Quantitative behaviour analysis requires the classification of behaviour to produce the basic data. This can be challenging when the theoretical taxonomy does not match observational limitations, or if a theoretical taxonomy is unavailable. Binary keys allow qualitative observation to be used to modify a theoretical taxonomy to produce a practical classification method, or to develop a pragmatic taxonomy in the absence of theory. Furthermore, inter-observer consistency can be maximized with less training, and construction of a binary key ensures that the categories into which behaviour is classified are complete and mutually exclusive. We give examples from special education research for the behaviour of children with developmental disabilities. (Contains 5 figures and 1 note.)