Why exponential growth cannot be the whole story
Two guides ago you met dP/dt = k P, the law of exponential growth: a population whose growth rate is proportional to its size. Its solution P(t) = P_0 e^(kt) is clean and, for a short while, often spot-on — a fresh bacterial culture really does double on schedule. But run it forward and it predicts a single colony outweighing the Earth within days. Something is missing, and that something is not a flaw in the calculus. It is a flaw in the assumption: real growth happens in a finite world.
Food runs short, space fills, waste accumulates, predators arrive. As a population swells, each individual competes harder, so the per-capita growth rate — the growth divided by the size, (1/P) dP/dt — should not stay fixed at k. It should fade as the crowd thickens. The exponential model assumes that per-capita rate is the constant k forever; the honest repair is to let it shrink toward zero as P approaches the largest population the environment can sustain.
One braking factor turns a line into an S
The simplest such repair is to make the per-capita rate decline straight-line in P: replace the constant k by k(1 - P/K). Here K is the [[carrying-capacity|carrying capacity]] — the ceiling population. Multiply through by P and you have the [[ode-logistic-equation|logistic equation]], dP/dt = k P (1 - P/K). Read the two factors as a tug of war. When P is tiny, the bracket (1 - P/K) is almost 1, so growth is nearly exponential — plenty of room. As P climbs toward K, the bracket shrinks toward 0, throttling the growth. At P = K it is exactly 0: the engine idles.
P region bracket (1 - P/K) dP/dt sign what happens ------------- ------------------- ----------- ------------------- 0 < P < K between 0 and 1 positive population rises toward K P = K exactly 0 zero steady — no change P > K negative negative population falls back to K
Trace a solution that starts small. Early on it rises faster and faster, just like an exponential — the curve is concave up. But the braking factor is always tightening, and somewhere around P = K/2 the growth rate dP/dt peaks and then begins to ease. The curve bends over, becomes concave down, and glides up to flatten against the ceiling K. That gentle reversal of curvature — speeding up, then slowing down — is the famous S-shaped (sigmoid) curve, the visual signature of logistic growth.
Reading it without solving: equilibria and their stability
Here is the move this whole rung is building toward. You can read everything important about a logistic population without ever solving the equation, because the right-hand side f(P) = k P (1 - P/K) depends on P alone — the equation is autonomous. First find where the population can hold still: set f(P) = 0. The product is zero when P = 0 or P = K, so those two constant solutions are the [[equilibrium-solution|equilibrium solutions]]. (Notice the catch from a previous guide: the constant solution P = 0 is exactly the kind of solution that separating variables can quietly drop, since you'd divide by P. Never lose it.)
Now do a [[sign-analysis-of-f-of-y|sign analysis]] of f(P) between and beyond the equilibria. For 0 < P < K, f is positive, so dP/dt > 0 and the population rises. For P > K, f is negative, so it falls. Picture P on a vertical axis with little arrows: below K, arrows point up; above K, arrows point down. From either side, solutions are driven toward K. That makes K a stable equilibrium (a sink): nudge the population and it returns. The same arrows point away from 0, so P = 0 is unstable (a source): the slightest population takes off and never comes back. This sign-arrow picture is the stability verdict, and it is the seed of the next guide's phase line.
The closed-form solution, and a warning about it
The logistic equation is one of the lucky few nonlinear ODEs that does have a closed form. It is separable, and a partial-fraction split of 1/(P(1 - P/K)) integrates to the sigmoid P(t) = K / (1 + A e^(-kt)), where the constant A is fixed by the starting population via A = (K - P_0)/P_0. You can confirm by inspection that as t grows the exponential dies, A e^(-kt) -> 0, and P -> K: the solution does climb to the carrying capacity, exactly as the sign analysis already promised.
But take the lesson to heart precisely because the formula exists: you did not need it. The qualitative reading — two equilibria, an S-curve rising to K — came faster and told you the long-run fate directly. That matters because the logistic is a rare exception. Most ODEs that come out of real models have no closed-form solution at all, and then the sign-and-arrow analysis is not a shortcut, it is the only tool you have. The logistic is the friendly training ground where you can check the qualitative method against an exact answer and watch them agree.
Fitting K, and what the model still hides
The equation supplies the *shape*; data must supply the *numbers* k and K. A clean trick: since the per-capita rate (1/P) dP/dt = k(1 - P/K) is a straight line in P, plot measured per-capita growth against population. If the points fall on a downward line, the logistic assumption is vindicated; the line's intercept gives k and its zero-crossing gives K. This is the fitting-and-validation step in action — and if the points clearly curve instead, that is honest evidence the simple logistic is the wrong shape for your system.
Be honest about what the tidy S-curve omits. It assumes K is a fixed constant, but a real environment's capacity shifts with seasons, rainfall, and human meddling. It assumes growth depends only on present size, ignoring time lags — a real population may overshoot K before it feels the crowding, and oscillate. And it assumes any positive head-start grows; many real species suffer an Allee effect, dying out below a minimum viable population. The plain logistic, with its single stable ceiling, cannot show that — a reminder that this elegant equation, like every model, has a domain of honesty beyond which it quietly misleads.
One honest extension is worth a peek because it shows the qualitative method earning its keep. Subtract a constant harvest h (fishing, hunting) to get the [[harvesting-model|harvesting model]] dP/dt = k P (1 - P/K) - h. Now the equilibria are the roots of a quadratic, and as you crank h up the two equilibria slide together and, past a critical harvest rate, vanish entirely — the population that had a safe ceiling abruptly has none and crashes to extinction. That sudden disappearance of equilibria as a parameter passes a threshold is a bifurcation, the dramatic next idea your phase-line skills will let you see coming.