(Anti-)That-Trace Effects Revisited: Labeling at the Interface via Externalization

Abstract

This paper reanalyzes the (anti-)that-trace effects in English within an interfacesensitive framework and proposes the Externalization-Based Labeling Condition, which requires that exactly one morphophonologically overt element appear at the edge of CP for successful labeling. I argue that the classic that-trace effect does not arise from labeling failure per se, but from anti-locality violations that are activated only when the CP phase is visible at the CI interface. Specifically, when the complementizer that is overt, labeling at the CP edge succeeds; however, this success renders the phase visible, thereby triggering anti-locality constraints on subject movement (Douglas 2017; Deal 2016; Erlewine 2020). In contrast, when that is covert, labeling fails due to the lack of an overt anchor, but the resulting structure becomes invisible at the interface, allowing the derivation to bypass anti-locality. The anti-that-trace effect in subject relatives, on the other hand, arises from the absence of any externalized element in configurations lacking that, resulting in labeling failure. Here, that becomes obligatory as the sole visible anchor for successful CP labeling. The optionality of that in object relatives is derived from the dual availability of operator-based and head-based labeling strategies (Ha 2020). By distinguishing interface-visible from interface-invisible configurations, this approach explains when and why anti-locality applies, offering a unified analysis of (anti-)that-trace effects grounded in externalization and the interface-based computation of labeling.