Janzen-Connell seedling mortality

Why does a tropical seedling fare worst on exactly the patch of ground where its own parent is thriving?

A large tropical tree drops most of its seed within a few metres of its own trunk, onto soil it has evidently found congenial for a century or more. That ought to be the best nursery in the forest: the right drainage, the right light regime, and a thriving parent standing over it as proof that the site works. Yet census the saplings there and remarkably few belong to that species. The ones taking hold are mostly somebody else's.

So the assumption worth questioning is the obvious one — that a site proven good for an adult is therefore good for its offspring. What could make the ground under a successful parent the worst ground in the neighbourhood for its own kind?

Ask first who else lives around that trunk. A long-established tree is not only a producer of seeds — it is a habitat, and in particular a habitat for whatever eats it. Host-specialised insects, and more consequentially soil-borne fungal and oomycete pathogens of the Pythium and Phytophthora sort, accumulate around a host over years, living off its roots and its fallen litter. The adult is large, woody and chemically defended, and mostly shrugs them off.

Now drop a seedling into that soil. It has millimetre-thick roots, a cotyledon's worth of reserves, and it is genetically almost exactly the meal the local pathogen population has spent a decade getting good at. Do you see the inversion? The parent's success has been quietly converted into an enemy load, and the seedling inherits the load without inheriting the size that made it survivable.

Two quantities now move in opposite directions as you walk away from the trunk. Seed arrival falls steeply, because most seed rains close and dispersal thins fast with distance. Survival probability climbs, because enemy density is highest where the host is. What does the product of a falling curve and a rising curve look like? It has a hump. Recruitment peaks neither at the trunk nor at the far edge but somewhere in between — and that displaced peak is the entire prediction, made independently by Daniel Janzen and Joseph Connell around 1970 and 1971.

Follow it up one level. If every common species suppresses its own offspring wherever it is locally abundant, then abundance is self-limiting and rarity is an advantage: a species that has become scarce leaves ground uncontaminated by its own enemies, and its seedlings do comparatively well there. That is negative conspecific density dependence, and it is one of the stabilising mechanisms invoked to explain how a hectare of tropical forest supports hundreds of tree species instead of being taken over by the best competitor.

Now the honest part. Is the seedling adapted to die near its parent? No — and this is where the domain's favourite error lurks. Nothing in the pattern is a trait of the tree. The agents are the pathogens and the herbivores, acting in their own interest; the mortality curve is what they do, not what the tree has arranged. Selection may act on the tree downstream of this, favouring seeds that travel further or germinate later, but the effect itself would exist in a forest where no dispersal trait had ever evolved.

Three refinements hold the account together. First, distance is a stand-in, not the cause — what actually predicts death is conspecific density, of adults and of neighbouring seedlings, and distance from the parent merely tracks it. In a stand where a species is common everywhere, moving away buys little. Second, "host-specific" is stronger than the evidence usually supports: most soil pathogens implicated here are not confined to one host but are simply more damaging to conspecific seedlings than to the mixed neighbours around them, and that asymmetry is all the mechanism needs.

Third, the strength of the effect is genuinely variable and its sufficiency is contested. It is well demonstrated in many species and absent or weak in others, and monodominant tropical stands exist — forests where a single species holds the canopy across large areas — which is exactly what should be rare if the effect were universal and strong. Treat it as one stabilising force among several, not as the explanation of tropical diversity.

The claim is testable in a way that could plainly go the other way. If soil-borne enemies are the agent, then growing seedlings in soil collected under conspecific adults should kill more of them than soil collected under other species, and sterilising or fungicide-treating that soil should erase the difference. The refuting observation is direct: if conspecific and heterospecific soils performed alike once the effect had been controlled for light and nutrients, or if sterilisation left the mortality gap intact, the pathogen account would be wrong and something abiotic — allelopathy, nutrient depletion, shade — would have to carry it instead.

• How mast seeding attacks the same enemies from the other direction — flooding them in one year rather than escaping them in space.
• Whether temperate forests show the same plant-soil feedback, and why it looks weaker there.

Every term the collection defines is gathered in the glossary.