The burgeoning empirical literature on constitutional drafting draws a distinction between constitutional amendment and constitutional replacement and, on the assumption that replacement is of greater magnitude and importance than amendment, treats the two as distinct phenomena. However, as an empirical matter, it is unclear whether this distinction is warranted, or whether replacements are in fact changes of greater magnitude than amendments. Using natural language processing (NLP) methods, this project empirically examines the magnitude of changes observed in both constitutional amendments and in constitutional changes (i.e. when a state introduces a new constitution). We analyse a text corpus we have constructed that enables us to compare “new” constitutions with “old” but amended constitutions. Our approach is to use the magnitude of textual change—which we can measure using NLP techniques—as a quantitative proxy for the magnitude of constitutional change.