Meridia Insight Science Breakthroughs Knowledge

Five species can obey the rules—and still collapse into a cycle

A five-species ecosystem can have a healthy-looking interior equilibrium and still be pulled toward a boundary cycle where species repeatedly approach

5 species suffice: a stable boundary cycle defeats two sign rules once thought to guarantee permanence.

A system with an interior equilibrium—where every species is present—can still lose species over and over, with some populations shrinking toward extinction before returning. In the five-species Lotka–Volterra system studied here, that is not a transient wobble: an attracting cycle on the boundary pulls in a set of interior trajectories. The result overturns a 2008 conjecture that two simple rules about interaction signs are enough to guarantee permanence.

That word, permanence, has a precise meaning in mathematical ecology. A permanent system eventually keeps every species bounded away from zero and from runaway growth, whatever positive starting population it began with. The paper’s counterexample exposes a gap between a system’s sign pattern—who helps or harms whom—and the long-term consequences of those interactions. The same qualitative rules that had seemed to rule out collapse do not.

The finding is a theorem about a mathematical model, not evidence that a particular ecosystem is collapsing. But its wider lesson matters anywhere we try to infer system-wide resilience from a diagram of pairwise effects: interaction signs can constrain outcomes without fully determining them. The arrangement of those links, the size of their effects, and the nonlinear sequence they trigger can matter just as much.

The Science

Beldiman and Halanay examine the Lotka–Volterra equations, a classic model in which each species’ growth rate depends on its own abundance and on the abundances of the other species. In the version considered, species abundance changes according to

Here, is the abundance of species , is its growth rate when alone, and describes the effect of species on species . A positive interaction coefficient means a benefit; a negative one means harm. The diagonal coefficients are negative, so each species is limited by its own increasing abundance.

In 2008, Hofbauer, Kon and Saito established two necessary conditions for a sign pattern to guarantee permanence under every choice of interaction strengths with those signs. First, every directed interaction cycle of length two or more must have a nonpositive product of its coefficients. Second, a cycle with a negative product must contain exactly one negative link. They proved the conditions sufficient for systems with up to three species and conjectured they would suffice more generally (Hofbauer et al., 2008).

The authors produce a five-species counterexample. Its interaction matrix has negative self-limitation on the diagonal and just seven nonzero interactions between different species. It meets both sign conditions. Yet for one explicitly specified set of growth rates, it has a positive interior equilibrium and also a stable boundary cycle that attracts an open set of initial conditions inside the system. An interior equilibrium is a state at which every species has positive abundance and all populations are stationary. Its existence does not guarantee that all populations starting nearby—or elsewhere—will approach it.

The network combines two motifs. Species 1 is a predator that consumes species 2 and 3; it can also grow in the absence of those prey. Separately, three one-way interactions create a loop: species 4 benefits species 2, species 2 benefits species 5, and species 5 harms species 4. None of these three effects is reciprocated. The loop has two positive links and one negative link, satisfying the conjecture’s cycle rule.

The boundary cycle visits four equilibria, each with a different combination of species present: . Species 1 remains throughout. At each stop, one absent species can invade; its arrival helps rearrange the community and sets up the next transition. The paper’s Figure 1 plots the logarithms of species abundances along one interior trajectory. Species 1 persists, while the others repeatedly become scarce and return; the successive excursions move closer to the boundary as the orbit approaches the cycle.

To establish the result, the authors verify the four connections between equilibria, then calculate whether perturbations near the cycle shrink or grow. Three connections follow from a Lyapunov argument, a way of showing that a quantity consistently decreases as the system approaches a target. The fourth lies in a face of the system containing the one-way loop, where that standard stability argument does not apply. For this case, the authors prove a lemma using an invasion graph: a directed graph describing which equilibrium states can replace which others through species invasion. They then show the invasion graph has no cycles and use it to establish the needed connection.

Finally, they compute the cycle’s stability. In technical terms, a characteristic matrix records how species absent at each equilibrium fare if they try to invade. A standard criterion shows the boundary cycle attracts nearby trajectories, including some that begin with all five species present. The proof uses exact rational arithmetic; the authors say the finite calculations can be checked by hand. Their numerical illustration tested 200 random interior starting conditions, all of which approached the cycle—but the theorem, not that experiment, establishes the counterexample.

What They Found

The core result is direct: conditions (C1) and (C2) do not guarantee qualitative permanence for five or more species. “Qualitative” means the guarantee must hold for every matrix with the same pattern of positive, negative, and zero interactions, and for growth rates that permit an interior equilibrium. The paper supplies a particular matrix and growth vector that defeat that guarantee.

Five species overturn a conjecture that held through three

The paper gives a counterexample for five species; the referenced result established sufficiency for systems with at most three. Four species remains an open question.

Five species overturn a conjecture that held through three
LabelValue
At most 3: proven sufficient3 species
Counterexample demonstrated5 species

The cycle is not a story of a single species disappearing and staying gone. Instead, the community moves through a sequence of invasions and losses. Species 4 invades the predator–prey pair of species 1 and 3. Species 2, helped by 4, arrives and displaces species 3. Species 5, helped by 2, then drives species 4 out. Without 4, species 3 returns, while 2 and 5 are lost. Then the sequence can begin again.

The mathematical distinction between “not permanent” and “all species go extinct” matters. The system has an interior equilibrium, but an open set of interior starting points is drawn toward a boundary cycle. Along those trajectories, some species’ abundances approach arbitrarily close to zero during repeated excursions. The paper does not claim that every initial condition does so, nor that all five species vanish together. The result is that the permanence guarantee fails.

The counterexample is also sparse: it needs only seven nonzero off-diagonal entries. Its five species are the minimum demonstrated here to break the conjecture, while the authors’ earlier theoretical context establishes sufficiency for up to three. They extend the counterexample to every larger system by adding species that do not interact with the original five and that settle at positive abundance. The problematic cycle remains, so merely increasing species richness does not repair the failure.

The counterexample’s four successive invaders

Positive invasion rates at each equilibrium identify the species that can enter next along the boundary cycle.

The counterexample’s four successive invaders
LabelValue
Species 4 at F₁₃1
Species 2 at F₁₃₄10.3333333333
Species 5 at F₁₂₄37
Species 3 at F₁₂₅1.75

The four cycle equilibria make the invasion sequence concrete. At , species 4 has an invasion rate of 1; at , species 2’s rate is ; at , species 5’s rate is 37; and at , species 3’s rate is . Each is positive, marking the next invader in the cycle. For example, the negative rate for species 4 at is , consistent with its loss at that stage.

Reported off-diagonal links in the five-species network

The interaction matrix contains seven nonzero effects between different species; the diagonal self-limitation terms are not included.

Reported off-diagonal links in the five-species network
LabelValue
Nonzero off-diagonal interactions7 links
Species in counterexample5 links

There is an important qualification to the paper’s generality. This is a counterexample to a universal sufficiency claim, not a prediction that every five-species food web with a one-way loop will fail to persist. The authors report that the example is robust to small changes in the matrix and growth rates: nearby systems retain an interior equilibrium and a stable cycle with the same supports. That strengthens the result beyond a finely tuned mathematical curiosity, while still leaving open how common such conditions are in real ecological communities.

Why This Changes Things

The conjecture was appealing because it asked whether the signs of interactions alone could certify persistence. A sign pattern is a coarse but often useful summary: it says who benefits or harms whom without requiring precise estimates of effect sizes. If two simple cycle rules were sufficient, ecologists might be able to draw a network and infer a broad resilience property from its structure.

This paper shows that the inference can fail. The sign rules still matter; they remain necessary. But in a five-species network, they do not capture every route through which populations can take turns opening opportunities for invaders and destabilizing previous combinations. A positive equilibrium can coexist with a separate attracting route toward the edge of ecological possibility.

The surprise is not simply that “complexity matters.” The authors identify a relatively small mechanism. A one-way three-species feedback loop, with two benefits and one harm, is combined with a predator of two prey. The predator and its prey create a persistent background, while the one-way loop changes which of the other species can establish. The community’s future depends on the sequence of states, not only on whether the product around each cycle passes a sign test.

That distinction is useful far beyond this particular model. In ecology, researchers often infer resilience from food-web structure, interaction networks, or local responses to perturbations. The result is a caution against treating a network’s qualitative wiring as a complete stability certificate. Quantitative strengths, growth rates, and the dynamics on particular subsets of species can make a difference. In engineering, epidemiology, or economics, the analogous warning is that individually acceptable feedback loops may combine into a system-level failure mode that only appears when the components interact over time.

The model is deliberately idealized. Real ecosystems have changing environments, spatial structure, stochastic shocks, adaptive behavior, and interactions that may not remain fixed. Those features could alter the dynamics in either direction. The paper does not measure the frequency of this mechanism in nature, and its mathematical result cannot tell us whether a particular food web faces it. Its contribution is more foundational: it tells researchers that a broad class of sign-based proofs cannot establish permanence in general.

The authors also show why mathematical proof can reveal something simulations alone might miss. Their 200 random starting points all moved toward the cycle, but finite numerical runs cannot establish attraction for an open set, much less rule out a hidden equilibrium or a slower alternative. The proof verifies the chain of invasions and its stability, turning a suggestive pattern into a counterexample. At the same time, the accessible ecology of the example makes the abstract result easier to inspect: a loop of one-sided effects interacts with a predator–prey pair to generate a repeatable succession.

What’s Next

The conjecture is settled as stated for five or more species, but the boundaries of the failure remain open. The authors ask whether the conditions may still guarantee permanence for four species, including cases with nongeneric growth rates. Their computer experiments suggest that, for four species and nonzero boundary invasion rates, the invasion graph may always be acyclic under the two sign conditions. That is a promising clue, not a proof.

A second question is what extra assumptions could restore sufficiency in larger systems. The example suggests possible restrictions on one-way loops: perhaps excluding them, or requiring some of their effects to be reciprocated, could help. But these are conjectural directions. Any proposed rule would need to distinguish genuinely protective structure from patterns that merely look reassuring.

A third task is ecological rather than purely combinatorial: determine for which growth rates a given interaction pattern loses permanence. The sign network alone is not enough, but the authors’ analysis shows how equilibrium invasion rates and invasion graphs can sharpen the picture. Better empirical estimates of interaction strengths and demographic rates could help connect such mathematical criteria to real food webs, though uncertainty in those estimates remains a serious challenge.

The broader payoff is a more honest theory of resilience. A system can pass every known sign-based test and still contain a route toward repeated near-extinction. Finding that route does not make ecological prediction hopeless; it identifies what the prediction must include. The next generation of permanence criteria will need to account not just for who affects whom, but for how the system moves from one community state to another—and whether those transitions eventually bring species back, or keep pushing them toward the edge.

Comments (0)

No comments yet. Be the first to share your thoughts.