Observing some asymmetries in cross-clausal negative polarity licensing in Korean, Lee (2017) proposes that, under the SVO word order hypothesis (Kayne 1994), negative polarity items (NPIs) are licensed at the “Spec domain” of a negated predicate, rather than being regulated by the clause mate condition (CMC, Choe 1988, Lee 1994, Sohn 1995, Kim 1995, Chung 2006, a.o.). This paper argues, however, that Lee’s Spec domain approach faces some non-trivial difficulties and that a CMC-based theory like Sohn’s (1995) in terms of NPI movement to Spec of NegP more appropriately accommodates the cross-clausal NPI licensing facts.

The cognitive attitude verb KNOW in most languages typically selects for a factive complement (Kiparsky and Kiparsky 1970). It is noted in the literature (Lee 1978, 1999; Kiefer 1978, Őzyildiz 2017, a.o.), however, that KNOW in some languages may take various forms of complements and that factivity varies depending on the complement types. An obvious generalization made is that nominalized complements tend to convey a factive reading, while non-nominal ones tend not to (Kastner 2015). This work makes it clear that for a clause selected by KNOW to have a factive reading, it not only bears a nominal feature but also carries a structural case. This paper additionally points out the following three issues and discusses their theoretical implications as to the syntax and semantics of attitudinal predicate constructions: (i) Cognitive attitude verbs may simultaneously take a nominalized clause and a predicational clause; (ii) The non-factive KNOW in the three languages commonly displays neg-raising and naturally anti-rogativity, siding with doxastic (belief) verbs; (iii) Lexically negated forms of these verbs select only for a nominalized (factive) clause.

Ko (2015b) classifies Korean resultative small clause (RSCs) into four sub-types, due to combinations of two factors, i.e., the RSC's functional status (complement vs. adjunct) and the RSC subject's phonological status (overt vs. covert). She accounts for typological differences, especially movement-related behaviors, RSCs display, in terms of the theory of cyclic spell-out (Fox and Pesetsky 2005, Ko 2005). It will be shown in this paper, however, that her system faces non-trivial problems. First, the RSC subject's phonological status in fact does not contribute to the RSC typology, as subjects in any type can be suppressed in principle, given an appropriate context. Second, -key RSCs may function as a complement (not unambiguously as an adjunct). Third, no proper theory is provided for the ellipsis restrictions that the RSC displays. The present work proposes a control based RSC typology and tries to provide a unified explanation of restrictions on ellipsis as well as movement in RSCs, basically following Chung’s (2007, 2009, 2011) constituency based account of the syntactic restrictions.

Despite a variety of accounts proposed to capture the grammatical properties that the so-called dummy plural maker (DPM) -tul in Korean displays, it has been commonly taken for granted that a DPMed phrase is to be c-commanded by its associate plural element (APE). This paper observes, however, that the APE itself can host DPM -tul. This fact calls for a novel licensing mechanism, as any account that resorts to a direct c-command relation between a DPMed phrase and its APE will fail. This paper proposes that there is a functional category that Agrees with the APE. This is plausible because the semantic interpretation of the DPM construction has more to do with a semantically plural element, i.e., the APE, rather than a DPMed category. Then DPM -tul on the APE can be naturally considered as a phonological realization of the Agree relation. DPM -tul on a non-APE element is viewed merely as a copy due to an operation called Spread, much like the distribution of negative morphemes in negative concord languages, except for the optionality of the phonological realization.