patches, and the risk of predation either favours larger groups or does not influence group size (Clutton-Brock 1989; Davies 1991). This ‘food abundance and quality hypothesis’ assumes that females range only as widely as needed to meet their nutritional requirements. Thus, under this hypothesis for the oribi female home range size should vary inversely with food abundance and quality, and female group size should vary positively with food abundance and quality.

Figure 1.

Hypothesized links between precipitation and habitat structure, habitat structure and female dispersion, and female dispersion and male behaviour.

The response of females to the abundance, quality and distribution of food is probably also influenced by the risk of predation. In particular, the ‘predation pressure’ hypothesis predicts that female home range and group size are a function of predation pressure, and that oribi, being an open-country species, should form larger groups with smaller home ranges where predation pressure is high (van Schaik & van Hoof 1983; Armitage 1988). However, Jarman (1974) pointed out that predator avoidance strategies also affect the home range and group size of female antelope, and that these traits should therefore vary depending on available cover. Thus, the ‘habitat structure’ hypothesis suggests that habitat structure interacts with predation pressure to affect female home range and group size (Walther 1977; Langbein & Thirgood 1989). This hypothesis predicts that where oribi occur in more closed habitats females should survive better when they occur singly and on small home ranges, than they would if they inhabited groups ranging over a larger area. This is because the advantages of protective cover and familiarity with hazardous and hiding sites for single females should be maximized on small home ranges. In open habitats, however, female oribi might be expected to form groups and occupy larger home ranges to take advantage of the dilution effect and advance warning of attack.

Finally, the ‘female dispersion/male behaviour’ hypothesis follows the first three hypotheses and posits that male behaviour, and ultimately the form of the social system, is determined by female dispersion (Emlen & Oring 1977). Thus, where females occur in stable groups on small home ranges, male oribi should defend territories encompassing the range of a female group (‘resource defence polygyny’; Clutton-Brock 1989). Where females occur solitarily on large home ranges, the increased cost of territory defence should favour a ‘following’ strategy, wherein male oribi guard a single female throughout its home range (‘female defence monogamy’; Clutton-Brock 1989).

Several predictions arise from this hypothesis as applied to male oribi. First, for males that defend and follow females rather than a territory, the average distance between males and females on the same home range should be less than where males defend territories. Second, earlier work has shown that scent marking by males is an integral part of territory maintenance in oribi (Brashares & Arcese 1999a). Thus, under the female dispersion/male behaviour hypothesis males should scent mark at higher rates where they defend territories as opposed to females. Third, related to the previous prediction, trespass on male territories should be greater, and the area used by a male less exclusive, where females occur singly on large home ranges.

In addition to testing the hypotheses listed above, we also compare some costs associated with the defence of females on large home ranges vs. the defence of territories. We predicted that males would spend more time active and would travel greater distances where females were dispersed singly on large home ranges.



The oribi is a small (c. 15 kg) savannah antelope (tribe Neotragini) that is found in every sub-Saharan country except Gabon (Kingdon 1997). Oribi are described as monogamous in parts of their range, and polygynous and polygynandrous in other areas (Rowe-Rowe, Everett & Perrin 1992; Arcese, Jongejan & Sinclair 1995). From one to four adult male oribi defend a year-round territory of 0·1–0·9 km2 (Brashares & Arcese 1999a), but males may also defend a female rather than a territory (Adamczak 1999). Female group size varies greatly across the oribi's range from one female per territory in northern Ghana (J. S. Brashares, this study) to a maximum of eight females per territory in northern Tanzania (Brashares & Arcese 1999a). The oribi feeds primarily on savannah bunch grasses and is Africa's smallest grazing ungulate (Hofmann 1989; Estes 1991).

Data were collected simultaneously on five oribi subpopulations in Ghana from January to June 1999 (Fig. 2). These were located in Kalakpa Resource Reserve (06°20′N, 00°28′E), Bui National Park (08°24′N, 02°13′W), south-central Mole National Park (09°21′N, 01°59′W), north-eastern Mole National Park (10°01′N, 01°22′W) and Red Volta Forest Reserve (10°46′N, 00°35′W). We identified 25–34 individual oribi on 11–14 territories at each study area (161 animals on 60 territories in total) at distances of 30–100 m with 10× binoculars and a 40× spotting scope. Oribi were identified using 14 naturally varying characteristics including horn length and shape (for males only), coat and tail colour, facial markings, the shape and size of the subauricular gland in relation to the eye, the pattern of marks in the ears, and natural scars (Jongejan, Arcese & Sinclair 1991; Brashares & Arcese 1999a,b). Male age was determined by estimating horn length in relation to ear length and number of annuli on horns (see Jongejan et al. 1991).

Figure 2.

The location of study sites and the range of oribi (shaded area) in Ghana.

In each subpopulation, we mapped territory borders using the position of 12–44 dung middens surrounding each territory using a global positioning system (GPS) receiver (see Komers 1996; Brashares & Arcese 1999a,b). We assessed the accuracy of mapped borders by plotting oribi sightings collected throughout the study period on each territory and by plotting the site of conflicts between neighbouring males. Borders mapped using the location of middens differed in size by <6% from those mapped using only the sightings of males and conflicts between neighbours (J. S. Brashares, unpublished research). Territory mapping and the plotting of data points were conducted using ArcView 3·0 geographical information systems software (ESRI 1996).


To quantify male territorial behaviour five 1-h focal watches were conducted bi-weekly for each of 60 males (300 h total). During each watch the behaviour of males was recorded at 5-min intervals as one of the following: sitting (head-up or down), walking, feeding, standing, courting/mating or antagonism towards other oribi (butting, chasing or threatening). We later calculated the total distance travelled by males during the focal period and pooled sitting and standing without motion as ‘inactive’ behaviour, and walking, running and feeding as ‘active’ behaviour. Also at each interval, we recorded the location of the focal male, and the approximate distance of the focal male from the nearest adult female. Focal periods were alternated between morning (6–10 am), mid-day (10 am–2 pm) and afternoon (2–6 pm) sessions. We also recorded all scent marking opportunistically during these periods; a scent mark was recorded each time a male tilted its head and inserted the distal portion of a plant stem into its preorbital gland (Brashares & Arcese 1999a). Last, to calculate rates of trespass at each study site, all movements of focal males across territory borders were recorded.


To measure oribi density and sex ratio, 15–20, 1-km transect counts were conducted in each study area (91, 1-km transects in total). For each oribi sighted within 100 m of the vehicle we recorded its distance, and group size and composition (Mduma & Sinclair 1994). The point of origin and bearing of drive counts were randomly selected. Oribi density was calculated as animals/km2 in each study area. Transect counts underestimated oribi group size by 25% in Serengeti, Tanzania (Arcese et al. 1995) and, in this study, transect-based group size estimates were 20–35% smaller than the size of groups of individually identified animals on the 60 focal territories. Nevertheless, estimated group size based on transects and group size on focal territories were highly correlated (r = 0·69, P < 0·001, n = 60).


Food abundance and habitat structure

For each of the 60 focal territories, twice a month we walked two 100-m transects to characterize vegetation type and abundance, and habitat structure (Krebs 1989). Eight transects were conducted per territory – four each in the dry and rainy seasons. On each transect we stopped at 10-m intervals to identify by genus all vegetation within a 1-m2 sample plot. At each plot we also estimated the percentage of ground covered by all vegetation and the percentage of ground covered by shrub or tree canopy (canopy cover), and we clipped all green grass leaves under 1 m in height to measure the fresh mass of grass available in the plot. Later, each sample was dried to estimate dry mass. Nearly all territories occurred in habitats dominated by one to four species of grass. Last, to measure openness of territories, a range finder was used to estimate at each 10-m interval the distance of visibility from a point approximately 60 cm above the ground (the average height of oribi). At each 10-m interval we alternated the viewing direction between north, east, south and west.

Food quality

Nitrogen (% total Kjeldahl) and neutral-detergent fibre (NDF) in green grass leaves were used as two separate indicators of forage quality (van Soest 1994). Twelve grass samples were collected for laboratory analysis at each of the five study sites once during both the dry (February) and rainy season (May). Samples represented only the green portion of leaf blades from 12 genera of grass that made up more than 95% of the plants eaten by oribi at the study sites (J. S. Brahares, unpublished research). Approximately 60% of the oribi diet comprised grasses of the genus Andropogon, predominately A. gayanus. To measure percentage nitrogen, leaves were ground and digested in boiling sulphuric acid with a copper sulphate catalyst. Nitrogen content was then determined from the resulting digests using colorimetric analysis (van Soest 1994). NDF was measured using the filter bag method and an Ankom 200 fibre analyser (Ankom, Fairport, NY, USA) (Goering & van Soest 1970).

Using the results of the above forage analyses as a reference collection, we then estimated the nitrogen and fibre content in each 1 m2 transect plot (see above) using values derived in the laboratory for each grass genus and weighted by the relative abundance of each genus in each 1 m2 plot. Finally, we calculated for each territory mean nitrogen and fibre values for all plots sampled in the dry (40 plots per territory from January to March) and rainy seasons (40 plots per territory from April to June).


We used 30 years of monthly census data for large mammals in Ghanaian wildlife reserves (see Brashares, Arcese & Sam 2001) to estimate the abundance of species likely to prey on oribi in our five study areas. These species included leopard (Panthera pardus Linnaeus), lion (Panthera leo Linnaeus), caracal (Felis caracal Schreber), serval cat (Felis serval Severtzov), spotted hyena (Crocuta crocuta Erxleben), side-striped jackal (Canis adustus Sundevall) and olive baboon (Papio anubis Linnaeus). We drew randomly 10 of a possible 98–880 wildlife counts conducted between 1995 and 1998 in each study area and tallied the total number of individuals of each species of predator observed during those counts. To allow a comparison of predator abundance among study sites, this process was repeated four additional times for each study area (without replacement) and these were included as independent counts in an analysis of variance. Because large carnivores have declined throughout Ghana in the last 25 years (Brashares et al. 2001), the census data were also used to estimate historic predator abundance using the same method as above for 1971–74. The length and intensity of wildlife counts were similar from month to month and park to park (M. K. Sam, personal communication).


Monthly rainfall data for the period 1993–97 were acquired from the Ghana Wildlife Division for four of our study sites (Kalakpa, Bui and Mole 1 and 2) and from the Ghana Department of Agriculture for the Red Volta study site. To estimate variation in rainfall an evenness-of-monthly rainfall measure was calculated using the Shannon diversity index for each year from 1993 to 1997 (see Bronikowski & Webb 1996). Potential evapotranspiration (PET; Budyko 1974) data were acquired from the UNEP Global Resources Information Database (UNEP-GRID 2001), but we found PET to be highly correlated with rainfall and variation in rainfall (r > 0·90) and a poorer predictor of ecological conditions in our study areas. Thus, we did not include evapotranspiration rate as an independent variable in statistical analyses.


systat 9·0 ( SPSS 1998 ) was used for statistical analyses. Parametric tests were employed when their assumptions were met. A Kruskal–Wallis anova and Spearman rank correlation ( Sokal & Rohlf 1981 ) were used for data that were not distributed normally and a Bonferroni adjustment applied for planned comparisons. The predicted relationships among ecological variables, female dispersion and male mating strategy were tested in two separate ways. First, we used standard univariate analyses to focus on individual links among variables. Second, we used forward and backward step-wise multiple regression analysis ( Draper & Smith 1998 ) to identify ecological variables that best predicted female home range and group size when correlations among independent variables are accounted for. Whenever possible we aggregated data by territory, or, for comparisons among males, by individual. We did not pool data by subpopulation because intraterritory variance was frequently smaller than between-territory variance (for a review of pooling see Leger & Didrichsons 1994 ). Data collected as percentages were transformed by √arcsine, but presented as untransformed values.




Annual rainfall between 1993 and 97 and variation in rainfall differed among subpopulations along a north–south gradient with rainfall highest in the south (one-way anova: F4,20 = 28·45, P < 0·001; F4,20 = 41·67, P < 0·001; annual rain and variation, respectively). The two southernmost subpopulations, Kalakpa and Bui, experienced on average 25–30% more rain annually than Red Volta and upper Mole to the north, with precipitation distributed more evenly throughout the year in the southern subpopulations (Table 1).

Table 1.  Mean (± SE) annual rainfall and variation in rainfall from 1993 to 1997, mean current and historic predator abundance, and population density and sex ratio of oribi at five sites in Ghana. Current predator abundance represents counts from 1995 to 1998 and historic predator abundance represents counts from 1971 to 1974 (see Methods ). Annual rainfall, variation in rainfall, and current and historic predator abundance differed among subpopulations (one-way anova s: F4,55 , P  < 0·01 for all)
 Annual rainfall (mm)aVariation in rainfallaCurrent predator abundanceaHistoric predator abundanceaPopulation density (km2)bAdult sex ratio (f:m)b
  • a

    n   =  5.

  • b

    Population density and sex ratio values are based on 42–78 group sightings during transect counts in each study area.

Kalakpa RR1179 ± 610·93 ± 0·160·82 ± 0·875·82 ± 1·903·201·85
Bui NP1197 ± 570·89 ± 0·173·20 ± 1·547·22 ± 1·853·061·29
Mole NP 11168 ± 480·87 ± 0·164·82 ± 2·139·28 ± 2·893·951·45
Mole NP 21002 ± 430·84 ± 0·143·80 ± 1·727·80 ± 2·343·051·11
Red Volta R  915 ± 360·79 ± 0·180·68 ± 0·602·80 ± 1·053·500·92

Food abundance

The biomass of grasses in the oribi diet, measured by dry masses of leaves, varied among subpopulations with the wettest site, Kalakpa, averaging 23·74 g m−2 of grass compared with 8·06 g m−2 at Red Volta, the most arid site (Kruskal–Wallis anova: H4 = 51·48, P < 0·001; H4 = 48·89, P < 0·001; dry and rainy season masses, respectively; Table 2). Food abundance was related closely to female home range and group size both between and within subpopulations (Tables 3 and 4; Figs 3 and 4). Females had the smallest ranges and occurred in the largest groups on territories where the biomass of green grass leaves was highest, particularly in the case of analyses of data collected during the dry season. Food abundance differed between seasons for all subpopulations pooled (Mann–Whitney U-test: U = 2944·00, P < 0·001, n1,2 = 60 territories).

Table 2.  Dry and rainy season habitat conditions of oribi territories in five subpopulations in Ghana (values are mean ± SE). Sample sizes are number of territories. Habitat conditions in each territory were assessed four times in the dry and rainy season (see Methods ). Leaf mass, and fibre and nitrogen content and habitat visibility and canopy cover differed among subpopulations (Kruskal–Wallis and standard one-way anova s: F4,55 , P  < 0·01 for all). Leaf mass, and fibre and nitrogen content differed also by season (Mann–Whitney U -tests: P  < 0·001, n1,2   =  60 territories, for all)
SubpopulationDry seasonRainy seasonVisibility (m)Canopy cover
Dry mass (g m−2)Leaf fibre (% NDF)Leaf nitrogen (% Kjeldahl N)Dry mass (g m−2)Leaf fibre (% NDF)Leaf nitrogen (% Kjeldahl N)
Kalakpa RR (n = 11)23·74 ± 2·3824·92 ± 4·200·94 ± 0·0439·31 ± 2·4915·84 ± 3·581·60 ± 0·0659·36 ± 6·112·23 ± 0·79
Bui NP (n = 11)20·76 ± 2·8033·07 ± 4·590·83 ± 0·0430·23 ± 1·7620·04 ± 0·901·36 ± 0·0347·22 ± 7·033·14 ± 0·71
Mole NP 1 (n = 12)19·48 ± 1·9132·37 ± 6·760·81 ± 0·0429·43 ± 2·5719·59 ± 1·361·35 ± 0·0546·70 ± 6·543·00 ± 0·91
Mole NP 2 (n = 12)12·85 ± 2·2439·74 ± 5·080·72 ± 0·0427·20 ± 2·3624·20 ± 3·351·20 ± 0·0455·46 ± 8·442·47 ± 0·43
Red Volta R (n = 14)  8·06 ± 2·4746·06 ± 5·040·64 ± 0·0220·95 ± 1·8224·32 ± 2·621·18 ± 0·0564·74 ± 5·472·21 ± 0·54
Table 3.  Correlation coefficients between female home range (HR) and group size and ecological variables. n   =  60 unless noted otherwise
VariableHR sizeGroup sizeGrass biomassa% Fibrea% NitrogenaVisibilityCanopy coverPredator Abun.bPopulation densitySex ratioAnnual rainfall
  • *

    P  < 0·05;

  • **

    P  < 0·01;

  • ***

    P  < 0·001.

  • a

    Represents dry season value.

  • b

    Represents predator abundance for period 1995–98.

  • c

    n   =  5.

Group size −0·58***          
Grass biomassa −0·69***0·55***         
% Fibrea0·78*** −0·64*** −0·60***        
% Nitrogena −0·64***0·61***0·68*** −0·73***       
Visibility0·16 −0·02 −0·36*0·27* −0·31*      
Canopy cover −0·17 −0·080·35* −0·38*0·30* −0·24     
Predator abundanceb −0·19 −0·070·31* −0·34*0·21 −0·46**0·18    
Population density0·110·010·200·05 −0·14 −0·120·090·39c   
Sex ratio −0·37*0·71***0·74*** −0·60***0·71*** −0·220·190·14c0·12c  
Annual rainfall −0·36*0·45***0·89*** −0·79***0·83*** −0·29*0·43**0·13c0·19c0·58c 
Rainfall variance −0·42**0·54***0·91*** −0·80***0·84*** −0·41**0·48**0·24c0·29c0·59c0·96*c
Table 4.  Regressions of female home range and group size and grass biomass, fibre and nitrogen in the dry and rainy seasons using simple least squares regression. Sample sizes represent number of territories. Values shown represent r
VariablenDry seasonRainy season
Biomass% Fibre% NitrogenBiomass% Fibre% Nitrogen
  • P  < 0·10;

  • *

    P  < 0·05;

  • **

    P  < 0·01;

  • ***

    P  < 0·001.

HR size
All subpopulations60 −0·69***0·78*** −0·64*** −0·64***0·54*** −0·61***
Kalakpa RR11 −0·520·66 −0·61 −0·290·48 −0·38
Bui NP11 −0·710·49 −0·58 −0·410·50 −0·49
Mole NP 112 −0·78*0·74 −0·72 −0·560·69 −0·47
Mole NP 212 −0·80*0·81* −0·70 −0·710·74 −0·66
Red Volta R14 −0·89**0·77* −0·79* −0·78* −0·75* −0·72*
Group size
All subpopulations600·55*** −0·64***0·61***0·51*** −0·48**0·44**
Kalakpa RR110·49 −0·540·440·40 −0·410·34
Bui NP110·57 −0·600·480·35 −0·280·29
Mole NP 1120·70 −0·82*0·650·56 −0·440·52
Mole NP 2120·76* −0·79*0·660·68 −0·610·48
Red Volta R140·88** −0·83*0·71*0·73* −0·590·55
Figure 3.

Plots of female home range size in relation to (a) grass biomass and (b) grass fibre content for 60 territories (see Methods ) separated by subpopulation. Subpopulations are Kalakpa (○), Bui (◊), Mole 1 (▪), Mole 2 (□) and Gbele (•). Trend lines represent the ordinary least-squares regression for each subpopulation. Associated statistics are provided in Table 4 .

Figure 4.

Box-plots of female group size in relation to grass (a) biomass, (b) fibre content and (c) nitrogen content for 60 territories (see Methods ). Box-plots show the median (horizontal line), interquartile range (box), and range of the data (whiskers). Associated statistics are provided in Table 4 .

Food quality

The fibre content (NDF) of preferred grasses varied among subpopulations along a rainfall gradient both in the dry (Kruskal–Wallis anova: H4 = 41·02, P < 0·001) and rainy seasons (H4 = 36·40, P < 0·001; Table 2) with the highest fibre values occurring in the northernmost subpopulations. Grass nitrogen content also varied among subpopulations for both dry and rainy seasons (H4 = 53·19, P < 0·001; H4 = 50·49, P < 0·001; dry and rainy, respectively; Table 2). At the maximum in Kalakpa, average nitrogen content was 50% higher and fibre content 45% lower than in Red Volta (dry season values, Table 2). Fibre and nitrogen content differed between seasons for all subpopulations pooled with fibre lowest and nitrogen highest in the rainy season (Mann–Whitney U-test: U = 3372·00 and 3442·00, respectively; P < 0·001, n1,2 = 60 territories, for both).

Food quality, particularly fibre and nitrogen content in the dry season, was related closely to female home range and group size both between and within subpopulations (Tables 3 and 4; Figs 3 and 4). Females had the smallest ranges and occurred in the largest groups on territories where grasses contained low fibre and high nitrogen (Table 4).

Habitat structure

Habitat structure as measured by visibility and fraction of canopy cover varied among subpopulations (Kruskal–Wallis anova: H4 = 32·30, P < 0·001; H4 = 9·39, P = 0·05; visibility and canopy cover, respectively; Table 2), and was related to rainfall and rainfall variation (Table 3). Kalakpa and Red Volta, the southern and northernmost sites, respectively, were the most open habitats with 20–30% greater visibility and up to 35% less canopy cover than sites in Mole (Table 2). Visibility and canopy cover were uncorrelated (Table 3). Moreover, visibility and canopy cover were unrelated to variation in female home range or group size among subpopulations (Table 3).

Predator abundance

Both current and historic estimates of predator abundance varied widely among subpopulations (one-way anova: F4,55 = 38·65 and 35·58, respectively, P < 0·001 for both) with the lowest abundances recorded in Red Volta and Kalakpa and the highest (by 4–6 times) in Mole and Bui (Table 1). Predator abundance was unrelated to rainfall and rainfall variation (Table 3). Current and historic predator abundance was also unrelated to the home range and group size of female oribi (Table 3). Historic (1971–74) predator communities were similar among subpopulations with olive baboon occurring at the highest densities in each study area followed in abundance by side-striped jackal, spotted hyena, lion, leopard, caracal and serval cat. The current (1995–98) structure of predator communities also was similar among subpopulations and to the historic data; however, many species observed in 1974 were less abundant or absent in study areas in 1998 (Brashares et al. 2001).

Density and sex ratio

The number of oribi observed per km2 in transect counts varied by 30% among subpopulations (Table 1), and was unrelated to rainfall and variation in rainfall (Table 3). Population density also was unrelated to female home range and group size (Table 3). The sex ratio of oribi groups observed in transect counts varied by 50% among subpopulations (Table 1), and was unrelated to both rainfall and rainfall variation (Table 3). Female group size, however, was largest and home range smallest in subpopulations where the sex ratio was skewed most toward females (Table 3).

Female dispersion and multiple regression analysis

The mean home range sizes of females varied from 0·12 to 0·95 km2 among subpopulations (one-way anova: F4,55 = 18·96, P < 0·001; Table 5). Home ranges were routinely smallest in Kalakpa and were on average 250% larger in Red Volta and northern Mole (Table 5). The size of female groups resident on focal territories also varied among subpopulations (Kruskal–Wallis anova: H4 = 12·35, P = 0·02) from an average of 2·27 females per group in Kalakpa to 1·03 in Red Volta (Table 5). Home range and group size were correlated negatively (r = −0·58, n = 60, P < 0·001; Table 3); females with small home ranges tended to occur in larger groups than females with larger ranges.

Table 5.  Mean (± SE) home range and group size of female oribi and mean scent mark rate, male–female distance, trespass rate, distance travelled and time active for male oribi in five subpopulations in Ghana. Sample sizes represent number of male territories. Female home range and group size and all measures of male behaviour varied among subpopulations (Kruskal–Wallis anova s: H4   =  12·35–36·62, P  < 0·02 for all)
SubpopulationFemale HR size (km2)Female group sizeMale behaviour
Mark rate (h−1)Male–female distance (m)Trespass rate (h−1)Distance travelled (m h−1)Percentage time active (h−1)
Kalakpa RR (n = 11)0·22 ± 0·112·27 ± 0·3012·20 ± 3·2835·70 ± 5·700·06 ± 0·05244·5 ± 67·924·43 ± 6·45
Bui NP (n = 11)0·34 ± 0·101·91 ± 0·29  8·69 ± 3·1111·30 ± 2·700·11 ± 0·10285·1 ± 63·037·07 ± 6·51
Mole NP 1 (n = 12)0·40 ± 0·131·58 ± 0·19  7·03 ± 2·0210·30 ± 2·600·13 ± 0·18374·7 ± 92·339·45 ± 5·91
Mole NP 2 (n = 12)0·54 ± 0·141·33 ± 0·16  4·90 ± 1·74  4·80 ± 3·300·58 ± 0·20409·9 ± 78·643·70 ± 8·03
Red Volta R (n = 14)0·59 ± 0·201·23 ± 0·14  1·91 ± 0·87  2·70 ± 1·000·51 ± 0·19614·5 ± 165·555·90 ± 9·98

The ecological and environmental variables that together described the greatest amount of variation in female home range and group size are shown in Table 6. Dry season grass fibre content was the best predictor of both female home range and group size among all variables, including grass biomass and nitrogen content, habitat structure, predator abundance, population density, sex ratio, and annual rainfall and rainfall variation (Table 6). Dry season grass biomass and nitrogen content also were significant predictors in the best-fitting models for female group size, and dry and rainy season grass biomass were good predictors of home range size (Table 6). We were unable to increase the percentage of variance accounted for by alternating the order of entry and exit of the variables. Correlations among variables selected in the best fitting models were often significant (r = 0·40–0·73; Table 3), but were not so collinear as to confound the interpretation of results (Draper & Smith 1998).

Table 6.  Best-fitting multiple regression models of female home range and group size against food abundance and quality, predator abundance, habitat structure, sex ratio, population density, rainfall and rainfall variation. Independent variables were selected by forward and backward step-wise analyses
  • a

    r2 =  0·794, n   =  60.

  • b

    r2 =  0·548, n   =  60.

Female group size:a
Dry season grass fibre content −0·007< 0·001
Dry season grass biomass0·1200·002
Dry season grass N content0·0540·012
Female home range size:b
Dry season grass fibre content0·013< 0·001
Dry season grass biomass −0·0360·004
Rainy season grass biomass −0·0080·042


Given that food abundance and quality were related to rainfall and variation in rainfall, and that these factors were also related to female home range and group size, we next asked if rainfall predicted female home range and group size directly. Rainfall and variation in rainfall were significant predictors of female home range size among subpopulations, but little of the observed variation in home range size was explained by these variables (simple least-squares regression: r2 = 0·13 and 0·18; rainfall and rainfall variation, respectively). Rainfall and variation in rainfall also were related to female group size among subpopulations, but were poorer predictors of group size than direct ecological measures such as grass biomass and grass fibre and nitrogen content (Table 3). Female home range size and group size remained unchanged through the dry and rainy season (P > 0·50: U-tests).


Scent marking, male–female distance and trespass

Rate of scent marking, male–female distance and rate of trespass each differed significantly between subpopulations (Kruskal–Wallis anovas: P < 0·001 for all anovas; Table 5) and each was closely correlated with female home range size (Spearman rank correlations: P < 0·001 for all correlations; Fig. 5). Male scent marking rate ranged by an order of magnitude from an average of 12·2 marks h−1 in Kalakpa to 1·91 h−1 in Red Volta. Similarly, the average distance between males and females sharing a territory in Kalakpa was 35·7 m compared with 4·8 m in northern Mole and 2·7 m in Red Volta (Table 5). Males in Kalakpa also experienced approximately 10 times fewer trespassers than males in northern Mole and Red Volta. Scent marking rate, male–female distance and trespass rate were similar in dry and rainy seasons (Mann–Whitney U-tests: P > 0·35 for all).

Figure 5.

Plots of female home range size in relation to (a) male scent mark rate, (b) male-female distance and (c) male trespass rate for 60 territories (see Methods ). Spearman rank correlations: rs   =  0·77, 0·75 and 0·66, for a, b and c, respectively; n   =  60, P  < 0·001 for all.

Male activity budgets

The average distance travelled by males and the percent of time that males were active varied among subpopulations (Kruskal–Wallis anovas: H4 = 31·64 and 48·45 for distance and time active, respectively; n = 60, P < 0·001 for both) with males in Kalakpa active less than 25% of the day on average compared with more than 50% of the day in Red Volta (Table 5). Similarly, males travelled an average distance of 244·5 m per hour of observation in Kalakpa vs. 614·5 m per hour in Red Volta (Table 5). Both distance travelled and time active were related positively to female home range size: males were most active and travelled greater distances where the home ranges of females were large (Spearman rank correlations: r = 0·52 and 0·66, home range × distance travelled and time active, respectively; n = 60, P < 0·001 for both correlations).



We have organized this paper around the resource, habitat and predation model (Fig. 1) to conceptualize links between: (a) ecological conditions and the dispersion of female oribi, and (b) female dispersion and the social behaviour of males. Ecological conditions and the behaviour of female and male oribi differed considerably among our five study sites. Our tests of relationships between ecology and female dispersion support the hypothesis that food abundance and quality, and not habitat structure and predator abundance, influence female home range and group size in oribi. Univariate analyses showed that female oribi formed larger groups and maintained smaller home ranges where grass was most abundant in the dry season and where it was relatively low in fibre and high in nitrogen (Table 3, Figs 3 and 4). Moreover, these results were obtained both across and within subpopulations (Table 4). In addition, regression analyses of female dispersion against the available ecological variables revealed that female home range and group size were primarily associated with dry season grass fibre content and secondarily with dry season grass biomass and nitrogen content, and rainy season grass biomass (Table 6).

Our results also support the hypothesis that female dispersion determines the social behaviour of male oribi. Where females had large home ranges males scent marked at lower rates, maintained closer proximity to a female, and suffered higher rates of trespass than where females had smaller home ranges (Fig. 5). These results suggest that male oribi actively defended territories where the home ranges of females were smaller, perhaps because it was economically feasible to do so (Brown 1964; Lowen & Dunbar 1994). Males were less active in territory defence and more active in defending a mate where females ranged more widely, suggesting that direct defence of females is incompatible with the effective defence of a territory (Jarman 1974; Gosling 1986; Clutton-Brock 1989). Males associated with wide-ranging females also spent less time resting and travelled greater distances than males associated with females that occurred on small home ranges.

Taken together, our results suggest that variation in mating system among and within oribi subpopulations in Ghana reflects female responses to the availability and quality of dry season food resources and male responses to the variable distribution and ranging behaviour of females (Rowe-Rowe et al. 1992). These findings match closely the long-standing theory that resource abundance and quality and female dispersion are the primary forces driving variation in male mating strategy between species among both antelope (e.g. Jarman 1974; Gosling 1986) and other vertebrates (e.g. Crook 1965; Emlen & Oring 1977; Clutton-Brock 1989; Davies 1991). Food availability also has been linked to variation in social behaviour within species (reviews in Lott 1991; Maher & Lott 2000), and studies that have measured food abundance have generally found it to be a good predictor of social organization (e.g. Davies & Houston 1983; Ostfeld 1986; Ims 1987; Maher 2000). Food quality is more difficult to quantify and, perhaps as a result of this, fewer studies have considered how the nutritional value and digestibility of forage affect social organization (reviewed in Maher & Lott 2000). Nevertheless, several authors have suggested that food quality is a key determinant of social behaviour in ungulates (Jarman 1974; Rubenstein 1981; Maher 2000).

Of the ecological variables considered, the fibre content of grasses in the dry season was the best predictor of both female home range and group size (Tables 3 and 6; Figs 3 and 4). This finding is perhaps best explained by the oribi's unusual ecophysiology. Because of their small forestomach and high energy requirements, small ruminants (<20 kg) typically are unable to rely on foods that require slow fermentation and long passage times for breakdown and absorption (Demment & Van Soest 1985; Hofmann 1989). As a result, small ruminants select foods of high nutritional value and low fibre content (Wenninger & Shipley 2000). The oribi is a notable exception to this pattern in that it feeds year-round on grasses that are low in protein and high in fibre (Hofmann 1989; Gagnon & Chew 2000). It is likely that oribi persist in this niche by locating and selecting new grass leaves that are relatively low in fibre and, thus, more quickly digested (Kingdon 1982). If fibre content in dry season grasses is a key component of oribi habitat selection this could also explain our finding that the dispersion of female oribi is linked closely to forage digestibility during the leanest time of the year (Rowe-Rowe et al. 1992).


Predation pressure has been linked to variation in social organization in antelope (Jarman 1974; Estes 1991). Predator abundance differed considerably among our study sites in Ghana. However, predator abundance was unrelated to variation in dispersion of female oribi. Measuring predation pressure is difficult in practice, and it is possible that by measuring only predator abundance we did not account adequately for the pressure placed on oribi by predators in Ghana. However, the census data we used to calculate predator abundance are unusual in their longevity and detail, and we therefore think it is likely that predation pressure in fact plays a secondary role in determining oribi social organization.

It is possible also that total predator abundance is too broad a measure and should be replaced by one in which species are weighted based on their propensity to prey on oribi. However, because the structure of predator communities was similar among our study areas, we think that total predator abundance is a reasonable estimate of relative predation pressure. Furthermore, information on predation rates is limited or non-existent for oribi and other small antelope and, thus, weightings applied to predators would be arbitrary.

We found no evidence to suggest that habitat structure, as measured by canopy cover and visibility, affects female group size or ranging behaviour. This finding contrasts with some conclusions for other antelope species (Lent 1969; Walther 1977). However, these studies did not quantify food abundance or quality. It is possible that if these traits had been measured, they could also have been linked to social organization in these species. It is also possible that visibility and canopy cover are inadequate measures of habitat structure, and that including other measures would have strengthened the link between habitat structure and the dispersion of female oribi. However, a qualitative examination of our study sites in Ghana shows that this is unlikely to be the case. The sites that were most structurally similar, Kalakpa and Red Volta, differed the most with regard to rainfall, the abundance and quality of preferred grasses, and female and male behaviour. Thus, we suggest that the influence of habitat structure on social behaviour is secondary to food considerations.


The density of oribi estimated across our study sites in Ghana was unrelated to female dispersion and male behaviour. This finding is contrary to our prediction and to the results of previous studies of oribi (Arcese et al. 1995; Adamczak 1999) and other ungulates (e.g. Warren 1974; Leuthold 1977; Langbein & Thirgood 1989). We did find, however, similar to previous studies of small antelope (Dunbar & Dunbar 1974; Arcese et al. 1995; Adamczak 1999), that sex ratio was related to both female home range and group size. While there are clear demographic correlates of variation in social organization, it is unclear if density and sex ratio cause variation in social organization or are the by-products of it. For example, polygyny and territoriality may co-occur when the sex ratio is skewed toward females or the sex ratio may become skewed toward females because the dispersion of females allows males to defend more than one female while excluding other males. We might also ask if polygyny occurs because animals occur at high density, or do polygyny and high density co-occur as a result of abundant resources? In each case above, we suggest that the latter explanation is more likely.


Many authors have used annual rainfall as an indirect measure of food quality and abundance for analyses of links between ecology and social behaviour (e.g. Jarman 1979; Maher 2000). Annual precipitation has been shown to predict primary productivity and the densities of several species of African antelope (Owen-Smith 1990; Fritz & Duncan 1993). Here, we found that annual rainfall explained 60–80% of the variance in food abundance and quality among study sites and was related to female home range and group size (Table 3). However, direct measures of food abundance and quality were considerably better at predicting female home range and group size than annual rainfall, explaining 40–60% of variance compared with only 13–20% explained by rainfall. Moreover, annual rainfall predicted not only food abundance and quality but also habitat structure and openness – traits that were not related to variation in social organization. We conclude from these findings that rainfall alone may allow only coarse-grained analyses. In contrast, more direct measures of habitat conditions should facilitate more informative fine-scale comparisons within sites.

It has also been suggested that monthly variation in rainfall may better explain variation in behaviour across geographical ranges where animals adjust their behaviour to the availability of resources in the most limiting season (Dunbar 1992; Bronikowski & Altmann 1996). In support of this idea, variation in rainfall was generally a better predictor of ecological conditions and the behaviour of female oribi than was annual rainfall (Table 3). Nevertheless, the explanatory powers of rainfall and rainfall variation were similar. We expect the relative utility of these variables will depend largely on the precipitation regime where studies are conducted (Bronikowski & Altmann 1996).


Identifying the difference between genotypic polymorphism (different genotypes adapted to different local conditions) and phenotypic plasticity (a single genotype capable of responding adaptively to local conditions) as mechanisms underlying variation is an important element in studies of intraspecific variation, particularly for making inferences about evolutionary processes (Futuyma 1998; Foster & Endler 1999). However, identifying which of these two mechanisms operates will often require common garden or reciprocal transplant experiments (Alcock 1998), which are not feasible for most large vertebrates because of logistic and other difficulties. Here, by identifying variation in social behaviour among contiguous subpopulations of oribi in a small geographical area, we have shown that phenotypic plasticity rather than genotypic polymorphism is the primary mechanism at work. Our finding that behavioural variation occurred within subpopulations as well as between them also supports this conclusion.


We are grateful to the Ghana Wildlife Division for permission to work in Ghana's national parks and reserves. In particular we thank M. Sam, N. Ankudey, W. Oduro and B. Volta for research permits, letters of introduction, access to historical data and advice. We thank C. Kresge, P. Kresge, J. Mason and the staff of the NCRC for logistical support in the field. P. Boateng and staff at the University of Science and Technology, Kumasi, Ghana, provided invaluable assistance with laboratory analyses. We thank A. Kilpatrick, K. Martin, A. Sinclair, J. Smith, R. Dunbar and S. Thurgood for helpful comments on the manuscript. We were supported by the Department of Wildlife Ecology at the University of Wisconsin, the Forest Sciences Department at the University of British Columbia, a grant from the Zoological Society of Milwaukee County to J.S.B., and grants to P.A. from NSF and the National Geographic Society. J.S.B. was supported also by a predoctoral fellowship from the Killam Trust.