A Meta-Analysis of Local Adaptation in Plants

Most plant populations are not locally adapted. They just look like they are. That's the finding at the center of this meta-analysis by Roosa Leimu and Markus Fischer, and it challenges one of the most deeply held assumptions in ecology. Local adaptation, in evolutionary ecology, means something specific: plants from a given population perform better in their home environment than do plants that originated elsewhere. The logic is straightforward. Natural selection acts on traits in local conditions, so over time, local genotypes should be shaped to fit local soil, local climate, and local competitors. Ecologists have treated this as the default expectation for so long that it's become background furniture—something you assume is happening when a plant thrives in its home patch. Leimu and Fischer decided to actually count how often it's true. The empirical test for local adaptation is the reciprocal transplant experiment. You take plants from two populations, swap them, grow each origin at each site, and measure who does better where in terms of reproduction, survival, biomass, and germination. Leimu and Fischer combed through the literature, screened 211 articles, and retained 35 papers covering 32 plant species and 1,032 pairwise comparisons of local versus foreign performance. To standardize results across different traits and studies, they calculated Hedges' d for each comparison. That's the difference in mean performance between local and foreign plants, divided by the pooled standard deviation, with a small correction for sample size. A positive Hedges' d means local plants are winning. But they also ran a second, stricter analysis. The conventional measure asks: do local plants beat foreign plants at a given site? The stricter measure asks: do local plants beat foreign plants at both sites in a reciprocal pair? Following Kawecki and Ebert's definition, true local adaptation requires crossing reaction norms, with each population outperforming the other in its own environment. Leimu and Fischer classified every pairwise comparison as POS-POS, meaning local wins at both sites; POS-NEG, meaning local wins at only one site; or NEG-NEG, meaning foreign wins at both. These two tests—the looser per-site measure and the stricter per-pair measure—are what make the results interesting. Here's the headline number: local plants outperformed foreign plants at their site of origin in seventy-one percent of the individual transplant sites. The overall effect size was a Hedges' d of zero point sixteen—small but statistically clear. On its face, that sounds like confirmation of the textbook story. Then comes the asterisk. When Leimu and Fischer applied the stricter pairwise test—do local plants win at both sites—only forty-five point three percent of the one thousand thirty-two population pairs qualified. In more than half of all pairwise comparisons, the reciprocal pattern that defines true local adaptation simply wasn't there. Instead, fifty-one point four percent of pairs showed the POS-NEG pattern, with one population outperforming the other at both sites, not each population thriving at home. That's not local adaptation. That's one population being generally more vigorous. This distinction is the conceptual core of the paper. The per-site seventy-one percent figure can be inflated by asymmetry—by one population that just happens to be better everywhere, raising the count of sites where "local" wins without any of that signaling divergent selection. Only the POS-POS cases provide real evidence that natural selection has shaped each population to its home environment. And those cases cover fewer than half the comparisons. Local adaptation in plants is less common than the field has assumed. Now for the variable that actually predicted local adaptation: population size. Large populations—defined here as more than one thousand flowering individuals—were locally adapted far more often than small populations. The pattern shows up in both analyses. In the frequency comparison of twenty-four large-population pairs versus eight small-population pairs, the likelihood of local adaptation was seventy-six percent for large populations and only forty-nine percent for small ones. That's a substantial gap, and it holds up even under conservative interpretation of ambiguous cases. The evolutionary logic is straightforward. Larger populations carry more genetic variation, which gives natural selection more raw material to work with. Small populations suffer stronger genetic drift—the random shuffling of allele frequencies that can overwhelm selection, causing beneficial variants to disappear by chance. Add inbreeding depression and founder effects, and you have multiple compounding mechanisms that all push small populations away from local adaptation. This finding reinforces something population geneticists have argued theoretically for decades: effective population size is not just a demographic concern; it's an evolutionary one. Equally important are the negative results. Leimu and Fischer found that local adaptation appeared to be independent of the plant life-history traits they considered—longevity, mating system, and clonality—and independent of spatial and temporal habitat heterogeneity. Geographic distance between sites showed no effect either. That lack of a distance effect held across a range spanning six orders of magnitude, from three meters to three thousand five hundred kilometers. Three meters and three thousand five hundred kilometers produced statistically indistinguishable likelihoods of local adaptation. The null results for those variables are almost as informative as the positive result for population size, because they indicate which intuitions to stop relying on. A few caveats are worth naming, because Leimu and Fischer mention them. The dataset covers primarily temperate herbaceous plants—the species most commonly studied in transplant experiments—so generalizing to woody plants, tropical species, or highly specialized plant communities should be done cautiously. The classification of population size used a single threshold, and the habitat heterogeneity categories were coarse. These are real limitations. But the magnitude of the population-size effect relative to the null effects for other variables suggests the signal is genuine, not an artifact of measurement precision. The implications are particularly relevant in conservation biology and climate change ecology. If local adaptation requires large populations, and most plant populations worldwide are small—fragmented by land use, isolated by development—then those populations may lack the evolutionary capacity to track changing conditions. Leimu and Fischer state this directly: the clear role of population size raises considerable doubt about the ability of small plant populations to cope with changing environments. That's not a rhetorical flourish; it's an inference from the data. Conservation strategies sometimes rely on the assumption that local genotypes are specially tuned to local conditions, which is why translocation programs have historically preferred to move plants from nearby source populations. But if fewer than half of pairwise comparisons show strict local adaptation, and small populations are especially unlikely to show it, then the premise underlying some of those strategies needs reexamination. The paper doesn't tell practitioners exactly what to do differently. It does indicate what not to take for granted. The broader takeaway is this: population size is not only a demographic variable affecting extinction risk; it fundamentally shapes evolutionary potential, and thus the capacity of plant populations to respond to a changing world. A population doesn't just risk extinction because it's small; it risks losing the capacity to evolve. Seventy-one percent sounds like good news for local adaptation. Forty-five percent tells you what the number actually means. 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.