A good scientific experiment, like a good story, has a beginning, a middle, and an end (Galison, 1987). It is satisfying to describe the scientific method as a linear narrative beginning with hypotheses to be tested and then proceeding to experimental design, execution (funding, equipment and material procurement, set-up and manipulations, measurement and data collection, compilation of results), evaluation of evidence, and formulation of new hypotheses. In the occasional blockbuster scientific story, this process culminates in the emergence of a transformative new insight into nature—the recognition of the cell as the basic unit of life, of mitochondria and chloroplasts as evidence of past symbioses, of plants’ ability to turn CO₂ and sunlight into O₂ and sugars. This is rarely the way it happens, however. Real empirical practices turn out to be a good deal more complicated and a good deal less linear. The traditional story of scientific method leaves as a mystery the important question “Where do new hypotheses come from?” But like a bad television screenplay, the mystery is dissipated by focusing the plot elsewhere, on the problem of confirming or falsifying hypotheses—the logic of justification—rather than the psychology of discovery (Popper, 1959).

Each of the steps in this narrative is treated as a black box, when in fact both historical contingency and scientific judgment (in other words, the theoretical and conceptual framework within which the scientists are operating) are at work throughout the narrative, connecting the testing of hypotheses with the generation of new theory. For example, the technologies, protocols, and instruments that are chosen as means of experimentation also appear to have “life cycles.” Their endings or disappearance, like experimental methods in the broad sense, can come from anything from a change of interest, to new discoveries that render them obsolete, to new inventions or procedures that replace them. Decisions to use new instruments, to carry out experiments in new ways, or to take notice of odd or puzzling results do not come out of nowhere but instead are informed by the scientists’ theoretical framework. The ways in which experimental approaches evolve again hints at more complexity than the standard plot allows.

Scientific observation is likewise complex, although it is often thought of as no more than merely “looking.” To count as observation in science, “looking” usually requires a sophisticated approach, involving instruments and elaborate protocols embedded in technical practices that frame and shape both the observations and the reports of the results (Hacking, 1983). The things scientists want to observe are rarely easy to see, hear, taste, smell, or touch unaided by instruments or concepts. The things biologists want to observe are not only complex in their own rights but are embedded in complex structures or communities. Indeed, merely choosing what to