This paper illustrates how an "experimental state of mind", i.e. principles of experimental design, can inform hypothesis generation and testing in structured fieldwork elicitation. The application of these principles is demonstrated with case studies in toneme discovery. Pike's classic toneme discovery procedure is shown to be a special case of the application of experimental design. It is recast in two stages: (1) the inference of the hidden structure of tonemes based on unexplained variability in the pitch contour remaining, even after other sources of influence on the pitch contour are accounted for, and (2) the confirmation of systematic effects of hypothesized tonal classes on the pitch contour in elicitations structured to control for confounding variables that could obscure the relation between tonal classes and the pitch contour. Strategies for controlling the confounding variables, such as blocking and randomization, are discussed. The two stages are exemplified using data elicited from the early stages of toneme discovery in Kirikiri, a language of New Guinea.