Although usage-based linguistics emphasises the need for studies of language change to take frequency effects into account, there is a lack of research that tries to systematically model frequency effects and their relation to diffusion processes in language change. This monograph offers a diachronic study of the change in Spanish perfect auxiliary selection between Old and Early Modern Spanish that led to the gradual replacement of the auxiliary ser `be' with the auxiliary haber `have'. It analyses this process in terms of the interaction between gradience, gradualness, and the conserving effects of frequency and persistence in language change. The study contributes to the theory and methodology of diachronic linguistics, additionally offering insights on how to explain synchronic grammatical variation both within a language and between languages. The book is of interest to the fields of Spanish and Romance linguistics, syntax, as well as historical and variationist linguistics.