The present paper surveys the diachronic development of the Ancient Greek perfect in four periods: Mycenaean, Archaic, Classical and post-Classical. At each stage the semantic evaluation of perfect is assessed in the context of the semantics of its predicate. While generally confirming the standard picture of increasing anteriority and past reference in the perfect correlating with greater numbers of verbs able to form perfects, the present study contributes empirical data to support this assertion. The article traces the growing paradigmatisation of the perfect form throughout its history. However, this development is not linear. Instead in the post-Classical language we witness a bifurcation along diglossic lines, with the literary language remaining much more conservative in terms of the perfect’s semantic range, while in lower-register material the perfect increasingly competes with the aorist to denote perfective semantics.