This dissertation analyzes an unusual grammatical pattern that I call locative determiner omission, which is found in several languages belonging to the Slavic, Romance, and Albanian families, but which does not appear to have been directly inherited from any individual genetic ancestor of these languages. Locative determiner omission involves the omission of a definite article in the context of a locative prepositional phrase, and stands out as a feature of the Balkan linguistic area for which there are few, if any crosslinguistic parallels. This investigation of the origin and diachronic spread of locative determiner omission serves the particular goal of revealing how the social context of language contact could have resulted in a pattern of grammatical borrowing without lexical borrowing, yielding a present distribution in which locative determiner omission appears in several Balkan languages no longer in direct contact with one another. A detailed structural and historical analysis of locative determiner omission in Albanian, Romanian, Aromanian, and Macedonian is used as a basis for comparison with other Balkan languages. The analysis pays particular attention to the sequence of grammatical changes necessary for the outcome of locative determiner omission in each language, and the specific sociocultural configurations between speaker communities at relevant historical stages that allowed for the spread of locative determiner omission without direct lexical borrowing. This makes it possible to establish that locative determiner omission arose from a period of early contact between proto-Albanian and Late Latin, resulting in the generalization of the structure across all branches of Balkan Romance. During a later period, contact between Aromanian and individual dialects of Albanian and Macedonian resulted in the transfer of this feature, which then spread throughout most, but not all dialects of these latter languages.
A methodological contribution of this dissertation is the demonstration that in-depth study of a grammatical feature that is suspected to have developed through language contact can yield important insights into the historical and social dynamics of a linguistic area that cannot be determined through synchronic observation of broad similarities alone. Even in the absence of documentation, careful reconstruction of the structural accommodations required for the adoption of a grammatical innovation can reveal new information about the process of language contact. This is particularly true for features that are not uniformly distributed across a linguistic area, as is the case with locative determiner omission in the Balkans. As a consequence, my proposal argues for an approach to linguistic areas that views them as an outcome of localized, layered clusters of convergence.