Reference time in the dynamics of temporal dependency in Korean

PDF

Download

Open document

Transcript (if available)

Content Chapter 1. REFERENCE TIME IN THE DYNAMICS OF TEMPORAL DEPENDENCY IN KOREAN by Hyuna Byun Kim A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (LINGUISTICS) May 2011 Copyright 2011 Hyuna Byun Kim ii ACKNOWLEDGEMENTS Writing a Ph.D. dissertation is much like bearing and giving birth to a baby. One day, you get one tiny seed of idea that catches you and grows inside, nurtured with endless thoughts and dialogues. Through the ups and downs that absolutely fascinate your soul but sometimes almost drown you in the pain, you come to the point where you have to do the final push with all your strength to give birth to your beloved. Then, you see the fragile little one coming out to the world, finally. However, we all know that it is not the end, but a beginning. So, I do not mind the immaturity of my dissertation in its infancy, but rather enjoy its birth with thanks to many persons in my life, professional and private. God led me to this point in my academic journey even though many times I suffered from following my own instinct, rather than His guidance. He blessed me tremendously and provided great helping hands. For this, I first give Him thanks and glory. Then, I want to thank Professor Chungmin Lee in Seoul National University who introduced me to Linguistics. Professor Seungho Nam in Seoul National University has a special place in my life in that he showed me how to live as a linguist and Christian. I express my sincere gratitude to my dissertation committee members at USC, my respected teachers and mentors. Professor James Higginbotham, my advisor and committee chair, has generously supported me through my tumultuous long-distance study interrupted by life‟s unexpected events. Online communication was not always easy, but I benefited so much from the valuable discussions that we had. I give special thanks to Professor Roumyana Pancheva for thorough reviews of my drafts and detailed discussions over numerous difficult topics. Her encouragement, advice, and iii comments were instrumental to the completion of this dissertation. I would like to thank Professors Barry Schein and Andrew Simpson whose insightful comments tremendously helped me improve my dissertation. As expected, Professor Schein raised difficult questions and offered discussions, which helped me sort out what I knew and what I didn‟t. Professor Simpson was generous to join the committee a short notice, and he brought a unique perspective as an expert on East Asian linguistics. The support of late Professor Jean Roger Vergnaud will be forever remembered. He influentially encouraged me to keep up with my study and finish it through the difficulties presented by the long distance. I regret not having the opportunity to express my gratitude one last time. Professors Hajime Hoji and student advisor Joyce Perez helped me go through the difficult time of the last year of my study. Without their helpful advice, I would not been able to get back on track with my study. Also, I am richly blessed to have good friends, colleagues, and teachers that I met and studied together at USC. I will miss the serious and lively discussions that I had with them there. Very special thanks should go to my family that sustained me through the past 8 years of my study. My father, Youngmin Byoun, and my mother, Moses Lee, gave me a heart unafraid to face difficult problems and the strengthto meet the challenge joyfully. I owe much to my mother-in-law, Ok Ja Kim, and my father-in-law, Jong Myung Kim. I received their unconditional love and support onmy frequent trips to Los Angeles, year after year. I would like to thank all my Mokjang family members who have been there for me, sharing my struggles in their prayer. My husband, Jin Kwang Kim, was a mentor during my study. He always knew what I needed to hear, sometimes bitter and sometimes iv sweet, and loved me enough to tell me. What a brave man! I am truly thankful to him. Finally, my grateful heart goes toward my two little daughters, Noelle and Bethany. My dissertation grew with them. Actually, they helped it grow in a sense, so it also is their achievement. v TABLE OF CONTENTS Acknowledgements ii List of Tables vii Abbreviations viii Abstract ix Chapter 1. Introduction 1 1.1. Purpose of the study 1 1.1.1. Toward a unified theory of tense 1 1.1.2. Investigating temporal interactions 3 1.2. Background Assumptions 4 1.2.1. Defining Tense beyond idexicality 4 1.2.1.1. Reference time: a temporal pro 4 1.2.1.2. Non-deictic definition of tense. 14 1.2.1.3. Indexicality on syntax 17 1.2.1.4. Summary 20 1.2.2. The morphological factors in tense semantics 21 1.2.3. Making a non-quotation setting 26 1.3. Structure of the proposal 35 Chapter 2. Korean facts and Previous analyses 37 2.1. Matrix tenses 37 2.1.1. Temporal categories in Korean 37 2.1.2. „-nun- /  - ‟ as a non-past tense marker 40 2.1.3. Two kinds of „-ess-‟: a past tense 56 2.1.3.1. Tense marker & Aspectual marker 56 2.1.4. A double form of „-ess-ess-‟ 80 2.2. Subordinate Tenses 85 2.2.1. The preliminary issue: Complementation or Quotation? 86 2.2.2. Types of temporal interpretation 92 2.2.2.1. Conjunctive clauses: Indexical interpretation only 93 2.2.2.2. Non-complement subordination: both anaphoric and indexical interpretations 99 2.2.2.3. Complement clauses: Anaphoric interpretation only 108 2.2.3. Summary 113 Chapter 3. Dynamics in Temporal Dependency 115 3.1. The skeleton of the proposal 115 3.1.1. The temporal anaphor Rx 115 vi 3.1.2. The Speaker at a CP layer 117 3.1.3. The unified system 119 3.2. Temporal dependency in matrix tenses 119 3.2.1. Past in the future 120 3.2.2. Present in the past: the historical present tense 122 3.2.3. The Double form of „-ess-ess-‟ 125 3.3. Temporal dependency in subordinate tenses 134 3.3.1. The asymmetry 134 3.3.2. Explanation 141 3.3.2.1. The Rx movement versus the Speaker 142 3.3.2.2. The Speaker in multiple embedding 148 3.4. Cross-linguistic implications 157 3.5. Summary 161 Chapter 4. Tense and Indexicality 163 4.1. Subordinate tenses and indexical temporal adverbials 163 4.1.1. Identifying the Restriction across Languages 164 4.1.2. The restriction on a temporal pronoun 169 4.1.3. The restriction explained 171 4.2. The Double Access Reading effect in Korean 173 4.2.1. The alleged Double Access Reading 173 4.2.2. Clearing out the confusion 175 4.2.3. A movement analysis of a DAR-ish reading 179 4.2.4. The Proposed Analysis 181 4.2.4.1. A semantic conflict with indexicals 181 4.2.4.2. A DAR effect as a pragmatic resolution 183 4.2.4.3. Defending the proposal 185 4.2.5. Summary 186 Chapter 5. Conclusions and Prospects 188 Bibliography 190 vii LIST OF TABLES Table 1: A paradigm of inflectional suffixes by S.O. Sohn (1995) 37 Table 2: The tense and aspect system of Korean #1(Chung 2005) 39 Table 3: The tense and aspect system of Korean #2(Chung 2005) 39 Table 4: Temporal interpretation across clauses 114 Table 5: The temporal interpretation of present under past 134 Table 6: The pattern in complement clauses 159 Table 7: The pattern in relative clauses 160 Table 8: The present tense & the long distance anaphor (LDA) 161 Table 9: Adverb Position and Temporal interpretation (from Dickey 2001) 166 viii ABBREVIATIONS ACC Accusative COMP complementizer CONJ Conjunctive DEC Declarative FRT Future HON Honorific IMP Imperative IMPF Imperfective IND Indicative INT Interrogative Int Intended meanig Lit Literal meaning LOC Locative MOD Modal NEG Negation NOM nominative PST past PST.IMPF past imperfective PERF perfective PFCT Perfect PRS Present PRS.IMPF present imperfective PROG progressive REL relative clause TOP topic ix ABSTRACT This dissertation investigates the temporal interpretation of subordinate clauses, aiming to provide a uniform account for the temporal interpretation of both matrix and subordinate clauses in Korean. To this end, I propose a non-deictic analysis of tense in Korean that treats tense as a function denoting a binary temporal relation, not necessarily tied to deiticity, the same way for both matrix and subordinate clauses. I claim that the important element of tense responsible for a deictic or anaphoric interpretation of clauses is a Reference time, a temporal argument of tense, which I treat as a temporal pro in this paper. Thus, the proposed tense system of Korean is simple and neat in that a tense morpheme refers to either an anterior or overlapping relation across the board, maintaining its meaning independent of the context where it appears. Yet, the choice of the antecedent of a temporal pro and the restrictions posed on it determine the temporal interpretation of matrix and subordinate tenses. Therefore, the semantic interpretation of tense heavily relies on what a Reference time of tense can take as its antecedent. The proposed tense system has advantages in accounting for the phenomena which are otherwise attributed to lexical idiosyncrasies of Korean tense morphemes. The peculiar future time interpretation of a past tense morpheme as well as the so-called historic present tense is well explained in terms of the choice of the temporal antecedent of a Reference time. Under the proposed analysis, the meaning of a past or present tense morpheme remains constant, not adding any idiosyncratic lexical meaning of the given morpheme: the peculiar meanings come from the fact that a context selects a certain (contextually) salient time as the antecedent of a Reference time over a Speech time, x which is the default reference time. Also, the proposed analysis, where the deictic versus anterior past tense distinction does not hold anymore, is able to account for the temporal meaning of a double form of „-ess-ess-‟ in terms of its temporal presupposition. More importantly, turning to the temporal interpretation of subordinate tenses, this dissertation provides a novel account for the long standing issue of the temporal difference between complement and relative clauses. The present analysis reaches the solution by combining Higginbotham‟s (2002a, 2006) temporal movement analysis along with Giorgi‟s (2010) theory of indexicality. The main difference between complement c and relative clause tenses is accounted for by the presence/absence of the obligatory movement of a temporal pro to the binding position in Spec of a CP, following Higginbotham (2002a, 2006). The movement is justified when we consider that the speaker‟s coordinates introduced to every sentence can bind the temporal pro, yielding an indexical interpretation, but the movement to the peripheral position in Spec of a CP gets the temporal pro out of the domain, resulting in an anaphoric interpretation only. In contrast, the movement is optional in relative clauses, resulting in a temporal ambiguity. Finally, two types of temporal interactions are discussed in this paper. The interaction between tense and temporal adverbials reveals that the restriction on deictic temporal adverbials inside a complement clause is tightly connected to the effect on the choice of a Reference time, given by the modification of temporal adverbs. The other is the interaction between temporal interpretation and pragmatic information. It is shown that Korean does not have a Double Access Reading in a semantic sense, but the alleged double access effect comes from the pragmatic inference available during the temporal xi interpretation. 1 Chapter 1. Introduction 1.1. Purpose of the study 1.1.1. Toward a unified theory of tense The overall goal of the present study is to provide a unified theory of tense in matrix and subordinate clauses in the Korean language. The temporal interpretation of matrix tenses is a well-studied subject in the Korean linguistic literature (e.g. Martin 1954, Choi 1965, Na 1971, Suh 1976, C. Lee 1985, Sohn 1995, Nahm 1996, Chung 1999 2005 and many others), but theories of tense devised to deal with matrix tenses are found inapplicable to the temporal interpretation of subordinate tenses because subordinate tenses do not act like matrix tenses, as illustrated in and (1), (2)-a, and (2)-b below: (1) Mary-nun shiaytul-ey sa(l)-n-ta. Mary-TOP Seattle-in live-PRS-DEC (2) a. John-un [Mary-ka shiaytul-ey sa(l)-n-ta]-ko malhay-ss-ta John-TOP Mary-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC „John said that Mary lived in Seattle (at the saying time).‟ b. John-un [e 1 shiaytul-ey sa(l)-nun]-eyca 1 -rul manna-ss-ta John-TOP Seattle-LOC live-PRS-REL-woman-ACC meet-PST-DEC Reading #1: „John met a woman who lived in Seattle (at the meeting time).‟ Reading #2: „John met a woman who is living in Seattle (at the speaker‟s now). 2 Due to the matrix present tense in (1), the event time of „Mary‟s living in Seattle‟ overlaps with (the speaker‟s) speech time. Whereas the embedded present tense in complement clauses in (2)-a makes the embedded event time overlap with the time set by the matrix (past) tense, i.e. John‟s speech time, in (2)-b where the present tense appears in a relative clause, it overlaps either (the speaker‟s) speech time or the matrix event time, that is, the time of „John‟s meeting with a woman.‟ Most theories of tense, which are based on the traditional definition of tense – namely, that tense is a temporal category locating an event in time in relation to speech time (Lyons 1968) – give a straightforward explanation of the matrix present tense in (1): a given event ongoing at speech time. However, they inevitably face a problem in dealing with subordinate tenses in (2) because in those examples „speech time‟ is not necessarily associated with the interpretation of subordinate tenses such that the meaning of subordinate tenses cannot be derived in the same way with that of matrix tenses which are tied to speech time as one of deictic elements. With respect to the difference between matrix and subordinate tenses, it is a widely adopted assumption in the literature since Comrie (1985) that there are different types of tense, i.e. „absolute versus relative tenses 1 ,‟ which are considered to account for the different treatment of matrix and subordinate tenses. For instance, according to the 1 Comrie (1985) states that while absolute tense is a tense which includes as part of its meaning the present moment as deictic centre, relative tense refers to one that does not include it as deictic centre. Yet, note that Comrie (1985) further makes a sub-distinction of „absolute-relative tense‟ vs. „pure relative tense.‟ According to Comrie (1985), unlike „pure relative tense,‟ „absolute-relative tense‟ combines absolute and relative time references, as in the English past perfect that requires another reference time point to be involved along with speech time; for pure relative tense, an absolute time reference, that is, speech time, is not involved at all, as seen in English non-finite verb forms. 3 assumption, while the matrix tense in (1) is an absolute present tense, the complement clause tense in (2)-a is a (pure) relative present tense; and the relative clause in (2)-b can have both of them, yielding the ambiguous readings. Such an account based on the familiar distinction, however, is not advantageous in that separate analyses for matrix and subordinate tenses that take the same form need to be provided. In this paper, departing from the traditional definition of tense and related assumption discussed above, I aim to provide a unified account of matrix and subordinate tenses. Re-defining tense as a relation between a given event time and reference time along the same line of Giorgi & Pianesi (1997) and Higginbotham (20002, 2006), I argue that the temporal meaning of a tense morpheme itself remains the same across sentential structure, and the difference in the temporal interpretation of matrix and subordinate tenses in (1) and (2) comes from what will be taken as the reference time of the tenses. It will be shown that structural property constraints the selection of reference time. This non-deictic analysis of tense pursued in this paper provides a simpler means to treat matrix and subordinate tenses in a uniform way. 1.1.2. Investigating temporal interactions This paper concerns two types of temporal interactions. The first interaction to be investigated in this paper is between matrix and subordinate tenses in Korean. Particularly, the temporal interpretation of three types of subordination will be examined in relation to matrix tenses: subordinated tenses in complement, relative and adjunct clauses. It will be shown how and why a complement clause tense in (1) differs from a 4 relative clause tense in (2) or from an adjunct clause one in (3): (3) [ Mary-ka ttena- ]-ki ttaymuney John-un ul-ess-ta. Mary-NOM leave-PRS-because John-TOP cry-PST-DEC Reading #1: „Because Mary left (at that time), John cried.‟ Reading #2: „Because Mary is leaving (today / tomorrow), John cried.‟ The present tenses in an adjunct clause in (3) can be interpreted as simultaneous to either the speech time or the matrix event time, patterning with the present tense in a relative clause in (2). The heart of this discussion will be on the issue of why indexical interpretation is not available for the present tense in a complement clause in (1). The second interaction in question is the one between tense and temporal adverbials. How positional temporal adverbials interact with tense with the same temporal domain will be discussed. 1.2. Background Assumptions 1.2.1. Defining Tense beyond idexicality 1.2.1.1. Reference time: a temporal pro Since Reichenbach (1947), „Reference time (R-time, hereafter)‟ has been in great use in theories of tense. Identifying and introducing the role of R-time would be considered as Reichenbach‟s main contribution to theories of tense, thereby leading to a significant improvement on Jesperson‟s (1924) traditional analysis of tense where tense is defined as 5 a relation between Event Time (E) and Speech Time (S). Although it is shown that R-time is an important tool to define tenses in Reichenbach (1947) 2 , he was not explicit about what R-time is (Borik 2006, Klein 2009). In Reichenbach (1947), tenses are not primitives corresponding to „past, present and future times‟ (Binnick 1991), but the relation between three time points 3 of E-time, S-time and R-time determines the meaning of tenses 4 . Yet, unlike E-time or S-time that has a firm base to be anchored in a sentence, it is hard to pinpoint what kind of linguistic element R-time is anchored to. Reichenbach (1947) states that temporal adverbials are to set up or establish R-time. However, given that every sentence including those without temporal adverbials is to have R-time in its tense system, temporal adverbials cannot be treated as the one that R-time is necessarily associated with. It is often understood that Klein‟s (1994) TOPic time (T-time) broadly corresponds to Reichenbach‟s (1947) R-time. In Klein (1994), T-time is the time about which something is asserted, i.e., the time about which the speaker expresses something. But, Klein‟s T-time does not help understand the temporal nature of R-time because 2 Reichenbach (1947) views tense as a relation of three time points, Speech Time (S), Event Time (E) and Reference Time (R). In Reichenbach (1947), there is no direct relation between E and S, but rather the two are mediated by the relation to R. The change adopted by Reichenbach (1947) made it possible to distinguish the English simple past (in „he left‟) from the English past perfect (in „he had left‟) by the different relation of E to R: E, R - S for English simple past and E-R-S for English anterior past. Going a step further, Hornstein (1977), as one of neo-Reichenbachian developments, claimed that tense is a linearly ordered relation among S, R, and E syntactically as well as semantically. Thus, it follows in Hornstein (1977) that E, R - S is distinguishable from R, E - S because of the different linear order between E and R. 3 As Borik (2006) points out, it would be fair to understand Reichenbach‟s (1947) reference time as a temporal interval rather than a punctual temporal point. It is so because Reichenbach (1947) states that reference time is set by temporal adverbs, such as „yesterday‟ or „in 1943,‟ which refers to a duration rather than a time point. 4 It is generally assumed in the literature that Event time is associated with a Verb Phrase containing a verb describing a given event. Speech time is the speaker‟s „now,‟ i.e. the time at which the speaker makes an utterance. Reference time is considered as a time with respect to which the speaker makes a time reference. 6 Klein‟s T-time is vaguely defined as well. Since R-time was introduced by Reichenbach (1947), R-time has been discussed in three lines of theoretical approach in the literature: the analysis of narrative discourse, the semantic analysis of temporal adverbs, and the syntactic analysis of tense and aspect 5 . Since a detailed discussion of proposals from each line of approach is beyond the scope of this study, we will instead focus on the points from an approach that help us understand the temporal nature of R-time. First, R-time became in the heart of discussion in the temporal interpretation of narrative discourse because it was shown to be a useful tool to account for the sequence of narrative discourse in Kamp (1981a), Hinrich (1981), Partee (1984), and Kamp & Ryle (1993). As Partee (1984) points out, this line of approach contributed to developing a system that incorporates the notion of Reichenbach‟s R-time into a formal semantic framework. More specifically, Partee (1984), based on Hinrich (1981), assumes that R- time is a temporal anaphor: semantically, the R-time for each sentence is a free variable bound by a discourse-level existential closure operation (Dickey 2001). Partee (1984) gave up the idea that tense itself is referential just like pronouns, voiding the direct analogy between tense and pronouns previously claimed in Partee (1973). Partee (1984) proposes that the analogy should be made between R-time and a pronoun because R-time is a temporal anaphor 6 : the referential property of tense comes from the referential 5 For a more detailed discussion on the development of the notion of R-time in the literature, reference Dickney (2001, Ch.1) and Borik (2006, Ch. 5). 6 Although the direct referential theory proposed by Partee (1973) was discarded in Partee (1984), Kratzer (1998) kept the original idea of Partee (1973), and it has been a standard assumption in the literature about 7 meaning of the R-time that tense is associated with. In this study, we take Hinrich (1981) and Partee‟s (1984) assumption of the notion of R-time and use it for the analysis of tense in Korean. Note that the notion of R-time as a temporal anaphor was not used for temporal interpretation of tenses within a sentence because Hinrich (1981) and Partee (1984) are mainly concerned with the temporal interpretation of narrative discourse 7 . We propose that R-time is a „temporal pronoun,‟ which can be bound structurally or contextually, and that it serves as an argument of tense. Second, the analyses of temporal adverbs exploited the notion of R-time. As mentioned before, Reichenbach (1947) treated temporal adverbs as to have a function of setting up the R-time of a sentence but not to involve with E-time. But, such an analysis turns out to be inappropriate because it fails to account for the ambiguous readings that the sentence in (4) has (an example from Dickey 2001): (4) John had left [at 5]. Reading 1: John already left as of 5 o‟clock [Reference Time] Reading 2: John‟s leaving took place at 5 o‟clock [Event Time] If the time adverb „at 5‟ serves to set up only an R-time for the sentence in (4) as tense. 7 In Hinrich (1981) and Partee (1984), the temporally ordered sequence of narrative discourse is accounted for in terms of „Reference time movement.‟ Hinrich (1981) proposed that eventive sentences, unlike stative ones, introduce a new reference time, thereby making a reference time of narrative discourse move forward, which results in a sequential interpretation of discourse. 8 Reichenbach (1947) assumes, Reading 2 would not be available. Thus, the existence of the second reading of (4) indicates that temporal adverbs can serve to modify E-time as well as establish the R-time of a clause. In general, the ambiguity of the sentence (4) is explained in terms of which syntactic level the temporal modification of the time adverb is involved with, namely, a Verb Phrase or Aspect Phrase, as shown in Kratzer (1998). We do not take the assumption of Reichenbach (1947) and others that temporal adverbs are the direct source in establishing R-time. In this analysis, an R-time of a clause is one of time arguments of tense that functions as the center of an evaluation. So, it is required by tense whether a sentence contains temporal adverbs or not; as a temporal pronoun, the content of R-time is provided by the binding relation to its antecedent. Given these, the R-time of the past perfect in (4) is S-time, rather than the time span established by the temporal adverb. This becomes clearer in a simpler case like (5): (5) John left [at 5]. It is generally agreed that the E-time of „John’ s leaving‟ in (5) is evaluated with respect to S-time, thereby being located in the past from S-time, which should mean, in the present analysis, S-time serves as an R-time of the past tense. The time span established by the time adverb [at 5] does not play a role in establishing R-time in (5). The idea that S-time is a default case of R-time is not new. It is a basic assumption taken in the Discourse Representation Theory (DRT) proposed by Kamp & Ryle (1993). And many authors, including Partee (1984), Prior (1967), Mittwoch (1988), 9 Smith (1991), Giorgi & Pianesi (1997), Hatav (1997), Higginbotham (1999, 2002), share the assumption. Specifically, Hatav (1997) states that speech time is the limiting case of R, marked as R s 8 . Given that R-time is nothing else an evaluation time of tense, it is natural to treat S-time as a default case of R-time, following a widely adopted assumption in the literature. If this assumption is right, we can replace two Reichenbachian (1947) terms of R- time and S-time with one single term of R-time that functions as an evaluation time. Then, it follows that the Reichenbachian definition of tense should be modified. Following Giorgi & Pianesi (1997) and Higginbotham (2002, 2006), tense is defined in this paper as a relation between R-time and E-time. What implication the new definition has to the present analysis of tense in Korean will be discussed in details in the following subsection. Then, how do time adverbs contribute to the temporal interpretation of (4) and (5)? In the proposed analysis, positional temporal adverbs modify the E-time of tense: the [at 5] in (5) is an E-time modifier. As for the preposed time adverb in (4), we analyze it as a temporal „topic,‟ as proposed by Reinhart (1983). A similar view was couched in a different term such as „frame setters‟ (Chafe, 1984) or a „scene setting element‟ (Kuno, 1975) 9 . The tense adverb [at 5] in (5) moves to or is base-generated in the topic phrase just like other topical elements to receive the topical meaning. 8 See Borik (2006). 9 Klein‟s (1993) term „Topic time‟ would fit for referring to the time span set by preposed time adverbs as in (5) even though he did not extend to do so. For Klein (1993), Topic time, i.e. the time which is talked about is a time point /interval that determines the meaning of tense. 10 The Topic analysis can easily explain the fact that the ambiguity found in (4) disappears in (6), where a time adverb is preposed, as below (Dickey 2001): (6) a. [At 5] John had left Reading 1: John already left as of 5 o‟clock [Reference Time] Reading 2: #John‟s leaving took place at 5 o‟clock [Event Time] b. [At 5], John left. Only Reading 1 is available or strongly preferred when the time adverb „at 5‟ is preposed in (6)-a. In the present analysis, because the preposed time adverb in the topic phrase occupies the highest position out of a vP or an AspectP, it cannot modify the event time of „John’ s leaving‟ in the structure. The same analysis is applied to the simpler case as in (6)-b. Lastly, let us turn to the syntactic approach to R-time. Stowell (2007) 10 assumes that tense imposes a temporal ordering between Reference Time (RT) and Event time (ET), and that RT and ET are syntactically realized as internal and external arguments of a Tense Phase (TP) in a tree structure as below: 10 Stowell‟s (2007) tense structure is based on Zagona (1990), but he made a modification of it. 11 (7) TP ZP 1 (RT) T‟ T ZP 2 (ET) As shown in (7), RT, one type of a Zeit Phrase (ZP) 11 , comes in the specifier position (Spec) of a TP while ET, another ZP, appears in the complement position of a TP. The important point of his proposal with respect to R-time is that he interprets R-time as a temporal PRO. It means that R-time is an empty element whose content should be provided by its controller through a control relation; ZPs can serve its controller. But, Stowell‟s (2007) syntactic account of R-time faces problems because there are issues that make it hard to incorporate his notion of R-time into the standard assumptions of syntactic theories. First of all, it is generally assumed that PRO is allowed only in an ungoverned position; but, the temporal PRO, RT, appears in the governed position, i.e. Spec of a TP in Stowell‟s (2007) structure in (7); it is hard to imagine that PRO can be allowed in the subject position to which a both theta role and a case are assigned. Even if the temporal PRO can appear in such a position anyhow, it still brings a trouble because it occupies the position to which a subject Noun Phrase (NP) is usually assumed to move. If we put an Agr S (Subject) Phrase to host it, the subject NP should 11 Stowell (2007) presents a ZP as a Zeit Phrase (ZP), a temporal analogue of a Determiner Phrase (DP). 12 jump over the Spec of a TP, the intermediate one between a VP and Agr S P, violating the principle of cyclic movement. Finally, Stwoell (2007) is forced to assume another higher level of clause above matrix clauses in order to feed his temporal PRO in matrix ones with a controller; otherwise, his temporal PRO in a matrix clause will be left without any controller, which is problematic unless it is interpreted as an arbitrary PRO. The controller of a temporal PRO in a matrix clause, however, is not arbitrary but already fixed because the R-time of a matrix clause is to be S-time. Thus, he needs to introduce an „invisible‟ superordinate clause headed by a performative verb „say‟ in order to provide S-time as a controller of the temporal PRO in a matrix clause. It is a possible move, but it is not preferable to introduce several null form elements to explain the matrix tense meaning only. In the proposed analysis, I focus on the property of a reference time at a syntactic level and treat it as an argument of a temporal relation, that is, Tense, even though a reference time can be associated with a temporal interpretation of a discourse beyond a sentential boundary, serving as a discourse anaphor, or with another type of temporal interpretation without a TP as within a NOMinal phrase. In the paper, I propose that R- time is a temporal pronoun that takes a null form, pro. The basic framework proposed here with respect to a reference time resembles that of Stowell (1995, 2007) in many aspects, but the significant difference between the two lies in the fact that it is proposed here that a reference time is a temporal pro rather than PRO. As one of implicit arguments, unlike PRO, a temporal pronoun has no problem with appearing in a governing position and it does not require a controller in a super-ordinate verb phrase 13 above it. The alternative approach is found in Higginbotham (2002, 2006) which is close to Stowell‟s (2007) analysis in basic matters. In Higginbotham,like Stowell (2007), tense is a binary relation between two temporal coordinates of tense, R-time and E-time. But, Higginbotham (2002, 2006) differs from Stowell (2007) in that he does not posit that R- time takes a syntactic position; to Higginbotham (2002, 2006), R-time is an implicit argument that is not necessarily realized as a syntactic argument. It is plausible to think as in Higgintbotham‟s (2002, 2006) that Reference Time is not tied to a specific syntactic position, i.e., Spec of a TP, considering that a reference point for an evaluation is conceptually required but does not need to be obligatorily realized in syntax, as in some comparatives or evaluative predicates 12 : However, in fact, Higginbotham‟s (2002, 2006) analysis of tense does not benefit much from such a view of R-time. Higginbotham (2002, 2006) argues that a second coordinate of tense of a complement tense, corresponding to R-time obligatorily moves to Spec of a CP. He is suggesting that a temporal element which is optionally present in syntax undergoes an obligatory movement in a complement clause that determines the interpretation of a complement clause tense. If the movement of R-time is obligatory, it would be better to posit that it is syntactically visible. That is the reason I take Stowell‟s 12 Consider the following examples: (i) a. I‟m better person because of you. b. The cookie is tasty. The common thing between (i)-a and (ii)-b is that there should be a reference point for an evaluation. Yet, it does not need to be realized as a syntactic argument. Likewise, one can say that the R-time of tense is a temporal argument of tense, but it does not occupy the syntactic position of the Spec of a TP. 14 (2007) position with some important modification. Given this, the syntactic representation of tense in Korean will be given in (8) with a seemingly trivial substitution of PRO with pro, which has nevertheless a desirable consequence: (8) TP ZP 1 (R x =pro) T‟ T ZP 2 (ET) As shown in the structure given above, a temporal pro takes an external argument position of a TP, which functions as a local evaluation time of a clause. The temporal pro, an implicit argument of Tense should be provided with its semantic content. The mechanism is mainly based on the c-commanding relation between the temporal pro and its possible antecedents. It will be discussed later in Chapter 3. 1.2.1.2. Non-deictic definition of tense. In this paper, following Giorgi & Pianesi (1997) and Higginbotham (2002a, 2006), we 15 view tense as a binary relation 13 between two temporal elements, R-time and E-time; S- time is treated as a default case of R-time. When R-time remains unbound or free as in a matrix sentence, S-time will be given for the R-time as a default case. The advantage of this approach is that it enables us to analyze matrix and subordinate tenses in a unified way. Both matrix and subordinate tenses are evaluated with respect to a reference time, as schematized below: (9) A unified analysis of Tense: a. Tense in root sentences: [s 1 E 1  R 1 ] b. Tenses in embedded sentences: [s 1 E 1  R 1 [s 2 E 2  R 2 ] ] 13 In spite of Reichenbach‟s (1947) contribution to theories of tense, the tripartite tense system has led researchers to raise some important questions about its in/adequacy in accounting for tense in natural language. Following the line of Lindstedt (1985), Vikner (1985) and others, Verkuyl (2008) presented in details the shortcomings of the Reichenbachian system. Putting aside specific details of his arguments such that one tense form like „will write‟ undesirably occupies three cells in Reichenbach‟s (1947) system, we want to pay an attention to general objections given as follows (Verkuyl 2008:11): (i) a. There are languages that occupy considerably less than the nine tense cells of the matrix. b. Languages that have eight or nice tense forms cannot harbor them appropriately in the nine cells. c. There are languages that have considerably more than nine tense forms and as shown above, they cannot be harbored either. By presenting three points in (i) above, Verkuyl (2008) argued that the Reichenbach‟s (1947) tense system with fixed nine possible tense forms lacks the flexibility of dealing with languages with a small size of tense forms as well as a large size of tense forms. Rejecting Reichenbach‟s (1947) ternary tense system, Verkuyl (2008) defends for Te Winkel‟s (1866) binary tense system where three appositions are used to make up the system: (a) Present vs. Past (b) Synchronous vs. Posterior, and (c) Action in Progress (Incompleted) vs. Completed Action. The possible number of tense forms decreases depending on whether some of the three appositions are missing or inactive; it can expand beyond 10 or 11 tense forms when more appositions or dimensions are added to the system. We are not going into the details of Te Winkel‟s (1866) three dimensional cubic system that Verkuyl (2008) is arguing for, but Verkuyl (2008) should be considered as a serious attempt in semantics to make good challenge to the dominant Reichenbachian tradition and to offer an interesting alternative analysis. 16 Note that the abstract tense, represented by the symbol  above, is a binary function to take two temporal arguments whether in matrix and subordinate tenses, and that the choice of the temporal arguments affects the temporal interpretation of a sentence. In this unified analysis of tense, Comrie‟s (1985) distinction of relative tenses versus absolute ones fades away and the issue becomes under what conditions R-time is to be restricted to have a default value, that is, S-time. Finally, along with advocating of the binary tense system, we assume a Davisonian event-based semantic analysis of tense where an event, rather than time, is treated as an atomic entity, as given in Higginbotham (2002a, 2006): (10) a. John said that Mary was happy b. [ e 1 : (e 1 )  (u)] say (j, ^[ e 2 : (e 2 )  (e 1 )] & happy (m,e 2 )], e 1 ) In (10),  is a function from an event to a time associated with it; (e 1 ) is the matrix E- time and (u) is the matrix R-time, i.e. S-time. The LF in (10) helps us clearly see that tense is a relation between not only two time arguments but also ones associated with two events. Given that temporal information is closely related to events and tense functions to situate events in time, it seems more adequate to keep an event variable, rather than to introduce another time variable. In this line of approach, tense is a relation with no resemblance of pronouns, but what goes into a binding or controlling relation as pronouns do is the temporal argument of tense, which is R-time. The present paper takes this view of tense where the 17 parallelism between tense and pronouns comes naturally from the assumption that one argument of tense, R-time, is a temporal pronoun. 1.2.1.3. Indexicality on syntax Before we close this subsection, we need to turn to the issue of how indexicality is reflected in the course of temporal interpretation. It is widely assumed that the indexical information of the speaker and hearer plays an important role in the interpretation of tenses. Many authors have treated this speaker and hearer‟s information in temporal interpretation differently. In the traditional formal semantics, indexical elements, i.e., Kaplan‟s „monsters,‟ are treated to be determined by a context, formally, an assignment function (Heim and Kratzer 1998). Yet, more often, following Reichenbachian (1947) tradition, the term of „Speech time‟ has been used informally as a certain temporal entity that is associated with temporal interpretation. Ogihara‟s (1996) view of „Speech time‟ is still vague but a familiar one based on a philosophical ground. He argues that the strong tie of matrix tenses to Speech Time is not due to the property of tense itself, but due to the semantics of the speech act of saying: “uttering a sentence involves the speaker‟s self-ascribing the property denoted by the sentence (Ogihara 1996, p.242).” In other words, the highest tense with a widest scope configuration is to be interpreted with respect to Speech Time “via the speaker‟s self ascription” as a speech act phenomenon. In the sense that any tense appearing the outermost layer in a sentence is to be evaluated with respect to a higher element provided by the speech act, in Ogihara‟s (1996) tense typology, Japanese is classified as a language 18 that has only relative tenses 14 . However, the elaboration of the idea, i.e., how the speaker‟s self ascription works in semantics or syntax for the Speech Time-related interpretation of matrix tenses is not found in Ogihara (1996). Adopting Ross‟s (1970) proposal of a performative speech predicate, Stowell (1995a, 2007) took a step to elaborate the idea of embedding of matrix tenses and reflected the speaker and hear‟s information in syntax. According to his analysis, there is another invisible syntactic layer above a matrix clause, namely, a covert superodrinate clause containing a covert performative speech predicate of “I say to you…..” It follows 14 As shown in the table below, Ogihara‟s (1996) claims that Japanese is a pure relative language which only has relative tenses; English is also a relative language with an exceptional absolute tense of the English present tense, as given below: English Japanese type/meaning -s(present) no correlate absolute present ø(empty) -(r)u (present) relative present -ed (past) -ta(past) relative past have (perfect) -ta(past) relative past woll (future) -(r)u (present) relative future A true distinction between absolute and relative tenses can emerge in embedded clauses where the speech act of a speaker‟s self ascription is not always involved. So, the English present tense is taken as the example of an absolute tense which is inherently associated with Speech Time because it requires Speech Time for its interpretation even in an embedded context such as in a complement clause, yielding a peculiar Double Access reading. However, the English present tense does not seem to be absolute all the time. When it is embedded under a future tense, the Speech Time requirement is optional so that the sentence in (i) below have two readings in one of which the present tense of the relative clause is interpreted simultaneous to the matrix future tense. (i) John will marry a woman who is living in Boston. The fact that the relative clause tense in (i) can be interpreted simultaneous to the matrix tense indicates that the English present tense can be a relative tense, which means that English has to be classified as a relative languages as well according to Ogihara (1996). 19 from it that every matrix tense is embedded in a complement clause of the covert verb of speech whose temporal argument is Speech Time. Given this, matrix tenses would be interpreted relatively as one of embedded tenses 15 so that a parallel treatment of matrix and embedded tenses is made possible. In this way, deictic information including Speech Time is syntactically realized in the highest clause, but it is not part of the meaning of tense in Stowell (1995a, 2007). Other authors more rigorously attempt to reflect speaker and hearer‟s information in syntax. For the interpretation of imperative sentences, Platzack & Rosengren (1994) and Portner (2004) introduced Addressee Phase (AddrP); Mauck & Zanuttini (2004) introduced an Addressee / Speaker‟s phrase in order to account for various clausal types of imperatives, exhortatives and promisives. On the other hand, Giorgi (2010) claims that Spec of a CP is the position where the speaker‟s temporal location is specified; it means the S-time of the speaker can be introduced into a complement clause by being specified in Spec of a CP. Given that the speaker‟s temporal coordinates is tied to a syntactic position of a CP, it is predicted that when a CP is deleted for some reason, the access to the speaker‟s information is denied and the indexical reading becomes very hard to get. It is attested in the complementizer deletion (CD) phenomenon with the subjunctive form of verbs in Italian (Giorgi 1997). Thus, Giorgi (2010) relates the CD facts to the Double Access effect in Italian: when a CP 15 Yet, to Stowell (1995a, 2007) it is not right to say that matrix tenses are also relative to their higher tenses in a superordinate clause of a performative speech predicate as Ogihara (1996) would posit, because Stowell (2007) assumes that there is no tense in the superordinate clause which matrix tenses can bear a relation to. In Stowell (1995a, 2007), a reference time, but not tense, is the element which bears a dependent relation to Speech Time appearing in the superordinate clause. 20 is present, a double access reading (DAR) is possible while when it is absent due to the CD deletion, a DAR is blocked 16 . In this paper, I take Giorgi (2010) position that the speaker‟s temporal / spatical location is specified in a CP layer; the indexical information of the speaker and hearer is realized syntactically. It will be discussed in Chapter 3 in detail. 1.2.1.4. Summary The followings are the main assumptions of the present analysis, regarding the definition of tense, which have been discussed so far: (11) a. Tense is a binary relation between R-time and E-time b. R-time is a temporal evaluation time c. S-time is a default case of R-time d. R-time is a temporal pronoun; it can be bound by a structural antecedent through a binding relation. Or, it can be free, referring to S-time. e. Positional temporal adverbs modify E-time. f. The speaker and hearer information occupies a syntactic position. 16 However, such a correlation is not found in Korean because a CD, although possible, does not bring out any significant difference in temporal interpretation in Korean. 21 1.2.2. The morphological factors in tense semantics One issue to emerge in dealing with various tense forms is how we separate morphological factors from a semantic interpretation, or how we reflect the difference in forms in the course of interpretation. A Sequence of Tense phenomenon in English presents such an issue more clearly. The simultaneous reading of the English past under past form in (12) below leads us to ask whether the past tense morphology on the embedded verb „was‟ is a semantically vacuous or not. (12) John said that Mary was in Seattle The simultaneous reading of (12) requires the embedded past tense to be interpreted not as holding the typical anterior relation of a past tense form but as holding the anaphoric relation to the matrix tense, yielding a Sequence of Tense phenomenon. So, the mismatch between the past tense makeup and the non-past tense semantic interpretation has drawn researcher‟s attention for a long time. Higginbotham (2002) puts forth to a lexical ambiguity account positing two kinds of past tense, that is, A-past responsible for a simultaneous reading and B-past for a back-shifted reading. This view assumes that the former one is not derived from the latter one so that they are unrelated syntactically and semantically even though they ended up with having the same form maybe resulted by a historical change. Stowell (1999, 2007) presents a similar view where the English past tense of the simultaneous reading is just a morphological disguise and there should be another abstract level for a tense meaning to strip away the confusing morphological 22 makeup. From a different perspective, Ogihara (1989, 1996) attempted to resolve this problem by deleting the embedded past tense under the condition such that the embedded tense and its locally c-commanding matrix tense at LF are tenses of the same kind, namely, past tense. Since Ogihara (1989, 1996) removed the embedded past tense at all by a syntactic deletion rule, resulting in a tenseless embedded clause at LF, there is no discrepancy between the form and its semantic interpretation at LF. Ogihara (1989, 1996) adopted this way of analysis because it can give him a uniform account for English as well as Japanese both of which end up with having the same LF after the optional deletion rule application. Note that Japanese takes a bare verb with no overt tense form for the simultaneous reading corresponding to the English one in (13), as below: (13) Japanese: Taroo-wa [Hanako-ga Siatoru-ni i- -ru]- to it-ta Taro-TOP Hanako-NOM Seattle-LOC be-DEC-COMP say-PST-DEC „Taroo said that Hanako was in Seattle.‟ In (13), the embedded clause with a simultaneous reading to the matrix clause appears, at least morphologically, to have no tense form attached to a verb. Ogihara (1989, 1996) took this lack of a tense form literally and proposed a tenseless analysis for Japanese so that the symbol, , in (13) represents a morphologically, syntactically, and semantically null tense. Given this, Japanese has a more simple structure in the sense that the surface 23 form is directly put into a semantic interpretation while English needs a syntactic tense deletion rule, in order to eliminate the form and meaning discrepancy and get the same input LF as Japanese. Thus, he attributes the difference between English and Japanese to the parametric difference of the presence / absence of the tense deletion rule. Japanese may give us such a neat pattern due to its rather simple tense morphology. However, there is a minor but important morphological difference in Korean even though Korean keeps the same temporal structure and interpretation as Japanese. Pattering with Japanese, for adjectives and a copular verb, no overt tense form appears in the embedded clause with a simultaneous reading as in (14)-a. However, for rest of verbs, overt tense form „-n-/-nun-‟ should be attached to a verb as in (14)-b: (14) Korean: a. John-un [Mary-ka apu- -ta]- ko malha-ytt-ta John-TOP Mary-NOM sick-DEC-COMP say-PST-DEC „John said that Mary was sick (at that time).‟ b. John-un [Mary-ka shiaytul-ey sa(l)-n-ta]-ko malha-ytt-ta John-TOP Mary-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC „John said that Mary lived in Seattle (at that time).‟ (14)-b shows that Korean verbs require an overt tense form, „-n-/-nun-‟, a so-called present tense form. It means, according to Ogihara (1989, 1996) that Korean does not have a tenseless embedded clause, at least for a verb class, unlike Japanese so that it 24 needs the tense deletion rule, in order to delete the present tense form and get a tenseless clause for a simultaneous reading of (14)-b. However, the problem is that the Korean case in (14)-b does not meet the condition for tense deletion because the upstair and downstairs tenses are NOT the same kind: the embedded present tense and the matrix past tense c-commanding it. The tense deletion rule is not applicable to (14)-b. Then, the question is how (14)-b can get a simultaneous reading if it remains tensed at LF. Ogihara‟s (1989, 1996) tense deletion theory cannot provide an explanation of how (14)- b actually gets the same simultaneous reading as the Japanese one in (13), with a presence of a present tense form. Also, there is another point suggesting that Ogihara‟s (1989, 1996) null tense account is not convincing enough. There is no way in Ogihara (1989, 1996) to treat the present tense in root and embedded clauses in a proper way. The root present tense has the same morphological property as the null tense in the embedded clause, but the two tenses are a totally different kind in Ogihara (1989, 1996): a root present tense form is a deictic tense while the embedded clause does not have any tense at all even though they take the same surface form: (15) Japanese: a. Hanako-wa Siatoru-ni i- -ru. Hanako-NOM Seattle-LOC be-DEC „Hanako is in Seattle.‟ 25 a‟.Taroo-wa [Hanako-ga Siatoru-ni i- -ru]- to it-ta Taro-TOP Hanako-NOM Seattle-LOC be-DEC-COMP say-PST-DEC „Taroo said that Hanako was in Seattle (at that time).‟ Ogihara (1989, 1996) assumes that Japanese has two kinds of present tense. One is an absolute present tense appearing in root sentences and the other is a relative one which is actually a null tense without any morphological and semantic content in it. The apparent sameness in their surface forms might not be a big problem to Ogihara (1989, 1996) if it is just assumed as lexical ambiguity, especially when there is no tense in (15)-a‟. But, it cannot be considered desirable when the same form of a present tense should be treated totally unrelated in root and embedded clauses as in Korean below: (16) Korean: a. Mary-nun apu- -ta Mary-TOP sick-DEC „Mary is sick.‟ a‟. John-un [Mary-ka apu- -ta]- ko malhay-att-ta John-TOP Mary-NOM sick-DEC-COMP say-PST-DEC „John said that Mary was sick (at that time).‟ b. Mary-nun shiaytul-ey sa(l)-n-ta. Mary-NOM Seattle-LOC live-PRS-DEC „Mary lives in Seattle.‟ 26 b‟. John-un [ Mary-ka shiaytul-ey sa(l)-n-ta]-ko malhay-att-ta John-TOP Mary-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC „John said that Mary lived in Seattle (at that time).‟ As in (16)-b, the present tense form remains the same in the root sentence and the embedded one. Of course, they convey different meanings: one is a deictic present tense meaning and the other an anaphoric meaning. But, it would fail to capture the form and meaning shared by them if we simply assume two different lexical entries for them. Rather, we want to propose in this paper that the meaning of a present tense, which is a simultaneous relation between E and R, remains the same as the present tense form does, but the difference between the deictic and anaphoric meaning of them comes from the different choice of R each of the present tenses in (16)-b and (16)-b‟ takes and the restriction(s) imposed on the choice of R. 1.2.3. Making a non-quotation setting Before we continue to go into the interpretation of embedded clauses, we need to eliminate some irrelevant factors in order to make our discussion simple and straightforward. One of them is a direct speech interpretation which becomes available, when the (in)direct speech marker „-ko‟ appears, along with the indirect speech one under consideration. See that the Korean sentence in (17) below is ambiguous between direct and indirect discourse: 27 (17) John-un [Nay-ka shiaytul-ey sa(l)-n-ta]- ko malhay-att-ta John-TOP I-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC Reading #1(Indirect Speech): „John said that I lived in Seattle (at that time).‟ Reading #2(Direct Speech): „John said, “I live in Seattle” ‟ =„John said that he lived in Seattle (at that time).‟ While in the indirect speech reading of (17) „Nay(=I)‟ refers to the speaker, in the direct speech reading, it ends up with referring to the author of the direct speech, that is, „John.‟ Because in Reading #2 of (17) the elements in direct speech are interpreted independently of the rest of the matrix sentence, being isolated by the means of „Direct Quotation,‟ they are not embedded by the matrix verb and then it is clear that Reading #2 is not to be considered in the present discussion of the interpretation of matrix and subordinate tenses. Given this, the fact that the irrelevant reading of Reading #2 as well as Reading #1 is available with the speech marker „-ko‟ might cause a confusion in processing the sentence (17). Hence, we want to remove Reading #2 from our discussion by putting the sentence (17) into a non-quotation setting. In order to make a non-quotation setting, as Schlenker (2000) suggested, we can build an anaphoric relation across a clausal boundary by putting a 3 rd person pronoun in the subject position of a subordinate clause as in (18) and (19): (18) John-un [ ku-ka shiaytul-ey sa(l)-n-ta]-ko malhay-att-ta John-TOP he-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC 28 Reading #1(Indirect Speech): „John 1 said that he 1 lived in Seattle (at that time).‟ Reading #2(Direct Speech): „John said, “He lives in Seattle” ‟ =„John 1 said that he 2 lived in Seattle (at that time).‟ (19) Modun Sonyun-i [ku-ka shiaytul-ey sa(l)-n-ta]-ko malhay-att-ta Every boy-NOM he-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC Reading #1(Indirect Speech): „Every boy 1 said that he 1 lived in Seattle (at that time).‟ Reading #2(Direct Speech): „Every boy 1 said, “He 2 lives in Seattle” ‟ Reading #1 of (18) / (19) does have an anaphoric link that Reading #2 of (18) / (19) lacks, but there is no big difference between (17) and (18) / (19) in the sense that Reading #2 is still there interfering. Secondly, as in Anand (2002), we can use Negative Polarity Items (NPIs) to show that a targeted clause is not quoted but syntactically embedded. We can provide an example where a NPI appearing inside a downstairs clause is licensed by the negation in the matrix clause, indicating that the embedded clause is not a direct quotation, as in (20) 17 : 17 More natural example will be like the following: (i) a. ?John-un [han saram-to Mary-rul chohaha-n-ta]- ko saynkakha-ciahn-att-ta John-Top one person-even Mary-Acc like-PRS-DEC-that think-NOT-PST-DEC Reading #1(Indirect Speech): „John didn‟t think that anyone likes Mary‟ Reading #2(Direct Speech): „*John didn‟t think, “anyone likes Mary‟” ‟ 29 (20) a. John-un han saram-to chohaha-cian-nun-ta. John-TOP one person-even like-NEG-PRS-DEC „John does not like anyone.‟ b.*John-un han saram-to chohaha-n-ta. John-TOP one person-even like-PRS-DEC . „*(Lit.) John likes “even not-one person”.‟ c. ?John 1 -un [ e 1 han saram-to chohaha-n-ta]-ko malhay-ciahn-att-ta John-TOP pro one person-even like-PRS-DEC-COMP say-NOT-PST-DEC Reading #1(Indirect Speech): „John didn‟t say that he likes anyone‟ Reading #2(Direct Speech): „*John didn‟t say, “I like anyone‟” ‟ As we can see in (20)-a and (20) –b, „han saram-to‟(anyone/even one person) is a NPI that has to be licensed by a negation whose immediate scope the NPI falls into. Even though (20)-c is not as natural as (20)-a because the NPI and a negation, its licenser, are separated by a clausal boundary, (20)-c is regarded acceptable in which a NPI is still licensed by a remote negation in the matrix clause. It shows that the clause headed by „- ko‟ is not an isolated speech but a clause embedded by the matrix verb „malha-ta(say).‟ Notwithstanding, examples including a NPI and a negation will not be used to block a direct quotation reading from being associated with the marker „-ko.‟ It is because natural and acceptable examples of a NPI and a remote negation are rather restricted so that it is not easy to keep the same structure and verb selection as (17) and so some changes are avoidable as you can see in (17) and (20)-c. Thus, we would like to leave it 30 aside and find another way to deal with the matter. Thirdly, a subject honorific suffix „-shi-,‟ which is a grammatical tool to show the speaker‟s respect toward the subject of a clause, can be used as an indicator of a dependency between matrix and subordinate clauses 18 . Consider the following examples where an honorific suffix „-shi-‟ appears in a complement clause: (21) a. Apunim i -un [pro i Chulswu-lul pwull-ess-ta]- ko malhay-att-ta Father-TOP Chulswu-ACC call-HON-PST-DEC-COMP say-PST-DEC Indirect Speech Reading: „My father said that he i called Chulswu.‟ Direct Speech Reading: „My father said, “I called Chulswu” ‟ b. Apunim i -un [pro i Chulswu-lul pwuru-shi-ess-ta]- ko malhay-att-ta Father-TOP Chulswu-ACC call-HON-PST-DEC-COMP say-PST-DEC Indirect Speech Reading: „My father i said that he i called Chulswu.‟ Direct Speech Reading: (Int.)„My father said “I [whom I respect] called Chulswu” ‟ Notice the clear contrast found between (21)-a and (21)-b when the subject of the complement clause is coreferential with that of the matrix subject: while the complement clause can be read as either direct or indirect speech in (21)-a with no honorific suffix 19 , the direct speech reading is not available for (21)-b where „-shi-‟ occurs in the 18 I want to thank Andrew Simpson for suggesting to include an honorific suffix in the discussion for a non- quotation setting. 19 The two readings are not distinctive in the sense that the direct reading ends up with meaning the same with the indirect readings in (21)-a. 31 complement clause. It is so because the otherwise quoted part would be a sentence that sounds very odd, as below: (22) */??? Nay-ka Chulswu-lul puru-shi-ess-ta. I-NOM Chulswu-ACC call-HON-PST-DEC (Int.) „ I [whom I respect] called Chulswu.‟ Given that the subject of the downstairs clause is the same with that of the matrix clause, „Apunim (father),‟ the subject pro should be specified as a first person pronoun „nay(I)‟ as in (22). But, (22) sounds very odd or unacceptable because with the presence of the subject honorific suffix „-shi-‟ the subject „nay(I)‟ will be the object of the speaker‟s respect so that it will end up with meaning that the speaker shows respect toward himself 20 . Therefore, the direct speech reading is unavailable. However, we cannot jump to the conclusion that a given clause including „-shi-‟ is not direct speech but indirect speech because a direct speech reading becomes available when the subject pro of the downstairs clause refers to someone else about whom „apunim(father)‟ can appropriately use the honorific suffix, for example „father‟s teacher.‟ Then, the contrast shown in (21) disappears, as below: (23) a. Apunim i -un [pro j Chulswu-lul pwull-ess-ta]- ko malhay-att-ta Father-TOP Chulswu-ACC call-HON-PST-DEC-COMP say-PST-DEC 20 It is socially avoided in general unless the speaker is joking. 32 Indirect Speech Reading: „My father said that he i called Chulswu.‟ Direct Speech Reading: (Int.)„My father said, “(he) called Chulswu” ‟ b. Apunim i -un [pro j Chulswu-lul pwuru-shi-ess-ta]- ko malhay-att-ta Father-TOP Chulswu-ACC call-HON-PST-DEC-COMP say-PST-DEC Indirect Speech Reading: „My father i said that he i called Chulswu.‟ Direct Speech Reading: (Int.)„My father said “(he) [whom I respect] called Chulswu” ‟ (23)-b shows that a direct speech reading is available for the downstairs clause with „-shi-‟ if the content of the pro is properly provided by the context that makes possible „apunim(father)‟ to use the honorific suffix toward it. Hence, the subject honorific suffix „-she-‟ cannot be a definitive tool to eliminate a direct speech reading from the examples to be considered in this paper. Forth, a pause placed between the declarative ending of the downstairs clause and the (in)direct marker „-ko‟ can signal that the downstairs clause is a direct quotation. (24) John-un [Nay-ka shiaytul-ey sa(l)-n-ta]-<PAUSE>-ko<H> malhay-att-ta John-TOP I-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC Reading #1(Indirect Speech): „John said that I lived in Seattle (at that time).‟ Reading #2(Direct Speech): „John said, “I live in Seattle” ‟ =„John i said that he i lived in Seattle (at that time).‟ 33 When there is a pause after the downstairs clause as in (24), the preferred reading of a given sentence would be a direct speech reading. The pause is often followed by a high pitch of a speech marker „-ko.‟ This super-segmental phonetic information can make a distinction, thereby contributing to the resolution of the ambiguity involved with (24), but it cannot get rid of the possibility of the indirect speech reading of (24). Hence, we turn to our final consideration, a long distance anaphor „caki(self).‟ A long distance anaphor builds an anaphoric relation to its antecedent across a clausal boundary, not having the context-dependent indexical interpretation that the pronoun „ku‟(he) in (18) / (19) carries which makes the direct speech reading possible in (18) / (19). (25) John-un [caki-ka shiaytul-ey sa(l)-n-ta]-ko malhay-att-ta John-TOP self-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC Reading #1(Indirect Speech): „John said that I lived in Seattle (at that time).‟ Reading #2(Direct Speech): „John said, “I live in Seattle” ‟ =„John said that he lived in Seattle (at that time).‟ The fact that „caki‟(self) in (25) needs an antecedent and „John,‟ the matrix subject, is the only possible candidate for it demonstrates that there should be a syntactic relation between the matrix clause and the downstairs one. In other words, the downstairs clause including „caki‟ cannot be a direct quotation but an embedded clause. Here, we limit ourselves not to go for the argument such that „caki‟ is a logophoric pronoun appearing 34 only in indirect speech or reported speech. As mentioned of a Japanese case of „jibun‟ by Schlenker (2000), strictly speaking, „caki‟ cannot be considered as a logophor because it can show up in a non-logophoric environment, which is, a root clause as follows: (26) a.*caki-ka shiaytul-ey sa(l)-n-ta self-NOM Seattle-LOC live-PRS-DEC „*Himself lives in Seattle.‟ b. John-un caki-lul mit-nun-ta. John-TOP self-ACC trust-PRS-DEC „John trusts himself.‟ As in (26), „caki‟ is not restricted to appear in indirect speech, but also allowed in a root clause as long as it has a suitable antecedent, „John,‟ in this case 21 . Thus, a logophoric use of „caki‟ is not a strong argument that lets (25) receive an indirect speech reading only. We will stick to the anaphoric relation set up between „caki‟ and its antecedent in (25). Note that (26)-a is ungrammatical as „caki‟ cannot find a suitable antecedent, but it becomes grammatical when it is embedded in (25) because by embedding, the matrix clause provides an antecedent for „caki.‟ Thus, when (26)-a is embedded inside direct speech headed by a direct speech marker „-lako 22 ,‟ it is predicted that the sentence becomes ungrammatical. It turns out that this is the case, as shown in (27) below: 21 There is another view. One can assume that there are two kinds of „caki‟: one is a logophor and the other is a locally bound reflexive. 22 „-lako‟ is generally considered as a direct speech marker (S.O. Sohn 1995, H.M. Sohn 1999) that is not involved with an indirect speech reading in contrast to „-ko.‟ 35 (27) *John-un [ caki-ka shiaytul-ey sa(l)-n-ta]-lako malhay-att-ta John-TOP self-NOM Seattle-LOC live-PRS-DEC-COMP say-PST-DEC Reading #2(Direct Speech): „John said, “I live in Seattle” ‟ =„John said that he lived in Seattle (at that time).‟ Because the direct speech marker blocks „caki‟(self) from taking „John‟ the matrix subject by isolating the embedded clause inside direct quotation, the sentence (27) turns out to be ungrammatical. So far, we have shown that a long distance anaphora is the best solution to get rid of a direct quotation reading associated with a (in)direct speech marker „-ko.‟ It will be used in all examples to give us only one relevant reading, that is, an indirect speech reading. 1.3. Structure of the proposal Chapter 1 is introduction, and Chapter 2 takes a look into Korean facts and related previous analyses of matrix and subordinate tenses in Korean. In Chapter 3, the basic structure of the proposed analysis will be proposed and it will be discussed in detail how the system works for the temporal interpretation across subordinate clauses; it is tested whether the analysis can be extended to explain Polish / Russian data. And, Chapter 4 is about the interaction between tense and temporal adverbs. It will include the restriction on indexical temporal adverbials inside a complement clause and the matter of the 36 alleged Double Access Reading in Korean which is associated with the indexical interpretation in nature. Ch. 5 is conclusion. An appendix follows. 37 Chapter 2. Korean facts and Previous analyses 2.1. Matrix tenses 2.1.1. Temporal categories in Korean Let us briefly introduce inflectional suffixes consisting of temporal categories in Korean and the discussion of major issues on them in the Korean literature. One example of a paradigm of inflectional suffixes in Korean is given in S.O. Sohn (1995; p 24), based on Martin (1954, 1963), as below: (26) Table 1: A Paradigm of inflectional suffixes by S.O. Sohn (1995) VERB  STEM +INFLECTIONAL CATEGORIES INFL. CAT.  (SUBJECT HONOR) + (ASPECT) + TENSE + (MODAL) + (ADDRESSEE HONOR) + MOOD + CLAUSE-ENDER 23 CLAUSE-ENDER  COMPLEMENTIZER; SENTENCE TYPE COMPLEMENTIZER  Conjunctive (-ese, etc.) Adverbial (-key, -e, etc.) Adnominal (-n, etc.) Nominal(-ki, um) ASPECT  Perfect (-ess) 23 Parenthesized inflectional categories are optional. 38 TENSE  Past (-ess, - - ) Non past (- -) MODAL  V olitive (-keyss) Presumptive (-keyss) Predictive (-(u)l(i)) MOOD  Indicative (-(n)un / -ni) Retropective (-te / -ti) Requestive (-si, ) As shown in (26), S.-O. Sohn (1995) assumes that Korean has both of tense and aspect categories, but unlike tense aspect is an optional. Specifically, Korean has a past tense marker „-ess-‟ and a perfect marker of the same form „-ess-‟; the double form of „-ess-ess-‟ is analyzed as the orderly combined form of a perfect marker and a past tense marker. And, the null tense form, , expresses a present tense meaning, but maps to a past tense meaning as well. Chung‟s (2005) approach to the tense and aspect system of Korean is quite different, as presented below: 39 (27) Table 2: The tense and aspect system of Korean #1(Chung 2005) Deictic Tense Non-Deictic Tense Past -ess-ess- -ess- Present -nun- or - - According to Chung (2005), the past tense form „-ess-‟ which is the subject of a controversy is analyzed as a non-deictic tense with an anterior relation, i.e., a present perfect; the double past tense form of „-ess-ess-‟ is the deictic past tense corresponding to the English simple past tense; „-nun-‟ which is viewed as a mood marker in S.-O. Sohn (1995), is analyzed as a Deictic present tense, along with the null tense form of „- -‟ in the table above 24 . And, for the simple deictic tense, Chung (2005) makes a further distinction between a „simple present‟(S-Present) and an „imperfective present (I-Present)‟ as well as a „simple past (S-Past)‟ and an „imperfective past (I-Past),‟ as shown in (28): (28) Table 3: The tense and aspect system of Korean #2(Chung 2005) Simple Deictic Tense Spatial Deictic Tense S-Present * * I-Present -nun- or - - -ney S-Past -ess-ess- * I-Past -te- -te- 24 The table in (27) is made on the basis of Chung‟s (2005) proposal. 40 Chung (2005) claims that Korean does not have a simple deictic present tense corresponding to the English one; the deictic present tense forms of „-nun-‟ or „- -‟ carry an imperfective meaning such that the event described by a given verb is still ongoing at the utterance time. And, while the double form of „-ess-ess-‟ is analyzed as a simple deictic past tense, Chung (2005) treats „-te-‟ as an imperfective past tense when it appears in an attribute clause, unlike a spatial deictic tense „-te-‟ elsewhere. In the following subsections, I will discuss four issues related to determining temporal categories in Korean, evaluating proposals on those issues, including S.-O. Sohn (1995) and Chung (2005): (29) a. About identifying „-ess-‟ as a past tense marker or a perfect / perfective marker b. About identifying „-nun- / - - ‟ a present/non-past tense marker c. About identifying „-ess-ess‟ as an tense marker of an anterior relation 2.1.2. ‘-nun- /  - ’ as a non-past tense marker Korean has two types of a present tense marker: a covert form „ ‟ and an overt form „- nun-.‟ While adjectival predicates take a covert present tense form, verbal predicates take an overt one, as follows: (30) Adjectival predicates (in formal speech) a. Individual level predicates: 41 John-un ttokttokha-  / (*-nun-)-ta. John-TOP smart-PRS-DEC „John is smart.‟ b. Stage-level predicates: John-un apu-  / (*-nun-)-ta. John-TOP sick-PRS-DEC „John is sick.‟ (31) Verbal predicates (in formal speech) a. Activity verbs: John-un chwum-ul chwu-(*) / (nu)n-ta. John-TOP dance-ACC dance- PRS-DEC „John dances.‟ b. Achievement verbs: John-un cuk-(* ) / nun -ta. John-TOP die- PRS-DEC „John dies.‟ c. Accomplishment verbs: John-un cip-ul cis-(* ) / nun-ta. John-TOP house-ACC build- PRS-DEC „John builds.‟ d. Stative verbs: 42 John-un LA-ey sa(l)-(* ) / (nu)n-ta. John-TOP LA-LOC live-PRS-DEC „John lives in LA.‟ As shown in (30) and (31), the covert present form „ ‟ can only appear with adjectival predicates, but not with verbal ones; the overt one „-nun-‟ occurs with verbal predicates exclusively. Yet, there is an exception to the complementary distribution of the covert and overt present forms. The existential verb „iss-(be)‟ can take both of the present forms, as below in (32): (32) be-verb (in formal speech) a. John-un cip-ey honca iss- -ta. John-TOP home-LOC alone be-PRS-DEC „John is at home by himself‟ b. ?John-un cip-ey honca iss -nun-ta. John-TOP home-LOC alone be-PRS-DEC „John is at home by himself‟ Even though the overt present form „-nun-‟ attached to the existential verb „iss-‟ in(32)-b is less natural, compared to (32)-b ted, there is no significant difference in term of meaning between the two sentences with the overt and covert present forms in (32)-a and (32)-b. 43 The same pattern is preserved in embedded clauses: (33) a. Adjectivs: Mary-nun [ John-i apu-  / (*-nun-)-ta]-ko malhay-(e)ss-ta. Mary-TOP John-NOM sick-PRS -DEC-COMP say-PST-DEC „Mary said that John was sick (at that time)‟ b. Activity verbs: Mary-nun [ John-i chwum-ul chwu-* / -nun -ta]-ko malhay-(e)ss-ta. Mary-TOP John-NOM dance-ACC dance-PRS-DEC-COMP say-PST-DEC „Mary said that John danced (at that time).‟ c. Achievement verbs: Mary-nun [ John-i cuk-(* )/ (nu)n-ta]-ko malhay-(e)ss-ta. Mary-TOP John-NOM die-PRS-DEC-COMP say-PST-DEC „Mary said that John was dying / is going to die.‟ d. Accomplishment verbs: Mary-nun [ John-i cip-ul cis-(* ) / (nu)n-ta]-ko malhay-(e)ss-ta. Mary-TOP John-NOM house-ACC build- PRS-DEC-COMP say-PST-DEC „Mary said that John was building a house.‟ e. Stative verbs: Mary-nun [ John-i LA-ey sa(l) -*  / (nu)n-ta]-ko malhay-(e)ss-ta. Mary-TOP John-NOM LA-LOC live-PRS-DEC-COMP say-PST-DEC „Mary said that John was living in LA.‟ 44 f. Be-verb: ?Mary-nun [ John-i cip-ey honca iss-  / -nun-ta ]-ko malhay-(e)ss-ta. Mary-TOP John-NOM home-LOC alone be-PRS-DEC-COMP say-PST-DEC „Mary said that John lived in LA (at that time).‟ The examples in (33) show that the overt and covert present forms can embedded under indirect speech, not losing the same pattern with (30) and (31). The difference from the two groups is that while the present forms in root sentences are evaluated with respect to the speech time, the ones in embedded sentences are interpreted anaphoric to the matrix ones. How the two types of present tense with the same form and different interpretation can be accounted for in a unified way will be discussed in a later section. Let us point out that there is a disagreement in the literature about the categorical identity of the overt form„-nun-.‟ S.-O. Sohn (1995) and H.M. Sohn (1999) classified „- nun-‟ as one of indicative mood marker, claiming that Korean has only one present tense marker, which is the covert form „ .‟ However, the obvious objection against this analysis is the fact that the alleged indicative mood marker cannot co-occur with a past tense marker 25 : 25 For the whole paradigm, see below: (i) Verbal predicates (in formal speech) a. Activity verbs: John-un chwum-ul chwu-ess-*nun-ta. John-Top dance-Acc dance- PST-IND?-DEC „(Lit) John danced.‟ b. Achievement verbs: John-un cuk-ess-*nun-ta. John-Top die-PST-IND?-DEC 45 (34) John-un chwum-ul chwu-ess-*nun-ta. John-TOP dance-ACC dance- PST-IND?-DEC „(Lit) John danced.‟ If „-nun-‟ is an indicative mood marker, there is nothing preventing it from appearing with a past tense. But, as we can see in (34), „-nun-‟ cannot be preceded by a past tense marker „-ess-‟ at all. Then, there should be an ad-hoc stipulation such that „-nun-‟ is an indicative marker only for „a present tense sentence‟ even though a separate indicative mood marker for a past tense is missing in Korean. Indeed, another indicative mood marker that S.-O. Sohn (1995) and H.M. Sohn (1999) listed, that is, „-ni-‟ accompanying an addressee honorific marker „-sup-,‟ can occur with a past tense as well as a present tense 26 : „(Lit) John died.‟ c. Accomplishment verbs: John-un cip-ul cis-ess-*nun-ta. John-Top house-Acc build-PST-IND?-DEC „(Lit) John built.‟ d. Stative verbs: John-un LA-ey sa(l)-ess-*nun-ta. John-Top LA-in live-PST-IND?-DEC „(Lit) John lived in LA.‟ e. be-verb: John-un cip-ey honca iss-ess-*nun-ta. John-Top home-at alone be-PST-IND?-DEC „(Lit) John was at home by himself‟ 26 The whole paradigm is given below: (i) Verbal predicates (in formal speech) a. Activity verbs: John-un chwum-ul chwu-ess-sup-ni-ta. John-Top dance-Acc dance- PST-AHNR-IND-DEC „John danced.‟ b. Achievement verbs: John-un cuk-ess-sup-ni-ta. 46 (35) John-un chwum-ul chwu-ess-sup-ni-ta. John-TOP dance-ACC dance- PST-AHNR-IND-DEC „John danced.‟ In contrast to (34), another indicative marker „-ni-‟ following an addressee honorific marker can appear with a past tense as in (35). It is not properly explained in S.-O. Sohn‟s (1995) analysis why only „-nun-‟ is not allowed with a past tense marker. Another side argument against the indicative marker analysis is that it poses an odd constraint such that the alleged indicative marker „-nun-‟ cannot be attached to adjectival predicates as in (30); whereas, the same kind of indicative marker „-ni-‟ does not exhibit such a restriction. The inconsistency associated with „-nun-‟ indicates that the indicative marker analysis is not feasible. As mentioned before, from a different angle, Chung (2005) proposes that „-nun-/- Ø -‟ is an imperfective present tense, „I-present,‟ whose reference includes the speech time, rather than a simple deictic present tense. Her proposal of the imperfective present tense „-nun-/-Ø -‟ is based the observation that „-nun-/-Ø -‟ can bear by itself an imperfective John-Top die-PST-AHNR-IND-DEC „John died.‟ c. Accomplishment verbs: John-un cip-ul cis-ess-sup-ni-ta. John-Top house-Acc build-PST-AHNR-IND-DEC „John built.‟ d. Stative verbs: John-un LA-ey sa(l)-ess-sup-ni-ta. John-Top LA-in live-PST-AHNR-IND-DEC „John lived in LA.‟ e. be-verb: John-un cip-ey honca iss-ess-sup-ni-ta. John-Top home-at alone be-PST-AHNR-IND-DEC „John was at home by himself‟ 47 meaning with no help of a progressive form „-ko iss-,‟ as we can see in (36) below (from Chung 2005: p. 97): (36) Mina-ka phenyci-lul ssu-n-ta. Mina-NOM letter-ACC write-PRES.IMPF-DEC 'Mina is writing a letter / Mina writes letters.' In addition to a habitual or generic reading of a present tense, „-nun-/-Ø -‟ in (36) can convey the imperfective meaning that Mina‟s writing event is still ongoing or incomplete at the speech time. Chung‟s (2005) proposal is attractive in that it can nicely account for the peculiar imperfective meaning associated with the Korean present tense„-nun-,‟ which is lacking in the English present tense. So, Chung (2005), adopting the imperfective present analysis of Giori and Pianesi (1997:115), classifies the present tense in Italian and Korean to the same side and the English present tense to the other. It explains the cross linguistic difference in the present tense meanings in the three languages with respect to a co- occurrence restriction with past time adverbs, as follows (Chung, 2005: 107): (37) a. English: *I love Mary since yesterday. b. Italian: Amo Mary da ieri c. Korean: Mina-ka ithul-cen-pwuthe aphu-ta. Mina-NOM two.day-before-from. be.sick-DEC 48 '(Lit.)*Mina is sick for two days now.' ='Mina has been sick for two days now.' While the English present tense that cannot come with „since‟ as in (37)-a, there is no crash in meaning between the past time adverb and the present tense in Italian and Korean as shown in (37)-b or (37)-c respectively. So, Chung (2005) claims that it is due to the imperfective meaning of the null present tense marker „-Ø -‟ in (37)-c. Chung‟s (2005) analysis of the Korean imperfective present tense, however, runs into a problem when it combines with a progressive marker expressing imperfectivity. According to Chung (2005), Korean has two temporal items with an imperfective meaning: the imperfective present tense and a progressive marker. The one is a tense marker and the other an aspectual marker. Then, it is predicted that Korean has two layers of imperfectivity in a sentence with a present progressive marker because the two markers should appear combined: (38) Mina-ka phenyci-lul ssu-ko-iss-Ø -ta. Mina-NOM letter-ACC write-PROG-PRES.IMPF-DEC 'Mina is writing a letter / Mina writes letters.' Given that the present tense as well as a progressive marker has an imperfective meaning, the sentence in (38) is to be analyzed to have two layers of imperfectivity, which is redundant, which is avoided in Russian that has an imperfect present tense, but lacks a 49 progressive marker. Note that in Chung (2005) the two types of imperfective aspect are distinguished; taking Comrie‟s (1976) distinction, Chung (2005) claims that the imperfective is one of situation-external aspect (SEA) while the progressive is one of situation-internal aspect (SIA) in Korean 27 . However, the distinction does not help explain the redundancy in (38). Even if we can assume that the two layers of imperfectivity are a different kind, Chung‟s (2005) cannot properly explain why two aspectual categories that have an opposite direction for viewing a given event should be involved in (38); there is no difference in meaning in terms of imperfectivity between the present progressive with two layers and the present tense with one layer. One might argue that the imperfective present tense form is needed for adjectival predicates, justifying the redundancy. Because a progressive marker is not allowed to occur with adjectival predicates, an imperfective present tense marker is to be used for the imperfective meaning of adjectival predicates, as illustrated below: (39) a. Adjectival predicates: individual level predicates John-un cak-*ko iss- -ta. John-TOP short-PROG-IMP.PRS-DEC „(Lit.) #John is being short.‟ b. Adjectival predicates: stage level predicates John-un apu-*ko iss- -ta. 27 In a similar fashion, but from a different perspective, Giorgi and Pianesi (1997) distinguish the progressive from the imperfective: the former is an intensional operator and the latter is an existential operator. 50 John-TOP sick-PROG-IMP.PRS-DEC. „(Lit.) #John is being sick.‟ c. Be-verb: John-un cip-ey honca iss-*ko iss- -ta. John-TOP hoem-LOC alone be-PROG-IMP.PRS-DEC „(Lit.) #John is being at home alone.‟ All of adjectival predicates and an existential „be‟ verb in (39) cannot be followed by a progressive form „-ko iss-‟; only a null form of a present tense is allowed with this type of predicates. The first question to ask will be whether the imperfective meaning in question is compatible this type of predicates, motivating the presence of the imperfective present tense in (39). It is a general tendency across languages that those predicates are generally not compatible with an imperfective meaning of a progressive marker, as we can see in the English glosses in (39). The facts indicate that the stative meaning of those predicates blocks the aspectual meaning of imperfectivity 28 . Given this, it follows that the present tense in (39) cannot have an „imperfective‟ meaning because it is compatible with adjectival predicates and a „be‟ verb in Korean. In this paper, we defend the analysis that Korean has a past versus non-past 28 The prohibition of the Korean progressive form in (39) is not due to any morphological or phonological constraint. Also, it is not because the Korean progressive form has a more strict restriction on predicate types that it takes; it is well known that in Japanese and Korean a progressive marker is more generous to allow before it even psyche verbs like „sayngkakha-ta (think),‟ „al-ta (know),‟ or „cohaha-ta (like)‟ which are considered to have rather non-eventive meanings. Considering these factors, it becomes clearer that the strong stative meaning of adjectival predicates and a „be‟ -verb is closely related to the incompatibility of the Korean progressive form in (39). 51 apposition and the non-past tense meaning is conveyed by the covert marker  and the overt one „-nun-‟: (40) a. Adjectives: John-un apu-Ø / ass-ta. John-TOP sick-PRS / PST-DEC „John is sick / was sick‟ b. Activity verbs: John-un chwum-ul chwu-(nu)n / ess -ta. John-TOP dance-ACC dance- PRS / PST-DEC „John dances / is dancing / John danced ‟ c. Achievement verbs: John-un cuk- nun / ess-ta. John-TOP die- PRS / PST-DEC „John dies / is dying / died‟ d. Accomplishment verbs: John-un cip-ul cis- nun / ess-ta. John-TOP house-ACC build- PRS / PST-DEC „John builds a house / is building a house / John built a house.‟ e. Stative verbs: John-un LA-ey sa(l)-(nu)n / ass -ta. John-TOP LA-LOC live- PRS / PST-DEC 52 „John is living / lives in LA / John lived in LA.‟ f. be verb: John-un cip-ey honca iss-Ø / ? nun / ess-ta. John-TOP hoem-LOC anlone be- PRS / PST-DEC „John is at home alone // John was at home alone.‟ All the present tense sentences in (40)-a through (40)-b convey a non-past meaning, contrasting with the past tense ones. As we have already discussed, the overt and covert present tense forms in (40) can express the actual present tense meaning that locates events at the speech time, with no help of a progressive form. In addition to it, they can have other meanings beyond a pure present tense meaning, such as a habitual or generic one. For instance, (40)-c can mean that John dances repeatedly in a certain day or John is a dancer. Like Japanese cases, Korean present tense forms, whether a covert or overt one, can be read as holding a so-called future tense meaning, without any future tense morpheme attached 29 : 29 The future tense meaning of a present tense form is not limited to motion verbs that are more likely associated with a future tense meaning. See a complete pattern for verbs: (i) a. Activity verbs: John-un nayil chwum-ul chwu-(nu)n-ta. John-Top tomorrow dance-Acc dance- PRS-DEC „John will dance tomorrow‟ b. Achievement verbs: (John is a main character in a drama and the speaker is talking about what will happen to John tomorrow in the drama.) John-un nayil cuk-nun-ta John-Top tomorrow die- PRS-DEC „John will die tomorrow‟ c. Accomplishment verbs: 53 (41) a. (John saw Mary walking across a street and spoke the following sentence) Mary-ka ceki oh-(nu)n-ta. Mary-NOM over there come-PRS-DEC „Mary‟s coming over there.‟ b. (John talked to Sue over the phone and found that Mary will be coming tomorrow) Mary-nun nayil oh-(nu)n-ta. Mary-TOP tomorrow come-PRS-DEC „Mary‟s coming tomorrow.‟ When there is a future time denoting adverb such as „nayil (tomorrow)‟ as in (41)-b, a future tense meaning of a present tense form becomes strong. But, such a meaning is still available with the absence of the adverb. The covert present tense form also has a future tense meaning 30 , as given below: ?John-un nayil cip-ul ci(s)-nun-ta. John-Top tomorrow house-Acc build- PRS-DEC „John will build a house tomorrow‟ d. Stative verbs: John-un naynyen-pute LA-ey sa(l)-(nu)n-ta.. John-Top next year-from LA-in live-DEC „John will live in LA from next year.‟ The examples in (i) might sound a bit odd or too formal with the formal declarative ending. For more natural examples, we can change the ending to an informal declarative one, „-e/a,‟ keeping all the other things the same. 30 A modal affix „-(u)l-‟ seemingly convey the same future tense meaning, but such a meaning obtained when the speaker‟s prediction is made on a non-past situation in (i) below: (i) a. (John talked to Sue over the phone and found that Mary will be coming tomorrow) Mary-nun nayil oh- -(u)l-kesi-ta. Mary-Top tomorrow come-PRS-PRD-thing-DEC „(I guess) Mary will be coming tomorrow.‟ 54 (42) a. Sue: John-i nayil eti ka- -ni? John-NOM tomorrow where go-PRS-Interrogatives „Is John going somewhere?‟ b. Mary: Ani, John-un (nayil) cip-ey iss- -e. No, John-TOP tomorrow home-in be-PRS-DEC „No, John will be at home tomorrow‟ The null tense form in (42)-a and (42)-b is responsible for a future tense meaning matching with the future time denoting adverbial. One thing to note is that the future tense meaning by a covert present tense marker is not available with adjectival predicates: (43) a. John-un ttokttokha- -ta. John-TOP smart-PRS-DEC „John is smart.‟ b. John-un apu- -ta. „-(u)l-‟ is generally treated as a „predictive‟ modal affix (S.O. Sohn 1995) that expresses the speaker‟s prediction or conjecture. Because it is not a tense marker but a modal item following it, it is expected to co- occur with a past tense, unlike a present tense form „-(nu)n-‟ or „ ,‟ as it turns out to be true in (ii): (ii) a. (John talked to Sue over the phone and found that Mary will be coming tomorrow) Mary-nun ecey oh-ass-(u)l-kesi-ta. Mary-Top yesterday come-PST-PRD-thing-DEC „(I guess) Mary came yesterday.‟ The compatibility of the modal affix with a past tense in (ii) demonstrates that it is not the modal affix but the covert present tense „ ‟ in (i) that is responsible for a future tense meaning. 55 John-TOP sick-PRS-DEC „John is sick.‟ A future tense reading does not arise in (43)-a and (43)-b with adjectival predicates. Only present tense meaning is available. Thus, when „nayil(tomorrow)‟ is inserted to (43)-a/b to modify the adjectival predicates, the sentences become ungrammatical or unacceptable. The semantic property of adjectival predicates blocks the future tense meaning which is otherwise available with the non-past tense marker „ ‟ in (43). Yet, see that a predictive modal affix is present, a future tense meaning becomes possible: (44) a. (Mary is speaking, watching a 1 year old baby John talking and understanding short words) John-un (ku-myen) ttokttokha- -(u)l kes-i-ta. John-TOP grow-Conditional smart-PRS-PRD-thing-COP-DEC „(I guess) that John will be smart when he grows up.‟ b. (Mary is speaking, watching a 1 year old baby John coughing badly while he is sleeping) ?John-un (ama) nayil apu- -(u)l kes-i-ta. John-TOP probably tomorrow sick-PRS-PRD-thing-COP-DEC „(I guess) John will be sick tomorrow.‟ With the supporting examples discussed so far, we can reach the conclusion that the overt 56 and covert tense markers of „-nun-‟ and „ ,‟ are a non-past tense marker that can have a present and future tense meaning as well. 2.1.3. Two kinds of ‘-ess-’: a past tense 2.1.3.1. Tense marker & Aspectual marker There has been a long-standing controversy in the Korean literature on the grammatical identity of „-ess-‟ and „-ess-ess-‟ as a temporal or aspectual suffix 31 . Let us consider the simple form „-ess-‟ first. Among many, the three major analyses will be examined in this section: the past tense analysis, the aspectual analysis, and the ambiguity analysis of the tense and the aspectual marker. The past tense analysis is a null hypothesis put forth by most traditional grammarians (Choi 1965; Gim 1980, 1985 and many others). This analysis presents „-ess-‟ as a past tense placing a given event before S-time. (45) a. John-un apu-ass-ta. John-TOP sick-PST-DEC „John is was sick‟ b. John-un chwum-ul chwu-ess-ta. John-TOP dance-ACC dance-PST-DEC „John danced.‟ 31 See Chung (2005) for detailed discussion. 57 The eventuality described by the predicates „apwu-ta (be sick)‟ or „chwu-ta (dance)‟ in (45) is placed before S-time due to the past tense marker „-ess-.‟ The past tense meaning is straightforward in this analysis, but it is easy to find examples where „-ess-‟ does not convey the simple past tense meaning, as below (from S.-O. Sohn 1995): (46) a. [Naynyen-ey Seoul-ey ka-ss-ul ttay], Kim-ul manna-l kes-i-ta. Next year Seoul-LOC go-?-REL time Kim-ACC meet-REL-fact-COP-DEC „Next year when I go to Seoul, I will meet Kim.‟ b. Sunhi-ka cikum tochakhay-ess-ta. Sunhi-NOM now arrive „Sunhi has just arrived.‟ As shown in (46), „-ess-‟ can come with a present time or future time denoting time adverb. In (46)-a, „-ess-‟ denotes the future time point / interval which is specified by the time adverb „nayneyen-ey (next year).‟; (46)-b shows that „-ess-‟ can be used with a present time adverb „cikum (now),‟ indicating that the „-ess-‟ in (46) is not a simple past tense 32 . 32 S.-O. Sohn‟s (1995) original example for (46) is slightly different. It contains „cikum mak (just now)‟ rather than „cikum (now),‟ as below: (i) Suni-ka cikum mak tochakhay-ess-ta. S.-Nm now just arrive „Suni has just now arrived.‟ The sentence in (i) is fine, but it does not actually support that „-ess-‟ in it is not a past tense marker. The assumption that „cikum mak (just now)‟ is a present time denoting adverb, taken by S.-O. Sohn (1995) is wrong because unlike „cikum (now)‟ it cannot come with a present tensed sentence as below: 58 Furthermore, the following example are generally considered as the supporting evidence for the proposal that „-ess-‟ is not a past tense: (47) a. Ney os-ey hulk-i mwut-ess-ta. your cloth-LOC mud-NOM stain-PER-DEC 'Mud has stained your clothes.' b. Mina-ka chengbaci-lul ip-ess-ta. Mina-NOM jeans-ACC wear-PER-DEC 'Mina has put on jeans.' c. Mina-nun nulk-ess-ta. Mina-TOP get.old-PER-DEC 'Mina is old.' Nahm (1978) The English glosses of the Korean sentences in (47) indicate that the Korean sentences with „-ess-‟ do not describe past situations but present ones. With these facts, Nam‟s (1978) argues that „-ess-‟ cannot be a past tense. Apparently, the examples given above show that the past tense analysis of „-ess-,‟ based on the traditional deictic definition of (ii) a. Suni-ka cikum tochakha-n-ta. S.-Nm now arrive „[Lit.] Suni is arriving now .‟ b. #Suni-ka cikum mak tochakha-n-ta. S.-Nm now just arrive „[Lit.] *Suni is arriving just now .‟ The difference shown in (ii) indicates that „cikum mak (just now)‟ should be considered as a past time denoting time adverb. Then, the grammaticality of the sentence in (i) only supports that „-ess-‟ in (i) is a simple past tense. That is the reason I changed the example in (46)-b into the one with „cikum (now).‟ 59 tense, is not feasible. But, when we define tense differently, as done in this thesis, (46)-a is not problematic anymore because „-ess-‟ is analyzed in this paper as denoting a relative past time from some future time, rather than S-time. Yet, for other examples, we need another explanation. So, we now turn to the second analysis, i.e., the perfect analysis of „- ess-.‟ Several authors including Na (1971), Kim (1974), Suh (1976), Nahm (1978), Huh (1987), Chung (2005) and many others advocate the (present) perfect analysis of „-ess-.‟ Even though the authors do not agree upon the aspectual meaning of „-ess-‟, resulting in proposing different analyses; Na (1971), Nahm (1978) and Huh (1987) argue that „-ess-‟ is a perfective marker 33 . For Kim (1974) and Suh (1976), „-ess-‟ is a combined category of tense and aspect, covering the dual function. Chung (2005) proposes that „-ess-‟ is a (present) perfect marker, which is viewed as a non-deitic / relative tense. Despite of the differences in the perspective and the details of the proposals, I put them all under this line of analysis because they deny that „-ess-‟ can have a past tense meaning only. The perfect or perfective analysis of „-ess-‟ can explain why the examples in (46) and (47) express non-past situations, but not past ones. More specifically, the surface form of „-ess-‟ in (46) and (47) is to be analyzed as the perfect aspect and the null form of present tense: „-ess-‟ + Ø 34 . The deictic meaning that associates the sentences with S-time comes from the presence of the deictic present tense of Ø . However, Chung (2005: p.30) 33 The term „perfective‟ is sometime used inter-changeably with another term „perfect‟ in those analyses, as pointed out by S.-O. Sohn (1995). 34 For (46)-a, the null form Ø should be treated as a non-deictic / anaphoric tense in the temporal adjunct clause while for other examples with the present perfect meaning it has to be analyzed as a deictic present tense in the matrix tense. 60 questions the suspicious identity of „-ess-‟ as a perfective marker, on the basis of the cross linguistic generalization of Bibyee et al. (1994) such that perfective has the effect of signaling a present state with stative verbs when it can be used with them. But, Chung (2005) points out that „-ess-‟ does not express a present situation when used with a stative predicate, as follows: (48) a. Mina-ka phyenci-lul ss-ess-ta. Mina-NOM letter-ACC write-PER-DEC 'Mina has written / wrote a letter.' b. Mina-ka aph-ass-ta. Mina-NOM be.sick-PER-DEC 'Mina has been / was sick.' „-Ess-‟ in (48)-b following a stative predicate „aphwu- (be sick)‟ can refer to a past state, just like in (48)-a with a non-stative predicate, not conforming to the generalization. Chung (2005) presents the lack of difference in meaning between (48)-a and (48)-b as the evidence proving that the perfective analysis is not on the right track. The past tense meaning of „-ess-‟ can be shown clearly when it accompanies a past time denoting adverb as in (49): (49) a. Mina-ka sam il cene phyenci-lul ss-ess-ta. Mina-NOM three day before letter-Acc write-PER-DEC 61 'Mina wrote a letter 3 days ago.' b. Mina-ka sam il cene aph-ass-ta. Mina-NOM three day before sick-PER-DEC 'Mina was sick three days ago.' (49)-b as well as (49)-a locates the event of „writing a letter‟ or „being sick‟ in a past time specified by the past time adverb „sam-il cene (three days ago).‟ The meaning of a present situation is not involved with the alleged perfective marker „-ess-‟ and the stative predicate in (49)-b, which indicates that the perfective analysis of „-ess-‟ is not convincing. To account for the above examples including (49), Chung (2005) claims that „- ess-‟ is a (present) perfect. Chung (2005) treats the perfect „-ess-‟ as a relative tense which signals an anterior relation between an event time and the reference time 35 , and following and modifying Kratzer (1998), gives a denotation of perfect as below: (50) [[ANTERIO(RPE RFECT)]] = P. t. e [ (e) < t & P (e)] Presupposition: p' at t, where p' is a consequent state of the perfect proposition p 36 . The denotation in (50) tells us that perfect carries a presupposition such that a consequent 35 Chung (2005) uses the term „perfect‟ interchangeably with an „anterior‟ relation. 36 (e): the running time of an event (cf. Kripka 1989) 62 state of the perfect proposition holds at the reference time of the perfect 37 . One should notice that by this definition Chung (2005) claims that the existential reading (Iatridou et al. 2003) is more general one for the present perfect or the interpretation of a non-past time situation. She argues that the Korean perfect „-ess-‟ is unambiguously the existential perfect so that it can have the result state reading, the experiential reading and the recent past reading but does not have the continuous reading of the universal perfect 38 . According to Chung (2005), „-ess-‟ in all the examples from (46) through (49) has an existential reading and the meaning related to a present situation comes from the presupposition of „-ess-‟; it is presupposed that the consequent state of the proposition including „-ess-‟ should hold at S-time in the matrix clauses. But, as for the asserted meaning of the sentences with „-ess-,‟ S-time is not included in the reference time. The analysis can nicely explain how the present perfect „-ess-‟ is compatible with the past time interpretation, occurring with the past time denoting adverb in (50), which is not allowed in English type languages thereby creating the well known „present perfect puzzle‟. If the present perfect analysis of „-ess-‟ is correct, Korean should be analyzed to behave Italian or German with respect to the co-occurrence constraint with time adverbs 37 Chung (2005) assumes that the basic denotation of perfect is the same for all languages, but within the boundary of the basic meaning, the meaning of perfect varies in languages depending on the meaning of a present tense in a given language. 38 Via Chung (2005): Iatridou et al. (2003) claims that the English perfect in (i) is ambiguous between the universal reading in (ii-a) and the existential reading in (ii-b): (i) Since 1990 I have been sick. (ii) a. U-reading: There is a time span (the perfect time span) whose LB (left boundary) is in 1990 and whose RB (right boundary) is the utterance time, and all the points of that time span are points of my being sick => i (LB= 1990 & RB= Now & t i (Eventuality (t))) b. E-reading: There is a time span (the perfect time span) whose LB is in 1990 and whose RB is the utterance time, and in that time span is an eventuality of my being sick => i (LB= 1990 & RB= Now & t  i (Eventuality (t))) 63 because perfect can appear with past time denoting adverbs in those languages, as illustrated in the Korean example in (50). Chung (2005) relates the cross linguistic difference in the meaning of perfect to the denotation of the present tense in a given language. Chung (2005) claims that perfect does not have a universal meaning in a language that has a durative present tense; a universal reading is available for perfect in a language that has an instantaneous present tense 39 . Korean, Italian and German belong to the former group, an A-type language while English the latter group, a B-type one, as shown below: (51) Chung (2005) More specifically, according to Chung (2005), because Korean has an imperfective 39 Chung (2005) adopts the generalization of Giorgi and Pianesi (1997) such that the present tense can be durative or instantaneous in language. 64 present tense (I-Present) whose reference time includes S-time, the perfect „-ess-‟ lacks the universal reading so that the present perfect puzzle does not arise 40 . However, Chung‟s (2005) proposal is not without problems. First of all, if „-ess-‟ denotes an anterior relation as a perfect then it needs a deictic tense to follow it for a temporal anchoring so that it is expected to appear with the deictic present tense „-nun-‟ for verbal predicates and with the null form „- -‟ for adjectival predicates, as below: (52) a. Mina-ka phyenci-lul ss-ess-  / (*nun)-ta. Mina-NOM letter-ACC Write-PER-PRS-DEC 'Mina has written / wrote a letter.' b. Mina-ka aph-ass- -ta. Mina-NOM sick-PER-PRS-DEC In (52), the perfect „-ess-‟ does not allow the present tense marker „-nun-‟ to follow it even in the example with the verbal predicate „ssu-ta (write).‟ It might be due to a hidden phonological constraint working behind in (52), but conversely can suggest that „-ess-‟ is 40 Chung‟s (2005) analysis aims to cover the non-past as well as the past time interpretation of „-ess-,‟ avoiding resorting to the ambiguous analysis of „-ess-‟; „-ess-‟ is not ambiguous and the „another reading‟ is explained by the presupposition involved with „-ess-.‟ Despite of differences in the details of the proposal, Chung‟s (2005) analysis is, in spirit, resembles Pancheva & Stechow‟s (2004) analysis of the English present perfect positing that the perfect is a weaker version of past, including the past tense meaning within in. In other words, the English present perfect introduces an interval which can include S-time as in (51)-a while the English past tense introduces a past time interval that excludes S-time as in (51)-b 40 . In Pancheva & Stechow‟s (2004) term, the Korean perfect „-ess-‟ has a non-strengthened meaning like the German perfect, which can correspond to the English past tense as in (51)-b. Then, there is no need to posit that Korean has a separate past tense meaning of „-ess-‟ because the perfect meaning covers the past tense meaning. 65 a past tense marker so that it does not come with a present tense marker „-nun-‟ because otherwise it causes a contradicting meaning. Also, Chung (2005) claims that the double form „-ess-ess‟ is a true deictic past tense. It follows from it that the double past tense should be able to form a past perfect, combining with the perfect „-ess-,‟ i.e. „-ess-essess-.‟ And, it is predicted according to Chung‟s (2005) analysis that there would be a difference in meaning between a present perfect of „-ess +  -‟ and a past perfect „-ess-essess-‟ or between the past „-essess-‟ and the past perfect „-ess-essess.‟ But, the prediction is not born out. The tripe form of „-ess- essess,‟ even though possible, is considered as an emphatically repeated form of the double form „-essess-.‟ That there is found no difference in meaning between the past „- essess-‟ and the past perfect „-ess-essess-‟ supports the point. Finally, Chung‟s (2005) perfect analysis of „-ess-‟ is not able to account for the present situation interpretation of „-ess-‟ in the following examples: (53) a. Mina-nun apeci-lul manhi talm-ass-ta. Mina-TOP father-ACC much resemble-?-DEC. „Mina resembles his father. b. Ku pontu-ey pyenci-ka tul-ess-ta. That envelope-LOC letter-NOM be contained-?-DEC. „A letter is contained in that envelope.‟ 66 It is clear that the sentences in (53) 41 describe present situations, but not past ones. When the present tense marker replaces the past tense one, the sentences in (54) become unacceptable: (54) a. #Mina-nun apeci-lul manhi talm-nun-ta. Mina-TOP father-ACC much resemble-?-DEC. „Mina resembles his father. b. #Ku pontu-ey pyenci-ka tu(l)-n-ta. That envelope-LOC letter-NOM be contained-?-DEC. „A letter is contained in that envelope.‟ The present tense marker „-nun-‟ or „- -‟ cannot be used to express a present situation in (54), which supports that „-ess-‟ is the one conveying the present time interpretation in (53). Chung‟s (2005) account where the universal reading is obtained only through presupposition is not able to explain the obligatory present time interpretation in (53). Now let us turn to the lexical ambiguity analysis. One version of it, such as S.O. Sohn (1995), says that there are two kinds of „-ess-‟: a past tense „-ess-‟ and a perfect „- ess-‟; it is a coincidence for them to take the same form or there might be a historical reason for the coincidence. According to the analysis, the sentence (55)-a, is ambiguous between a past and a (present) perfect 42 : 41 The example from Nam (1978) in (34c) is a repeated in (37a). 42 In S.O. Sohn (1995), the two terms „a perfect marker‟ and „a perfective one‟ are used interchangeably. 67 (55) a. A past tense ‘-ess-’: Ecey Mina-ka chengbaci-lul ip-ess-ta. Yesterday Mina-NOM jeans-ACC wear-PST-DEC 'Yesterday, Mina put on jeans.' b. A present perfect ‘-ess-’: Mina-ka chengbaci-lul ip-ess- -ta. Mina-NOM jeans-ACC wear-PER-PRS-DEC 'Mina has put on jeans.' „-Ess-‟ in (55)-a is a past tense so that the past event of „wearing jeans‟ can be modified by a past time adverb „ecey‟ (yesterday). Whereas, „-ess-‟ in (55)-b is a perfect form, followed by a covert present tense , which is responsible for the perfect meaning that is associated with a present situation of „wearing jeans.‟ Even though the ambiguity account of a past tense and a present perfect seems to be able to cover most of the data with „-ess-,‟ we find the analysis unsatisfactory and are forced to suggest a slight different ambiguity account of „-ess-.‟ The heart of the objection lies in the suspicious grammatical status of „-ess-‟ as a perfect marker. Our claim that Korean does not have a perfect marker as a grammatical category and „-ess-‟ is a different species with a similar meaning is grounded on the following facts. First, the alleged perfect marker „-ess-‟ takes only a small portion of verbs for the prefect meaning. Most of achievement verbs or verbs with a meaning of change of states can make of a present perfect meaning with „-ess-‟ when there is no other past tense 68 marker: (56) a. „tochakha-ta‟ (arrive), „LA-e ka-ta‟ (go to LA), b. „nulk-ta‟ (get old), „kay-ci-ta‟ (get broken) c. „ip-ta‟ / „ssu-ta‟ (put on), On those verb cases, the ambiguity of a past tense marker and a perfect one becomes clearly. The narrow range of verb selection is unexpected as a perfect marker and it needs explanation of why a perfect meaning is heavily dependent on verb meanings. And, in other cases like activity verbs, a present perfect meaning is very vague and it seems that the researchers in the literature are content with having finding a meaning corresponding to the English present perfect: (57) a. Cikum-un (na-nun) chwungpwurnhi mek-ess-e. Now-TOP (I-TOP) enough eat-PFR-DEC 'Now I have eaten enough.' Chung (2005: 69) b. Samsip-pun ceney (na-nun) chwungpwurnhi mek-ess-e. Thirty-min ago (I-TOP) enough eat-PER-DEC '(Lit.) *I have eaten enough thirty minutes ago.' Chung (2005) treats the data in (57)-a to have a present perfect meaning such as the English gloss of „now I have eaten enough,‟ arguing that it is supported by the fact that 69 the time adverb „cikum‟ (now) can be used with the perfect marker „-ess-,‟ as in English. However, the sentence (57)-b with a past time adverb „samsip-pun ceney‟ (thirty minutes ago) does not differ from the overall meaning of (57)-a except for the adverb, actually entailing that of (57)-a, which is not available with the English present perfect. In other words, (57)-b with a past time adverb also has the so-called same present perfect meaning with (57)-b such that the speaker ate enough and has a full stomach now. According to S.O. Sohn‟s (1995) ambiguity account, we have to say that (57)-a has a present perfect while (57)-b has a past tense, failing to capture the same perfect meaning that bears a present time relevance. So, both of them turn out to be undesirable and unsatisfactory accounts on this matter. In this paper, it will be proposed that both (57)-a and (57)-b has a past tense marker only. The meaning in question that has present time relevance is to be accounted for by a pragmatic inference obtained by the past tense meaning in Korean. Because Korean does not have a past tense versus present perfect apposition, it becomes possible so that a past tense is not blocked from having present time relevance unlike English. It will be discussed in details shortly. Secondly, if one instance of „-ess-‟ is a perfect form, it is expected to be able to appear with a progressive form, maintaining the alleged ambiguous readings of (58)-a below, but it turns out that the perfect meaning disappears when a progressive form inserted as in (58)-b: 70 (58) a. Mina-nun cemshim-ul mek-ess-ta. Mina-TOP lunch-ACC eat-??-DEC Reading #1(a past tense „-ess-‟): 'Mina had a lunch.' Reading #2(a perfect „-ess-‟): „Mina has had a lunch.‟ b. Mina-nun cemshim-ul mek-koiss-ess-ta. Mina-TOP lunch-ACC eat-PRG-??-DEC Reading #1'Mina was having a lunch.' Reading #2: „Mina has been having a lunch.‟ Note that Reading #1, a past progressive meaning, is available for (58)-b where a progressive form „-koiss-‟ is added to, while Reading #2, a present progressive + perfect meaning disappears. The unavailability of Reading #2 in for (58)-b is hard to explain if we assume that (58) has a perfect marker „-ess-‟ with Reading #2 43 . Third, time adverbials or other time phrases such as „(time)-cenpute‟ (since), „(time)-tongan‟ (for) or „cikum-kkaci‟ (until now) are often regarded as time phrases that indicate that „-ess-‟ co-occurring with them is a perfect marker, as shown in (59): (59) a. ?Mina-nun seshikan-cenpute cemshim-ul mek-ess-ta. Mina-TOP three hours-before-since lunch-ACC eat-??-DEC 'Mina has been eating a lunch since three hours before.' 43 We are not going to see other examples such as „ip-ta‟ (wear) or „tul-ta‟ (be contained). As for „ip-ta,‟ „-koiss-ta‟ is notoriously ambiguous between a progressive form and a result state marker, which makes things complicated and confusing. And, as for „talm-ta‟ (resemble) or „tul-ta‟ (be contained), a progressive form is not allowed to be attached to those predicates. Hence, those examples are not discussed. 71 b. ?Mina-nun cikum-kkaci emshim-ul mek-ess-ta. Mina-TOP now-until unch-ACC eat-??-DEC 'Mina has been eating until now.' c. Mina-nun sechikan-tongan cemshim-ul mek-ess-ta. Mina-TOP three hours-for lunch-ACC eat-??-DEC 'Mina has been eating for 3 hours.' At a first glance, the fact that the time phrases such as „seshikan-cenpute‟, „cikum-kkaci‟ or „sechikan-tongan‟ in (59) that have a time interval connected to the speech time can co-occur with „-ess-‟ seems to demonstrate that „-ess- is not a past tense but a present perfect.‟ However, those time phrases are not guaranteed to reach to the speech time when they are used with „-ess-,‟ as in (60): (60) a. Mina-nun seshikan-cen-pute cemshim-ul mek-ess-nunte, Mina-TOP three hours-before-since lunch-ACC eat-??-and/but. pangkum-ceney ta mek-ko ka-ss-ta. just now -before all eat-and go-PST-DEC 'Mina started eating a lunch since three hours ago, and she finished it and left just now.' b. Mina-nun cikum-kkaci (seshikan-tongan) cemshim-ul mek-ess-nunte. Mina-TOP now-until lunch-ACC eat-??-and/but. pangkum-ceney ta mek-ko ka-ss-ta. 72 just now -before all eat-and go-PST-DEC (Lit.) 'Mina had been eating a lunch (until now), but she finished it and left just now. c. Mina-nun seshikan-tongan cemshim-ul mek-ess-nunte. Mina-TOP three hours-for lunch-ACC eat-??-and/but. pangkum-ceney ta mek-ko ka-ss-ta. just now -before all eat-and go-PST-DEC 'Mina ate a lunch for 3 hours, and she finished it and left just now.' As the second conjuncts of (60)-a through (60)-c show, the events described by the first conjuncts with „-ess-‟ can be a past event that is disconnected from the speech time. It means that time phrases given above do not act like English ones and cannot serve as a tester that is closely tied to a perfect meaning. What is interesting is that in Korean a perfect reading is obtained when a present progressive marker accompanies those time phrases: (61) a. Mina-nun seshikan-cen-pute cemshim-ul mek-koiss- -ta. Mina-TOP three hours-before-since lunch-ACC eat-PRG-PRS-DEC (Lit.) „#Mina is eating a lunch since 3 hours before.‟ (Int.) 'Mina has been eating a lunch since three hours before.' b. Mina-nun cikum-kkaci cemshim-ul mek-koiss- -ta. Mina-TOP now-until lunch-ACC eat-PRG-PRS-DEC 73 (Lit.) „#Mina is eating a lunch until now.‟ (Int.) 'Mina has been eating until now.' c. Mina-nun sechikan-tongan cemshim-ul mek-koiss- -ta. Mina-TOP three hours-for lunch-ACC eat-PRG-PRS-DEC (Lit.) „#Mina is eating a lunch for 3 hours.‟ (Int.) 'Mina has been eating for 3 hours.' In (61), the event of Mina‟s eating a lunch should be continued from the time point marked by the time phrases to the speech time, resulting in a present perfect meaning. The perfect meaning of the sentences in (61) cannot be nullified by a second conjunct describing a past event as in (60). See (62): (62) a. #Mina-nun seshikan-cen-pute cemshim-ul mek-koiss- -ciman, Mina-TOP three hours-before-since lunch-ACC eat-PRG-PRS-but pangkum-ceney ta mek-ko ka-ss-ta. just now -before all eat-and go-PST-DEC (Lit.) „#Mina has been eating a lunch since three hours before, and she finished it and left just now ' b. #Mina-nun cikum-kkaci cemshim-ul mek-koiss- -ciman. Mina-TOP now-until lunch-ACC eat-PRG-PRS-but pangkum-ceney ta mek-ko ka-ss-ta. just now -before all eat-and go-PST-DEC 74 (Lit.) '#Mina has been eating until now, and she finished it and left just now.' c. #Mina-nun sechikan-tongan cemshim-ul mek-koiss- -ciman. Mina-TOP three hours-for lunch-ACC eat-PRG-PRS-but pangkum-ceney ta mek-ko ka-ss-ta. just now -before all eat-and go-PST-DEC (Lit.) '#Mina has been eating for 3 hours, and she finished it and left just now.' In (62), when the second conjunct clearly describes a past event, the whole sentence becomes uninterpretable or yields a temporal contradiction because the first conjunct means that Mina‟s eating event is still ongoing at the speech time, but the second one coveys the meaning that the event was already completed. Hence, we have to say that Korean does not express a perfect meaning by a perfect marker, the alleged on „-ess-,‟ but by a different means. The discussion so far leads us conclude that the ambiguity account for „-ess-‟ is appealing option to take, but that the given data show that another instance of „-ess-‟ (besides a past tense one) cannot be analyzed as a perfect marker for several reasons. Thus, it is proposed in this paper that the other instance of „-ess-‟ in question is not a perfect marker but an auxiliary verb construction of „-(e/a) iss-‟ with a resultative meaning. Consider the meaning difference between the auxiliary verb construction and a present tense: 75 (63) a. Mina-nun uica-ey anc-a iss- -ta. Min-TOP chair-LOC sit-BE-PRS-DEC „Mina is sitting on a chair.‟ b. Mina-nun uica-ey anc-nun-ta. Min-TOP chair-LOC sit-PRS-DEC (Lit.)„Mina is in the course of sitting.‟ (63)-a with an auxiliary verb construction of „-(e/a) iss-ta‟ means that the state of Mina‟s sitting on a chair continues up to the speech time. The auxiliary verb„-(e/a) iss-ta‟ signals a resulative state of the event described by a verb. Whereas, (63)-b with a present tense „- nun‟ only relates the given event to the speech time so that it means that Mina is doing the action of sitting on a chair and the action is incomplete at the speech time. The difference becomes clearer when a durational time phrase such as „for three hours‟ is added to them: (64) a. Mina-nun uica-ey seshikan-tongan anc-a iss- -ta. Min-TOP chair-LOC three hours-for sit-BE-PRS-DEC „Mina has been sitting on a chair for 3 hours.‟ b. #Mina-nun uica-ey seshikan-tongan anc-nun-ta. Min-TOP chair-LOC three hours-for sit-PRS-DEC (Lit.)„Mina is in the course of sitting for 3 hours.‟ 76 While the state of Mina‟s sitting on a chair in (64)-a does not cause a problem with the durational time phrase „seshikang-tongan‟ (for three hours), the action of Mina‟s sitting on a chair in (64)-b brings a problem to the interpretation of such a time phrase. The idea for another version of an ambiguity analysis of „-ess-‟ to be put forth in this paper is that „-ess-‟ is a shortened or grammaticalized form of the resultative auxiliary verb construction of „-(e/a) iss-ta.‟ In other words, „-ess-‟ that was claimed to be a perfect form in S. O. Sohn (1995), Chung (2005), and many others, is actually an auxiliary verb construction with a similar meaning, that is, a resultative meaning that continues to the speech time. Given this, many issues that arose with a perfect analysis of „-ess-‟ are accounted for by the current analysis. See the following example: (65) Mina-nun uica-ey anc-ass-ta. Min-TOP chair-LOC sit-BE.PRS-DEC Reading #1: „(Int.) Mina sat on a chair and still is sitting on it.‟ Reading #2: „Mina sat on a chair.‟ The difference between the two readings of (65) can be presented as follows: (66) a. Mina-nun uica-ey anc-ass- -ta. Min-TOP chair-LOC sit-BE-PRS-DEC „(STATE): Mina is sitting on a chair.‟ b. Mina-nun uica-ey anc-a iss- -ta. Min-TOP chair-LOC sit-BE-PRS-DEC 77 „(STATE): Mina is sitting on a chair.‟ c. Mina-nun uica-ey anc- -ass-ta. Min-TOP chair-LOC sit-PRS-PST-DEC „(Action): Mina sat on a chair.‟ As in (66)-a and (66)-b, the first resultative reading with a strong present time relevance is to be associated with the resultative auxiliary verb construction of „-a iss-ta,‟ which means it is claimed that (66)-a and (66)-b end up with having the same LF; there is no past tense involved in (66)-a and (66)-b that makes a possible for the resultative state of „sitting on a chair‟ continues to the speech time. Whereas, in (66)-c, „-ess-‟ is a simple past tense that relates the given event to a certain past time. This analysis can explain why the other „-ess-‟ (not a past tense) should show a narrow range of predicate selection. The auxiliary verb construction is restricted to a certain predicate type. It cannot be with adjectival predicates and all the transitive verbs. Among intransitive verbs, it only allows verbs of a change of state including achievement verbs. See below: (67) Adjectival predicates (in formal speech) a. Individual level predicates: *John-un ttokttokha-y iss- -ta. John-TOP smart-BE-PRS-DEC „(Lit.) John is in the state of being smart.‟ 78 b. Stage-level predicates: *John-un apu-a iss- -ta. John-TOP sick-BE-PRS-DEC „(Lit.) John is in the state of being sick.‟ (68) Verbal predicates (in formal speech) a. Activity verbs: *John-un chwum-ul chwu-e iss- -ta. John-TOP dance-ACC dance-BE-PRS-DEC „(Lit.) John is in the state of having danced.‟ b. Achievement verbs: John-un cuk-e iss- -ta. John-TOP die- PRS-DEC „(Lit.) John has died. (John is in the state of having died)‟ c. Accomplishment verbs: *John-un cip-ul cis-e iss- -ta. John-TOP house-ACC build-BE-PRS-DEC „(Lit.) John is in the state of having built a house)‟ d. Stative verbs: *John-un LA-ey sa(l)-e iss-  -ta. 44 44 The auxiliary verb construction of „sal-a iss-ta‟ is possible for another meaning, „being alive.‟ But, it is not relevant to the current discussion so that will not be considered. 79 John-TOP LA-LOC live-BE-PRS-DEC „(Lit.) John is in the state of having lived in LA.‟ So, the highly restricted distributional fact of the auxiliary verb construction explains why the alleged perfect form of „-ess-‟ is so selective in choosing verbs. Also, the auxiliary verb construction „-(e/a) iss-ta‟ is not compatible with a progressive meaning because of the resultative state meaning: (69) ?Mina-nun uica-ey anc-a iss-ko iss- -ta. Min-TOP chair-LOC sit-BE-PRG-PRS-DEC „(STATE): Mina is sitting on a chair.‟ The unavailability of (69) with a progressive form and a resultative auxialiary verb accounts for why in a perfect meaning of „-ess-‟ becomes unavailable when a progressive form is added in (68) before. Finally the double form „-ess-ess-‟ is treated nicely as a sequence of the auxiliary verb construction and a paste tense. (70) Mina-nun uica-ey anc-a iss-ess-ta. Min-TOP chair-LOC sit-BE-PST-DEC „(STATE): Mina had been sitting on a chair. / Mina was sitting on a chair‟ 80 The discussion so far shows that the ambiguity analysis of „-ess-‟ as a past tense and an auxiliary verb construction is appealing and satisfactory. 2.1.4. A double form of ‘-ess-ess-’ The double form of „-ess-ess-‟ has received much attention and various proposals have been presented in the literature 45 . The crucial difference between „-ess-‟ and „-ess-ess-‟ is found with respect to the co-occurrence fact with temporal adverbs; while „-ess-‟ occurs with non-past time adverbs such as „cikum (now)‟ or „nayil (tomorrow),‟ „-ess-ess-‟ cannot (Chung 2005): (71) a. Chulswu-nun cikum ttuna-ess-ta. Chulswu-TOP now leave-PST-DEC „Chulswu left just now.‟ b. *Chulswu-nun cikum ttuna-ss-ess-ta. Chulswu-TOP now leave-?-PST-DEC „*Chulswu left now.‟ As shown in the contrast between (71)-a and (71)-b, „-ess-ess-‟ cannot be used for a situation that has current relevance indicated by the time adverb „cikum (now) 46 .‟ Also, „- 45 See Chung (2005) for the detailed discussion. 46 However, the co-occurrence restriction may not be rigid criteria differentiating the meaning of „-ess-ess-‟ 81 ess-ess-‟ has a meaning of discontinuity (C. Lee 1985) or no-current relevance (Chung 2005), as illustrated in the following examples: (72) a. Chulswu-nun LA-e sal-ess-ta. Chulswu-TOP LA-at live-PST-DEC „Chulswu lived in LA.‟ b. Chulswu-nun LA-e sal-ess ess-ta. Chulswu-TOP LA-at live-?-DEC „Chulswu had lived in LA.‟ In contrast to (72)-a, (72)-b with „-ess-ess-‟ means that Chulswu had lived in LA before, but is not living there anymore. So the event described by the verb „sal-ta (live)‟ has to be ceased at the time of speech, which is often couched in the term „discontinuity‟ or „no from „-ess-‟ because, though restricted, there is an example where „cikum (now)‟ can appear with „-ess-ess-‟: [Context: a dance teacher is watching a video of his student‟s performance. Pointing out one scene that his student performed excellently, he said the following sentence] (i) Cikum coh-ass ess-e. Now good-?-DEC (Int.)„Now was good ‟ In the sentence above, „cikum (now)‟ can be used with the double form of „-ess-ess-.‟ Yet, it might be because the video watching situation is an unusual situation where the speaker‟s temporal reference can be separate from the temporal reference of the event described by the verb. The past event that took place before is being repeated „now,‟ at the speech time so that the given event is associated with two distinct time points, past and present. Given this, the time adverb „cikum (now)‟ is allowed in relation to the ongoing present event in the video even though the actual dancing event was performed before for which the past tense form of „-essess-‟ is used. If this is right, the example given above would not be considered as a counter example suggesting that „-ess-ess‟ can co-occur with „cikum (now)‟ 82 current relevance 47 .‟ The naï ve analysis positing that the double form of „-ess-ess-‟ is a mere duplicate of a deictic past tense faces a problem in explaining how the inner „-ess-‟ is evaluated with respect to S-time and what contribution it has for a temporal interpretation. Such an issue does not arise with the pluperfect analysis where „-ess-ess-‟ is analyzed as a past perfect form of „aspect + tense.‟ However, the past perfect analysis of „-ess-ess-‟ makes a wrong prediction about the interpretation of „-ess-ess-‟ with a time adverb. (73) a. Chulswu-nun ttuna-ss-ess-ta. Chulswu-TOP leave-?-PST-DEC „Chulswu left.‟ b. Chulswu-nun ?seshi-e ttuna-ss-ess-ta. Chulswu-TOP 3 at leave-?-PST-DEC „Chulswu left at 3 o‟clock.‟ 47 Things go a bit differently with achievement verbs. When an achievement verb such as „cuk-ta (die)‟ is followed by „-ess-ess-,‟not the event itself but the result (state) of the event become the object of discontinuation. (i) a. Chulswu-nun ne shi-e cuk-ess-ta. Chulswu-TOP 4 o‟clock-at die-PST-DEC „Chulswu died at 4 o‟clock.‟ b. Chulswu-nun ne shi-e cuk-essess-ta. Chulswu-TOP 4 o‟clock-at die-PST-DEC „Chulswu died at 4 o‟clock, but was brought back to life again.‟ In contrast to (i-a) with „-ess-,‟ (i-b) including „-ess-ess-‟ has an interpretation such that Chulswu died at the past time point, but he was (somehow) brought back to life and is alive now. In (i-b), the result of Chuswu‟s death does not continue to the speech time, but ceased at some past time point. Thus, it has been often pointed out in the literature that „-ess-ess-‟ has a peculiar meaning of a „discontinued event.‟ 83 If „-ess-ess-‟ is a „perfect + past tense,‟ it is predicted that (73)-b is ambiguous depending on which level the adverbial modification is involved with, namely a VP or AspectP. But, the prediction is not born out because (73)-b is not ambiguous; the time adverb „seshi-e (at 3 o‟clock) modifies the VP, the event time only. The past perfect analysis cannot account for the absence of another reading. Note that when an adverb „imi (already)‟ is inserted, as pointed out in Chung (2005), the pluperfect meaning becomes available: (74) ?Chulswu-nun se shi-e imi ttuna-ss-ess-ta. Chulswu-TOP 3 o‟clock-at already leave-?-PST-DEC „Chulswu had already left at 3 o‟clock.‟ The presence of „imi (already)‟ renders the pluperfect meaning for (74). However, it does not provide a support for the pluperfect analysis of „-ess-ess-‟ because it seems to right to say that the pluperfect meaning depends on the meaning of the aspectual adverb „imi (already).‟ As in (73)-b, without the presence of „imi (already),‟ only the reading where „seshi-e (at 3 o‟clock) modifies the event time of (74) is available; moreover, when „imi (already)‟ is present, the simple past form of „-ess-‟ can have the pluperfect meaning of (74), as in (75) below: (75) a. Chulswu-nun seshi-e imi ttuna-ss-ta. Chulswu-TOP 3 at already leave-PST-DEC 84 „Chulswu had already left at 3 o‟clock.‟ b. Mina-ka wo-ass-ul ttay Chulswu-nun imi ttuna-ss-ta. Mina-NOM come-PST-REL when Chulswu-TOP already leave-PST-DEC „When Mina came in, Chulswu had already left.‟ As in the examples in (75), even the past tense marker „-ess-‟ can convey the past perfect meaning with the aspectual adverb „imi (already), ‟ which means that the past perfect meaning of (74) comes not from „-ess-ess-‟ but from „imi (already).‟ Chung (2005) solves this problem by proposing that „-ess-ess-‟ is as a deictic past tense and „-ess-‟ is a non-deictic anterior past tense. Because „-ess-ess-‟ is analyzed as a (morphologically simple) deictic past tense and there is no double layer of tense or aspect in Chung (2005), the absence of ambiguity is expected. However, Chung‟s (2005) proposal makes a wrong prediction about the temporal meaning of „-ess-ess-‟ in a subordinate clause, as below: (76) a. Kim-un [ Chulswu-ka ttuna-ssess-ta]-ko malhay-ss-ta. Kim-TOP Chulswu-NOM leave-PST-DEC-COMP say-PST-DEC „Kim said that Chulswu had left.‟ b. Kim-un [ Chulswu-ka ttuna-ss-ta]-ko malhay-ss-ta. Kim-TOP Chulswu-NOM leave-PST-DEC-COMP say-PST-DEC „Kim said that Chulswu had left.‟ 85 If „-ess-ess-‟ is a simple deictic tense and „-ess-‟ is an anterior past tense, it is expected that there should be a difference between the temporal meanings of the complement clauses in (76)-a and (76)-b. More specifically, if the past tense of the complement clause in (76)-a is to be evaluated with respect to S-time, as a deictic past tense, the embedded event can either overlap with or follow the matrix event because both past tenses are deictic; any order of two past events should be possible. However, (76)-a has the only one reading where the embedded past tense precedes the matrix one, just like the anterior reading of „-ess-‟ in (76)-b. In the proposed analysis, following Chung (2005), it is argued that „-ess-ess-‟ is not a combined form of a perfect and a past tense or any other kinds; it is a morphologically simple form of a past tense. Yet, going against Chung‟s (2005) proposal that „-ess-ess-‟ is a deictic tense, I claim that it is another anterior past tense with a pragmatic implicature with respect to the current relevance mentioned above. The proposal naturally accounts for the absence of a meaning difference between (76)-a and (76)-b, and opens a way to explaining the redundant tripling case of „-ess-ess-ess‟ or more. 2.2. Subordinate Tenses We started to deal with tenses in matrix clauses and now will go into the more complicated temporal interpretation of tense markers in embedded clauses in this chapter. Three types of clauses will be on the discussion: complement, relative, and adverbial clauses (other subordinate clauses). 86 2.2.1. The preliminary issue: Complementation or Quotation? Let us consider an example of an embedded tense in Korean: (77) John-nun [ caki-ka LA-ey iss- -ta]-ko malha-(e)ss-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC. „John 1 said that he 1 was in LA (at that time).‟ (77) receives a simultaneous reading where the embedded event temporally overlaps with the matrix event. In (77), „-ko-‟ attached to the ender of the subordinate clause analyzed as a clausal complementizer taking the complement clause. However, let us point out that it is not an all agreed-view in the traditional Korean linguistic literature, even though it is commonly assumed by many Korean generative grammarians. S.O. Sohn (1995) and H.M. Sohn (1999) stand for the other view, proposing that the subordinated clause marked by a bracket in (77) is an „indirect quotative construction.‟ It seems that the main reason for them to take the classification is that they wanted to distinguish it from verbal complement clauses that are to be treated as „true‟ complement clauses 48 according to their analyses. One example of a verbal complement clause is given below: 48 Korean has various kinds of verbal complementizer including the infinitive „-e/a,‟ the gerundive „-ko,‟ the suspective „-ci-‟ and so on. Considering all auxiliary verb or serial verb constructions can be treated as including verbal complement clauses (H.M. Sohn 1999), a verbal complement construction is very productive in Korean. 87 (78) John-un Mary-lul [ pro ecey hakkyo-ey ka- -key ] hay-ss-ta. John-TOP Mary-ACC yesterday school-LOC go-PRS-COMP cause-PST-DEC „John made Mary go to school yesterday.‟ (S.O. Sohn 1995:152) In (78), the so-called empathic „-key‟ (H.M. Sohn 1999) is a verbal complementizer, consisting of a causative construction along with the matrix verb „ha-ta (do).‟ The main characteristics of a verbal complement clause are that its subject is to be realized as a pro form and the embedded and matrix verbs show a strong cohesion, not allowing adverbials to intervene; the important property of a verbal complement construction with respect to temporal interpretation is that most of a verbal complementizer, except for a few cases 49 , do not allow a past tense maker to appear before it, as shown below: (79) a. ‘e/a-po-ta’ (try to do…) John-un [ pro hakkyo-ey ka-  / *ss-a ] po-ass-ta. John-TOP school-LOC go-PRS/PST-COMP try-PST-DEC „John went to school (to find out….).‟ b. ‘ko-sihp-ta’ (want to do…) John-un [ pro hakkyo-ey ka-  / *ss-ko ] siph-ess-ta. John-TOP school-LOC go-PRS/PST-COMP want-PST-DEC „John wanted to go school.‟ c. ‘a-iss-ta’ (be in the state of…) 49 For example, a past tense marker „-e/ass-‟can appear before a verbal complementizer „-eya ha‟ (should/must). 88 John-un [ pro hakkyo-ey ka--  / *ss -a ] iss-ess-ta. John-TOP school-LOC go-PRS/PST-COMP be-PST-DEC „John was at school.‟ d. ‘ci-anh-ta’ (not...) John-un [ pro hakkyo-ey ka--  / *ss -ci ] anh-ess-ta. John-TOP school-LOC go-PRS/PST-COMP not-PST-DEC „John didn‟t go to school.‟ (S.O. Sohn 1995:152) 50 The examples in (79) show that a paste tense marker cannot precede a verbal complementizer, which leads S.O. Sohn (1995) to give the generalization that no overt tense form is allowed in verbal complement clauses. But, the fact makes us go further and suspect that there is no tense form, whether overt of covert one, in verbal complement clauses in (79). If we assume that verbal complement clauses are literally tenseless in Ogihara‟s (1996) sense, many things are explained easily. First, if the embedded verb and matrix verb make one VP or vP, it is expected that a VP external adverb such as „ecey‟ (yesterday) is not to come between the two verbs. Secondly, the pro subject would be the best strategy because there is no TP structure which can host an overt subject NP. Finally, the mysterious adverbial fact that S.O. Sohn (1995) presented surprises us no more: 50 There are some of examples that are taken from S.O. Sohn (1995: 152). English glosses some descriptions were added to them for clarification. 89 (80) a. John-un Mary-lul [ pro ecey hakkyo-ey John-TOP Mary-ACC yesterday school-LOC ka- -key ] hay-ss-ta go-PRS-COMP cause-PST-DEC. „John made Mary go to school yesterday.‟ b. John-un Mary-lul [ pro nayil hakkyo-ey John-TOP Mary-ACC tomorrow school-LOC ka- -key ] hay-ss-ta go-PRS-COMP cause-PST-DEC. „(Lit.) *John made Mary go to school tomorrow.‟ S.O. Sohn (1995) presented the data in (80)-b in contrast to (80)-a, arguing against Kwon‟s (1985) analysis that tenses in verbal complements clauses are always anaphoric to matrix tenses. For the fact that a future time denoting adverb „nayil‟ (tomorrow) can appear in verbal complement clauses in (80)-b, she suggests that verbal complement clauses should have anaphoric and indexical tenses. But, if verbal complement clauses are infinitive ones lacking a tense marker as we suppose, „nayil‟ (tomorrow) will not be prevented from occurring in verbal complement clauses unless there are other semantic restrictions associated with verbal complementizers or matrix verbs. After all, verbal complement clauses are in many ways different from clausal complement clauses headed by „-ko,‟ as S.O. Sohn (1995) and H.M. Sohn (1999) assume. But, it cannot be a right reason to set aside for the „-ko‟ clause a new category, „a 90 quotative construction.‟ The traditional term „direct / indirect quotation marker‟ brings an unnecessary confusion to discussion, giving an illusion that the indirect quotation marker „-ko‟ is somehow related to a quotation. The confusion doubles because the interpretation of direct quotation as well as indirect quotation is available with „-ko, ‟ as mentioned in Chapter 1 before. Thus, we want to clarify that „-ko‟ accompanying a full clause can be associated with „direct / indirect speech,‟ using a pragmatic term, but not with „direct / indirect quotation.‟ And, our assumption that „-ko‟ in question is a clausal complementizer is justified when we consider the following basic syntactic facts about „- ko.‟ First of all, the „-ko‟ complement clause has an overt subject and overt present and past tense markers as follows: (81) a. Overt Subject: John-nun [ caki-ka LA-ey iss- /-ta]-ko malha-(e)ss-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC. „John 1 said that he 1 was in LA (at that time).‟ b. Overt present tense: John-nun [ caki-ka LA-ey sa(l)-n/-ta]-ko malha-(e)ss-ta. John-TOP self-NOM LA-LOC live-PRS-DEC-COMP say-PST-DEC. „John 1 said that he 1 lived in LA (at that time).‟ c. Overt past tense: John-nun [ caki-ka LA-ey sal-ass/-ta]-ko malha-(e)ss-ta. 91 John-TOP self-NOM LA-LOC live-PST-DEC-COMP say-PST-DEC. „John 1 said that he 1 had lived in LA.‟ The examples in (81) show that the clause followed by „-ko‟ is a fully tense clause with an overt subject, as are other clausal complements. Secondly, the fact that a long distance anaphora „caki‟ (self) in (81) does take the matrix subject as its antecedent, setting up an anaphoric relation across a clause, demonstrate that the „-ko‟ clause is not part of quotation, whether indirect or direct, but a syntactically embedded clause. Third, the „-ko‟ clause is not only embedded by saying verbs such as „malha-ta‟ (say) but also by other kinds of verbs, indicating that the term „quotative construction‟ has a limitation to account for it and is rather misleading. See the examples below: (82) a. saying verbs: malha-ta(say), cwucanha-ta (argue), kopaykha-ta (confess), senenha-ta (declare), palpyoha/kongpyoha-ta (announce), sorichi-ta(shout)… John-nun [ caki-ka LA-ey iss-  / ta]-ko malha-(e)ss-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC. „John 1 said that he 1 was in LA (at that time).‟ b. believing verbs: mit-ta(believe), sayngkakha-ta(think), chuceukha-ta(guess), kacengha-ta(suppose)… John-nun [ caki-ka LA-ey iss-  / -ta]-ko mit-(e)ss-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP believe-PST-DEC. 92 „John 1 believed that he 1 was in LA (at that time).‟ c. knowing verbs: al-ta(know), kiekha-ta(remember), ihayha-ta(understand)…. John-nun [ caki-ka LA-ey iss-  / -ta]-ko alko-iss- -ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PROG-PRS-DEC. „John 1 thinks/knows that he 1 is in LA (at that time).‟ 51 As in (82)-a and (82)-b, saying verbs and believing verbs can take a „-ko‟ clause as a complement. Also, it can serve as a complement of knowing verbs in (82)-c with some restriction and meaning alteration. It is obvious that the verbs in (82)-b and (82)-c) are not involved with quotation at all and still take a „-ko‟ clause as its complement. Hence, it turns out that the term „indirect quotation marker‟ is inappropriate and misleading. In short, the issue of the term „Direct / Indirect quotation marker‟ for „-ko‟ has been discussed and there are presented the reasons that „-ko‟ is to be analyzed as a clausal complementizer. 2.2.2. Types of temporal interpretation In conjunctive clauses, the tenses of both conjuncts are interpreted indexically. No anaphoric interpretation is available for each conjunct. Yet, for some subordinate clauses such as relative and adverbial one, both anaphoric and indexical tense readings are 51 When these knowing verbs come with a progressive form, taking a „-ko‟ complement clause, it seems that they do not behave factive verbs, but can be understood as belonging to believing verbs. A factive meaning of those verbs requires a nominalized complement clause headed by „kes.‟ 93 available. And, only an anaphoric tense interpretation is allowed in a complement clause. We will examine examples of each embedding type in details and show that the pattern presented above is correct. 2.2.2.1. Conjunctive clauses In a strict sense, tenses in conjunct clauses do not interact, being interpreted independently. It is assumed that there is no temporal ordering between the first and the second conjunct so that switching the order of them does not affect the meaning of the whole sentence, as given below: (83) a. [ John-un cikum-to ai-ka eps- ]-ko, John-TOP now-even child-NOM not have-PRS-CONJ Mary-un caknyen-e tulccay-ai-lul nah-ass-ta. Mary-TOP last year-LOC second-child-ACC give birth to-PST-DEC „John doesn‟t have a child even now, and Mary gave birth to her second child last year.‟ b. [ Mary-un caknyen-e tulccay-ai-lul nah-ass]-ko Mary-TOP last year-LOC second-child-ACC give birth to-PST-CONJ John-un cikum-to ai-ka eps- -ta, John-TOP now-even child-NOM not have-PRS-DEC „John doesn‟t have a child even now, and Mary gave birth to her second child last year.‟ 94 As in (83)-a, the first coordinated conjunct is a non-past tense sentence while the second one is a past tense sentence. The tense meaning of each conjunct does not change when their order is reversed as in (83)-b and the meaning of the whole sentence remains the same. More specifically, the first conjunct of (83)-a with an embedded non-past tense is interpreted indexically, meaning that John does not have a child at the speech time. It cannot be anaphoric to the second conjunct clause because the adverb „cikum-to‟ (even now) in the first conjunct rules out the anaphoric interpretation where the event of the first conjunct overlaps with that of the second conjunct that happened in last year, as indicated by the second conjunct adverb „caknyen-e‟ (in last year). Hence, it becomes clear that a non-past tense receives only an indexical interpretation and that tenses of the two conjuncts are interpreted independently of each other. One might tempt to think that the non-past tense of the first conjunct is anaphoric to that of the second conjunct, as in S.O.Sohn (1995), based on the fact that the first conjunct of (84)-a can have a past tense interpretation in spite of a non-past tense form: (84) a. [ John-un hakkyo-ey ka- ]-ko, John-TOP school-LOC go-PRS-CONJ Mary-un yuckiwon-ey ka-ss-ta. Mary-TOP kindergarten-LOC go-PST-DEC „John went to school, and Mary went to kindergarten.‟ b. [ John-un hakkyo-ey ka-ss]-ko, John-TOP school-LOC go-PST-CONJ 95 Mary-un yuckiwon-ey ka-ss-ta. Mary-TOP kindergarten-LOC go-PST-DEC „John went to school, and Mary went to kindergarten.‟ As in the English gloss of (84)-a tells, the first conjunct has a past tense reading. However, the null tense form  of the first conjunct is not a non-past tense, but it is a trace left after a PF tense deletion. The past tense of the first conjunct in (84)-b is deleted at PF and appears just as (84)-a); (84)-a and (84)-b have the exactly same meaning. It is supported by the evidence that a past time denoting adverb „ecey‟ (yesterday) can co- occur with the null form as in (85)-a and moreover the each conjunct of can be modified by adverbs that denotes different past times: (85) a. [ John-un ecey hakkyo-ey ka- ]-ko, John-TOP yesterday school-LOC go-PRS-CONJ Mary-un onul achim-e yuckiwon-ey ka-ss-ta. Mary-TOP today morning-at kindergarten-LOC go-PST-DEC „John went to school yesterday, and Mary went to kindergarten in the morning today.‟ b. [ John-un ecey hakkyo-ey ka-ss]-ko, John-TOP yesterday school-LOC go-PST-CONJ Mary-un onul achim-e yuckiwon-ey ka-ss-ta. Mary-TOP today morning-at kindergarten-LOC go-PST-DEC 96 „John went to school yesterday, and Mary went to kindergarten in the morning today.‟ When conjoined by „-ko‟ the two conjunct do not need to be associated with the same past time so that the event of the first conjunct occurred „yesterday‟ and the that of the second one happened „in the morning today,‟ as (85)-a shows. If the tense of the first conjunct is anaphoric to the second one, such an interpretation cannot be obtained so that different time adverbs are not allowed to modify each conjunct as in (85)-a. The same story goes for (85)-b, indicating that the temporal interpretation of the first conjunct is independent of that of the second one. Especially, (85)-b shows that a past tense is also interpreted indexically in conjunctive constructions, without bearing any temporal relation to the second conjunct. In other words, the past event of the first conjunct needs not be anterior to the second one. Thus, the example (78) below where the adverb „onul achim-e‟ (this morning) modifies the first conjunct, indicating that the first conjunct event follows that of the second one, is just fine: (86) [ John-un onul achim-e hakkyo-ey ka-ss]-ko, John-TOP today morning-at school-LOC go-PST-CONJ Mary-un ecey yuckiwon-ey ka-ss-ta. Mary-TOP yesterday kindergarten-LOC go-PST-DEC „John went to school yesterday, and Mary went to kindergarten in the morning today.‟ 97 Hence, the first conjunct of (86) does not have a past shifted reading, but both of conjuncts have an indexical reading where they are evaluated with respect to the speech time independently. Let us point out that we treat clause enders with a temporal ordering as subordinate clause markers, but not as conjunctive particle as in S.O. Sohn (1995) and H.M. Sohn (1999), even though some of them are very close to conjunctive particles in terms of meaning. One typical example of „-e/ase‟ (and then) is given below: (87) a. John-un [ pro hakkyo-ey ka- ]-se, Mary-lul manna-ss-ta. John-NOM school-LOC go-PRS-CONJ Mary-ACC meet-PST-DEC „John went to school, and then he met Mary.‟ b.*John-un [ pro hakkyo-ey ka-ss]-se, Mary-lul manna-ss-ta. John-NOM school-LOC go-PST-CONJ Mary-ACC meet-PST-DEC (Int.)„John went to school, and Mary went to kindergarten.‟ When two clauses are conjoined by the clausal ender „-se‟ (and then) classified as a „temporal conjuctive‟ particle in S.O. Sohn (1995), (87)-a receives a reading that the first clause event precedes the second one temporally, as the English gloss „and then‟ indicates. But, we exclude clausal enders with a temporal ordering like „-se‟ (and then) from conjunctives because the clauses headed by such enders are structurally subordinated rather than simply conjoined. See that a past tense is not allowed in the „-se‟ clause in (87)-b indicating that it is not an independent clause by itself. Also, the subject of the 98 clause should be a pro form. Otherwise, the ender cannot keep its original temporal meaning as in (88)-b: (88) a. John-un [ pro hakkyo-ey ka- ]-se, Mary-lul manna-ss-ta. John-NOM school-LOC go-PRS-CONJ Mary-ACC meet-PST-DEC „John went to school, and then met Mary.‟ b. [ John-i hakkyo-ey ka- ]-se, Mary-ka ul-ess-ta. John-NOM school-LOC go-PRS-CONJ Mary-NOM cry-PST-DEC „Because John went to school, Mary cried.‟ When a sentence conjoined by „-se‟ has two distinct overt subject in both clauses as in (88)-b, a temporal meaning disappears and a casual meaning is associated with the clausal ender „-se. 52 ‟ The two facts above show that the „-se‟ clause is structurally dependent on the second clause. Hence, the clause headed by „-se‟ cannot be regarded as an independent clause, consisting of a conjunct construction. Rather, it is to be considered as a subordinate clause or small one with no tense in it. 52 In (80b), the tense of „-se‟ does not need to be a past tense. It can be a non-past tense as in the following example: (i) [ John-i nayil hakkyo-ey ka- ]-se, Mary-ka ul-ess-ta. John-NOM tomorrow school-LOC go-PRS-CONJ Mary-NOM cry-PST-DEC „Because John will go to school tomorrow, Mary cried.‟ The „-se‟ clause allows „nayil‟ (tomorrow) to appear in it as in (i) above, which mean that the „-se‟ clause in (80a) cannot have an overt subject in it and it differs from the one in (80b). 99 2.2.2.2. Non-complement subordination 2.2.2.2.1. Temporal adjunct clauses: Some of temporal adjunct clauses with a temporal ordering do not allow a tense marker inside them: (89) a. John-un [ Mary-ka o]-ki cene, ttena-ss-ta. John-NOM Mary-NOM come-before leave-PST-DEC „Before Mary came, John left.‟ a'. *John-un [ Mary-ka o-ass]-ki cene, ttena-ss-ta. John-NOM Mary-NOM come-PST-before leave-PST-DEC (Int.) „Before Mary came, John left.‟ b. John-un [ Mary-ka o]-n hue, ttena-ss-ta. John-NOM Mary-NOM come-after leave-PST-DEC „After Mary came, John left.‟ b'. *John-un [ Mary-ka o-ass]-n hue, ttena-ss-ta. John-NOM Mary-NOM come-PST-before leave-PST-DEC (Int.) „After Mary came, John left.‟ c. John-un [ Mary-ka o]-nun tongan, ttena-ss-ta. John-NOM Mary-NOM come-while/during leave-PST-DEC „While Mary was coming, John left.‟ c'. *John-un [ Mary-ka o-ass]-nun tongan, ttena-ss-ta. John-NOM Mary-NOM come-PST-while/during leave-PST-DEC 100 (Int.) „While Mary was coming, John left.‟ As in (89)-a', b', c', a past tense marker cannot be followed by adverbial clausal enders. In (89)-a,b,c, there is no covert present tense marker because the adverbial clauses do not have an anaphoric reading as well as an indexical reading. For example, if the covert / null tense in (89)-a is a simultaneous present tense, (89)-a should have an anaphoric reading where the embedded event overlaps with the matrix event or an indexical reading where it overlaps with the speech time. However, none of the readings are available for (89)-a because the adverbial clausal ender „-ki cene‟ (before) requires the embedded event of (89)-a to occur prior to the matrix event. The same thing goes for (89)-b and (89)-c. Therefore, we conclude that there is no overt or covert tense marker in the subordinated adverbial clauses in (89). Yet, there is another type of temporal adjunct clauses which allow tense markers as in (90) below: (90) a. John-un [ Mary-ka LA-ey sa(l)- ]-l ttay LA-ro isaha-yss-ta. John-NOM Mary-NOM LA-LOC live-PRS-when LA-to move-PST-DEC „When Mary lived in LA, John moved to LA.‟ a'. ?John-un [ Mary-ka LA-ey sal-ass]-ul ttay LA-ro isaha-yss-ta. John-NOM Mary-NOM LA-LOC live-PST-when LA-to move-PST-DEC „When Mary lived in LA, John moved to LA.‟ b. John-un [ Mary-ka LA-ey sal- ]-ki ttaymuney LA-ro isaha-yss-ta. 101 John-NOM Mary-NOM LA-LOC live-PRS-because LA-to move-PST-DEC „Because Mary lived in LA, John moved to LA.‟ b'. John-un [ Mary-ka LA-ey sal-ass]-ki ttaymuney John-NOM Mary-NOM LA-LOC live-PST- because LA-ro isaha-yss-ta. LA-to move-PST-DEC „Because Mary had lived in LA, John moved to LA.‟ c. John-un [ Mary-ka LA-ey sa(l)- ]-nunteto pulkuhako John-NOM Mary-NOM LA-LOC live-PRS-even if LA-ro isaha-yss-ta. LA-to move-PST-DEC „Even if Mary lived in LA, John moved to LA.‟ c'. John-un [ Mary-ka LA-ey sal-ass]- nunteto pulkuhako John-NOM Mary-NOM LA-LOC live-PST -even if LA-ro isaha-yss-ta. LA-to move-PST-DEC „Even if Mary had lived in LA, John moved to LA.‟ In (90)-a, b,c, the embedded adverbial clauses have a simultaneous present tense because those sentences are interpreted simultaneous to the matrix past tenses; it might be unclear in (90)-a since the simultaneous reading is somewhat forced by the adverbial clausal ender with the function of a simultaneous temporal ordering. But, the clausal enders, „-ki 102 ttaymuney‟ (because) and „-nunteto pulkukaho‟ (even if) in (90)-b and (90)-c respectively do not carry any temporal ordering, and still the sentences have a simultaneous reading. Also, the second set of data in (90)-a', b', c' shows that a past tense marker is allowed to appear before the adverbial clausal markers, resulting in a back shifted reading in (90)-b and (90)-c 53 . The facts so far indicate that the embedded tenses in adverbials clauses in (90) are interpreted in related to the matrix tenses, indicating that an anaphoric interpretation of embedded tenses is available for (90). Then, let us take a close look at the examples to see if an indexical temporal interpretation is also available for the embedded tenses in temporal adjunct clauses. (91) a. [ Mary-ka ttena- ]-ki ttaymuney John-un ul-ess-ta. Mary-NOM leave-PRS-becauseJohn-un John-TOP cry-PST-DEC „Because Mary left (at that time), John cried.‟ b. [ Mary-ka (today / nayil) ttena- ]-ki ttaymuney Mary-NOM (today/tomorrow) leave-PRS-becauseJohn-un John-un ul-ess-ta. John-TOP cry-PST-DEC „Because Mary is leaving (today / tomorrow), John cried.‟ In (91)-a, the embedded present tense in the adverbial clause is interpreted simultaneous 53 In (82a), the back shifted reading is not available because of the simultaneous ordering that the adverbial clausal ender has as its functional meaning. 103 to the matrix past tense. Whereas, in (91)-b, it has an indexical non-past meaning so that the embedded tense is interpreted with respect to the speech time: the embedded event follows the speech time, but not the matrix event time. Thus, the data in (91) demonstrates that a simultaneous present tense can be interpreted indexcially as well as anaphorically. Is the same pattern found in a past tense embedded in adverbial clauses? The answer is yes. Even though the typical example of it, given in (92), often appears to have a relative interpretation only, namely a past shifted reading, as in (92), we can construction examples where it receives an indexical reading as in (85): (92) [ Mary-ka ttena-ss]-ki ttaymuney, John-un ul-ess-ta. Mary-NOM leave-PST-becauseJohn-un John-TOP cry-PST-DEC „Because Mary had left, John cried.‟ Because the clausal ender „-ki ttaymuney (because)‟ imposes a casual relation between the matrix and subordinated clauses, the cause clauses headed by it is usually understood to precede the effect one, the matrix clause event. However, it does not need to be that way always. See the following examples: (93) a. [ Mary-ka kyulhonha-yss]-ki ttaymuney, John-un senmul-ul pona-yss-ta. Mary-NOM leave-PST-because John-un gift-ACC send-PST-DEC „Because Mary got married, John sent a gift (to her).‟ b. [ Mary-ka samwol-e kyulhonha-yss]-ki ttaymuney, 104 Mary-NOM March-in get married-PST-because John-un sawol-e senmul-ul pona-ss-ta. John-un April-in gift-ACC send-PST-DEC „Because Mary got married in March, John sent a gift (to her) in April.‟ c. [ Mary-ka sawol-e kyulhonha-yss]-ki ttaymuney, Mary-NOM April-in get married-PST-because John-un samwol-e senmul-ul pona-ss-ta. John-un March-in gift-ACC send-PST-DEC „Because Mary got married in April, John sent a gift (to her) in March.‟ The sentence in (93)-a is multi way ambiguous in terms of temporal interpretation. The subordinated event can precede or follow the matrix event as in (93)-b and (93)-c respectively. And, it can even overlap with the matrix event, as we already saw in (90)-a' and (90)-b', repeated below: (94) a. ?John-un [ Mary-ka LA-ey sal-ass]-ul ttay John-NOM Mary-NOM LA-LOC live-PST-when LA-ro isaha-yss-ta. LA-to move-PST-DEC „When Mary lived in LA, John moved to LA.‟ b. John-un [ Mary-ka LA-ey sal-ass]-ki ttaymuney John-NOM Mary-NOM LA-LOC live-PST- because 105 LA-ro isaha-yss-ta. LA-to move-PST-DEC „Because Mary had lived in LA, John moved to LA.‟ Because the subordinated tense is an indexical past tense, as is the matrix tense, the two past tenses can be interpreted independently, resulting in several possible readings, even though the two clauses are not completely independent because they are connected by a causal relation so that the causal relation affects or put some restriction on temporal interpretation. In sum, the data examined so far lead us to conclude that embedded tenses in adverbial clauses can have both anaphoric and indexical readings whether they have a temporal ordering or not. 2.2.2.2.2. Relative clauses: It has been often noticed that relative clause tenses in Japanese is ambiguous between an anaphoric reading and an indexical one (Ogihara 1996): (95) Taroo-wa [ nai-te i-ru otoko]-o mi-ta. Taro-TOP cry-PROG-PRS man-ACC see-PAST Reading #1: “Taro saw a man who WAS crying (at the time of the meeting)” Reading#2: “Taro saw a man who IS crying (now-the speech time)” 106 The covert present tense in the relative clause can be interpreted simultaneous to the matrix past tense, or to the speech time, being independent of the matrix tense. The same pattern is observed in Korean: (96) John-nun [ LA-ey sa(l)- -]-nun yeca-lul manna-ss-ta. John-TOP LA-LOC live-PRS-REL woman-ACC meet-PST-DEC Reading #1: „John met a woman who lives in LA (now-the speech time).‟ Reading #2: „John met a woman who lived in LA (at that time).‟ The simultaneous present tense in the relative clause puts an overlapping temporal relation between the relative and matrix events as in Reading #2, or it is just simultaneous to the speech time, yielding an independent reading. Things turn out a bit complicated with a past tense embedded in a relative clause because there are more readings available, but the bottom line is the same as in a relative present tense. See the following example: (97) Mary-nun [ Olympic-eyse kum meytal-ul tta]-n Mary-TOP Olympic-at gold medal-ACC win-PST.REL namca-wa teyitu-lul ha-yss-ta. man-with date-ACC do-PST-DEC Reading #1: „Mary dated a man who had won an Olympic gold medal.‟ Reading #2: „Mary had dated a man who won an Olympic gold medal.‟ 107 The natural and preferred reading of (97) is Reading #1 where the relative past tense receives a past shifted reading where the relative clause event precedes the matrix one. So, in Reading #1, the relative tense is interpreted in relation to the structurally higher tense, namely, the matrix tense. Yet, another reading is also available. In reading #2, the event of the relative clause can be preceded by the matrix event, which becomes clearer with temporal adverbs as in (98): (98) Mary-nun [ ipun hay Olympic-eyse kum meytal-ul tta]-n Mary-TOP this year Olympic-at gold medal-ACC win-PST.REL namca-wa cak kyun-e teyitu-lul ha-yss-ta. man-with last year-in date-ACC do-PST-DEC Reading #2: „Mary (had) dated with a man last year who won an Olympic gold medal in this year.‟ As in (98), the independent reading in question is also available so that the embedded event can follow the matrix event. In fact, it is not clear from (97) whether a past shifted reading of a past tense is really available in a relative clause or not because the same reading is obtained when a past tense is interpreted indexically / independently, as in (97). In other words, as an independent temporal ordering of the two clauses in (97) can result in a past shifted reading, the real past shifted reading obtained in relation to a higher tense is obscured. Yet, we maintain our position that a past tense in a relative clause also has an anaphoric / 108 relative interpretation to a matrix tense. 2.2.2.3. Complement clauses In Korean, a present tense embedded in a complement clause is interpreted simultaneous to a matrix tense. See the following examples: (99) a. present under past John-nun [ caki-ka LA-ey iss- -ta]-ko malha-yss-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC „John 1 said that he 1 was in LA (at that time).‟ b. present under non-past John-nun [ caki-ka LA-ey iss- -ta]-ko malha-n-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC „John saves that he 1 is in LA.‟ The data in (99) show that the embedded overt present tense, , is anaphoric to its structurally higher tense in a matrix clause, whatever higher tense comes. In the same manner, an embedded past tense is interpreted relative to the matrix tense, bearing an anterior relation: locating the embedded event anterior to the matrix event: 109 (100) a. past under past John-nun [caki-ka LA-ey iss-ess-ta]-ko malha-yss-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC „John 1 said that he 1 had been in LA (at that time).‟ b. past under non-past John-nun [caki-ka LA-ey iss-ess-ta]-ko malha-n-ta. John-TOP self-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC „John says that he 1 was in LA.‟ (100)-a shows that the embedded past tense is not interpreted indexically, namely, with respect the speech time. Otherwise, the embedded tense and the matrix one might end up with being simultaneous to each other when they happen to denote the same past time when they are interpreted indexically. The fact that the embedded past event of (100)-a is interpreted prior to the matrix past event demonstrates that the embedded past tense is always interpreted in relation to the matrix tense. Before we move on, let us informed that Ogihara (1999, 2008) claims that Japanese present tense behaves differently in a complement clause. According to Ogihara‟s (1996) null tense theory, the English sentence in (101)-a and the Japanese one in (101)-b have the same logical form as in (101)-c: (101) a. English: past under past Taroo said that Hanako was in Seattle 110 b. Japanese: present under past Taroo-wa [ Hanako-ga Siatoru-ni i- -ru]- to it-ta Taro-TOP Hanako-NOM Seattle-LOC be-PRS-that say-PST-DEC „Taroo said that Hanako was in Seattle.‟ c. An LF for (79a) and (79b): Taroo Past say [ CP Hanako be in Seattle ] The surface difference between the embedded past and present tenses in (101)-a and (101)-b respectively is eliminated by the Tense Deletion Rule 54 which optionally deletes the embedded past tense in the English case in (101)-a. Hence, Japanese and English sentences end up with obtaining the identical semantic interpretation as in (101)-c. He proposes that the complement clause in (101)-c actually does not have any tense form, that is, it is literally tenseless so that it is a natural consequence that the indexical interpretation of the embedded tense is not available for (101)-c because there is no indexical tense in (101)-c at all. Thus, the null tense functions like a relative tense that is to be evaluated with a syntactically higher tense, resulting in a simultaneous reading of 54 The so-called Sequence of Tense rule is stated in Ogihara(1996) as below: If a tense A is locally c-commanded by another tense B at LF and A and B are occurrences of the same tense (i.e. either present or past), A is optionally deleted. ▶ .. T B .. […. T A ….]   The deletion rule applied ▶ .. T B .. […. …. ] As we can see above, the complement clause with the empty tense, , left by the Tense Deletion Rule is tenseless by itself and relative to a higher tense. 111 (101)-c 55 . This was the original position that Ogihara (1996) had took, but in the modified version of his view in Ogihara (1999a) and his subsequent work in Ogihara (2008), he proposes that Japanese has not only a relative tense but also an absolute one, in parallel to English so that (101)-b is ambiguous between a simultaneous reading and an indexical reading of the embedded present tense. The former simultaneous reading is accounted for by the same way as in Ogihara (1996) while the latter indexical reading leads to a double accessibility reading, as in English. Because Ogihara (1996, 1999a, 2008) strictly sticks to Comrie‟s (1985) idea that natural language employs an absolute tense and a relative one, he finds it as a more systematic and straightforward analysis to posit that Japanese as well as English has all types of the absolute and relative tenses: a relative / anaphoric present, an absolute present, a relative / anaphoric past, and an absolute past, even though Ogihara (1999a) admits that there is no strong evidence for a relative past tense in Japanese and that the absolute present tense, responsible for a double access reading, is obscured by the other very salient relative tense reading, appearing very weak. 55 Yet, more precisely speaking, that the complement clause is tenseless does not guarantee a simultaneous reading for (79c). The way for (79c) to get a simultaneous reading is provided by semantics. In Ogihara(1996), the temporal variable created by the deletion rule in the embedded clause is to be semantically bound by a  operator. It means that the simultaneous meaning is obtained by „ -abstraction‟ over the embedded temporal variable, which can be illustrated as follows: (i) a. Taroo said that Hanako was in Seattle. b. Taroo Past say that Hanako  be in Seattle. c. t 1 [t 1 < s* say'(t 1 , Taroo, ^ t 1 [ be in Seattle' (t 1 , Hanako) ] 55 d. ^ t 1 [ be in Seattle' (t 1 , Hanako)] = {<w,t 1 >  Hanako is in Seattle in w at t 1 } Ogihara(1989, 1996) treats indirect discourse verbs, such as „say,‟ in the same way as propositional attitude verbs like „think,‟ which means that they all denotes relations between persons and sets of world-time pairs, as we can see from 0. Hence, a tenseless embedded clause „Hanako be in Seattle’ is translated as a lamda expression that denotes sets of world-time pairs, as in 0-d above. 112 After all, Ogihara (1999a, 2008) tries to establish a parallelism between a complement clause tense and relative clause one in that relative and absolute tenses can appear in both of clauses. However, it turns out that a complement clause tense differs from a relative clause one because an absolute present tense should be realized as a double access reading, as claimed in Ogihara (1999a, 2008), unlike the one in a relative clause. So, in order to account for the special feature of the absolute present tense of a complement clause, Ogihara (1999a) needs to make a use of his generalization of Temporal Directionality Isomorphism 56 . After all, it is obvious that a complement place a certain constraint, whether syntactic, semantic or pragmatic, on a temporal interpretation of embedded tenses. Plus, the basic interpretation of past under past in a complement clause faces a problem in Ogihara (1999a). See the example below: (102) Japanese: past under past Taroo-wa [ Hanako-ga Siatoru-ni i-ta]- to it-ta Taro-TOP Hanako-NOM Seattle-LOC be-PST-COMP say-PST-DEC „Taroo said that Hanako had been in Seattle.‟ Putting aside an issue of a relative past tense in Japanese, put forth by Ogihara (1999a), 56 I cite a short summary of his Temporal Directionality Isomorphism provided by Ogihara (1999a): The basic idea is that a de re report about a temporal entity can only be made when the speaker‟s viewpoint and the attitude holder‟s viewpoint match with respect to their temporal directionality. For example, an interval that is described as a future interval by the attitude holder must also be reported as a future interval in a de re report. 113 the complement clause in (102) has an absolute past tense form, yielding a past shifted reading, according to Ogihara (1996, 1999a). Yet, the absolute past tense does not automatically yield the past shifted reading because the embedded past event could overlap with the matrix past event if it is interpreted independently of the matrix past tense as in a relative clause. Thus, once again, Ogihara (1996, 1999a) needs the semantic constraint of „Temporal Directionality Isomorphism‟ to block such a reading. Because a relative tense only means a simultaneous or anaphoric temporal relation in Ogihara (1996), it is hard to capture the true relative interpretation of a past shifted reading between the matrix event and the embedded one. But, in the proposed analysis where the reference time of the complement clause takes the matrix event as its antecedent, both the anaphoric relation and the shifted meaning are explained by one mechanism. In Korean, we do not find any good reason to assume that there is an absolute present tense, yielding a double access reading as in English. The alleged double access reading that we might get in present under past in Korean when we follow Ogihara (1999, 2008) is not a real double phenomenon, but it is obtained by pragmatic inferences available with a given assertion. It will be discussed in Chapter 5 in details. Hence, we stick to our & Ogihara‟s (1996) original position that only the anaphoric reading is available for present under past in a complement clause in Korean. 2.2.3. Summary The types of temporal interpretation discussed so far can be summarized as follows: 114 (103) Table 4: Temporal interpretation across clauses 57 Conjunctive C. Subordinated Clauses Non-Complement C. Complement C Adverbial C Relative C Anaphoric Reading * √ √ √ Indexical Reading √ √ √ * 57 The symbol * stands for the unavailability of a given reading while √ means that a certain type of clauses can the reading. 115 Chapter 3. Dynamics in Temporal Dependency In the previous chapter, I introduced theories of tense in Korean and discussed the important issues, evaluating those theories. In this chapter, I turn to the proposed analysis and will show how it works and what advantages it has in accounting for matrix and subordinate tenses in Korean. 3.1. The skeleton of the proposal 3.1.1. The temporal anaphor Rx Let us start with the basic scheme of the proposed analysis of tense in Korean laid out as follows. As discussed in Chapter 1, it is assumed that tenses are a binary relation between two temporal elements, E-time and R-time: a present tense conveys a simultaneous relation and a past tense an anterior one. E-time that is one of time arguments of tense is the time interval associated with an event described by a verb in a clause, and R-time is the time with respect to which the given event is evaluated in terms of time in a clause. Given these, the definition of tense will be presented formally, as follows: (103) a. [[Past ]] = λP. λy. [ e 1 : (e 1 ) < R x ] & P(y, e 1 ) b. [[Present ]] = λP. λy. [ e 1 : (e 1 ) Θ R x ] & P(y, e 1 ) In (103),  is a function from an event to time and R x R x is R-time that is a temporal pronoun. It is assumed that tense, as a temporal ordering relation, is not deictic or 116 relative by itself; tenses in matrix and subordinate clauses or in discourse are related through the binding relation of R-time of a given clause 58 . Now, we need to further specify what R x is. Following Partee (1984), it is assumed that R x is a temporal pronoun of a covert form that carries temporal features. Syntactically, it originates in Spec of a TP, following Stowell (2007), as below: (104) TP R x (RT) T‟ T vP (ET) past/present Because R x is a temporal pronoun, like a third person pronoun „he/she,‟ it can have a bound variable reading and an indexical reading as well. When there is another temporal elements, a ZP in Stowell‟s (2007) term within a sentence, R x is bound by it; when there is no co-referring temporal element in the structure above it, the default value, S-time, is given to it. Given this, it is predicted that while S-time generally provides the value for R x 58 Tense can be anaphoric to another tense because it contains an anaphoric element which is R-time. So, the distinction between absolute versus relative tenses is treated as an epiphenomenon as in Kubota (2008) which does not play an important role in the present analysis. 117 in matrix clauses, yielding an unambiguous reading, a sentence with subordinate tenses can be able ambiguous to have because there are more than one temporal element available as an antecedent of R x . 3.1.2. The Speaker at a CP layer The deictic information of the speaker and hearer plays an important role in the interpretation of tenses. Yet, so far, the term of „Speech time‟ has been informally used in the proposed analysis by saying the S-time is a default case of R-time, along the line of Giorgi & Pianese (1997) and Higginbotham (2002, 2006). Now, we need to be more specific about it. Adopting Rizzi‟s (1998) Split CP hypothesis and Giorgi‟s (2010) hypothesis of indexicality, we present the syntactic structure of a subordinate clause as follows: 118 (105) CP e(U) C TP R x T‟ T vP (ET) past/present The structure above shows that there are multiple CP layers and a Specifier position of one of the CPs holds the speaker and hearer‟s information, e(U). Even though not fully elaborated yet in Giorgi (2010), the presence of the indexical information about the speaker and hearer‟s temporal / spatial location at a CP layer accounts for the interpretation of all indexical items within its domain. In other words, the contextual information of the speaker and hearer is introduced into the syntactic structure so that the speaker‟s temporal location specified at Spec of a CP enables a subordinate clause to have an indexical interpretation of tense, by binding a temporal pronoun locally. In that sense, the antecedent of R x is locally provided in this system. 119 3.1.3. The unified system As discussed in Ch. 1, the proposed analysis presents a simple and unified system of tenses in Korean 59 . It is argued in this paper that Korean has past and non-past tenses 60 which are all relative in that they are to be evaluated with respect to the R-time of a given clause. The R-time, R x , is the local evaluation time that every TP should have, and S-time is assumed as one kind of R-time. Given this, the proposal is able to treat matrix tenses, often viewed as a deictic element, in the same way the embedded tenses are interpreted, not failing to capture the strong connection to the speaker‟s temporal location, i.e., S-time. In this unified system, both matrix and subordinate tenses are defined as a temporal relation between R-time and E-time. In the following subsections, it will be shown how the system works to account for the interpretation of matrix and subordinate tenses in Korean. 3.2. Temporal dependency in matrix tenses The basic structure of matrix tenses is as given in (104). The interpretation of matrix tenses is simple and clear in the sense that there is only one antecedent for a R-time of 59 There is no morphological agreement involved and verbal aspect (aktionart) does not play much role, not posing any restriction on verbs even for a simultaneous reading, unlike English. 60 The inventory of Korean tense is given as follows: Past -ess- Non-past -nun- / -Ø - A past tense conveys an anterior relation while a non-past tense an overlapping one. 120 matrix tenses, i.e., S-time so that past and non-past tenses are evaluated with respect to it: with a past tense an given event precedes S-time while with a non-past tense the event overlaps with S-time. Because there is no other tense above matrix tenses in the structure, it might be hard to imagine any temporal dependency relation in matrix tenses under the traditional deictic approach to tenses. However, there exists a temporal dependency „within‟ and „across‟ a sentence boundary. Let us take a close look at such cases. 3.2.1. Past in the future Context plays a role providing an appropriate antecedent for a temporal pronoun when it is unbound syntactically. As mentioned before, the speaker‟s information about time and location is the immediately available information to speakers and hearers. But, it does not need to be always that way, and context can provide a suitable antecedent for a temporal pronoun. As one of the examples, there is a case where a past tense is used for a non-past situation in matrix clauses. Though highly restricted, a past tense morphem „-ess-‟ can co- occur with a future time adverb in a matrix clause, as below: (106) Chulswu-nun nayil cuk-ess-ta. Chulswu-TOP tomorrow die-PST-DEC „(Lit.) *Chulswu was dead tomorrow.‟ „(Int.) (I bet) Chulswu will be dead tomorrow.‟ 121 In order to get the intended reading, the sentence in (106) should be uttered with a peculiar intonation in which the matrix verb receives a focal accent. Otherwise, it is ungrammatical or unacceptable due to the temporal mismatch between the past tense and the future time adverb. To the analyses based on the deictic definition of tenses, (106) is problematic because (106) with the past tense morpheme „-ess-‟ describes an event in the future. However, the proposed analysis has room to accommodate the idiosyncratic meaning of the past tense „-ess-.‟ Normally, the R-time of matrix clauses is set as S-time at default. Yet, if there is a more salient time interval available that is contextually provided, the R-time of a matrix clause can be bound by it. For (106), its R-time is not S- time but a certain future time that the speaker has in mind. Then, the anterior past tense meaning can place the event time of (106) after S-time, thereby allowing the future time adverb to modify the event time. The peculiar interpretation of (106) becomes possible when the speaker treats the non-past event as the past event that has already taken place by choosing the past tense form „-ess-.‟ In other words, when the speaker is very sure that an event is going to take place, she considers it as a past event, which is reflected in the choice of a tense morpheme. This change of the speaker‟s temporal perspective can be accounted for within the proposed analysis because „-ess-‟ is an anterior past tense, but not a deictic past tense that has to be evaluated with S-time only. When the speaker chooses a contextually salient future time, rather than S-time, as the R-time of the matrix clause in (106), the shift of the speaker‟s reference time takes place. Then, because the speaker takes a reference position in the future, the non-past event of (106) can be anterior to the speaker‟s reference time so that the anterior past „-ess-‟ is chosen to be 122 used for the interpretation. 3.2.2. Present in the past: the historical present tense Sometimes, an antecedent of R-time can be sought across a sentence boundary. A so- called historical present is an example of it. The „historical‟ present tense is found cross- linguistically in many languages and has been considered as an exceptional case, challenging the analyses based on the deictic definition of tense. The example of a discourse below illustrates it: (107) I was walking home from work one day. All of a sudden this man comes up to me and says.... In the second sentence of (107), the present tense does not pick up the speech time as its reference time. Otherwise, it should mean that the event of the man‟s coming up is taking place at the speech time. But, the historical present is understood as not being associated with the speech time. Rather, the present tense in the second sentence conveys the simultaneous meaning to the previous event across the sentence boundary in a discourse. The same pattern is found in Korean, as follows: (108) a. Chulswu-nun ul-ess-ta. Chulswu-TOP cry-PST-DEC „Chulswu cried.‟ 123 b. Ku-nun amu mal-to haci-an-nun-ta. He-TOP any word-even do-NEG-PRS-DEC „Chulswu does not say a word.‟ The present tense of the second sentence in (108) is interpreted as simultaneous to the previous sentence with a past tense. The phenomenon is not confined t a written text but also found in an oral dialogue: (109) (Sumi is telling her sister about her dream last night: ) a. Nay-ka ku mun-ul yel-ess-e. I-NOM he door-ACC open-PST-DEC „I opened the door.‟ b. Kurenika, calsangkin namca-ka nao-nun-kes-i-ya „Then, handsome guy-NOM come-PRS-fact-COP-DEC „Then, (to my surprise) a handsome guy came in. As the English gloss shows, the present tense of the second sentence is evaluated with respect to S-time, but the event time of the previous sentence. The proposed analysis explains this straightforwardly. The simultaneous meaning of the present tense of the second sentence in (108) or (109) remains the same; the important difference from the usual present tense meaning is that not the speech time but another contextually salient event time, that is, the preceding event time, is picked up for the R-time of the second 124 sentence in (108) or (109). In other words, the temporal pronoun, R x , of the second sentence takes the event time of the previous sentence as its antecedent. If this is right, the fact that the second sentences takes the present form is not meaningless or just a coincidence, but is very consistent in conveying a simultaneous meaning of a present tense. Here again, the choice of tense reflects the shift of the speaker‟s temporal perspective. Let me note in passing that the historical present tense proves that the temporal pronoun analysis presented in this paper is advantageous, compared to a similar view of Stowell‟s (2005) temporal PRO analysis. Stowell‟s (2005) temporal PRO analysis fails to account for this phenomenon; according to Stowell‟s (2005) temporal PRO analysis, the best candidate for the controller of a temporal PRO of the second sentences would be S- time because it is a temporal element in a minimal distance which is provided by a superordinate performative speech predicate. The event time of the preceding sentence is syntactically further away from the temporal PRO of the following sentence, and is not suitable for its controller 61 . Nevertheless, the event time of the preceding sentences is chosen to be a controller, contra to Stowell‟s (2005) prediction, yielding the interpretation of the historical present tense. 61 If Stowell (2005) wants to overcome the technical problem in explaining the peculiar present tense meaning of (109), he might have to posit another sub-part of the superordinate performative verb concerning a past dream, like „I say to you about my dream….‟ Yet, the controlling relation would not look neat. 125 3.2.3. The Double form of ‘-ess-ess-’ A double form of „-ess-ess-‟ can be treated properly in the proposed analysis. I propose that there are two kinds of „-ess-ess-‟ in Korean. The first type of „-essess-‟ is a morphologically simple past tense form with a specific temporal presupposition. The other type of „-ess-ess-‟ is a morphologically complex one that composes of an aspectual auxiliary „-ess-‟ and a past tense „-ess-.‟ For the first type of „-essess-,‟ it is claimed that it is a morphologically simple past tense form conveying an anterior relation. Several crucial evidence against the past perfect analysis of „-ess-ess-‟ point out that „-essess-‟ is not a combined form of a present perfect and past tense 62 . Unlike the English past perfect in (110), no two (distinct) time references are allowed, leading to the absence of temporal ambiguity in Korean below: (110) He had left at 3pm Reading 1: „At 3pm, he had already left.‟ Reading 2: „he left at 3pm and it was before some specific past time.‟ (111) Ku namca-nun sey shi-ey ttena-essess-ta. the man-TOP three o‟clock-at leave-?-DEC „The man left at 3pm and it was before some specific past time.‟ In contrast to the ambiguity of (110), the time adverb „sey shi-ey (at 3 o‟clock)‟ is only 62 See Chung (2005) for detailed discussion. 126 associated with the event time of the man‟s leaving in (111). Also, as mentioned in 2.4.1 and repeated below, „-essess-‟ can receive a simultaneous reading to the subordinate past tense: (112) Mina-ka wo-ass-ul ttay Chulswu-nun ttuna-ssess-ta. Mina-NOM come-PST-REL when Chulswu-TOP leave-PST-DEC „When Mina came in, Chulswu left.‟ When „-essess-‟ appears in the matrix clause embedding a past tensed adjunct clause, a simultaneous interpretation is obtained so that (112) can mean that Chulswu‟s leaving event can coincide Mina‟s coming event in the past. If „-ess-ess-‟ is a past perfect, such a simultaneous reading. The main evidence supporting the proposed analysis that „-essess-‟ is a past tense morpheme conveying an anterior relation comes from the existence of an anterior interpretation of „-essess-‟ in subordinate clauses. When embedded under a non-past tense, though restricted, it is possible to get the interpretation where „-essess-‟ is evaluated with respect to some future time, rather than the speech time, yielding the past- in-the future reading, as follows: (113) [Context: Kim‟s son is sick now and doesn‟t seem to get better by tomorrow when an important meeting for Kim to attend is scheduled. So, Kim is considering not 127 attending the meeting, thinking what he has to tell his boss.] a. Kim-un [ adul-i ap(u)-ass-ta]-ko malha-l kes i-ta. Kim-TOP son-NOM sick-?-DEC-COMP say-FTR-DEC „Kim will say that that his son was sick.‟ b. Kim-un [ adul-i ap(u)-assess-ta]-ko malha-l kes i-ta. Kim-TOP son-NOM sick-?-DEC-COMP say-FTR-DEC „Kim will say that that his son was sick.‟ There is no significant difference in meaning between (113)-a and (113)-b in that both of the complement clause tenses can convey an anterior relation. If „-essess-‟ is a deictic past tense, as Chung (2005) claimed, such a past-in-the future interpretation would not be available for „-essess-‟ in (113)-b. Therefore, the availability of the reading supports the (anterior) past tense analysis of „-essess-‟ defended in this paper. If „-essess-‟ is another past tense 63 , then, how does it differ from the regular past tense „-ess-‟? It is proposed in this paper that unlike „-ess-,‟ „-essess-‟ has a specific temporal presupposition that makes it as a rigid anterior relation, as put below: (114) ‘-Essess-’ presupposes: every part of E or a result state of E should precede R-time 64 . 63 Since a past tense refers to an anterior relation where E-time precedes R-time under the present analysis, the distinction between a „deictic past tense‟ and an „anterior past tense‟ does not hold. 64 E stands for an event. 128 What the presupposition of „-essess-‟ tells us is that any part of the event described by a given verb cannot overlap with the R-time of a clause. Thus, a result state of the event as well as the event itself should precede the R-time of a clause. In contrast to it, „-ess-‟ does not have such a restriction stemming from a presupposed meaning so that a result state of a given event can extend from the past time set by „-ess-‟ up to the point overlapping with the reference time. Given this, the contrast between „-ess-‟ and „-essess-‟ with respect to the co- occurrence restriction with „cikum (now)‟ is well explained. The example in (71) in the section 2.4.1. is repeated below: (115) a. Chulswu-nun cikum ttuna-ess-ta. Chulswu-TOP now leave-PST-DEC „Chulswu left just now.‟ b. *Chulswu-nun cikum ttuna-ss-ess-ta. Chulswu-TOP now leave-?-PST-DEC „*Chulswu left now.‟ When it is assumed that the event „ttuna-ta (leave)‟ happened in the past very near to the speech time, (115)-a turns out acceptable, meaning that the event happened right before the speech time and the result state of the event continues at the speech time. But, even though such an assumption is given, (115)-b remains unacceptable. It is so because the temporal presupposition that „-essess-‟ carries requires the event time of the sentence or 129 its result state to strictly precede the speech time in (115)-b, preventing the event time from overlapping with the speech time. Hence, it cannot co-occur with the adverb „cikum (now)‟ referring to the speech time. On the other hand, in (115)-a with „-ess-‟ lacking such a presupposition, it is allowed that the result state of the event overlaps with the speech time, and so there is no conflict in meaning in (115)-a. Such a meaning of a discontinued event form the speech time forced by „-essess-‟ is accounted for as a presupposition in this paper. But, there are different approaches to it. C. Lee (1985) treats the peculiar meaning associated with „-essess-‟ as part of asserted / truth conditional meaning of „-essess-.‟ He accounts for this fact by proposing that „- essess-‟ has the discontinuity meaning of a given event or a result state of the event from the speech time as its truth conditional meaning along with a regular past tense meaning; whereas, „-ess-‟ has the discontinuity meaning as its pragmatic implicature. On the other hand, Chung (2005), the peculiar meaning of „-ess-ess-‟ i.e., no current relevance in Chung‟s (2005) term, is explained as a pragmatic implicature that „-essess-‟ has. The proposed analysis positing that the meaning of a discontinued event should be attributed to a presuppositional meaning of „-essess-‟ is based on the scope relation of „-essess-‟ to negation. The examples below show that the discontinuity meaning of „- essess-‟ is not negated, indicating that the meaning is presuppositional. (116) a. Chulswu-nun tampay-lul piwu-essess-ta. Chulswu-TOP cigarette-ACC smoke-PST-DEC „Chulswu used to smoke a cigarette‟ 130 b. Chulswu-nun tampay-lul piwu-ci ahn-assess-ta. Chulswu-TOP cigarette-ACC smoke-NEG-PST-DEC NEG Narrow Scope Reading: „Chulswu used not to smoke a cigarette.‟ NEG Wide Scope Reading: „It is not the case that Chulswu used to smoke a cigarette = „Chulswu has never smoked a cigarette.‟ In (116), „-essess-‟ is typically associated with a habitual reading as the English glosses show. When a sentence with „-essess-‟ is negated as in (116)-b, negation takes a narrow scope and the sentence means a past habitual event of „not doing something.‟ Hence, the interpretation where negation takes a wide scope, as in a sentential negation above, is not available for (116)-b. Otherwise, (116)-b should be able to mean that Chulswu has never smoked a cigarette, but it is not the case 65 . Rather, (116)-b means that the habitual event of „Chulswu‟s not smoking a cigarette‟ holds for some past time, but it does not continue to the speech time, so that it eventually means that Chulswu is smoking at the speech time. The absence of the wide scope interpretation of negation indicates that the discontinuity meaning of „-essess-‟ is presuppositional because it is not negated, scoping over negation. 65 The same thing goes for a short negation „ahn‟ and negation always takes a narrow scope in relation to „- essess-,‟ as below: (i) a. Chulswu-nun tampay-lul piwu-essess-ta. Chulswu-TOP cigarette-ACC smoke-PST-DEC „Chulswu used to smoke a cigarette‟ b. Chulswu-nun tampay-lul ahn piwu-essess-ta. Chulswu-TOP cigarette-ACC NEG smoke-PST-DEC „Chulswu used not to smoke a cigarette‟ (i)-b is not ambiguous between the wide and narrow scope interpretation of negation. 131 The presuppositional meaning of „-essess-‟ becomes clearer when it is contrasted to „-ess-,‟ as follows: (117) a. Chulswu-nun tampay-lul piwu-ess-ta. Chulswu-TOP cigarette-ACC smoke-PST-DEC „Chulswu smoked a cigarette‟ b. Chulswu-nun tampay-lul piwu-ci ahn-ess-ta. Chulswu-TOP cigarette-ACC smoke-NEG-PST-DEC NEG Narrow Scope Reading: „Chulswu didn‟t smoke a cigarette.‟ NEG Wide Scope Reading: „It is not the case that Chulswu smoked a cigarette.‟ As in (117)-b, when negated, a past tensed sentence by „-ess-‟ with no temporal presupposition can take a narrow or wide scope in relation to negation. Now, let us go back to the beginning of the discussion on „-essess-‟ and recall that it is proposed in this dissertation that there is another kind of „-essess-‟ that is morphologically complex. It is a combined form of an aspectual auxiliary „-ess-‟ and a past tense „-ess-.‟ It is argued in the previous chapter that the auxiliary „-ess-‟ that does not convey an anterior relation when attached to a certain class of predicates 66 . Consider following examples: 66 This special class of predicates includes „talm-ta(resemble),‟ „tul-ta(be contained),‟ „kaci-ta(possess),‟ „imshinha-ta(be pregnant),‟ „ip-ta / ssu-ta / kki-ta / shin-ta (wear: cloth / hat / glasses / shoes)‟ etc. 132 (118) a. ??/*Chulswu-nun Minho-lul talm-nun-ta. Chulswu-TOP Minho-ACC resemble-PRS-DEC (Int) „Chulswu resembles Minho.‟ b. Chulswu-nun Minho-lul talm-ass-ta. Chulswu-TOP Minho-ACC resemble-?-DEC „Chulswu resembles Minho.‟ „-Ess-‟ in (118)-b does not express a past situation when it appears with a stative predicate „talm-ta (resemble).‟ As the English gloss shows, (118)-b has the same temporal meaning with a present tensed sentence. Note that when a present tense marker „-nun-‟ is attached to the predicate, it makes the sentence ungrammatical or unacceptable unless it receives a generic reading 67 . In this dissertation, this type of „-ess-‟ is analyzed as an aspectual auxiliary, but not a tense marker. It means that (118)-b has a present tense marker of a null form Ø , which is responsible for the present tense meaning 68 . The non- 67 A sentence whose predicate is „talm-ta (resemble) followed by „-nun-‟ should have a generic reading, as below: (i) PwuPwu-nun (sero-lul) talm-nun-ta. Husband and wife-TOP (each other-ACC) resemble-PRS-DEC „Husband and wife become alike each other.‟ As the English gloss shows in (i), the predicate, when followed by „-nun-‟ is no longer a stative predicate but can be said to belong to accomplishment verbs. 68 Another example of the same kind: (i) a. ??/*Ku pongthu-e phyunci-ka tu(l)-n-ta. The envelop-in letter-NOM be contained-PRS-DEC (Int)„A letter is contained in the envelope.‟ b. Ku pongthu-e phyunci-ka tul-ess-ta. The envelop-in letter-NOM be contained-PST-DEC 133 past tense meaning with „-ess-‟ is shown clearly when embedded: (119) a. *Kim-un [ Chulswu-ka Minho-lul talm-nun-ta]-ko malha-yss-ta. Kim-TOP Chulswu-NOM Minho-ACC resemble-PRS-DEC say-PST-DEC (Int.) „Kim said that Chulswu resembles Minho.‟ b. Kim-un [ Chulswu-ka Minho-lul talm-ass-Ø -ta]-ko Kim-TOP Chulswu-NOM Minho-ACC resemble-ASP-PRS-DEC malha-yss-ta. say-PST-DEC „Kim said that Chulswu resembles Minho.‟ (119)-b containing „-ess-‟ in a complement clause has a simultaneous reading while (119)-a is ungrammatical for a non-generic reading. Hence, „-ess-‟ in (119)-b is not a past tense morpheme but an aspectual one. Given this, it would be natural to conclude that „- ess-ess-‟ attached to „talm-ta (resemble)‟ is another kind of „-ess-ess-‟ that should be distinguished from the morphologically simple one. The second „-ess-‟ in „-ess-ess-‟ in question is a past tense marker which is responsible for the anterior meaning of „-ess-ess-.‟ See the following example: „A letter is contained in the envelope.‟ The sentence in (i-b) describes a present situation, but not a past tense one. 134 (120) Kim-un [ Chulswu-ka Minho-lul talm-ass-ess-ta]-ko Kim-TOP Chulswu-NOM Minho-ACC resemble-ASP-PRS-DEC malha-yss-ta. say-PST-DEC „Kim said that Chulswu resembled Minho.‟ As expected, the complement clause in (126)-b has a back shifted reading, for which the second „-ess-‟ of the complement tense is responsible. 3.3. Temporal dependency in subordinate tenses 3.3.1. The asymmetry Now let us turn to the pattern of temporal interpretation in subordinate clauses, repeated below: (121) Table 5:the temporal interpretation of present under past Conjunctive C. Subordinated Clauses Non-Complement C. Complement C Adjunct C Relative C Anaphoric Reading * √ √ √ Indexical Reading √ √ √ * It is expected that conjunctive clauses get an indexical interpretation only because both of 135 tenses in conjunction are not structurally embedded and they should be behave like matrix tenses. But, the asymmetric pattern of interpretation in complement clauses lets us raise a question of what prevents an indexical reading in complement clauses, but not in relative or temporal adjunct clauses, exemplified below: (122) a. John-un [ Mary-ka LA-ey sa(l)-n-ta]-ko malha-yss-ta] John-TOP Mary-NOM LA-LOC love-PRS-DEC-COMP say-PST-DEC “John said that Mary lived in LA (at that time).” b. John-un [ e 1 shiaytul-ey sa(l)-nun]- eyca 1 -rul manna-ss-ta John-TOP Seattle-in live-PRS-Rel.-woman-ACC meet-PST-DEC Reading 1: „John met a woman who lived in Seattle (at the meeting time).‟ Reading 2: „John met a woman who is living in Seattle (at the speaker‟s now). If a temporal pronoun, R x , can be interpreted either anaphorically or indexically and there is further constraints working behind, an indexical reading as well as an anaphoric one should be available for the complement clause in (122)-a. The asymmetric pattern of temporal interpretation given in (122) reveals that there is a certain constraint on the interpretation of an embedded tense in a complement clause which rules out an indexical interpretation of the embedded tense. There have been several solutions sought for. In Ogihara‟s (1996) null tense theory, the complement clause does not have any interpretable tense in Japanese even though it takes a present tense form morphologically so that only the simultaneous / anaphoric interpretation is possible; 136 the relative clause in (122)-b is analyzed to have two kinds of tenses: a null tense and a deictic one which are responsible for the ambiguous interpretation. Yet, it is not explained why a deictic tense is not allowed to appear in a complement clause in Japanese type languages. Abusch‟s (1997) Upper Limit Constraint (ULC) can be understood to deal with it to some extent. The contsraint states that the tense of the embedding clause is an upper bound on the tenses in subordinate clauses 69 . According to Abusch (1993), the indexical interpretation of the present tense of a complement clause in (123) violates the ULC because the embedded present tense follows the matrix past tense, thereby surpassing its upper bound 70 . (123) a. John-nun [ Mary-ka LA-ey iss- -ta]-ko malha-yss-ta. John-TOP Mary-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC „John 1 said that Mary was in LA (at that time).‟ b. [ e 1 : (e 1 )  (u)] say ( j, ^[ e 2 : (e 2 )  (e 1 )] & be-in-LA (m,e 2 )], e 1 ) As in (123)-b, the local evaluation time of the complement clause, which is a temporal pronoun R x in our term, is fixed to the matrix event time (e 1 ). If the embedded present tense is a deictic present tense overlapping with the speech time, it turns out that the 69 Originally, Abusch (1993) proposes this constraint to account for the absence of the forward shifting reading of English past under past in a complement clause. 70 Heim showed in her note that the ULC can be understood as a temporal presupposition. 137 embedded present tense should follow the local evaluation time which is the time of a past event from the speech time. Then, the indexical interpretation in question violates the ULC and will be ruled out. It also works for an embedded past tense in a complement clause, as below: (124) a. John-nun [ Mary-ka LA-ey iss-ess-ta]-ko malha-yss-ta. John-TOP Mary-NOM LA-LOC be-PRS-DEC-COMP say-PST-DEC „John 1 said that Mary had been in LA (at that time).‟ b. [ e 1 : (e 1 )  (u)] say ( j, ^[ e 2 : (e 2 )  (e 1 )] & be-in-LA (m,e 2 )], e 1 ) c. *[ e 1 : (e 1 )  (u)] say ( j, ^[ e 2 : (e 2 )  (u)] & be-in-LA (m,e 2 )], e 1 ) When the embedded past tense in a complement clause interpreted indeixcally, we will get an LF in (124)-c which can give us an interpretation pictured in the diagram in (124)- c' where the embedded event follows the matrix event, as both of them preceding the speech time. Since the ULC rules out such an interpretation, (124)-a can have a back shifted reading and a simultaneous reading only. Because a simultaneous reading of an embedded past tense is not an available option for non-Sequence of Tense languages such *c'. John‟s saying Mary‟s being in LA the speech time 138 as Korean, (124)-a will get a back shifted reading only. Abusch‟s (1997) ULC nicely accounts for the absence of the indexical interpretation of Korean present under past in the complement clause in (122)-a, working with the property of intensional verbs that requires such a fixed local evaluation time. And, as there is nothing to force the ULC to be observed in relative or temporal adjunct clauses, an indexical interpretation is not ruled out, as in (122)-b. However, note that Abusch‟s (1997) ULC fails to accounts for the interpretation of a multiple embedding case, as below: (125) John-un [ [ Mary-ka manna-koiss- ]-nun namca-ka John-TOP Mary-NOM meet-PROG-PRS-REL man-NOM LA-ey sal-ass-ta ]-ko malha-yss-ta. LA-in live-PST-DEC-COMP say-PST-DEC Reading #1: „John said that the man whom Mary is dating (at S-time) had lived in LA.‟ Reading #2: „John said the man whom Mary was dating (at that time) had lived in LA. ‟ In (125), the relative clause present tense is embedded by the complement clause past tense, which is in turn embedded by the matrix past tense. Reading #1 shows that the deeply embedded relative clause present can get an indexical interpretation, referring to S-time. Even though Abusch‟s (1997) ULC does not play a role within the innermost 139 relative clause, it is predicted that the constraint should be observed in the intermediate complement clause in the next step of interpretation, blocking the indexical interpretation of the relative present tense. However, the prediction is not born out. After all, Abusch‟s (1997) ULC does not do it job when an embedding layer is added, failing to explain the asymmetry between complement clause and relative clause tenses. On the other hand, the context shifting analysis by Schlenker (1999), Anand (2006) and others is not able to provide an appropriate account for the asymmetry as well. According to their analysis, the indexical shifting optionally takes place by the force of intensional verbs such as „say‟ or „think.‟ Hence, if we assume that the verb „malha-ta (say)‟ in Korean is one of the intensional verbs containing a shifting operator, the absence of the indexical interpretation is easily accounted for because the shifting operator re- writes all of indexical items including a present tense and only the anaphoric interpretation is available for (122)-a; the indexical interpretation of a relative present tense in (122)-b results due to the absence of the shifting operator. But, as Anand (2006) admits, the indexical shifting hypothesis runs into a problem when it deals with the anaphoric interpretation of (122)-b. Given that the anaphoric interpretation is the outcome of the indexical shifting, it is expected that a relative clause which is a non-intensional environment, thereby not triggering the shifting, does not have the anaphoric interpretation in (122)-b. The Korean facts in (122), however, show that it is not the case: (122) has an ambiguous reading, including an anaphoric interpretation. Yet, it might not be a serious problem if one posits that the anaphoric interpretation of a relative clause tense comes from a difference source, rather than from an indexical 140 shifting, as put forth by Kusumoto (1999) 71 . The real problem emerges with the multiple embedding case in (125). Under the indexical shifting analysis, the relative present tense is under the scope of the shifty operator introduced by the saying verb so that it has to be shifted, resulting in an anaphoric interpretation only. But, note that, once again, the indexical interpretation is available for (122)-b. The following example will show the problem clearer. (126) John-un [ [ Mary-ka manna-koiss- ]-nun namca-ka John-TOP Mary-NOM meet-PROG-PRS-REL man-NOM LA-ey sa(l)-n-ta ]-ko malha-yss-ta. LA-LOC live-PRS-DEC-COMP say-PST-DEC Reading #1: „John said that the man whom Mary is dating (at S-time) lived in LA (at that time)‟ Reading #2: „John said the man whom Mary was dating (at that time) live in LA (at that time)‟ The sentence with a multiple embedding structure in (126) is two-way ambiguous, depending on the interpretation of the present tense in the innermost relative clause. The present tense of the intermediate complement clause is unambiguously interpreted as simultaneous to the matrix past tense. Whereas, the innermost present tense of the relative 71 Kusumoto (1999) treats the anaphoric interpretation of a complement clause and relative one in a different way. In Kusumoto (1999), a relative clause, unlike a complement clause, lacks a TP layer so that it is a tensless clause, resulting in an anaphoric interpretation. 141 clause can be simultaneous to S-time or the event time of the intermediate complement clause. If an indexical shifting takes place within the complement clause, it is expected that the present tense of the relative clause should be shifted as well, yielding Reading #2 only. But, once again, (126) has two readings, indicating that the shifting mechanism does not properly handle the asymmetry between complement and relative tenses 72 . 3.3.2. Explanation Hgginbotham (2002, 2006) explains the asymmetry of the English past under past between complement and relative clauses, by positing that an obligatory movement of a temporal element takes place in a complement clause. Specifically, he argues that the first coordinate of the past tense under past should move to Spec of a CP in a complement clause, yielding a simultaneous / anaphoric reading. In a relative clause, the movement takes place optionally. About the same syntactic position, i.e., Spec of a CP, Giorgi (2010) makes a different claim that while the Spec of a CP position of a complement clause is occupied by the speaker‟s temporal location in Double Access Reading (DAR) languages such as English, it is not the case in non-DAR languages such as Chinese. Given that Korean is one of non-DAR languages, it follows from it that the speaker‟s temporal location is not 72 Let us add briefly the discussion on Krazter‟s (1998) zero tense analysis. Setting aside the issue that a so- called Korean zero tense / null tense does not inherit its morphological contents through the agreement relation with the binder unlike the English zero tense, the fact that a zero tense can appear in complement and relative clauses while a real tense version of an indexical tense only occurs in a relative clause remains unexplained. 142 specified in Spec of a CP of a complement clause in Korean. The proposed analysis attempts to combine Higginbotham(2002, 2006) and Giorgi (2010) in order to account for the asymmetry between subordinate tenses in complement and relative clauses in Korean, as illustrated in (122). I take the obligatory R x movement in a complement clause from Higginbotham (2002, 2006), which blocks the specification of the speaker‟s temporal location within a CP. On the other hand, exploiting Giorgi‟s (2010) idea of syntactic indexicality, I propose that the speaker‟s temporal location is specified in Spec of a CP in relative or temporal adjunct clauses has so that it can serve as an antecedent of the R x of those clauses. Let us go into the details of the proposal. 3.3.2.1. The Rx movement versus the Speaker Under the proposed analysis, a complement clause of the verb „malha-ta (say)‟ has the following structural representation. 143 (127) CP C TP -lako R x (RT) T‟ T vP (ET) past/present The basic tree structure above shows that the complement clause of the saying verb is headed by the complimentizer „-lako‟ which selects a TP that takes two temporal elements, i.e., RT and ET. Note that there is no means in the structure to prevent R x from getting an indexical interpretation of the subordinate tense in a complement clause because there are two possible antecedents for R x , E-time of the matrix clause and S-time above the CP. Thus, adopting Higginbotham‟s (2002, 2006) approach, I claim that the R x of a complement clause obligatorily moves to the Spec of a CP position, as shown in the structure below: 144 (128) CP C TP -lako R x (RT) T‟ T vP (ET) past/present The structure in (128) results when we apply Higginbotham‟s (2002, 2006) analysis of the English past tense embedded in a complement clause to the Korean present under past case. Yet, one could soon notice that there is no strong motivation to justify the movement in Higginbotham (2002, 2006) 73 . Hence, in the present work, following Giorgi and Pianesi (2001a) 74 , we argue that the temporal feature carried by the complementizer „-lako‟ has to be checked so that it forces the movement of a temporal 73 The left periphery position can be viewed as the syntactic position where an access to the matrix clause is most possible, which leads to the simultaneous / anaphoric interpretation between the complement and matrix clauses. However, given that the movement renders the R x of the complement clause to get an access to the matrix clause, there is no reason why S-time, the R x of the matrix clause, should not be accessible. Then, R x still can be interpreted indexically. To avoid this problem, another means to rank the possible antecedents for R x , such as Rosenbaum‟s „a minimal distance principle‟ adopted in Stowell (1c 993, 2007) would be needed. 74 According to Giorgi and Pianesi (2001a), T is merged to C and the T and C must agree with each other. 145 pronoun R x to Spec of a CP 75 . It means that the property of a complementizer „-lako‟ induces the movement, resulting in a constraint on a possible antecedent of the R x of a complement clause. The movement creates a CP with the Spec of a CP position already filled with R x , thereby making no room for Giorgi‟s (2010) speaker‟s temporal coordinate to appear in Spec of a CP. In other words, the speaker‟s temporal location cannot be specified in Spec of a CP due to the obligatory movement of R x to Spec of a CP in a complement clause in Korean, as seen again in (129) below: (129) CP R x (RT) C TP -lako t T‟ T vP (ET) past/present It is claimed that in Korean a complement clause has a CP structure that lacks the speaker‟s temporal location, resulting in the absence of an indexical interpretation; the 75 Yet, any morphological features are not required to be checked between R x and C so that a morphological agreement does not arise in Korean. 146 verb „malha-ta (say)‟ takes such a CP structure. In contrary to the complement clause case, the R x movement takes place optionally in relative and temporal adjunct clauses 76 , along the same line of Higginbotham (2002, 2006), which is attributed to the property of the complementizer. Because the head C of a CP in relative or temporal adjunct clauses does not have a strong temporal feature to be checked, it does not require the R x movement to take place obligatorily 77 . When the movement takes place, a relative or adjunct clause will get the same CP structure as in (129); when R x remains in situ, the Spec of a CP is not occupied so that the speaker‟s temporal coordinates can merged into the position, having R x bound by the speaker‟s coordinate, which leads to the indexical interpretation, as shown in the structure below: 76 The examples of Korean relative clauses and adverbial ones are repeated below: (i) a. Relative Clauses: John-nun [ LA-ey sa(l)- -]-nun yeca-lul manna-ss-ta. John-TOP LA-LOC live-PRS-REL woman-ACC meet-PST-DEC Reading #1: „John met a woman who lives in LA (now-the speech time).‟ Reading #2: „John met a woman who lived in LA (at that time).‟ b. [ …. Ry … PRS .. REL ] ….. Rx PAST (ii) a. Adverbial Clauses: [ Mary-ka ttena- ]-ki ttaymuney John-un ul-ess-ta. Mary-NOM leave-PRS-becauseJohn-un John-TOP cry-PST-DEC Reading #1: „Because Mary lives / is leaving, John cried.‟ Reading #2: „Because Mary left (at that time), John cried.‟ b. [ …. Ry … PRS .. ADV ] ….. Rx PAST The sentences in (i) and (ii) are ambiguous between an anaphoric and an indexical interpretation. 77 The lexical meaning of the head of an adjunct clause certainly affects the temporal interpretation of a given sentence, but it has nothing to do with feature checking and will not be discussed further in the present study. 147 (130) CP S C NP TP R x T vP (ET) As in (130), the CP structure of a relative clause that contains R x in situ has an empty Spec of a CP position, which can host the speaker‟s temporal location, i.e., S (-time) above. The dotted line indicates a binding relation between the S-time and R x , yielding an indexical interpretation. Whereas, when it moves to Spec of a CP, the anaphoric interpretation is obtained for in the same way the complement clause tense is interpreted, as in (129). At this point, I have to admit that the CP structure for a complement clause in (129) as well as the one for a relative or adjunct clause in (130) does not give rise to an unambiguous anaphoric interpretation of the subordinate present tense. It is so because there are still two possible antecedents for the R x of the complement clause in the matrix clause-namely the matrix event time and the speech time. Hence, a separate rule or 148 constraint is needed to ensure to pick out the matrix event time as the antecedent of the R x . In order to achieve the goal, as in Stowell (2007), I adopt Rosenbaum‟s (1970) Minimal Distance Principle to account for the choice of the binder of R x in general. The principle states that only a temporal element in a minimum distance from R x can serve as its antecedent. Then, working with the principle, the structure in (129) gives us an anaphoric interpretation only. Because the matrix even time is the temporal element in a minimum distance from R x in (129), it serves as an antecedent for the R x of the complement clause, yielding an anaphoric reading. Whereas, for the indexical interpretation of the relative clause present tense, the speaker‟s temporal location specified in Spec of a CP serves as an antecedent for the in situ R x of the complement clause because it is closer than the matrix event time in (130) 78 . Hence, the indexical interpretation in question does not involve with movement, contra Ogihara (1996) or Stowell (2007). Yet, one would ask why one has to make the system more complicated by introducing the information of the speaker‟s temporal location to a CP. The answer is found in the payoff of the system in dealing with multiple embedding cases, which will be discussed in the following section. 3.3.2.2. The Speaker in multiple embedding The study of the temporal interpretation in a multiple embedding construction is rare 78 For the anaphoric reading of a relative clause tense, the structure in (129) where the R x movement takes place is to be given. 149 because the basic interpretation of the embedding structure with one layer is already complicated and challenging 79 . However, a multiple embedding construction is very commonly used in language and moreover it can be used as an important test case to evaluating theories of tense as we saw in the previous section, because many theories turn out not to work for the complex type of temporal interpretation. First, let us identify the temporal interpretation(s) that a multiple embedding construction can get and see how the proposed system works for various types of multiple embedding. 3.3.2.2.1. Complement under Complement: an anaphoric interpretation only An embedded present tense in a double layer of a complement clause will be exemplified as follows: (131) [ John-i [ Mary-ka LA-ey sa(l)-n-ta]-ko malha-yss-ta]- John-NOM Mary-NOM LA-LOC love-PRS-DEC-COMP say-PST-DEC- -ko Sam-un malha-yss-ta. COMP Sam-TOP say-PST-DEC „Sam said, “John said that Mary lived in LA (at that time).” ‟ Given that only an anaphoric interpretation of the most embedded present tense is 79 Anand (2006) showed the intervention effect on indexical shifting in a complement clause embedded by another complement one. Also, Anand and Hacquard (2008) claim that the indexical interpretation of a present tense under past is allowed by the presence of a future oriented auxiliary „woll‟ that functions as a polarity intervener under which an anti polity item „PRES‟ can be licensed even in the scope of PAST. 150 available and the past tense in an intermediate complement clause receives a back shifted reading, the LF of (131)) will be given as in (132)-a: (132) a. [ e 1 : (e 1 )  (u)] say ( s, ^[ e 2 : (e 2 )  (e 1 )] & say ( j, ^[ e 3 : (e 3 ) ≈ (e 2 )] & be-in-LA (m,e 3 )], e 2 ) , e 1 ) b. [ [ [ E 3 ≈ R 3 ]-ko E 2 < R 2 ]-ko E 1 < R 1 ] The temporal interpretation in (132)-a can be represented in a simplified way as in (132)- b. The arrows in (132)-b express the anaphoric relation between the reference time of each clause and its possible antecedent. The sentence in (131)) is unambiguous because only the anaphoric interpretation is available for the embedded present tenses 80 . Under the present analysis, the R x movement takes place in each subordinate clause, accounting for the anaphoric reading together with the minimum distance principle. 3.3.2.2.2. Complement under Relative: two-way ambiguous Now let us turn to the second type of embedding: a complement tense embedded under a relative clause. A present complement clause under a present relative clause with a past 80 There is no difference in terms of interpretation in an embedded past tense in a double complement clause, except that the innermost tense is a past tense, being an interior relation, as below: (i) a. [ John-i [ Mary-ka LA-ey sal-ass-ta]-ko malha-yss-ta]-ko John-NOM Mary-NOM LA-LOC love-PST-DEC-COMP say-PST-DEC-COMP Sam-un malha-yss-ta. Sam-TOP say-PST-DEC „Sam said, “John said that Mary had lived in LA.” ‟ b. [ e 1 : (e 1 )  (u)] say ( s, ^[ e 2 : (e 2 )  (e 1 )] & say ( j, ^[ e 3 : (e 3 ) < (e 2 )] & be-in-LA (m,e 3 )], e 2 ) , e 1 ) 151 matrix tense has only two options of interpretation since only the present tense in the intermediate relative clause is ambiguous between an anaphoric interpretation and an indexical one, as shown below: (133) [ caki-ka taytongryeng-i- -ra]-ko e 1 cucangha-nu]-n self-NOM president-be-PRS-DEC-COMP insist-PRS-REL yeca 1 -ka LA-ey sal-ass-ni? woman-NOM LA-in live-PST-INTR. Reading #1: „Did the woman who is insisting that she is president live in LA?‟ Reading #2: „Did the woman who was insisting that she was president live in LA?‟ The interpretation of the sentence in (133) varies depending on whether the R x movement takes place in the relative clause. When it takes place, Reading #2 results; when it does not, Reading #1 is obtained 81 ). 81 For other embedded tenses of this construction see the following examples: (i) A past complement tense embedded in a present relative clause: [ caki-ka taytongryeng-i-ess-ra]-ko e 1 cucangha-nu]-n self-NOM president-be-PST-DEC-COMP insist-PRS-REL eyca 1 -ka LA-ey sal-ass-ni? Woman-NOM LA-LOC live-PST-INTR. Reading #1: „Did the woman who is insisting that she was president live in LA?‟ Reading #2: „Did the woman who was insisting that she had been president live in LA?‟ (ii) A present complement tense embedded in a past relative tense [ caki-ka taytongryeng-i- -ra]-ko e 1 cucangha-a]-n self-NOM president-be-PST-DEC-COMP insist-PST-REL eyca 1 -ka LA-ey sal-ass-ni? Woman-NOM LA-loc live-PST-INTR. 152 3.3.2.2.3. Relative under Complement: two-way ambiguous When a relative clause is embedded by a complement clause, two readings are obtained due to the ambiguity brought by the most embedded relative clause: (134) John-un [ [ Mary-ka manna-koiss- ]-nun namca-ka John-TOP Mary-NOM meet-PROG-PRS-REL man-NOM ku il-ul ha-yss-ta ]-ko malha-yss-ta. that thing-ACC do-PST-DEC-COMP say-PST-DEC Reading #1: „John said that the man whom Mary is dating (at the speech time) had done it.‟ Reading #2: „John said, “the man whom Mary was dating (at that time) did it.” ‟ This type of embedding construction presents us an interesting case from a theoretical point of view. In (134), the relative clause tense is embedded by the domain of a verb of saying. In the proposed analysis, there is no place for the speaker‟s temporal location inside the complement clause due to the R x movement. Because the complement clause tense has only one anaphoric reading where it is interpreted simultaneous to the matrix past tense, the relative clause tense embedded by it should have a way to access to the speaker‟s temporal location in order to get an indexical interpretation, if we assume the minimum distance principle. The speaker‟s temporal location introduced at the Spec of a Reading #1: „Did the woman who insisted that she was president (at that time) live in LA?‟ Reading #2: „Did the woman who had insisted her being president live in LA?‟ 153 CP position inside the relative clause does the job. By allowing to introduce the speaker‟s information at every clause level, the proposed analysis is able to account for the indexical interpretation of tenses in non-complement clauses that seems to be given freely no matter how deeply embedded it is in the structure. 3.3.2.2.4. Relative under Relative: three-way ambiguous The case of relative clauses embedded relative ones is a bit complicated because the tense in each relative clause can be ambiguous, yielding 4 possible readings as below: (135) Na-nun [ [ New York shicang-kwa sakwi-nu]-n I-TOP New York mayer-with date-PRS-REL daughter ttal-ul e 1 calangha- ]-n yeca 1 -lul manna-ess-ta daughter brag about-PRS-REL woman meet-PST-DEC Reading #1: Ind. PRS + Ind.PRS 82 „I met a woman who is bragging about her daughter who is dating the New York mayor.‟ Reading #2: Ind. PRS + Ana. PRS „I met a woman who bragged about her daughter who is dating the New York mayor.‟ Reading #3: Ana. PRS + Ind. PRS (=Reading #1) „I met a woman who is bragging about her daughter who 82 Ind.PRS=Indexical Present tense; Ana. PRS=Anaphoric Present tense. 154 is dating the New York mayor.‟ Reading #4: Ana PRS + Ana. PRS „I met a woman who bragged about her daughter (at that time) who dated the New York mayor (at that time).‟ Because a present relative tense in both the innermost clause and intermediate one can be interpreted indexically and anaphorically, we get four combinations among which Reading #1 and Reading #2 turn out to be identical temporally so that three readings are available for the sentence in (135). Once again, at every CP layer of the relative clauses in (135), the speaker‟s temporal location is specified unless the R x movement takes place as. 3.3.2.2.5. Complement under Relative under Complement: two-way ambiguous The case of triple embedding looks very complicated, but such an example in (136) has only two readings depending on whether the intermediate relative clause has an indexical or anaphoric interpretation. (136) John 1 -un [ pro 1 [[ caki-ka taytongryeng-i- -ra]-ko e 2 cucangha-nu]-n John-TOP self-NOM president-be-PRS-DEC-COMP insist-PRS-REL yeca 2 -lul manna-ss-ta]-ko malha-yss-ta woman-ACC meet-PST-DEC-COMP say-PST-DEC 155 Reading #1: „John said “I met a woman 1 who is insisting that she 1 is president.” ‟ Reading #2: „John said “I met a woman 1 who was insisting that she 1 was president.” ‟ Once again, what is interesting in (136) is that when the intermediate relative clause is interpreted indexically, the innermost complement clause tense receives the same indexical interpretation through the simultaneous temporal relation that the complement tense holds, even though it cannot receive an indexical interpretation by itself. The two complement clauses have the R x movement, blocking the speaker‟s temporal location from being specified in Spec of their CPs. However, the relative clause in between hosts the speaker‟s temporal location in its CP level so that the indexical interpretation of the relative present tense is made possible. 3.3.2.2.6. Relative under Complement under Relative: four-way ambiguous When the embedding of two relative clauses is involved, things get more complicated, giving us four possible readings for (137): (137) John-un [ [ Mary-ka manna-ko iss- ]-nun namca-ka John-TOP Mary-NOM meet-PROG-PRS-REL man-NOM ku il-ul ha-yss-ta ]-ko e 1 cucangha-nu]-n that thing-ACC do-PST-DEC-COMP insist-PRS-REL yeca 1 –lul mana-ss-ta. 156 woman-ACC meet-PST-DEC Reading #1: Ind. PRS + Ana.PST 83 + Ind.PRS „John met a woman who is insisting that the guy whom Mary is dating did it.‟ Reading #2: Ind. PRS + Ana.PST + Ana.PRS „John met a woman who was insisting that the guy whom Mary is dating had done it.‟ Reading #3: Ana. PRS + Ana.PST + Ana.PRS „John met a woman who was insisting that the guy whom Mary dated had done it.‟ Reading #4: Ana. PRS + Ana.PST + Ind.PRS „John met a woman who is insisting that the guy whom Mary was dating did it.‟ Because both of the tenses in the two relative clauses in (137) can be interpreted independently of other higher tenses or relatively to them, the four readings are obtained. In the proposed analysis, the embedded tense in a relative clause receives an indexical interpretation in situ without any movement because the temporal pronoun R 2 or R 4 is bound by the Speaker‟s coordinates (SC) introduced to every clause. After all, the relative present tense is interpreted independently. Hence, the number of a layer of embedding a relative clause tense involved does not matter as long as it receives an 83 By the term „Ana. PST’ I mean a back shifted reading of an embedded past tense. It is anaphoric in the sense that the temporal pro of the past tense is anaphoric to the higher event time. 157 indexical interpretation. It almost freely receives the indexical reading even though it is located in a triple or quandary embedding construction because the speaker‟s coordinates can be introduced to every sentence. Finally let us note that this pattern found in relative clause tenses also applies to adverbial clause tenses even though we do not have space for the discussion here. 3.4. Cross-linguistic implications The analysis of embedded tenses and their interaction with other temporal elements in Korean is the main Topic of this dissertation. Here, one can raise a fair question of whether the analysis can be extended to account for tense in other languages of the same type, e.g. Japanese, or of another type, e.g. English. I do not attempt to provide a full- fledged account of cross-linguistic variations in temporal interpretation but do show what implication the analysis of tense in Korean could bear for it from a theoretical perspective. There have been proposed several analyses to account for cross-linguistic differences in the interpretation of tense. Ogihara (1996) is considered as the first serious attempt to find the underlying mechanism working. It is a well-known fact that the English past tense, when embedded, can have a simultaneous interpretation to the matrix past tense, which is called a Sequence of Tense phenomenon; but the same reading is carried by the embedded present tense in Japanese, as shown below in (153): 158 (138) a. Taroo said that Hanako was in Seattle.(at that time) b. Taroo-wa [ Hanako-ga Siatoru-ni i- -ru]-to it-ta. Taroo-TOP Hanako-NOM Seattle-in be-DEC-that say-PST -DEC „Tarro said that Hanako was in Seattle (at that time).‟ Ogihara (1997) attributed the surface difference between English in (138)-a and Japanese in (138)-b to the presence or absence of the tense deletion rule: while English – an SOT language – has the tense deletion rule, Japanese – a non-SOT languages – lacks it. Ogihara accounts for the fact that (138)-a and (138)-b have the same meaning despite the difference in their form in terms of the parametric difference between the two languages. However, Kusumoto (1999) pointed out that Ogihara‟s (1997) SOT versus non- SOT distinction does not suffice to explain a cross linguistic variation when Polish or Russian cases are also considered. While Russian and Polish behave like Japanese, which is a non-SOT language, in a complement clause, they pattern with English, which is an SOT language. Specifically, a present tense embedded in a complement clause gets a simultaneous reading to a matrix tense, as in Japanese. In contrast, a present tense in a relative clause cannot receive such a reading, but only an indexical interpretation is available for it, as in English, Kusumoto (1999) explains the difference between Japanese and Polish / Russian in terms of a structural difference of relative clauses in those languages: while Polish / Russian relative clauses have a full-fledged CP, Japanese relative clauses have an IP-like tenseless construction. In this paper, we add Korean to the system of comparison, which makes 159 Kusumoto‟s (1999) solution less promising. This is because it is hard to claim that Korean, which shares with Japanese syntactic and semantic properties in the interpretation of relative clause tenses, has a tenseless construction but is not a CP. Unlike Japanese with no relative clause marker, Korean relative clauses have overt suffixes carrying a temporal meaning corresponding to tense morphology in complement clauses. Hence, treating a relative clause as an IP lacking temporal information by itself is not suitable for Korean relative clause cases. Thus, we need another solution for Russian / Polish data. Why an anaphoric interpretation of present under past is not restricted in relative clauses in Russian / Polish will be answered with respect to the interaction between a relative clause tense and the associated aspectual meaning. The pattern can be summarized in the following table: (139) Table 6: The pattern in complement clauses Present under past Past under past Present under future English Double Access Reading: Simultaneous & Indexical Rs Simultaneous R or Back shifted R DAR? Korean / Japanese Simultaneous R Back shifted R Simultaneous R Polish / Russian Simultaneous R Back shifted R Simultaneous R 160 (140) Table 7: The pattern in relative clauses Present under past Past under past Present under future English Indexical R IR (or SR/BSR?) SR or IR Korean / Japanese SR or IR IR (or BSR?) SR or IR Polish / Russian Indexical R IR (or BSR?) SR or IR The rough idea taken in this dissertation in order to account for the pattern in complement and relative clauses in the three language groups is to make an analogy between a tense morpheme and a long distance anaphora. In other words, a tense morpheme can be viewed as a long distance temporal anaphor, more precisely, a present tense is to be considered as a long distance anaphor. In English, there is no long distance anaphor and as such a present tense cannot be anaphoric to a past tense in any case. In Korean and Japanese with long distance anaphors, a present tense can be anaphoric to a past tense both in complement and relative clauses. As a language group in the middle, Russian / Polish has a long distance anaphor, but only in a complement clause. Thus, a present tense in a complement clause can be anaphoric to a past matrix tense; however, such a long distance biding does not apply across a relative clause in those language so that a present tense cannot be anaphoric to a past tense in relative clauses 84 . The correlation pattern is given in the table below: 84 I owe Barry Schein the account for the three way pattern. 161 (141) Table 8: The present tense & the long distance anaphor (LDA) Complement Clauses Relative Clauses English *LDA-*SR *LDA-*SR Korean / Japanese LDA-SR LDA-SR Polish / Russian LDA-SR *LDA-*SR The cross linguistic comparison and the explanation of the pattern need further research. 3.5. Summary In this chapter, I presented the skeleton of the proposed analysis and showed that how the system works for the temporal interpretation across complement and non-complement clauses. Furthermore, it is elaborated to the point that the proposed analysis can better handle more complex embedding structure such as double or triple embedding constructions. Many specific and carefully chosen examples are given to complete the picture proposed in the analysis. What is most important ingredient of the proposal is a temporal pronoun R x that is an argument of a binary relation of tense. It goes into a binding relation to higher temporal elements like a higher event time. In fact, R x needs an antecedent to feed its temporal content so that the binding relation is necessary for the interpretation of tense, it is claimed in this paper. The main difference between 162 complement clause tenses and relative clause ones is accounted for by the presence / absence of the obligatory movement of R x to Spec of a CP in a complement clause, according to Higginbotham‟s (2002a, 2006) analysis. The motivation of the movement is accounted for by means of the feature checking mechanism of the complimentizer „-lako.‟ The nature of the speaker‟s coordinates, to what extent and in what way it will be realized syntactically or just contextually, should be the part to be more elaborated later on. 163 Chapter 4. Tense and Indexicality In the previous chapter, I proposed the non-deictic analysis of tense in Korean. Provided that tense is treated separable from temporal indexicality that always involves the deictic center of the speaker‟s „now,‟ it is predicted that tense does not need to behave in accordance with other indexical items like indexical pronouns or adverbials contra Schlenker‟s (1999) claim. Yet, there should be some pattern of interaction between tense and indexical items even though the interpretation of tense takes a different route from that of indexical items. As one of a pattern of the interaction, a restriction on a temporal adverbial in a complement clause will be presented and the source of the restriction will be discussed in the first sub section. And, the issue of a Double Access Reading where an embedded present tense is always interpreted in relation to the speech time will addressed and the claim will be put forth that Korean does not have an English type Double Access Reading, but it can have a Double Access Reading effect. 4.1. Subordinate tenses and indexical temporal adverbials The way of dealing with temporal adverbials in an embedded environment within the theories of tense remains unexplored even though it has been touched by several researchers such as Dowty (1979) and Klein (1994). It seems nearly impossible to embrace within a theoretical account the lexical variety and syntactic flexibility of temporal adverbials. Even a lot of insightful accounts of embedded tenses, proposed by Abusch (1997), Krazter (1998), Ogihara (1996), von Stechow (1995), Schlenker (2000), 164 Higginbotham (2002a), and Stowell (1997, 2007), could not go beyond enough to provide an account of how temporal indexical adverbials interact with embedded tenses, and how they are constrained in an embedded context across languages. Thus, in this chapter, limiting ourselves to indexical temporal adverbials such as yesterday or tomorrow, it will be investigated under which syntactic or semantic conditions temporal adverbials are restricted in the course of the interpretation of embedded tenses. 4.1.1. Identifying the Restriction across Languages Like other Sequence of Tense (SOT) languages, English exhibits such characteristics that an embedded past tense is ambiguous between a back-shifted reading and a simultaneous one, as below: (142) John said that Mary was in LA. Past 1 Past 2 a. Back-shifted Reading: „John said, “Mary was in LA.” ‟ b. Simultaneous Reading: „John said, “Mary is in LA.” ‟ The past tense on was in the embedded clause of indirect speech can be interpreted as prior to the matrix past tense on said as indicated by the back shifted reading in (142)-a, or as simultaneous to it as shown in the reading (142)-b. Let us consider what happens when we add an indexical temporal adverbial yesterday to the embedded clause of (142). Note that it makes a difference because it 165 removes the second reading of (142)-b 2 , as we can see in (143): (143) John said [that Mary was in LA yesterday.] 1 a. Back-shifted Reading: „John said, “Mary was in LA yesterday.” ‟ b. Simultaneous Reading: „John said, “*Mary is in LA yesterday.” ‟ When yesterday is inserted, the simultaneous reading such as (143)-b disappears or becomes very hard to get. The intuition on the English data is supported by Dickey‟s (2001) questionnaire study conducted on 52 English native speaker participants (pp. 196~199). Participants were given the following four sentences 3 and were asked whether they could get a simultaneous reading for them or not. (144) a. At the party, John admitted that he was drunk. b. John admitted that at the party, he was drunk. c. John admitted that he was drunk at the party. d. John admitted that he was drunk. 166 (145) Table 9: Adverb Position and Temporal interpretation (from Dickey 2001) Adverb position Ambiguous items Eventive controls 4 (3a) Preposed 0.580 0.113 (3b) Post-COMP 0.163 0.036 (3c) Sentence-final 0.120 0.031 (3d) No adverb 0.635 0.059 As we can see in the table above, for no-adverb or matrix adverb environments such as (3a) and (3d), a simultaneous reading is slightly preferred: 0.580 for (3a) and 0.635 for (3d). However, when the adverbs appear in the embedded clauses, the preference drops significantly, as in (3b: 0.163) and (3c: 0.120), indicating that a simultaneous reading is hard to get with embedded temporal adverbs. It is concluded in other words that temporal adverbials are restricted in the simultaneous reading of an embedded clause of indirect speech. The same restriction is found in non Sequence of Tense (non-SOT) languages such as Korean and Japanese. In Korean and Japanese, the simultaneous reading under consideration is obtained in a different way, as follows: (146) Korean: John-un [ Mary-ka LA-ey iss- -ta]-ko malha-ess-ta. John-TOP Mary-NOM LA-LOC be-PRS-DEC-that say-PST-DEC Simultaneous Reading: „John said, “Mary is in LA.” ‟ 167 (147) Japanese: John-wa [ Mary-ga LA-ni iru- -]-to yu-tta. John-TOP Mary-NOM LA-LOC be-PRS-that say-PST.DEC Simultaneous Reading: „John said, “Mary is in LA.” ‟ The null tenses embedded in indirect speech in (146) and (147) are to be interpreted as simultaneous to the matrix past tense. Unlike the English past tense in (142), embedded past tenses in Korean / Japanese do not show an ambiguity so that they are responsible only for a back shifted reading while the null tenses in (146) and (147) only for a simultaneous reading. In spite of the difference in their tense morphology, however, the same restriction found in the English case in (143) also applies to Korean and Japanese. (148) Korean: *John-un [ Mary-ka ecey LA-ey iss- -ta]-ko malha-ess-ta. John-TOP Mary-NOM yesterday LA-LOC be-PRS-DEC-that say-PST-DEC Simultaneous Reading: „John said, “ *Mary is in LA yesterday.” ‟ (149) Japanese: *John-wa [ Mary-ga kinoo LA-ni iru- ]-to yu-tta. John-TOP M-NOM yesterday LA-LOC be-PRS-that say-PST.DEC Simultaneous Reading: „John said, “*Mary is in LA yesterday.” 168 The sentences (146) and (147) with a simultaneous reading only become ungrammatical / unacceptable when a past time denoting adverbial such as ecey (Kor. „yesterday‟) or kinoo (Jap. „yesterday‟) is added as in (148) an (149). One might be attempted to say that the embedded tense in (148) takes the same morphological form with a present tense so that it cannot go with the past time denoting adverb, ecey (yesterday), thereby resulting in the ungrammaticality of (148); the same thing applies to (149). But, the embedded tenses in (148) and (149) are not a deictic present tense which mismatches with past time denoting adverbs, but an anaphoric tense with a null form. With deictic tenses, there is no way to get the simultaneous readings of (146) and (147) because the matrix past tense and embedded present tense do not agree, that is, the two tenses cannot be interpreted as simultaneous. Therefore, the embedded tenses of indirect speech in (148) and (149) are an anaphoric tense which is responsible for the simultaneous reading under consideration. The cross-linguistic restriction discussed so far can be illustrated as follows: (150) [[…*yesterday/ *ecey/ *kinoo…Tense i … ] …Past i …]: Eng., Kor., Jap. (150) says that in the three languages certain temporal adverbials are restricted in the simultaneous reading between a matrix tense and embedded one of indirect speech, indicated by the same index on the tenses. The question to arise is what restricts them. More specifically, the past time denoting temporal adverbials are not allowed to appear with the embedded tense simultaneous to the matrix past tense. If the embedded anaphoric tense ends up with being interpreted as past through the anaphoricity, why are 169 past time denoting temporal adverbials in (150) not able to occur with them? The proposed analysis accounts for it by placing a restriction on a temporal pronoun in a complement clause which moves to Spec of a CP. This restriction also applies to the anaphoric interpretation of a relative clause tense because the movement of a temporal pronoun also takes place. 4.1.2. The restriction on a temporal pronoun It is in this paper posited that a RT is a temporal pronoun so that it can be interpreted anaphorically (bound) or deictically (free) as other pronouns. Kratzer (1998), along the same line of Partee (1973), argues for the parallelism between tenses and pronouns. Taking a slight different position, it is assumed in this paper that there is no direct parallelism between them but the parallelism is to be established between a RT and a pronoun. With this added to the previous assumptions, embedded tenses can be given as follows: (151) RT a … ET b … …[ S … RT a/b … ET c … ] The embedded reference time RT a/b can be interpreted bound or free, depending which temporal antecedent it takes. As a temporal pronoun, it builds an anaphoric relation to its possible antecedents, namely RT a or RT b . But, the embedded event time ET c is just referential that it does not connect itself to preceding temporal elements. So, even though the embedded event time ET c happens to overlap with the matrix event time ET b , tenses 170 in natural languages do not express a direct relation between the two event times. Then, the LFs of two readings of the sentence in (142) will be represented as below: (152) John said that Mary was in LA a. Back Shifted Reading: [ e 1 : (e 1 )  RT 0 ] say  (j, ^[ e 2 : (e 2 )<RT 1 ] be-in-LA (m, e 2 ), e 1 ) b. Simultaneous Reading: [ e 1 : (e 1 )  RT 0 ] say  (j, ^[ e 1 : (e 2 ) RT 1 ] be-in-LA (m, e 2 ), e 1 ) As we can see above, the back shifted reading in (152)-a differs from the simultaneous reading in (152)-b with respect to the way the embedded event time is related to the bound reference time. Following Dowty (1979), non-anaphoric temporal adverbials are regarded to serve as a restriction to an event time, as represented below: (153) John said [that Mary was in LA yesterday.] a. Back shifted Reading: RT a < ET b … [ S … RT b < ET c …yesterday ] [ e 1 : (e 1 )  RT 0 ] say  (j, ^[ e 2 : (e 2 )<RT 0 & (e 2 )  yesterday ] be-in-LA (m, e 2 ),e 1 ) b. Simultaneous Reading: *RT a < ET b … [ S … RT b  ET c …yesterday ] [ e 1 : (e 1 )  RT 0 ] say  (j, ^[ e 1 : (e 2 ) RT 1 & (e 2 )  yesterday] be-in-LA (m, e 2 ), e 1 ) 171 It means that most of temporal adverbials except for an anaphoric adverbial such as “at that time” are an event time modifier, and does not affect a reference time at all. Thus, as we can see in the back shifted reading of (153)-a, the bound reference time remains intact. Along with this, let us take the following principle on a reference time. (154) A principle of a reference time (Hyuna Kim, 2008) A bound reference time RT i cannot be altered or modified. The principle above says that a reference time, when bound, is fixed so that it cannot be altered or assigned a new value for it. Given that temporal adverbials such as “yesterday” do not affect a bound reference time, it is expected that temporal adverbials do not bring any trouble by violating the principle given in (154), which turns out to be attested in (153)-a. Then, the initial question arises again: where does the restriction in question, given in (153)-b, come from? 4.1.3. The restriction explained In short, the answer to the question above is due to the characteristics of an overlapping relation expressed by the embedded anaphoric tense. In (155)-a, the temporal adverbial yesterday modifies the embedded event time and does not affect a reference time so that there is nothing wrong with the interpretation of the adverbial and tense in the LF. However, as (155)-b reveals, because through the overlapping relation of the embedded anaphoric tense, the second event time and the bound reference time fall into the same 172 time interval, the temporal modification of the former by yesterday ends up with applying to the latter. (155) John said [that Mary was in LA yesterday.]: Simultaneous Reading a. [ e 1 : (e 1 )  RT 0 ] say  (j, ^[ e 1 : (e 2 ) RT 1 & (e 2 )  yesterday] be-in-LA (m, e 2 ), e 1 ) Event Time Modification b. #[ e 1 : (e 1 )  RT 1 ] say  (j, ^[ e 2 : (e 2 ) RT 1 & RT 1  yesterday ] be-in-LA (m, e 2 ),e 1 ) Reference Time Modification In other words, even though the temporal modification by „yesterday‟ only targets the event time (e 2 ) in (155)-b, the adverb eventually comes to modify the bound reference time RT 1 as well, as the substitution of (e 2 ) by RT 1 indicates, because the event and reference time occupy the same time interval. Therefore, the bound reference time is altered by the adverbial modification in (155)-b, violating the principle of a reference time in (154), which causes the restriction in question. This account goes for the cross- linguistic restriction illustrated in Section 2. 173 4.2. The Double Access Reading effect in Korean 4.2.1. The alleged Double Access Reading English present under past sentences have a peculiar reading of „Double Accessibility,‟ which is illustrated in the following Abusch‟s examples (1991): (156) a. John believed that Mary is pregnant. b. John said that Mary lives in Chelsea. The basic intuition is that the embedded present tense in (156)-a or (156)-b is not only evaluated with the Utterance time, but also involved with the time of the believing or saying event of the matrix clause, as Comrie (1985) and Smith (1978) already pointed out. More specifically, in (156)-a, the event of Mary’ s being pregnant holds at two time intervals, the time of John’ s believing and the Utterance Time, yielding the peculiar reading where Mary was pregnant at a certain past time when John had such a belief and her pregnancy continues up to the Utterance time. So the term a „Double Access Reading (DAR hereafter)‟ well represents the characteristics of English present under past. Note that the DAR of (156)-a is obligatory in the sense that (156)-a would turn out to be ill- formed or uninterpretable if those sentences do not get the DARs: if the embedded clauses in (156)-a is only about the present / Utterance time, it cannot be the belief which „John‟ had in the past time, without a help of an auxiliary verb „would.‟ The question to arise is whether the same phenomenon can be found in other language group such as Japanese and Korean. Ogihara (1996) affirmed by providing the 174 Japanese data below: (157) Taroo-wa konoo Hanako-ga ima Taroo-TOP yesterday Hanako-NOM now Tookyoo-ni i-ru-to it-ta-yo. Tokyo-LOC be-PRES-that say-PAST ending Reading#1: „Yesterday Taro said that Hanako was in Tokyo at that time‟ Reading#2: „Yesterday Taro said that Hanako was in Tokyo at that time and (the speaker assumes that) she is still in Tokyo at the utterance time‟ According to Ogihara (1996), the sentence (157) has two readings, one of which, Reading #2, is the so-called DAR. Ogihara (1996) claims that Japanese has an optional DAR while English has an obligatory DAR because the Japanese present tense is not inherently deictic unlike English. Note that Korean shows a very similar pattern as Japanese: (158) John-un [ Mary-ka cikum LA-ro oh-koiss- -ta]-ko John-TOP Mary-NOM now LA-to come-PRG-PST-DEC-COMP malhay-ss-ta say-PST-DEC Reading #1: „John said, “Mary is coming to LA.” ‟ =„John said that Mary was coming to LA.‟ 175 Reading #2: „John said that Mary was coming to LA at that time and (the speaker assumes that) she is still coming at the utterance time.‟ Reading #2 of (158) seems to be associated with two time points. Then, can we say that Korean has an optional DAR as Japanese? Our short answer is NO. In this paper, we will cast a question to the status of the optional DAR and approach it from a different perspective. What has been assumed in the line of study by Abusch (1991) and Ogihara (1996) is that a DAR is a semantic phenomenon found across languages. However, it is claimed in the present paper that while a DAR is a semantic phenomenon in English, it is not in Korean. In other words, we propose that there is no optional DAR in Korean and the so-called optional DAR should be attributed to a pragmatic effect: the double accessibility meaning in question is obtained by a pragmatic adjustment triggered by the presence of temporally conflicting indexicals. 4.2.2. Clearing out the confusion One might be attempted to say that the Korean example in (158) has a kind of DAR because it could correspond to the English DAR in (156). In other to clear out the confusion, let us start with a sentence with no temporal adverbs as follows: (159) John-un [ Mary-ka LA-ro oh-koiss- -ta]-ko malhay-ss-ta John-TOP Mary-NOM LA-to come-PRG-PST-DEC-COMP say-PST-DEC 176 Reading #1: „John said, “Mary is coming to LA.” ‟ =„John said that Mary was coming to LA.‟ Reading #2: „John said that Mary was coming to LA at that time and (the speaker assumes that) she is still coming at the utterance time.‟ The sentence (159) with no temporal adverbs also gets Reading #2, an alleged DAR. What becomes clearer without temporal adverbs is that Reading #1 is much more preferred to Reading #2 and also Reading #2 is inclusive of Reading #1. Let us see this from a different perspective such that Reading #1 is the only available (semantic) reading for (159) and we can get Reading #2, which is in fact not a separate reading, when some inferences are added to it, as given below: (160) John-un [Mary-ka LA-ro oh-koiss- -ta]-ko malhay-ss-ta. John-TOP Mary-NOM LA-to come-PRG-PST-DEC-COMP say-PST-DEC Reading #1: „John said, “Mary is coming to LA.” ‟ Possibilities: (i) The event of Mary‟s coming to LA might have been going on up to the speech time, (ii) Or, it might have terminated before the utterance time. With (160), we can think of two logical possibilities (i) and (ii) about the event said by „John:‟ Mary‟s coming event might go on or not at the Utterance time. When the inference based on the logical possibility (i) is taken by the speaker so that it is assumed 177 that Mary‟s coming is still going on at the utterance time, we will get Reading #2, the DAR-ish reading. Especially, when there is a certain context supporting the inference such that the Speaker knows that John said it 1 minute ago, the DAR-ish reading becomes stronger. Yet, in this case, weak or strong, Reading #2 is not a separate reading from Reading #1. We get it only when the pragmatic inference in (i) is added to Reading #1. That is why we do not regard Reading #2 as a DAR in a semantic sense. The same story goes for the English sentence in (161): (161) John said that Mary was coming to LA. R #1: John said that Mary was coming to LA at that time. Possibilities: (i) The event of Mary‟s coming to LA might have been going on up to the speech time, (ii) Or, it might have terminated before the utterance time. Let us ignore another back-shifted reading available for (161) and focus on the Sequence of Tense reading, given in R#1, where the embedded past tense is interpreted as anaphoric to the matrix past tense. One might think that the logical possibility (i) naturally follows the semantic meaning of (161), R#1, given a certain context available for the Speaker and Hearer which is more compatible with the possibility / inference. However, note that a DAR is not involved with the past under past sentence in (161), unlike the present under past sentence in (156). It means that a DAR should be distinguished from a DAR-ish meaning that is obtained by an added pragmatic inference 178 as given in Possibility (i). Once again, the logical possibility in (161)-i is not a part of the semantic meaning of (161), but should be considered as a pragmatic inference involved with the assertion of (161). The parallel pattern is found in (160): the embedded present under past is anaphoric to the matrix sentence as in (161) and when a certain context enhances the inference stemming from the logical possibility in (160)-i, a DAR-ish reading become available. Given this, it follows from it that the Korean example in (160) has no DAR; only the DAR-ish reading is available. Now, let us go back to the Korean example with temporal adverbs, repeated in (162) below: (162) John-un [ Mary-ka cikum LA-ro oh-koiss- -ta]-ko John-TOP Mary-NOM now LA-to come-PRG-PST-DEC-COMP malhay-ss-ta. say-PST-DEC Reading #1: „John said, “Mary is coming to LA.” ‟ Reading #2: „John said that Mary was coming to LA at that time and (the speaker assumes that) she is still coming at the utterance time.‟ There should be more story to be provided to account for (162) because a DAR-ish reading seems to be rather enhanced or forced when a temporal adverb „cikum‟ is present. So, the real research question of this paper is why a DAR-ish reading or the 179 pragmatic inference associated with the Utterance time is strengthened when a temporal adverb appears. We will see whether the movement analysis, the traditional account for a DAR in English and Japanese can provide an explanation for it in the next section. 4.2.3. A movement analysis of a DAR-ish reading In Abusch (1991, 1997), Double Accessibility is accounted for in terms of a temporal de re and her Upper Limit Constraint. The access to the Utterance time of the embedded present tense in (156) is achieved by a „Scoping out‟ mechanism, as shown in (163): (163) John PAST 1 say [ s PRES 3  3  2 [ s Mary t 3 be pregnant ]] Abusch (1991, 1997) moves the present tense, a temporal de re, out of the intensional domain of  2 to the position where the present tense can be evaluated with respect to the Utterance time, t 0 . As for the other access to the past time of John‟s saying event, Abusch (1997)‟s „Upper Limit Constraints‟ do its work: the time denotation of the trace of Pres 3 , t 3 inside the intensional domain cannot exceed the time of John’ s saying, the upper limit. Hence, the denotation of present tense also has to include the time interval including „the past time of John‟s saying‟ through the interpretation of the trace remaining in the intensional domain, resulting in a DAR. Following a similar line of Abusch (1991, 1997), Ogihara (1996) proposes another 180 movement analysis of a DAR. Given his assumption that Utterance time oriented indexicals cannot appear in the translation of the intensional argument of an attitude verb (Ogihara 1996, p 211), he claims (i) in English, an absolute present tense, an indexical element in nature, obligatorily moves out of an intensional domain at LF. Whereas, (ii) in Japanese, a relative present tense, not necessarily tied to the speech time, optionally moves out of an intensioal domain at LF. In his analysis, the double accessibility is accounted for as follows: the access to the Utterance time is obtained by a tense movement out of the intensional domain as below: (164) [ CP PRES 2 [ S John PAST 1 say t 2 [ CP that [ S Mary t 1 be pregnant ]]]] A cyclic movement to the empty Comp position, leaving a trace, t 1 and t 2 at LF makes PRES 2 /t 2 accessible to the Utterance time, as in (164). As for the access to the time of John’ s saying, by stipulation, t 2 , a tense that moves out of an intensional domain, obtains a new index distinct from the initial one t 1 ; t 1 is accessible to the time of John‟s saying. These tense movement analyses of a DAR run into several problems when we try to deal with Japanese or Korean DAR examples, given in (156) and (160). First of all, the movement analyses are based on the assumption that the DAR-ish reading of (156) and (160) is a syntactic phenomenon which is to be accounted for by a LF movement. However, as we have found in (159) through (160), the DAR-ish reading in question is a 181 byproduct of a pragmatic inference so that there is no need of any syntactic or semantic amendment. Thus, Ogihara (1996)‟s movement analysis would be able to give an explanation for the English DAR, but not for the Korean DAR-ish reading. Also, the tense movement analyses fail to account for the case where a tense movement should be involved but no DAR is found. The example in (165) below shows that tense movement does not necessarily involve with a DAR: (165) John said that this guy who is now sitting in front of you killed the woman at the bar. The deeply embedded present tense inside the relative clause has to move out up to the matrix tense so that it could get an access to the Utterance time, with which it is to be evaluated. However, the tense movement does not result in a DAR for (165). It seems that the present tense just freely moves out to the most outer clause, violating Abusch (1991, 1997)‟s ULC; the present tense does not hold at two time points, but it is associated with a single time point, that is, the Utterance time. Thus, the movement analyses turn out to run into a problem in dealing with this case. 4.2.4. The Proposed Analysis 4.2.4.1. A semantic conflict with indexicals Let us recall that we cleared out the confusion about the DAR-ish reading in Korean for (5), and concluded that Korean has no Double Access phenomenon found in the English 182 example in (156) and that the DAR-ish reading available for (160) is obtained when a certain pragmatic inference is added to the meaning of (160). Now, we want to go back to the Korean example with temporal adverbs, repeated in (166) below: (166) John-un [ Mary-ka cikum LA-ro oh-koiss- -ta]-ko John-TOP Mary-NOM now LA-to come-PRG-PST-DEC-COMP malhay-ss-ta say-PST-DEC Reading #1: „John said, “Mary is coming to LA now.” ‟ =„John said that Mary was coming to LA at that time.‟ Reading #2: „John said that Mary was coming to LA at that time and (the speaker assumes that) she is still coming at the utterance time.‟ (166) gets a DAR-ish reading as in R#2, but it is not obtained through the same mechanism in (160). When an indexical item like „cikum‟ is present inside indirect speech of (166), the whole sentence becomes uninterpretable. Because „cikum‟ refers to the Utterance time, it cannot modify the embedded tense which is anaphoric to the matrix past tense. The ungrammaticality of the English sentence in (167) could correspond to it: (167) *John said that Mary was coming to LA now. One the other hand, if we assume that both of the adverb „cikum‟ and the embedded tense 183 are deictic, being associated with the Utterance time so that there is no conflict between the two, the indirect speech interpretation of (166) still faces an interpretation problem. It is so because „John‟ cannot say in a past time about his future, using a present tense in Korean. Therefore, there should be more story to be provided to account for (166) because a DAR-ish reading is obtained even though the configuration of (166) seems to be uninterpretable, having a semantic conflict, as shown below: (168) Indirect Speech: * …[anaphoric present… cikum(indexical present)….]-ko P AST... We find the solution to the conflict in its pragmatics, which will be elaborated in the following section. 4.2.4.2. A DAR effect as a pragmatic resolution Given that the indeixical adverbial „cikum‟ in (166) cause a conflict with the embedded tense in their semantic interpretation, and that notwithstanding (166) gets a DAR-ish reading, the question would be what resolves the conflict and makes (166)get a DAR-ish reading. We propose a pragmatic resolution rather than a semantic repair. As for the pragmatic resolution, there would be two ways to solve the problem. One would be to make a change on the embedded tense, and the other on temporal adverbs. We take the second way and apply a pragmatic resolution to the meaning of temporal adverbs. And, we will show McCord‟s Extended Now(EXN) theory can be used to account for the 184 accommodation. (169) McCord (1978)‟s pragmatic Extended Now theory a. English Preterite: a marker of prior events which is concluded and separate from the overall period of the present, „Extended now.‟ [+then] b. English Perfect: a marker of prior events which is included within „Extended now.‟ [-then] According to McCord (1978), the difference between the two with respect to their relation to the „Extended Now‟ comes from different pragmatic inferences associated with them, which unfortunately is not explicitly presented in McCord (1978). Adopting his idea, von Stechow (1999) and many others developed a semantic Extended Now theory, as follows: (170) a. von Stechow(1999)’ s semantic Extended Now theory A perfect marker denotes an Extended Now interval XNP(r) whose right boundary is the reference time r and whose left edge is a contextually salient time. XNP(r) is a restrictor of an adverb of quantification b. Musan(2002)’ s representation: EX t [IN (EN (t))] [VP-(t)] =There is a time t properly included in the ExtendedNow such that VP at t. c. Pancheva(2004)’ s representation: 185 [[PERFECT]] = p<i,t>. t i . t i [XN(t ,t) & p(t )] where XN(t,t) iff t is a final Subinterval of t  To interpret an indexical item, it is introduced in pragmatics an Extended Now interval XN(r) whose left boundary is structurally provided and whose right boundary is the Utterance time indicated by indexical items. In other words, the denotation of „cikum‟ is temporally extended from the Utterance time to the time of saying, by pragmatic resolution. 4.2.4.3. Defending the proposal The proposed analysis can account for (165), repeated in (171) below: (171) a. John said that this guy who is now sitting in front of you killed the woman at the bar. b. [this guy who is now sitting in front of you] John said that t killed the woman at the bar. The fact that there is no DAR involved with (171)-a is accounted for by the tense movement without a pragmatic accommodation. Also, it can account for a DAR-ish reading triggered by other non-temporal indexical elements such as a demonstrative „i(this)‟: 186 (172) John-un [ i yeca-ka LA-ro o-koiss-ta]-ko malhay-ss-ta John-TOP this woman-NOM LA-to come-PRG-PST-COMP say-PST-DEC Reading #1: „John said, “Mary is coming to LA.” ‟ Reading #2: a DAR-ish reading Finally, the pragmatic accommodation can occur even across the direct speech boundary: (173) John-un [Mary-ka cikum LA-ey iss-e]-lako malhay-ss-ta John-TOP Mary-NOM now LA-LOC be-PST-COMP say-PST-DEC Reading #1: <Direct Speech> „John said, “Mary is now in LA” ‟ Reading #2: <Indirect Speech> „John said that Mary was in LA and still in LA now.‟ The examples in (171), (172) and (173) indicate that the pragmatic resolution is triggered by indexical items. 4.2.5. Summary The following three points about the interaction between tense and indexical adverbials have been discussed and proposed in this sub section. 187 a) A Reference Time, a temporal pronoun, has to be considered as a main factor in order to solve the present puzzle of the restriction of temporal adverbials. b) When temporal adverbials result in altering a bound reference time, their modification is not allowed across languages. c) A simultaneous reading, but not a back shifted reading, provides a semantic environment to cause such alteration through its overlapping relation. Also, as for a Double Access Reading effect in Korean, we conclude that there is no obligatory / optional Double Access Reading in Korean, and claim that the DAR effect in question is obtained by a pragmatic resolution triggered by uninterpretable indexical items in indirect quotation. For the pragmatic resolution, it is proposed that the uninterpretable indexical Items introduce in pragmatics an Extended Now interval XN(r) whose left boundary is structurally provided and whose right boundary is the utterance time indicated by indexical items. 188 Chapter 5. Conclusions and Prospects In this dissertation, I have developed an anaphoric theory of tense in Korean. However, I intended to defend neither a quantificational theory of tense nor a referential theory. The long standing debate between the two camps helps to elaborate theories of tense, but it is also true that it has added complications to the big topic of tense, without making much contribution to the empirical aspect. Rather, I have attempted to show the real source of the temporal dependency between tenses across a clausal boundary or sentential one sometimes. I proposed that tense itself is not an anaphoric element but merely denotes a temporal relation. A Reference time, i.e. an argument of tense, is the one bearing the anaphoric property. I claim that it is a temporal pro functioning as an evaluation time of a given clause. Thus, the anaphoric meaning associated with tense stems from the fact that its argument, a temporal pro, needs an antecedent to receive its linguistic content. More precisely speaking, I presented an anaphoric theory of a Reference time in Korean. In Chapter 1, I addressed the theoretical assumptions that I take in the present analysis in order to avoid confusion coming from the flood of undefined and misused terminology. Chapter 2 is devoted to providing a proper introduction to the Korean tense system and important issues worth mentioning. It includes critical reviews of previous analyses of tense in Korean. In Chapter 3, how the proposed tense system works for the temporal interpretation across several types of embedded clauses is shown. I presented the framework of the proposed analysis and explained the difference in the temporal interpretations between complement and non-complement clauses. Furthermore, I elaborated on the proposed analysis so that it can better handle more complex embedding 189 structures such as double or triple embedding constructions. Chapter 4 treats the interaction between tense and indexciality, including the restriction on indexical temporal adverbials inside a complement clause and the matter of the alleged Double Access Reading in Korean, which is associated with the indexical interpretation in nature. Regarding the restriction on indexical adverbials, I have shown that when temporal adverbials result in altering a bound reference time, a temporal pronoun R x in Spec of a CP, their modification is not banned across languages. And, for a Double Access Reading effect, I conclude that there is no obligatory/optional Double Access Reading in Korean and that the DAR effect in question is obtained by a pragmatic resolution triggered by uninterpretable indexical items in a complement clause. For a pragmatic resolution, I have shown that the uninterpretable indexical Items introduce in pragmatics an Extended Now interval XN(r) whose left boundary is structurally provided and whose right boundary is the utterance time indicated by indexical items. 190 Bibliography Abusch, Dorit. 1997. Sequence of tense and temporal de re. Linguistics and Philosophy 20:1-50. Abusch, Dorit. 1997. Remarks on the state formulation of de re present tense. Natural Language Semantics 5: 303-313. Bartsch, Renate. 1976. The grammar of adverbials. North-Holland. Binnick. Robert.I. 1991. Time and the verb. Oxford: Oxford University Press. Borik Olga. 2006. Aspect and Reference Time. Oxford: Oxford University Press. Bybee, Joan, Revere Perkins, and William Pagliuca. 1994. The evolution of grammar:Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press. Comrie, Bernard. 1984. Tense. Cambridge. Dickey, Michael. W. 2001. The processing of tense: psycholinguistic studies on the interpretation of tense and temporal relations, Kluwer Academic publishers. Dowty, David. R. 1979. The syntax and Semantics of tense and time adverbials in English: English fragment. Word meaning and Montague Grammar. Ch. 7, D. Reidel Publishing company. Enç, Murvet. 1987. Anchoring Conditions for Tense. Linguistic Inquiry 18. pp. 633-657. Higginbotham, James. 2002. Why is sequence of Tense Obligatory? Logical Form and Language, Oxford University Higginbotham, James. 2005. Sequence of tense extended. manuscript. Higginbotham, James. 2006. The anaphoric theory of tense. Proceedings of 16th Semantics and Linguistic Theory. Hornstein, Nobert. 1990. As time goes by. MIT Press. Giannakidou, Anastasia. 2003. A Puzzle about until and the Present Perfect. In Perfect Explorations, eds. Artemis Alexiadou, Monika Rathert, and Arnim von Stechow, 101-132. New York: Mouton de Gruyter. 191 Giorgi, Alessandra. 2003. The day after: the distribution of anaphoric temporal location. manuscript. Giorgi, Alessandra. 2010. About the Speaker: towards a syntax of indexicality. Oxford: Oxford University Press pect: From semantics to morphosyntax. Oxford: Oxford University Press. Giorgi, Alessandra, and Fabio Pianesi. 1997. Tense and as Giorgi, A. & Pianesi, F. 2000. On the Morphosyntax of Sequence of Tense in Italian.Lyon Institute for Cognitive Science. Kratzer, Angelika. 1998. More structural analogies between pronouns and tense. Proceedings of SALT VIII. CLC publication, Ithaca, NY. Kusumoto. Kiyomi. 2005. On the quantification over times in natural language. Natural Language Semantics. 13: 317-357. Ludlow, Peter. 1999 Semantics, Tense and Time, MIT Press. Ogihara, Toshiyuki. 1989. Temporal Reference in English and Japanese, Ph.D. Dissertation, University of Texas at Austin. Ogihara. Toshiyuki. 1999b. Double-Access Sentences Generalized. Proceedings from Semantics and Linguistic Theory IX (SALT9), CLC Publications, Cornell University, pp. 224-236. Ogihara, Toshiyuki. 1996. Tense, Attitudes, and Scope, Studies in Linguistics and Philosophy 58. Kluwer. Ogihara. Toshiyuki. 2003.The scope theory of tense and adnominal modifiers.” Empirical and theoretical investigation into language, ed. by S. Chiba et al. Tokyo. Pancheva. Roumanya. & Arim von Stechow. 2004. On the Present perfect puzzle. Proceedings of NELS 34. Pancheva. Roumanya. 2004. Another perfect puzzle. WCCFL 2, ed. G. Garding and M. Tsujimura, pp. 101-114. Somerville, MA: Cascadilla Press. Parsons, Terence. 1990. Event in the semantics of English, MIT Press Partee. Barbara. 1973. Some structural analogies between tenses and pronouns in English. Journal of Philosophy 70: 601-609. 192 Partee. Barbara. 1984. Nominal and Temporal anaphora. Linguistics and Philosophy 7: 243-286. Portner, Paul. 2003. The Temporal Semantics and Modal Pragmatics of the Perfect. Linguistics and Philosophy 26: 459-510. Reichenbach, Hans. 1947. Element of Symbolic Logic, Berkeley: University of California Press Smith, Carlota. 1991. The parameter of Aspect. Ch. 1-4. pp. 3-133. Kluwer Academic Publishers. Sohn, Sung-Ok. 1995. Tense and Aspect in Korean, Center for Korean studies. Schlenker, Philippe. 2000. Propositional attitude and indexicality. Ph.D. dissertation. MIT. Sharvit, Yael. 2003. Embedded Tense and Universal Grammar. Linguistic Inquiry 34:669-681. Sharvit, Yale. 2003. Tense and Identity in copular constructions. Natural Language Semantics 11: 363-393. Song, M. (2000). “A semantics of Sequence of Tense without a Sequence-of-tense Rul.” Language and Informantion: 4.2, 98-105 Stowell, Tim. 1995a “The phrase structure of Tense.” in J. Rooryck & L. Zaring(eds.) Phrase structure and Lexicon, Kluwer. Trautwein, M. 2005. Language context & cognition, Walter de Gruyter. Vendler, Zeno. 1967. Verbs and Times in Linguistics in Philosophy. Ch. 4. pp. 97-121. Cornell University Press von. Stechow, Arim. 1995. On the proper treatment of tense. Proceedings of Semantics and linguistic theory 5. CLC publication. Cornell university, Ithaka, NY. von. Stechow, Arim. 2003. Feature Deletion under Semantic Binding: Tense, Person, and Mood under Verbal Quantifier. MELS 33.

Abstract (if available)

Abstract This dissertation investigates the temporal interpretation of subordinate clauses, aiming to provide a uniform account for the temporal interpretation of both matrix and subordinate clauses in Korean. To this end, I propose a non-deictic analysis of tense in Korean that treats tense as a function denoting a binary temporal relation, not necessarily tied to deiticity, the same way for both matrix and subordinate clauses. I claim that the important element of tense responsible for a deictic or anaphoric interpretation of clauses is a Reference time, a temporal argument of tense, which I treat as a temporal pro in this paper. Thus, the proposed tense system of Korean is simple and neat in that a tense morpheme refers to either an anterior or overlapping relation across the board, maintaining its meaning independent of the context where it appears. Yet, the choice of the antecedent of a temporal pro and the restrictions posed on it determine the temporal interpretation of matrix and subordinate tenses. Therefore, the semantic interpretation of tense heavily relies on what a Reference time of tense can take as its antecedent.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Building phrase structure from items and contexts

PDF

Exploring the effects of Korean subject marking and action verbs’ repetition frequency: how they influence the discourse and the memory representations of entities and events