Educational studies that attempt to identify variables contributing significantly to student learning of science concepts often rely on careful construction of operational definitions as a means of controlling treatment variables. It is argued that this is not sufficient in a behavioral experiment such as in a science classroom because students may not behave as planned. In such research, quantitative treatment verification is necessary. A survey of the use of treatment verification in science teaching research is presented as well as one such procedure in detail. If the researcher finds significant differences between treatment groups on dependent measures, and if differences for student‐teacher behaviors between treatments are quantified and determined to differ significantly, the researcher has a much stronger case for arguing causation than if treatment verification had not been employed.