Adaptation, Plasticity, and Extinction in a Changing EnvironmentTowards a Predictive Theory

Species today are being pressed by sustained environmental change, such as climate shifts, habitat loss, and overexploitation, faster than our ability to predict which populations will survive. Chevin, Lande, and Mace set up the problem starkly: habitat fragmentation prevents many species from simply tracking suitable conditions across the landscape. So, populations must adapt in place or face extinction. The question isn't just where suitable climates will exist on a future map. It’s whether a given population can maintain positive growth as its local environment shifts beneath it. Two modeling traditions try to answer this, and both fall short. Climate-envelope models, also called niche models, are correlative. They use records of species presence and absence alongside environmental variables to infer where a species can persist, then project that envelope onto future climates. But as Chevin and colleagues point out, these models rely on equilibrium assumptions, are sensitive to spatial scale, cannot account for source-sink dynamics, and, critically, ignore evolution entirely. At the other end, mechanistic population models combine evolutionary genetics and demography to track ecologically important traits and their effects on population growth. That approach can identify the biological limits to persistence, but without an explicit environmental variable built in, you cannot project what happens under climate change. Chevin, Lande, and Mace propose unifying these two traditions. Building on earlier theory by Lynch and Lande, they introduce a model with two often-missing ingredients: phenotypic plasticity and an explicit environmental variable—say, temperature—that drives selection on traits. The core of their framework is a single threshold: the critical rate of environmental change, which they call g sub c. Think of it as the speed limit. It's the maximum sustained pace of environmental change—degrees Celsius per year, for example—that an isolated population can tolerate indefinitely while its mean phenotype tracks the shifting optimum. Exceed g sub c, and the population's phenotype lags so far behind the optimum that intrinsic growth turns negative. Extinction follows. Four classes of parameters set this threshold. First, demography: generation time T and the maximum intrinsic rate of increase, r sub max. Longer generation times mean fewer evolutionary opportunities per unit time, and because r sub max is roughly inversely proportional to T across species, slow breeders face double jeopardy—they evolve slowly and have less demographic buffer. Second, evolutionary potential: the additive genetic variance and the strength of stabilizing selection, written as c. Stabilizing selection is the force that penalizes deviation from an optimum—c measures how steeply fitness falls as the phenotype strays. Higher genetic variance and stronger stabilizing selection both enable faster evolutionary tracking. Verbally, g sub c scales with the square root of heritability multiplied by the strength of stabilizing selection. Take those two numbers, multiply them, and take the square root—that quantity is proportional to the population's speed limit. The third and fourth parameters are where plasticity enters, and they are the most novel part of the model. The environmental sensitivity of selection, B, is the slope of the relationship between the environment and the optimum phenotype—how far the optimum shifts per unit of environmental change. Phenotypic plasticity, b, is the slope of the individual reaction norm—how far an individual's phenotype shifts per unit of environmental change. These two quantities are directly comparable. When b is close to B, the organism's plastic response tracks the moving optimum almost automatically. The critical rate g sub c increases as the gap between B and b narrows. Plasticity reduces the lag between the population mean and the moving optimum, and that reduction is what matters—it buys persistence under faster change. That's the optimistic version of plasticity. Now for the complications. Chevin and colleagues are careful to show that plastic responses carry costs, and those costs can bite back. They distinguish two kinds. A constitutive cost reduces fitness simply by virtue of having plastic machinery, like sensory systems, regulatory networks, and the infrastructure of developmental flexibility. An induced cost reduces fitness in proportion to how much phenotypic change is actually expressed. Either way, the cost reduces r sub max, which in turn reduces g sub c. The paper states it directly: r sub max equals r sub zero minus a cost that depends on plasticity b, where r sub zero is the growth rate of a population sitting at the optimum with no plasticity. There's even a threshold: when the cost of plasticity rises above a critical level, only intermediate values of b maximize the critical rate. Too much plasticity, and the costs overwhelm the benefit. These costs also generate the generalist-specialist trade-off that shows up in thermal tolerance curves—a framework the paper develops in detail. Under linear reaction norms with stabilizing selection, the reaction-norm slope sets tolerance breadth and the intercept sets the optimal environment. Add a cost that increases with slope, and you get a trade-off: broader tolerance comes at the price of lower peak fitness—the classic jack-of-all-trades outcome. Chevin and colleagues demonstrate this numerically with specific parameters—r sub max of zero point one four, heritability of zero point five, stabilizing selection c equal to one fifty-first—and show how increasing costs shift the range of b values that sustain long-term persistence. Thermal tolerance curves are also typically skewed, they note, because nonlinear reaction norms in underlying traits produce saturation or amplification at physiological extremes. And critically, in novel or extreme environments, the linear reaction norm assumption breaks down entirely. Plastic responses can saturate, be disrupted, or take unexpected shapes where they haven't been tested by selection. Plasticity buys time. It has a ceiling. Given all of this, what do biologists actually need to measure? Chevin and colleagues are specific. The classical tool remains essential: regress individual fitness on phenotype to quantify selection. But most studies estimate relative fitness, which predicts evolutionary change, while the paper stresses that absolute fitness and r sub max are also required to link trait evolution to population growth and extinction risk. Selection analyses should be extended to age- and stage-structured life histories because the parts of the life cycle where phenotype most strongly shapes demography vary by species and environment. Beyond classical selection analysis, one parameter stands out as the keystone. The environmental sensitivity of selection, B—the slope of the relationship between the environment and the optimum phenotype—has been almost entirely neglected empirically. Most field studies compare selection between a small number of discrete conditions: drought versus non-drought, warm year versus cool year. Discrete contrasts don’t tell you how the optimum shifts continuously along an environmental gradient. And continuous gradients are exactly what climate projections deliver. Chevin and colleagues call for measuring phenotypic selection along continuous environmental gradients so that B can be estimated directly, not guessed. Once you have B alongside b, genetic variance, c, generation time, and r sub max, you can compute g sub c for a population and compare it against projected rates of environmental change to get an actual extinction-risk forecast. That's the payoff of the whole framework. A unified model that integrates genetic evolution, phenotypic plasticity, demography, and physiology can be connected to climate projections to ask a pointed question: will this population's critical rate exceed the rate of change its environment will experience? Chevin and colleagues acknowledge the approach is more expensive than niche modeling—it demands extensive field and lab measurement—but their results are more useful precisely because they include the mechanisms that determine whether populations adapt or collapse. They point to encouraging examples: Deutsch and colleagues using thermal tolerance curves, and Kearney and colleagues combining biophysical models with heritability, both pushing toward this integration. What's still missing is the combination of all the pieces—plasticity, genetic evolution, and demography—in a single empirical program. We are the cause of the environmental change these populations are experiencing. That’s the context Chevin, Lande, and Mace never let drift far from the surface. The model they've built is a scaffold for the harder empirical work ahead. Identify the traits, measure selection along environmental gradients, quantify plasticity and its costs, estimate genetic variance and demographic rates, and then ask whether the population's speed limit exceeds what the climate will demand of it. For the species that face that test, knowing the answer in advance is exactly what conservation requires. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.