How Cordyceps Fungi Control Insect Behavior: The Strange Neuroparasitology of Zombie Ants

A fungus that takes over an ant's nervous system to spread itself sounds like fiction. It is documented biology, dating back to the 19th-century descriptions of Carpenter ants found high in tropical canopies with fungal stalks growing from their heads, in positions and at heights that would not occur if the ants had died of ordinary causes. The basic phenomenon has been understood for over a century; the mechanism is beginning to be characterized in the last decade. The story is one of the strangest examples of cross-kingdom behavioral manipulation in biology.

The basic phenomenon

The fungus is Ophiocordyceps unilateralis (one species within a genus of several hundred related insect-pathogenic fungi). It infects carpenter ants in the genus Camponotus in tropical forests across Asia, South America, and Africa. The infection begins when an ant walks under a fungal-spore drop zone in the forest understory and a spore lands on its exoskeleton. The spore germinates, the hyphae penetrate the cuticle, and the fungus begins growing inside the ant's body, consuming the soft tissues while leaving the nervous system and muscle structure relatively intact.

Over the next several days, the ant's behavior changes in stages. The ant leaves its normal foraging territory and begins climbing. It bypasses the canopy where it usually forages. It descends to a layer about 25 centimeters above the forest floor (a height with stable temperature and humidity that the fungus prefers for fruiting). It attaches itself to a leaf vein, biting down with its mandibles in what biologists call the "death grip" because the muscles lock and the ant cannot release. Then it dies, the fungus consumes the remaining soft tissue and grows a fruiting body out of the ant's head, and after one to three weeks, the fruiting body releases new spores that fall to the forest floor to infect new ants.

The behavior is reliable and specific. Carpenter ants of the susceptible species in infected territories show distinctive death-grip postures on leaves about 25 cm above the ground, with fungal fruiting bodies growing from their heads. The David Hughes lab at Penn State documented dozens of consistent behavioral signatures across multiple geographic populations.

The mechanism question

The straightforward hypothesis would be that the fungus enters the ant's brain and directly controls behavior. This turns out to be wrong. The Hughes lab's 2017 PNAS paper (Fredericksen et al.) used serial-block-face scanning electron microscopy to reconstruct the distribution of fungal cells in infected ant heads at high resolution. The result was unexpected: fungal cells fill the soft tissue of the ant's body and head but largely avoid the brain itself. The behavior change happens with the ant's central nervous system substantially intact.

This complicates the mechanism story considerably. The fungus is changing behavior without direct neural infection. The current best-guess explanation is chemical: the fungus releases bioactive compounds that affect the ant's muscle control, sensory inputs, or modulatory neurotransmitter systems from the periphery. Several candidate compounds have been identified in infected ants that are not present in healthy ants, including alkaloids and small peptides. The Charissa de Bekker lab (now at Utrecht) has done substantial work on the molecular interface.

The muscle-control angle is particularly interesting because the "death grip" itself appears to be a kind of fungal hijacking of the mandibular muscles. Microscopy of the muscle tissue in the mandibles of infected ants shows the muscle fibers atrophied and replaced by fungal hyphae, but the fibers remaining are locked in contracted position. The fungus appears to physically replace the muscle while preserving its tension, producing the lock-jaw posture that holds the ant in position even after death.

The behavioral repertoire

The "zombie ant" behavior is not a single behavior but a sequence: leave the colony, climb downward to a specific height, grip a leaf vein, hold position. Each element is precise. The 25 cm height is consistent enough that researchers can predict where infected ants will be found. The leaf-vein grip is mechanically optimal for holding the ant in position against rain and wind. The timing of the final death grip correlates with solar noon, possibly so the fungus has maximal stable conditions for the next fruiting-body growth phase.

The Hughes lab also documented that the infected ants do not bite random leaves. They bite the underside of leaves of specific species, in positions that maximize the surface area available for the fungal fruiting body and the drop zone for spores. The behavioral specificity is fine-grained.

What is not clear is how much of this behavior is fungal-directed and how much is normal ant behavior expressed in a sick ant. Healthy ants do not climb downward to fixed heights and bite leaves to death, so at least the gross outline of the behavior is fungal. But ants do have stereotyped responses to illness (leaving the nest to die, climbing to high points) and the fungal manipulation may be partly exploiting existing pathological responses rather than creating new behaviors from scratch.

Convergent evolution in other taxa

Ophiocordyceps is the most famous case but not the only one. Behavioral manipulation by parasites is widespread in biology.

The nematomorph horsehair worm (Spinochordodes tellinii) develops inside terrestrial crickets and grasshoppers, and when ready to reproduce, manipulates the host into seeking water and jumping in. The worm then emerges and reproduces in the water; the host typically drowns.

The trematode Dicrocoelium dendriticum (the lancet liver fluke) cycles through snails, ants, and grazing mammals. The infected ant climbs to the top of a blade of grass at dusk and waits there motionless, where it is more likely to be eaten by a grazing cow or sheep that will become the next definitive host. The mechanism appears to involve a single fluke larva that lodges in the ant's subesophageal ganglion and modulates motor output.

The Toxoplasma gondii single-celled parasite cycles through rodents and cats; infected rodents lose their fear of cat urine and become more easily caught. Human Toxoplasma infection (carried by perhaps 30% of the global human population, mostly asymptomatically) has been claimed to affect human behavior, with weakly supported correlations to risk-taking and to schizophrenia in vulnerable subpopulations. The human-behavior claims are controversial; the rodent-behavior changes are well documented.

The pattern across taxa is that behavioral manipulation by parasites evolves repeatedly when the parasite's life cycle benefits from the host being in a specific situation that the host would not otherwise seek. The mechanisms vary (fungal chemistry, larval lodging in ganglia, single-cell modulation of dopamine signaling), but the recurring pattern is that nervous systems are manipulable from the periphery in ways that biology has discovered multiple times independently.

The medical and applied angle

The Ophiocordyceps phenomenon is interesting medically because it represents an existence proof for behavioral manipulation by external compounds acting on a nervous system from the periphery. The compounds the fungus uses are unknown in detail, but they are by definition small enough to cross the insect cuticle and reach the relevant neural targets, and they produce specific behavioral outputs rather than general impairment. Characterizing these compounds is an active research area with potential applications in understanding behavioral pharmacology in general.

A separate genus, Cordyceps sinensis (recently renamed Ophiocordyceps sinensis), is one of the most expensive natural commodities in the world, used in traditional Chinese medicine and selling for $50,000 per kilogram. The medicinal claims are weakly supported but the demand has caused severe overharvesting in the Tibetan plateau where it grows. The species infects caterpillars rather than ants and has been a subject of mycological and pharmacological research for decades.

The science-fiction angle (popularized by the video game and television series The Last of Us, which imagines a cordyceps-like fungus jumping to humans) is not a serious biomedical concern. Cordyceps species are highly specialized to specific insect hosts; the molecular machinery the fungi use to infect insect cuticles and manipulate insect physiology does not work in mammalian biochemistry. Mammalian body temperatures are also outside the optimal range for most cordyceps species. The fiction works because it draws on a phenomenon that is genuinely strange enough to be unsettling.

Three observations

First, the textbook story of "fungus controls ant brain" turned out to be wrong in detail. The fungus controls ant behavior without directly infecting the brain. This is a recurring pattern in biology: the obvious explanation is sometimes wrong and the real mechanism is stranger than the initial guess. The 2017 PNAS paper that established this was substantially more interesting than it would have been if the fungus had infected the brain as expected.

Second, behavioral manipulation by parasites is widespread across multiple taxa with multiple independent mechanisms. The pattern is not exotic. Once you start looking, parasitic manipulation of host behavior shows up in nematodes, trematodes, cestodes, single-celled parasites, viruses, bacteria, and fungi. The recurring evolution of these mechanisms suggests that nervous systems are more manipulable than the canonical "brain controls behavior" framing implies.

Third, the long time gap between phenomenon and mechanism is typical of biology. Carpenter-ant zombification was described in the 19th century, the basic life cycle was worked out in the early 20th century, the high-resolution anatomy is from the 2010s, and the molecular mechanism is still under active investigation in 2026. The pattern of "observe a strange phenomenon → characterize it taxonomically and ecologically → eventually figure out the mechanism over decades" is more typical of biology than the rapid-discovery narrative that physics provides.

Deeper observation

Zombie ants are one of those cases that should be more famous for what they reveal about the universe than for the science-fiction-like quality of the phenomenon. Behavior is not always controlled from inside the organism by mechanisms the organism owns. Nervous systems are interfaces that can be manipulated from outside by compounds and processes that have evolved precisely to exploit them. The boundary between organism and environment is more permeable than the canonical model-organism-centered biology curriculum suggests, and the inventory of biological mechanisms for crossing that boundary is much larger than the casual scientific imagination tends to assume.