Exponential Population Growth
In 1859, a farmer named Thomas Austin released 24 wild rabbits onto his estate. It was in Victoria, Australia, and done "for sport." Within six years, he could shoot 14,253 rabbits on his land. This was in a single season. By the 1920s, an estimated 10 billion rabbits occupied the continent. The population had spread across nearly the entire southern half of Australia. No new rabbits were being imported. The entire explosion came from unrestrained reproduction of a handful of founders. This is the signature of exponential growth. A population increases by a constant proportion of itself each interval. The absolute number added accelerates, even though the underlying rate per individual never changes. It is the single most important baseline model in Population Ecology. Every real population grows exponentially whenever resources are effectively unlimited. This applies to bacteria in a fresh culture, a reintroduced endangered species, a new pathogen strain, and the human species over most of its history. It happens whenever no other species checks the population. Understanding this model is not just a mathematical exercise. It explains why small numbers of invasive species become catastrophic within a few years. It explains why an epidemic can go from a handful of cases to millions in weeks. And it explains why the human population, having only reached one billion around 1804, added its most recent billion in scarcely over a decade. This article builds the geometric (discrete-time) and continuous exponential models from first principles. It defines the intrinsic rate of increase r, and derives doubling time. It walks through the classic empirical demonstrations that establish exponential growth as a genuine, testable biological phenomenon. These range from Gause's Paramecium cultures to real-world invasions, showing the phenomenon is not an abstraction. It builds directly on population definition and age structure. It also sets up the necessary contrast with Logistic Growth and Carrying Capacity, where resource limits finally bend the curve.
- Exponential growth occurs when births and deaths are constant proportions of population size, independent of density — the per-capita growth rate stays fixed while absolute numbers accelerate.
- The discrete (geometric) model uses a finite rate of increase, λ = Nₜ₊₁/Nₜ; the continuous model uses the intrinsic rate of increase, r = b − d, in dN/dt = rN.
- The closed-form solution Nₜ = N₀eʳᵗ produces a J-shaped curve; doubling time t₂ = ln(2)/r ≈ 0.693/r.
- No real population sustains exponential growth indefinitely — resources, predators, and disease eventually impose density dependence (foreshadowing the logistic model).
- Classic evidence comes from microbial cultures (Gause's Paramecium), controlled invasions (Australian rabbits, ring-necked pheasants on Protection Island), and reintroduction programs.
- r is a summary statistic of a population's entire life table (age-specific survival and fecundity), not an independent, freely chosen parameter.
Exponential growth means the population's per-capita growth rate is constant, so absolute growth accelerates purely because the base (N) keeps getting larger.
The core idea
A population grows exponentially when its rate of increase per individual — not the number of individuals added — stays constant over time. Formally, in continuous time this is written as
dN/dt = rN where N is population size, t is time, and r is the intrinsic rate of increase (also called the intrinsic rate of natural increase or the Malthusian parameter), defined as the instantaneous per-capita birth rate b minus the instantaneous per-capita death rate d: r = b − d. Crucially, r has units of 1/time (e.g., per day, per year) and is a property of a population's age-specific survival and fecundity schedule under given environmental conditions — it is not a fixed constant of a species, but it can be treated as constant when resources are unlimited, mates are freely available, and no predators, competitors, or diseases regulate the population. In discrete time — appropriate for organisms that breed in seasonal pulses (many insects, annual plants, temperate birds) — the equivalent model is geometric growth:
N_{t+1} = λ N_t, so N_t = N_0 λ^t where λ (lambda), the finite rate of increase, is the ratio of population size in one generation (or time step) to the size in the previous one. λ > 1 means growth, λ = 1 means stasis, λ < 1 means decline. The continuous and discrete parameters are related by λ = eʳ (equivalently r = ln λ), so the two frameworks describe the same underlying phenomenon at different time resolutions — geometric growth is what you get if you sample the continuous exponential curve only once per generation. The defining biological feature of both models is density independence: the per-capita rate (r or λ) does not change as N changes. This is an idealization, valid only over the range of population sizes and time spans where resources are not yet limiting — which is why the model is most accurate for founder populations, recovering populations far below their historical numbers, and organisms colonizing genuinely new, resource-rich habitat.
Mechanism
Step 1 — Start from individual vital rates. Every individual in the population has some probability of surviving to reproduce and some expected number of offspring, summarized in a life table (age-specific survivorship lₓ and fecundity mₓ, covered under Age Structure and Population Pyramids). If a population's environment offers unlimited food, space, and mates, and no predators or pathogens depress survival, these individual rates do not depend on how many other individuals are present.
Step 2 — Aggregate to per-capita birth and death rates. Averaging across all individuals gives an instantaneous per-capita birth rate b (offspring produced per individual per unit time) and per-capita death rate d (probability of dying per individual per unit time). Both are constants under the density-independent assumption.
Step 3 — Write the population-level balance equation. The total number of new individuals added to the population per unit time is bN (each of N individuals contributes at rate b), and the total lost is dN. The net rate of change is therefore: dN/dt = bN − dN = (b − d)N = rN
Step 4 — Solve the differential equation. Separating variables (dN/N = r dt) and integrating from time 0 to time t gives the exponential solution:
N_t = N₀ e^{rt} This is a J-shaped curve: population size increases slowly at first (in absolute terms) because N₀ is small, then increases ever more steeply because the same proportional rate r is now acting on a much larger base. The curve never bends — that is precisely what makes it "unregulated."
Step 5 — Doubling time. Because the proportional rate is fixed, the time required for the population to double is also fixed, independent of the starting size. Set N_t = 2N₀ and solve: 2N₀ = N₀ e^{rt₂} → ln 2 = rt₂ → t₂ = ln(2)/r ≈ 0.693/r A population growing at r = 0.10 per year (10% per year) doubles roughly every 6.93 years; one growing at r = 0.02 per year doubles roughly every 35 years. This single formula explains why small changes in r — driven by, say, a slight increase in average family size or a slight drop in infant mortality — produce dramatically different long-run outcomes.
Step 6 — The discrete analogue. For organisms with distinct, non-overlapping generations, the same logic proceeds generation-by-generation: N_{t+1} = λN_t, giving N_t = N_0 λ^t. Because λ = e^r, a population with r = 0.10 per generation has λ ≈ 1.105 — each generation is about 10.5% larger than the last, compounding multiplicatively exactly as compound interest does in finance.
Step 7 — Why it always eventually fails. Exponential growth requires that b and d remain constant regardless of N. As N grows, individuals increasingly compete for the same finite food, space, mates, and are exposed to more predators and faster-spreading disease — b falls, d rises, and the realized per-capita rate declines. This is the transition point toward the logistic model; exponential growth is best understood as the initial phase of growth for any population released into a new or recovering environment, valid only until density-dependent feedbacks engage.
The evidence
Gause's Paramecium microcosms (1934). The Russian ecologist Georgii Gause grew the ciliate protozoan Paramecium caudatum in tubes of fixed-volume oat-medium culture, counting individuals daily. In the first several days after inoculating a few hundred cells into fresh medium, population size increased in a textbook J-shaped curve, with the per-capita division rate essentially constant — exactly matching N_t = N₀e^{rt}. Only after several days, as bacteria (the Paramecium's food) became depleted and metabolic wastes accumulated, did the growth rate fall and the curve bend over into a plateau. This single experiment is historically important for two reasons: it empirically confirmed the exponential phase as a real, reproducible regime (not merely an equation), and it simultaneously demonstrated its limit, motivating the logistic model that follows exponential growth in the population-ecology curriculum. Gause could estimate r directly from the slope of ln(N) versus time during the early linear-log phase, a technique still used in microbiology today (the "log phase" of a bacterial growth curve is precisely this regime).
The Australian rabbit invasion (1859 onward). Thomas Austin released 24 European rabbits (Oryctolagus cuniculus) on his property near Geelong, Victoria. Australia had no native lagomorph predators calibrated to rabbits, unlimited pasture, and a climate allowing year-round breeding. Rabbit numbers increased so fast that a self-generated "rabbit-proof fence" over 3,250 km long was built across Western Australia by 1907 in an attempt to halt the advancing front — it failed, because rabbits had already crossed by the time construction finished. Contemporary and retrospective estimates put the continental population at roughly 10 billion by the 1920s, spreading at an average front velocity of about 110 km/year — one of the fastest documented mammal invasions, and now a standard case study of what happens when b stays high and d stays low with no density-dependent brake for decades. The population only came under partial control after the deliberate introduction of the myxoma virus in 1950, which suddenly (and artificially) elevated d — direct proof that it was the lack of mortality-imposing checks, not some innate reproductive ceiling, that had sustained the exponential phase.
Ring-necked pheasants on Protection Island (1937). A cleaner, quasi-controlled ecological experiment: eight ring-necked pheasants (2 males, 6 females) were introduced to Protection Island off the coast of Washington State, an island with no pheasant predators and abundant food. By 1938 the population had reached 1,325 birds. Fitting the data gives an intrinsic rate of increase of roughly r ≈ 1.6–2.0 per year, consistent with the calculated doubling time of only a few months — one of the cleanest field demonstrations that a small founder population, released into a resource-rich, predator-free environment, follows the exponential model closely before density dependence (in this case, food and space limitation) set in within a few years.
Human population data as a natural experiment. Global human population size follows a curve well approximated by exponential growth over long stretches punctuated by shifts in r — for instance, r rose sharply after the mid-20th-century decline in death rates (the "epidemiological transition," driven by antibiotics, vaccination, and sanitation) even before birth rates fell in response, producing the fastest doubling time in human history (roughly 37 years around 1970, corresponding to r ≈ 0.019/year). This is treated in depth under Human Population Growth, but it stands as a large-scale, non-manipulated confirmation that r is a joint outcome of birth and death rates, and that changes in either one — not some fixed species-level ceiling — drive changes in growth rate.
Worked examples
Example 1 — Bacterial culture, generation-time and r. A strain of E. coli is inoculated into fresh nutrient broth at a density of 500 cells/mL. Under optimal conditions it divides every 24 minutes (its generation time), and division is essentially synchronous enough to treat as continuous exponential growth over short intervals. Step 1: Convert generation time to r. Doubling time t₂ = 24 min = 0.4 h, so: r = ln(2)/t₂ = 0.693/0.4 h = 1.73 per hour Step 2: Predict density after 5 hours. N_t = N₀e^{rt} = 500 × e^{(1.73)(5)} = 500 × e^{8.65} ≈ 500 × 5,720 ≈ 2.86 × 10⁶ cells/mL Step 3: Sanity-check via doublings. In 5 hours at a 24-minute doubling time, there are 5h / 0.4h = 12.5 doublings. 500 × 2^{12.5} ≈ 500 × 5,793 ≈ 2.90 × 10⁶ cells/mL — consistent (small difference from rounding). This is exactly the calculation microbiologists perform to predict when a culture will reach a target optical density, and it is the same arithmetic underlying predictions of early-stage epidemic spread, where "generation time" becomes the serial interval between infections.
Example 2 — Reintroduced population with asymmetric survival and fecundity (life-table route to r). Conservation biologists reintroduce 40 female sea otters (Enhydra lutris) to a previously extirpated coastal stretch with abundant, unexploited shellfish beds and no immediate predation pressure. Suppose age-specific data give an average per-capita birth rate of b = 0.28 female offspring per female per year, and a per-capita death rate of d = 0.09 per year (favorable because food is abundant and predators are currently sparse). Step 1: Compute r. r = b − d = 0.28 − 0.09 = 0.19 per year Step 2: Compute doubling time. t₂ = ln(2)/r = 0.693/0.19 ≈ 3.6 years Step 3: Project population size after 10 years assuming the exponential (density-independent) phase holds throughout — a reasonable first approximation while the population remains far below the habitat's carrying capacity. N₁₀ = 40 × e^{(0.19)(10)} = 40 × e^{1.9} ≈ 40 × 6.69 ≈ 268 female otters Step 4: Contrast with a naïve linear extrapolation. If a manager mistakenly assumed constant absolute annual increase (say, the observed first-year increase of 40 × 0.19 ≈ 7.6 individuals per year applied linearly), they would predict only 40 + 7.6×10 ≈ 116 otters after 10 years — under-forecasting the true exponential trajectory by more than a factor of two. This is precisely the kind of error the exponential model is designed to prevent: intuitive linear thinking chronically underestimates compounding growth, whether in ecology, epidemiology, or finance. Real reintroduction programs (e.g., otters reintroduced to Washington State in 1969–1970) have shown exactly this pattern — rapid early exponential-like expansion before leveling off as suitable habitat filled.
Common misconceptions & traps
| Misconception | Why it's wrong |
|---|---|
| "Exponential growth means the population grows by the same number of individuals each period." | That is linear growth. Exponential growth adds the same proportion each period, so the absolute number added increases over time. |
| "r is a fixed biological constant for a species, like a genetic trait." | r depends on the current environment (food, predators, disease, temperature) acting on b and d; the same species can have very different r values in different conditions — it is not a species constant, though textbooks sometimes report a species' "maximum" or "intrinsic" r under ideal conditions as a reference point. |
| "Exponential growth can continue indefinitely if the species is prolific enough." | No population sustains it indefinitely; resource limitation, predation, and disease always eventually impose density dependence — exponential growth describes an initial phase, not an endpoint. |
| "λ and r are two unrelated growth-rate ideas — pick whichever textbook uses." | They describe the same phenomenon on different timescales: λ = e^r exactly relates the discrete finite rate to the continuous intrinsic rate. |
| "A population with r close to zero is 'stable' in the sense of being at carrying capacity." | r ≈ 0 only means births balance deaths at that instant under current density-independent conditions; it says nothing about whether density-dependent regulation is occurring at all — a population well below carrying capacity can transiently have r ≈ 0 due to, say, a cold snap, with no bearing on K. |
The scientist behind it
The mathematical seed of exponential population growth was planted by Thomas Robert Malthus, whose 1798 An Essay on the Principle of Population argued that populations, left unchecked, increase geometrically while food supply increases only arithmetically — a mismatch he believed would inevitably produce famine, disease, and war as checks on growth. Malthus was a clergyman and political economist, not a biologist, and his purposes were social and political — he was arguing against utopian schemes for perpetual social improvement — but his core mathematical insight (that reproduction compounds multiplicatively) is exactly the r of the modern exponential model. Notably, Charles Darwin explicitly credited reading Malthus in 1838 as the trigger that let him see how limited resources would force a "struggle for existence," directly seeding the concept of natural selection. The rigorous ecological formalization came later, especially through Raymond Pearl and Alfred Lotka in the early 20th century, who fit growth curves to real population data (Pearl's fruit-fly cultures, national census data) and formalized r as the intrinsic rate of natural increase computable from an age-specific life table — connecting the abstract Malthusian parameter to measurable birth and survival schedules, the same life-table logic used today.
Origins · 1 minThomas Robert Malthus and how this was discovered
An English cleric and scholar whose Essay on the Principle of Population (1798) first formalized the idea that populations grow multiplicatively while resources grow additively — the founding insight, later mathematized by ecologists, behind the intrinsic rate of increase r and behind Darwin's theory of natural selection.
Where this matters today
Epidemiology and public health. Early-stage outbreak growth (before susceptible depletion or intervention) is modeled as exponential, with the basic reproduction number R₀ playing a role directly analogous to λ; estimating the epidemic's growth rate and doubling time from early case counts — exactly the calculation in Example 1 — is how health agencies forecast hospital demand and decide when interventions must begin (COVID-19 case-count doubling times of 3–4 days in early outbreaks translated directly into overwhelmed ICU capacity forecasts using this model).
Conservation biology. Reintroduction and recovery programs (California condors, gray wolves in Yellowstone, sea otters) explicitly track whether a small founder population is achieving the exponential growth predicted from its life table; a lower-than-expected r signals unrecognized mortality, poor habitat quality, or Allee effects, prompting intervention before a population's fate is sealed. Conversely, unchecked exponential growth is the signature of a runaway population, whether desirable (a recovering endangered species) or catastrophic (an invasive species establishing a beachhead).
Invasive species management and biosecurity. Detecting an invader while it is still in its exponential phase — few individuals, but with r unmistakably positive — is vastly cheaper than eradication after it has spread across a continent, as the rabbit and cane toad invasions in Australia illustrate; biosecurity agencies use estimated r and doubling times to prioritize which new detections warrant emergency eradication funding.
Agriculture and pest control. Aphid, locust, and pathogen populations expand exponentially between generations under favorable weather; agronomists use estimated λ per generation to time pesticide application to the early exponential window, when a small absolute reduction in population still translates into a large delay in reaching damaging densities, rather than waiting until numbers are already enormous.
Biotechnology and industrial microbiology. Fermentation processes (antibiotic production, biofuel synthesis, recombinant protein expression) are timed and scaled using exactly the log-phase exponential model from Example 1, choosing when to harvest a culture, feed additional nutrients, or switch bioreactors based on measured r and predicted doubling time.
Ask the tutor: How does the exponential model's assumption of a stable age distribution break down right after a population crash, and what does growth look like before it re-stabilizes?
What is the difference between the geometric and exponential growth models?
Both describe density-independent, unregulated growth, but they differ in timescale treatment. The geometric model (N_{t+1} = λN_t) applies to organisms with discrete, non-overlapping generations — annual plants, many insects, seasonally breeding birds — and tracks population size generation by generation. The exponential model (dN/dt = rN, solved as N_t = N₀e^{rt}) applies to organisms with continuous, overlapping generations, like humans or bacteria dividing asynchronously, and tracks instantaneous rates of change. They are mathematically equivalent representations of the same underlying process: λ = e^r converts between them, so a population's growth pattern doesn't depend on which model you use, only which better matches its actual breeding biology.
Why can't exponential growth continue forever in nature?
Exponential growth requires per-capita birth and death rates to stay constant no matter how large the population gets — an assumption that only holds when resources, space, and mates are effectively unlimited relative to current numbers. As a population grows, individuals increasingly compete for the same finite food and space, predators and parasites find prey more easily, and waste or disease accumulates faster, all of which raise the death rate and/or lower the birth rate. This density-dependent feedback is precisely what the logistic growth model adds to the exponential one, producing an S-shaped curve that levels off at the environment's carrying capacity (K) instead of climbing forever.
Is r the same thing as the population growth rate you'd calculate from a census?
Not exactly. The intrinsic rate of increase r is technically defined under a specified, often idealized, set of environmental conditions (frequently "unlimited resources, no predators or competitors") and is calculated formally from a population's age-specific life table (survivorship and fecundity schedule). The realized per-capita growth rate observed in a real census at a given moment — sometimes written r' or simply "observed growth rate" — reflects whatever density-dependent and environmental constraints are actually operating. Only when a population is genuinely growing at its maximum, unconstrained rate (as in Gause's early-phase Paramecium cultures or newly founded island populations) does the observed rate closely approximate the true intrinsic r.
Why did the Australian rabbit population grow so explosively compared to rabbits in their native European range?
In their native range, European rabbit populations are held well below their reproductive potential by predators (foxes, weasels, birds of prey), parasites, and diseases that co-evolved with them over long timescales — these factors elevate d and depress realized population growth even though the rabbits' intrinsic reproductive capacity (b) is similarly high. In Australia, none of these regulating factors were present when rabbits arrived in 1859: no evolved predators recognized them as prey, and myxoma virus and rabbit hemorrhagic disease had not yet been introduced. With b unchanged and d artificially low, the realized r approached the theoretical maximum sustainable by rabbit reproductive physiology, producing the extraordinary exponential expansion actually observed.
How do you estimate r from real population count data?
If growth is genuinely exponential, taking the natural log of population size converts the curve N_t = N₀e^{rt} into a straight line: ln(N_t) = ln(N₀) + rt. Plotting ln(population count) against time and fitting a straight line (by linear regression) gives r directly as the slope — this is exactly how Gause estimated Paramecium growth rates and how microbiologists read bacterial "log-phase" growth curves. A genuinely exponential phase shows up as a straight line on this semi-log plot; curvature (a bending slope) signals that density-dependent effects have begun and the exponential assumption no longer holds.
Does a positive r always mean a population is thriving in a healthy, sustainable sense?
No — positive r only means births currently exceed deaths under present conditions; it says nothing about whether that state is sustainable. An invasive species with a large positive r may be depleting resources it will soon exhaust, causing a later crash (as some introduced rabbit and reindeer populations have shown after outstripping food supply). Similarly, a recovering endangered population with positive r is a good sign only if the habitat can support continued growth toward a healthy population size; ecologists always interpret r alongside information about carrying capacity, resource trends, and genetic diversity, not as a standalone verdict on population health.
How does age structure affect whether a population can even show exponential growth?
Exponential growth as derived here assumes a population's age distribution has settled into a stable age distribution, where the proportion of individuals in each age class stays constant even as total N grows — this is what lets a single scalar r (or λ) validly describe the whole population's dynamics. A population with a highly skewed age structure (for example, one dominated by post-reproductive individuals after a mass die-off of juveniles) can show transient growth or decline that doesn't match the simple exponential prediction until the age structure stabilizes. This is why age structure, covered in the prerequisite on age pyramids, is not a separate topic but the underlying scaffolding that makes a single-number r a valid summary of population dynamics at all.