This dissertation develops a comprehensive theory of 'selective opacity', syntactic configurations in which one and the same syntactic domain is transparent to some operations, but opaque to others. The prime example of selective opacity are finite clauses in English, which are transparent to A'-movement, but opaque to A-movement. Following and extending the previous literature, I argue that selective opacity extends beyond the A/A'-distinction and even to syntactic dependencies that do not involve movement. Empirically, I argue that selective opacity exhibits intriguing meta-generalizations, which become evident once selective opacity across constructions and languages is treated as a uniform phenomenon. These two meta-generalizations are what I call 'Upward Entailment' and the 'Height-Locality Connection'. Upward Entailment states that if a clause of a given structural size is opaque to some operation, then structurally larger clauses are likewise opaque to this operation. The Height-Locality Connection states that the locality of a movement type is related to the height of the landing site of that movement type within the clausal spine. Specifically, movement types that land in a structurally high position are able to escape more domains than movement types that target a lower position. The core theoretical proposal is that selective opacity is the manifestation of a constraint on the locality of probes. I propose that probes have characteristic 'horizons', which delimit their search space. Crucially, horizons can differ between probes. As a result, the opacity of a domain can be relative to the probe conducting the search. I argue that this is what underlies locality differences between movement types, between movement and agreement, and between types of dependencies that do not involve movement. I demonstrate how a wide array of selective opacity effects and complex interactions between them can be derived. I also demonstrate how meta-generalizations of selective opacity emerge from the account. Finally, I explore the consequences of horizons for more familiar concepts of syntactic locality like phases. I show that horizons coexist with CP phases, but that they are incompatible with vP phases. Independent psycholinguistic evidence for this conclusion is provided and I reassess previous arguments in support of vP phases.