In this dissertation I return to an old question within the generative grammar tradition, the that-trace effect: the incompatibility of the overt complementizer that with extraction from a subject position in an English embedded declarative clause. In common with other approaches to the that-trace effect, my analysis will place the that-trace effect within the larger context of the complementizer system of English. However, my main comparison will be with a similar pattern which I will call the ‘anti-that-trace effect: in non-wh relative clauses the null realization of the complementizer is incompatible with relativization over a subject. The two phenomena display a common characteristic which should not go unnoticed: a subject/non-subject asymmetry which is quite unique amongst the various phenomena that display restrictions in the distribution of complementizers. I argue that this empirical observation begs for a uniform analysis of the two phenomena. To achieve this I propose that the left periphery of an embedded clause encodes the semantic property of (non)-referentiality, which is structurally determined by the movement of the subject to the CP domain and which is phonologically realized in the overt/null form of the complementizer. Therefore, subject/non-subject asymmetries in the distribution of complementizers appear when the independent movement operations of extraction or relativization interfere with this mechanism. Furthermore, the contrast between the null-C requirement of the that-trace effect versus the overt-C requirement of the anti-that-trace effect is captured by the proposal that relativized nominals are featurally impoverished in comparison to a regular DP or a wh-phrase.