publications

Network theory predicts that losing highly connected species should trigger cascades of coextinction, but evidence from nature is scarce. We built an interaction network between 179 dung beetle species and East African mammals, then compared simulated elephant loss with a 15-year exclusion experiment and real-world elephant declines. Elephant dung was the network's most connected resource, and its loss reduced beetle diversity, biomass and dung removal, providing experimental evidence that elephants are true keystone species.

Do plant–pollinator interactions become more specialized toward the tropics? Analyzing more than 3,400 quantitative pollination networks worldwide, we find no simple latitudinal gradient. Network and pollinator specialization peak near the northern tropical–subtropical boundary, plant specialization differs between hemispheres, and climate, especially temperature and precipitation, predicts specialization better than latitude or biodiversity, with strong differences among pollinator groups such as birds and insects.

How does losing large herbivores affect savanna trees? Combining 16 years of growth data with measurements of leaf water potential and gas exchange in a Kenyan exclusion experiment, we show that removing large herbivores increases tree growth but also increases water stress, especially in dry periods. Species with the most acquisitive water use suffered most, suggesting herbivores make savanna trees more resilient to drought.

Can we predict how plant communities respond when large herbivores disappear? Using herbivore diet data, plant functional traits and size-selective exclosures in Kenya, we identified traits that predict which plants each herbivore group selects and tested whether selectivity predicts plant responses to herbivore removal. Plant responses depended not only on being eaten but on traits and plant–plant interactions, refining simple first-principles predictions.

Trees can help understory plants by providing shade that eases water stress, or by shielding them from browsers. Studying the association of the forb Hibiscus meyeri with tree canopies, with and without large browsers across an aridity gradient in Kenya, we found that both mechanisms operate and interact: protection from browsing was amplified in arid sites, where shade does the most to relieve water stress.

Why are widespread species also locally abundant? Combining population genetics and dietary DNA metabarcoding for a community of East African small mammals, we tested dispersal-based and consumer-resource explanations for the abundance-occupancy relationship. Dispersal did not explain the pattern, but common, widespread species had broader diets that better matched available food plants, and individuals whose diets reflected local food availability survived better.

How does losing large herbivores affect small mammals? Using the UHURU exclosure experiment across a rainfall gradient, we found that excluding large herbivores broadened the dietary niche of fringe-tailed gerbils, shifted their use of nitrogen-fixing Acacia, and altered their gut microbiomes, enriching fiber-degrading bacteria at the most productive sites, showing that vegetation changes after defaunation cascade into the diets and microbiomes of smaller animals.

Demographic rates such as survival and reproduction can change in opposite directions along environmental gradients, buffering population growth. Testing for this demographic compensation in the forb Hibiscus meyeri across a rainfall gradient with size-selective herbivore exclosures, we found compensation only when megaherbivores were excluded. Large herbivores therefore alter how plant populations respond to aridity across the landscape.

A broad review of the ecosystem services and functions that insects provide to people, from pollination and pest control to nutrient cycling, decomposition and food. The review synthesizes how insects support human well-being and highlights why insect declines matter for society.

Do shaded savanna saplings invest more below ground? In a screenhouse experiment with saplings of Acacia (Vachellia) drepanolobium, we tested how reduced light affects carbon allocation to roots, shoots, storage and defense. Shade shifted allocation toward belowground growth and storage, suggesting that carbon limitation from competing grasses shapes how young savanna trees prepare for regrowth after fire and browsing.

Ant defenders are costly, so how do they influence a tree's investment in roots? Using minirhizotrons over a dry and a wet year, we measured root surface area of ant-acacias in relation to ant activity, rainfall, herbivore exclusion and invasion by the big-headed ant. Root growth tracked mutualist activity and rainfall in season-dependent ways, and invasion and herbivore loss altered these responses.

Do plant–pollinator interactions become more specialized toward the tropics? Using thousands of quantitative pollination networks from around the world, we show that specialization does not simply increase toward the Equator. Instead, it peaks outside the tropics and is shaped primarily by climate, with distinct patterns for different groups of pollinators.

An invasive ant, the big-headed ant, displaces the native ants that defend Acacia (Vachellia) drepanolobium from browsing elephants. Combining landscape-scale surveys, elephant-exclusion experiments and GPS-collared lions, we show that invaded areas lose tree cover, which removes the concealment lions use to ambush zebras. Lions killed fewer zebras in invaded areas and shifted toward buffalo, showing how disrupting one mutualism can cascade through an entire food web.

Ecosystem restoration often underperforms or fails. We outline how ten foundational ecological theories, from facilitation hierarchies to macroecology, could be integrated into restoration planning, and propose a simple framework for choosing which theories best match a project's goals, aiming to make restoration more efficient, successful and scalable.

How much do grazing herbivores reduce fuel loads and fire? Combining herbivore exclosure experiments across African savannas with remotely sensed fire and herbivore density data, we show that grazers substantially reduce grass biomass and burned area, with effects scaling with grazer density and remarkably consistent across environments. Including grazing could markedly improve fire and carbon-cycle models.

How do herbivores of different sizes shape the structure and diversity of savanna vegetation? Combining high-resolution LiDAR with field surveys in size-selective exclosures in Kenya, we show that large herbivores reduce the structural complexity and alter the diversity of plant communities, with effects that depend on herbivore body size and plant type, and that herbivores reshape the relationship between structural complexity and diversity.

Do generalist hosts carry more variable gut microbiomes? Studying three co-occurring rodent species across an environmental gradient in a Kenyan savanna, we compared the taxonomic, phylogenetic and functional composition of their gut microbiomes over space and time, testing how host species identity, population size and habitat breadth shape microbiome diversity.

A short, accessible primer on acacia ants: the specialized defensive symbionts that live in the swollen-thorn domatia of acacia trees and feed on extrafloral nectar, how these partnerships work in Africa and the Americas, and what they teach us about cooperation, competition and the ecological consequences of mutualism.

Restoration often assumes plants will recover once the cause of degradation is removed. Synthesizing field experiments from ecosystems across the globe, on land and in coastal oceans, we show that herbivores strongly limit the success of vegetation restoration, and that protecting plants from herbivory substantially improves restoration outcomes, arguing for integrating top-down control into restoration practice.

Are there general predictors of soil microbial communities? Across a global network of 18 grasslands, we found that bacterial communities were associated with particular soil properties, and both bacteria and fungi with plant community composition, both across and within sites. A consistent microbial signal distinguished high- from low-productivity soils worldwide.

Biotic interactions should influence where trees send their carbon. Piloting a ¹³C pulse-labeling method on adult Acacia (Vachellia) drepanolobium in the field, we traced newly fixed carbon into tree tissues and into ant partners. Trees exposed to browsing herbivores appeared to allocate more new carbon to their native ant defenders, demonstrating a practical way to measure how mutualisms and herbivory shape the carbon cycle.

Do acacia ants influence where birds build their nests? Surveying tree-nesting birds in a Kenyan savanna, we found that birds preferentially nested in trees occupied by ant species that vigorously defend against browsing mammals, suggesting that birds gain protection from the plant-ant mutualism, and showing how mutualisms can cascade to other members of the community.

Do gut microbiomes mirror host evolutionary history? Analyzing 14 co-occurring small mammal species in a Kenyan savanna, we found strong phylosymbiosis: microbiome dissimilarity increased with host phylogenetic distance. Among closely related rodents, body size, diet breadth and dietary overlap better explained microbiome differences, showing that deep evolutionary divergence and ecological traits act at different scales.

How do invasive ants change what happens below ground? Excavating the coarse roots of Acacia (Vachellia) drepanolobium in invaded and uninvaded areas, we found that trees in areas invaded by the big-headed ant had altered lateral root distributions, with less root biomass near the soil surface, showing that ant invasions can reshape tree root systems.

Can we predict how evolution responds to new species interactions? In a multigeneration experiment on 16 Bahamian islands, we added predators and competitors to brown anole populations and measured locomotor and habitat-use traits and genome-wide allele frequencies before and after roughly five generations. Despite well-understood ecological effects of predators and competitors, evolutionary outcomes were difficult to predict, highlighting the challenge of forecasting evolution in a changing world.

Do wild herbivores help savannas resist plant invasions? Using three long-term exclosure experiments in central Kenya and landscape-scale surveys, we found that invasive Opuntia cacti, among the world's worst invaders, were far less dense wherever wild herbivores had access and increased more rapidly when herbivores were excluded, with elephants playing a leading role. Conserving wild herbivores can provide biotic resistance to invasion.

Do different ant partners leave different signatures on their host's resources? Using a five-year ant removal experiment with Acacia (Vachellia) drepanolobium, we show that the four acacia ants differentially affect tree carbon fixation, sugars and water stress. Protective, non-pruning ants increased carbon fixation, a sterilizing ant reduced water stress, and a poor defender lowered carbon fixation, evidence of partner fidelity feedback in a multispecies mutualism.

What are the demographic consequences when an invader disrupts a mutualism? Experiments in a Kenyan savanna show that invasion by the big-headed ant, which eliminates the native ants that defend Acacia (Vachellia) drepanolobium from browsers, exacerbates tree population declines, mainly by reducing adult survival and reproduction, and that these effects are amplified by browsing large mammals.

How do invasive ants affect tree carbon dynamics? In a screenhouse experiment with Acacia (Vachellia) drepanolobium saplings, soil nesting by the invasive big-headed ant, rather than its tending of sap-feeding insects, disrupted sapling photosynthesis, carbohydrates and growth. Field surveys showed that invasion and browsing together reduce sapling crown size, revealing a below-ground pathway by which invasive ants harm trees.

A data paper releasing seven years of data from the UHURU experiment in Kenya, in which size-selective exclosures remove different groups of large mammalian herbivores across a rainfall gradient. The data include measurements of plants, small mammals and other community and ecosystem responses, supporting research on the consequences of large-herbivore loss.

Large herbivores are rarely considered part of pollination systems. Using 1-hectare exclosures in Kenya that remove large mammalian herbivores, we found that exclusion increased the abundance and diversity of flowers and their insect visitors, producing larger, less specialized and more robust plant-pollinator networks. Species outside a network can strongly reshape its structure, linking the global decline of large mammals to the functioning of pollination.

Lianas are increasing in many ecosystems, but what keeps them in check? Using long-term herbivore exclosures in a Kenyan savanna, we show that large mammalian herbivores strongly suppress liana infestation of trees through browsing, and that liana prevalence increases when herbivores are removed, revealing an overlooked consequence of large-herbivore decline.

Megaherbivores have been proposed for rewilding, but how do they affect smaller herbivores? Using long-term exclusion experiments and multispecies hierarchical models of dung counts, we tested how elephants and giraffes affect the occurrence and use intensity of mesoherbivores, and whether species' traits such as diet and height predict their responses. Megaherbivore effects on smaller herbivores were partly predictable from these traits.

Do invasive ants affect fundamental ecosystem processes? We show that the big-headed ant, which displaces the native ant defenders of Acacia (Vachellia) drepanolobium, sharply reduces whole-tree photosynthesis in long-invaded areas through an interaction between invasion and browsing by vertebrate herbivores that damages leaves and canopies. Invasion of a foundation mutualism can thus reduce carbon capture by an entire savanna.

How do pollutants move through coastal food webs? In a field experiment manipulating saltmarsh invertebrates, we show that suspension-feeding mussels concentrate PCBs, including toxic dioxin-like compounds, in sediments and drive their biomagnification through the food web, revealing that non-trophic interactions can control contaminant bioavailability.

How do supercolonial invasive ants spread? Tracking big-headed ant abundance along transects spanning the invasion front over 20 months, we show that population growth is negatively density dependent and influenced by rainfall and tree density, allowing us to model the rate of spread of an invader that is disrupting a keystone acacia–ant mutualism in East Africa.

How do invaders reassemble native communities? Invasion by the predatory big-headed ant eliminated the three nectar-dependent Crematogaster ants on Acacia (Vachellia) drepanolobium, but strongly increased the competitively subordinate Tetraponera penzigi, which does not depend on nectar. Long-term monitoring and experiments show the invader indirectly facilitates this formerly rare ant, with consequences for the function of the mutualism.

Allometric equations often fail for small, heavily browsed trees. We developed a method that estimates above-ground biomass from standardized photographs of small savanna trees, showing that pixel area in photos predicted the measured biomass of harvested trees better than published allometric relationships, offering a non-destructive tool for tracking herbivore effects on woody plants.

Classic theory holds that top predators promote coexistence among competing prey. In whole-island experiments in the Bahamas, we staged invasions by a competitor (green anoles) and a new top predator (curly-tailed lizards) on islands where brown anoles were the native top predator. Fear of the predator collapsed the spatial and dietary niche differences that let the two anole species coexist, intensifying competition in predator-free refuges and contributing to the extinction of green anoles on two islands.

How do organisms respond in the first moments of colonizing a new environment? Experimentally introducing brown anole lizards to new islands, we measured genome-wide DNA methylation and found rapid epigenetic changes after colonization, suggesting that epigenetic mechanisms may contribute to fast phenotypic responses to novel environments.

Does temperature influence the protection ants provide to plants? Measuring patrolling and aggression of the four ant mutualists of Acacia (Vachellia) drepanolobium from dawn to dusk, we found that aggression toward a simulated herbivore increased with branch temperature, especially in the two most aggressive Crematogaster species, suggesting trees may be poorly defended during cool periods when many browsers feed.

Neighbors can either shield a plant from herbivores or draw herbivores to it. Studying a savanna plant in Kenya, we found that these opposing processes, associational resistance and associational susceptibility, both strongly shaped the plant's investment in physical defenses, producing predictable variation in defense phenotype across the landscape.

Do large mammals affect pollinators? In 1-hectare plots manipulating large-herbivore density in a Kenyan savanna, we show that large mammals alter the composition of available flowers, and that these changes cascade to floral visitor communities, an extended trophic cascade that links large herbivores to insects that never interact with them directly.

Can a third partner strengthen a mutualism? The dominant acacia ant Crematogaster mimosae tends scale insects for honeydew in addition to feeding on extrafloral nectar. We show that scale insects increase the ant's colony size and resource base, enhancing the protection it provides to Acacia (Vachellia) drepanolobium, so a seemingly parasitic third partner strengthens a keystone mutualism.

Why do herbivores increase plant diversity in some places and decrease it in others? Analyzing herbivore exclusion experiments from grasslands and savannas worldwide, we show that the key predictor is how herbivores change plant dominance: when herbivores reduce the abundance of dominant species, diversity increases, and when they increase dominance, diversity declines.

An invited reply to comments on our study showing that ticks proliferate when large herbivores are excluded. We evaluate alternative explanations for this pattern, including changes in host availability and microclimate, and discuss what they imply for disease risk as wildlife declines.

A review of lessons for conservation from three long-term large-herbivore exclusion experiments in Kenya, KLEE, UHURU and GLADE. Together they show how large mammals shape vegetation, smaller animals, disease and ecosystem function, how these effects vary with rainfall and herbivore identity, and how wildlife and livestock interact in shared landscapes.

Why do individuals of the same plant species differ so much in defense? We show experimentally that physically defended plants growing in the shelter of spiny Acacia trees, which protect them from browsing, invest less in their own defenses. Associational refuges and induced resistance interact to create predictable variation in defense within species.

An overview of the diverse interactions among plants, microbes and insects, from antagonism to mutualism, and how these interactions reconfigure plant physiology to favor beneficial partners and deter pathogens and herbivores. The review highlights open questions about the molecular dialogue among partners and how environmental change will reshape these interactions.

Darwin proposed that species interactions limit populations most in benign environments, and abiotic stress in harsh ones. Experimentally manipulating competition, herbivory and pollination for the forb Hibiscus meyeri across an aridity gradient in Kenya, we found that these interactions affected population growth more strongly in wetter areas. Surprisingly, this was not because interacting species were more abundant there, but because aridity made population growth less sensitive to them.

Do neighboring plants alter the dynamics of an ant-plant mutualism? By factorially manipulating grasses and ant occupants on Acacia (Vachellia) drepanolobium saplings, we tested whether competition from grasses limits a tree's ability to support costly, nectar-dependent ant partners. Contrary to prediction, the high-cost, high-reward Crematogaster mimosae survived better on saplings with grass neighbors, showing that plant–plant interactions can shape mutualism in unexpected ways.

A book chapter synthesizing decades of work on ant-acacia interactions in East African savannas. We argue that the simple '+/+' view of mutualism conceals great complexity: costs and benefits to partners shift across space, time, environmental conditions, the guild of competing ant partners, and the wider community of herbivores and other species.

Do wildlife loss and climate interact to affect disease risk? Measuring ticks and tick-borne pathogens within size-selective large-herbivore exclosures across a rainfall gradient in Kenya, we found that excluding wildlife sharply increased tick abundance, with the strongest effects at the driest, least productive sites, demonstrating interactive effects of defaunation and aridity on disease risk.

Are plant-fungal symbioses locally adapted? Termite mounds create small-scale contrasts in water availability. In a greenhouse experiment, the native grass Themeda triandra inoculated with arbuscular mycorrhizal fungi from termite mounds tolerated drought better than plants with fungi from off-mound soils, contrary to expectation, revealing habitat-specific benefits of mycorrhizal partners.

How do large herbivores influence the behavior of small mammals? Using four years of trapping and vegetation data from the UHURU exclosure experiment, we show that where large herbivores are present, small mammals strongly select tree cover and avoid bare ground, especially in the dry season, but these preferences weaken when large herbivores are excluded, revealing climate-dependent indirect effects on behavior.

How does colony size influence aggression and competition among social insects? By experimentally increasing and decreasing worker density in colonies of the acacia ant Crematogaster mimosae, we show that denser colonies were more aggressive, more successful at expanding onto neighboring trees, including trees held by a rival species, and more likely to survive battles.

How do land use and climate jointly shape rodent-borne disease? Surveying rodents and their pathogens across landscapes in central Kenya that differ in land use and aridity, we show that these factors interact to influence rodent abundance and pathogen prevalence, with implications for zoonotic disease risk as land use intensifies.

An opinion article arguing that restoration should account for interactions among species. Many species have 'backup plans', traits that allow them to persist when a partner is lost, such as switching to new partners. Recognizing which interactions are resilient and which are fragile can guide restoration priorities.

Could a fungal biocontrol agent for termites harm other insects? We tested the entomopathogenic fungus Metarhizium anisopliae on ants associated with Odontotermes termite mounds in laboratory bioassays and in an 18-month field trial, and found no negative effects on ant survival or diversity, supporting its use as a targeted alternative to chemical insecticides.

Elephants topple trees, but how do they affect small plants in the understory? Using seven years of data on more than 100 understory taxa in 36 experimental plots, we found that elephants had neutral effects on most species, with positive and negative responses about equally common. Positive effects arose from ecosystem engineering, as elephant damage to trees created refuges that offset direct consumption.

Do invasive ants disrupt ecosystem functions as well as diversity? Comparing sites invaded and not invaded by the big-headed ant in a Kenyan savanna, we found lower arthropod diversity in invaded sites, but no change in dung-decomposing insects or in rates of dung decomposition. Invaded savannas can lose diversity while retaining some key functions, so diversity alone is an incomplete measure of invasion impact.

Why do savanna trees segregate so sharply across soil types? In Kenya, the ant-defended Acacia (Vachellia) drepanolobium dominates clay soils while poorly defended species dominate sandy soils. Transplant and exclusion experiments show that without large herbivores both species establish on both soils; browsers maintain the pattern, so herbivory, not physiology alone, drives habitat specialization and beta diversity.

A book chapter on how mutualisms operate within broader ecological communities. Using the acacia-ant system and other examples, we show how herbivores, competitors, predators and environmental context shape the costs, benefits and stability of mutualisms, and how mutualisms in turn structure communities.

Is Earth entering a sixth mass extinction? Using deliberately conservative assumptions, we compared modern vertebrate extinction rates with background rates from the fossil record. Even under these assumptions, recent extinction rates are far higher than background, indicating that a sixth mass extinction is under way and that urgent conservation action is needed.

How does plant species richness relate to productivity? Using standardized sampling across grasslands on six continents, we found a consistent hump-shaped relationship, with richness highest at intermediate productivity, resolving a long-standing debate and supporting the idea that both stress and competition limit diversity.

How do multiple disturbances interact? Factorially manipulating fire and elephants in a Kenyan savanna, we show that fire and elephants together synergistically increased the density of an arboreal gecko, because fire made trees more susceptible to elephant damage, creating better gecko habitat. Fire also restructured the acacia-ant community, and these synergies weakened over time.

Do recovering large carnivores restore trophic cascades? After African wild dogs recolonized central Kenya, their main prey, dik-dik, declined, and dik-dik browsing suppressed trees. Yet comparing herbivore exclusions set up before and after wild dog recovery, we found no trophic cascade to trees, cautioning that carnivore restoration may not automatically restore ecosystem functions.

We asked whether the invasive big-headed ant (Pheidole megacephala) disrupts the defensive acacia-ant mutualism in Laikipia, Kenya. Across five properties, invasion virtually eliminated the three aggressive Crematogaster species, while the non-aggressive Tetraponera penzigi persisted by retreating into its domatia. Losing their defenders left trees far more vulnerable to catastrophic elephant damage, suggesting the invasion can reduce tree cover across whistling-thorn savannas.

How do different forms of large-wildlife loss affect small mammals? Comparing wildlife removal by fencing, replacement by livestock, and conversion to agriculture across landscapes in central Kenya, we found that wildlife loss generally increased small-mammal density, reduced species richness and shifted community composition, with effects strongly dependent on rainfall and type of land use.

Can predators shape the plant community through fear? We show that impala in Kenya avoid dense woody cover where they are most vulnerable to leopards and wild dogs, concentrating browsing in open areas. This pattern favors heavily thorned Acacia in open habitats and allows less-thorny species to thrive in denser bush, so large carnivores ultimately make the tree community less thorny.

Why do sterilizing parasites leave hosts partially fertile? We combine theory and experiments showing that Acacia (Vachellia) drepanolobium reproduces more when occupied by small colonies of the sterilizing ant Crematogaster nigriceps. Battles between ant colonies keep colonies small, opening intermittent windows for host reproduction, so ecological dynamics, not just evolution, can explain incomplete sterilization.

Are mammalian browsers interchangeable in controlling an encroaching shrub? Using size-selective exclosures and experiments in Kenya, we show that different browser species affect the native shrub Solanum campylacanthum in distinct, non-substitutable ways, so there is little functional redundancy among browsers and the loss of particular species can promote shrub encroachment.

A data paper releasing the first five years of plant and small-mammal data from the UHURU experiment, in which size-selective exclosures remove different groups of large mammalian herbivores across a rainfall gradient in central Kenya, providing a resource for studying the community consequences of large-herbivore loss.

Why do some plants grow better near neighbors? Testing two mechanisms for a savanna forb in Kenya, we show that neighboring plants help mainly by hiding it from herbivores rather than by ameliorating abiotic stress such as heat or drought, highlighting the importance of herbivory in shaping plant–plant facilitation.

How do nutrient conditions affect decomposition in savannas? Decomposing grasses grown under factorial nitrogen and phosphorus additions in a Kenyan savanna, we found that decomposition and nutrient release were slow in the dry season and rapid in the wet season, and were influenced by the nutrient conditions under which the grasses grew.

How can several competitors coexist on a single limiting resource? In the four-species guild of acacia ants, we show that the dominant ant outcompetes two intermediate species while tolerating the most subordinate one, creating openings for it to establish. Trees held by the subordinate species are later taken over by intermediates and then by the dominant, a cyclical succession driven by indirect facilitation that can promote coexistence.

What happens to tropical forests emptied of wildlife? Comparing Central African forest sites affected by logging and hunting, logging alone, or neither, we found that both reduced large mammals and shifted animal communities toward squirrels and small birds. Experiments show this defaunation disrupts seed dispersal and predation, with potential consequences for tree recruitment and forest carbon storage.

Do herbivores matter more in stressful or benign environments? Using exclosures across an aridity gradient in Kenya, we found that herbivores reduced individual performance of the subshrub Hibiscus meyeri most in arid areas, but had the strongest effects on population size structure and abundance in wetter areas, where plants released from herbivory grow fastest.

Does carbohydrate-rich food fuel ant dominance? Manipulating carbohydrate access for the plant-ant Crematogaster nigriceps over ten months, we found that supplementation quickly increased aggressive defense of the host tree, but did not increase protein foraging or colony growth; instead, carbohydrate-deprived colonies foraged more intensively for protein.

How do the costs and benefits of a mutualism scale with partner investment? Manipulating densities of the acacia ant Crematogaster mimosae on Acacia (Vachellia) drepanolobium, we found that protection from elephants and galling insects saturated at moderate ant abundance, while maintaining large colonies reduced fruit production. Beyond a point, more ants bring costs without added benefit: enough is enough.

Introduces the UHURU experiment (Ungulate Herbivory Under Rainfall Uncertainty), a replicated series of 1-hectare exclosures across a rainfall gradient in central Kenya that progressively remove three size classes of large mammalian herbivores, designed to reveal the functional roles of different herbivores and how rainfall shapes their effects.

Why do some ant colonies have multiple queens? Using microsatellite genotypes, we found that colonies of the acacia ant Crematogaster mimosae are often secondarily polygynous, with multiple closely related queens and occasionally unrelated queens, a strategy that may help colonies compete for scarce nesting sites on host trees.

How do plant communities respond to wildlife declines in real landscapes? Comparing exclosure experiments with sites where human activity has replaced wildlife with livestock, we show that plant responses to herbivore loss are stronger in less productive environments and differ considerably between exclosures and real-world land-use change, so exclosures alone may not predict the consequences of defaunation.

African spiny mice can shed patches of skin to escape predators. We show that these mice heal and regenerate the lost tissue, including hair follicles, glands, dermis and cartilage, with little scarring, revealing a mammal with remarkable regenerative abilities and a new model for studying tissue regeneration.

How is cooperation maintained in underground networks where partners can cheat? Manipulating cooperation in plants and arbuscular mycorrhizal fungi, we show that plants detect and preferentially reward more cooperative fungi with carbohydrates, and fungi reward plants that provide more carbon with more phosphorus, a reciprocal reward system that stabilizes cooperation.

Do small, cryptic herbivores alter the effects of large ones? Combining a large-scale ungulate exclusion experiment with a five-year manipulation of rodents, we quantified the impacts of wild ungulates, cattle and rodents on every life stage of a savanna tree. Wild ungulates shifted the population from growth to decline, but the mechanisms depended strongly on rodents, which limited seedling survival.

How much does protection by ants actually benefit a tree? Experimentally removing each of four ant symbionts from Acacia (Vachellia) drepanolobium for four and a half years, we found that protection against herbivores poorly predicted net effects on trees. Removing the three Crematogaster species increased tree growth or reproduction, while the poorly protective but low-cost Tetraponera penzigi had no net negative effect.

A book chapter reviewing some fifteen years of experiments at Mpala Research Centre on how large mammals, including cattle, indirectly affect less conspicuous species such as plants, small mammals, reptiles and invertebrates. Excluding ungulates generally increases plants and many small consumers, but these indirect effects vary with environmental context and differ among herbivore species.

Natural selection acts on lifetime fitness, yet most studies of mutualism measure the costs and benefits of a single partner at a single moment. Using long-term demographic data on Acacia (Vachellia) drepanolobium and its four ant symbionts, we show that trees achieve the highest lifetime fitness by hosting a sequence of partners over their lives, including partners that provide little protection. Partner diversity, rather than partner quality alone, can stabilize cooperation in multispecies mutualisms.

Mutualisms bind species to a common fate, so their breakdown could accelerate the effects of global change. We develop an evolutionary perspective on mutualism breakdown, focusing on shifts from mutualism to antagonism, switches to novel partners and abandonment of mutualism, and identify the factors that make particular mutualisms especially vulnerable.

Savanna tree cover is notoriously unstable, but symbiotic ants can stabilize it. In feeding trials and field experiments, elephants avoided whistling-thorn trees with ants but severely damaged trees from which ants were removed. As elephant numbers rose, cover of ant-defended trees remained stable while cover of undefended trees declined sharply, showing that a mutualism can buffer landscapes against megaherbivore-driven change.

Do regular spatial patterns in nature boost ecosystem function? We show that evenly spaced termite mounds in an African savanna are hotspots of plant productivity and of animal abundance, from insects to their predators, and that the regular spacing of mounds enhances productivity across the whole landscape compared with random or clumped patterns.

How do termites shape nitrogen cycling? Using stable isotopes, we show that Acacia (Vachellia) drepanolobium trees near termite mounds, where soils are nutrient rich, rely less on nitrogen fixation by their bacterial symbionts than trees farther away, with consequences for neighboring grasses and forbs, so termites create spatial structure in a key ecosystem process.

Which method best samples beetles in savanna tree canopies? Sampling 1,440 Acacia (Vachellia) drepanolobium trees quarterly over 14 months, we compared branch beating with canopy fogging for surveying beetle communities, as a step toward measuring how grazing and browsing by large mammals affect canopy insects.

What limits tree establishment in savannas? Combining seed addition and predator exclusion with long-term large-herbivore exclosures, we found that wild herbivores and cattle facilitated tree establishment by suppressing rodents, the main seed and seedling predators. Wild herbivores also reduced seed production, offsetting this benefit, while cattle further helped by reducing grass cover.

What determines fruiting in a dominant savanna tree? Using observations and long-term experiments, we found that Acacia (Vachellia) drepanolobium trees near termite mounds had higher leaf nitrogen and were more likely to fruit, regardless of whether large herbivores were excluded. Soil enrichment by termites, rather than dung from herbivores, drives fruiting near mounds, and fruiting also depended on the identity of the resident ant.

A published exchange in Science. Responding to a letter suggesting that fire, rather than browsing, might explain our finding that ant-acacia mutualisms break down when large herbivores are removed, we explain why fire cannot account for our results in upland Kenya, describe how acacia ants respond to fire, and note that very different pathways can lead to similar shifts in the ant community.

Do mutualisms depend on the herbivores they protect against? After a decade of excluding large herbivores in Kenya, Acacia (Vachellia) drepanolobium trees reduced the nectar and housing they provide to their ant partners. The dominant mutualist ant declined, while a less beneficial ant that associates with wood-boring beetles increased, and trees grew more slowly and died more often, showing that losing large herbivores can unravel a mutualism.

An accessible account in the Bulletin of the Ecological Society of America of our work on the acacia-ant mutualism, describing how each partner pursues its own interests and how what looks like cooperation can be viewed as reciprocal exploitation, with costs and benefits that shift with herbivores, competitors and environment.

Do ant defenders protect all parts of a plant equally? Comparing Acacia (Vachellia) drepanolobium trees occupied by four ant species, we found that ants recruited strongly to new leaves but very weakly to developing fruits. Aggressive ants reduced damage to leaves and stems, but not seed predation, and fruit production was unrelated to ant identity, so plant-ants defend foliage but not reproductive structures.

What does browsing cost a savanna tree in reproduction? Using a long-term exclosure experiment, we found that Acacia (Vachellia) drepanolobium trees protected from large herbivores were twice as likely to reproduce and produced more seed. Trees respond to browsing by growing longer spines, an induced defense that reduces reproduction but partly offsets the negative effects of herbivory where browsers are present.

A chemical analysis identifying the volatile compounds in the mandibular glands of Tetraponera penzigi, one of the four ant symbionts of the whistling-thorn acacia, compounds likely involved in alarm and defense behavior.

How does fire affect savanna birds? Using a replicated experimental burn in central Kenya, we found that bird diversity was higher on burned plots almost two years after the fire, with some bird families found only on burned plots, while total abundance and richness did not change.

What limits the founding of new ant colonies? Surveying saplings of Acacia (Vachellia) drepanolobium, we found intense competition among founding queens for swollen thorns, with combat the most common cause of queen death. Species differed in their success in these contests, and a parasitic wasp frequently killed the brood of solitary queens, revealing barriers to colony establishment that influence coexistence.

Do livestock and wildlife compete? Using the long-term KLEE exclosure experiment in Kenya, we found that zebras used plots more heavily where cattle were excluded, that cattle fully compensated for the absence of wildlife in their effects on grass cover, and that elephants mitigated the strong competition between cattle and zebras.

What determines dominance among competing ant species? Studying battles among the four acacia ants for host trees, I show that conflicts are wars of attrition with roughly equal losses on both sides, that competitive outcomes track average colony size, and that experimentally manipulating colony size reverses the dominance hierarchy, the first experimental demonstration that size asymmetries underlie dominance across an entire guild.

Should plants reduce costly defenses when herbivores disappear? We found that Acacia (Vachellia) drepanolobium trees protected from browsing mammals for seven years produced fewer nectaries and fewer swollen thorns, the rewards that attract defensive ants, than trees exposed to browsers, the first demonstration that induced indirect defenses relax when herbivory declines.

Can spatial variation in resources promote coexistence? Termite mounds create a mosaic of productivity in which host trees grow faster and prey are more abundant. I show that dominant acacia ants are more likely to take over trees near mounds, while subordinate species do better away from mounds, so termite-generated habitat heterogeneity helps maintain coexistence among competing ants.

A review of competition and coexistence within guilds of mutualists that compete for the same partner, drawing on plant–pollinator and ant-plant systems. We examine evidence for competition for partners, the mechanisms that allow competing mutualists to coexist, and how interactions with the shared partner influence these dynamics.

How does a weak competitor hold its ground? We show that the least aggressive acacia ant, Tetraponera penzigi, modifies its host tree to resist takeover: it destroys the tree's nectaries and keeps entrances to its swollen thorns too small for rival ants. Experimentally restoring nectar and enlarging entrances increased takeovers, showing that 'burning bridges' helps this subordinate species persist.

How do patch shape and boundary contrast affect animal movement in fragmented habitats? In a California grassland experiment, we manipulated patch shape and the contrast of patch boundaries by mowing, and found that boundary contrast strongly influenced beetle movement, with high-contrast boundaries producing net emigration, and that patch shape mattered in interaction with boundary type.

Can trade-offs between competing and colonizing explain coexistence? Among the four acacia ants, the competitive hierarchy among mature colonies is offset by differences in the ability to colonize newly available trees, through colony expansion and through founding queens, so competition–colonization trade-offs help maintain diversity in this guild.

A chemical comparison of the volatile compounds in the mandibular glands of three coexisting Crematogaster acacia ants. The species differ markedly in their gland chemistry, differences that may underlie species recognition and species-specific alarm and defense responses during competitive interactions.

How do savanna trees respond to herbivory? Simulating shoot and leaf browsing on Acacia (Vachellia) drepanolobium with different histories of large-mammal herbivory, we found strong compensatory regrowth, but past herbivore pressure was associated with reduced net shoot growth, suggesting long-term costs of herbivory even when plants appear to compensate.

How do four ant species share a limiting resource? Using staged conflicts, natural transitions on nearly 1,800 tagged trees and tree-size patterns, we confirm a linear dominance hierarchy among the acacia ants. During a dry period, dominant ants abandoned slowly growing small trees, and takeovers tracked host growth, suggesting that tree dynamics help maintain coexistence.

Why would an ant sterilize its host plant? We show that the acacia ant Crematogaster nigriceps prunes the flower buds and lateral shoots of its host tree, sterilizing it and reshaping its canopy. Pruning reduces contact with neighboring trees, protecting the colony from invasion by competitively superior ants, revealing a strategy that benefits the ant at the plant's expense.

Introduces the Kenya Long-term Exclosure Experiment (KLEE), a set of replicated semi-permeable barriers in Laikipia, Kenya, that differentially exclude cattle, large wild herbivores and megaherbivores, designed to understand interactions between livestock, wildlife and the wider ecosystem.

How influential are microcosm 'bottle' experiments in ecology? Surveying journals and textbooks, we assessed how common microcosm studies of species interactions are, how they differ from field studies in the number of species and duration, and how often they are cited and taught.

How does patchiness affect animal growth? In laboratory streams, I manipulated resource distribution and current velocity and found that mayfly growth depended on both, with drift behavior, the way nymphs move among patches, explaining the effects of spatial heterogeneity on growth.

What do young ostriches eat? Observing penned and free-ranging chicks in an East African savanna, we found that juvenile ostriches show innate dietary preferences, favoring newly emerging grass leaves and certain woody forbs, with little change in diet as they gain experience.

Does intense pollen competition produce more vigorous offspring with better pollen? Manipulating pollen load size in Brassica campestris, we tested whether stronger competition among pollen grains in one generation leads to offspring with greater vigor and more competitive pollen in the next.