Chapter 10 attempts to offer a number of historical hypotheses that could explain why the geographical distribution of certain agreement patterns appears to be a priori unexpected. Since verbal agreement has been shown to be very common and evenly distributed among the world's languages, the crucial fact in need of an explanation is the distribution of languages with adnominal agreement, which is areally rather limited. We look at a number of well documented or reasonably well reconstructed cases, including Zande (an Ubangian language), Nilo-Saharan, Daly languages of North Australia, Proto-Indo-European, etc., and discuss the attested and probable paths in the development of adnominal agreement. It is argued that agreement often spreads from the clausal domain, where it is pragmatically motivated, to the domain of the NP, where it is largely redundant.