1 Introduction

Finiteness is a morphosyntactic notion inherited from traditional grammar. Clauses are classified into finite and non-finite, typically based on verbal morphology. Traditionally, finite clauses are characterized by their full inflection for tense and agreement, hold independent propositional content (Rizzi 1997; Kornfilt 2007; Nikolaeva 2010; Eide 2016; among others) and set syntactic operation domains with an association to the subject (Kornfilt 2007; Nikolaeva 2010).

Rizzi (1997) locates finiteness in the CP-domain’s Fin head, which is tied to T(ense). This suggestion reflects a tight interconnection between semantic (i.e., temporal) and syntactic notions of finiteness.1 A clause is either finite or non-finite and this contrast correlates to the type of T as tensed or non-tensed, given that Fin selects for T. For example, in English, C in clausal complements, headed by that and for select finite and non-tensed infinitival clauses, respectively (Chomsky & Lasnik 1977: 444–450), while in Italian che selects tensed finite clauses (Rizzi 1997).

However, tense marking and subject agreement do not have to be realized together in a single functional head. Languages with separate heads for subject and tense raise an interesting question about associating finiteness with T. In theories where tense and agreement occupy separate functional heads, T and Agr (Pollock 1989; Cinque & Rizzi 2009), Agr’s presence without T predicts clausal finiteness differently. In European Portuguese (Raposo 1987) and Turkish (Kornfilt 2007), embedded clauses without a tense-marked predicate are non-finite in morphological and semantic senses but are syntactically finite in terms of subject licensing and constituting local domains for syntactic operations. In other words, the properties of syntactic finiteness, i.e., subjecthood and opacity, go together, but they do not necessarily align with semantic finiteness, as characterized by independent temporal interpretation. Furthermore, following the seminal work of Wurmbrand (2001), subsequent studies (Adger 2007; Kornfilt 2007; Todorović & Wurmbrand 2020; among others) observe that smaller-than-TP phrases show misalignment between syntactic and semantic notions of finiteness. They argue that T does not solely decide finiteness (pace Rizzi 1997) and domains without T can still be syntactically finite.

This paper focuses on syntactic finiteness and investigates the minimal domain for finiteness with data from Korean, drawing parallels with Turkish, where Agr’s presence dictates finiteness regardless of T (Kornfilt 2007). Analyzing various clausal complements in Korean, this study finds that subject agreement, specifically in Hon(orific) P(hrase), marks a clause as syntactically finite.

As a brief sketch of the items under discussion, subject agreement in Korean is realized as an honorific marker -si as in (1) (Choe 1988; Ryu 1995; Choi 2010; among others). The honorific marker is suffixed to the verbal when the subject has the feature of honourability. For instance, the subject apeci ‘father’ in (1a) is an honourable referent, and the verb agrees with the noun in honorification by suffixing -si to the stem. A non-honourable subject like kkoma ‘kid’ does not induce the -si marking. As shown in (1b), suffixing -si is ungrammatical here.

    1. (1)
    1. Korean honorific agreement
    1.  
    1. a.
    1. Honourable subject agrees with the verbal, and -si marked
    1. Apeci
    2. Father
    1. ket-si-ess-ta.
    2. walk-h-pst-decl
    1. ‘Father walked.’
    1.  
    1. b.
    1. Non-honourable subject agrees with the verbal, and -si is not marked
    1. Kkoma
    2. Kid
    1. ket(*-si)-ess-ta.
    2. walk(*-h)-pst-decl
    1. ‘A kid walked.’

As for the working definition of syntactic finiteness, I follow Kornfilt (2007) in assuming that a domain is syntactically finite if it shows opacity in terms of anaphor binding and licensing of negative polarity items (NPI). A clause is non-finite if an element outside of it can bind anaphors or license NPI across the clause boundary. This study shows that the syntactic domains that contain the equivalent of [Agr] in Korean, i.e., HonP and larger phrases, are syntactically opaque, but a domain smaller than HonP is transparent. Cross-clausal licensing of NPIs and anaphor binding are impossible if the embedded clause is HonP or larger. In other words, HonP is the locus of finiteness and the smallest finite domain.

The organization of the paper is as follows. In section 2, I provide an analysis of Korean phrase structure and introduce the types of clausal complement selected by different verb classes. Then, I present the functional head Hon as the licenser of the nominative subject. Sections 3 and 4 deal with clausal complements that are larger than TP. I show that they are finite, based on the application of NPI and binding domain tests. Section 5 provides the main argument that HonP is the smallest domain for syntactic finiteness in Korean by showing that the syntactic diagnostics provide different results between HonP and vP complements. Finally, Section 6 summarizes the study.

The Korean data presented in this paper are either drawn from previous studies, with citations provided where applicable, or created based on the grammatical judgments collected from three native speakers of Korean (all female, aged 31–36). The consultants were asked to judge the grammaticality and acceptability of Korean sentences presented in written form. For cases involving adverbial ordering, intended contexts were provided. When a sentence was judged either ungrammatical or unacceptable, the consultants were asked to elaborate on their linguistic intuitions. While the judgments were collected informally, all consultants are native speakers with no significant exposure to foreign languages, and they were unaware of the purpose of the study. The informal nature of the elicitation process and the small sample size may introduce certain limitations. Nonetheless, their grammatical judgments generally aligned with prior observations in the literature. I explicitly describe instances of misalignments in the text.

2 Background

2.1 Functional projections in Korean

Korean, an agglutinative head-final language with an SOV word order, presents a complex verbal morphology involving the verbal stem and inflectional elements. There is no conventional agreement in φ-features between the verb and its arguments, but previous studies view subject honorification as an instance of agreement (Choe 1988; Ryu 1995; Choi 2010; among others). Honorification as agreement is an essential assumption in this paper and is discussed in Section 2.2.

Following previous studies (Kang 1988; Ryu 1995; Cinque 1999: 55–56; Koopman 2005; Choi 2010; Suh 2017; among others), I assume that verbal forms in Korean are syntactically complex and each verbal suffix is an instance of a functional projection. A schematic representation is given in (2). The verbal stem tonguiha- ‘agree’ is suffixed by -si, which is one of the two phonetic forms of the Hon(orific) head. Hon is realized as either -si or the null form (∅). -si is selected if the subject is a person to be honoured, and ∅ is selected otherwise. -si is used in (2) because the subject apeci ‘father’ is an honourable noun. Hon is followed by -ess for the past tense, -keyss for epistemic modality (conjecture) and -ta for speech act mood (declarative).2 If the order of verbal morphology reflects the syntactic hierarchy in the verbal and inflectional domains (Baker 1985), the Hon head is positioned right above the verb stem, but below T and Mod.

    1. (2)
    1. Functional projections in Korean (Ryu 1995; Kim et al. 2018: 101–124)3
    1. [[[[HonP
    2.  
    1. Apeci-kkeyse [Hon'
    2. Father-hnom
    1. [tongui-ha vP]
    2.   agree-vblz
    1. -si
    2. -h
    1. Hon'] HonP]
    2.  
    1. -essTP]
    2. -pst
    1. -keyss ModP]
    2. -mod
    1. -taMoodP]
    2. -decl
    1. ‘Father might have agreed.’

This study compares different embedded clauses and investigates what the smallest finite domain is. It builds on the idea proposed in Wurmbrand (2001), developed in her subsequent work and further surveyed cross-linguistically in Laszakovits and Shen (2021), that different types of matrix verbs select different-sized clausal complements, and the complement size can range from VP to CP. In Korean, clausal complements of different sizes are suffixed with different morphemes. For example, malha- ‘tell’ takes a clausal complement that ends with -ko, pala- ‘wish’ takes a nominalized complement that ends with -ki, and mantul- ‘make’ and po- ‘try’ take complements that end with -key and -e, respectively. Table 1 summarizes these morphological distinctions.4 Since Korean is a head-final language, these morphemes can be the functional head that dominates the whole embedded domain. Previous studies (Kang 1988; Ryu 1995; Suh 2017; Kim et al. 2018: 101–124; among others) regard all of them as complementizers. If they are on the right track, all verbs take CP complements, but some embeddings lack lower functional projections beneath C. I will crucially show that not all of them are CPs, as well as that some smaller-than-CP complements are finite.

Table 1

Verb types and their clausal complements.

Verb type

Clausal complement
malha- ‘to say’ [CP ___ -ko]                           V
pala- ‘to wish’ [DP [TP [HonP ___ ] ]-ki]      V
ha-/mantul- ‘to make (long causative)’ [HonP [vP ___ ] ]-key             V
po- ‘to try’ [vP ___ ]-e                              V

Examples of different embeddings are given in (3). (3a) shows malha- ‘say’ with a finite CP complement. A wish-type verb in Korean takes a nominalized clause as in (3b). I follow previous studies (Kang 2001; Park 2009; Park 2016; Kim et al. 2018: 136–140; among others) assuming that nominalized clauses in Korean are DPs. -ko in (3a) is a complementizer and -ki in (3b) a nominalizer (Sohn 1994: 345–346; Suh 2006: 140). Apart from these two cases, I will not attempt to identify the nature of the subordinating heads.5 More important than identifying heads is that only some types of the smaller clausal complements allow cross-clausal syntactic operations. I argue that some smaller-than-CP complements are syntactically finite while others are not. The finiteness of the embedded clause depends on its size; particularly, on whether it is large enough to include an element for subject agreement and licensing. To this end, I will ascertain for each type of complement clauses whether (i) it is as large as CP, and (ii) whether it is finite. It will turn out that HonP and larger domains, e.g., (3a–c) below, are finite, while a domain which is smaller than HonP, e.g., (3d), is not. In other words, the contrast between (3c) and (3d) is critical.

    1. (3)
    1. Clausal complements in Korean
    1.  
    1. a.
    1. Full CP embedding, containing MoodP and everything below it
    1. Na-nun
    2. 1sg-top
    1. [cp
    2.  
    1. apeci-kkeyse
    2. father-hnom
    1. pap-ul
    2. meal-acc
    1. mek-si-ess-keyss-ta
    2. eat-h-pst-mod-decl
    1. -ko]
    2. -c
    1. malha-∅-ess-ta.
    2. say-nh-pst-decl
    1. ‘I said that my father might have had a meal.’
    1.  
    1. b.
    1. Nominalized clausal complement, containing TP and everything below it
    1. Na-nun
    2. 1sg-top
    1. [dp [tp
    2.  
    1. apeci-kkeyse
    2. father-hnom
    1. pap-ul
    2. meal-acc
    1. mek-si-ess]
    2. eat-h-pst
    1. (*-keyss) (*-ta) -ki]-lul
    2. (*-mod) (*decl)-nmlz-acc
    1. pala-∅-ess-ta.
    2. wish-nh-pst-decl
    1. ‘I wished my father had a meal.’
    1.  
    1. c.
    1. HonP embedding with -key, containing HonP and everything below it
    1. Mina-nun
    2. Mina-top
    1. [HonP
    2.  
    1. apeci-kkeyse
    2. father-hnom
    1. pap-ul
    2. meal-acc
    1. mek-si] (*-ess/*-l) (*-keyss)
    2. eat-h (*-pst/*-fut) (*-mod)
    1. (*-ta) -key
    2. (*-decl)-key
    1. mantul-∅-ess-ta.
    2. make-nh-pst-decl
    1. ‘Mina made her father have a meal.’
    1.  
    1. d.
    1. vP embedding with -e
    1. Apeci-kkeyse
    2. Father-hnom
    1. [vP
    2.  
    1. (*Mary-ka/*ku-ka/*caki-casin-i)
    2. (*Mary-nom/*(s)he-nom/*self-nom)
    1. pule-lul
    2. French-acc
    1. kongpu-ha]
    2. study-vblz
    1. (*-si) -e]
    2. (*-h) -e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1. ‘Father tried learning French.’

2.2 Korean subject-verb agreement and Hon as the nominative Case licenser

Subject honorification in Korean, realized by the nominal marker -kkeyse and the verbal suffix -si, has been regarded as an instance of agreement in both theoretical studies (Choe 1988; Kang 1988; Ryu 1994; 1995; Choi 2010) and empirical research (Kwon & Sturt 2016; 2023).6 Choe (1988: 112–113) shows that whether the subject is honourable or not aligns with the presence of -si and claims that the subject and predicate should agree in the [Hon(oured)] feature. The examples in (4) provide a snapshot of Korean subject-verb agreement. They illustrate the interaction between nominal [Hon] features and the verbal suffixes. In (4a), the subject apeci ‘father’ contains a [+Hon] feature. Therefore, the same feature of Hon on the verb should also be [+Hon], in agreement with the subject. The [+Hon] in the verbal suffix is phonetically realized as -si. In contrast, (4b) is ungrammatical because [Hon] values do not match between the subject and verb.7 The -si suffix is not allowed, as the subject Chelswu is [–Hon]. (4b) is repaired in (4c), where the null form is used instead of -si. The null form is the realization of [–Hon].8

    1. (4)
    1. Korean honorific agreement
    1.  
    1. a.
    1.   The subject contains [+Hon] and Hon on the verb is realized as -si.
    1.   Apeci-kkeyse
    2.   [+H]
    3.   father-hnom
    1. cip-ey
    2.  
    3. home-at
    1. o-si-ess-ta.
    2. [+H]
    3. come-h-pst-decl
    1.   ‘(My) father has come home.’             (= (35a) in Choe (1988: 113))
    1.  
    1. b.
    1.   Subject contains [–Hon] and -si is not allowed
    1. *Chelswu-ka
    2.   [–H]
    3.   Chelswu-nom
    1. Nyuyok-ey
    2.  
    3. New.York-to
    1. ka-si-ess-ta.
    2. [+H]
    3. go-h-pst-decl
    1.   ‘Chelswu went to New York.’             (= (36b) in Choe (1988: 113))
    1.  
    1. c.
    1.   The null form for the Hon head
    1.   Chelswu-ka
    2.   [–H]
    3.   Chelswu-nom
    1. Nyuyok-ey
    2.  
    3. New.York-to
    1. ka-∅-ess-ta.
    2. [–H]
    3. go-nh-pst-decl
    1.   ‘Chelswu went to New York.’

Building on the demonstration of the subject-verb agreement in (4), I now present the distribution of the honorific adverb chinhi ‘in person’ to emphasize the structural (rather than pragmatic) nature of -si. The adverb chinhi is a pragmatic marker used for honouring the agent of an action, emphasizing that the action was performed by a person of higher status. In contrast to -si, it is sensitive to the pragmatic context. Most importantly, it can appear when -si is ungrammatical, as long as the pragmatic context requires it.9 Example (5a) illustrates chinhi’s use without -si for the first-person subject. When the speaker is higher than the hearer in social hierarchy, there is a pragmatic need for honouring the subject, making chinhi applicable. However, since the honorific agreement feature is tied to third person, -si is not allowed with the first-person subject. (5b) shows that chinhi is infelicitous when the subject is kkoma ‘kid.’ Finally, (5c) shows that the honorific adverb can co-occur with -si as apeci ‘father’ is an honourable noun. In short, the honorific adverb’s distribution in (5) rules out the possibility that the verbal suffix -si is merely a pragmatic device for signalling dynamic and performative social relations.

    1. (5)
    1. Honorific agreement vs. honorific adverb
    1.  
    1. a.
    1. Context: father talking to his son about his experience
    1. Nay-ka
    2. 1sg-nom
    1. chinhi
    2. in.person
    1. phyenci-lul
    2. letter-acc
    1. ssu (*-si) -ess -ta.
    2. write(*-h)-pst-decl
    1. ‘I wrote the letter myself.’             (See (210c) in Koo et al. (2015: 297))
    1.  
    1. b.
    1. The honorific adverb is infelicitous with a non-honourable agent
    1. Kkoma-ka
    2. Kid-nom
    1. (*chinhi)
    2. (*in.person)
    1. phyenci-lul
    2. letter-acc
    1. ssu-ess-ta.
    2. write-pst-decl
    1. ‘The kid wrote the letter himself/herself.’
    1.  
    1. c.
    1. The honorific adverb occurs with -si
    1. Apeci-kkeyse
    2. 1sg-nom
    1. chinhi
    2. in.person
    1. phyenci-lul
    2. letter-acc
    1. ssu-si-ess -ta.
    2. write-h-pst-decl
    1. ‘Father wrote the letter himself.’

In addition to distributional evidence described in (4) and (5), Kwon & Sturt (2016; 2023) provide experimental evidence that the distribution of Korean honorifics is structural rather than pragmatic. Kwon & Sturt’s (2016) reading-task experiments showed slower processing for honorific mismatches and an attraction effect in embedded constructions, akin to subject-verb agreement effects in English. The results indicate a structural basis of Korean honorifics. Furthermore, Kwon & Sturt’s (2023) neurolinguistic study reinforces this through the P600 effect observed with honorific mismatches, alongside other observable reactions to ungrammaticality such as lower naturalness ratings and longer reading times. P600 is an event-related potential (ERP) signal evoked when trying to process an ungrammatical sentence.10 The absence of signals like the N400 wave in honorific feature mismatches suggests that the use of honorific markers alongside honourable subjects in Korean is fundamentally structural, distinguishing it from Japanese and Chinese honorifics.

Having confirmed the structural nature of the subject honorification, I now turn to its role in nominative Case licensing. In Korean, Hon, rather than T licenses nominative Case on the subject (Ryu 1995; Choi 2010; Kim et al. 2018: 233). Evidence comes from the presence of a nominative subject in T-less clauses in (6) below. (6a) illustrates HonP embedding, and (6b–c) are instances of the tenseless construction (cf. Choe (1988: 114)). The clausal complement in (6a) must be smaller than TP because the presence of a tense (*-ess/*-l) marker makes it ungrammatical. Even in the absence of T, the embedded clause has a nominative subject apeci ‘father.’ Likewise, the clause in (6b) must be smaller than TP because adding a tense marker is ungrammatical, as in (6c). Still, the nominative subject apeci ‘father’ is licensed, since (6b) contains Hon.

    1. (6)
    1. a.
    1.   Causative construction (= 3c)
    1.   Mina-nun
    2.   Mina-top
    1. [keyP [HonP
    2.  
    1. apeci-kkeyse
    2. father-hnom
    1. pap-ul
    2. meal-acc
    1. mek-si] (*-ess/*-l)
    2. eat-h (*-pst/*-fut)
    1. -key]
    2. -key
    1. mantul-∅-ess-ta.
    2. make-nh-pst-decl
    1.   ‘Mina made her father have a meal.’
    1.  
    1. b.
    1.   ‘Tenseless infinitival’ construction
    1.   [Kulim-ey
    2.   painting-to
    1. apeci-kkeyse
    2. father-hnom
    1. hungmi-lul
    2. interest-acc
    1. kaci-si-ki]-lan
    2. have-h-nmlz-as.for
    1. tangyenhan
    2. natural
    1. il-i-ta.
    2. matter-cop-decl
    1.   ‘As for father’s having an interest in painting, that is natural.’    (see (37a) in Choe (1988: 114))
    1.  
    1. c.
    1.   (6b) is ungrammatical with Tense
    1. *[Kulim-ey
    2.   painting-to
    1. apeci-kkeyse
    2. father-hnom
    1. hungmi-lul
    2. interest-acc
    1. kaci-si-ess-ki]-lan
    2. have-h-pst-nmlz-as.for
    1. tangyenhan
    2. natural
    1. il-i-ta.
    2. matter-cop-decl
    1.   ‘It is natural for the father to have had an interest in painting.’ (intended)

The Korean absolute construction (Im 1983; 2007; Jeong 2020) provides another piece of evidence that the subject is not licensed by T.11 (7a–b) are instances of the absolute construction. Even though it is a root clause, the absolute construction is smaller than TP for morphological and semantic reasons. Morphologically, the absolute constructions lack explicit tense marking.12 Semantically, if we assume that a fixed temporal interpretation comes from T, the lack of temporal reference in (7) indicates that there is no T in (7a–b). However, the nominative subject is still allowed.

    1. (7)
    1. The Korean absolute constructions
    1.  
    1. a.
    1. Apeci-kkyese
    2. father-hnom
    1. cip-ey
    2. home-to
    1. o-si-ta.
    2. come-h-decl
    1. ‘Father comes/came/will come home (event time not denoted).’
    1.  
    1. b.
    1. Aki-ka
    2. baby-nom
    1. cip-ey
    2. home-to
    1. o-∅-ta.
    2. come-nh-decl
    1. ‘The baby comes home (event time not denoted).’

Furthermore, T in Korean has been argued to license a topic, instead of the subject (Ryu 1995: 57–61), as the contrast between (8a–b) shows. (8b) is tensed and (8a) is a tenseless clause. In both, the subject apeci ‘father’ is licensed by Hon. The topic Yengswu, on the other hand, is not allowed in the absolute construction (8a), whereas it is allowed in the tensed construction (8b) with the past tense marker -ess and with a topic marker -nun.

    1. (8)
    1. T licenses the topic
    1.  
    1. a.
    1. *Yengswu-nun
    2. *Yengswu-top
    1. apeci-kkyese
    2. father-hnom
    1. cip-ey
    2. home-to
    1. o-si-ta.
    2. come-h-decl
    1.   ‘As for Yengswu, his father comes home (intended, event time is not denoted).’   (= (46a) in Ryu (1995: 59))
    1.  
    1. b.
    1.   Yengswu-nun
    2.   Yengswu-top
    1. apeci-kkyse
    2. father-hnom
    1. cip-ey
    2. home-to
    1. o-si-ess-ta.
    2. come-hon-pst-decl
    1.   ‘As for Yengswu, his father came home (event time denoted).’

This section has introduced my assumption that Korean verbs select clausal complements of varying sizes and that the functional head Hon, structurally positioned below T, agrees with the subject noun and licenses nominative Case. If subject licensing is a sufficient condition for syntactic finiteness (Kornfilt 2007), a projection of Hon should constitute the smallest finite clause even without T. The following sections will demonstrate that HonP is opaque for cross-clausal operations, and by hypothesis, that it is syntactically finite.

3 Fully finite embedded clauses as the baseline

In this section, I analyze the embedded domain introduced by -ko as the baseline for a finite clause. Following the literature, these -ko embeddings are assumed to be full-fledged CPs and therefore finite. I apply two diagnostics of syntactic finiteness, i.e., cross-clausal negative polarity item (NPI) licensing and reflexive binding, to these -ko clauses. As expected, the results show that -ko CP complements are finite, validating the two diagnostic tests.

3.1 Clause-external negation cannot license an NPI in an embedded CP

My first diagnostic for finiteness is cross-clausal NPI-licensing. I use the negative polarity item (NPI) amuto ‘anyone.’ In Korean, NPIs are licensed by clausemate negation a within CP (Choe 1988; Kuno 1998). The clausemate requirement stipulates that the NPI and negation must be in the same CP domain. That is to say, an NPI cannot be licensed by superordinate negation outside of its minimal CP. The example in (9) satisfies the clausemate requirement with the NPI amuto ‘anyone’ and the negation an in the same clause.13

    1. (9)
    1. Negation and NPI in the simple sentences
    1. Apeci-kkyese
    2. father-hnom
    1. amuto
    2. anyone
    1. an
    2. neg
    1. pulu-si-ess-ta.
    2. call-h-pst-decl
    1. ‘Father did not call anybody.’

Consider now -ko clauses in (10a–b) below. In the grammatical example in (10a), both NPI and negation are in the embedded clause. In (10b), however, the NPI is in the embedded clause, while the negation is in the matrix clause. The ungrammaticality of (10b) indicates that the embedded NPI cannot be licensed by the matrix negation, and that the matrix and embedded clauses do not count as the same clausal level for NPI licensing. The embedded clause is syntactically opaque for polarity licensing. If, by hypothesis, such opacity is an indicator of finiteness, then -ko embedding is finite.

    1. (10)
    1. Negation and NPI in complex sentences
    1.  
    1. a.
    1.   Clausemate condition satisfied (both negation and NPI in the embedded clause)
    1.   Na-nun
    2.   1sg-top
    1. [CP
    2.  
    1. Mina-ka
    2. Mina-nom
    1. amuto
    2. anyone
    1. an
    2. neg
    1. pulu-∅-ess-ta-ko]
    2. call-nh-pst-decl-c
    1. malha-∅-ess-ta.
    2. say-nh-pst-decl
    1.   ‘I said that Mina did not call anyone.’
    1.  
    1. b.
    1. Clausemate condition violated (negation in the matrix clause and NPI in the embedded clause)
    1. *Na-nun
    2.   1sg-top
    1. [CP
    2.  
    1. Mina-ka
    2. Mina-nom
    1. amuto
    2. anyone
    1. pulu-ess-ta-ko]
    2. call-pst-decl-c
    1. an
    2. neg
    1. malha-∅-ess-ta.
    2. say-nh-pst-decl
    1.   ‘I didn’t say that anybody had a meal.’

3.2 Embedded CP is the local domain for binding purposes

The next syntactic diagnostic for finiteness is anaphor binding, using the local anaphor caki-casin ‘self’ (Kim & Yoon 2009). A clause is judged finite if the anaphor is bound clause-internally while a pronoun is free. In this section, I show that the clausal complement of malha- ‘say’ is an opaque domain for binding, and it is thus finite.

Consider first the simple clauses in (11). The examples in (11a–b) show how Binding Conditions A and B are satisfied in a simple clause: the anaphor is bound, and the pronoun is free in this finite domain. In (11a), the anaphor caki-casin is c-commanded by the antecedent Naun, a proper noun that refers to an individual. Therefore, the antecedent Naun can bind caki-casin. On the other hand, the pronoun ku ‘(s)he’ in (11b) cannot refer to the subject.

    1. (11)
    1. a.
    1. Anaphor in a simple sentence (Condition A)
    1. Nauni-i
    2. Naun-nom
    1. caki casini-ul
    2. self-acc
    1. cohaha-∅-n-ta.
    2. like-nh-prs-decl
    1. ‘Naun likes herself.’
    1.  
    1. b.
    1. A pronoun in a simple sentence (Condition B)
    1. Nauni-i
    2. Naun-nom
    1. ku*i/j-lul
    2. 3sg-acc
    1. cohaha-∅-n-ta.
    2. like-nh-prs-decl.
    1. ‘Nauni likes himj/herj.’

Now, consider the complex sentences in (12a–c). In (12a), only Naun in the embedded clause, not Mina, can bind the anaphor caki-casin. Contrastingly, (12b) shows that the third-person pronoun kunye ‘she’ cannot refer to Naun but can to Mina, reversing (12a)’s binding dynamics. Finally, (12c) underscores that the matrix subject Naun cannot bind an embedded subject anaphor. In short, the embedded clause is the domain within which the anaphor should be bound (12a, c) and the pronoun should be free (12b). Therefore, the clausal complement of malha- ‘say’ is the local domain for binding. If closing off the binding domain correlates with finiteness, then this is an indication of the finiteness of the -ko clause.

    1. (12)
    1. a.
    1. Minai-ka
    2. Mina-nom
    1. [CP
    2.  
    1. Naunj-i
    2. Naun-nom
    1. caki casin*i/j-ul
    2. self-acc
    1. cohaha-∅-n-ta-ko]
    2. like-nh-prs-decl-c
    1. malha-∅-n-ta.
    2. say-nh-prs-decl
    1. ‘Mina says that Naunj likes herselfj.’ but not ‘Minai says that Naun likes heri.’
    1.  
    1. b.
    1.   Minai-ka
    2.   Mina-nom
    1. [CP
    2.  
    1. Naunj-i
    2. Naun-nom
    1. kunyei/*j/k-lul
    2. 3sg-acc
    1. cohaha-∅-n-ta-ko]
    2. like-nh-prs-decl-c
    1. malha-∅-n-ta.
    2. say-nh-prs-decl
    1. ‘Mina say that Nauni likes him/herj.’
    1.  
    1. c.
    1. *Nauni-i
    2.   Naun-nom
    1. [CP
    2.  
    1. caki casini-i
    2. self-nom
    1. kay-lul
    2. dog-acc
    1. cohaha-∅-n-ta-ko]
    2. like-nh-prs-decl-c
    1. malha-∅-n-ta.
    2. say-nh-prs-decl
    1. ‘Nauni says that shei likes the dog.’ (intended)

4 TP embedding is syntactically opaque

In this section, I discuss the nominalized clausal complement of the verb pala- ‘wish.’ Examples in (13) show the complement of pala- is DP with an embedded TP. The nominalizer -ki turns the TP into a nominal (Sohn 1994: 345–346; Suh 2006: 140). This section provides morphological and semantic evidence that this nominalized complement cannot contain higher phrases like ModP or MoodP.14 Then, I apply the adverb ordering test to show that the complement DP does not include a CP domain. Importantly, I will show that this type of embedding behaves as finite in terms of NPI licensing and binding.

The nominalized complement of the verb pala- ‘wish’ includes TP and anything below it. In (13a), the past tense marker -ess is attached to the embedded verbal stem o- ‘come.’ Semantically, the embedded event time is different from the matrix time. The time of coming is independent of the time of wishing, which is shown by the temporal adverb ecey ‘yesterday.’ As for grammatical elements assumed to be higher than T, (13a) shows that modality and force markers cannot be attached to the embedded verbal stem. However, consider the contrast between (13a) and (13b). The root clause in (13b) expresses the same event as the embedded clause in (13a), and the modality marker -keyss and the mood marker -ta are suffixed after the tense marker -ess. The availability of these two markers in (13b) indicates that there is no independent reason for blocking them after the verb. Given that Modality and Mood occupy higher positions than T in Korean (Ryu 1995: 35; Cinque 1999: 53–54; Suh 2017: 173), I take the absence of these elements in complements of pala- as evidence that these complements cannot be larger than TP.

    1. (13)
    1. No Modality and Force in -ki embedding
    1.  
    1. a.
    1. Na-nun
    2. 1sg-top
    1. [dp [tp
    2.  
    1. ecey
    2. yesterday
    1. apeci-kkeyse
    2. father-hnom
    1. o-si-ess] (*-keyss) (*-ta)-ki]-lul
    2. come-h-pst(*-mod)(*-decl)-nmlz-acc
    1. pala-∅-n-ta.
    2. wish-nh-prs-decl
    1. ‘I wish that my father had come yesterday.’
    1.  
    1. b.
    1. Ecey
    2. yesterday
    1. apeci-kkeyse
    2. father-hnom
    1. o-si-ess-keyss-ta.
    2. come-h-pst-mod-decl
    1. ‘Perhaps, my father came yesterday.’

As shown in (14a) below, the accusative Case marker -lul is attached to the domain headed by -ki. The same marker is attached to the DP, tayka ‘reward,’ in a simple clause in (14b). This indicates that clausal complements of pala- pattern together with nominals. I assume that Korean nominals are DPs, following the literature (Kang 2001; Park 2016; Kim et al. 2018: 128; among others).

    1. (14)
    1. The accusative case marker attached to -ki embedding
    1.  
    1. a.
    1. Na-nun
    2. 1sg-top
    1. [dp [tp
    2.  
    1. ecey
    2. yesterday
    1. apeci-kkeyse
    2. father-hnom
    1. o-si-ess]-ki]-lul
    2. come-h-pst-nmlz-acc
    1. pala-∅-n-ta.
    2. wish-nh-prs-decl
    1. ‘I wish that my father had come yesterday.’
    1.  
    1. b.
    1. Na-nun
    2. 1sg-top
    1. [dp
    2.  
    1. tayka]-lul
    2. reward-acc
    1. pala-∅-n-ta.
    2. wish-nh-prs-decl
    1. ‘I wish for rewards.’

To confirm that -ki embedding is not as large as CP, I apply a diagnostic test involving the ordering of adverbs. According to Cinque (1999), adverbs are universally generated in the specifier positions of different functional heads, a principle that applies to Korean as well (Lee 2000).15 Following Cinque (1999) and Lee (2000), I assume that adverbs of a particular class are generated in the specifier position of a specific functional head, and they follow a strict hierarchy. My assumptions concerning their structural positions are given in (15). The manner adverb ppalli ‘quickly’ is base generated in [Spec, vP] and wancenhi ‘completely’ is generated in [Spec, AspP]. That is, ppalli ‘quickly’ should always take a higher position than wancenhi ‘completely,’ within a clause.

    1. (15)
    1. Adverb in the specifier position of a functional head (Cinque 1999: 77; Lee 2000)
    1. … [vP
    2.  
    1. ppalli
    2. quickly
    1. [AspP
    2.  
    1. wancenhi
    2. completely
    1. [VP …]
    2.  
    1. Asp ] v]
    2.  

The projection of a silent inner aspectual (AspP) between vP and VP is relevant (Travis 2010). It is supported by different scopes of the adverbs ppalli ‘quickly’ and wancenhi ‘completely.’ In (16a), ppalli ‘quickly’ cannot modify the lower verb mek- ‘eat.’ It must have the scope over the causative. It cannot mean ‘Mincwun caused the event in which the child quickly ate.’ In contrast, wancenhi ‘completely’ (16b) can modify the caused event only. ‘Mincwun caused the event in which the child ate up the meal’ is a possible interpretation of (16b). The ordering constraint between the two as shown in the ungrammatical example (16c) indicates that ppalli ‘quickly’ should be higher than wancenhi ‘completely.’ Thus, I assume that ppalli ‘quickly’ is base generated in [Spec, vP] and wancenhi ‘completely’ in [Spec, AspP].

    1. (16)
    1. a.
    1.   Mincwun-i
    2.   Mincwun-nom
    1. ai-eykey
    2. baby-to
    1. [vP
    2.  
    1. pap-ul
    2. meal-acc
    1. ppalli
    2. quickly
    1. [AspP
    2.  
    1. pap
    2. meal
    1. mek]-i]-ess-ta.
    2. eat-caus-pst-decl
    1.   ‘Mincwun quickly hadcaus the child fed.’
    2. *‘Mincwun hadcaus the child quickly fed.’
    1.  
    1. b.
    1.   Mincwun-i
    2.   Mincwun-nom
    1. ai-eykey
    2. baby-to
    1. [vP
    2.  
    1. pap-ul
    2. meal-acc
    1. [AspP
    2.  
    1. pap
    2. meal
    1. wancenhi
    2. completely
    1. mek]-i]-ess-ta.
    2. eat-caus-pst-decl
    1.   ‘Mincwun hadcaus the child completely fed.’
    1.  
    1. c.
    1. *Mincwun-i
    2.   Mincwun-nom
    1. ai-eykey
    2. baby-to
    1. pap-ul
    2. meal-acc
    1. wancenhi
    2. completely
    1. ppalli
    2. quickly
    1. mek-i-ess-ta.
    2. eat-caus-pst-decl
    1.   ‘Mincwun completely hadcaus the child quickly fed.’ (intended)

Now, a comparison between (17a–b) provides empirical evidence that the adverb ordering is strict within the same clause. (17b) is ungrammatical because ppalli ‘quickly’ is lower than wancenhi ‘completely’ within the same clause, given that the causative construction in (17b) is monoclausal (Son 2006; Lee 2009).

    1. (17)
    1. The relative hierarchy of adverbs in Korean
    1.  
    1. a.
    1.   Minswu-ka
    2.   Minswu-nom
    1. aki-eykey
    2. baby-to
    1. [vP
    2. [
    1. ppalli
    2. quickly
    1. wancenhi
    2. completely
    1. pap-ul
    2. meal-acc
    1. mek-i]-ess-ta.
    2. eat-caus]-pst-decl
    1.   ‘Minswu quickly made the baby completely eat the meal’ (see (48) in (Son 2006: 49))
    1.  
    1. b.
    1. *Minswu-ka
    2.   Minswu-nom
    1. aki-eykey
    2. baby-to
    1. wancenhi
    2. completely
    1. ppalli
    2. quickly
    1. pap-ul
    2. meal-acc
    1. mek-i-ess-ta.
    2. eat-caus-pst-decl
    1.   ‘Minswu completely made the baby quickly eat the meal.’ (intended)16

I use adverb fronting as a diagnostic for CP-hood. Adverbs can undergo focus movement and move to a [Spec, CP] position (Cinque 1999; Lee 2000). If an embedded clause contains a CP domain, an embedded adverb can move to the embedded [Spec, CP]. This makes it linearly adjacent to an in-situ matrix adverb. When focus fronting takes place, the two adverbs are not subject to the ordering constraint because they are in different domains. In contrast, if an embedded clause does not have a CP position, embedded adverbs cannot be fronted to the embedded clause-initial position. In (17b) above, both adverbs are in the same clause, and the strict ordering is required. The lower adverb wancenhi ‘completely’ can only be followed by ppalli ‘quickly’ when wancenhi ‘completely’ is in the matrix clause in-situ, and ppalli ‘quickly’ is fronted to the embedded [Spec, CP] position.

Based on this, I argue that the availability of surface wancenhi ppalli ‘completely quickly’ entails that there exists a CP boundary between the two. Such an order arises only when the lower adverb is generated in the matrix clause and the higher adverb in the embedded clause. Two adverbs can become adjacent if the embedded higher adverb is fronted to the embedded [Spec, CP] while the matrix adverb stays in-situ. For an embedded (higher) adverb to be fronted inside the embedded clause, the embedding must be as large as CP. Therefore, the availability of this adverb order indicates that the embedded clause is a CP.

Now, consider the contrast between (18) and (19). The different orderings of ppalli ‘quickly’ and wancenhi ‘completely’ are applied to the -ko clause in (18) and the -ki clause in (19). If the -ko clause is a CP, we expect that an adverb can move to [Spec, CP] within the embedded domain. (18b) provides crucial evidence that -ko clauses are CPs. The grammaticality of (18b), notwithstanding the apparent violation of the hierarchy constraint, can only be accounted for if the two adverbs are in different domains. For this to be possible, -ko clauses should be CPs with an independent specifier position to which an embedded adverb moves. If -ko clauses were not CPs, there could not be a [Spec, CP] to which an embedded adverb could move, predicting (18b) to be ungrammatical contrary to fact. Therefore, the adverb ordering test reiterates that the -ko clause is an independent CP domain.

    1. (18)
    1. Adverb fronting test confirms that -ko embedding is CP
    1.  
    1. a.
    1. Miaka
    2. Mia
    1. ppalli
    2. quickly
    1. [cp
    2.  
    1. wancenhi
    2. completely
    1. apecikkeyse
    2. father
    1. wancenhi
    2. completely
    1. silphayhasyessta-ko]
    2. failed-c
    1. malhayssta.
    2. said
    1. ‘Mia quickly said that father completely failed.’
    1.  
    1. b.
    1. Miaka
    2. Mia
    1. wancenhi
    2. completely
    1. [cp
    2.  
    1. ppalli
    2. quickly
    1. apecikkeyse
    2. father
    1. ppalli
    2. quickly
    1. silphayhasyessta-ko]
    2. failed-c
    1. malhayssta.
    2. said
    1. ‘Mia completely said that father quickly failed.’

In contrast, the same test tells us that a -ki clause is not a CP. In (19a) below, ppalli ‘quickly’ cannot occur right after wancenhi ‘completely.’ Compare this to (19b). (19b) is grammatical with ppalli ‘quickly’ located in its base-generated position below the embedded subject. The contrast between (19a) and (19b) suggests that the embedded adverb cannot be fronted in -ki clauses, and I argue that this is evidence that there is no [Spec, CP] position in this type of embedded clause. In other words, -ki clauses are smaller than CP.17

    1. (19)
    1. Adverb fronting test shows that -ki clauses are smaller than CP
    1.  
    1. a.
    1. *Miaka
    2.   Mia
    1. wancenhi
    2. completely
    1. [ppalli
    2. [quickly
    1. apecikkeyse
    2. father
    1. ppalli
    2. quickly
    1. ttenasyess-ki]-lul
    2. left-ki]-acc
    1. palassta.
    2. wished
    1.   ‘Mia completely wished that father quickly left.’ (intended)
    1.  
    1. b.
    1.   Miaka
    2.   Mia
    1. wancenhi
    2. completely
    1. [
    2. [
    1. apecikkeyse
    2. father
    1. ppalli
    2. quickly
    1. ttenasyess-ki]-lul
    2. left-ki]-acc
    1. palassta.
    2. wished
    1.   ‘Mia completely wished that father quickly left.’

Having established that the clausal complements involving -ki are smaller than CP, I now investigate whether they are syntactically finite. To do so, I apply the two syntactic tests introduced in Section 3, i.e., NPI licensing and reflexive binding. Although smaller than CP, the clausal complements of pala- ‘wish’ do not allow cross-clausal syntactic operations.

First, NPI licensing is not possible between the matrix and -ki clauses. In the grammatical baseline (20a), Neg an and NPI amuto ‘anyone’ are both in the embedded clause. In (20b), on the other hand, the negative expression and NPI are located in different domains. The matrix negation cannot license an NPI inside a -ki clause. Therefore, a nominalized subordinate clause is syntactically opaque in terms of NPI licensing, and it thus patterns with other finite clauses.

    1. (20)
    1. NPI licensing with -ki embedding
    1.  
    1. a.
    1.   Nanun
    2.   1sg
    1. [dp [tp
    2.  
    1. Minaka
    2. Mina
    1. amuto
    2. anyone
    1. an
    2. neg
    1. pulless]-ki]-lul
    2. called-nmlz-acc
    1. palanta.
    2. wish
    1.   ‘I wish that Mina did not call anyone.’
    1.  
    1. b.
    1. *Nanun
    2.   1sg
    1. [dp [tp
    2.  
    1. Minaka
    2. Mina
    1. amuto
    2. anyone
    1. puluess]-ki]-lul
    2. called-nmlz-acc
    1. an
    2. neg
    1. palanta.
    2. wished
    1.   ‘I did not wish that Mina called anyone.’ (intended)

Second, cross-clausal binding is not allowed with -ki clauses. The matrix antecedent Mina cannot bind the anaphor caki-casin located inside the embedded clause in (21a) and (21c) below. In (21a), the embedded anaphor can only be bound by Naun in the embedded clause and not by Mina in the matrix clause. Similarly, in (21c), the matrix subject Mina cannot bind caki-casin in the embedded clause. On the other hand, the pronoun ku in (21b) can refer to the matrix subject or any other referent except Naun. To summarize, the anaphor is bound, and the pronoun is free within the DP domain. Therefore, the clausal complement of this wish-type verb is the binding domain, and it patterns with other embedded domains that are considered finite.

    1. (21)
    1. Binding with -ki embedding
    1.  
    1. a.
    1.   Minai-ka
    2.   Mina-nom
    1. [dp [tp
    2.  
    1. Naunj-i
    2. Naun-nom
    1. caki casin*i/j-ul
    2. self-acc
    1. cohahaess]-ki]-lul
    2. liked-nmlz-acc
    1. palassta.
    2. wished
    1.   ‘Mina wished Naunj liked herselfj.’ but not ‘Minai wished Naunj liked heri.’
    1.  
    1. b.
    1.   Minai-ka
    2.   Mina-nom
    1. [dp [tp
    2.  
    1. Naunj-i
    2. Naun-nom
    1. kui/*j/k-lul
    2. 3sg-acc
    1. cohahaess]-ki]-lul
    2. liked-nmlz-acc
    1. palassta.
    2. wished
    1.   ‘Minai wished Naunj liked him/heri/*j/k.’
    1.  
    1. c.
    1. *Minai-ka
    2.   Mina-nom
    1. [dp [tp
    2.  
    1. caki casini-i
    2. self-nom
    1. kay-lul
    2. dog-acc
    1. cohahaess]-ki]-lul
    2. liked-nmlz-acc
    1. palassta.
    2. wished
    1.   ‘Minai wished that shei liked the dog.’ (intended)

5 Sub-TP embeddings in Korean

In this section, we discuss sub-TP clausal complements. I focus on the two verbs mantul- ‘make (causative)’ and po- ‘try’ and show that they take smaller-than-TP clausal complements. Given that a functional head other than T can license the subject in Korean, this section tests whether these complements can be finite in the sense of Kornfilt (2007). The syntactic tests show that the complement of mantul- is syntactically finite, while the complement of po- is not.

5.1 Introduction

Based on the availability of verbal morphology and a clause-internal subject, I will show that the causative verb mantul- ‘make’ takes HonP as its complement, while the verb po- ‘try’ takes vP. Although both HonP and vP are smaller than TP, only the former contains the subject.

5.1.1 Causatives take HonP and try takes vP as the complement

Example (22b) below is an instance of the causative construction in Korean. The simple clause in (22a) is embedded under the verb mantul- ‘make’ in (22b). The matrix verb selects the subordinating element -key. (22b) shows that the suffixes for tense, modality and mood are not available in this type of embedding (Ryu 1995: 38; Kim et al. 2018: 120–122), in contrast to the simple clause in (22a). The lack of tense, modality and mood morphology in (22b) constitutes evidence that complements of the causative verbs contain Hon but not T or higher functional heads. I assume such heads do not exist without overt morphology, especially for T, in line with Todorović (2016: 241–297). Note that an embedded subject, apeci ‘father,’ is licensed in (22b) and (23) even without T. This is not surprising because Hon, rather than T, licenses nominative Case in Korean, as discussed in Section 2.2. In these two examples, the embedded Hon on the verb is realized as -si, indicating that it agrees with the embedded subject. In short, the complement of the causative verb is HonP and contains a subject.

    1. (22)
    1. A -key clause contains Hon but not T
    1.  
    1. a.
    1. Apeci-kkeyse
    2. father-hnom
    1. ecey
    2. yesterday
    1. cip-ey
    2. home-at
    1. o-si-ess-keyss-ta.
    2. come-h-pst-mod-decl
    1. ‘Father came home yesterday.’
    1.  
    1. b.
    1. Na-nun
    2. 1sg-top
    1. [apeci-kkeyse
    2. father-hnom
    1. ecey
    2. yesterday
    1. cip-ey
    2. home-at
    1. o-si(*-ess)(*-keyss)(*-ta)]
    2. come-h(*-pst)(*-mod)(*-decl)
    1. -key
    2. -key
    1. mantul-ess-ta.
    2. make-pst-decl
    1. ‘I made father come home yesterday.’ (Yesterday is the time when father came.)

In (23), there are different temporal adverbs in the matrix and embedded clauses. The temporal reference is denoted by kucey ‘the day before yesterday’ in the root clause and ecey ‘yesterday’ in the embedded clause. The matrix time is two days earlier than the utterance time. However, even when the embedded time is different from the matrix time, a tense marker in the embedded clause is ungrammatical. I take this as evidence for the absence of TP.

    1. (23)
    1. Different temporal reference between the matrix and embedded clauses
    1. Kucey
    2. 2.days.ago
    1. na-nun
    2. 1sg-top
    1. [apeci-kkeyse
    2. father-hnom
    1. ecey
    2. yesterday
    1. cip-ey
    2. home-at
    1. o-si(*-ess)]-key
    2. come-h(*-pst)-key
    1. mantul-ess-ta.
    2. make-pst-decl
    1. Lit. ‘Two days ago, I made father come home yesterday.’
    2. (An action that happened two days ago caused the event yesterday in which father came.)

Like the causative verb mantul- ‘make,’ po- ‘try’ takes a smaller-than-TP clausal complement. I argue that it takes vP as its complement (cf. Lee 2009: 190 ff.).18 In (24a–b), the verbalizer -ha is assumed to be the head of vP (Jung 2016; Kim et al. 2018: 176–179; among others). The complement cannot have the honorific marker or any other suffixes hierarchically higher than Hon. The suffix -e does not have any modal or temporal meaning.19 If Hon licenses nominative Case, the embedded vP must not have an overt subject.20 (24b) shows this prediction holds true. Finally, (24c) shows that the embedded event cannot have independent temporal reference. In short, the vP embedding includes neither Hon nor T.

    1. (24)
    1. -e embedding
    1.  
    1. a.
    1.   Apeci-kkeyse
    2.   father-hnom
    1. [
    2. [
    1. PRO
    2.  
    1. nakse-ha (*-si)(*-ess/*-n)(*-kyess)(*-ta)]-e
    2. scribble-vblz (*-h)(*-past/*-prs)(*-mod)(*-mood)]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father tried scribbling.’
    1.  
    1. b.
    1.   Apecii-kkeyse
    2.   father-hnom
    1. [
    2. [
    1. (*Apeci-ka/*casini-i/*kui-ka)
    2. father-nom/self-nom/3sg-nom
    1. nakse-ha]-e
    2. scribble-vblz]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father tried scribbling.’                                                    (see (40) (Lee, 2009:192))
    1.  
    1. c.
    1. *Kucey
    2.   Two.days.ago
    1. Minci-ka
    2. Minci-nom
    1. [ PRO
    2. [
    1. sakwa-lul
    2. apple-acc
    1. ecey
    2. yesterday
    1. kkakk]-e
    2. peel]-e
    1. poassta.
    2. tried
    1.   Lit. ‘Two days ago, Minci tried peeling an apple yesterday.’ (intended)
    2.   (Two days ago, Minci tried something, and that is to peel an apple yesterday.)

The try construction should be distinguished from a serial verb construction, despite similarly juxtaposing two verbs. I show that po- ‘try’ embeds a vP while the serial verb construction has no embeddings involved.

First, an embedded vP can be topicalized but a component of the serial verb construction cannot (Kim Gyeong-yeol 2019). Consider the contrast between the try construction in (25a) and serial verb construction in (26b).21 In (25a), the vP-e part forms a constituent that can be extracted from the phrase headed by po- ‘try.’ In (25a), the topic marker -nun shows that [sakwalul kkakk]-e is topicalized. By contrast, in a true serial verb construction in (25b), the topicalization marker in between the two verbs is ungrammatical (Kim Gyeong-yeol 2019). This implies that in the try construction (25a), the two verbs form separate predicates, while in the serial verb construction in (25b), the two verbs form a single predicate and can thus not be separated.

    1. (25)
    1. The clausal complement of po- ‘try’ can be topicalized
    1.  
    1. a.
    1. Apecikkeyse
    2. father
    1. [vP
    2. [
    1. sakwalul
    2. apple
    1. kkakk]-e-nun
    2. peel]-e-top
    1. posyessta.
    2. tried
    1. ‘Father tried peeling an apple (and not chopping it).’
    1.  
    1. b.
    1. Apecikkeyse
    2. father
    1. [
    2. [
    1. sakwalul
    2. apple
    1. pa]-e (*-nun)
    2. dig]-e(*-top)
    1. mutusyessta.
    2. buried
    1. ‘Father dug (the soil) and buried an apple.’ (ungrammatical if ‘dig’ is topicalized).

Second, the try construction involves a hierarchy between the two predicates: the verb po- ‘try’ is in a higher position and selects vP as its complement. The ‘do-so’ test clarifies this: kuliha- ‘do so’ can substitute for vP (Kim Kyeong-min 2019). In (26a), [vP dulessu-lul ip] ‘wear a dress’ in the first clause is substituted by kuliha- in the second clause. This form does not replace HonP, as the honorific -si marking is still required in the second clause. Now, consider (26b–d). One try clause is coordinated with another one. The suffix -ko that is suffixed to poass- ‘tried’ coordinates two hierarchically equal phrases. Notably, the verbal pro-form kuliha- can replace either the smaller vP with only kkakk- ‘peel’ as in (26b), or both kkakk- ‘peel’ and the main verb po- ‘try’ as in (26c). However, as shown in the ungrammatical example (26d), it cannot replace po- ‘try’ only and leave behind the other predicate kkakk- ‘peel.’ This indicates that the smaller vP is embedded in a larger phrase that is headed by po- ‘try.’

    1. (26)
    1. Pro-form kuliha cannot substitute for higher verb only
    1.  
    1. a.
    1.   Apeci-kkeyse
    2.   Apeci-hnom
    1. [vP
    2. [
    1. tulessu-lul
    2. dress-acc
    1. ip]-si-ess-ko,
    2. wear]-h-pst-and
    1. ameni-to
    2. mother-also
    1. kuliha*(-si)-ess-ta.
    2. do *(-h)-pst-decl
    1.   ‘Father wore a dress, and so did mother.’
    1.  
    1. b.
    1.   Mincika
    2.   Minci
    1. [vP
    2. [
    1. sakwalul
    2. apple
    1. kkakk]-e
    2. peel]-e
    1. poass-ko,
    2. tried-and
    1. Naun-to
    2. Naun-also
    1. kuliha-e
    2. do-e
    1. poass-ta.
    2. tried-decl
    1.   ‘Minci tried peeling an apple, and Naun tried it too.
    1.  
    1. c.
    1.   Mincika
    2.   Minci
    1. [vP
    2. [
    1. sakwalul
    2. apple
    1. kkakk-e
    2. peel-e
    1. po]-ess-ko,
    2. try]-pst-and
    1. Naun-to
    2. Naun-also
    1. kuliha-ess-ta.
    2. do-pst-decl
    1.   ‘Minci tried peeling an apple, and Naun did too.
    1.  
    1. d.
    1. *Mincika
    2.   Minci
    1. sakwalul
    2. apple
    1. kkakk-e
    2. peel-e
    1. [po]-ess-ko,
    2. [try]-pst-and
    1. Naun-to
    2. Naun-also
    1. hopakul
    2. pumpkin
    1. ssel-e
    2. chop-e
    1. kuliha-ess-ta.
    2. do-pst-decl
    1.   ‘Minci tried peeling an apple, and Naun tried chopping a pumpkin.’ (intended)

In contrast, partial substitution is not possible in serial verb constructions. In (27a), kuliha replaces the whole [sakwalul pae mut] ‘dig and bury an apple’ in the second clause. In (27c–d), on the other hand, the pro-form cannot be used in place of the first verb nor the second verb separately. Therefore, the pro-form test provides evidence that two verbs in serial verb constructions are inseparable. On the other hand, po- ‘try’ and its complement are not.

    1. (27)
    1. Pro-form kuliha can substitute for both verbs only in the serial verb construction.
    1.  
    1. a.
    1.   Mincika
    2.   Minci
    1. [
    2. [
    1. sakwalul
    2. apple
    1. pa-e
    2. dig-e
    1. mut]-ess-ko,
    2. bury]-pst-and
    1. Naun-to
    2. Naun-also
    1. kuliha-ess-ta.
    2. do-pst-decl
    1.   ‘Minci dug and buried an apple, and Naun did too.’
    1.  
    1. b.
    1. *Mincika
    2.   Minci
    1. [
    2. [
    1. sakwalul
    2. apple
    1. pa]-e
    2. dig]-e
    1. mut-ess-ko,
    2. bury-pst-and
    1. Naun-to
    2. Naun-also
    1. [kuliha]-e
    2. [do]-e
    1. mut-ess-ta.
    2. bury-pst-decl
    1.   Lit. ‘?Minci dug and buried an apple, and Naun did so and buried it too.’
    1.  
    1. c.
    1. *Mincika
    2.   Minci
    1. sakwalul
    2. apple
    1. pa-e
    2. dig-e
    1. [mut]-ess-ko,
    2. [bury]-pst-and
    1. Naun-to
    2. Naun-also
    1. paylul
    2. pear
    1. kki-e
    2. put.together
    1. [kuliha]-ess-ta.
    2. [do]-pst-decl
    1.   ‘Minci dug and buried an apple, and Naun buried pear with the apple. (intended)’22

So far in this section, I have argued that the causative verb mantul- ‘make’ takes an HonP complement, and the verb po- ‘try’ takes a vP complement: both lack a TP. Since Hon is the nominative Case licensor, an embedded subject can appear in HonP but not in vP. Note that I have made two claims with respect to the vP complement: (i) the -e embedding lacks Hon, and therefore (ii) it does not license an overt subject. I now address counterarguments raised in the anonymous reviews, beginning with the ones to the first claim.

A seemingly problematic case to my first claim comes from the use of suppletive forms. Verbs such as mek- ‘eat’ and ca- ‘sleep’ take irregular honorific forms tusi- and cwumusi-, respectively, and they can appear inside the -e embedding as in (28).

    1. (28)
    1. Suppletive verbs in -e embedding
    1.  
    1. a.
    1. Apeci-kkeyse
    2. father-hnom
    1. [ PRO
    2.  
    1. ku
    2. that
    1. umsik-ul
    2. food-acc
    1. tusi]-e
    2. eat.h]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1. ‘Father tried (eating) the food.’
    1.  
    1. b.
    1. Apeci-kkeyse
    2. father-hnom
    1. [ PRO
    2.  
    1. ku
    2. that
    1. chimtay-eyse
    2. bed-on
    1. cwumusi]-e
    2. sleep.h]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1. ‘Father tried sleeping on the bed.’

At first glance, this seems to challenge the claim that the -e embedding lacks Hon. However, a diagnostic involving the existential verb iss- ‘be’ demonstrates that suppletive verbs do not provide reliable evidence for the presence of syntactic Hon. This verb allows both types of honorific forms as in (29): one with the regular honorific marker, iss-si ‘be-h’, and another as a suppletive honorific form, kyeysi- ‘be.h.’

    1. (29)
    1. Suppletive verbs allowed because they lack Hon
    1.  
    1. a.
    1.   Apeci-kkeyse
    2.   father-hnom
    1. [ PRO
    2.  
    1. yeki-ey
    2. here-at
    1. iss]-e
    2. be]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father tried staying here.’
    1.  
    1. b.
    1. *Apeci-kkeyse
    2.   father-hnom
    1. [ PRO
    2.  
    1. yeki-ey
    2. here-at
    1. iss-si]-e
    2. be-h]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father tried staying here.’ (intended)
    1.  
    1. c.
    1.   Apeci-kkeyse
    2.   father-hnom
    1. [ PRO
    2.  
    1. yeki-ey
    2. here-at
    1. kyeysi]-e
    2. be.h]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father tried staying here.’

Although both (29b) and (29c) include an honorific verb form of iss ‘be,’ it is not grammatical to include the explicit honorific marker as in (29b), while the example with the suppletive form, i.e., (29c), is grammatical. This fact is accounted for if the suppletive form and the canonical -si attached form are not structurally equivalent, notwithstanding the functional similarity. Assuming that -si is the realization of Hon, it follows that the suppletive form does not include Hon syntactically. Such suppletive forms can be listed in the lexicon as distinct verb entries.23

Turning to the second point, i.e., whether an overt subject can be licensed inside the -e embedding, an anonymous reviewer presents a counterexample, reproduced in (30a) below.

    1. (30)
    1. The apparent overt subject in the -e embedding is part of adverbial
    1.  
    1. a.
    1.   Apeci-kkeyse
    2.   father-hnom
    1. [tangsin-i
    2. [self-nom
    1. capalcekulo
    2. voluntarily
    1. o]-e
    2. come]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father voluntarily tried coming.’
    1.  
    1. b.
    1. *Apeci-kkeyse
    2.   father-hnom
    1. [tangsin-i
    2. [self-nom
    1. o]-e
    2. come]-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father tried coming himself.’ (intended)
    1.  
    1. c.
    1.   Apeci-kkeyse
    2.   father-hnom
    1. [AdvP
    2. [
    1. tangsin-i
    2. self-nom
    1. capalcekulo
    2. voluntarily
    1. ] PRO
    2. ]
    1. o-e
    2. come-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘Father voluntarily tried coming.’

On the surface, tangsin ‘self’ appears to be licensed as the subject inside the clausal complement. However, if tangsin ‘self’ were indeed a subject, we would expect (30b) to be grammatical, contrary to fact. An alternative analysis would be to treat tangsin ‘self’ as part of an adverbial adjoined to the structure, as schematized in (30c). Under this analysis, the -e embedding does not contain an overt subject, consistent with the claim that it lacks Hon.

In the following section, I provide further justification for the claim that these complements also do not contain a CP, as has been previously argued.

5.1.2 The causative and try constructions do not involve CP complements

The elements -key and -e, which appear in between the matrix verb and the clausal complements, have been previously identified by some researchers as complementizers (Ryu 1995: 30–36; Sells 1995: 283–285; Kim et al. 2018: 120–121). If they are on the right track, the clausal complements of mantul- ‘make’ or po- ‘try’ can be as large as CP without intermediate projections. To test whether these clausal complements are CPs or reduced complements in Wurmbrand’s (2001) sense, I apply the same adverb ordering test as used in Section 4.

Consider first -key complements. The adverb ordering test shows that -key clauses do not include a CP domain. Recall that adverbs can move to a [Spec, CP] position (Cinque 1999; Lee 2000: 73). The movement alters the otherwise fixed linear order of ppalli ‘quickly’ and wancenhi ‘completely.’ Within the same clause, ppalli ‘quickly’ should always be in a higher position than wancenhi ‘completely’ (Lee 2000: 71). Therefore, ppalli ‘quickly’ precedes wancenhi ‘completely.’ The opposite order is grammatical only when the two adverbs are in different domains. They appear linearly adjacent because the second adverb can move to the initial position of the embedded clause via focus movement, while the first one remains in its base-generated position in the matrix clause. Now consider (31a) and (31b) below. In both, the adverb ppalli ‘quickly’ is base-generated in the embedded -key clause. In (31a), it moves to the clause-initial position while in (31b), it stays in the base position. Focus movement is expected to be grammatical if -key clauses are CPs. However, (31a) is ungrammatical, indicating that -key clauses are not CPs. The grammaticality of (31b) also suggests that it is specifically focus movement which is not available: the adverbs can take different scope if no movement take place. Based on this, I argue that -key embedding does not contain a CP domain.

    1. (31)
    1. Adverb fronting test shows that -key embedding is not CP
    1.  
    1. a.
    1. *Minaka
    2.   Mina
    1. wancenhi
    2. completely
    1. [ppalli
    2. quickly
    1. apecikkeyse
    2. father
    1. papul
    2. meal
    1. ppalli
    2. quickly
    1. meksi]-key
    2. eat-key
    1. mantulessta.
    2. made
    1.   ‘Mina completely made father have a quick meal.’ (intended)
    1.  
    1. b.
    1.   Minaka
    2.   Mina
    1. wancenhi
    2. completely
    1. [
    2.  
    1. apecikkeyse
    2. father
    1. papul
    2. meal
    1. ppalli
    2. quickly
    1. meksi]-key
    2. eat-key
    1. mantulessta.
    2. made
    1.   ‘Mina completely made father have a quick meal.’

Similarly, the adverbial fronting test for CP-hood is applied to the complements of the verb po- ‘try’ in (32) below, indicating that -e clauses are smaller than CP. Recall that an embedded adverb can only be fronted if there is a [Spec, CP] position. In (32a), the embedded clause does not have an additional [Spec, CP] position to which ppalli ‘quickly’ can move. (32b) shows that the ungrammaticality has nothing to do with the relative scope of the two adverbs. The sentence is grammatical if the embedded adverb stays in situ.

    1. (32)
    1. Adverb fronting test shows that -e embedding is not CP
    1.  
    1. a.
    1. *Nanun
    2.   1sg
    1. wancenhi
    2. completely
    1. [ppalli
    2. [quickly
    1. kimpapul
    2. maki
    1. ppalli
    2. quickly
    1. mek]-e
    2. eat]-e
    1. poassta.
    2. tried
    1.   Intended: ‘I wholeheartedly tried so that I could eat maki rolls quickly.’
    2.   (Lit. ‘I completely tried to eat maki rolls quickly.’)
    1.  
    1. b.
    1.   Nanun
    2.   1sg
    1. wancenhi
    2. completely
    1. [
    2. [
    1. kimpapul
    2. maki
    1. ppalli
    2. quickly
    1. mek]-e
    2. eat]-e
    1. poassta.
    2. tried
    1.   I wholeheartedly tried so that I could eat maki rolls quickly.
    2.   (Lit. ‘I completely tried to eat maki rolls quickly.’)

Therefore, the embeddings are smaller than CP, and both causative and the try-type verbs in Korean take a reduced clausal complement.

Sections 5.2 and 5.3 discuss whether the two smaller-than-TP embeddings can be finite in terms of NPI licensing and binding domain.

5.2 Hon without Tense: -key construction

This section discusses whether the complement in the causative construction is syntactically finite. Notably, the HonP complement selected by the causative verb mantul- ‘make’ licenses a clause-internal subject. I use NPI licensing and anaphor binding to test whether the causative construction allows cross-clausal syntactic operations.

5.2.1 NPI

As I discussed above in 3.2, an NPI can only be licensed by clausemate negation. This requirement is satisfied in the baseline sentence presented in (33a). The NPI amuto ‘anyone’ is in HonP, and so is the Neg. In (33b–c), on the other hand, the clausemate condition is not satisfied. The Neg in the root clause cannot license an embedded NPI in either the subject or object position.

    1. (33)
    1. NPI licensing with -key embedding
    1.  
    1. a.
    1.   Baseline: both negation and NPI are in the embedded clause
    1.   Nanun
    2.   1sg
    1. [HonP
    2.  
    1. amuto
    2. anyone
    1. Minci-lul
    2. Minci-acc
    1. an
    2. neg
    1. manna]-key
    2. meet-key
    1. mantulessta.
    2. made
    1.   ‘I made nobody meet Minci.’
    1.  
    1. b.
    1.   Matrix negation and embedded subject NPI
    1. *Nanun
    2.   1sg
    1. [HonP
    2.  
    1. amuto
    2. anyone
    1. Minci-lul
    2. Minci-acc
    1. manna]-key
    2. meet-key
    1. an
    2. neg
    1. mantulessta.
    2. made
    1.   ‘I did not make anyone meet Minci.’ (intended)
    1.  
    1. c.
    1.   Matrix negation and embedded object NPI
    1. *Nanun
    2.   1sg
    1. [HonP
    2.  
    1. Minci-ka
    2. Minci-nom
    1. amuto
    2. anyone
    1. manna]-key
    2. meet-key
    1. an
    2. neg
    1. mantulessta.
    2. made
    1.   ‘I did not make Minci meet anyone.’ (intended)

Based on the observation that the NPI cannot be licensed by the matrix negation in (33b), I claim that an element inside the embedded HonP is not a clausemate with another one in the matrix clause. The embedded HonP is thus syntactically finite as diagnosed by NPI licensing.

5.2.2 Binding domains

Examples in (34) test whether the HonP complement of the causative verb mantul- ‘make’ acts as the binding domain. In (34a), the embedded anaphor caki-casin must be bound by the closest antecedent Naun. In (34b), the embedded anaphor is replaced by the third person pronoun ku. It cannot refer to the embedded subject Naun but can refer to the matrix subject Mina. Sentence (34c) is an example where the anaphor is in the embedded subject position. The matrix subject cannot bind the anaphor inside the HonP. (34b) and (34c) are accounted for if HonP is opaque.

    1. (34)
    1. Binding with -key embedding
    1.  
    1. a.
    1.   Minai-ka
    2.   Mina-nom
    1. [HonP
    2.  
    1. Naunj-i
    2. Naun-nom
    1. caki casin*i/j-ul
    2. self-acc
    1. cohaha]-key
    2. like-key
    1. mantulessta.
    2. made
    1.   ‘Mina made Naunj like herselfj.’ but not ‘Minai made Naunj like heri.’
    1.  
    1. b.
    1.   Minai-ka
    2.   Mina-nom
    1. [HonP
    2.  
    1. Naunj-i
    2. Naun-nom
    1. kui/*j/k-lul
    2. 3sg-acc
    1. cohaha]-key
    2. like-key
    1. mantulessta.
    2. made
    1.   ‘Minai made Naunj like him/heri/*j/k.’
    1.  
    1. c.
    1. *Nauni-i
    2.   Naun-nom
    1. [HonP
    2.  
    1. caki casini-i
    2. self-nom
    1. kay-lul
    2. dog-acc
    1. cohaha]-key
    2. like-key
    1. mantulessta.
    2. made
    1.   ‘Nauni made herselfi like the dog.’ (intended)
    1.  
    1. d.
    1. *Nauni-i
    2.   Naun-nom
    1. [HonP
    2.  
    1. kui-ka
    2. he/she-nom
    1. kay -lul
    2. dog-acc
    1. cohaha]-key
    2. like-key
    1. mantulessta.
    2. made
    1.   ‘Nauni made him/her*i like the dog.’

The unexpected ungrammaticality of (31d), given HonP is a binding domain, results from an independent reason. The ungrammaticality of both (31c) and (31d) is related to the semantics of this causative construction. The -key construction denotes directive causation, where the causer influences an action without control over the outcome. It is semantically not possible to indirectly cause oneself to do an action, so the causer and causee cannot have the same referent: the causee needs to bring about the caused event independently (Shibatani 1976: 31–38; Suh 2006: 1120–1123). In short, (31c–d) are not good because the embedded subject must not have the same referent as the matrix subject semantically, rather than syntactically. Notwithstanding (31c–d), I claim that HonP patterns with other finite domains in terms of binding, based on (31b).

So far, I have applied two syntactic tests to the embedded HonP. According to the tests, an NPI inside a -key clause cannot be licensed by matrix negation, and -key embedding is a binding domain where an anaphor must be bound, and a pronoun must be free. These results tell us that the HonP complement is syntactically opaque for cross-clausal operations, and in turn, syntactically finite.

5.3 Smaller than Hon: vP embedding

The next question investigates whether a reduced complement as small as vP, as in the try construction, can be finite. Before we apply the NPI and binding tests, recall that the vP complement clause lacks Hon. As shown in (35a), attaching -si, the overt realization of Hon, to the embedded verb stem makes it ungrammatical. Due to the absence of Hon, there can be no nominative subject in the embedded clause as in (35b).

    1. (35)
    1. a.
    1.   Sensayngnim-kkeyse
    2.   Teacher-hnom
    1. [
    2.  
    1. Jihye-lul
    2. Jihye-acc
    1. pulu (*-si)]-e
    2. call (*-h)-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘The teacher tried calling Jihye.’
    1.  
    1. b.
    1. *Sensayngnim-kkeyse
    2.   Teacher-hnom
    1. [
    2.  
    1. Naun-i
    2. Naun-nom
    1. Jihye-lul
    2. Jihye-acc
    1. pulu]-e
    2. call-e
    1. po-si-ess-ta.
    2. try-h-pst-decl
    1.   ‘?The teacher tried Naun calling Jihye.’ (intended)

5.3.1 NPI

Examples (36a–c) provide combinations of the NPI amuto and negation in the matrix and embedded clauses. (36a) is the baseline sentence where the clausemate NPI licensing condition is satisfied. (36b–c) are instances where the NPI amuto is in the vP and Neg in the matrix. As shown in (36b), the matrix negation can license the embedded NPI in the try construction. Therefore, the -e complement patterns with non-finite domains regarding NPI licensing.

    1. (36)
    1. NPI licensing with -e embedding
    1.  
    1. a.
    1.   Baseline: both negation and NPI are in the embedded complement
    1.   Nanun
    2.   1sg
    1. [vP
    2.  
    1. amuto
    2. anyone
    1. an
    2. neg
    1. pulu]-e
    2. call-e
    1. poassta.
    2. tried
    1.   ‘I tried not calling anyone.’
    1.  
    1. b.
    1.   Matrix negation and embedded NPI (long form)
    1.   Nanun
    2.   1sg
    1. [vP
    2.  
    1. amuto
    2. anyone
    1. pulu]-e
    2. call-e
    1. po-cianh-ess-ta.
    2. try-neg-pst-decl
    1.   ‘I did not try calling anyone.’
    1.  
    1. c.
    1.   Matrix negation and embedded NPI (short form)
    1. *Nanun
    2.   1sg
    1. [vP
    2.  
    1. amuto
    2. anyone
    1. pulu]-e
    2. call-e
    1. an
    2. neg
    1. po-ess-ta.
    2. tried
    1.   ‘I did not try calling anyone.’ (intended)

(36c) is like (36b) in that both have the negation in the matrix clause and the NPI inside the -e clause, but (36c) is ungrammatical. However, this does not undermine the argument that -e complements are non-finite. The difference between (36b) and (36c) is in the types of negation in Korean. (36b) uses the long form negation, while (36c) uses the short form.24 In (36c), the negative adverb an occurs between the two verbs, pulu- ‘call’ and po- ‘try,’ whereas in (36b), the negative verbal suffix -cianh is suffixed after the verb stem instead of an. The ungrammaticality of (36c) results from a constraint against adverb insertion. In the try construction, no adverb is allowed between the two verbs, including the negative adverb an (Kim 2014; Kim Gyeong-yeol 2019). (37b) below shows that no adverb can be inserted between pulu- ‘sing’ and po- ‘try’. Therefore, (36c) is bad not because the external negation fails to license amuto in vP but for independent reasons that disallow adverb insertion in these constructions.

    1. (37)
    1. Adverbs are not allowed in the try construction
    1.  
    1. a.
    1.   The try construction with vP embedding
    1.   Chelswu-nun
    2.   Chelswu-top
    1. [nolay-lul
    2. song-acc
    1. pulu]-e
    2. sing-e
    1. po-ess-ta.
    2. try-pst-decl
    1.   ‘Chelswu tried singing a song.’
    1.  
    1. b.
    1.   An adverb is inserted between the matrix verb and the embedded verb
    1. *Chelswu-nun
    2.   Chelswu-top
    1. [nolay-lul
    2. song-acc
    1. pulu]-e
    2. sing-e
    1. cal
    2. well
    1. po-ess-ta.
    2. try-pst-decl
    1.   ‘Chelswu tried well to sing a song.’                                       (= (37) in Kim (2014: 58))

5.3.2 Binding domains

Now, I show that the binding test gives us the same result regarding the syntactic finiteness of the embedded vP. There is only an internal argument in the complement of po- ‘try,’ because it does not contain Hon, which is required to license an overt subject. Cross-clausal binding is possible in the try construction, meaning that the vP is not a syntactically finite domain. In (38a), the potential binder Mina in the matrix clause can bind the anaphor inside the vP. On the other hand, the third person pronoun ku in the embedded vP must refer to a different entity than the matrix subject Mina. This indicates that the embedded vP does not constitute a binding domain. Therefore, the embedded vP is non-finite.

    1. (38)
    1. Binding with -e embedding
    1.  
    1. a.
    1. Minai-ka
    2. Minai-nom
    1. [vP
    2.  
    1. caki casini/*j-ul
    2. selfi/*j-acc
    1. cohaha]-e
    2. like-e
    1. poassta.
    2. tried
    1. ‘Minai tried to like herselfi/*j.’
    1.  
    1. b.
    1. Minai-ka
    2. Minai-nom
    1. [vP
    2.  
    1. ku*i/j-lul
    2. 3sg*i/j-acc
    1. cohaha]-e
    2. like-e
    1. poassta.
    2. tried
    1. ‘Minai tried to like him/her*i/j.’

5.3.3 vP embedding is not finite

Both the NPI and binding tests indicate the embedded vP in the try construction is transparent for cross-clausal syntactic operations. Therefore, the embedded vP embedding is non-finite. This result corroborates the conclusions of Lee (2009), who argues that Korean complex verb constructions with control verbs (e.g., po- ‘try’) are infinitival, and Sohn (1994), who classified verbs as infinitival when their root is suffixed by -e.

6 Conclusion

This study investigated finiteness in Korean across various types of embedded complements. I first showed that clausal complements in Korean come in different sizes, with several smaller than CP (Wurmbrand 2001). This contrasts with earlier analyses, which assume that all clausal complements project up to CP (Kang 1988; Ryu 1995; Suh 2017; Kim et al. 2018: 101–124; among others). Then, two diagnostics for syntactic finiteness, namely cross-clausal NPI licensing and binding, were applied to each type of clausal complements (Kornfilt 2007). The analysis showed that clausal complements, even those smaller than TP, are finite if they are large enough to contain the Hon(orific) head. Hon is a functional head in Korean related to subject agreement and nominative Case licensing (Ryu 1995; Choi 2010; Kim et al. 2018: 233). These findings are summarized in Table 2.

Table 2

Results of diagnostics on clausal complements.

Complement of… Contains Tests for finiteness Finite?
Tense Agreement NPI? Binding?
‘say’ (CP) ✓ ✓ ✓ ✓ Yes
‘wish’ (TP) ✓ ✓ ✓ ✓ Yes
‘makecaus’ (HonP) × ✓ ✓ ✓ Yes
‘try’ (vP) × × × × No

A comparison between two kinds of smaller-than-TP complements showed that an embedded complement without Hon, and thus without a clause-internal subject, fails to show properties of syntactic finiteness. This shows that a domain can be syntactically finite without T(ense), supporting Kornfilt’s (2007) argument that overt subject agreement, rather than tense, is the primary determinant of (syntactic) finiteness. This contrasts with the view that T is necessary for finiteness (Rizzi 1997). Recall that similar observations supporting Kornfilt’s approach have been reported across languages such as Turkish (Kornfilt 2007) and Brazilian Portuguese (Raposo 1987).

The results of this study may also be connected to the morphological approach to finiteness in some descriptive grammars of Korean. Martin (1992) and Sohn (1994) classify Korean clauses as finite or non-finite depending on verb forms and argue that the -e suffix to the verb stem is an infinitival marker. These reference grammars define the term ‘infinitival’ as something interchangeable with the citation form or verb stem. However, given that the only non-finite clausal embedding in this study turned out to be the one that ends with -e, the form of the subordinating element, the morphological claim in descriptive studies may have some bearing on syntactic finiteness/opacity.

Additionally, the findings resonate with cross-linguistic studies on domains that have traditionally been understood as CP. For instance, Messick & Alok (2021) analyze Hindi complements introduced by the element ki, which is traditionally assumed to head CP. However, the ki embedding in Hindi allows stripping unlike English that-clause complements. On this basis, they argue ki is a subordination marker that heads Subordinate Phrase (SubP), a functional projection high in the clausal left periphery. Similarly, this study shows that various clausal complements in Korean do not project to CP, and they should be reanalyzed likewise.

Abbreviations

1/2/3 = first/second/third person; acc = accusative; c = complementizer; caus = causative; decl = declarative; h = honorific; hnom = honorific nominative; mod = modality; neg = negation; nh = non-honorific; nmlz = nominalizer; nom = nominative; pass = passive; prs = present; pst = past; sg = singular; top = topic; vblz = verbalizer.

Acknowledgements

I thank Neda Todorović, Henry Davis and Gunnar Ólafur Hansson for guidance on the early stages of this work, with special thanks to Neda for invaluable support and encouragement. I am grateful to the audience at UBC for their feedback on its initial presentation. I benefited from insights of Nayoung Kwon and Changguk Yim, as well as from their helpful pointers to relevant literature. This work would not have been possible without the support from all of them. I am also grateful to the anonymous reviewers for constructive comments and suggestions, and to my consultants Jieun Choi, Jihye Jeong and Hyojeong Kim for their intuitions. All remaining errors are mine.

Competing interests

The author has no competing interests to declare.

Notes

  1. Specifically, Rizzi (1997: 284) assumes that finiteness in verbal morphology reflects the abstract tense distinction. [^]
  2. Ryu (1995) and Kim et al. (2018) classify -ta as a complementizer, but this paper does not follow this convention and simply label it as a declarative suffix, following a reference grammar (Sohn 1994). As in (3a) below, -ta can occur together with -ko, a complementizer that introduces the embedded clause. Therefore, -ta should be something other than a complementizer. Another option can be S(peech) A(ct) head as Lee (2019: 116–127) suggests, but its syntactic behaviour differs from SA in languages like Japanese (Speas & Tenny 2003; Miyagawa 2012; 2017). Unlike -ta in Korean, Japanese SA cannot be included in an embedded clause since it takes a position higher than C (Miyagawa 2012). Moreover, Wiltschko (2019) and Lee (2020) propose that there is a different silent SA in Korean that corresponds to Japanese SA. [^]
  3. I follow Jung (2016), Kim et al. (2018: 176–179) and others in analyzing -ha as a phonetic realization of v. [^]
  4. The classification is based on Ryu (1995: 38). [^]
  5. Especially, it is not the concern of this paper what kind of functional heads -key and -e are. One might speculate, however, that the projections are SubP, similar to Hindi ki clauses as discussed in Messick & Alok (2021), albeit not as high as the Hindi subordination marker. [^]
  6. As an anonymous reviewer has pointed out, Korean also has an addressee honorific marker, morphologically encoded as -supnita on the verb. Although various aspects of the addressee honorific, especially its embeddability and cross-linguistic comparisons are interesting, they are not the central focus of this paper. Interested readers may refer to Portner et al. (2019) for discussion of the embeddability of the Korean addressee honorific clause, and to Alok (2020; 2021) and Jou (2024) for cross-linguistic studies. [^]
  7. An anonymous reviewer proposed that such constructions like (4b) might be infelicitous rather than ungrammatical, given that honorific relations are often dynamic and performative. However, (4b) is only acceptable in special contexts that serve pragmatic functions such as humour or irony. See Alok & Baker (2022, specifically fn. 15) for a similar case in Magahi. [^]
  8. Notably, subject honorification as syntactic agreement is not without controversy. Kim & Sells (2007) argue subject honorification is a discourse-driven expressive which functions as a marker of social status. However, see Choi (2010) for a theoretical argument against this perspective, and refer to Kwon & Sturt (2016; 2023) for empirical evidence undermining it. [^]
  9. In a contemporary newspaper corpus (Kim et al. 2011), most instances of the adverb are found with a political or military leader or monarch in the subject position. [^]
  10. See beim Graben et al. (2008) for an overview of linguistically relevant ERP components, including P600 and N400. [^]
  11. The Korean absolute construction is different from the absolute constructions in English such as ‘everything being equal,’ or ‘weather permitting.’ Korean absolutes are root clauses and do not form part of a larger structure. [^]
  12. Adding a tense marker to (7a–b) does not make them ungrammatical. However, it gets interpreted with a fixed temporal reference. [^]
  13. In the parametric theory of polarity phenomena propounded by Progovac (1994), Korean patterns with Catalan, Mandarin and Italian in that NPI raises and adjoins to IP (Progovac 1994: 84–88). [^]
  14. As illustrated in (2), ModP is between TP and MoodP. [^]
  15. An anonymous reviewer questions cross-linguistic validity of Cinque’s (1999) adverb hierarchy, echoing Ernst’s (2001: 110–113) critique. While the cross-linguistic support for the hierarchy remains debated, I base the argument on Lee (2000) who claims that Korean adverb ordering aligns with the sequence in Cinque (1999). This assumption is also adopted by Ko (2005) and Jeon & Shin (2022). [^]
  16. The intended reading is impossible in Korean. My consultants report that ‘completely’ can only modify ‘quickly,’ resulting in a similar interpretation to ‘very quickly.’ [^]
  17. An anonymous reviewer suggests (19a) is acceptable if the adverb wancenhi is removed or placed after the -ki embedding, as in (i) below. Although I do not evaluate (i) here, I acknowledge that targeted judgment experiments regarding adverb ordering would be a valuable avenue for future research.
      1. (i)
      1. a.
      1. Miaka [ppalli apecikkeyse ttenasyess-ki]-lul palassta.
      1.  
      1. b.
      1. Miaka [ppalli apecikkeyse ttenasyess-ki]-lul wancenhi palassta.
    [^]
  18. Lee (2009) claims that po- ‘try’ takes a VP complement. However, morphological evidence in (24a–b) and long passive data in (ii) below suggest that the complement is vP. The embedded object nolaytul ‘songs’ in (ii-a) cannot be the nominative subject in the passive example (ii-b), indicating that the try construction does not allow long passive movement. Lee (2009) reports that the same long passive construction as in (ii-d) is grammatical, but all my consultants judged (ii-d) ungrammatical. If the embedded domain were a VP, long passivization should be possible (Wurmbrand 2001). Note in (ii-b) and (ii-d), the passive morpheme is on try, not on the embedded verb.
      1. (ii)
      1. Long passivization
      1.  
      1. a.
      1.   Wuli-ka
      2.   1pl-nom
      1. [nolaytul-ul
      2. songs-acc
      1. pulu]-e
      2. sing-e
      1. poassta.
      2. tried
      1.   ‘We tried to sing songs.’
      1.  
      1. b.
      1. *Noraytul-i
      2.   songs-nom
      1. [
      2.  
      1. pulu]-e
      2. sing-e
      1. po-ci-ess-ta.
      2. try-pass-pst-decl
      1.   Lit. ‘Songs were tried to be sung.’
      1.  
      1. c.
      1.   Sensayngnim-kkeyse
      2.   teacher-hnom
      1. [Bil-ul
      2. Bill-acc
      1. mopem-ulo
      2. model-as
      1. sam]-e
      2. make-e
      1. po-si-ess-ta
      2. try-hon-pst-decl
      1.   ‘The teacher tried to make Bill as an example.’                      (= (65a) in (Lee 2009))
      1.  
      1. d.
      1. *Bil-i
      2.   Bill-nom
      1. [
      2.  
      1. mopem-ulo
      2. model-as
      1. sam]-e
      2. make-e
      1. po-ci-ess-ta.
      2. try-pass-pst-decl
      1.   Lit. ‘Bill was tried to be made an example.’                           (= (65b) in (Lee 2009))
    [^]
  19. Lee (2009) and Suh (2006), in the same vein, claim that -e simply combines the verb with another verb (Lee 2009: 190, 214), or that it is the ‘connector’ between the complement and the verb (Suh 2017: 631–663). [^]
  20. The non-overt subject in the embedded clause is tentatively assumed to be PRO but further investigation of its properties is beyond the scope of this paper. For discussion, see Sundaresan (2014) and Pesetsky & Wurmbrand (2021). [^]
  21. Contrary to the try construction, we can see that in the serial verb construction in (25b), the two verbs are semantically related, and they denote a single event happening as a sequence of two sub-events. Most importantly, the two composite verbs do not assign theta roles separately (Kim Gyeong-yeol 2019: 9). In (25b), sakwa ‘apple’ is assigned a theta role by one verbal complex composed of pa- ‘dig’ and mut- ‘bury.’ [^]
  22. Compare this to the grammatical sentence below, which has the same meaning but without the pro-form.
      1. (iii)
      1. [Mincika
      2. Minci
      1. sakwalul
      2. apple
      1. pa-e
      2. dig-e
      1. mutess]-ko,
      2. buried-and
      1. [Naun-to
      2. Naun-also
      1. paylul
      2. pear
      1. kki-e
      2. put.together-e
      1. mutess]-ta.
      2. buried-decl
      1. ‘Minci dug and buried an apple, and Naun buried pear with the apple.’ 
    [^]
  23. Although tangential to the main focus of the paper, one possibility is that the Hon head is merged separately with these suppletive verbs in the derivation. In such cases, Hon might be realized either as the null form or -si, which is later phonetically deleted. [^]
  24. For the distinction between the two negation types, see Martin (1992: 315) and Sohn (1994: 132–133). [^]

References

Adger, David. 2007. Three domains of finiteness: A minimalist perspective?. In Nikolaeva, Irina (ed.), Finiteness: Theoretical and empirical foundations, 23–58. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780199213733.003.0002

Alok, Deepak. 2020. Speaker and addressee in natural language: Honorificity, indexicality and their interaction in Magahi. New Brunswick, NJ: Rutgers University dissertation.

Alok, Deepak. 2021. The morphosyntax of Magahi addressee agreement. Syntax 24(4). 264–296. DOI:  http://doi.org/10.1111/synt.12213

Alok, Deepak & Baker, Mark. 2022. Person and honorification: Features and interactions in Magahi. Glossa: a Journal of General Linguistics 7(1). 1–35. DOI:  http://doi.org/10.16995/glossa.7675

Baker, Mark. 1985. The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16(3). 373–415.

beim Graben, Peter & Gerth, Sabrina & Vasishth, Shravan. 2008. Towards dynamical system models of language-related brain potentials. Cognitive Neurodynamics 2(3). 229–255. DOI:  http://doi.org/10.1007/s11571-008-9041-5

Choe, Hyon Sook. 1988. Restructuring parameters and complex predicates. Cambridge, MA: Massachusetts Institute of Technology dissertation.

Choi, Kiyong. 2010. Subject honorification in Korean: In defense of Agr and Head-Spec agreement. Language Research 46(1). 59–82.

Chomsky, Noam & Lasnik, Howard. 1977. Filters and control. Linguistic Inquiry 8(3). 425–504.

Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective [Oxford Studies in Comparative Syntax]. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780195115260.001.0001

Cinque, Guglielmo & Rizzi, Luigi. 2009. The cartography of syntactic structures?. In Heine, Bernd & Narrog, Heiko (eds.), The Oxford handbook of linguistic analysis, 51–66. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oxfordhb/9780199544004.013.0003

Eide, Kristin Melum. (ed.) 2016. Finiteness matters: On finiteness-related phenomena in natural languages. Linguistik Aktuell. Amsterdam: John Benjamins Publishing Company. DOI:  http://doi.org/10.1075/la.231

Ernst, Thomas. 2001. The syntax of adjuncts [Cambridge Studies in Linguistics]. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511486258

Im, Hong-Pin. 1983. Kwukeuy celtaymuney tayhaye [On the unconditional construction of the Korean language]. The Chin-tan Hakpo 56. 97–136.

Im, Hong-Pin. 2007. Celtaymunun way seyweya hanunka (1): Koyengkun (2006)ey tapham [Justifying the unconditional construction (1): a reply to Ko (2006)]. Morphology 9(2). 331–340. DOI:  http://doi.org/10.1385/1-59259-929-X:331

Jeon, Yoonho & Shin, Eun Young. 2022. Adverb positions in Korean EFL learners’ interlanguage. Yengeyengmunhak21 35(4). 341–368. DOI:  http://doi.org/10.35771/engdoi.2022.35.4.014

Jeong, Haegwon. 2020. Time reference and zero morph in Korean unconditional sentences. Linguistic Typology Research 5(1). 1–19.

Jou, Eunsun. 2024. Honorification as Agree in Korean and beyond. Glossa: a Journal of General Linguistics 9(1). 1–48. DOI:  http://doi.org/10.16995/glossa.9565

Jung, Hyun Kyoung. 2016. On the verbalizing suffixes in Korean and their implications for syntax and semantics. Lingua 179. 97–123. DOI:  http://doi.org/10.1016/j.lingua.2016.05.012

Kang, Myung-Yoon. 1988. Topics in Korean syntax: Phrase structure, variable binding and movement. Cambridge, MA: Massachusetts Institute of Technology dissertation.

Kang, Myung-Yoon. 2001. DP hypothesis and Korean nominal phrases. Korean Linguistics 13. 15–51.

Kim, Gyeong-yeol. 2014. A study on characteristics and argument structures of complex predicates in contemporary Korean. Cheongju, KR: Chungbuk National University dissertation.

Kim, Gyeong-yeol. 2019. The type classification of serial verb complex predicate constructions. The Journal of Yeongju Language & Literature 41. 5–27. DOI:  http://doi.org/10.30774/yjll.2019.02.41.5

Kim, Heunggyu & Kang, Beom-mo & Lee, Do-Gil & Chung, Eugene & Kim, Ilhwan. 2011. Trends 21 corpus: A large annotated Korean newspaper corpus for linguistic and cultural studies. Paper presented at the 23rd Joint International Conference of the Association for Literary and Linguistic Computing and Association for Computers and the Humanities. Stanford University.

Kim, Ji-Hye & Yoon, James H. 2009. Long-distance bound local anaphors in Korean—An empirical study of the Korean anaphor caki-casin. Lingua 119(5). 733–755. DOI:  http://doi.org/10.1016/j.lingua.2008.10.002

Kim, Jong-Bok & Sells, Peter. 2007. Korean honorification: a kind of expressive meaning. Journal of East Asian Linguistics 16(4). 303–336. DOI:  http://doi.org/10.1007/s10831-007-9014-4

Kim, Kyeong-min. 2019. The syntax of Korean anaphora: An experimental investigation. Vancouver, BC: Simon Fraser University dissertation.

Kim, Yong-Ha & Park, So-Young & Lee, Chung-Hoon & Choi, Kiyong. 2018. Hankwuke sayngseng thongsalon [Korean generative syntax]. Seoul, KR: Youkrack.

Ko, Heejeong. 2005. Syntax of why-in-situ: Merge into [Spec,CP] in the overt syntax. Natural Language & Linguistic Theory 23(4). 867–916. DOI:  http://doi.org/10.1007/s11049-004-5923-3

Koo, Bon-Kwan & Park, Jaeyon & Yi, Seonung & Lee, Jin-Ho. 2015. Hankwuke munpep chonglon [A general introduction to Korean grammar]. Seoul, KR: Jipmoondang.

Koopman, Hilda. 2005. Korean (and Japanese) morphology from a syntactic perspective. Linguistic Inquiry 36(4). 601–633. DOI:  http://doi.org/10.1162/002438905774464359

Kornfilt, Jaklin. 2007. Verbal and nominalized finite clauses in Turkish?. In Nikolaeva, Irina (ed.), Finiteness: Theoretical and empirical foundations, 305–332. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780199213733.003.0009

Kuno, Susumu. 1998. Negative polarity items in Korean and English?. In King, Ross (ed.), Description and explanation in Korean linguistics, 87–131. Ithaca, NY: Cornell University.

Kwon, Nayoung & Sturt, Patrick. 2016. Attraction effects in honorific agreement in Korean. Frontiers in Psychology 7. DOI:  http://doi.org/10.3389/fpsyg.2016.01302

Kwon, Nayoung & Sturt, Patrick. 2023. When power matters grammatically: Investigation of the processing of honorifics in Korean. Ms. DOI:  http://doi.org/10.2139/ssrn.4481367

Laszakovits, Sabine & Shen, Zheng (eds.). 2021. The size of things I: Structure building. Berlin: Language Science Press.

Lee, EunHee. 2019. Korean syntax and semantics. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/9781108265041

Lee, Ju-Eun. 2000. Korean adverbs. Rivista di Grammatica Generativa 25. 67–80.

Lee, Kum Young. 2009. Finite control in Korean. Iowa City, IA: University of Iowa dissertation.

Lee, Soo-Hwan. 2020. Speech act phrases in Korean nominal structures?. In Barrie, Michael (ed.), Japanese/Korean linguistics 27, 1–12. Stanford, CA: CSLI Publications.

Martin, Samuel E. 1992. A reference grammar of Korean. Tokyo, Japan: Charles E. Tuttle Company.

Messick, Troy & Alok, Deepak. 2021. Stripping in Hindi: Does clause size matter? In Laszakovits, Sabine & Shen, Zheng (eds.), The size of things I: Structure building, 247–263. Berlin: Language Science Press.

Miyagawa, Shigeru. 2012. Agreements that occur mainly in the main clause?. In Aelbrecht, Lobke & Haegeman, Liliane & Nye, Rachel (eds.), Main clause phenomena: New horizons, 79–112. Amsterdam: John Benjamins Publishing Company. DOI:  http://doi.org/10.1075/la.190.04miy

Miyagawa, Shigeru. 2017. Agreement beyond phi [Linguistic Inquiry Monographs]. Cambridge, MA: The MIT Press. DOI:  http://doi.org/10.7551/mitpress/10958.001.0001

Nikolaeva, Irina. 2010. Typology of finiteness. Language and Linguistics Compass 4(12). 1176–1189. DOI:  http://doi.org/10.1111/j.1749-818X.2010.00253.x

Park, Hyun-Jung. 2016. Consideration on DP properties of Korean NPs: with respect to extraction and DP acquisition by Korean children. Proceedings of 2016 joint international conference: From convergence to development, 197–202. Seoul, KR: The Association of English Language & Literature in Korea.

Park, So-Young. 2009. The syntax of numeral classifiers: A small clause inside a DP. Language Research 45(2). 203–230.

Pesetsky, David & Wurmbrand, Susanne. 2021. David Pesetsky and Susanne Wurmbrand deal with Complementation. https://www.youtube.com/watch?v=At_uH9UCXN0

Pollock, Jean-Yves. 1989. Verb movement, universal grammar, and the structure of IP. Linguistic Inquiry 20(3). 365–424.

Portner, Paul & Pak, Miok & Zanuttini, Raffaella. 2019. The speaker-addressee relation at the syntax-semantics interface. Language 95(1). 1–36. DOI:  http://doi.org/10.1353/lan.2019.0008

Progovac, Ljiliana. 1994. Negative and positive polarity. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511554308

Raposo, Eduardo. 1987. Case theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18(1). 85–109.

Rizzi, Luigi. 1997. The fine structure of the left periphery?. In Haegeman, Liliane (ed.), Elements of grammar: Handbook in generative syntax, 281–337. Dordrecht: Springer. DOI:  http://doi.org/10.1007/978-94-011-5420-8_7

Ryu, Dong-Suk. 1994. Agreement in Korean. Studies in Generative Grammar 4. 211–248.

Ryu, Dong-Suk. 1995. Parameter grammar of Korean. Seoul, KR: Seoul National University dissertation.

Sells, Peter. 1995. Korean and Japanese morphology from a lexical perspective. Linguistic Inquiry 26(2). 277–325.

Shibatani, Masayoshi. 1976. The grammar of causative constructions: A conspectus?. In Shibatani, Masayoshi (ed.), The grammar of causative constructions, 1–40. Leiden: Brill. DOI:  http://doi.org/10.1163/9789004368842_002

Sohn, Ho-Min. 1994. Korean. London: Routledge. DOI:  http://doi.org/10.4324/9780429286889

Son, Minjeong. 2006. Causation and syntactic decomposition of events. Newark, DE: University of Delaware dissertation.

Speas, Peggy & Tenny, Carol. 2003. Configurational properties of point of view roles?. In Di Sciullo, Anna-Maria (ed.), Asymmetry in grammar, 315–344. Amsterdam: John Benjamins Publishing Company. DOI:  http://doi.org/10.1007/s11049-013-9216-6

Suh, Cheong-Soo. 2006. Kwukemunpep [A Korean grammar]. Seoul, KR: Jipmoon.

Suh, Chung-Mok. 2017. Hankwukeuy muncang kwuco [Korean sentence structures]. Seoul, KR: Youkrack.

Sundaresan, Sandhya. 2014. Making sense of silence: Finiteness and the (OC) PRO vs. pro distinction. Natural Language & Linguistic Theory 32(1). 59–85. DOI:  http://doi.org/10.1007/s11049-013-9216-6

Todorović, Neda. 2016. On the presence/absence of TP: Syntactic properties and temporal interpretation. Storrs, CT: University of Connecticut dissertation.

Todorović, Neda & Wurmbrand, Susi. 2020. Finiteness across domains?. In Radeva-Bork, Teodora & Kosta, Peter (eds.), Current developments in Slavic linguistics. Twenty years after (based on selected papers from FDSL 11), 47–66. Frankfurt am Main: Peter Lang.

Travis, Lisa deMena. 2010. Inner aspect: the articulation of VP. Dordrecht: Springer. DOI:  http://doi.org/10.1007/978-90-481-8550-4

Wiltschko, Martina. 2019. Towards a formal typology of formality. Paper presented at the 12th Generative Linguistics in the Old World (GLOW) in Asia. Dongguk University.

Wurmbrand, Susanne. 2001. Infinitives: Restructuring and clause structure. Berlin: De Gruyter Mouton. DOI:  http://doi.org/10.1515/9783110908329