“Truths We Don’t Suspect”: Binary Frameworks, Epistemic Excess, and the Production of Data About Nature Across Five Fields

Binary oppositions used in scientific practice—such as male/female for classifying organisms, natural/social for delimiting causal explanations, signal/noise as variations recorded by measurements, and normal/anomalous for sorting meaningful observations—do not merely describe a preclassified world. They shape what researchers expect to find, what instruments and databases are designed to record, what counts as reliable evidence, and what is relegated to error or exception. This exploratory, theory-building study draws on 10 in-depth interviews (2 per participant) with 5 researchers in paleoecology, microbial ecology, astronomy, glaciology, and human geography/climate science. Each participant’s interviews were integrated into a field-specific case, yielding five cases for cross-case analysis. I examined reported binary distinctions and related classificatory boundaries across three interconnected moments—observation, instrumentation, and data production—and distinguished their evidentiary status in each case. Binary thinking constitutes a particular form of categorization: it organizes phenomena into two mutually exclusive positions, often treated as exhaustive. When embedded in research procedures and infrastructures, these distinctions can shape what becomes observable and recordable, while rendering phenomena that do not fit either position harder to recognize, preserve, or investigate. Through abductive comparison, I conceptualize observations, ambiguities, and possibilities that do not fit previously established categories as epistemic excess and propose epistemic accountability as the practice of documenting classificatory decisions, preserving uncertainty and the production history of data, and keeping categories open to revision. The findings generate propositions for field-specific research rather than estimates of disciplinary prevalence and suggest that making classificatory choices visible may support more inclusive and reliable knowledge and policy.