In many experimental sciences, like particle physics or molecular biology, the proper place for establishing facts is the laboratory. In the sciences of population biology, however, the laboratory is often seen as a poor approximation of what occurs in nature. Results obtained in the field are usually more convincing. This raises special problems: it is much more difficult to obtain stable, repeatable results in the field, where environmental conditions vary out of the experimenter’s control, than in the laboratory. We examine (...) here how this problem affected an influential experimental research tradition in community ecology, the study of the ecology of the rocky seashores. In the 1960s, a handful of North-American ecologists, most notably Joseph Connell, Robert Paine and Paul Dayton, made the rocky seashores a model study system for experimenting in the field. Their experiments were deceptively simple: they removed species living on the seashore and described the resulting effects on the local ecology. These experiments exerted a deep influence on community ecology. They provided evidence for speculative developments concerning the theory of interspecific competition, the factors responsible for species richness and the ecology of food webs. They also stimulated novel conceptual developments. In particular, Paine developed the predation hypothesis, which states that the presence of predators can favour species richness, before introducing the keystone species concept, according to which some species exert disproportionate effects on ecological systems. More broadly, these experiments gave support to a methodological trend in favour of field experimentation. Only controlled perturbations in the field, it seemed, provided a reliable method to get insights into the structure of ecological communities. However, as experiments were continued in time and repeated in different sites, divergent results appeared. We analyse here how intertidal researchers coped with the variability of environmental conditions and tried to stabilize their results. In the process, they reconsidered not only their early conclusions, but also the exclusive status given to field experiments. Expanding on this case study, we discuss some significant differences between laboratory and field experiments. (shrink)