Skip to content
NatureHQ

Fungifungus

The ant-manipulating fungus

The fungus is not in the ant’s brain. It fills the body and wraps the muscle fibres — and the brain is left alone.

The fungus makes an infected ant climb, orient and bite into a leaf vein at a consistent height and time of day — and then it fruits from the body. What it does not do is invade the brain. It fills the body and wraps individual muscle fibres, and the brain is left essentially untouched.

The behaviour is real and it is precise, which is what makes the mechanism worth getting right. An infected worker leaves the colony, descends from the canopy to a narrow band of height, orients in a consistent direction, and bites into a leaf vein around the middle of the day. The bite locks; the ant dies there; and days later a stalk grows from the back of its head and releases spores over the trail below. The precision is not incidental. When researchers moved dying and dead ants outside that band of height and microclimate, the fungus frequently failed to develop and produce spores at all. The ant is dying somewhere the fungus needs it to die, which is what makes this an extended phenotype rather than an illness with a coincidental location. Now the part that the popular telling gets backwards. Three-dimensional reconstruction of infected ants — serial sectioning combined with machine-learning segmentation — mapped where the fungal cells actually are. They fill the body cavity and form networks around and between individual muscle fibres, including in the mandibular muscles that produce the locking bite. The brain is essentially free of them. Whatever is directing this animal, the fungus is not sitting in the control room operating it; it appears to be working on the muscles, and on the ant’s physiology, from outside the nervous system. A chemical route is likely and unproven. Cultured with the brain of its natural host, the fungus secretes a different set of compounds than it does with a brain from a species it cannot infect — which fits a chemical mechanism tuned to a particular host, and stops well short of identifying a molecule that causes a behaviour. NatureHQ describes what is known and marks the mechanism as unresolved, because it is.

Early coverage · 55% complete · reviewed 2026-09-03

What this page covers

A genus of many species, each highly specific to one ant species or a small group. The famous behaviour is documented in a handful of host–parasite pairings, and generalising from one to the genus is the standard error.

Often confused with: Cordyceps, a related genus often used loosely for all of them; A fungus that controls a brain, which is not what the anatomy shows; One species with many hosts, when these fungi are highly host-specific

Quick facts

Not in the brain
Fungal cells fill the body and wrap muscle fibres; the brain is essentially free
A precise death site
Consistent height, orientation and time of day
And it matters
Ants moved elsewhere often produced no spores at all
Host-specific
Many species, each tied to one ant — not one fungus with many victims

What actually happens

A precise sequence, ending somewhere the fungus needs.

Infected ants die in a narrow band of height, orientation and microclimate — and when researchers moved them somewhere else, the fungus often failed to produce spores at all.

Well supported

Good evidence backs this, though some details remain open.

Ophiocordyceps-infected Camponotus ants die at a consistent height, orientation and microclimate, biting into vegetation around solar noon. Experimental transplantation of dying and dead ants outside this band substantially reduced or eliminated fungal fruiting-body development and spore production.

Who this applies to
One host–parasite pairing in one forest; Ophiocordyceps species are highly host-specific.
Studied in
Camponotus leonardi, Ophiocordyceps unilateralis
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

The transplant is what makes this a demonstration rather than a description: moving the ants and watching the fungus fail shows the location matters and is not incidental to being ill.

Caveats

  • One system in one forest, and transplantation is itself a disturbance that could affect development independently.
  • It shows the death site is favourable for the fungus without establishing how the ant is directed there.

Still unanswered

  • How the ant is steered to a specific height and orientation, which is the central unanswered question of the whole system.

Last reviewed 2026-09-03

The evidence (2 studies)

How we know

Moving the ant the fungus placed

Infected ants die at a consistent height, orientation and time of day. Is that precision doing anything for the fungus, or is it incidental?

Manipulated ants were located in the field and their death sites recorded, establishing the pattern. Ants were then transplanted — to the canopy above the usual band, and to the forest floor below it — and left to develop. Whether the fungus went on to produce a fruiting body and spores was scored at each location, against ants left where they had bitten.

What happened

Ants left where they bit developed normally. Transplanted ants frequently failed to produce a fruiting body at all, with development impaired both above and below the natural band.

What it shows

That the location is load-bearing. The fungus is not merely making a sick animal wander; it is placing the ant where its own reproduction is possible, and getting the placement wrong costs it everything. That is what makes the behaviour an extended phenotype rather than a symptom.

What it does not show

It does not identify which feature of the site matters — humidity, temperature and light co-vary with height, and this design cannot separate them. It also says nothing about how the ant is steered there.

The controls — what makes this evidence rather than a story
  • Ants left undisturbed at natural death sites as the comparison, so handling alone cannot explain a difference.
  • Transplants in both directions, up and down, distinguishing a specific band from a simple gradient.
  • Death sites mapped before any transplant, so the pattern was established independently of the test.

From The life of a dead ant: the expression of an adaptive extended phenotype

The transplant is what converts this from a description into evidence. A sick animal wandering and dying would produce some distribution of death sites, and one could tell a story about any of them. Moving the ants and watching the fungus fail shows that the location is load-bearing — the manipulation is achieving something specific for the parasite, and doing it wrong costs the fungus its reproduction.

The part everyone gets backwards

Where the fungus actually is, once somebody looked.

The fungus that makes an ant climb and bite does not invade its brain. It fills the muscles and leaves the brain alone.

Well supported

Good evidence backs this, though some details remain open.

Three-dimensional reconstruction of Camponotus castaneus infected with Ophiocordyceps unilateralis found fungal cells forming an interconnected network throughout the body cavity and surrounding and penetrating muscle fibres, including mandibular muscle, with the brain conspicuously free of fungal cells.

Who this applies to
One fungus and one ant host, in laboratory infections.
Studied in
Ophiocordyceps unilateralis, Camponotus castaneus

You may have heard

“The zombie-ant fungus takes over the ant’s brain.”

It leaves the brain untouched and takes the muscles instead — which is stranger. The ant’s nervous system is intact while its body is carried somewhere it did not choose.

Why we rate it this way, and what the caveats are
Well supportedModerate confidence

A careful anatomical reconstruction with a clear and unexpected result, but a single host–parasite pair and a description of location rather than of mechanism.

Caveats

  • Other manipulating fungi work differently; this establishes nothing about fungal manipulation in general.
  • Where the fungus is does not establish the chemistry by which the behaviour changes.

Still unanswered

  • Which compounds produced by the fungus act on the host, and where they act.

Last reviewed 2026-08-30

The evidence (1 study)

In culture with the brain of the ant it infects, the fungus secretes different compounds than with a species it cannot infect. That fits a chemical mechanism tuned to one host — without naming it.

Emerging evidence

Real findings exist, but too few or too recent to be settled.

Ophiocordyceps species cultured in the presence of host and non-host ant brain tissue produce differing secreted metabolite profiles, consistent with host-specific chemical interaction. No individual compound has been shown to produce the manipulated behavioural sequence.

Who this applies to
Demonstrated in culture for particular fungus–ant pairings; not a statement about the genus as a whole.Do not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Ophiocordyceps, Camponotus
Why we rate it this way, and what the caveats are
Emerging evidenceModerate confidence

The differential secretion is a real measured result. The step from it to a mechanism of behaviour is an inference, and is marked as one.

How far it can be extended

These fungi are highly host-specific, and results from one pairing are not evidence about another.

Caveats

  • A culture dish is not an infected ant, and a compound secreted in one may not be present or active in the other.
  • No compound has been given to an uninfected ant and shown to produce any part of the behaviour.

Still unanswered

  • Which secreted compounds, if any, are responsible for the behaviour — and what they act on, given the fungus is not in the brain.

Last reviewed 2026-09-03

The evidence (1 study)

Diagram

Where the fungal cells actually are

A schematic of the reconstruction result: dense through the body and around muscle, absent from the brain.

Where the fungal cells actually areHead — brainEssentially freeHead — muscleNetworks presentThoraxDenseAbdomenDenseSchematic, not a specimen. Shaded blocks carry fungal networks.Reconstructed by serial sectioning with machine segmentation, which is the onlypractical way to map a network this fine through a whole animal. The brain — theorgan the popular story assigns the work to — is the one place it is not.
The same explanation in words

Four blocks represent regions of an infected ant: the brain within the head, the muscle of the head, the thorax and the abdomen. The three carrying fungal networks — head muscle, thorax and abdomen — are shaded and outlined; the brain is left plain and marked as essentially free of fungal cells. The diagram is labelled as a schematic rather than a depiction of a specimen. A note records that the mapping was done by serial sectioning with machine segmentation, the only practical way to trace a network this fine through a whole animal, and that the brain — the organ the popular story assigns the work to — is the one place the fungus is not.

The reconstruction is worth describing because the result depends on the method. Infected ants were serially sectioned and imaged, and the fungal material identified through the whole body by a trained segmentation model rather than by eye — which is the only practical way to map a network that fine through an entire animal. What emerged was fungal cells packed through the body cavity and forming interconnected networks around individual muscle fibres, and the brain almost entirely free of them.

That does not make the manipulation less remarkable; it relocates it. Something is producing a highly specific behavioural sequence in an animal whose nervous system the parasite is not occupying — which is a harder problem than direct neural control, not an easier one. The likely route is chemical, and the evidence for it is that the fungus secretes a different suite of compounds when cultured with its own host’s brain than with a species it cannot infect. Which molecules matter, and what they act on, is unknown.

The comparison that makes this land

  • Parasitism and predation

    Where a wasp that really does operate on a brain sits beside this one

  • Fungi

    What a fungal body is, and what it can do

  • Ants

    The animal on the other end of this

What is not known

  • How is the ant steered, if not through the brain?

    Why it matters: The behaviour is specific to a height, an orientation and a time of day, and the parasite is not in the organ that would ordinarily produce such a sequence. Whatever mechanism achieves that is more interesting than direct neural control, and it is entirely unidentified.

    What would settle it: Identifying secreted compounds present during manipulation and testing them individually on uninfected ants — which requires culturing a fastidious fungus and delivering candidate molecules in realistic amounts.

  • Why is each fungus tied to one ant species?

    Why it matters: Host specificity is extreme in this group, and the manipulation appears tuned to a particular species’ physiology. Understanding what makes a host suitable would say a great deal about how the manipulation works at all.

Claims about this, checked

Things people have heard, and what the evidence actually supports.

The research behind this page

5 studies, newest first. Each one has a page explaining what it found and what it could not show.

2017Proceedings of the National Academy of Sciences

Three-dimensional visualization and a deep-learning model reveal complex fungal parasite networks in behaviorally manipulated ants

Fungal cells filled the ant’s body and formed an interconnected network surrounding and penetrating muscle fibres, including in the mandibles.

2014BMC Evolutionary Biology

Species-specific ant brain manipulation by a specialized fungal parasite

The fungus produced different sets of compounds depending on whose brain it was exposed to, secreting a distinct profile in the presence of its natural host’s brain compared with a non-host species.

2011BMC Ecology

Behavioral mechanisms and morphological symptoms of zombie ants dying from fungal infection

Infected ants leave the canopy, descend to a consistent height and orientation, and bite into a leaf vein around solar noon, dying there.

2009The American Naturalist

The life of a dead ant: the expression of an adaptive extended phenotype

Ants died in a narrow band of height, orientation and microclimate.

2009Annual Review of Entomology

Manipulation of host behavior by parasitic insects and insect parasites

Behavioural manipulation is achieved by several distinct routes — direct injection into the nervous system, secretion of neuroactive compounds, and alteration of host physiology — and the neural mechanism is well characterised in very few systems.

Where to go from here

Each of these follows from something on this page — a relationship in the evidence, a claim people ask about, or the next mechanism along.

How complete this page is, and what it is still missing

NatureHQ publishes its own gaps. This page is at 55% completeness against what we would call a finished subject, and was last reviewed on 2026-09-03. It carries 3 claims and answers 10 mapped search questions.

  • 5 high-priority search intent(s) not yet covered
  • no research from the last few years is attached — check for newer work
  • more experiments could be explained in plain English
  • The mechanism of manipulation is unidentified and is presented as unidentified.
  • Most work is on one or two host–parasite pairings out of many species in the genus.
  • The fungus’s wider ecology, including its effect on ant colonies, is treated only briefly.