The paper discusses how the oldest manuscripts (L, C, V) of De civitate Dei are related to the more recent ones. The problems that emerge concern theexistence of an archetype; the relationship between L and C; the question whether the earlier manuscripts may be right (they sometimes are). In quite a few passages the readings preferred by editors are questionable, and others are proposed. It is shown that interpolation and revision of the text play an important role, and that the (...) context must take priority in attempting to establish the text. We should not put excessive trust in the older manuscripts, yet, at the same time, we should be aware of the uncertainty of our choices. (shrink)