Predators in motion: fine-scale dyadic movements reveal interference competition and asymmetric coordination in African lions and spotted hyenas

Abstract

Introduction:

Interference competition among sympatric large carnivores is a central process structuring African predator guilds. However, the moment-by-moment movement decisions that underlie it are difficult to observe directly in elusive species ranging over remote landscapes.

Methods:

We applied a relative-motion analysis to high-frequency GPS telemetry from 17 lions (Panthera leo) and 14 spotted hyenas (Crocuta crocuta) tracked at 5-min intervals across two southern African ecosystems, the Etosha National Park in Namibia and the Chobe–Linyanti system in northern Botswana. We deliberately restricted the dyadic analysis to timesteps with real, concurrent fixes from both individuals rather than continuous-time-imputed positions. Across the 23 dyads that met this criterion, we classified each individual's movement at each timestep as toward, orthogonal to, or away from its partner. We then tested the resulting toward-versus-away proportion against a null of no directional bias using three complementary inferential layers that explicitly handle the within-dyad autocorrelation and between-dyad individual reuse: per-dyad exact binomial confidence intervals, an across-dyad Wilcoxon signed-rank test, and a generalized linear mixed-effects model on the per-timestep data with a random intercept for dyad.

Results:

The retreat-by-hyena signal is supported across all three layers. None of the three inferential layers supported a uniform population-level lion approach to hyenas, but a directionally consistent pattern indicated an approach bias for male lions and no consistent directional bias for female lions. Intraspecific lion dyads in Etosha resolve mating-pair and satellite-pride coordination at scales as fine as 15 m; hyena dyads in either ecosystem produced no concurrent fixes below 1 km, recovering the exclusive clan-territorial structure of the species.

Discussion:

The relative-motion framework demonstrates that fine-scale telemetry can substitute for behavioral observation in landscapes where direct observation is logistically infeasible. It also reveals that the asymmetric interference-competition signature predicted by guild theory operates through the subordinate species' reactive avoidance at multi-kilometer scales rather than through symmetric two-species coordination.

Introduction

Apex carnivore guilds in African savannas are structured by complex interactions among species that share both prey and space. Interference competition between sympatric carnivores has long been recognized as a central mechanism by which guild members partition shared resources and coexist on the same landscape (Linnell and Strand, 2000; Amarasekare, 2002; Vance, 1984). Classical theory, derived from the compression hypothesis and self-organizing spatial-coexistence models (Schoener, 1974; Hassell et al., 1994), predicts that subordinate guild members partition their use of contested space and time in response to the proximity, activity, and signaling of dominant competitors and that this partitioning can occur at multiple spatiotemporal scales. African lions (Panthera leo) and spotted hyenas (Crocuta crocuta), the two largest predators across most of their range, occupy similar trophic positions and exhibit substantial overlap in diet, habitat use, and activity periods (Hayward and Kerley, 2008; Périquet et al., 2015a; Cozzi et al., 2012). Where the two species are sympatric, asymmetric interference competition dominates their interactions, with lions typically displacing hyenas at carcasses, killing hyenas opportunistically, and limiting hyena reproductive output through cub mortality (Trinkel and Kastberger, 2005; Périquet et al., 2015a). In the African large-carnivore guild, lion–hyena interference has been documented through kleptoparasitism (Cooper, 1991), through reproductive-fitness consequences for hyenas (Watts and Holekamp, 2008), and through eavesdropping and risk assessment among guild members (Webster et al., 2010; Lehmann et al., 2017), with the operating mechanism interpreted variously as a “landscape of fear” exerted by lions (Périquet et al., 2015a) and as coexistence under reduced fear in the same study systems (Swanson et al., 2016). Spatiotemporal partitioning at the activity-period and landscape scales has been documented across African carnivore guilds (Hayward and Slotow, 2009; Cozzi et al., 2012), and scavenging–predation linkages are increasingly recognized as a structuring force in these communities (Moleón et al., 2014). The subordinate species, the spotted hyena, must therefore navigate a landscape that includes both the resources it shares with lions and the risks that lions pose. This navigation is expected to leave a measurable signature in fine-scale movement behavior.

A substantial body of work has documented how competitively subordinate large carnivores respond to dominant ones. Using direct observation, telemetry, audio-playback and olfactory-manipulation experiments, and high-frequency accelerometry, these studies have repeatedly demonstrated strong responses across a wide range of spatial and temporal scales, and have shown that subordinate carnivores respond differently to cues of immediate, short-term risk than to cues of longer-term variation in risk, a distinction often framed as the difference between avoiding risky places and avoiding risky times (Durant, 2000; Creel and Creel, 2002; Webster et al., 2012; Haring et al., 2023; Creel et al., 2025; Goodheart et al., 2022, 2024). Against this background, what has remained comparatively difficult to characterize is the immediate, moment-by-moment movement geometry of a subordinate-dominant encounter: which individual moves toward which, and at what distance, when the two are simultaneously in proximity. Home-range overlap statistics, occupancy modeling, and coarse-resolution telemetry are powerful for describing space use and temporal partitioning (Swanson et al., 2014; Vanak et al., 2013; Périquet et al., 2021), but they integrate movement decisions over hours or days and so average over the individual approach-and-retreat decisions that make up an encounter. High-frequency, concurrent relocation data from both members of a pair can resolve that geometry directly. Recent advances in dyadic movement analysis have made it possible to extract directional information from such data, classifying each timepoint by whether an individual is moving toward, away from, or laterally relative to a known conspecific or heterospecific (Luisa Vissat et al., 2021; Long and Nelson, 2013).

We use dyadic movement analysis to test three predictions about fine-scale dyadic movement in sympatric lions and spotted hyenas. First, if hyenas are avoiding lions through reactive movement, then at close distances we expect hyenas to retreat from approaching lions significantly more often than by chance. Second, if lions do not perceive hyenas as a threat, then lions should not show a corresponding retreat from hyenas; instead, lion movement relative to hyenas may be neutral, or even biased toward approach if lions investigate hyena activity as a potential cue for scavenging opportunity. Third, intraspecific dyads should display coordination patterns that reflect each species’ social organization: lion dyads from within the same pride or mating pair should show coordinated joint movement, whereas hyena dyads from within the same clan should show patterns consistent with the more variable cohesion characteristic of spotted hyena clan structure (Smith et al., 2008).

To test these predictions, we apply the relative-motion method of Luisa Vissat et al. (2021) to GPS telemetry data from 17 lions and 14 spotted hyenas tracked across two southern African ecosystems, the Etosha National Park in north-central Namibia and the Chobe National Park with adjacent Linyanti Conservancy in northeastern Botswana. We restrict the dyadic analysis to timestamps where both individuals in a pair have real concurrent GPS fixes at five-minute sampling intervals, a conservative choice that excludes model-imputed positions and bases all interaction findings on actual animal locations. We complement the binomial confidence intervals with a circular-statistics check on directional concentration, addressing potential concerns about the inferential strength of binary toward-versus-away classifications. By documenting which collared individuals were mating pairs, satellite group members, or members of different prides and clans, we are able to distinguish coordinated movement that reflects shared group membership from coordinated movement that reflects something else.

The patterns we describe here speak to the broader question of how subordinate large carnivores persist in landscapes shared with dominant ones, and how movement behavior itself functions as a mechanism of coexistence between species that share too many resources to be ecologically independent. What the relative-motion approach adds to this already substantial multi-scale literature is a direct, per-dyad characterization of the immediate-encounter geometry: the bidirectional, moment-by-moment approach-and-retreat structure of a lion–hyena pairing resolved at the five-minute scale. This is a complement to, not a replacement for, analyses of space use, habitat selection, and temporal partitioning at coarser scales. Whether these fine-scale, immediate-risk responses scale up to the broader patterns of space use and demography that ultimately shape fitness is an open question, and one that fine-scale dyadic data are well positioned to help address.

Methods

Study sites

This research was conducted in two protected-area systems in southern Africa (Figure 1). The first is Etosha National Park (ENP), a 22,270-km² fenced semi-arid savanna reserve in north-central Namibia (Huntley and Walker, 1982), where perennial water is available only at natural artesian springs and pumped boreholes (Auer, 1997). The present study did not encompass the entire park; the ENP study site, 9,000 km² in extent, occupied the central regions of Okaukuejo and Halali, with one collared individual additionally dispersing approximately 120 km outside the southern park boundary (collar successfully retrieved). The area is interspersed with dolomite inselberge and sandy bushveld, comprising mainly grasslands, short-grass plains, shrublands, and dwarf-shrub and mixed tree savannas including mopane (Colophospermum mopane) (le Roux et al., 1988). Seasonal rainfall occurs mainly between November and April, with the highest monthly rainfall in January and February (Engert, 1997).

Figure 1

The second system is the Chobe–Linyanti complex in northeastern Botswana, comprising the unfenced 11,700-km² Chobe National Park (CNP) and the adjacent 1,660-km² community-managed Linyanti Conservancy (LC), both of which form part of the core component of the Kavango-Zambezi (KAZA) Transfrontier Conservation Area. The 7,500-km² Botswana study site occupies the floodplains of the northern Kalahari, bounded by the Chobe and Linyanti rivers, with papyrus swamps, lagoons, and marshes. The Chobe and Linyanti rivers, fed by the Kwando and Zambezi rivers, and the Cubango River that feeds into the Okavango Delta are the main sources of permanent water for this ecosystem (Fox et al., 2017). Vegetation includes riparian and mopane woodlands, Baikiaea plurijuga forest, open grasslands, mixed Vachellia savanna, and sandveld communities of Terminalia sericea and Philenoptera nelsii (Murray-Hudson et al., 2011; Sianga and Fynn, 2017). Seasonal rainfall in the Botswana system occurs mainly between November and March, with a mainly dry season from April to October. Data collection in both ecosystems spanned 2013 to 2018. Further details of the study sites, including vegetation classification and prey assemblages, are provided in Supplementary Materials 1.

Animal capture, collaring, and social context

Seventeen adult lions (13 females and four males) and 14 adult spotted hyenas (10 females and four males) were captured and fitted with GPS-accelerometer collars (IridiumTrackM, Lotek Wireless Inc., Newmarket, Ontario, Canada) following the procedures, governmental authorizations, and animal welfare committees detailed in Supplementary Materials 1. In each ecosystem, extensive pre-collaring observations were made to identify distinct lion prides and spotted hyena clans (approximately 8 months in Etosha and 5 months in Chobe–Linyanti). Individuals were then selected for collaring such that each known pride or clan was represented by a single collared animal. The exception to this design was the Linyanti pride, where low lion density meant that two collared lions (AM-36714 and AF-34308) belonged to the same pride. Pride and clan membership were determined from these direct observations prior to and during collaring. Coordinated-movement associations and pair status were determined post-collaring from the GPS records and continued night-follow observations. In Etosha, three intraspecific lion associations were identified after collaring: two confirmed mating pairs (OK-33863 with OJ-33867 in the Okaukuejo region, and RE-33864 with SU-33868 in the Halali region), and one satellite-pride association in which the lioness NU-33865 operated as a peripheral satellite of the pride that included the lioness MO-33866 (Halali region). All other collared Etosha lions and all collared Etosha spotted hyenas were from distinct prides or clans. In the Chobe-Linyanti system, all collared spotted hyenas were from different clans (the Chobe Eastern, Chobe Western, Chobe Central, Linyanti Eastern, and Linyanti Western clans); collared lions came from the Chobe Eastern, Chobe Western, Chobe Central, and Linyanti Eastern prides, with two lions (AM-36714 and AF-34308) representing the Linyanti Western pride. The complete per-individual social context, including pride or clan membership and pair or group association, is provided in Table 1 with additional information in Supplementary Table 1. This explicit documentation of social context allows readers to distinguish coordinated movement attributable to shared group membership from coordinated movement that arises among individuals from different groups.

GPS data collection

Collars were programmed to record GPS fixes every 30 minutes during nocturnal periods (1800 h to 0600 h, or 1700 h to 0800 h in some deployments) and every five minutes for the two-hour periods following sunset and preceding sunrise (1900 h to 2100 h and 0400 h to 0600 h). All collars additionally recorded diurnal fixes at 1000 h and 1400 h to aid in locating individuals prior to night follows. This sampling regime was chosen to maximize temporal resolution during the dusk and dawn activity peaks of both species while limiting battery consumption. Losses across the tracking period were attributable to collar failure, mortality, and depleted-battery collars that could not be physically recovered owing to field logistical constraints, including periods without veterinary support and limited access for collar recovery, rather than animal dispersal. Per-individual tracking summaries are provided in Supplementary Table 1.

Behavioral observations and night follows

Satellite-uploaded locations of all collared lions and spotted hyenas were monitored on a daily basis to ensure that individuals were alive and active. Throughout the tracking period, night-follow observations were made on collared individuals in both study areas, both for ground-truthing of the accelerometer activity data used in companion analyses on inter- and intraspecific movement and activity patterns, recursion and duration within home ranges, and the influence of landscape features, and biological and environmental drivers on those patterns (Barker et al., 2023), and to record any predation events or competitive interactions that occurred during direct observation. Each night, focal individuals were selected from among those whose satellite-uploaded locations earlier that day were accessible by the road network and were showing active movement. Focal animals were located in the field by radio telemetry, using a handheld 3-element directional Yagi antenna (Lotek) and a custom-built vehicular-mounted 5-element antenna paired with an SRX 600 telemetry receiver. Individuals were then approached by vehicle to a distance permitting visual observation under ambient light, supplemented when necessary by red-filtered spotlights to minimize disturbance. During each session, the activity state of the focal animal (e.g., resting, moving, feeding, vigilant, social behaviors), the identity and number of conspecific group members present, the presence of any heterospecific individuals nearby, and discrete behavioral events of interest including mating-pair associations, carcass possession or sharing, and direct interspecific encounters were recorded. These night-follow observations were used to establish the social grouping of each collared individual during tracking, including identification of the mating pairs and satellite-pride associations among Etosha lions described above, and provide the context for behavioral observations reported in the Results.

Dyad construction and the real-position-only criterion

Pairs of individuals were considered as candidate dyads if their tracking periods overlapped and both collars were operational. For each candidate dyad we identified all timestamps at which both individuals had a real, non-imputed GPS fix at the five-minute sampling rate, and at which a real fix also existed for both individuals at the next five-minute timestep. Only pairs with such synchronized location data entered the relative-motion analysis. We deliberately did not use continuous-time movement model (CTMM) imputed positions for dyad construction, even though such imputation was applied for other analyses of the same data (Barker et al., 2023). This conservative choice substantially reduced the effective sample size relative to the full relocation record, but it ensured that every inferred approach or retreat event corresponded to actual animal positions observed by the collars rather than to model-imputed locations whose error structure could plausibly induce spurious directional patterns. For each dyad we computed inter-individual distance at every concurrent timestep, and classified each individual’s movement relative to its partner across nine cumulative distance thresholds spanning 0 m to 10 km (i.e., upper bounds of 50, 100, 200, 500, 1,000, 2,000, 3,000, 5,000, and 10,000 m).

Relative motion analysis

All analyses were conducted in R 4.1.2 (R Core Team, 2021). For each retained timestep within a dyad, we computed four angular quantities using the geosphere R package (Hijmans, 2022). For each individual, the absolute heading at time t was defined as the angle (clockwise from geographic north) between the vector pointing north and the vector joining the individual’s GPS position at time t and at time t + 1 (Figure 2a). The relative heading for each individual at time t was defined as the angle from that individual’s position at time t to the partner’s position at time t, again measured clockwise from north. We then computed, separately for each individual, the difference between the absolute and relative headings at time t and took its absolute value. This heading-difference angle quantifies the extent to which the individual’s movement at time t was directed toward or away from the location of its partner (Figure 2a).

Figure 2

Luisa Vissat et al. (2021), © 2021 British Ecological Society, with permission.

Following Luisa Vissat et al. (2021), we classified each timestep into one of three categories on the basis of the absolute heading-difference angle: moving toward if the magnitude was less than 67.5° (or, equivalently on the full 0°–360° circle, between 0 and 67.5° or between 292.5° and 360°), moving away if the magnitude was between 112.5° and 247.5°, and orthogonal if the magnitude was between 67.5° and 112.5° or between 247.5° and 292.5°. These thresholds correspond to dividing the circle into eight equal 45° sectors, with the three sectors aligned with the bearing toward the partner (the sector centered on that bearing and its two adjacent sectors) treated as toward, the three sectors aligned with the opposite bearing (centered on the opposite direction with its two adjacent sectors) treated as away, and the two remaining lateral sectors (centered perpendicular to the bearing) treated as orthogonal.

For each dyad we computed the proportion of toward versus away movements, restricted to non-orthogonal timesteps, at each of nine cumulative distance thresholds (50, 100, 200, 500, 1,000, 2,000, 3,000, 5,000, and 10,000 m). At each threshold we calculated the 95% exact binomial confidence interval on this proportion using the binom R package (Dorai-Raj, 2014). The null hypothesis is no directional bias, meaning the focal individual moves toward and away from its partner with equal frequency, giving an expected toward-versus-away proportion of 0.5. A confidence interval that excludes 0.5 indicates significant directional bias toward or away from the partner across all timesteps at distances up to that threshold. We additionally computed the simultaneous-movement classification of Luisa Vissat et al. (2021), which categorizes each dyadic timestep into one of five joint-movement types depending on the directional movement of both individuals at once (Figure 2b).

To address potential concerns about treating directional data as if it were Bernoulli-distributed, we performed a robustness check using Watson’s one-sample test against a uniform circular null (Jammalamadaka and SenGupta, 2001) on the full circular distribution of heading-difference angles, comparing the observed distribution against a uniform null distribution that would arise from random movement. Sensitivity of the results to inclusion of orthogonal timesteps was evaluated by repeating the analysis with and without these timesteps.

Statistical analysis

Two sources of statistical non-independence in these data motivated the layered inferential strategy adopted here. The first is temporal autocorrelation within dyad: consecutive five-minute GPS fixes of the same individual are not independent observations of its behavior, because an animal’s position five minutes beyond a location in time depends on its location at that time. The second is between-dyad non-independence introduced by individual reuse: a single collared lion or spotted hyena participates in multiple dyads (a single hyena may be paired with several different lions and with conspecifics in turn), so summary statistics aggregated across dyads cannot be treated as independent observations either. Standard chi-square and t-tests applied to pooled per-timestep data address neither of these non-independence structures. We therefore report three complementary inferential analyses for the lion–hyena interspecific dyads.

First, we report per-dyad two-sided 95% exact (Clopper–Pearson) binomial confidence intervals (Dorai-Raj, 2014) on the toward-versus-away proportion at each of the nine cumulative distance thresholds. Per-dyad CIs are the conservative within-dyad statement: each CI is computed on the timesteps from one focal individual within one dyad, treating non-orthogonal timesteps as independent Bernoulli trials. A CI that excludes 0.5 indicates significant directional bias within that specific dyad.

Second, we report an across-dyad one-sample Wilcoxon signed-rank test on the per-dyad ratios, computed separately at each cumulative threshold for the hyena and lion focal animals. Treating each dyad’s ratio as one observation against a null of 0.5 sidesteps the within-dyad autocorrelation entirely, because the within-dyad structure has been collapsed to a single summary. A one-sample t-test against 0.5 was computed alongside the Wilcoxon for comparability.

Third, we report a generalized linear mixed-effects model fit using the glmer function from the lme4 package (Bates et al., 2015), with the per-timestep binary outcome (toward = 1, away = 0, orthogonal timesteps excluded) regressed on a fixed intercept plus a random intercept for dyad: towards ~ 1 + (1 | dyad) (see Bates et al., 2015, for an explanation of this notation). The mixed-effects intercept estimate is the population-level log-odds of moving towards versus away from the partner, and an intercept significantly less than zero would indicate population-level retreat bias. The random intercept by dyad explicitly models the between-dyad variance that produced different ratios in different dyads.

An equivalent model was also fit with (1 | individual) and with (1 | individual) + (1 | dyad) as a sensitivity check; the individual random effect collapsed to zero (boundary singular fit) in both, reflecting that with only three unique collared spotted hyenas and a comparable number of unique lions across the interspecific data, the model could not partition variance between within-individual and between-individual sources. A larger collared sample would be required to resolve the individual random effect, and our preferred specification uses dyad alone as the random effect for the interspecific lion–hyena results. A per-unique-individual sensitivity check, in which all timesteps for each focal individual were pooled across the dyads in which that individual participated and a single per-individual ratio was computed, was used as a further check against the across-dyad result being driven by repeated entries of any one individual.

Where the lion focal results were further stratified by sex, we fit towards ~ sex + (1 | dyad) and compared it against the no-sex model towards ~ 1 + (1 | dyad) by likelihood ratio test, with sex coded as a binary fixed effect (female as reference). Female-only and male-only versions of the model were also fit separately to obtain the population-level towards probability for each sex.

Data and code availability

The relative motion analysis was implemented in R using the method of Luisa Vissat et al. (2021) and the code provided therein. The processed dyad-level data and the per-individual social context table are provided in Supplementary Materials 1–3, Supplementary Tables 1, 2. Raw GPS relocations are available from the corresponding author on reasonable request, with restrictions where necessary to protect the known ranges of these species, which remain vulnerable to poaching.

Results

Data collection

Across all deployments, 57% of possible relocations were retrieved, with losses attributable to mortality and to collars whose physical recovery was prevented by field logistical constraints (n = 3 lions and 6 hyenas in total). The retrieved dataset comprised 349,163 GPS relocations from the two study areas, of which 55% were from lions and 45% from spotted hyenas (Supplementary Table 1). Collared lions were sometimes observed together in both Etosha (51% of fixes within ≤50 m to another collared lion) and Chobe-Linyanti (6% of fixes within ≤50 m), while only female collared hyenas occurred together in Chobe-Linyanti (1% and 2% of fixes within ≤50 m for the two female-female pairings, respectively). Of the collared lions observed together within ≤50 m in Etosha, 65% were female-female pairs and 35% were male-female pairs; in Chobe-Linyanti, by contrast, nearly all close-distance lion pairs were male-female, with only 1% female-female. These fix-level co-occurrence values describe the wider sampling regime within which the per-dyad relative-motion analysis below operates.

Dyad sample composition

Where each pair of collared individuals from the same study area had tracking periods that overlapped, that pair was considered a candidate dyad. Across 11 collared lions and 9 collared spotted hyenas in Etosha and 6 collared lions and 5 collared spotted hyenas in Chobe-Linyanti, the upper-bound number of possible within-site dyads was 190 in Etosha and 55 in Chobe-Linyanti; the number realized in practice was smaller because tracking-period overlap and the requirement for both collars to produce a real five-minute fix at the same instant excluded many candidate pairs. Under the real-positions-only criterion, this produced 8 interspecific lion–hyena dyads in Etosha and 2 in Chobe-Linyanti, 8 intraspecific lion–lion dyads in Etosha and 1 in Chobe-Linyanti, and 4 intraspecific hyena–hyena dyads in Etosha and 0 in Chobe-Linyanti, for a total of 23 dyads with usable concurrent fixes (per-dyad concurrent-fix counts in Supplementary Table 2). The number of intraspecific lion-lion dyads in Etosha (8) reflects unanticipated post-collaring associations rather than within-pride sampling: collared individuals were chosen one per identified pride or clan in both ecosystems following extended pre-collaring observation, and the associations that produced close-distance lion-lion dyads in Etosha were identified after collaring as one satellite-pride association (NU-33865 satellite of MO-33866’s pride) and two confirmed mating pairs (OK-33863 with OJ-33867; RE-33864 with SU-33868). In Chobe-Linyanti the only same-pride pairing in the collared sample was AM-36714 and AF-34308, both from the Linyanti Western pride; however, AM-36714’s collar was not retrieved at the end of tracking, so for that individual only the satellite-uploaded subset of relocations was available rather than the full five-minute stored record, limiting the same-pride intraspecific dyad detection in Chobe-Linyanti.

The fix-level co-occurrence patterns reported in the Data collection summary above (51% of Etosha lion fixes within ≤50 m of another collared lion, against 6% in Chobe-Linyanti) provide the population-level context within which this real-positions-only dyad sample sits.

Interspecific dyads: an asymmetric retreat-by-hyena pattern at multiple inferential layers

Per-dyad confidence intervals

Per-dyad exact binomial confidence intervals on the towards-versus-away proportion at the cumulative 10 km threshold reveal an asymmetric pattern between hyenas and lions that is detectable in a minority of dyads when each dyad is inspected individually. Of the 10 interspecific lion–hyena dyads, four hyena focal animals had binomial CIs that significantly excluded 0.5 below at the cumulative 10 km threshold, indicating a retreat bias; no hyena focal animal had a CI that significantly exceeded 0.5; the remaining six contained 0.5 (Table 2). Lion focal animals in the same dyads showed the inverse asymmetry: three had CIs that significantly exceeded 0.5, indicating an approach bias, no lion had a CI that significantly excluded 0.5 below, and the remaining seven contained 0.5 (Table 2). At closer cumulative thresholds the per-dyad signal weakens because fewer concurrent fixes are available within tighter distance bins: at the cumulative 5 km threshold two of 10 hyena focal animals had CIs excluding 0.5 below, and at the cumulative 1 km threshold a single hyena focal animal did so. No lion focal animal had a CI excluding 0.5 in either direction at thresholds at or below 5 km.

The OF-34309/SA-33872 dyad in Etosha (Figure 3a) is the cleanest single example of the asymmetric pattern: at distances up to 10 km the lion had a towards-versus-away ratio of 0.687 (95% CI 0.605 to 0.761), significantly above 0.5, while the hyena had a ratio of 0.299 (95% CI 0.229 to 0.378), significantly below 0.5. By contrast the SU-33868/SA-33872 dyad in Etosha (Figure 3b), which had three times as many concurrent fixes (8,567 versus 1 606), showed neither individual departing significantly from 0.5 at any distance threshold. The full set of per-dyad results is provided in Supplementary Materials 2 and Supplementary Table 2.1. The patterns from the Chobe-Linyanti dyads are consistent with those in Etosha: in the KW-36716/SR-34310 dyad the hyena had a ratio of 0.465 (95% CI 0.442 to 0.489), significantly below 0.5, while in the SW-33950/SR-34310 dyad the lion had a ratio of 0.534 (95% CI 0.515 to 0.552), significantly above 0.5.

Figure 3

Across-dyad Wilcoxon signed-rank tests

Treating each dyad’s focal-individual ratio as one observation and testing against a null of 0.5 across the 10 interspecific dyads produces an across-dyad inferential result that does not depend on within-dyad autocorrelation. For the hyena focal animal the across-dyad ratio was significantly below 0.5 at the cumulative 2 km, 5 km, and 10 km thresholds, with the strongest signal at the cumulative 5 km threshold where all nine dyads with usable data had hyena ratios below 0.5. At ≤2 km, seven of eight dyads with data had hyena ratios below 0.5; at ≤5 km, all nine dyads with data had hyena ratios below 0.5; and at ≤10 km, eight of 10 dyads were below 0.5 (all p values < 0.05; Table 3). For the lion focal animal, no cumulative distance threshold produced a significant departure from 0.5 (all p values > 0.05; Table 3), confirming the heterogeneity observable in the per-dyad CIs.

Mixed-effects logistic regression with random intercept by dyad

Pooled across all per-timestep observations at distances under 10 km, the mixed-effects logistic regression on the binary towards/away outcome with random intercept by dyad estimated a population-level towards probability of 0.461 (95% Wald CI 0.421 to 0.501) for the hyena focal animal in lion–hyena dyads, with the logit intercept −0.158 (SE 0.082, z = −1.92, two-sided p = 0.054; Table 4). The intercept estimate falls just inside conventional significance under the one-sided alternative (retreat bias) and the upper confidence limit just touches 0.5. For the lion focal animal the corresponding estimate was 0.517 (95% Wald CI 0.491 to 0.542), logit intercept 0.067 (SE 0.052, z = 1.29, p = 0.20; Table 4), no significant departure from 0.5. The dyad random-effect standard deviation was 0.228 for hyenas and 0.133 for lions, reflecting modest between-dyad heterogeneity in directional bias.

Per-unique-individual sensitivity check

Because spotted hyena SA-33872 participated in six of the 10 interspecific dyads, the across-dyad Wilcoxon could in principle be biased by the repeated entry of one individual. As a sensitivity check, all timesteps for each unique focal individual were pooled across that individual’s interspecific dyads to produce a single per-individual ratio. With only three unique collared hyenas in the interspecific data (TJ-33870, SA-33872, SR-34310) and eight unique collared lions, the per-individual sample is too small for non-parametric tests to reach conventional significance on their own, but every unique hyena’s pooled ratio at the cumulative 10 km threshold falls below 0.5 (TJ 0.487, SA 0.477, SR 0.483; one-sample t = −6.04, df = 2, p = 0.026), and every unique hyena’s ratio at the cumulative 5 km threshold also falls below 0.5 (TJ 0.483, SA 0.438, SR 0.473), although the per-individual t-test at 5 km is not significant given the smaller per-individual sample sizes (t = −2.60, df = 2, p = 0.12). The across-dyad pattern is therefore not driven by SA-33872 alone.

Lion-focal sex stratification

A field-observation prediction prior to the analysis was that male lions would track and approach hyenas more consistently than female lions, with female lions showing more variable responses contingent on group composition. Four male lions were collared in total in this study (SU-33868, OF-34309, OJ-33867, AM-36714), of which three had collars retrieved with the full five-minute stored record (SU, OF, OJ) and one (AM) had only the satellite-uploaded subset of relocations available because his collar was not retrieved. Among the 10 interspecific lion–hyena dyads with concurrent five-minute fixes, the male-lion sample consequently reduces to two unique individuals contributing three dyads: SU-33868 in two dyads (SU/TJ, SU/SA) and OF-34309 in one dyad (OF/SA), all in Etosha. OJ-33867 is paired in the intraspecific OJ/OK mating-pair dyad and does not appear in any interspecific dyad in the present dataset. AM-36714 produced no concurrent five-minute interspecific dyad with any collared hyena because of his satellite-only subset. Six unique female lions contributed the remaining seven interspecific dyads. Across these three male-lion dyads the mean towards-vs-away ratio at the cumulative 10 km threshold was 0.573, with all three above 0.5 (range 0.515 to 0.687). Across the seven female-lion dyads the mean was 0.486, with five of seven below 0.5 (range 0.430 to 0.541). A mixed-effects model with sex as a fixed effect (towards ~ sex + (1 | dyad)) on the per-timestep lion-focal data at distances under 10 km estimated a female population-level towards probability of 0.501 and a male probability of 0.548, with the sex effect in the expected direction but marginal in significance (χ² = 2.77, df = 1, p > 0.05; Table 4). The pattern is directionally consistent with the field-observation prediction (males tracking, females variable), and the male-lion sample of two unique individuals across three dyads is the principal constraint on statistical resolution; the OF-34309/SA-33872 dyad in particular provides the strongest single example of male-lion approach. The field-observation records of male lions actively following hyenas during night follows corroborate the directional pattern recovered from the GPS data.

Intraspecific lion dyads: close-distance coordination in mating and satellite pairs

Concurrent five-minute fixes at distances below 200 m were available in four of the nine intraspecific lion–lion dyads: the two confirmed mating pairs OK-33863 with OJ-33867 (236 paired fixes below 200 m) and RE-33864 with SU-33868 (1,384 paired fixes below 200 m); the satellite-pride pair NU-33865 with MO-33866 (2,378 paired fixes below 200 m, the satellite female NU being a peripheral member of the larger pride that included MO); and a small number of close-distance fixes between SU-33868 and MO-33866 (26 paired fixes below 200 m; Supplementary Materials 3, Supplementary Table 2.2). The remaining lion–lion dyads, including the single Chobe-Linyanti dyad (KW-36716 with SW-33950, pairing the Chobe Eastern and Chobe Western prides), produced no concurrent five-minute fixes at distances below 200 m and consequently no per-dyad CIs are reported at those distances.

For the satellite-pride pair (NU-33865 and MO-33866), both individuals had binomial CIs significantly above 0.5 at the 0–50 m bin (MO ratio 0.601, 95% CI 0.580 to 0.622; NU ratio 0.557, 95% CI 0.535 to 0.578), indicating symmetric directional co-movement consistent with coordinated joint travel within a pride subgroup. For the two mating-pair dyads at the 50–100 m and 100–200 m bins, the focal-individual CIs showed a sex-asymmetric pattern: the male (OJ-33867 or SU-33868) had a CI significantly above 0.5 in every bin, while the female (OK-33863 or RE-33864) had a CI significantly below 0.5 in the same bin. For the SU-RE mating pair, at 50–100 m the male’s ratio was 0.840 (95% CI 0.782 to 0.887) and the female’s was 0.177 (95% CI 0.127 to 0.237); at 100–200 m the male’s ratio was 0.926 (95% CI 0.875 to 0.961) and the female’s was 0.140 (95% CI 0.091 to 0.203). At distances below 15 m for these mating pairs the pattern reversed (Supplementary Table 2.4), with females moving toward males more often than away, consistent with female presenting at close range and male-led pursuit at slightly greater distances. The single Chobe-Linyanti intraspecific lion–lion dyad (KW-36716 with SW-33950) pairs lions from two different Chobe prides and the absence of a close-distance signal in it is consistent with cross-pride sampling, not with a between-ecosystem difference in pride coordination. Of the collared Chobe-Linyanti lions, AM-36714 and AF-34308 were from the same Linyanti Western pride; however, AM’s collar was not retrieved at the end of tracking, so only the satellite-uploaded subset of AM’s relocations was available, and this subset was insufficient to produce a usable concurrent five-minute dyad with AF. The within-pride coordination that the Etosha data resolved cannot therefore be tested directly within Chobe-Linyanti from the present sample.

Intraspecific hyena dyads: clan separation at short distances, variable at long distances

Hyena–hyena dyads produced a markedly different distance signature from lion–lion dyads. No concurrent five-minute fixes between any pair of collared hyenas in either ecosystem were recorded at inter-individual distances below 1 km; the closest paired-fix observations were in the 1–2 km bin (14 paired fixes across all four hyena–hyena dyads). This contrasts with the lion–lion dyads in Etosha that were observed at distances down to 50 m and below (in the mating pairs and the satellite-pride pair specifically), and confirms that the collared spotted hyenas in our sample maintained substantial separation from collared conspecifics from other clans even at the home-range scale. At distances up to 10 km, the per-dyad binomial CIs showed mixed directional signals: the 5–10 km bin recovered both retreat and approach signals depending on dyad, with no consistent pattern across the four hyena dyads (Supplementary Materials 3, Supplementary Table 2.3). All four hyena–hyena dyads in the analysis were within Etosha, so no within-Botswana hyena dyad results are available. However, during night-follow observations in Chobe-Linyanti, hyenas from different clans were sometimes observed feeding at the same carcass within minutes of each other, a tolerance behavior not observed during night follows in Etosha, where rare close-range encounters between hyenas from different clans terminated in mutual vigilance and retreat.

Additional robustness checks

Two additional robustness checks on the per-dyad CI inferences complement the per-unique-individual sensitivity check described above. First, Watson’s one-sample U² test against a uniform circular null was applied to the full heading-difference distribution of each individual at each cumulative distance threshold (a total of 208 individual-threshold combinations). At the cumulative 10 km threshold, 31 of 46 individual-dyad combinations (67%) showed circular distributions departing significantly from uniform at the p < 0.05 level, with 24 of those at p < 0.01. By pairing type at the 10 km threshold, 11 of 20 interspecific individual-dyad combinations (55%), 13 of 18 intraspecific lion combinations (72%), and 7 of 8 intraspecific hyena combinations (88%) departed significantly from uniform. The Watson result confirms non-random movement in the dyads where the binomial CI itself was significant.

Second, the sensitivity of the per-dyad ratios to inclusion of orthogonal timesteps was evaluated by recomputing the ratio with orthogonal timesteps included in the denominator. Including orthogonals systematically lowered the ratio (mean reduction of 0.12 across the 46 individual-dyad combinations at the 10 km threshold) because orthogonal timesteps add to the denominator without contributing to towards counts. The direction of any departure from 0.5 (above versus below) was preserved in 26 of 46 cases; the remaining 20 cases all flipped from above 0.5 to below 0.5, reflecting the structural downward bias that arises from including orthogonals in the denominator. We therefore retain the Luisa Vissat et al. (2021) convention of restricting the towards-vs-away ratio to non-orthogonal timesteps as the primary analysis, treating the orthogonals-included re-coding as a downward-biased reference rather than an equivalent recoding.

Discussion

Asymmetric interference competition operates through reactive retreat by the subordinate species

Our analysis of fine-scale dyadic movement in lion–hyena pairs provides population-level evidence consistent with the long-postulated mechanism by which spotted hyenas mediate the risk of interference competition with African lions. The retreat-by-hyena signal is supported across three complementary inferential layers. At the per-dyad layer, four of ten hyena focal animals showed significant directional bias toward retreat (that is, towards-versus-away ratios significantly below the no-bias value of 0.5, the value at which an individual moves toward and away from its partner with equal frequency). At the cumulative 10 km threshold, no hyena focal animal showed significant approach bias, and the asymmetric pattern was particularly visible in the OF-34309/SA-33872 dyad, in which the lion showed significant approach bias and the hyena showed significant retreat bias. At the across-dyad layer, treating each dyad’s hyena ratio as one observation against a null of no directional bias, the Wilcoxon signed-rank test was significant at the cumulative 2 km, 5 km, and 10 km thresholds, with the strongest signal at the 5 km cumulative threshold where all nine dyads with usable data showed retreat bias. At the population-modeling layer, the mixed-effects logistic regression with random intercept by dyad recovered a marginal retreat bias at the population level. The mixed-effects estimate is more conservative than the across-dyad Wilcoxon at ≤5 km because it uses the full distance range under 10 km (where between-dyad heterogeneity is greater) and weights dyads by fix count rather than equally. The two analyses converge in direction (population-level hyena retreat; no population-level lion approach) and the across-dyad evidence is most consistent at the 2–5 km range where the per-dyad data are densest and least diluted by dyads with very few close-distance observations. The three layers converge in direction and together support a population-level retreat bias in spotted hyenas relative to lions, which the broader literature on reactive avoidance among African carnivores (Broekhuis et al., 2013; Vanak et al., 2013; Swanson et al., 2014) and the more specific reports on Etosha lion–hyena interactions (Trinkel and Kastberger, 2005) predict but had not previously been demonstrated with per-dyad directional resolution at the five-minute timescale. Coarse-scale habitat selection analyses have reported that spotted hyenas do not systematically avoid lions at the landscape level (Périquet et al., 2025); the fine-scale dyadic movement analysis we report here reveals the behavioral avoidance mechanism that operates at the within-encounter scale and would not be detected by habitat selection methods. The behavioral plasticity of spotted hyenas in the face of lion proximity has been further documented through foraging-strategy switches in response to intraguild interactions (Périquet et al., 2015b) and through interference-mediated den-site selection (Périquet et al., 2016), both of which are consistent with reactive retreat operating at the dyadic scale we resolve here.

The lion-focal signal is more heterogeneous and warrants careful interpretation. None of the three inferential layers supported a uniform population-level lion approach to hyenas: the across-dyad Wilcoxon did not reach significance at any threshold, and the mixed-effects model placed the lion towards probability consistent with no directional bias. What the data do show is dyad-level variability, with three specific lions (OF-34309 in OF/SA, RE-33864 in RE/SA, and SW-33950 in SW/SR) approaching their hyena partners significantly more often than chance, and the remaining seven lions showing no significant directional bias. Individual-level heterogeneity in carnivore responses to dominant competitors has been documented previously for spotted hyenas confronting lions (Watts et al., 2010), and the dyad-level variability we observe in the lion-focal signal is consistent with that broader pattern of within-species behavioral variation. The strength of lion-driven effects on co-occurring carnivores has been shown to vary across species pairs and with resource availability and lion abundance (Everatt et al., 2026), consistent with the dyad-level heterogeneity we observe in lion-focal directional bias.

Stratifying lion focal animals by sex recovered a directionally consistent pattern with field-observation expectations, with male-lion dyads showing approach bias and female-lion dyads showing no consistent directional bias (Table 4). The limiting factor is the small number of unique collared male lions that produced interspecific dyads (two unique males across three dyads). The pattern is consistent with first-author night-follow observations across both ecosystems, in which male lions were repeatedly observed actively tracking hyenas while female lions showed more variable responses contingent on whether other pride members or males were present. This sex difference is consistent with the established asymmetry between male and female lions in their interactions with hyenas: adult male lions are the more dangerous and dominant competitors, accounting for most lion-caused hyena mortality and preventing kleptoparasitism at kills in Etosha (Trinkel and Kastberger, 2005), and spotted hyenas accordingly treat groups containing male lions as more dangerous, mobbing female-only groups far more readily (Lehmann et al., 2017). It is also consistent with the intraspecific pattern in which male lions more often close distance while females maintain separation (Chakrabarti et al., 2021). The convergence between the GPS-derived directional pattern and the direct night-follow observations strengthens the inference that the lion approach signal is real for collared males specifically, even though the small collared-male sample precludes a definitive statistical test.

The lion signal. where present. is informative in what it does not show. If hyena activity were aversive to lions, we would expect lion movement relative to hyenas to be biased toward retreat in at least some dyads. We see none. Instead, in three of the 10 dyads (and in all three male-lion dyads). lions move toward hyenas more often than chance, consistent with lions using hyena activity as an investigative cue, plausibly for scavenging opportunity or assessment of clan territory boundaries (Périquet et al., 2015a; Honer et al., 2002). Dynamic-interaction analyses of African apex predators have similarly recovered contrasting seasonal attraction patterns rather than uniform avoidance (Périquet et al., 2021), reinforcing the inference that lion–hyena spatial relations cannot be reduced to a single-direction avoidance signal across all individuals and contexts.

Lion intraspecific dyads reveal sexual-pair and satellite-pride coordination

The three Etosha lion dyads in which both individuals shared a sexual or social pair bond (the two mating pairs OK-33863 with OJ-33867 and RE-33864 with SU-33868, and the satellite-pride pairing NU-33865 satellite of MO-33866’s pride) all showed directional co-movement at distances below 100 m, consistent with the fission–fusion sociality of African lions (Benhamou et al., 2014; Mosser and Packer, 2009). African lion sociality is grounded in cooperative pride defense and group-living benefits documented across multiple ecosystems (Packer and Pusey, 1997; Bailey et al., 2013), and within-pride coalitions function in both intergroup contest (McComb et al., 1994) and infanticide defense by females (Grinnell and McComb, 1996). The mating-pair signature at fine distance scales is particularly informative: at 50–200 m the female moved away from the male more often than toward him while the male moved consistently toward the female, and at distances below 15 m (Supplementary Table 2.4) the pattern reversed with the female moving toward the male. This bidirectional asymmetry recovers the expected geometry of male-led pursuit at intermediate range and female presenting at close range observed directly in carnivore mating sequences (Schaller, 1972), and demonstrates that the relative-motion method can distinguish behavioral states at scales fine enough to recognize mating context from movement geometry alone.

The Etosha-versus-Chobe-Linyanti contrast in the intraspecific lion analysis reflects the sampling design rather than a definitive between-ecosystem difference in pride coordination. In Etosha three special associations (two mating pairs, one satellite-pride pairing) were identified post-collaring from the GPS records and continued night-follow observations, and these produced the close-distance lion–lion fixes that resolved the within-pride and within-pair pattern. In Chobe-Linyanti, AM-36714 and AF-34308 were from the same Linyanti Western pride, but AM’s collar was not retrieved at the end of tracking and only the satellite-uploaded subset of AM’s relocations was available, which was insufficient to produce a usable concurrent five-minute dyad with AF. The within-pride coordination that the Etosha data resolved cannot therefore be tested directly within Chobe-Linyanti from the present sample, and the absence of a close-distance signal in the single Chobe-Linyanti intraspecific lion–lion dyad (KW-36716 with SW-33950, which pairs the Chobe Eastern and Chobe Western prides) reflects cross-pride sampling rather than a real absence of within-pride coordination in Chobe-Linyanti lions. Fission-fusion social structure has been documented in lions across multiple ecosystems including both arid savannas and riparian systems (Schaller, 1972; Mosser and Packer, 2009), and our prior observations indicate that fission-fusion dynamics characterize lions in both study areas; the asymmetric resolution of these dynamics in our data is a sampling rather than a biological feature. Reintroduced-lion studies that quantified resource selection alongside intergroup interactions have similarly emphasized the dependence of within-pride coordination on the spatial context of neighboring prides (Yiu et al., 2019), and further sampling that includes within-pride collared pairs in Chobe-Linyanti would clarify how closely the close-distance coordination signature there resembles that recovered for Etosha pride subgroups.

Hyena intraspecific dyads reflect inter-clan separation, with night-follow evidence of ecosystem contrast

Hyena intraspecific dyads differed from lion intraspecific dyads in one diagnostic respect: no concurrent five-minute fix between any pair of collared hyenas in either ecosystem was recorded at inter-individual distances below 1 km, in contrast to the lion–lion dyads in Etosha that were observed at distances down to 15 m in the mating and satellite pairs. The absence of close-distance hyena-hyena fixes confirms that the clan-territorial structure of spotted hyenas, with largely exclusive territories between neighboring clans (Boydston et al., 2001; Smith et al., 2008), is recoverable from fine-scale telemetry alone in the absence of direct social observation, given that the relative-motion method is applied. The clan-level organization of spotted hyenas in southern African dry savannas has been described from direct observation in Chobe-Savuti (Cooper, 1989) and from long-term studies in productive ecosystems where clans commute over substantial distances to track migratory prey (Hofer and East, 1993). Inter-group competition is itself a regulator of stable group sizes in social vertebrates (Zemel and Lubin, 1995), and the inter-clan separation we recover in the dyadic data is consistent with this broader pattern. The contrast between the two ecosystems in inter-clan tolerance is documented in our complementary night-follow records rather than in the per-dyad CIs: during night follows in Chobe-Linyanti, hyenas from different clans were sometimes observed feeding at the same carcass within minutes of each other, while in Etosha rare close-range encounters between hyenas from different clans terminated in mutual vigilance and retreat. This contrast tracks the known difference in hyena clan territorial structure between dry savanna systems with exclusive territories (Etosha; Trinkel and Kastberger, 2005) and more productive riparian systems where overlapping foraging ranges and mosaic territorial structure are more common (Chobe-Linyanti, similar to riparian systems documented elsewhere in southern African hyenas; Boydston et al., 2001).

Methodological strengths and limitations

The relative-motion analysis we apply here sits within a broader landscape of methods for quantifying dyadic interaction from animal tracking data. Foundational frameworks for testing spatial and temporal interaction between paired animals were developed for radio-tracking data (Minta, 1992; Kenward et al., 1993) and have been extended to GPS-era datasets through dynamic-interaction indices and movement-similarity measures (Long and Nelson, 2013; Long et al., 2014, 2015; Dodge et al., 2009), entropy-based time-series measures adapted from physiology (Richman and Moorman, 2000), path-segmentation approaches that detect changes in movement state (Edelhoff et al., 2016; Gurarie et al., 2016), and dyadic-movement metrics reviews that compare across these traditions (Joo et al., 2018). Compared with these approaches, the Luisa Vissat et al. (2021) relative-motion method we use computes a per-timestep classification of each individual’s movement direction with respect to its partner, which gives direct categorical resolution of approach versus retreat at the time scale of the underlying fix rate rather than a summary index of similarity over a tracking window. This per-timestep categorical resolution is what enables the per-dyad Clopper-Pearson confidence intervals and the across-dyad Wilcoxon framework used here, and is the principal methodological reason the dyadic-scale interference signal we report is recoverable from these data.

The real-positions-only criterion for dyad construction is a deliberate conservatism. By requiring real, concurrent GPS fixes from both individuals at each timestep, we ensured that every inferred approach or retreat reflects actual animal positions rather than model-imputed locations whose error structure could plausibly induce spurious directional patterns. This reduced the effective sample size relative to a continuous-time movement model imputed analysis as in Barker et al. (2023), but for the kind of inference made here, which is about the direction of an animal’s movement relative to a partner whose position must also be known accurately, the trade-off favors real positions. The explicit documentation of pride and clan membership and post-collaring pair association in Table 1 addresses the structural concern that apparent conspecific “interactions” may be artifacts of which individuals happened to be collared in the same group, and allows readers to evaluate the intraspecific coordination findings against the specific social context of each dyad.

The principal limitations are sample size and individual-level non-independence. The total number of dyads across both ecosystems is small for any species-pair combination (10 interspecific, 9 intraspecific lion, 4 intraspecific hyena), and many of the same individuals contribute to multiple dyads. We addressed the non-independence through the layered inferential strategy described in the Methods, with the per-dyad CI, across-dyad Wilcoxon and mixed-effects model converging in direction for the hyena retreat signal. One specific consequence of the small individual-level sample appeared in the mixed-effects fits: when both individual and dyad random effects were included in the model the individual random effect collapsed to zero variance (boundary singular fit), reflecting that with only three unique collared spotted hyenas across the interspecific data (and likewise eight unique collared lions, including two males), the model could not partition variance between within-individual and between-individual sources. A larger collared sample, particularly more unique collared individuals of each species and especially more collared male lions producing concurrent five-minute interspecific dyads, would resolve the individual random effect and would also provide the statistical power needed to confirm the marginal sex difference among lions at the conventional significance threshold. The OF-34309 dyad in particular suggests that male-lion approach to hyenas may be considerably stronger than the pooled three-dyad average implies, and dedicated collaring of additional male lions in either ecosystem would test this directly. Studies of elusive carnivores over large remote landscapes, where direct behavioral observation is logistically infeasible across most of the home-range, are inherently sample-size-constrained, and the value of the relative-motion method demonstrated here is precisely that high-resolution telemetry data can substitute for behavioral observation that cannot be obtained at scale, while the convergence between the GPS-derived sex pattern and direct night-follow observations in this study illustrates how the two evidence streams can reinforce one another.

Implications for coexistence

Lions and spotted hyenas occupy overlapping ecological roles across most of their African range, and the persistence of the two species in the same protected areas depends on behavioral mechanisms that reduce the costs of interference competition. The fine-scale movement data presented here resolve one such mechanism directly: reactive retreat by the subordinate species at the spatial scales at which a direct encounter would impose a cost. This operates alongside, rather than in place of, the spatiotemporal partitioning at the home-range and activity-period scales that other work has documented in this guild (Swanson et al., 2014; Vanak et al., 2013; Périquet et al., 2021); the moment-by-moment movement adjustment of hyenas in response to lion proximity is one of several complementary behavioral routes by which the two species share much of the same landscape. Our data show that this immediate-encounter response is present and directional, but they do not establish that it is the mechanism most consequential for fitness. Whether short-term encounter avoidance scales up to the longer-term patterns of space use and demography that ultimately govern coexistence, or is instead contingent on them, remains an open question, and the finding in other African carnivores that responses to immediate and to longer-term risk need not coincide (Durant, 2000; Goodheart et al., 2022, 2024; Creel et al., 2025) cautions against assuming that any single scale is decisive. For management, the implication is that interventions which compress the spatial scale at which lions and hyenas operate, including water-point concentration during drought, fence construction that constrains dispersal, and carcass aggregation through anthrax control, may reduce the operating distance over which the reactive avoidance mechanism functions. Quantifying the distance-dependence of this mechanism, as we have done here, provides a framework for evaluating those interventions in terms of their effect on guild-level coexistence rather than only on individual species’ viability. Large-carnivore guilds globally are subject to compounding anthropogenic pressures that reshape the ecological context in which interference mechanisms operate (Ripple et al., 2014), and climate-driven shifts in niche overlap among competing carnivores have been projected to intensify these pressures at range margins (Zielinski et al., 2017). Conserving the dyadic-scale behavioral mechanism by which African lions and spotted hyenas presently coexist therefore requires landscape-level management that preserves the spatial resolution at which the mechanism operates.

Conclusion

Asymmetric interference competition between sympatric African lions and spotted hyenas operates through reactive retreat by hyenas at the dyadic scale, sustained across distances of up to ten kilometers and across both savanna and riparian ecosystems. Lions show neither retreat nor consistent avoidance, and tend instead to track hyena activity, consistent with using hyena presence as an investigative cue. Within-species movement coordination is fundamentally different between the two species and is recoverable from telemetry data alone: lions exhibit strong fission–fusion coordination at fine spatial scales, including diagnosable mating-pair and satellite-group signatures in Etosha, while hyenas maintain largely exclusive clan territories with ecosystem-dependent inter-clan tolerance.

The dyadic relative-motion framework demonstrated here can be extended to other apex predator dyads where dual high-frequency telemetry is either currently available or will become available in the near future. More generally, as we enter movement ecology’s “big data era” (Nathan et al., 2022), with order-of-magnitude increases in the sample size and time resolution (i.e., step-size diminution; Getz, 2022) of movement data, we anticipate the methods applied here will become increasingly useful in using movement behavior to uncover behavioral elements that mediate interference competition. In particular, these methods offer a way to test specific predictions about the geometry of interference competition that occupancy-based and home-range-based methods cannot resolve.

References

• 1

  AmarasekareP. (2002). Interference competition and species coexistence. Proc. R. Soc Lond. B.269, 2541–2550. doi: 10.1098/rspb.2002.2181

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  AuerC. (1997). Chemical quality of water at waterholes in the Etosha National Park. Madoqua20, 121–128.

  • Google Scholar
  • View reference in article

• 3

  BaileyI.MyattJ. P.WilsonA. M. (2013). Group hunting within the Carnivora: physiological, cognitive and environmental influences on strategy and cooperation. Behav. Ecol. Sociobiol.67, 1–17. doi: 10.1007/s00265-012-1423-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  BarkerN. A.JoubertF. G.KasaonaM.ShatumbuG.StowbunenkoV.AlexanderK. A.et al. (2023). Coursing hyenas and stalking lions: the potential for inter- and intraspecific interactions. PloS One18, e0265054. doi: 10.1371/journal.pone.0265054

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  BatesD.MächlerM.BolkerB.WalkerS. (2015). Fitting linear mixed-effects models using lme4. J. Stat. Software67, 1–48. doi: 10.18637/jss.v67.i01

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  BenhamouS.ValeixM.Chamaillé-JammesS.MacdonaldD. W.LoveridgeA. J. (2014). Movement-based analysis of interactions in African lions. Anim. Behav.90, 171–180. doi: 10.1016/j.anbehav.2014.01.030

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  BoydstonE. E.MorelliT. L.HolekampK. E. (2001). Sex differences in territorial behavior exhibited by the spotted hyena (Hyaenidae, Crocuta crocuta). Ethology107, 369–385. doi: 10.1046/j.1439-0310.2001.00672.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  BroekhuisF.CozziG.ValeixM.McNuttJ. W.MacdonaldD. W. (2013). Risk avoidance in sympatric large carnivores: reactive or predictive? J. Anim. Ecol.82, 1098–1105. doi: 10.1111/1365-2656.12077

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  ChakrabartiS.BumpJ. K.JhalaY. V.PackerC. (2021). Contrasting levels of social distancing between the sexes in lions. iScience24, 102406. doi: 10.1016/j.isci.2021.102406

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  CooperD. S. M. (1989). Clan sizes of spotted hyaenas in the Savuti region of the Chobe National Park, Botswana. Bots. Notes Rec.21, 121–133.

  • Google Scholar
  • View reference in article

• 11

  CooperS. M. (1991). Optimal hunting group size: the need for lions to defend their kills against loss to spotted hyaenas. Afr. J. Ecol.29, 130–136. doi: 10.1111/j.1365-2028.1991.tb00993.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  CozziG.BroekhuisF.McNuttJ. W.TurnbullL. A.MacdonaldD. W.SchmidB. (2012). Fear of the dark or dinner by moonlight? Reduced temporal partitioning among Africa’s large carnivores. Ecology93, 2590–2599. doi: 10.1890/12-0017.1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  CreelS.CreelN. M. (2002). The African Wild Dog: Behavior, Ecology, and Conservation (Princeton, NJ: Princeton University Press).

  • Google Scholar
  • View reference in article

• 14

  CreelS.RedcliffeJ.GoodheartB.de MerkleJ. R.MwapeH.MatsushimaS.et al. (2025). Prey depletion, interspecific competition, and the energetics of hunting in endangered African wild dogs, Lycaon pictus. Proc. Natl. Acad. Sci. U.S.A.122, e2414772122. doi: 10.1073/pnas.2414772122

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  DodgeS.WeibelR.ForootanE. (2009). Revealing the physics of movement: comparing the similarity of movement characteristics of different types of moving objects. Comput. Environ. Urban Syst.33, 419–434. doi: 10.1016/j.compenvurbsys.2009.07.008

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  Dorai-RajS. (2014). Binom: Binomial Confidence Intervals for Several Parameterizations. Available online at: https://cran.r-project.org/package=binom (Accessed June 3, 2026).

  • Google Scholar
  • View reference in article

• 17

  DurantS. M. (2000). Living with the enemy: avoidance of hyenas and lions by cheetahs in the Serengeti. Behav. Ecol.11, 624–632. doi: 10.1093/beheco/11.6.624

  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  EdelhoffH.SignerJ.BalkenholN. (2016). Path segmentation for beginners: an overview of current methods for detecting changes in animal movement patterns. Mov. Ecol.4, 21. doi: 10.1186/s40462-016-0086-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  EngertS. E. (1997). Spatial variability and temporal periodicity of rainfall in the Etosha National Park and surrounding areas in northern Namibia. Madoqua20, 115–120.

  • Google Scholar
  • View reference in article

• 20

  EverattK. T.AndresenL.MooreJ. F.HinesJ. E.KerleyG. I. H. (2026). The functional effects of African lions on co-occurring carnivores differ across species pairs and with changes in resource availability and lion abundance. Oecologia208, 21. doi: 10.1007/s00442-025-05855-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  FoxJ. T.VandewalleM. E.AlexanderK. A. (2017). Land cover change in northern Botswana: the influence of climate, fire, and elephants on semi-arid savanna landscapes. Land6, 73. doi: 10.3390/land6040073

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  GetzW. M. G. (2022). A hierarchical path-segmentation movement ecology framework. Ecol. Process.11, 56. doi: 10.1186/s13717-022-00399-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  GoodheartB.CreelS.SchuetteP.DrogeE.BeckerJ. A.BandaK.et al. (2024). Spatial risk effects from lions compound impacts of prey depletion on African wild dogs. Ecol. Evol.14, e70401. doi: 10.1002/ece3.70401

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  GoodheartB.CreelS.VinksM. A.BandaK.de MerkleJ. R.KuslerA.et al. (2022). African wild dog movements show contrasting responses to long and short term risk of encountering lions: analysis using dynamic Brownian bridge movement models. Mov. Ecol.10, 16. doi: 10.1186/s40462-022-00316-7

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  GrinnellJ.McCombK. (1996). Maternal grouping as a defense against infanticide by males: evidence from field playback experiments on African lions. Behav. Ecol.7, 55–59. doi: 10.1093/beheco/7.1.55

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  GurarieE.BracisC.DelgadoM.MeckleyT. D.KojolaI.WagnerC. M. (2016). What is the animal doing? Tools for exploring behavioural structure in animal movements. J. Anim. Ecol.85, 69–84. doi: 10.1111/1365-2656.12379

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  HaringR. D.BeverleyG.ThompsonP. N.TaylorA.O’DellJ. H. (2023). Evaluation of lion (Panthera leo) scat as a wild dog (Lycaon pictus) deterrent on game farms. Wildl. Res.50, 1021–1030. doi: 10.1071/WR22084

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  HassellM. P.CominsH. N.MayR. M. (1994). Species coexistence and self-organizing spatial dynamics. Nature370, 290–293. doi: 10.1038/370290a0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  HaywardM. W.KerleyG. I. H. (2008). Prey preferences and dietary overlap amongst Africa’s large predators. S. Afr. J. Wildl. Res.38, 93–108. doi: 10.3957/0379-4369-38.2.93

  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  HaywardM. W.SlotowR. (2009). Temporal partitioning of activity in large African carnivores: tests of multiple hypotheses. S. Afr. J. Wildl. Res.39, 109–125. doi: 10.3957/056.039.0207

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  HijmansR. J. (2022). Geosphere: Spherical Trigonometry. Available online at: https://cran.r-project.org/package=geosphere (Accessed June 3, 2026).

  • Google Scholar
  • View reference in article

• 32

  HoferH.EastM. L. (1993). The commuting system of Serengeti spotted hyaenas: how a predator copes with migratory prey. I. Social organization. Anim. Behav.46, 547–557. doi: 10.1006/anbe.1993.1222

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  HonerO. P.WachterB.EastM. L.HoferH. (2002). The response of spotted hyaenas to long-term changes in prey populations: functional response and interspecific kleptoparasitism. J. Anim. Ecol.71, 236–246. doi: 10.1046/j.1365-2656.2002.00596.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  HuntleyB. J.WalkerB. H. (1982). Ecology of Tropical Savannas (Berlin: Springer-Verlag).

  • Google Scholar
  • View reference in article

• 35

  JammalamadakaS. R.SenGuptaA. (2001). Topics in Circular Statistics (Singapore: World Scientific).

  • Google Scholar
  • View reference in article

• 36

  JooR.EtienneM.-P.BezN.MahévasS. (2018). Metrics for describing dyadic movement: a review. Mov. Ecol.6, 26. doi: 10.1186/s40462-018-0144-2

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 37

  KenwardR. E.MarcstromV.KarlbomM. (1993). Post-nestling behaviour in goshawks, Accipiter gentilis: II. Sex differences in sociality and nest-switching. Anim. Behav.46, 371–378. doi: 10.1006/anbe.1993.1199

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 38

  LehmannK. D. S.MontgomeryT. M.MacLachlanS. M.ParkerJ. M.SpagnuoloO. S.VandeWeteringK. J.et al. (2017). Lions, hyenas and mobs (oh my!). Curr. Zool.63, 313–322. doi: 10.1093/cz/zox043

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 39

  le RouxC. J. G.GrunowJ. O.MorrisJ. W.BredenkampG. J.ScheepersJ. C. (1988). A classification of the vegetation of the Etosha National Park. S. Afr. J. Bot.54, 1–10. doi: 10.1016/S0254-6299(16)31355-2

  • CrossRef
  • Google Scholar
  • View reference in article

• 40

  LinnellJ. D. C.StrandO. (2000). Interference interactions, co-existence and conservation of mammalian carnivores. Divers. Distrib.6, 169–176. doi: 10.1046/j.1472-4642.2000.00069.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 41

  LongJ. A.NelsonT. A. (2013). Measuring dynamic interaction in movement data. Trans. GIS.17, 62–77. doi: 10.1111/j.1467-9671.2012.01353.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 42

  LongJ. A.NelsonT. A.WebbS. L.GeeK. L. (2014). A critical examination of indices of dynamic interaction for wildlife telemetry studies. J. Anim. Ecol.83, 1216–1233. doi: 10.1111/1365-2656.12198

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 43

  LongJ. A.WebbS. L.NelsonT. A.GeeK. L. (2015). Mapping areas of spatial-temporal overlap from wildlife tracking data. Mov. Ecol.3, 38. doi: 10.1186/s40462-015-0064-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 44

  Luisa VissatL.BlackburnJ. K.GetzW. M. (2021). A relative-motion method for parsing spatio-temporal behaviour of dyads using GPS relocation data. Methods Ecol. Evol.12, 2256–2269. doi: 10.1111/2041-210X.13700

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 45

  McCombK.PackerC.PuseyA. E. (1994). Roaring and numerical assessment in contests between groups of female lions, Panthera leo. Anim. Behav.47, 379–387. doi: 10.1006/anbe.1994.1052

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 46

  MintaS. C. (1992). Tests of spatial and temporal interaction among animals. Ecol. Appl.2, 178–188. doi: 10.2307/1941774

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 47

  MoleónM.Sánchez-ZapataJ. A.SelvaN.DonázarJ. A.Owen-SmithN. (2014). Inter-specific interactions linking predation and scavenging in terrestrial vertebrate assemblages. Biol. Rev.89, 1042–1054. doi: 10.1111/brv.12097

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 48

  MosserA.PackerC. (2009). Group territoriality and the benefits of sociality in the African lion, Panthera leo. Anim. Behav.78, 359–370. doi: 10.1016/j.anbehav.2009.04.024

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 49

  Murray-HudsonM.CombsF.WolskiP.BrownM. T. (2011). A vegetation-based hierarchical classification for seasonally pulsed floodplains in the Okavango Delta, Botswana. Afr. J. Aquat. Sci.36, 223–234. doi: 10.2989/16085914.2011.636904

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 50

  NathanR.MonkC. T.ArlinghausR.AdamT.AlósJ.AssafM.et al. (2022). Big-data approaches lead to an increased understanding of the ecology of animal movement. Science375, eabg1780. doi: 10.1126/science.abg1780

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 51

  PackerC.PuseyA. E. (1997). Divided we fall: cooperation among lions. Sci. Am.276, 52–59. doi: 10.1038/scientificamerican0597-52

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 52

  PériquetS.CourbinN.BandaE.LoveridgeA. J.MacdonaldD. W.RevillaE.et al. (2025). Don’t mind the cat: potential for intraguild interactions with lions does not affect spotted hyaena habitat selection. J. Zool.326, 1–15. doi: 10.1111/jzo.70004

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 53

  PériquetS.FritzH.RevillaE. (2015a). The Lion King and the Hyaena Queen: large carnivore interactions and coexistence. Biol. Rev.90, 1197–1214. doi: 10.1111/brv.12152

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 54

  PériquetS.FritzH.RevillaE.MacdonaldD. W.LoveridgeA. J.MtareG.et al. (2021). Dynamic interactions between apex predators reveal contrasting seasonal attraction patterns. Oecologia195, 51–63. doi: 10.1007/s00442-020-04802-w

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 55

  PériquetS.MapendereC.RevillaE.BandaJ.MacdonaldD. W.LoveridgeA. J.et al. (2016). A potential role for interference competition with lions in den selection and attendance by spotted hyaenas. Mamm. Biol.81, 227–234. doi: 10.1016/j.mambio.2015.10.005

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 56

  PériquetS.ValeixM.ClaypoleJ.Drouet-HoguetN.SalnickiJ.MudimbaS.et al. (2015b). Spotted hyaenas switch their foraging strategy as a response to changes in intraguild interactions with lions. J. Zool.297, 245–254. doi: 10.1111/jzo.12275

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 57

  R Core Team (2021). R: A Language and Environment for Statistical Computing (Vienna: R Foundation for Statistical Computing). Available online at: https://www.R-project.org/ (Accessed June 3, 2026).

  • Google Scholar
  • View reference in article

• 58

  RichmanJ. S.MoormanJ. R. (2000). Physiological time-series analysis using approximate entropy and sample entropy. Am. J. Physiol. Heart Circ. Physiol.278, H2039–H2049. doi: 10.1152/ajpheart.2000.278.6.H2039

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 59

  RippleW. J.EstesJ. A.BeschtaR. L.WilmersC. C.RitchieE. G.HebblewhiteM.et al. (2014). Status and ecological effects of the world’s largest carnivores. Science343, 1241484. doi: 10.1126/science.1241484

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 60

  SchallerG. B. (1972). The Serengeti Lion: A Study of Predator-Prey Relations (Chicago: University of Chicago Press).

  • Google Scholar

• 61

  SchoenerT. W. (1974). The compression hypothesis and temporal resource partitioning. Proc. Natl. Acad. Sci. U.S.A.71, 4169–4172. doi: 10.1073/pnas.71.10.4169

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 62

  SiangaK.FynnR. (2017). The vegetation and wildlife habitats of the Savuti-Mababe-Linyanti ecosystem, northern Botswana. Koedoe59, a1406. doi: 10.4102/koedoe.v59i2.1406

  • CrossRef
  • Google Scholar

• 63

  SmithJ. E.KolowskiJ. M.GrahamK. E.DawesS. E.HolekampK. E. (2008). Social and ecological determinants of fission-fusion dynamics in the spotted hyaena. Anim. Behav.76, 619–636. doi: 10.1016/j.anbehav.2008.05.001

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 64

  SwansonA.ArnoldT.KosmalaM.ForesterJ.PackerC. (2016). In the absence of a “landscape of fear”: how lions, hyenas, and cheetahs coexist. Ecol. Evol.6, 8534–8545. doi: 10.1002/ece3.2569

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 65

  SwansonA.CaroT.Davies-MostertH.MillsM. G. L.MacdonaldD. W.BornerM.et al. (2014). Cheetahs and wild dogs show contrasting patterns of suppression by lions. J. Anim. Ecol.83, 1418–1427. doi: 10.1111/1365-2656.12231

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 66

  TrinkelM.KastbergerG. (2005). Competitive interactions between spotted hyenas and lions in the Etosha National Park, Namibia. Afr. J. Ecol.43, 220–224. doi: 10.1111/j.1365-2028.2005.00574.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 67

  VanakA. T.FortinD.ThakerM.OgdenM.OwenC.GreatwoodS.et al. (2013). Moving to stay in place: behavioral mechanisms for coexistence of African large carnivores. Ecology94, 2619–2631. doi: 10.1890/13-0217.1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 68

  VanceR. R. (1984). Interference competition and the coexistence of two competitors on a single limiting resource. Ecology65, 1349–1357. doi: 10.2307/1939115

  • CrossRef
  • Google Scholar

• 69

  WattsH. E.BlankenshipL. M.DawesS. E.HolekampK. E. (2010). Responses of spotted hyenas to lions reflect individual differences in behavior. Ethology116, 1199–1209. doi: 10.1111/j.1439-0310.2010.01833.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 70

  WattsH. E.HolekampK. E. (2008). Interspecific competition influences reproduction in spotted hyenas. J. Zool.276, 402–410. doi: 10.1111/j.1469-7998.2008.00506.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 71

  WebsterH.McNuttJ. W.McCombK. (2010). Eavesdropping and risk assessment between lions, spotted hyenas and African wild dogs. Ethology116, 233–239. doi: 10.1111/j.1439-0310.2009.01729.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 72

  WebsterH.McNuttJ. W.McCombK. (2012). African wild dogs as a fugitive species: playback experiments investigate how wild dogs respond to their major competitors. Ethology118, 147–156. doi: 10.1111/j.1439-0310.2011.01992.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 73

  YiuS. W.KarczmarskiL.ParriniF.KeithM. (2019). Resource selection in reintroduced lions and the influence of intergroup interactions. J. Zool.307, 111–124. doi: 10.1111/jzo.12620

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 74

  ZemelA.LubinY. (1995). Inter-group competition and stable group sizes. Anim. Behav.50, 485–488. doi: 10.1006/anbe.1995.0263

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 75

  ZielinskiW. J.TuckerJ. M.RennieK. M. (2017). Niche overlap of competing carnivores across climatic gradients and the conservation implications of climate change at geographic range margins. Biol. Conserv.209, 533–545. doi: 10.1016/j.biocon.2017.03.016

  • Pubmed Abstract
  • CrossRef
  • Google Scholar