This article offers a new, transparent method to construe morphosyntactic categories for cross-linguistic research. It avoids the problem of categorial confusion attested in major post-Greenbergian studies in morphosyntactic typology, in particular in probabilistic typological investigations, which tend to mix up semantic and formal criteria and marginalize ‘statistically insignificant’ morphosyntactic variants. These and other problems are avoided by using functional criteria as the starting point in identifying comparable forms and constructions in different languages. Subsequently formal and semantic criteria are employed to arrive at a morphosyntactic category whose members are sufficiently similar in terms of function, form and meaning.