Cordyceps became a household word through The Last of Us, and the premise the game builds on is roughly accurate as far as it goes: there is a fungus that infects insects, alters their behavior in specific and reproducible ways, and kills them in a location that benefits the fungus. The mind control is real. The scientific literature on it is stranger than the fictional version.

Where the popular account goes wrong is in the mechanism it assumes. The intuitive model — the one the game and most explainers use — is that the fungus invades the brain and drives it. That model is almost the opposite of what happens. And the fungus's inability to do what the fiction requires is not a matter of degree. It is structural.

What Actually Infects Ants

First, a naming problem worth clearing up. The species that manipulates carpenter ants is Ophiocordyceps unilateralis, not Cordyceps proper. The genus was split in 2007 on molecular phylogenetic grounds, and the Cordyceps that turns up in supplement aisles — C. militaris, or the caterpillar fungus O. sinensis — is a different lineage doing different things. The zombie-ant fungus is in Ophiocordycipitaceae.

The infection cycle is well characterized. A spore lands on a foraging ant, penetrates the cuticle enzymatically, and proliferates inside the body as yeast-like cells circulating in the hemolymph. Over one to two weeks the fungus grows through the ant's tissues. Then the behavior changes.

An infected Camponotus ant leaves the colony, descends from the canopy, and climbs a plant to a characteristic height — roughly 25 cm above the forest floor in the Thai populations Hughes and colleagues studied. It orients to a specific side of a leaf, relative to the sun. It bites down on a major leaf vein and locks its mandibles there. This is the "death grip." The ant dies in that position, and days later a stroma erupts from the back of its head and rains spores onto the foraging trails below.

The precision is the striking part. The height, the humidity, the orientation, the vein — these are the conditions under which the fungus fruits best. The ant is being positioned.

The Brain Is Untouched

Here is the finding that reorganizes the whole picture.

In 2017, Hughes's group at Penn State published a three-dimensional reconstruction of infected ant tissue using serial block-face scanning electron microscopy, imaging fungal cells throughout the ant's body at the moment of the death grip. Fungal cells filled the head capsule, surrounded the mandibular muscles, invaded muscle fibers, and formed an interconnected network.

The brain was free of fungus.

This was not a partial finding. The fungal cells were densely packed around neural tissue and absent from inside it. Whatever the fungus is doing, it is not physically occupying the central nervous system.

Fredericksen and colleagues followed with machine-learning-assisted segmentation to quantify it: fungal cells accounted for roughly 40% of the biomass in the head of an infected ant, and the pattern held — muscles colonized, brain avoided. They also found the fungal cells connected to one another by tube-like structures, suggesting the network functions as a coordinated system rather than as independent invaders.

The mandibular muscle in a death-gripping ant shows atrophy and disorganized fibers, with fungal cells threaded between them. The ant is not clenching its jaw by decision. The muscle has been physically altered so that it cannot release.

Two Different Problems

Once you see the brain result, the manipulation splits into two mechanistically distinct questions.

The death grip itself is peripheral and largely mechanical. Fungal invasion of the mandibular adductor muscle, atrophy, and possibly direct chemical action at the neuromuscular junction produce a sustained contraction that outlives the ant. This is not mind control. It is closer to hijacking the actuator.

The summiting behavior is harder. An ant leaving the nest, descending, climbing a plant, and stopping at a particular height and orientation is executing a sequence, and a sequence requires the nervous system. If the fungus is not in the brain, it has to be acting on it from outside — secreting compounds into the hemolymph that cross into or act upon neural tissue.

There is evidence for this. Transcriptomic work from de Bekker and colleagues found that O. unilateralis upregulates candidate secreted effectors specifically when co-cultured with brains of its natural host ant species, and not with brains of non-host ants. The fungus is producing something targeted, and the targeting is species-specific. Enterotoxins, protein tyrosine phosphatases, and a guanine-specific ribonuclease appear among the differentially expressed genes.

Other work has implicated ergot alkaloids and related compounds, and there is evidence for disruption of biogenic amine signaling — dopamine and serotonin pathways that govern locomotion and foraging in insects. Sporulation-related genes and circadian genes are differentially regulated at the time of manipulation, which fits the observation that the death grip occurs around solar noon with considerable temporal precision.

None of this amounts to a complete account. Nobody has identified a single compound and shown that administering it produces summiting. The honest state of the field is that the fungus secretes a targeted cocktail into the hemolymph, and the specific molecules and receptors remain open.

Why It Cannot Work On Us

The fictional premise requires a jump to humans, usually justified by climate change selecting for heat tolerance. The problems with this are not close calls.

Thermal tolerance. Most fungi cannot grow at mammalian body temperature. O. unilateralis is adapted to tropical forest understory conditions — roughly 20–25°C and high humidity. Human core temperature is 37°C, and a fever pushes it higher. This is a substantial part of why fungal pathogens of mammals are rare relative to fungal pathogens of insects and plants: endothermy is a broad antifungal defense.

Immune systems. Insects have innate immunity — antimicrobial peptides, phenoloxidase cascades, hemocytes. They lack adaptive immunity entirely. No T cells, no antibodies, no immunological memory. A fungus that spends two weeks proliferating in an open hemolymph cavity is exploiting a defense architecture that mammals do not have. In a human, that same fungus would face a coordinated cellular and humoral response with memory.

The fungal infections that do kill humans are almost entirely opportunistic and almost entirely in immunocompromised people. Aspergillus, Cryptococcus, Candida — these become lethal when the immune system is already gone. Candida auris is genuinely concerning as a drug-resistant hospital pathogen, and it is nothing like a behavior-manipulating parasite.

Host specificity. O. unilateralis is a species complex, and individual lineages are specialized to individual ant species. The de Bekker co-culture result showed the secreted effector response is host-specific — the fungus produces a different profile against a non-host brain. Millions of years of coevolution produced compounds tuned to particular receptors in particular nervous systems. That specificity is the mechanism, not an incidental limitation.

Nervous system architecture. An ant brain has on the order of 250,000 neurons. Insect behavior includes a large repertoire of fixed action patterns — stereotyped sequences that run to completion once triggered. Summiting behavior may work by tripping an existing program: the escape or climbing response that ants show when infected or dying, which is a real phenomenon called adaptive social isolation. The fungus may be pushing a button that already exists.

A human brain has roughly 86 billion neurons and very few fixed action patterns of that kind. There is no "climb a tree and bite a leaf" module to trip.

What It Actually Illuminates

The interesting thing about this system is not that it is nearly a horror movie. It is that it is a demonstration of how much behavior is chemically addressable from outside the nervous system.

The fungus never touches the brain. It sits in the hemolymph and the muscles and secretes molecules, and a complete behavioral sequence follows — leave the colony, descend, climb, orient, bite, hold. That sequence involves navigation, motor coordination, and timing. All of it is being driven by extracellular chemistry.

That is a claim about how nervous systems work, not just about fungi. Behavior we would describe as decision-making turns out to be modulable by compounds acting on neuromodulatory systems that were already there. The fungus did not build a control system; it found the existing one and learned its inputs.

There is a related literature that makes the same point from other directions. Toxoplasma gondii reduces rodent aversion to cat odor, which increases predation and returns the parasite to its definitive host. Hairworms drive crickets into water. Baculoviruses induce caterpillars to climb before liquefying them — the same summiting behavior, in an entirely different pathogen, which suggests summiting is a convergent solution to the problem of spore dispersal rather than a Cordyceps quirk.

The convergence is the tell. Multiple unrelated parasites independently discovered that manipulating host height improves transmission. That means the behavioral levers are accessible enough to be found repeatedly by evolution, which is a stronger statement about nervous systems than any single case would be.

None of this makes a human zombie fungus plausible. What it makes plausible is a more uncomfortable and better-supported idea: that the boundary between an organism's behavior and its chemical environment is thinner than it feels from the inside, and that a sufficiently coevolved parasite can find the seams.