Constructing interdisciplinary knowledge is particularly difficult because scientific knowledge is situated in its discipline. Researchers must find common ground to share, and this causes high transaction costs. This article reports a method of conceptual analysis to elicit, analyse and compare conceptual models used by individual researchers, with the ultimate aim to facilitate researchers in sharing these models. Using this method, an analyst derives concepts and their mutual relations from documents, and defines them in natural language. We qualitatively analysed all writings of four researchers from an interdisciplinary research project and validated the resulting conceptual models in semi-structured interviews. The method was found to be effective in eliciting concepts, even those used implicitly. The interviews also revealed mechanisms by which researchers adopt new concepts or choose particular terms. The analysis costs are high but not prohibitive, and are shown to diminish with each researcher added to the analysis.