How do plants respond to environmental conditions? A common approach to finding an answer is to do an experiment in which the plant’s environment is varied. Experiments with plants can be simulated on a computer, performed in a greenhouse or growth cabinet, undertaken through field manipulations or conducted by observing what nature has come up with along environmental gradients or contrasts. The last of these, which I call ‘experiments by nature’, are the strongest in terms of realism, because they yield a long-term ‘ground truth’ beyond all initial effects of manipulative experiments. But, they are the most difficult to match into standardised protocols that permit statistics (Körner, 2018). One may, for example, have 1,000 mini-pots with Arabidopsis (rockcress) in a common garden, but whatever plant study one replicates on Kilimanjaro is pseudo-replicated because there is only one Kilimanjaro (n = 1), while all the garden pots are great from a purely statistical point of view.