Module 3: Cooperative & Social Predators

The African lion (Panthera leo) is the only felid to form large stable social groups. A pride consists typically of\(3\text{--}12\) related females, their dependent young, and a coalition of \(1\text{--}6\) unrelated adult males that defend the pride’s territory. Almost all hunting is performed by the females. Understanding the emergence of role-specialised cooperative hunting in lions required a combination of long-term behavioural ecology (the Serengeti Lion Project) and direct spatial observation (Etosha Pan).

Stander 1992: Positional Roles

Philip Stander’s (1992) study in Etosha National Park is the canonical demonstration of positional role specialisation. By tracking hundreds of hunts in open habitat, he showed that particular adult females consistently occupied one of two spatial roles:

• Wings: flanking hunters that circle wide and stampede the prey towards the ambush. Lions playing the wing role show lower stalk success in isolation but disproportionately contribute to pride-level success.
• Centres: ambushers that wait in cover downwind of the flushed prey and make the final capture. Centres are typically the heaviest individuals and dominant within the pride.

In Stander’s data, cooperative hunts with recognised role specialisation succeeded \(\approx 2.8\times\) more often than uncoordinated hunts with the same number of lionesses. Role assignment appeared consistent across hunts, implying individual-level specialisation rather than stochastic role assignment.

Packer 1990: The Pride-size Paradox

Craig Packer’s Serengeti Lion Project (Packer et al. 1990) measured per-capita hunting success across pride sizes and found a striking non-monotonicity. The optimum was \(5\text{--}7\)females for medium-sized prey (zebra, wildebeest); larger prides showed declining per-capita yield. This “pride-size paradox” is a classical illustration of the tension between group-level efficiency and per-capita fitness.

Payoff Matrix: Cooperate vs. Solo

The strategic choice each lioness faces can be cast as a simple two-player game. If two lionesses approach a 250 kg zebra, the payoff matrix in expected edible energy (MJ) is approximately:

Mutual cooperation strictly dominates: there is no defection incentive because zebra prey is large enough that a solo lioness rarely succeeds. The structure fails for small prey such as Thomson’s gazelle, where mutual defection (individual hunting) is the unique Nash equilibrium — consistent with field observations that lions rarely cooperate on small, fast prey.

The African wild dog has the highest reliably measured hunt success rate of any social carnivore: \(\sim 80\%\)of chases end in a kill (Creel & Creel 1995; Fanshawe & FitzGibbon 1993). By contrast, lion prides succeed\(\sim 25\text{--}35\%\) of hunts, cheetahs\(\sim 40\text{--}50\%\), and tigers \(\sim 10\%\).

Sink-and-Sprint Strategy

Wild dogs employ a specific pursuit profile: they trot at\(\sim 4\) m/s for several kilometres in a loose spread formation, close to within \(\sim 200\) m of selected prey (usually an individually-targeted impala or young wildebeest), then sprint coordinately at 15–18 m/s. Metabolic cost of transport in the trot phase is\(C_T \approx 1.5\) J/(kg·m); in the sprint phase approximately \(3.0\) J/(kg·m). A single hunt can cover 5–8 km.

Quorum Sensing via Sneezes

Walker et al. (2017) discovered that wild dogs use audible sneezes as a quorum-voting mechanism to decide whether to begin a hunt. The group commits only once a threshold number of sneezes is reached, and that threshold is lower when the dominant breeding pair sneezes first. The result is a decentralised decision process that implements something close to the statistical optimum: hunts are initiated only when enough pack members are ready to participate.

Food Sharing and Kleptoparasitism

Wild dogs regurgitate for pups and yearlings, an extreme form of paternal and alloparental investment. Kleptoparasitism by lions and hyenas is the leading ecological pressure on wild dog energetics: prey carcasses are consumed quickly (\(<15\) min per impala) and social rank is inverted during feeding bouts so that pups feed first. This departs sharply from lion feeding hierarchy.

The killer whale (Orcinus orca) is the most cosmopolitan mammal on Earth, yet local populations are so specialised in diet and behaviour that they qualify as incipient ecological species (Foote et al. 2016). The Southern-Ocean ecotypes, first formalised by Pitman & Ensor (2003) and expanded by subsequent workers, are:

• Type A (open-water Antarctic): specialises on Antarctic minke whale (\(Balaenoptera\ bonaerensis\)). Largest ecotype, offshore.
• Type B1 pack-ice: hunts Weddell and crabeater seals on ice floes using coordinated “wave washing” (Visser 2007). Medium-sized, extensively studied.
• Type B2 Gerlache: smaller B-type near the Antarctic Peninsula, takes penguins and small prey.
• Type C Ross Sea: smallest orca ecotype on Earth; almost exclusively piscivorous, specialising on Antarctic toothfish (Dissostichus mawsoni).
• Type D sub-Antarctic: morphologically distinctive (tiny eye patch, bulbous melon). Piscivorous; depredates longline fisheries.

Visser 2007: The Rope-and-Pulley Wave

Ingrid Visser documented the Type B1 wave-washing hunt in quantitative detail. A pod of 3–7 orcas swim in rank formation towards an ice floe on which a seal is hauled out. At a distance of approximately 20–30 m, the orcas dive in synchrony with upward-angled tail flukes, generating a travelling wave that crests at the floe. The wave is steep enough to wash the seal into the water. Secondary pod members intercept the seal or drown it by preventing resurfacing.

The hydrodynamics of wave generation require precise spacing and timing: Visser’s measurements showed sub-second synchronisation across pod members. This kind of coordinated collective motion is unparalleled among mammals outside humans and implies a shared internal timing representation, though its neural substrate remains unknown.

Bowhead Hunting in the Arctic

Transient orcas in the Bering, Chukchi, and Beaufort Seas occasionally take bowhead calves, a behaviour documented by indigenous hunters for centuries (George et al. 2017). An attack typically involves 3–6 orcas working to separate a calf from its mother, then drowning the calf by preventing surfacing for 15 min or more. This is a textbook example of the relay attack pattern.

Driver ants (Dorylus spp., Dorylinae) and their Neotropical counterparts the army ants (Eciton) are the best-known examples of group-raiding social predators. Raids are conducted by columns of \(10^4\text{--}10^6\) individual workers, preceded by a scouting line, with major castes acting as defenders. Raid columns progress at 1–4 cm/s and sweep through roughly \(100\) m\(^2\) of forest floor per hour, capturing every invertebrate and small vertebrate they encounter.

The key collective-behaviour mechanism is stigmergic trail laying: workers deposit pheromone along paths that successfully produce prey, amplifying via positive feedback into stable raid columns. Solenopsis fire ants use an analogous mechanism with shorter timescales to overwhelm large prey via chemical recruitment.

Group Hunting in Spiders

Most spiders are solitary, but roughly 25 species in the generaAnelosimus, Stegodyphus, Agelena and Mallos have evolved cooperative sociality. Social spiders construct communal webs and cooperate in subduing prey far larger than any individual could handle. Critically, unlike ants, social spiders are inbred with\(F \sim 0.3\text{--}0.5\): Hamilton’s rule is easily satisfied because \(r\) is large.

Harris’s Hawk: A Rare Raptor

The Harris’s hawk (Parabuteo unicinctus) is the only raptor known to hunt cooperatively in a structured way. Bednarz (1988) documented four distinct hunting “packages” used by Sonoran Desert groups, including surprise pounce, relay attack, and flushing. Groups of 4–6 birds—typically parents and older offspring—capture jackrabbits that would be impossible for a solitary hawk. Cooperative raptor foraging has evolved here because jackrabbits are large, fast and cover-dwelling: prey characteristics that independently favour group hunting across taxonomic classes.

Why should a predator share prey? Three mechanisms, introduced by Hamilton and formalised further by Maynard Smith, Axelrod and others, explain the evolutionary stability of cooperative predation.

Kin Selection and Hamilton’s Rule

Hamilton (1964) showed that a gene for cooperative behaviour spreads whenever:

Lion prides (\(r\approx 0.25\text{--}0.5\)), wild dog packs (\(r\approx 0.3\)), wolf packs (\(r\approx 0.4\)) and driver ant colonies (\(r = 0.75\) for full sisters of a singly-mated queen, owing to haplodiploidy) all satisfy Hamilton’s rule when the collective benefit of hunting together exceeds individual cost by a factor \(> 1/r\).

Axelrod & Hamilton 1981: Reciprocal Cooperation

When relatedness is low but repeated interactions are common, reciprocity can sustain cooperation. Axelrod’s iterated Prisoner’s Dilemma tournaments showed that Tit-for-Tat is evolutionarily stable against invading defector strategies, provided that the discount factor \(w\) is large enough that future encounters outweigh the one-time gain from defection:

Selfish Herd (Hamilton 1971)

Hamilton’s “Geometry for the Selfish Herd” (1971) shows that prey aggregation can emerge even without any cooperative benefit: each individual simply tries to reduce its domain of danger—the Voronoi tile that places it closer to a predator than to other group members. This purely geometric argument predicts that prey herds contract towards the centre under predator pressure, a prediction matched by field observations of sardine balls under tuna attack, red-necked phalarope wheels, and wildebeest river crossings.

For \(N\) independent hunters each with capture probability \(p_1\), the group success probability is \(p(N) = 1 - (1-p_1)^N\). The per-capita gain is:

which always has a unique interior maximum at finite\(N^*\): the optimum pack size under independence. For \(p_1 = 0.3\), \(N^* \approx 2\text{--}3\); for \(p_1 = 0.05\), \(N^* \approx 10\text{--}15\). This explains the observed trend that cooperative species evolve when solo success is low.

Super-Additive Synergy

The independence assumption fails for role-specialised hunters. Define a synergy exponent\(\alpha\) via:

Fanshawe & FitzGibbon (1993) estimated\(\alpha \approx 1.15\) for African wild dogs; MacNulty et al. (2011) found \(\alpha \approx 1\) for wolf packs hunting elk (but \(\alpha > 1\) for wolves hunting bison). Lions show \(\alpha \approx 1.2\text{--}1.3\)on medium-large prey.

Public-Goods Free-riding

The other side of super-additivity is the free-rider problem. Heinsohn & Packer (1995) quantified individual-level contribution to lion pride territorial defence and found a small number of consistent “laggards” who followed the lead lioness rather than initiating. Analogous free-riding has been documented in wild dog hunts, where a fraction of individuals contribute less than expected based on body size alone.

Coordinated predation requires bandwidth-efficient communication. Three signalling modalities are documented:

• Vocal: wolves use chorus howls for rendezvous and low rolling growls during stalking; orca “call types” are matrilineally inherited and function as group membership badges; spotted hyenas “whoop” with distance-dependent frequency modulation to recruit from up to 5 km.
• Postural and visual: lion tail tips are highly conspicuous against savannah grass and function as semaphore signals during stalking. Wolf scapular raise during hunts is a widely-recognised alert posture.
• Chemical: eusocial predators (ants) rely on stigmergic pheromone trails; vertebrate predators emit urine marks at kill sites that can recruit distant pack members.

Information-Theoretic Bounds

The classical Shannon bound on channel capacity constrains how quickly coordinates can be transmitted during a hunt. For lion pride vocalisations, Pfefferle et al. (2016) estimated an effective capacity of \(\sim 2\) bits per call, sufficient to encode four discrete signals (approach, flank left, flank right, freeze). This matches the observed repertoire size: when the coordination problem requires \(> 2\) bits, lions resort to visual cues rather than expanding vocal repertoire.

• Spotted hyena (Crocuta crocuta): matriarchal clans of 10–80 individuals; hunts of buffalo require \(\sim 8\) adults. Holekamp et al. (2007) showed hyena clans exceed chimpanzee social complexity on several axes (rank reversal stability, coalition politics).
• Chimpanzee (Pan troglodytes): Taï forest chimps hunt red colobus with role specialisation (drivers, blockers, chasers, ambushers) and share kills along grooming-based reciprocity networks (Boesch 1994). Gombe chimps show weaker coordination.
• Meerkat (Suricata suricatta): obligate cooperative breeders; subordinates serve as sentinels while dominants forage. Clutton-Brock et al. (1999) showed sentinel duty is a net benefit to the sentinel because the first to spot a predator escapes first.
• Bottlenose dolphin (Tursiops): Shark Bay dolphins use mud-ring feeding, sponge-tool foraging, and cooperative “fish-whacking”. Culturally transmitted; Mann et al. (2013) showed vertical transmission from mother to calf.
• Groupers with moray eels: Red Sea coral trout (Plectropomus pessuliferus) signal to giant moray eels (Gymnothorax javanicus) by shaking the head, recruiting the eel to flush prey from coral crevices (Bshary et al. 2006). First documented interspecific cooperative hunting with referential signalling outside primates.
• Honey badger & honeyguide: the greater honeyguide (Indicator indicator) leads humans—and arguably honey badgers—to bee nests, then shares the exposed comb. A rare documented case of between-class cooperation (bird + mammal).

In eusocial species, individuals permanently forgo reproduction to aid relatives. Army ants, social wasps, honeybees, naked mole-rats, and some termites are the classical examples. Eusociality creates extreme cooperative predation: Eciton burchellii swarm raids deploy workers as dedicated scouts, killers, transporters, and defenders, each caste physically specialised. The division-of-labour efficiency\(\eta_\text{eusoc}\) exceeds anything achievable by facultatively cooperative vertebrates.

Bourke (2011) summarises the transition to eusociality as requiring three conditions: lifetime monogamy (so sisters are on average as related as full-siblings, \(r=0.5\)), an overlap of generations, and parental-care infrastructure that permits alloparental helping. Where these are satisfied, repeated independent origins of eusociality are observed (at least 11 times in insects, twice in crustaceans, and twice in mammals).