The Lotka-Volterra Competition Model
Two species of flour beetle, Tribolium confusum and Tribolium castaneum, are placed together in a jar of flour. Sometimes confusum wins, sometimes castaneum wins, and occasionally both persist. Why does the outcome flip? Can we predict it before the experiment even starts? Interspecific Competition is a qualitative idea. Two species using the same limited resource suppress each other's growth. The Lotka-Volterra competition model turns that idea into arithmetic. It takes the logistic growth equation for a single population. Then it adds a term for how much a competitor's presence subtracts from your own carrying capacity. It does this symmetrically for both species. The result is a pair of coupled differential equations. Their long-term behavior can be stable coexistence, exclusion of one species, or a knife-edge outcome. That knife-edge outcome depends on starting numbers. All of this can be read directly off a simple graph. This model matters because it converts the vague statement "competition matters" into testable, falsifiable predictions about population trajectories. It also gives ecologists their first quantitative handle on the Competitive Exclusion Principle. The model shows precisely under what mathematical conditions exclusion is guaranteed. It also shows under what conditions exclusion is not guaranteed. This is exactly the loophole through which Resource Partitioning later enters the story. The same graphical machinery — zero-growth isoclines, vector fields, phase planes — will reappear almost unchanged in the Lotka-Volterra Predator-Prey Model. This article builds the model from the logistic equation upward. It derives the isoclines and shows all four qualitative outcomes graphically. It walks through the Gause and Park experiments that gave the model empirical teeth. Finally, it works two numerical examples so the abstract inequalities become concrete numbers.
- The model extends logistic growth to two species, each suppressing the other via competition coefficients α₁₂ and α₂₁.
- Zero-growth isoclines for each species are straight lines in N₁–N₂ space; their crossing pattern determines the outcome.
- Four qualitative outcomes exist: species 1 always wins, species 2 always wins, stable coexistence, or an unstable outcome decided by starting densities.
- Coexistence requires each species to limit itself more than it limits the other (intraspecific > interspecific competition).
- Gause's Paramecium experiments and Park's Tribolium experiments gave direct empirical tests of the model's predicted outcomes.
- The model is a simplifying caricature — no time lags, no age structure, constant coefficients — but it remains the conceptual foundation for modern coexistence theory.
Two competing species can coexist indefinitely only if each species inhibits its own population growth more than it inhibits its competitor's — otherwise one species drives the other to extinction.
The core idea
The Lotka-Volterra competition model, developed independently by Alfred J. Lotka (1925) and Vito Volterra (1926), describes how two species sharing a limited resource affect each other's population growth over time. It starts from the logistic growth equation for a single species, which caps growth as population size N approaches a carrying capacity K: dN/dt = rN(K − N)/K To model competition, each species' equation gains an extra term representing how much the other species' population subtracts from the first species' available "room to grow." This subtraction is scaled by a competition coefficient, α, which converts one competitor's individuals into equivalent units of the other species. The paired equations are: dN₁/dt = r₁N₁(K₁ − N₁ − α₁₂N₂)/K₁ dN₂/dt = r₂N₂(K₂ − N₂ − α₂₁N₁)/K₂ Here α₁₂ is the per-capita effect of species 2 on species 1 (how much one individual of species 2 "counts as" in species-1 currency), and α₂₁ is the reciprocal effect. If α₁₂ = 0.5, each individual of species 2 depresses species 1's growth exactly as much as half an individual of species 1 would. Critically, α₁₂ and α₂₁ need not be equal — competition is rarely symmetric in nature, since the two species differ in resource requirements, body size, and efficiency. The outcome of the interaction depends entirely on how these four numbers — K₁, K₂, α₁₂, α₂₁ — relate to one another. The model's genius is that this comparison can be visualized on a single two-dimensional graph, the N₁–N₂ phase plane, using zero-growth isoclines: lines along which one species' population is momentarily unchanging (dN/dt = 0).
Mechanism
Step 1 — Derive species 1's isocline. Set dN₁/dt = 0. Since r₁, N₁, and K₁ are nonzero in general, this requires: K₁ − N₁ − α₁₂N₂ = 0 → N₁ = K₁ − α₁₂N₂ This is a straight line in the N₁ (x-axis) vs N₂ (y-axis) plane. It intercepts the N₁-axis at N₁ = K₁ (when N₂ = 0, species 1 alone sits at its own carrying capacity) and intercepts the N₂-axis at N₂ = K₁/α₁₂ (the number of species-2 individuals whose competitive pressure alone would suffice to halt species 1's growth). Everywhere below-left of this line, species 1 still has "room," so N₁ increases; everywhere above-right, resources are oversubscribed and N₁ decreases.
Step 2 — Derive species 2's isocline. By the same logic, setting dN₂/dt = 0 gives: N₂ = K₂ − α₂₁N₁ This line intercepts the N₂-axis at K₂ and the N₁-axis at K₂/α₂₁. Below this line N₂ increases; above it, N₂ decreases.
Step 3 — Superimpose the two isoclines. Because both are straight lines on the same axes, they can cross zero times or once. There are exactly four geometrically distinct arrangements, corresponding to the four biological outcomes. The position of each line's intercepts — whether K₁ > K₂/α₂₁ or K₁ < K₂/α₂₁, and the analogous comparison for species 2 — decides which arrangement occurs.
Step 4 — Read the vector field. In each of the four regions carved out by the two isoclines, both dN₁/dt and dN₂/dt have definite signs (increasing or decreasing), so an arrow can be drawn showing the joint trajectory of (N₁, N₂). Following these arrows from any starting point reveals where the system ends up.
Outcome A — Species 1 always wins. Species 1's isocline lies entirely outside (further out than) species 2's isocline: K₁ > K₂/α₂₁ and K₂ < K₁/α₁₂. Species 1 can tolerate more of species 2's competition than species 2 can tolerate of species 1's. Every trajectory converges to N₁ = K₁, N₂ = 0. This is competitive exclusion, deterministic and starting-point independent.
Outcome B — Species 2 always wins. The mirror image: species 2's isocline dominates entirely. Every trajectory ends at N₂ = K₂, N₁ = 0.
Outcome C — Stable coexistence. The isoclines cross, with each species' isocline intercept on its own axis (K) exceeding the other species' isocline intercept on that same axis expressed in its units — concretely, K₁ < K₂/α₂₁ and K₂ < K₁/α₁₂. Biologically: intraspecific competition exceeds interspecific competition for both species. Every trajectory spirals or curves into the interior crossing point, a stable equilibrium where both species persist indefinitely.
Outcome D — Unstable equilibrium, winner depends on initial densities. The isoclines cross the opposite way: K₁ > K₂/α₂₁ and K₂ > K₁/α₁₂. Interspecific competition exceeds intraspecific competition for both species. The interior crossing point exists mathematically but is unstable — any small perturbation pushes the system toward one axis or the other. Whichever species starts with a numerical or positional advantage wins; this is sometimes called a founder-control outcome.
Step 5 — Interpret the biology behind the inequalities. The condition for stable coexistence, written out fully, is: K₁/α₁₂ > K₂ and K₂/α₂₁ > K₁ which rearranges to the memorable statement: α₁₂α₂₁ < 1 together with each species being limited more by its own density than by the other's. Ecologically this is the mathematical signature of niche differentiation — the two species must not be perfect substitutes for one another's resource. If α₁₂ = α₂₁ = 1 (species are ecologically identical, competing exactly as strongly with each other as with themselves), coexistence collapses to a knife-edge case, and the Competitive Exclusion Principle emerges as the generic, structurally stable outcome unless niches diverge.
The evidence
Gause's Paramecium experiments (1934). The Russian ecologist Georgii Gause was the first to test the model empirically, using two ciliate protozoan species, Paramecium aurelia and Paramecium caudatum, grown in test tubes with a constant bacterial food supply. Grown alone, each species reached its own stable carrying capacity following a logistic curve, exactly as the model's single-species term predicts. Grown together, P. aurelia — the faster-reproducing, more efficient competitor for the shared bacterial food — consistently drove P. caudatum to extinction within about 16-20 days, in every replicate. This matched Outcome A/B of the model: deterministic exclusion, independent of starting ratios. Gause also showed that when he used a different pair — P. caudatum and P. bursaria — the two species partitioned the tube spatially (bursaria concentrated at the bottom feeding on yeast, caudatum higher up), and both persisted, matching Outcome C. Gause's results are why the exclusion principle is sometimes called "Gause's Principle."
Park's Tribolium experiments (1948-1957). Thomas Park at the University of Chicago ran a much longer and more systematic program using two flour beetle species, Tribolium confusum and Tribolium castaneum, competing for flour as both food and habitat. Park varied temperature and humidity across six combinations (hot-dry, hot-moist, temperate-dry, temperate-moist, cold-dry, cold-moist) and ran dozens of replicate jars for each combination, always started from the same initial numbers. The outcome was strikingly climate-dependent:
| Condition | Typical winner | Win frequency |
|---|---|---|
| Hot, moist | T. castaneum | ~100% |
| Cold, dry | T. confusum | ~100% |
| Temperate, moist (intermediate) | either species | roughly 30:70 split, stochastic |
In the intermediate conditions, replicate jars split unpredictably between the two species — sometimes confusum won, sometimes castaneum won, from what appeared to be identical starting conditions. This variability was a direct empirical demonstration of Outcome D, the unstable equilibrium: when competition coefficients are close to the exclusion boundary, chance events (which eggs happen to be eaten by cannibalistic adults, minor demographic fluctuations) tip the outcome, exactly as the unstable interior fixed point predicts. Park's work showed the model's qualitative categories are not just mathematical curiosities — they map onto reproducible, if stochastic, laboratory outcomes, and that environmental conditions shift a real system's parameters between the four regimes.
What these experiments jointly proved: (1) that interspecific competition alone, without predation or disease, is sufficient to drive extinction in a closed system; (2) that the direction of exclusion is not fixed by "which species is generally superior" but depends on environmental context, matching the model's dependence on K and α values that themselves depend on conditions; (3) that stable coexistence and unpredictable, founder-controlled outcomes both occur in nature, exactly as the model's four-way classification predicts, not just the simple story of "the fittest always wins."
Worked examples
Example 1 — Deciding the outcome from coefficients. Two barnacle species compete for space on a rocky shore. Species 1 (Semibalanus): K₁ = 800 individuals per m², α₁₂ = 0.4 Species 2 (Chthamalus): K₂ = 500 individuals per m², α₂₁ = 1.3 Compute the two threshold comparisons. For species 1's isocline: does K₁ exceed K₂/α₂₁? K₂/α₂₁ = 500/1.3 = 384.6 K₁ = 800 > 384.6 → species 1's isocline extends past species 2's isocline on the N₁ axis. For species 2's isocline: does K₂ exceed K₁/α₁₂? K₁/α₁₂ = 800/0.4 = 2000 K₂ = 500 < 2000 → species 2's isocline does not extend past species 1's on the N₂ axis. Both inequalities point the same direction: species 1 dominates on both axes. This is Outcome A — Semibalanus always excludes Chthamalus, regardless of starting densities. (This mirrors Joseph Connell's classic 1961 field experiment showing Semibalanus physically crushes and undercuts Chthamalus on the lower shore — competitive exclusion realized as interference rather than pure resource competition, but the isocline logic is the same.)
Example 2 — Testing for coexistence and finding the equilibrium. Two annual plant species compete for soil nitrogen in an old field. Species 1: K₁ = 1000 plants/m², α₁₂ = 0.5 Species 2: K₂ = 600 plants/m², α₂₁ = 0.3 Check the coexistence condition: K₁/α₁₂ > K₂ and K₂/α₂₁ > K₁. K₁/α₁₂ = 1000/0.5 = 2000 > K₂ = 600 ✓ K₂/α₂₁ = 600/0.3 = 2000 > K₁ = 1000 ✓ Both hold, so this is Outcome C, stable coexistence. Find the equilibrium densities by solving the two isocline equations simultaneously: N₁ = K₁ − α₁₂N₂ = 1000 − 0.5N₂ N₂ = K₂ − α₂₁N₁ = 600 − 0.3N₁ Substitute the first into the second: N₂ = 600 − 0.3(1000 − 0.5N₂) = 600 − 300 + 0.15N₂ = 300 + 0.15N₂ N₂ − 0.15N₂ = 300 → 0.85N₂ = 300 → N₂ = 352.9 N₁ = 1000 − 0.5(352.9) = 1000 − 176.5 = 823.5 So the two species stabilize at roughly 824 plants/m² of species 1 and 353 plants/m² of species 2 — each below its solo carrying capacity (1000 and 600 respectively), because each pays a competitive tax to the other, but neither goes extinct.
Common misconceptions & traps
| Misconception | Why it's wrong |
|---|---|
| "The species with the bigger K always wins." | Outcome depends on the interaction of K₁, K₂, α₁₂, α₂₁ together — a species with smaller K can still exclude a larger-K competitor if its α is large enough (see Example 1, reversed roles are possible). |
| "α₁₂ = α₂₁ always." | Competition coefficients are asymmetric in general; one species can suppress the other far more than the reverse (e.g., allelopathic plants). |
| "Coexistence in the model means the two species are ecologically similar." | The opposite: coexistence requires niche divergence — intraspecific competition must exceed interspecific competition for both species. |
| "An unstable equilibrium (Outcome D) means the model failed." | It is a real, model-predicted category, empirically confirmed by Park's Tribolium results — it predicts stochastic, initial-condition-dependent exclusion, not model breakdown. |
| "This model proves competitive exclusion always happens in nature." | The model is a deliberately simplified caricature (constant K, α, no spatial structure, no time lags); real systems can stably coexist via mechanisms (resource partitioning, disturbance, predation) that violate the model's assumptions. |
The scientist behind it
Origins · 1 minVito Volterra and how this was discovered
Volterra was a mathematical physicist and senator, already famous for work on integral equations and elasticity, when his son-in-law Umberto D'Ancona — a marine biologist — brought him a puzzle: Adriatic fish catch data from before, during, and after World War I showed predatory fish species had become relatively more abundant during the war years, when fishing intensity dropped sharply. Volterra, with no prior training in biology, derived a system of coupled differential equations to explain oscillating predator-prey abundances, publishing his results in 1926. He extended the same coupled-equation logic to competing species shortly after, unaware that the American physical chemist and demographer Alfred J. Lotka had derived essentially identical equations independently in 1925, originally in the context of chemical reaction kinetics and general population dynamics, in his book Elements of Physical Biology. The shared name "Lotka-Volterra" reflects this genuinely independent, near-simultaneous discovery — a case where two very different starting points, one from fisheries statistics and one from chemical kinetics, converged on the same mathematical structure.
Where this matters today
Invasion biology and biocontrol. Predicting whether an introduced species will exclude a native competitor, coexist with it, or fail to establish uses exactly this isocline logic — estimating α's from resource-use overlap lets managers forecast invasion outcomes before releasing a biocontrol agent or before an invasive species becomes entrenched.
Conservation of endangered competitors. When a native species is losing ground to an introduced competitor (e.g., red squirrels displaced by grey squirrels in the UK, or native crayfish displaced by invasive signal crayfish), measuring the asymmetry in competition coefficients tells conservationists whether habitat modification (changing effective K's) could tip the system from exclusion toward coexistence.
Microbial ecology and the gut microbiome. Modern microbiome research uses generalized Lotka-Volterra models with many interacting taxa to predict which bacterial strains will dominate a gut community, informing probiotic and fecal-transplant strategies — the pairwise competition coefficients here are literally α terms estimated from co-culture experiments.
Agricultural weed science and intercropping. Crop scientists use competition coefficients to design intercropping systems (e.g., maize-bean polycultures) that land in the stable-coexistence region of parameter space, maximizing total yield rather than having one crop competitively suppress the other.
Foundation for predator-prey and multi-species theory. The same isocline-and-phase-plane machinery generalizes directly to the Lotka-Volterra Predator-Prey Model and to n-species community matrices, making this two-species case the essential training ground for all of quantitative community ecology.
Ask the tutor: How would adding a time lag to the competition terms change the stability of the coexistence equilibrium?
Why are the zero-growth isoclines straight lines rather than curves?
Each isocline comes from setting one species' growth equation to zero and solving for one population variable in terms of the other. Because the underlying logistic-with-competition equation is linear in N₁ and N₂ once dN/dt is set to zero (the equation K − N₁ − α₁₂N₂ = 0 has no squared or cross terms), the resulting relationship between N₁ and N₂ is a straight line. This is a direct consequence of the model's simplifying assumption that competitive effects add up linearly — each additional competitor subtracts a constant increment of "resource room," regardless of how many competitors are already present. Real competitive effects can be nonlinear, which is one reason more advanced models replace these straight isoclines with curves.
Can the competition coefficients α₁₂ and α₂₁ be measured empirically, or are they purely theoretical?
They can be estimated, though it takes careful experimentation. One common approach grows each species alone to estimate r and K, then grows the two species together at a range of starting densities and fits the joint growth trajectories to the coupled equations to back out α₁₂ and α₂₁. Resource-use overlap (e.g., diet similarity indices, niche overlap measured from stable isotopes) gives an independent, faster proxy, though it correlates only loosely with the dynamically measured α's because coefficients also capture interference competition, not just resource overlap. Both Gause's and Park's experimental programs effectively estimated these coefficients indirectly, by observing which of the four qualitative outcomes occurred.
Does the model account for predators, disease, or environmental fluctuation?
No — the basic two-species Lotka-Volterra competition model is deliberately minimal. It assumes a closed system with only the two competitors, constant carrying capacities, constant competition coefficients, and no time lags in the response of growth rate to density. Predation, parasitism, disturbance, and environmental variability are all excluded, though each has been added in extended versions of the model. This is why the model's predictions (especially deterministic exclusion) sometimes fail to match nature: in real communities, a predator that disproportionately eats the winning competitor, or periodic disturbance that resets both populations, can permit coexistence the bare competition model would rule out.
How does this model relate to the Competitive Exclusion Principle?
The Competitive Exclusion Principle states that two species with identical ecological niches cannot stably coexist indefinitely. The Lotka-Volterra model makes this principle mathematically precise: if the two species are ecologically identical, α₁₂ = α₂₁ = 1, and the coexistence inequality (intraspecific > interspecific competition) collapses to equality, which is a structurally unstable, measure-zero case. Any real-world deviation from perfect identity pushes the system into one of the exclusion outcomes. The model essentially explains why the exclusion principle should be generically true, and simultaneously identifies the exact escape route — sufficient niche differentiation — that permits Resource Partitioning to allow real coexistence.
Why did Park get different winners depending on temperature and humidity?
Temperature and humidity alter each Tribolium species' physiological performance — egg survival, developmental rate, cannibalism rates on eggs and pupae — which changes the effective r, K, and α values for each species under each climate regime. T. castaneum performed better (higher effective K, lower vulnerability) in hot, moist flour, while T. confusum performed better in cool, dry flour. This shows that competition coefficients are not fixed biological constants of a species but depend on the physical environment, meaning the same pair of species can occupy any of the model's four outcome categories depending on abiotic context — a key lesson that competitive outcomes are context-dependent, not fixed rankings of "superior" and "inferior" species.
What happens mathematically at the unstable equilibrium in Outcome D?
The interior point where the two isoclines cross still satisfies dN₁/dt = dN₂/dt = 0, so it is mathematically a fixed point of the system. But a stability analysis (examining the Jacobian matrix's eigenvalues at that point) shows that small perturbations grow rather than decay — trajectories starting exactly at the equilibrium stay there in principle, but any infinitesimal displacement is amplified, driving the system toward one axis or the other. This is a saddle point in phase-plane terms. Biologically, it means any stochastic fluctuation — a chance run of extra births or deaths in one species — determines which species eventually excludes the other, exactly the pattern Park observed in his intermediate-climate Tribolium replicates.
Is the Lotka-Volterra competition model still used in modern research, or has it been superseded?
It remains foundational, both pedagogically and as a building block for more sophisticated frameworks. Modern coexistence theory (e.g., Chesson's framework distinguishing "niche differences" from "fitness differences") is explicitly built on generalizations of the same isocline logic, reframing α₁₂ and α₂₁ in terms of niche overlap and relative competitive ability. Multi-species generalized Lotka-Volterra models are widely used in microbiome and community ecology research today to fit interaction networks from time-series abundance data. The original two-species version is rarely used to make precise field predictions on its own, given its simplifying assumptions, but it remains the conceptual starting point every more advanced model builds from.