Place nine wood blocks in a circle on a soil plate. Place nine more in a cross. Inoculate both with Phanerochaete velutina, a cord-forming fungus. Wait 116 days.
Fukasawa and colleagues at Tohoku University found that the mycelial networks that grew between the blocks were structurally different depending on the arrangement. In the circle, connectivity was uniform across all blocks — but the dead center of the ring remained empty. Hyphae grew outward from the perimeter, not inward through the shortest path between opposite blocks. In the cross, the four outermost blocks developed significantly greater connectivity than the inner blocks, functioning as exploration outposts. The fungus grew different architectures for different geometries.
The researchers framed this as pattern recognition: the organism distinguished a circle from a cross and responded differently to each. The framing invites the obvious objection — recognition implies a recognizer, and a fungus has no nervous system, no sensory organs, no central processor. The objection is correct but misses the structural point.
The fungus does not have a map of the circle. The fungus IS the map.
Each hyphal tip is a sensor-actuator that reads local conditions — chemical gradients, substrate topology, the presence of conspecific hyphae — and adjusts its growth direction in real time. When inward-growing hyphae from the circle's perimeter encounter hyphae growing inward from the opposite side, they detect conspecific saturation and no new resource. The cost-benefit calculation tilts against center growth. Outward-growing hyphae encounter unexplored territory with potential resources. The network reinforces outward cords and abandons inward ones.
No single hypha knows the arrangement is a circle. The circle emerges at the network level because the local decisions of thousands of tips, each responding to its immediate neighbors, produce a global pattern that reflects the spatial arrangement of resources. The network's architecture — which cords thicken, which are abandoned, where branching increases — encodes the environment's geometry as a structural fact about the organism's own body. The representation and the response are the same physical structure. There is no separate information layer.
This is different from brainless decision-making, where an organism without a nervous system produces a single binary output (the ant pupa signaling its own destruction, the immune cell choosing phagocytosis). The fungus produces a continuous, spatially extended representation — a network topology that mirrors the resource topology. The dimensionality of the response matches the dimensionality of the stimulus. And the representation persists: the cord architecture is the organism's memory of what it encountered and where.
The deeper point is about the separation of map and territory. In systems with brains, there is a physical distinction between the representation (neural activity pattern) and the thing represented (the environment). The map is inside; the territory is outside. The fungus dissolves this distinction. Its body grows into a shape that is simultaneously the territory (it IS the physical network connecting resources) and the map (the network's topology encodes the spatial relationships between those resources). When you look at the mycelium, you are looking at both the organism and its model of the world. They are the same object.
The question of whether this constitutes cognition depends entirely on the definition. If cognition requires a model that is separate from the thing modeled — a representation that can be manipulated independently of the environment — then the fungus does not think. If cognition requires only sensing, integration, and adaptive response, then it does. The definitional question is interesting but less interesting than the structural one: the assumption that maps must be separate from territories is an artifact of how brains work, not a requirement of information processing.