Growth Patterns and Nonlinear Behavior

Summary

This chapter examines how systems grow and change state, from linear and exponential growth through S-curves and diminishing returns. It introduces nonlinear behavior, tipping points, chaos theory, and overshoot-and-collapse as patterns that defy simple prediction. Students completing this chapter will be able to recognize exponential growth and tipping-point behavior in real systems and explain why they are easy to underestimate.

Concepts Covered

This chapter covers the following 25 concepts from the learning graph:

Concept	Concept Impact Score
Linear Growth	202
Linear Relationship	201
Exponential Growth	162
Exponential Blind Spot	3
Compound Interest	2
Compounding Effect	88
Nonlinear	198
Nonlinear Behavior	182
Nonlinear Feedback	15
Threshold Effect	90
Tipping Point	4
Bifurcation	54
Chaos Theory	22
Butterfly Effect	2
Attractor	2
Phase Transition	31
Edge Of Chaos	15
Sharpness Of Transition	15
Unpredictability Of Emergence	2
Oscillation	1
Overshoot	5
Overshoot And Collapse	4
S-Curve	38
Carrying Capacity	37
Diminishing Returns	2

Prerequisites

This chapter builds on concepts from:

Introduction

Chapter 4 showed that a system with more than one loop behaves like whichever loop currently dominates, and that dominance can shift. Chapter 5 showed that a stock's rate of change is just its net flow. Put those two ideas together and you get this chapter's subject: growth curves that are almost never simple straight lines, because the flows feeding a stock rarely stay constant, and the loops driving those flows rarely stay dominant forever. This chapter is a field guide to the shapes growth actually takes — and to why nearly every one of those shapes is easy for a person to underestimate the first time they meet it.

The Shapes Growth Actually Takes

Sage waving welcome Get ready to meet the growth curves that fool almost everyone at first glance — the ones that look tame for a while and then absolutely aren't. Let's zoom out and see the whole system!

Growing by the Same Amount: Linear Growth

Linear growth is a pattern in which a quantity increases (or decreases) by the same fixed amount during each equal time period — add 10 units this month, add 10 more next month, add 10 more the month after that, forever. The relationship this produces between time and the quantity is a linear relationship: a relationship in which one quantity changes at a constant rate with respect to another, which is precisely why plotting linear growth against time always produces a perfectly straight line.

[ y = mx + b ]

In that familiar equation, ( b ) is the starting amount, ( x ) is elapsed time, and ( m ) is the fixed amount added each period — the slope of the straight line. A savings plan that deposits exactly $50 every month, starting from $200, is linear growth: after ( x ) months, the balance is ( 200 + 50x ). After 10 months, that's $700; after 20 months, $1,200 — the balance always grows by the same $50, regardless of how large it has already become.

Growing by the Same Percentage: Exponential Growth

Exponential growth is a fundamentally different pattern: a quantity increases by the same percentage of its current value during each equal time period, rather than by the same fixed amount. This is exactly the reinforcing-loop behavior Chapter 3 introduced with compounding interest — each period's growth is calculated on an already-larger base, so the actual amount added keeps increasing even though the percentage stays fixed.

[ N(t) = N_0 (1 + r)^t ]

Here ( N_0 ) is the starting amount, ( r ) is the constant growth rate per period (as a decimal), and ( t ) is the number of periods elapsed. Compound interest is the financial name for exactly this pattern applied to money: interest earned in one period is added to the balance, so the next period's interest is calculated on a larger base. The compounding effect is the general name for that same mechanism outside of finance — any situation where a quantity's own growth feeds back into calculating its next round of growth, from a viral video's views driving more recommendations, to bacteria dividing, to a rumor spreading through a school.

Compare a $200 balance earning a flat $50 a month (linear) against that same $200 growing by 25% a month (exponential, ( r = 0.25 )). After 10 months, the linear balance is $700, just as before. The exponential balance is ( 200 \times 1.25^{10} \approx \$1{,}863 ) — already more than double. After 20 months, linear is $1,200; exponential is ( 200 \times 1.25^{20} \approx \$19{,}481) — over sixteen times as much. Notice something important about when the two curves diverge: in the first month or two, the exponential balance barely looks different from the linear one. The two curves feel almost interchangeable near the start, and it's precisely that early resemblance that causes the mistake this chapter's next term describes.

The exponential blind spot is the well-documented human tendency to underestimate how large an exponentially growing quantity will eventually become, because early exponential growth looks deceptively similar to linear growth and human intuition defaults to a linear mental model. A famous illustration: fold a sheet of paper in half 42 times (physically impossible, but bear with the math), and its thickness would reach the Moon — each fold merely doubles the thickness, yet doubling, repeated enough times, outruns almost any linear guess by an enormous margin.

The MicroSim below makes the exponential blind spot visible rather than just described: watch how a straight-line, a curved-polynomial, and an exponential quantity that all look similar for a while completely separate once you zoom out far enough.

Diagram: Growth Rate Comparison

Run the Growth Rate Comparison MicroSim fullscreen

Growth Rate Comparison (reused MicroSim)

Type: chart sim-id: growth-rates
Library: p5.js
Status: Reused
Source: https://dmccreary.github.io/calculus/sims/growth-rates/
Source Repo: https://github.com/dmccreary/calculus/tree/main/docs/sims/growth-rates

Reused from the MicroSim catalog (WHAT match score 0.76). Plots logarithmic, linear/polynomial, and exponential quantities on the same axes and lets the viewer zoom the x-axis from x=1-10 out to x=100-1000, showing that functions which look comparably sized up close separate dramatically at scale. Learning objective: given several growth patterns that appear similar over a short interval, the learner will predict which one dominates over a much longer interval (Bloom: Analyzing).

The MicroSim below shows the same idea from the other side — not comparing exponential growth to other patterns, but letting you drive an exponential reinforcing loop yourself and watch the compounding effect build.

Diagram: Population Growth Simulator

Run the Population Simulator fullscreen

Population Growth Simulator (reused MicroSim)

Type: microsim sim-id: population-simulator
Library: p5.js
Status: Reused
Source: ../../sims/population-simulator/main.html
Source Repo: local — docs/sims/population-simulator

Reused from this book's own sims collection, previously described in a Chapter 3 lesson plan around the reinforcing "Population-Births" loop. Here the same simulator illustrates compounding directly: raise the birth-rate slider and watch how a higher constant percentage rate produces a dramatically steeper — not just proportionally steeper — curve. Learning objective: given different constant growth-rate settings, the learner will predict the resulting population at a fixed future time (Bloom: Applying).

Why Your Gut Underestimates Compounding

Sage thinking with a raised wing Notice that the exponential blind spot isn't a math error — it's a mismatch of mental models. Your intuition is built for linear addition ("a little more each time"), but exponential growth is multiplication repeated ("a little more, of an already-bigger amount, each time"). Once you catch yourself reasoning "it only grew a little last period," ask whether the percentage stayed the same — if it did, expect the next period's amount to be bigger still.

Nonlinear Behavior and Nonlinear Feedback

Both linear and exponential growth are, in one sense, orderly: a linear relationship changes by a constant amount, and an exponential one changes by a constant percentage — both are entirely predictable once you know their one governing number. Nonlinear describes any relationship that is not linear — where the change in output is not simply proportional to the change in input, so doubling an input doesn't necessarily double (or produce any fixed multiple of) the output. Nonlinear behavior is the broader term for the actual patterns of change a system displays when its underlying relationships are nonlinear — patterns that can accelerate, decelerate, reverse, or shift entirely, sometimes several of those in succession.

Exponential growth is technically already nonlinear (its graph curves, not a straight line), but systems thinkers usually reserve close attention for cases where the feedback itself changes character, not just its magnitude. Nonlinear feedback is feedback in which the strength of the loop's effect is not constant but instead depends on the current value of the variable involved — unlike the fixed growth rate ( r ) in the exponential equation above, a nonlinear feedback's effective "r" can shrink, grow, or flip sign as conditions change. A crowded highway shows this well: at low traffic density, adding one more car barely slows anyone down, but past a certain density, each additional car slows everyone significantly more than the last one did — the feedback from "more cars" to "more delay" isn't a fixed multiplier, it gets stronger as the road fills up.

Crossing a Line: Thresholds and Tipping Points

Some nonlinear effects don't just change gradually in strength — they switch on abruptly at a specific point. A threshold effect is a nonlinear pattern in which a system shows little or no response to a changing input until that input crosses a specific critical value, after which the response appears suddenly and often substantially. Water is a familiar example: raising its temperature from 90°C to 99°C changes very little about how it behaves, but crossing 100°C at sea level flips it from liquid to boiling steam — the response was negligible right up until the threshold, then dramatic.

A tipping point is the specific critical value at which a threshold effect occurs — the exact input level beyond which the system's behavior changes qualitatively, not just quantitatively. Continuing the water example, 100°C is the tipping point; everything below it belongs to one behavioral regime (liquid), everything above it to another (gas). Because so little changes right up until the tipping point itself, systems near one are notoriously easy to misjudge — recall Chapter 4's loop-dominance lesson: a system can look stable for a long stretch of shifting dominance right before the previously weak loop finally takes over.

Diminishing Returns: When More Stops Helping as Much

Not every departure from a straight line involves compounding growth. Diminishing returns describes a pattern in which each additional unit of some input produces a smaller increase in output than the unit before it — output still rises, but at a shrinking rate. Studying for a test shows this clearly: the first hour of review typically raises a score substantially, because it covers unfamiliar material; by the fifth or sixth consecutive hour, exhaustion and material already mastered mean each additional hour adds much less. Diminishing returns is the mechanism that most often puts the bend in a growth curve that would otherwise keep compounding forever — which is exactly the mechanism behind the shape in the next section.

The S-Curve: Growth That Meets a Limit

Reinforcing loops compound without limit only on paper. In the real world, something eventually pushes back, and the resulting shape has its own name. An S-curve is a growth pattern that starts out looking exponential — slow at first, then rapidly accelerating — but then bends and flattens as the growing quantity approaches some upper limit, producing a curve shaped, unsurprisingly, like the letter S. A carrying capacity is that upper limit itself: the maximum level a population or quantity can sustain given the constraints of its environment or system, beyond which growth cannot continue at its earlier pace.

Chapter 4's epidemic example was already an S-curve in disguise: infections grew exponentially at first (the reinforcing loop dominating), then slowed and flattened as the pool of susceptible people — the epidemic's own carrying capacity — ran out. The same shape appears whenever a reinforcing loop feeds a stock that some other part of the system limits: bacteria multiplying in a petri dish until they exhaust the available nutrients, a new product's adoption climbing steeply and then leveling off once it has reached nearly everyone who wants it, a fast-growing city's population plateauing once it runs out of buildable land or water supply.

Mathematically, an S-curve is the solution to the logistic growth equation, which starts from the exact same reinforcing-growth idea as exponential growth but multiplies it by a term that shrinks toward zero as the stock ( N ) approaches the carrying capacity ( K ):

[ \frac{dN}{dt} = rN\left(1 - \frac{N}{K}\right) ]

When ( N ) is small compared to ( K ), the term ( \left(1 - \frac{N}{K}\right) ) is close to 1, so growth looks almost exactly exponential — this is the diminishing-returns term barely engaged yet. As ( N ) climbs toward ( K ), that same term shrinks toward zero, throttling growth down to nothing exactly at the carrying capacity. This is nonlinear feedback from a few sections ago made mathematically explicit: the growth rate's effective strength isn't the constant ( r ), it's ( r\left(1 - \frac{N}{K}\right) ) — a quantity that depends on how close the stock already is to its limit.

The local causal loop diagram below shows this same structure as two competing loops rather than as an equation: a reinforcing "Growth" loop pushing a system condition upward, and a balancing "Slowing" loop that strengthens as that condition nears a limiting condition.

Diagram: Limits to Growth Causal Loop Diagram

Run the CLD Viewer fullscreen

Limits to Growth Causal Loop Diagram (reused MicroSim)

Type: graph-model sim-id: cld-viewer
Library: vis-network
Status: Reused
Source: ../../sims/cld-viewer/main.html?file=limits-to-growth-cld.json
Source Repo: local — docs/sims/cld-viewer (examples/limits-to-growth-cld.json)

Reused from this book's own CLD Viewer, loaded with a dual-loop example built specifically to accompany this chapter's growth-and-limits material: a reinforcing loop driving a shared "System Condition" upward and a balancing loop that strengthens as that condition nears a "Limiting Condition." Hovering over each node shows real-world examples (population size, revenue, market penetration) it could represent. Learning objective: given a reinforcing growth loop and a balancing limiting loop sharing a stock, the learner will explain why their combined effect produces an S-curve rather than unbounded exponential growth (Bloom: Understanding).

What happens if the growing quantity doesn't gently approach its carrying capacity, but keeps compounding past it before the balancing loop's effect has time to catch up? An overshoot is exactly that: a growing quantity briefly exceeding its carrying capacity or sustainable limit before being pulled back down — a delay (Chapter 4 again) between crossing the limit and the balancing loop actually reining growth back in. Overshoot and collapse is the more severe version, where the excess isn't gently absorbed but instead damages the very limit the system depends on, causing the quantity to crash well below the carrying capacity rather than settling near it — a fish population that, having overshot the ocean's sustainable yield, degrades the breeding stock so badly that the resulting population collapses to a level far lower than the original carrying capacity, rather than a smooth S-curve landing.

The MicroSim below lets you set a growth rate and a carrying capacity and watch three outcomes side by side: a smooth S-curve when the response is quick, an overshoot that settles back near the limit when there's a moderate delay, and overshoot and collapse when the delay is long enough for real damage to occur first.

Diagram: Logistic Growth S-Curve Explorer

Run the Logistic Growth S-Curve Explorer fullscreen

Logistic Growth S-Curve Explorer

Type: microsim sim-id: logistic-growth-explorer
Library: p5.js
Template: https://github.com/dmccreary/ecology/tree/main/docs/sims/population-growth
Status: Specified

Purpose: Let learners compare pure exponential growth against logistic (S-curve) growth, and see how response delay turns a smooth S-curve into overshoot or overshoot-and-collapse.

Bloom Taxonomy Level: Analyze Bloom Taxonomy Verb: Compare / Predict

Learning Objective: Given a carrying capacity and a feedback-delay setting, the learner will predict whether a growing quantity settles into a smooth S-curve, overshoots and recovers, or overshoots and collapses (Bloom: Analyzing).

Canvas: 700x460 default, responsive — recompute plot axes as a proportion of canvas.width inside a windowResized() handler.

Visual elements: - A single time-series chart plotting quantity (y-axis) against time (x-axis), with a dashed horizontal line marking the current carrying capacity K. - Three selectable trace colors: pure exponential (light gray, for reference, uncapped), logistic S-curve (green), and the current delay-affected run (blue), all drawn on the same axes for direct comparison. - A live readout showing current value, percent of carrying capacity, and elapsed time steps.

Controls: - A growth-rate slider (createSlider(), range 0.05-0.5) setting ( r ). - A carrying-capacity slider (range 50-500) setting ( K ). - A response-delay slider (range 0-15 time steps) determining how many steps elapse between the stock exceeding K and the balancing term fully engaging; at delay 0 the curve is a textbook smooth S-curve, at moderate delay it overshoots and settles back near K, and at high delay combined with a "fragile limit" checkbox it overshoots and collapses to well below K (simulating capacity damage). - A "fragile limit" checkbox (createCheckbox()) that, when checked and combined with a high delay, permanently lowers K after an overshoot, producing the overshoot-and-collapse pattern rather than mere overshoot. - A "Run" button and a "Reset" button built with createButton().

Interactivity requirement: every slider and checkbox immediately changes the plotted curve and live readout on the next simulation step, giving direct, visible feedback tying delay and limit fragility to the resulting growth shape.

Color scheme: exponential reference trace in neutral gray, logistic S-curve in the book's balancing-loop green, delay-affected run in the reinforcing-loop red once it exceeds K (to visually flag overshoot), carrying-capacity line as a dashed dark gray.

Implementation: p5.js sketch computing the logistic difference equation ( N_{t+1} = N_t + rN_t(1 - N_t/K)\,\Delta t ) each frame, with a delay buffer storing past ((1-N/K)) terms so the currently applied balancing strength lags behind the current stock by the slider's delay amount; a "fragile limit" flag reduces K once N exceeds it by more than a set margin.

Sudden Shifts: Phase Transitions and the Edge of Chaos

A tipping point marks a single variable crossing a critical value. Sometimes what changes at that crossing is the entire character of a system's behavior, not just one number. A phase transition is a sudden, qualitative change in a system's overall state or organization, triggered by a smooth, gradual change in some underlying condition — the same water-to-steam example from earlier in the chapter is the textbook physical case, but the term applies equally to an unstructured social network suddenly organizing into rigid factions, or a smoothly flowing traffic stream suddenly locking into stop-and-go gridlock.

The sharpness of transition describes how narrow or wide the range of the underlying condition is over which a phase transition actually takes place — some transitions are razor-sharp (water boils across a fraction of a degree at a given pressure), while others are gradual, spreading the shift across a wide range of the underlying variable rather than a single crisp point. The edge of chaos is a special zone found in some systems, sitting between a highly ordered, predictable regime and a highly disordered, chaotic one, where the system is unusually rich in structure, adaptability, and complex behavior — neither frozen into rigid order nor dissolved into pure randomness. Researchers studying everything from cellular automata to evolving ecosystems and even neural networks have found this boundary zone to be where the most interesting, information-rich behavior tends to concentrate, rather than at either extreme.

Sharp Transitions Deserve Extra Caution

Sage pointing helpfully When you're monitoring a system that might have a sharp phase transition ahead, don't trust a smoothly changing dashboard number to warn you in time. A sharp transition can look completely calm right up until it isn't — build in a safety margin well before the suspected threshold, not right up against it.

Into Chaos: Bifurcation, Sensitivity, and Unpredictable Emergence

Some nonlinear systems don't just shift once at a threshold — they undergo a whole cascade of qualitative changes as a single parameter is turned up. A bifurcation is a point at which a small, smooth change in a system's parameter causes a sudden qualitative change in its long-term behavior — for instance, a system that used to settle into one stable value instead starts oscillating between two. Oscillation, formally, is a behavior pattern in which a variable repeatedly rises and falls rather than settling at a constant value or growing without limit — the same pattern Chapter 4's delayed thermostat produced, but here arising from the shape of a feedback rule itself rather than from a time lag.

The classic mathematical illustration is the logistic map, a deceptively simple discrete version of this chapter's growth equation:

[ x_{n+1} = r\, x_n (1 - x_n) ]

At low values of the parameter ( r ) (below about 3), this equation settles down to one fixed value no matter where it starts — orderly and entirely predictable. Push ( r ) past about 3, and a bifurcation occurs: the system stops settling on one value and instead oscillates forever between two. Push ( r ) further, and it bifurcates again into a four-value cycle, then eight, faster and faster, until around ( r \approx 3.57 ) the behavior becomes what mathematicians call chaos theory's namesake subject — a chaotic regime in which the sequence never repeats and never settles, despite being generated by a rule with no randomness in it whatsoever.

That last detail is the genuinely surprising part: chaos, in this technical sense, isn't randomness — it's fully deterministic behavior that is nevertheless practically unpredictable in the long run, because of extreme sensitivity to starting conditions. The butterfly effect is the name given to exactly that sensitivity: in a chaotic system, an arbitrarily tiny difference in starting conditions grows over time into a completely different outcome, so named from the whimsical image of a butterfly's wingbeat in one location eventually being linked to a distant storm forming or not forming weeks later. Run the logistic map at ( r = 3.9 ) twice, starting from ( x_0 = 0.500000 ) and ( x_0 = 0.500001 ) — a difference of one part in half a million — and within a few dozen steps, the two sequences are completely unrelated to each other, even though each was computed with the exact same, entirely deterministic rule.

Don't Mistake Chaos for Randomness

Sage holding up a wing in caution A chaotic system and a genuinely random one can look identical on a graph, but they are not the same thing, and mixing them up leads to the wrong response. A random process can't be made more predictable by better measurement. A chaotic one, in principle, could be — the problem isn't randomness, it's that you'd need impossibly precise knowledge of the starting conditions.

Not every dynamic system wanders forever, even a chaotic one. An attractor is a state, or a bounded set of states, that a dynamic system tends toward over time regardless of a range of different starting points. A system settling on one fixed value (like the logistic map below ( r = 3 )) has a single-point attractor; a system cycling between two or four values has a small-loop attractor; and a chaotic system like the logistic map at ( r = 3.9 ) has a so-called "strange attractor" — it never repeats exactly, yet it also never leaves a specific bounded region of possible values, wandering unpredictably but always within recognizable bounds.

The table below reinforces this progression from order to chaos, now that every stage has been defined and worked through above:

Approximate ( r ) range	Behavior	Type of attractor
Below ~3.0	Settles at one fixed value	Single point
~3.0 to ~3.44	Oscillates between 2 values	Small loop (2-cycle)
~3.44 to ~3.57	Repeatedly bifurcates (4, 8, 16 values…)	Growing loop
Above ~3.57	Chaotic — never exactly repeats	Strange attractor

The MicroSim below lets you drag the ( r ) slider through exactly this progression and watch the bifurcation diagram build up live, alongside a side-by-side butterfly-effect comparison of two nearly identical starting points.

Diagram: Logistic Map Bifurcation Explorer

Run the Logistic Map Bifurcation Explorer fullscreen

Logistic Map Bifurcation Explorer

Type: microsim sim-id: logistic-map-bifurcation-explorer
Library: p5.js
Status: Specified

Purpose: Let learners drag a growth-rate parameter through the logistic map's route to chaos and directly observe fixed points, periodic bifurcations, and chaotic behavior, alongside a live demonstration of sensitivity to initial conditions.

Bloom Taxonomy Level: Analyze Bloom Taxonomy Verb: Classify / Compare

Learning Objective: Given a range of parameter values for the logistic map, the learner will classify the resulting long-term behavior as a fixed point, a periodic cycle, or chaos (Bloom: Analyzing).

Canvas: 700x460 default, responsive — recompute both panels' widths as a proportion of canvas.width inside a windowResized() handler.

Visual elements: - Left panel: a live bifurcation diagram (r on the x-axis from 2.4 to 4.0, long-run x-values on the y-axis), plotting a faint point for each of the last 60 iterations at the current r, with a vertical marker line showing the currently selected r. - Right panel: two trajectory line charts stacked vertically, one starting at (x_0=0.5000) and one at (x_0=0.5001), both run at the current r, initially overlapping almost exactly and visibly diverging within a few dozen steps whenever r is in the chaotic range. - A live label beneath the bifurcation diagram reading the current classification: "Fixed point," "Period-2 cycle," "Period-4 cycle," or "Chaotic," computed from the spread of the last 60 iterated values.

Controls: - An r-slider (createSlider(), range 2.4 to 4.0, step 0.001) that immediately recomputes both panels. - A "Reset Trajectories" button, built with createButton(), that reseeds both right-panel trajectories at their starting values without changing r. - Clicking anywhere on the bifurcation diagram jumps the r-slider to that x-position, letting learners explore visually rather than only by dragging the slider.

Interactivity requirement: dragging the slider or clicking the bifurcation diagram immediately updates both the live classification label and the diverging trajectory panel, giving continuous visible feedback linking a single parameter to a qualitative change in long-term behavior.

Color scheme: bifurcation diagram points in the book's neutral node blue, the two trajectory lines in contrasting reinforcing-red and balancing-green so their divergence is easy to track, chaotic-range background tinted a very light red to visually flag the zone.

Implementation: p5.js sketch iterating ( x_{n+1} = r x_n(1-x_n) ) each frame for both the bifurcation diagram (discarding an initial transient before plotting) and the two trajectory panels; classification logic checks how many distinct values the last 60 iterations cluster into, within a small tolerance, to label fixed/periodic/chaotic.

However far a chaotic system's precise trajectory may be from predictable, that doesn't mean nothing useful can be said about it — which is exactly the nuance behind this chapter's final term. Unpredictability of emergence is the recognition that a complex or chaotic system can produce emergent, system-level patterns that are real and often statistically describable, even though the system's exact, moment-to-moment trajectory is not predictable in detail. Weather is the standard example: no forecaster can tell you the exact temperature in a specific city 90 days from now, because weather is chaotic in precisely this technical sense — yet climate scientists can say a great deal, with real confidence, about that season's average temperature and rainfall pattern. Unpredictable in detail and well-described in aggregate are not contradictions; they're both true at once, and telling them apart is one of the more valuable habits this chapter can leave you with.

Chaos Feels Unsettling at First — That's Normal

Sage giving an encouraging nod If "a deterministic rule that's still unpredictable" feels like a contradiction, you're in good company — it startled professional mathematicians when they first encountered it too. Play with the bifurcation slider above slowly, one small step at a time, and the transition from order to chaos will start to feel less like magic and more like a pattern you can recognize on sight.

Key Takeaways

You can now recognize the shape of growth, not just its direction:

Linear growth adds a constant amount each period; exponential growth — driven by compounding and, in finance, compound interest — adds a constant percentage, which is why the exponential blind spot so reliably fools human intuition.
Nonlinear relationships break the constant-multiplier assumption behind linear growth; nonlinear feedback is a loop whose strength itself changes with the variable's value.
A threshold effect produces little visible change until a tipping point is crossed, after which behavior shifts abruptly; diminishing returns instead bends a curve gradually as each additional input yields less.
An S-curve is what reinforcing growth looks like once it meets a carrying capacity; a delayed response to that limit produces overshoot, and a damaged limit produces the more severe overshoot and collapse.
A phase transition is a sudden qualitative shift in a system's whole state, whose sharpness of transition can range from razor-sharp to gradual; the edge of chaos is a uniquely rich zone between order and disorder.
A bifurcation is a parameter change that suddenly alters long-term behavior — from a fixed point, to oscillation, toward full chaos theory's namesake regime, where the butterfly effect makes precise long-term prediction practically impossible even though an attractor still bounds where the system can go, and unpredictability of emergence reminds you that aggregate patterns can still be knowable even when exact trajectories aren't.

You Can Now Spot the Shape Behind Any Growth Story

Sage celebrating with wings raised Whoo-hoo! Straight lines, compounding curves, S-curves that bend at a limit, and the strange, beautiful edge where order gives way to chaos — you can now recognize every one of them on sight, and explain why each one so often catches people by surprise. Chapter 7 asks what it takes for a system to survive being surprised like that.