This paper investigates how the Korean past marker -(e)ss (–ESS) is realized in non-final clauses preceding verbal connectives, focusing on temporal mismatch (a past first clause with a present/future second clause). Revisiting the debate between phrasal-affix analyses of –ESS in -ko coordination (J.-M. Yoon 1990, 1996; H.-S. Yoon 1993, 1994) and Chung’s (2005) null-tense proposal, I argue that neither uniform VP-coordination nor a simple coordination/subordination split captures the full distribution. A survey reveals a three-way pattern: only –ko and –kena (and –nikka) allow optional –ESS, other coordinators require it, and several subordinate connectives either require –ESS or categorically block it. I propose a connective-sensitive theory of tense-domain licensing in which connectives lexically select the T-profile of the non-final clause. -ko and -kena (and –nikka) are bifurcated between an overt-T coordination and a null-T dependent linkage, while -ese has a single null-T adverbial profile that blocks local past spell-out.
Myung-Kwan Park (Thu,) studied this question.