1. Introduction
This squib demonstrates that in Korean, depending on the selectional relations between tense and complementizers, and the restriction on the subject under certain complementizers, first- and second-person subjects can be dropped without being construed with a discourse topic. Therefore, the contents of pro are not always recovered in the same way in discourse-oriented null subject languages (Huang Reference Huang1984, Holmberg et al. Reference Holmberg, Nayudu and Sheehan2009, Barbosa Reference Barbosa2011).
The morphosyntactic contexts in which the person of a pro subject can only be interpreted as a discourse participant are shown in (1).Footnote 1 The well-formed clauses in (1) are marked by l-key and l-lay – what I call periphrastic irrealis complementizers consisting of irrealis l plus complementizer key or lay – which enables pro-drop, where pro must be understood as first person in (1a) and (1b) and second person in (1c).Footnote 2
Unlike -key, which can only introduce assertions, as in (1a), -lay can introduce either assertions (with an unmarked falling intonation), as in (1b), or questions (with ↑ rising intonation), as in (1c). Assuming that the null subject in (1) is syntactically present, the goals of this squib are to describe the morphosyntactic contexts of pro-drop in (1), and to show that the content of the null subject, particularly with respect to person, is restricted by the morphosyntactic markers -l-key and -l-lay.Footnote 3 By showing that the null subject must be a restricted element under these T-C combinations, I argue that pro-drop can be enabled by an irrealis C element selecting an irrealis T element in Korean, although the analysis of exactly how irrealis T-C licenses person features on the null subject will be a subject for future research. There are two sources of evidence to support this claim: (i) Constraints on the interpretation of null subjects, as illustrated in (1), and (ii) a ban on third-person referents as the subject of l-key and l-lay clauses; this restriction limits the interpretation of full DPs to either the speaker or the addressee.
This squib is organized as follows: after briefly introducing some core properties of Korean pronouns and complementizers in section 2, section 3 discusses the prohibition of third-person referents in l-key and l-lay clauses. Section 4 briefly reviews two null-subject licensing conditions that have been proposed for Korean. Section 5 discusses the compositionality of periphrastic irrealis complementizers and section 6 offers some conclusions.
2. Person in concord with matrix complementizers
Although Korean has various personal pronouns and their usage is pragmatically restricted, pronouns tend to be dropped when the referent is a speech act participant. For instance, semantically any first-person pronoun, including nay ‘I’, cey ‘I (humble)’, wuli ‘we’, and cehuy ‘we (humble)’, can be an overt subject in the clause in (2), yet only nay ‘I’ and wuli ‘we’ are simultaneously syntactically and pragmatically well-formed. The pragmatically ill-formed pronouns cey ‘I (humble) ’ and cehuy ‘we (humble)’ are indicated by “#”.
The person restriction and the interpretation of subject pro discussed in this squib are irrelevant to the pragmatic properties of pronouns and complementizers, because all of the pragmatically ill- and well-formed subject pronouns in (2) must carry [ +1st, –2nd].
Korean is a head-final language with complementizers in the right periphery; the basic order of a canonical clause is SOV. T and C elements are all morphological suffixes that attach to verb stems (i.e., V-T-CTYPE-(CINDIRECT)). Each morphophonological spellout of CTYPE is associated both with syntactic properties (i.e., clause-type distinctions) and with pragmatic properties (i.e., speech style distinctions) (see Ceong Reference Ceong2019 and references cited there). In this squib, I focus on the morphosyntactic properties of complementizers – their clause-typing properties and compatibility with other functional elements – rather than on their precise semantic and pragmatic functions.
3. The ban on third-person subjects and constraints on the interpretation of full DPs
In addition to the fixed interpretation of the person of null pro described in (1), l-key and l-lay ban third-person referents from being the subject of the same clause. The third-person pronouns ku ‘he’ and kunye ‘she’ are disallowed in both clauses in (3); overt second-person pronouns, such as ne ‘you’ and caki ‘you’ in (3a), and overt first-person pronouns, such as nay ‘I’ in (3b), are also prohibited.
As the person feature on the subject cannot be [3rd] under l-key and l-lay, even full nominals must be interpreted as first or second person, which yields imposter-like constructions (Collins and Postal Reference Collins and Postal2012). Kinship or title terms such as emma ‘mommy’, halmeni ‘granny’, and sensayngnim ‘teacher’ can refer to the first, second, or third person, depending on the choice of COMP. Consider the clause in (4), with acessi ‘uncle’, a familiar title or term of address for any adult man, which must be interpreted as an R-expression.
However, the same nominal must be interpreted as the speaker in l-key clauses such as (5a), and as the addressee in a l-lay clause with a rising intonation, such as (5b).
The first-person interpretation of acessi in (5a) is also confirmed by the incompatibility of the honorific subject marker si, as indicated. Si cannot occur with a first-person subject, but is compatible with second person (5b) and third person (4) subjects.Footnote 4 This phenomenon, in which the nominal loses its third-person interpretation in the clause, is referred to as an imposter construction (Collins and Postal Reference Collins and Postal2012). If an R-expression subject is not interpreted as either the speaker or the addressee, the clause will be ill-formed under l-key and l-lay. This observation supports my claim that l-key and l-lay require a null subject to be a discourse participant.
This restricted distribution of pro and the default person under l-key and l-lay must be explained differently from assuming a null topic under complementizers that have free selectional properties for the person feature on the subject; this default interpretation of a null subject in a clause is similar to what Zanuttini et al. (Reference Zanuttini, Pak and Portner2012) propose for so-called jussive clauses in Korean.Footnote 5 The differences between these two licensing conditions for pro-drop, namely a null-discourse-topic approach in tensed clauses (Huang Reference Huang1984) and a subject-agreement approach in tenseless clauses (Zanuttini et al. Reference Zanuttini, Pak and Portner2012), are reviewed briefly in the next section.
4. Discourse pro-drop and subject agreement in T-JussiveP
Huang (Reference Huang1984) proposes a theory of null discourse topics for pro-drop in discourse-oriented languages. Huang claims that pro-drop is possible when a null discourse topic binds the subject of a tensed matrix clause in discourse-oriented languages, including Korean. For instance, a null topic binds pro in most interactional utterances in Korean conversation, as in (6). The questions are marked by interrogative ni, as in (6a) and (6c), while the answer, in (6b), is marked by the underspecified expletive-like clause-typing complementizer e (Ceong Reference Ceong2019).
Unlike their English counterparts, not only are the Korean clauses without overt subjects in (6) well-formed, the null subjects are also unambiguously understood as the addressee (6a), the speaker (6b), and that class (6c), respectively. As the content of pro can be interpreted differently under the identical interrogative-typing complementizer ni, as in (6a) and (6c), evidently ni does not carry a particular person feature. This lack of agreement morphology in tensed clauses distinguishes Korean from the Portuguese type of pro-drop shown in (7), where the content of pro is licensed by and recovered through subject-verb agreement on the predicate chegaram.
The radical pro-drop phenomenon in Korean, such as in (6), forms the basis of the null discourse topic theory of Huang (Reference Huang1984) and others, such as Holmberg et al. (Reference Holmberg, Nayudu and Sheehan2009) and Barbosa (Reference Barbosa2011), in which a topic in the discourse determines the content of null subjects in tensed clauses.
The null-discourse-topic analysis can also account for the second person interpretation of a null subject in imperatives marked by the complementizer e, as in (8a). Although (8a) is interpreted as an imperative, and the null subject is therefore interpreted as second person, the interpretation of the null subject must be recovered by a topic (i.e., the addressee). This is because a first-or third-person subject is also compatible with the complementizer e, as in (8b).Footnote 6
This contrasts with the clause marked by la in (9). The null subject in (9a) is not licensed by a topic per se, because la itself restricts the subject to second person, and the clause to imperative. In la-clauses, the second-person pronoun ne ‘you’ is optional and first- and third-person subject pronouns are disallowed, as shown in (9b).
Based on the restricted interpretation of pro under la, Zanuttini et al. (Reference Zanuttini, Pak and Portner2012) argue that an interpretable second-person feature on a T-jussive head licenses a null second-person subject and makes the clause imperative; la is the overt realization of the Jussive head with second-person features. Having compared two imperatives with distinct head elements, we can conclude that the interpretations of the null second-person subjects of (8a) and (9a) are licensed and recovered differently within the same language. Either a topic or a morphosyntactic marker, such as a subject agreement marker on T-JussiveP (Zanuttini et al. Reference Zanuttini, Pak and Portner2012), can identify the content of the null subject.Footnote 7 In that sense, the morphosyntactic context of null subjects in l-key and l-lay clauses is somewhat similar to that of jussive clauses, in which morphosyntactic markers restrict the person of null pro subjects. These functional markers l-key and l-lay, however, cannot be seen as a single T or C element.
5. The properties of l-key and l-lay : TIRR-CIRR
I call l-key and l-lay “periphrastic irrealis complementizers” because the combination of two functional markers and intonation is what restricts the interpretation of null subjects.Footnote 8 Although the modal irrealis function of l ‘will’ is semantically transparent, and the clause-typing function of key is syntactically transparent, neither l nor key restrict the content of the subject by itself. The properties of l and key in clauses are indirectly identifiable by the form of the corresponding echo questions. The clause in (10b) is a well-formed echo question response to the statement in (10a).
As shown in (10b), one must use the modal keyss expressing volition (Koo and Lehmann Reference Koo and Lehmann2010), declarative ta, and the indirect complementizer ko to form the echo question response to (10a).
The morphemic status of l in l-key and l-lay is evident in minimal pair clauses where it does not appear. Although the relationships between key in (10a) and (11a) and lay in (10a) and (11b) are unclear, the minimal pairs with or without irrealis l are interpreted differently: Without irrealis l, (11a) is interpreted as an imperative, while (11b) is interpreted as a hearsay imperative.
The irrealis l is homophonous with a suffix l that occurs in noun complements or relative clauses. The l in (12a) has been analyzed as an adnominal marker associated with future tense/imperfective aspect/irrealis mood, which contrasts with the adnominal n associated with past tense/perfective aspect/realis mood in this domain (Kaplan and Whitman Reference Kaplan and Whitman1995, see footnote 2), shown in (12b). The irrealis l in (10a) and the adnominal l in (12a) are both translated as ‘will’ in English.
Based on its morphophonological form and its irrealis indication, I assume that the l in the irrealis negative marker mal should be identified as the same morpheme.
I therefore claim that l-key and l-lay are composed of elements from two distinct domains: TAM and C. The table in (14) shows that in matrix clauses, key and lay subcategorize for TAM differently from declarative ta, interrogative ni, and the underspecified complementizer e (cf. Ceong Reference Ceong2019).
When l co-occurs with either key or lay and a falling intonation, pro must be interpreted as the speaker; while when it co-occurs with lay and a rising intonation, pro must be interpreted as the addressee. Assuming that l is a T element and key and lay are C elements, the restricted interpretation of null subjects in clauses marked by the periphrastic irrealis complementizers discussed in this squib is a third strategy available in Korean for recovering the content of pro. Whether or not pro-drop in l-key and l-lay clauses is amenable to analysis along the lines of the T-JussiveP proposal of Zanuttini et al. (Reference Zanuttini, Pak and Portner2012) is left to future research.
6. Conclusions
Ever since Huang's (Reference Huang1984) conjecture that Korean, like Chinese and Japanese, is an agreement-lacking discourse pro-drop language, null-subject constructions in Korean have generally been treated as a homogeneous phenomenon (Holmberg et al. Reference Holmberg, Nayudu and Sheehan2009, Barbosa Reference Barbosa2011). There has been little recognition of the fact that subjects receive restricted interpretations for person depending on certain clause-type complementizers. This squib calls for detailed examination of the morphosyntactic properties of complementizers in pro-drop languages.
