Holmberg’s Generalization (Holmberg 1986) was originally stated to describe the “object shift” phenomena found in the modern Scandinavian languages. This dissertation argues that object shift is merely a subcase of scrambling, a type of adjunction, and that Holmberg’s Generalization is a subcase of a universal constraint, the “Generalized Holmberg Constraint” (GHC), which prohibits leftward scrambling across c-commanding functional heads. The existence of such a constraint turns out to have ramifications far beyond the analysis of scrambling itself, and the predictions it makes ultimately form an extended argument in favor of a universal antisymmetric approach to phrase structure (Kayne 1994). The most important evidence for the GHC comes from diachronic data. The study presents quantitative data from the history of Yiddish and English to show that, in cases where a language undergoes major changes in its clause structure, the GHC remains an active and stable constraint in the language, indicating its status as a universal. Once a phrase structure change begins, the resulting variation within a single speech community, and even within individuals, immediately shows the effect of the GHC on scrambling. The latter portion of the study argues that the GHC is not merely a constraint on scrambling, but rather a much more general constraint on the way syntactic computations progress, the “Conservation of C-Command.” The Conservation of C-Command finds a natural cross-linguistic formulation only if we adopt an antisymmetric approach to languages with head-final phrase structures. This approach turns out to have consequences for a variety of other problems of syntactic analysis, including the West Germanic Verb (Projection) Raising construction and Heavy NP Shift. This dissertation accounts for the typology of scrambling found in the world’s languages and during periods of language change, and shows that the way in which scrambling is constrained provides insight into basic properties of phrase structure. In addition, it constitutes an extended argument for the autonomy of syntax: while prosodic and pragmatic considerations favor leftward scrambling in a number of contexts, a language’s inventory of functional heads puts a strict upper bound on whether scrambling can respond to these considerations.