MIT Working Papers in Linguistics #61

Proceedings of the 6th Workshop on Altaic Formal Linguistics (WAFL6)

Hiroki Maezawa and Azusa Yokogoshi, 2010

for $36.95 x
or
+

Save $16.95 (25%) if bought together with Proceedings of the 1st Workshop on Altaic Formal Linguistics (WAFL 1)

for $49.95 x
or
+

Save $25.95 (38%) if bought together with Proceedings of FAJL 3: Formal Approaches to Japanese Linguistics

for $40.95 x
or
+

Save $26.95 (35%) if bought together with Proceedings of the 8th Workshop on Altaic Formal Linguistics (WAFL8)

for $49.95 x
Show more deals...

The Acquisition of Head-final Relative Clauses in Turkish

Gabriella Hermon, Jaklin Kornfilt and Özge Öztürk
University of Delaware, Syracuse University and New York University

This study examines the acquisition pattern of head-final relative clauses (RCs) in Turkish using an elicited production experimental technique. A large body of work on the acquisition of head-initial relative clauses has found that children perform better with subject-gapped relatives (SGRs) than with object-gapped relatives (OGRs), but the source of this asymmetry is unclear in languages with head-initial RCs.

In such languages (like English), the head and the gap are both linearly and structurally closer in SGRs than they are in OGRs. However, Turkish provides a clear test case for these competing hypotheses. If linear distance is critical, Turkish children should find SGRs more difficult, while if structural distance is critical, they should find OGRs more difficult.

Since naturalistic data provide very few cases of this construction in Turkish (as discussed in Slobin, 1986), we conducted an elicited production experiment with 20 Turkish children (mean age=4;8) and 10 adult controls.

The task in the experiment was to prompt the participants to produce RCs by asking them to describe the changes in the pictures presented to them. The pictures and the questions were carefully designed to elicit RCs with the intended gap position and the type of embedding (see Zukowski, 2001; Hsu, 2006; Hsu, Hermon and Zukowski, 2009).

Our results show that children perform better with SGRs than with OGRs for head-final RCs in Turkish: (i) Children produced more correct RC responses in the subject-gap condition across all three types of embedding conditions than in the object-gap condition; (ii) Children produced more RC-related ungrammatical responses in the object-gap condition than in the subject-gap condition; (iii) Children avoided using RC structure more often in the object-gap condition than in the subject-gap condition.

These findings in Turkish not only argue against the linear-distance-based hypothesis but also provide crucial evidence to support a structural-distance-based account for the subject-object asymmetry in children’s acquisition of relative clauses.

Moreover, since the preference for SGRs in Turkish child language mirrors the preferences found in English, Indonesian and Chinese (all languages in which subject and object gapped RCs do not differ in morphological complexity), we can also argue against a proposal (cf. Slobin, 1986) which ties the asymmetry directly to the complex morphology of OG relative clauses. This last point gains further support from our observations based on our experiments, which show that children do use the morphology of RCs correctly; in other words, sequences of –DIK + agreement are uttered as well as –(y)An without agreement, i.e. just the way an adult would. While the thematic relations may sometimes be wrong (with respect to the experimental situation), there is no doubt that the children do have, in their competence as well as in their production, the ability of using the complex morphology required by head-final RCs in Turkish. Clearly, Slobin (1986 and related work) was wrong in claiming children under 5 lack the morpho-syntax for RCs.

Keywords: RC processing, RC production, subject-object asymmetries in RC acquisition, Turkish, morpho-syntax in Turkish RC acquisition

The Onset of Complex NPs in Child Production

Keiko Murasugi
Nanzan University

Root Infinitives (RIs) are the “default” forms which children, at around two, use in root clauses. Murasugi, Fuji and Hashimoto (2007), at Asian GLOW 2007, discuss that there is a RI stage in Japanese (contra Sano 1995 and Kato et al. 2003), though RIs have not a specific infinitival form: Japanese RIs have past-tensed verb forms (V-ta). Like other languages, I-related (finite-be and Nominative-Case-marker) and C-related (Complementizer and wh-phrases) elements are not found then, and they could denote irrealis meaning.

Based on the detalied analysis of Sumihare database (Noji 1974-1977), this paper provides supportive evidence for this finding, arguing that RIs in Japanese, an agglutinative language, show peculiar characteristics with respect to the age, optionality, and Case-marking at the RI-stage. We show evidence for Phillips’ (1996) analysis that RIs are produced because of children’s deficit of syntactic representation. The multiple head-movement inside the TP projection is acquired step by step.

A Note on the Structure of Relative Clauses in Child Japanese

Koji Sugisaki
Mie University

Within the Principles-and-Parameters approach to UG (including the recent Minimalist Program), a theory of syntactic parameters is simultaneously a theory of the child’s “hypothesis space” during language acquisition. The task for a child is to identify the correct grammar for the community’s language from among the possibilities permitted by the parameters of UG. In principle, then, we can gain insight into the nature of UG parameters by investigating how the child’s grammar changes during the course of acquisition.

In light of this background, this study attempts to provide evidence from child Japanese for the parameter proposed by Murasugi (1991), which regulates the structure of relative clauses. Relative clauses in English and Japanese exhibit several syntactic differences: (i) While English relatives can be accompanied by an overt complementizer, Japanese relatives can never be; (ii) while relativization from an embedded adjunct position is possible in English, this is not permitted in Japanese. In order to account for these differences, Murasugi (1991) proposed the following parameter as a critical point of cross-linguistic variation: Relative clauses are {CP, IP}. According to Murasugi, English takes the CP value, while Japanese takes the IP value. Then, a question arises as to whether Japanese-speaking preschool children have adult-like knowledge of the structure of relative clauses, namely, the knowledge that Japanese relatives are IPs. Our experiment with twenty Japanese-speaking children demonstrated that children around the age of four and five already have the knowledge that relativization from an adjunct position is strictly clause-bound in Japanese, which in turn indicates that these children already know that Japanese relative clauses are IPs. Furthermore, the results of the analysis of child-directed speech suggested that this restriction cannot be directly learned from the input data. These findings argue for Murasugi’s (1991) view that the ban on long-distance relativization is parametrically linked to a different, more prominent property of Japanese relatives, such as the absence of an overt complementizer. A broader implication of this study is that child language acquisition is a potentially valuable source of evidence concerning the parameters of UG.

Keywords: (i) relative clause, (ii) parameters, (iii) parameter-setting, (iv) child Japanese, (v) the acquisition of syntax

The Properties of Ato: From Temporal Expression to Adverbial Expression

Motoko Kawabata
Aichi Institute of Technology

This paper deals with ato “after” and considers the ways in which it is used. The phenomenon of a temporal adverb deviating from its original meaning and acquiring new ones is seen in languages other than Japanese, but my intention is to explicate the framework in which this adverb acquired its new meanings in Japanese. Specifically, I take its fundamental meanings of “remaining amount,” “from the baseline time onward,” and “something continuing from a state that obtained at the baseline time,” and their common meaning of “a resulting situation established due to a certain circumstance having obtained at the baseline time” and consider their effect on the word’s acquiring adverbial and conjunctive uses. Furthermore, I take the construction “ato + expression of quantity,” considered an adverbial usage, and compare it to nokori “remainder,” mô “more,” and ima “now.” I confirm that the meaning of ato as “remaining amount” is one aspect of the situation resulting from a previous state and that their relative positions on the temporal axis are ordered by defining the previous state. When we envision an overall image of a state, these are the difference between perceiving the result of a certain circumstance at the baseline time as a quantity (i), or as a time (ii), or as a relationship (iii). When we perceive it as a quantity, it serves as an indication of the incompleteness of the situation at the baseline time when viewed as a whole. In that sense, ato has uses similar to those of ima “now” and mô “more,” and ii) and iii) have taken on adverbial uses in order to complete the ordering of two states that are being referred to in terms of their ordering.

Predicates with a genitive subject in Old Japanese : A descriptive study of their morphological and semantic characteristics

Eunju Kim
Nagoya University

In Old Japanese, the genitive subject marker no was used as a subject marker in conjunctive clauses. This paper focuses on predicates in conjunctive clauses in which genitive subject marker no appears in Old Japanese (no conjunctive clauses). I especially give careful attention to the morphological and semantic characteristics of predicates in no conjunctive clauses. The main point discussed in this article can be summarized as follows. (1) In no conjunctive clauses, predicates tend to be connected with a conjunctive form which indicates that the event or state described by the predicate actually exists, and there are relatively few examples that convey an uncertain situation. (2) As to what kind of suffixal auxiliaries can be connected to the stem of predicates in no conjunctive clauses, the majority of examples do not contain any such auxiliaries. If any suffix is connected to the stem of a predicate at all, it is most likely to denote past or perfect. We can answer why this holds as follows; In order to make sure that an event certainly exists, there are only two options to choose from; one way is to indicate that the event is taking place at present, and the other is to state that it has already happened. (3) There is no example where m(u), ram(u), or kem(u) is suffixed to the predicate stem of a no conjunctive clause.We can answer why this holds as follows: These suffixal auxiliaries express a conjecture and set up a virtual (i.e., uncertain or unrealistic) world. This conjectural meaning is not compatible with the semantic property that the predicates in no conjunctive clauses should have. (4) On the one hand, no conjunctive clauses impose a morphological constraint upon conjugated forms of predicates. Since the paradigmatic mizen, ren'yoo and izen +ba forms are missing in m(u), ram(u), and kem(u), they cannot occur in no conjunctive clauses. On the other hand, the "size" of no conjunctive clauses is limited to the stage which convey propositional meanings and exclude those which express modality. (5) Predicates in no conjunctive clauses show a strong affinity for "stativity", and the higher degree of lexical transitivity predicates have, the fewer examples we find in no conjunctive clauses. (6) I propose that the characteristics observed in predicates of no conjunctive clauses are correlated to the usage of no as a genitive case marker; that is, the genitive subject marker no modifies the nouniness of a predicate.

keywords: genitive subject, morphological constraint, stativity, transitivity restriction, nouniness

Japanese Traditional Language Study and Modernization of Japanese Linguistics

Toru Kuginuki
Nagoya University

Japanese language study has a couple of main streams. One stream flows from traditional study since 12th century. And the other is western linguistics. The traditional study is composed of Teniwoha and Kanazukai. Teniwoha and Kanazukai had constituted the basic framework of the Japanese classical study. This study had been shouldered by the urban literary aristocracy and upper-class Samurai, urban military nobility. This study had continued from Kamakura period to the first half of Edo period, i.e. the 17th century. For the future from the second half of Edo period, the cutting-edge academic support shifted to provincial lower-class Samurai from the traditional urban aristocracy and upper Samurai. Provincial Samurai’s study of the second half of Edo period had been very demonstrated and rational and practical. Modern language studies in Japan directly depend on the study in this period. This pre-modern Japanese language study provided the modern scientific methods. Modern Japanese linguistics has been based on the thick accumulation of these academic streams.

After the Meiji period (1868~) the introduction of linguistics opened the new stage of Japanese language study. At first modern Japanese grammar was shouldered by scholars out of government since Fumihiko Otsuki (1847~1928). Yoshio Yamada (1873~1958) and Daizaburo Matsushita (1878~1935) are both top grammarians out of main academic stream of government school. Hideyo Arisaka (1900~1952) and Yamada had profoundly considered the theoretical definition on the object of study. Arisaka had studied in classical phonetics, and Yamada in grammar. Arisaka thought what is sound as recognized mentally. Yamada thought what is sentence. At the same time, Motoki Tokieda (1900~1967) thought what is a language. Commonly they referred the metaphysics in Germany. Yamada’s concept of apperception “tokaku” depends on the theory of Immanuel Kant (1724~1804). Tokieda and Arisaka depend upon phenomenology of Edmund Husserl (1859~1938) for their definition “Language as a process” and “sound as an idea of pronunciation.”

Keywords: tokaku as a concept of defining “sentence”, language as a process, on’in as an idea of pronunciation, German metaphysics

Revisiting Beautiful Dancers and Complete Fools in Japanese

Seiki Ayano
Mie University

This paper shows that prenominal adjectives (As) and adjectival nouns (ANs) in Japanese can directly modify nouns (Ns) just like those in English. This direct modification for As and ANs allows for non-intersective interpretation. This study is in sharp contrast to Baker (2003), in which he has observed that while beautiful in Olga is a beautiful dancer is ambiguous between intersective (‘Olga is a dancer’ and ‘Olga is beautiful’) and non-intersective (‘Olga dances beautifully’), the equivalent A utukusi-i ‘beautiful’ in Japanese only permits intersective interpretation, and has proposed a relative-clause analysis of prenominal As and ANs. Further, he argues that the principle behind it is a lack of phi-feature agreement between Ns and modifying As/ANs. Contrary to Baker, this study supports Yamakido's (2000, 2005, 2007) findings of non-intersective As and ANs in Japanese, and provides further evidence for their presence. Further, I propose that such non-intersective As and ANs occupy a position lower than intersective ones, which parallels facts in English (Larson 2000). The present study also addresses two further issues concerning adjective ordering within DP and prenominal ANs with two kinds of suffixes, namely, -na and -de-aru. First, recent studies such as Svenonius (2008) and Truswell (2009) propose that subsective As, which are a subclass of non-intersective As, merge higher than intersective As. Since their proposals and relevant arguments contradict the present study, I attempt to defend my analysis by pointing out a couple of problems concerning their definition of subsective As and a relative position within DP for non-intersective As. Second, a contrast in grammaticality between non-intersective ANs with -na and those with -de-aru suggests that they are derived differently. Because of the ungrammaticality of non-intersective ANs suffixed by -de-aru, this study proposes a Kaynian relative-clause analysis for them.

keywords: adjective, Japanese, intersective, non-intersective, DP-internal syntax

Wh-Reduplication in Altai

Elizaveta Bylinina
Institute for Linguistics OTS, Utrecht University

The paper is concerned with a particular indefinite series in Altai and the questions these items raise for the theory of distributivity and semantics of indefinites. Tuba possesses a number of indefinite pronoun series. Most of them consist of a wh-root and a series marker – a particle (de/da, le/la etc.). One of the series though is derived from the wh-root not by adding a particle but via an operation of reduplication (kem-kem ‘who.REDUPL’, qandu:-qandu: ‘which.REDUPL’ etc.).

Altai wh-reduplication is unique as far as we know: it patterns with lots of other known indefinites in non-specific contexts, but shows unusual behavior in referential specific episodic contexts. When not under an operator, reduplicated wh-words in Altai have plural indefinite reading with collective interpretation (‘several individuals at once’). Known indefinite pronouns that are acceptable both in non-specific and episodic specific contexts (Altai -de/-da series, Russian -to series, English some etc.) do not show any signs of plurality when they are specific. In Altai, to the contrary, the only available specific interpretation is plural. Known wh-reduplicated indefinites in other languages are ungrammatical in these contexts.

Following the idea of a connection between indefiniteness and distributivity, one can view indefinite pronouns as markers of co-variation introducing a dependent variable (a variable the values assigned to which co-vary with those assigned to another variable, the so-called ‘domain variable’). We try to pursue this line.

The derivation of sentences with wh-reduplication in Altai we believe involves the following steps: 1) as usual, the wh-word introduces Hamblin set of alternatives; 2) operation of reduplication ‘catches’ the Hamblin alternatives and makes them an ordinary, non-Hamblin set; 3) in the case of narrow scope – i.e. non-specificity – the plural participant gets interpreted distributively, as a sum; in the other case, it shifts to group and gets interpreted collectively.

The analysis leaves many issues open, but we believe that the kind of indefinites we are discussing provide good grounds for the analogy between indefinites and distributivity, which needs to be explored further.

Keywords: Altai, reduplication, indefinites, distributivity, wh-words.

Description and Explanation in Inflectional Morphophonology: Notes on the Japanese Verb

Brent de Chene
Waseda University

In morphophonology, descriptive adequacy cannot be attained merely on the basis of analyzing patterns of distribution and alternation, given the existence of multiple observationally adequate accounts of a single data set; nor can explanatory adequacy be attained simply by adopting a priori a particular definition of simplicity, given the existence of cases where speakers have evidently arrived at analyses distinct from those predicted by the standard assumptions of phonologists (Hale 1973). The present paper, set within the research program (Kenstowicz and Kisseberth 1977:3) that seeks to determine the explanatory principles of morphophonology through examination of a critical mass of cases for which the descriptively adequate analysis is known from external evidence, first adduces ongoing change in the system of Japanese verbal inflection to show that the descriptively adequate analysis of that system is the one that, out of three observationally adequate analyses, would be least highly valued according to the classical feature-counting evaluation metric. It next shows that the choice of the descriptively adequate analysis follows from two principles based on proposals of Albright (2002), namely that (1) base forms are surface alternants drawn from a constant morphologically or phonologically defined environment across a given lexical category and (2) consistent with (1), base forms are chosen so as to allow prediction of the maximal number of inflected forms from the paradigm of the maximal number of stems. Reanalyses from the history of Portuguese and Korean are then adduced to argue that an implicit assumption of this account, namely that speakers evaluate base forms and the associated rules together as a unit, is false, and that the choice of the descriptively adequate analysis in the Japanese case is better understood as the result of two distinct principles, namely that (3) candidate analyses are evaluated not as complexes of base forms and rules, but sequentially, with base forms being chosen first, and (4) where the type frequency of a morpheme alternant or an alternation is the percentage of eligible morphemes with which that alternant or alternation occurs, basic alternants and regular alternations are those with the highest type frequency. In conclusion, it is pointed out that the Japanese case counterexemplifies the hypothesis that speakers obey surface phonotactic constraints in choosing basic forms (Hale 1973:419) and that, because it involves suffix alternations whose conditioning is purely phonological, it counterexemplifies the hypothesis that speakers derive inflected forms from one another within the paradigm of a single stem (Albright 2002, Bybee 1985:49ff.) and the hypothesis that speakers choose basic forms on semantic criteria (Bybee and Brewer 1980:203).

Keywords: description, explanation, inflection, morphophonology, Japanese

The Function of Accusative Case in Mongolian

Dolgor Guntsetseg
University of Stuttgart

In Mongolian, direct objects of transitive clauses and subjects of some embedded clauses can occur either in morphologically unmarked (nominative) form or they can be marked with the accusative case marker -(i)g. The former refers to the Differential Object Marking (DOM), the latter is a special type of Differential Subject Marking (DSM). This paper first investigates the conditioning factors for the accusative case marking in both instances, DOM and DSM, and presents some results of an experimental study that is conducted to find out the conditions for the DSM. And secondly, it deals with the question: what function has the accusative case marker in Mongolian: identifying and distinguishing function (cf. Comrie 1989, Song 2001, and among others)? More precisely, how the data in Mongolian might be analyzed within the framework of case function theory associated with the argument strength scale, which is suggested by de Hoop & Narasimhan (2005)?

DOM in Mongolian depends mainly on the referential features (there are also other weak factors) of the related arguments and does not reveal the function of the accusative case marker. Since the conditions for DSM were not clear, a written questionnaire was designed. The results of this study indicate that the accusative case marking of subjects of embedded clauses is more likely to be required if the two subjects, the main and embedded subjects, are in immediately adjacent position in the surface structure and if the embedded subjects are highly referential.

This leads to the assumption that accusative case marking in Mongolian clearly has the distinguishing function. It distinguishes i) between the direct object and the subject in a main clause and ii) between two subjects in the complex clauses; in both instances the related two arguments are closer to each other on the argument strength scale. Furthermore, I argue that the argument strength scale should be extended to include arguments across clause boundaries.

Keywords: case function, argument strength, differential case marking, Mongolian, accusative subjects

Relative Clauses Processing before the Head Noun: Evidence for Strong Forward Prediction in Turkish

Barış Kahraman, Atsushi Sato, Hajime Ono and Hiromu Sakai
Hiroshima University, Japan Scociety for Promotion of Science/Hiroshima University, Kinki University and Hiroshima University

Previous studies have shown that filler-gap dependencies (structures like scrambling sentences, in which the filler precedes its gap-site) are processed incrementally in a head-final language like Japanese (e.g., Aoshima, Philips and Weinberg, 2004, Miyamoto & Takahashi, 2004). In the case of structures such as relative clauses, the gap precedes the filler in head-final languages. We refer a dependency where the gap precedes the filler as gap-filler dependencies. From the existing data in Japanese, it is hard to determine whether gap-filler dependencies are also processed incrementally as in filler-gap dependencies.

In the present study, using Turkish subject and object relative clauses (SRCs / ORCs), we investigated whether gap-filler dependencies are processed incrementally. We particularly examined whether the parser postulates a filler as soon as the gap-site is determined, or it starts to construct a gap-filler dependency after the filler is encountered. In order to answer this question, we conducted two sentence-fragment completion experiments and one self-paced reading experiment. The results of the sentence-fragment completion experiments revealed that when Turkish native speakers encountered the RC-verb, they produced relative clauses almost at the rate of 100%. Furthermore, the results of the self-paced reading experiment revealed that SRCs were read faster than ORCs, and the processing asymmetry was observed only at the embedded verb position.

These results demonstrate that there is a strong forward prediction for RC-head (filler) at the embedded verb position, and this strong forward prediction is rapidly reflected in the online sentence processing as well. These findings suggest that in order to construct a gap-filler dependency, Turkish native speakers do not wait until the RC-head, and as soon as the gap-site is determined at the RC-verb position, they immediately postulate a filler, and start to construct a gap-filler dependency even before the filler appears. Therefore the processing of gap-filler dependencies is also incremental as in the case of filler-gap dependencies, and verb morphology is a crucial source for decision making and incremental processing of gap-filler dependencies in head-final languages.

Keywords: Relative clause processing, gap-filler dependencies, incrementality, Turkish, Japanese

The Structures of Modality in Korean

Shin-Sook Kim
SOAS, University of London

Korean has certain nouns which are used for expressing ability, possibility and necessity, such as swu (‘ability/possibility’), li (‘(circumstantial) possibility’), and philyo (‘necessity’). Following Palmer (2001) I use the terms ‘epistemic’ (based on our knowledge) and ‘dynamic’ (internal properties of an individual; ability) to describe the two types of modal interpretation relevant in this paper. As with modal expressions familiar from other languages, for example English can, swu allows both dynamic (ability) and epistemic (possibility) interpretations. In contrast, li only has an epistemic interpretation. The modal nouns all appear in construction either with iss-ta (‘exist’) or eps-ta (‘not exist’).

In this paper I consider the structures that these Korean modal nouns appear in and propose two clearly distinct structures correlating with different interpretations: a ‘propositional complement’ structure for epistemic modals and a ‘control’ structure for dynamic modals. There is also an independent subject-to-subject raising structure (with the raising predicate kes kath-ta (‘seem’)). I argue that the epistemic structure in Korean does not involve subject-to-subject raising, and the initial NP is part of an embedded CP.

Evidence for these structures comes from the interpretation of subjects (diagnostics following Kuno 1973 on Japanese), honorific predicate marking, and the interaction of modal structures with negation and NPI licensing, as follows:

(i) The nominative-marked subject in a dynamic modal structure necessarily has an exhaustive interpretation, while the subject in an epistemic modal structure may have a neutral interpretation. This shows that the former is a matrix subject and the latter is an embedded subject. The same contrast holds with topic marking on the subject. The subject of a dynamic modal is very natural with the topic marker nun. However, with an epistemic modal, as the initial subject NP is embedded, nun must receive a contrastive interpretation; no topic interpretation is possible.

(ii) Honorific marking on the matrix predicate iss-ta/eps-ta also shows that only the dynamic modal has a matrix referential NP subject, and that the initial NP with an epistemic modal is embedded. The subject does not undergo raising into the matrix clause and does not trigger honorific agreement on the matrix predicate.

(iii) Interactions with quantifiers, negation and NPI licensing also diagnose the two different structures. The subject in an epistemic modal structure tends to scope low, as it is structurally low, while the subject in a dynamic modal structure scopes high, and these scopes can be diagnosed by NPIs in Korean, which must be clause-mates with their licensing negation.

Keywords: modals, dynamic, epistemic, control, propositional complement

A Contrastivist View of the Evolution of the Korean Vowel System

Seongyeon Ko

Cornell University

This paper provides a unified formal account of the synchrony and diachrony of the vowel system in Korean. The framework adopted in this paper is what I call the Contrastive Hierarchy theory (Dresher 2009), which views phonology operates only on contrastive features (Hall 2007) that are organized into a hierarchical ordering.

This contrastivist view will shed new light on a variety of controversial issues in the phonology of Middle Korean vowels such as the asymmetry of the vowel inventory, the harmonic feature of the vowel harmony, the so-called discrepancy between the vowel system and harmony (K-M Lee 1972), and the neutrality of the vowel /i/ to the vowel harmony. It is shown that all these issues are resolved together within the contrastive hierarchy [coronal] > [low] > [labial] > [RTR], established on the independent basis of the phonological patterns in Middle Korean. In view of this hierarchy, the Middle Korean vowel system is analyzed as a RTR-based two-height vowel system (J Kim 1999, J-K Kim 2000) despite the apparent three-way height distinction. The difference between the phonetic and the phonological vowel system is attributed to the phonetic effect of sympathetic/antagonistic feature combination between tongue root and height features (Archangeli and Pulleyblank 1994, J-K Kim 2000).

The contrastivist approach also explains the historical development of the Korean vowel system, from Late Middle Korean through Early Modern Korean to modern regional dialects. It is argued that the changes in the vowel systems are best accounted for in terms of the changes in the contrastive hierarchy. For example, the change from Middle Korean to Early Modern Korean is analyzed as the change from the Middle Korean contrastive hierarchy [coronal] > [low] > [labial] > [RTR] to the Early Modern Korean contrastive hierarchy [coronal] > [low] > [high] > [labial], explaining the two different phonological merger patterns of the so-called alay a as neutralization of minimal contrast (RTR neutralization in Middle Korean vs. labial neutralization in Early Modern Korean) under the different hierarchies. In a similar vein, the synchronic variations in the vowel systems of modern dialects receive a comparable explanation: the two directional tendencies in the development of vowel system in modern dialects (Kwak 2003), i.e., the loss of labial contrast in North Korean dialects and the decrease of the number of height distinction in South Korean dialects, are ascribed to the flux in the relative hierarchy between [high] and [labial] in Early Modern Korean.

The present research, inspired by the analysis of Manchu vowel systems in Dresher and Zhang (2005), seems to provide a typologically plausible treatment of the Korean vowel system within the areal context of Northeast Asia, particularly in relation to Tungusic, although it calls for further research into other Northeast Asian languages such as Mongolic.

Keywords: Middle Korean, Early Modern Korean, vowel system, vowel harmony, contrastive hierarchy

Focus Particles and Suru-support in Japanese

Sachie Kotani
Tezukayama University

It is well-known that the English dummy verb, do, appears in questions, negations, VP-ellipsis cases, and emphatic cases. On the other hand, there is a specific case in Japanese where a dummy verb, suru ‘do,’ necessarily appears. When a verb is associated with a focus particle, sae ‘even,’ the verb stem is suffixed with sae, which must be followed by suru along with the tense morpheme, as shown in (1).

(1) Ano kin-medarisuto-wa uta-o DASI-sae si-ta.

that gold-medalist-Top song-Acc release-even do-Pst

‘In addition to doing something else with a song (e.g. writing or producing one), that gold-medalist even released a song.’

This paper characterizes the environment in which we see the appearance of a Japanese dummy verb, suru ‘do’. I propose two types of head movement in the syntax: the ordinary head movement and Partial Head Movement. I then develop a morpho-syntactic theory of the derivation of suru-support, where, in the syntax, Partial Head Movement applies to v, leaving √ behind, and derives the partial complex head v-T, which is realized as a tensed dummy verb in the morphology.

Without a focus particle in a derivation, there is no distinctive difference observed between the two types of head movement since they derive the same linear order in the morphology. With a focus particle, however, they derive different linear orders in the morphology, which results in the realization of different lexical items.

This paper also explains the cases where multiple predicate morphemes appear and the cross-linguistic effects triggered by a focus particle.

Focus particles do not take part in argument structures. They do not trigger movement. They are usually considered adjuncts and seem to have no effects in the syntax. This paper shows, to the contrary, that they actually affect the appearance of suru-support and even affect the word order in focus association.

Japanese Light Verb Voicing as a Connective Morpheme

Kazutaka Kurisu

Kobe College

Japanese has a morphological construction consisting of a Sino-Japanese stem (SJS) and the light verb (LV) /suru/. The initial consonant of LV undergoes voicing occasionally. In this paper, I argue that LV voicing is a phonological manifestation of the lexical status of SJS+LV morphological complexes. I analyze this observation with the notions of allomorph listing and cophonology. Also, the analysis is further extended to nominal and verbal compounds.

Evidence from accentuation, adverbial reduplication, and gemination strongly shows that SJS and LV are combined at the lexical level. The crucial observation for claiming the lexicality of SJS+LV complexes is that accentuation and LV voicing are in complementary distribution. Initial voicing of LV does not take place when the SJS+LV complex is accented on its penultimate syllable, and vice versa. Penultimate accentuation is in full conformity with the accentual pattern of monomorphemic verbs and verbal compounds. Monomorphemic verbs may remain unaccented, exactly like SJS+LV morphological complexes. Compound accentuation denotes lexicality of combined members, so the complementary distribution of penultimate accentuation and LV voicing indicates that LV voicing is a phonological expression of the lexicality of SJS+LV complexes.

The complementary distribution of LV voicing and penultimate accentuation receives a straightforward analysis with the notions of allomorph listing and cophonology. As a quite similar analytical possibility, lexical indexation is also discussed. It turns out that this alternative approach is not satisfactory since the complementary distribution of penultimate accentuation and LV voicing is missed.

The proposed analysis is sharply different from Rice (2005) and Kurisu (to appear) who maintain that LV voicing is postnasal voicing. These earlier analyses encounter two problems. First, they fail to reflect the fact that penultimate accentuation and LV voicing are in complementary distribution. Second, they require an unmotivated stipulation that LV is underlyingly /zuru/ in some SJS+LV complexes with a vowel-final SJS. Crucially, postnasal voicing would not be required once both /suru/ and /zuru/ are postulated. In addition, my comprehensive statistical survey of one of the largest Japanese dictionaries reveals that postnasal voicing is not a predominant factor. These considerations suggest that the earlier analyses appealing to postnasal voicing are not on the right track.

The analysis of SJS+LV complexes is further extended to nominal and verbal compounds. Voicing occurs in these two types of compounds although the frequency is different. Voicing is quite productive in nominal compounds. As a result, both voicing and accent assignment normally appear on the surface. On the other hand, voicing is sporadic and only exceptional in verbal compounds. I argue that nominal and verbal compounds contain both [accent] and [voice] in their lexical representation. This means that SJS+LV complexes, nominal compounds, and verbal compounds are alike in the sense that both [accent] and [voice] are present in the three morphological constructions. The different frequency of the occurrence of voicing is analyzed by taking advantage of cophonology theory again. The three morphological constructions are associated with different constraint hierarchies such that the three morphological constructions exhibit different degrees of voicing.

Keywords: allomorph listing, cophonology, light verb, Sino-Japanese, voicing

The Nominative/Genitive Alternation in Modern Mongolian

Hideki Maki[1], Lina Bao[1], Qing-Yu Wu[1], Wurigumula Bao[1], Asako Uchibori[2], Fumikazu Niinuma[3], and Kenichi Goto

[1]Gifu University, [2]Nihon University, and [3]Kochi Gakuen College

This paper aims to report thus far unreported phenomena regarding the nominative/genitive alternation in the Eastern dialect of modern South Mongolian (Mongolian, hereafter) spoken in Inner Mongolia, an Altaic language, and to discuss what they suggest for the theory of (comparative) syntax. We report seven phenomena in this paper. First, while genitive subjects are disallowed in simple sentences, both nominative and genitive subjects are allowed, when they appear in sentential modifiers to nouns. Second, genitive subjects are allowed in sentential modifiers without overt nominal heads in Mongolian, just as in Watanabe’s (1996)/Hiraiwa’s (2001) Japanese examples. Third, the Transitivity Restriction, which prohibits co-occurrence of an accusative DP with the genitive DP, does not apply to the distribution of genitive subjects in Mongolian. Fourth, a genitive subject is allowed in a non-local relation with the external nominal element. Fifth, in contrast to Dagur, one of its related languages, Mongolian allows complete optionality in agreement between a genitive subject and the external nominal head, and agreement between a genitive possessor and the host nominal. Sixth, the nominative/genitive alternation is completely optional in Mongolian, as long as the Conditions on Genitive Subject Licensing in Mongolian are met, in sharp contrast to Turkish, another Altaic language, in which there is no “genuine” nominative/genitive alternation. Seventh, and finally, there is a restriction on the occurrence of genitive subjects in Mongolian, depending on the types of the predicates. Namely, genitive subjects are disallowed with some predicates expressing “existence/possession.” Based on some of the findings, we argued for the conditions on genitive subject licensing in Altaic languages: a genitive subject must be c-commanded by a nominal element, and must be in a local relation with a particular form of the predicate. Based on the fact that a genitive subject is allowed in a non-local relation with the external nominal element in Mongolian, we also provided further evidence for Bošković’s (2007) claim on the nature of Agree, that is, the locality of Move and Agree is radically different in the sense that Agree is free from mechanisms constraining Move, such as the Phase Impenetrability Condition.

Keywords: agree, case alternation, genitive, Mongolian, nominative

ATB-Movement and Resumption: The Case of Japanese Left Node Raising

Chizuru Nakao
Daito Bunka University

It is widely acknowledged that island violations are ameliorated when a resumptive pronoun is inserted in place of a gap/trace (Ross 1967). Also, it is often claimed that a resumptive pronoun is a ‘last resort’ to save otherwise illicit movement (Aoun, Choueiri and Hornstein 2001, McDaniel and Cowart 1999, and Shlonsky 1992). That is, the use of resumption is limited to the case where a gap violates a locality constraint (e.g. island constraints). This paper defends such a view by showing that a null resumptive pronoun appears only when movement is blocked by an island. Specifically, I will focus on the null resumptive pronoun employed in the ATB-movement that derives the construction in Japanese which I call Left Node Raising (LNR). I will examine properties of Japanese LNR and claim that the ‘shared element’ of LNR undergoes ATB-movement from both sentences of a conjoined structure. Differences between LNR and Null Object Construction (NOC), such as existence of Case matching effects, show that LNR should not be treated as a variant of NOC. However, some of the properties of LNR disappear when the second gap position of LNR is included inside an island. I argue that this type of example involves a resumptive pro strategy. The second gap position is interpreted as a null resumptive pronoun only when ATB-movement is blocked due to the existence of an island, which leads us to conclude that the null resumptive pronoun in Japanese LNR is a last resort strategy. This is the same conclusion as Ishii’s (1991) argument on resumption in Japanese relative clauses. However, it is also observed that resumption in LNR is only available for the gap in the second conjunct of ATB-movement, but not the gap in the first conjunct. This fact may suggest that conjunction has an asymmetrical structure, and/or that linear distance in addition to the existence of an island violation determines the availability of a resumptive pronoun.

Keywords: Resumption, Left Node Raising, null pronouns, ATB-movement, islands

Possible syntactic subject positions in Turkish: Evidence from phonology

Öner Özçelik and Miho Nagai
McGill Universtiy and CUNY Graduate Center

The question of where in the sentence nominative arguments can appear has been well-studied within the fields of syntax (e.g. Heycock 1993; Ura 1996 for Japanese) and semantics (e.g. Diesing 1992; Kratzer 1996 for English and German). Most of the debate has centered around the issue of whether a nominative phrase has to be licensed in SpecTP (e.g., Chomsky 1991) or if it may remain in its base position (i.e., internal to vP/VP, Agree model in Chomsky 2000). In particular, it has been suggested, for several languages such as German, Greek, Japanese and Turkish, that, in these languages, certain subjects might be vP/VP-internal, never raising to SpecTP (see e.g. Haider 2005 and Wurmbrand 2006 for German; Alexiadou & Anagnostopoulou 2001 for Greek; Tateishi 1994 for Japanese; Kornfilt 1984 and Öztürk 2004, 2005 for Turkish).

In this paper, we provide, for the first time, prosodic evidence in support of this position: we argue, based on Turkish, that the latter option is true for this language; that is, Turkish allows certain nominatives to stay in situ. In particular, we show that indefinite subjects remain in a vP/VP internal position in this language whereas definite subjects must raise to SpecTP. We support this argument with prosodic evidence, which is based, mainly, on the following two facts: (a) In a simple sentence composed of an indefinite subject and a predicate, a single phonological phrase (PPh) is created, and this PPh is shared by these two constituents, signaling that the two are under the same syntactic projection, i.e. in a vP/VP internal position in this case. (b) For a similar sentence with a definite subject, on the other hand, two different PPhs are created, one for the subject and one for the predicate, implying that the two are not any more under the same vP/VP projection. (1) and (2) below respectively illustrate these two facts:

(1) [ [Adám gel-dí]PPh ]I (2) [ [Adám]PPh [gel-dí]PPh ]I

man arrive-PAST man arrive-PAST

“A man arrived.” “The man arrived.”

We also present independent evidence for this account: First, bare nouns in unergatives cannot have an indefinite interpretation, as shown in (3):

(3) a. *[ [Adám gül-dú]PPh ]I b. [ [Adám]PPh [gül-dú]PPh ]I

man laugh-PAST man laugh-PAST

“A man laughed.” “The man laughed.”

This is not surprising since, under the standard assumption, arguments of unergative verbs have to be external, introduced by a functional head such as v (e.g. Kratzer 1996; Chomsky 1995), while, as we argued above, Turkish indefinites, must be vP/VP-internal.

In existential constructions, we see the mirror image of what is happening with unergatives: Because of the Definiteness Effect (DE) (Milsark 1977), subjects of existentials cannot have a definite interpretation, and the prosodic representation with two PPhs is ungrammatical, too, for this would imply, for Turkish, that the subject would have to raise to SpecTP, i.e. the position for definite subjects:

(4) a. [ [Adám vár]PPh ]I b. *[ [Adám]PPh [vár]PPh ]I

man exist(ent) man exist(ent)

“There is a man.” “There is the man.”

Keywords: Multiple subject positions, Syntax-prosody interface, Turkish syntax, Turkish prosody, Phonological and Intonational phrase in Turkish

Korean Affricates and Consonant-Tone Interaction

Jeremy Perkins and Seunghun Julio Lee
Rutgers University and Central Connecticut State University

Korean has a three-way laryngeal contrast (lax, aspirated, and tense (Kim, 1965)) in obstruents which interacts with tone on a following vowel (Silva, 2006, Wright, 2007). Silva and Wright showed that lax obstruents occur with low tone in a following vowel, whereas aspirated and tense obstruents occur with high tone in a following vowel. These studies investigated consonant-tone interactions with stops only. The present study extends this investigation to include affricates. We argue that affricates are strident stops based on their phonological and phonetic behaviour.

First, while affricates and stops display the full three-way laryngeal contrast, fricatives do not. Lax fricatives are absent in Korean. This can only be explained by treating affricates as [+strident, –continuant], following Jakobson et al (1952), and that they are crucially not [+continuant] (cf. Sagey, 1986; Lombardi, 1990, who posit affricates containing both [–continuant] and [+continuant] values). Lax fricatives are banned by an independent constraint against lax sounds that are [+continuant]. Consonant-tone interaction in Korean involves only laryngeal features on consonants then, as is observed cross-linguistically.

Second, a phonetic experiment was conducted to measure VOT and F0 following obstruents, including affricates. The results showed that 1) F0 is lower for both lax stops and affricates when compared to tense stops and affricates and 2) there is no VOT difference between lax and aspirated stops, nor between lax and aspirated affricates. Therefore, our findings agree with Silva (2006) and Wright (2007): The contrast between lax and aspirated obstruents resides in the effect on F0 on a following vowel. We have extended the coverage of this finding to include affricates, in addition to stops.

Finally, an OT analysis of consonant-tone interaction for Korean affricates is proposed based on Lee (2008). The account allows any segment to associate with tone, utilising a constraint that is violated by a segment not linked to a tone (ROOTNODE→T). The faithfulness constraint, IDENT-T only applies to tones that are linked to moras, correctly deriving the cross-linguistic observation that non-moraic segments can never bear a tone contrast. Consonant-tone interaction, such as that seen in Korean, is achieved by ranking the markedness constraints, ROOTNODE→T and constraints banning tense and aspirated sounds with L tone (*[+c.g.]/L and *[+s.g.]/L) above the faithfulness constraint, IDENT-T.

Keywords: Consonant-Tone Interaction, Affricates, Voice-Onset Time (VOT), Optimality Theory (OT), Korean

A Comparative Syntax of Ellipsis in Japanese and Korean

Mamoru Saito and Duk-Ho An
Nanzan University and Konkuk University

This paper examines ellipsis in Japanese and Korean, focusing on argument ellipsis, “sluicing” and N’-ellipsis. It shows that N’-ellipsis obtains in Japanese but not in Korean, and attributes this difference to a micro-parameter in the genitive marker insertion rule. It also presents evidence for the LF copying analysis of N’-ellipsis as well as argument ellipsis.

It has been known since Kim 1999 and Oku 1998 that argument ellipsis applies in the same way in the two languages. We first discuss Oku’s argument for argument ellipsis and Shinohara’s 2006 evidence that it is interpreted through LF copying. On the other hand, the analysis of “sluicing” has been controversial for both languages, which has made the comparison difficult. Building on Takahashi 1994 and Nishiyama, et al. 1996, we argue that the “sluicing” structure in both languages is derived by applying argument ellipsis to a cleft sentence, eliding the CP subject that expresses the presupposition. Then, developing the discussion in Sohn 2000, we show that there are cases where “sluicing” is grammatical despite the fact that the corresponding cleft sentence is not. This discrepancy poses an apparent problem for our analysis because cleft should provide the “source” for “sluicing.” But we argue that this discrepancy is expected under the LF copying analysis, and hence, provides further supporting evidence for it.

Finally, we turn to the difference between Japanese and Korean with respect to N’-ellipsis. This difference is somewhat surprising, given that the noun phrase structures of the two languages are otherwise very similar. We first argue on independent grounds that there is a micro-parameter dictating the distributions of genitive markers in Japanese and Korean. Genitive is not licensed when the head noun is not overtly expressed in Korean, but there is no such restriction in Japanese. Then, we show that this micro-parameter explains the absence of N’-ellipsis in Korean as well as its presence in Japanese. Having presented an analysis for the absence/presence of N’-ellipsis, we compare N’-ellipsis with “right-node raising,” which is argued to be a PF process in An 2007. We examine the difference between the two with respect to the distributions of genitives, and demonstrate that the result leads to evidence that N’-ellipsis also involves LF copying.

Keyword: comparative syntax, ellipsis, sloppy interpretation, genitive insertion, LF copying

Coordination, Dependency, and Gapping in Japanese

Yosuke Sato
National University of Singapore

Within the theory of coordination and ellipsis laid out by Williams (1997) and further articulated by Ackema and Szendröi (2002), this paper proposes a new analysis of gapping in Japanese as the product of two elliptic processes, Coordinate Ellipsis and Dependent Ellipsis. I demonstrate that, with a slight revision of the Dependent Ellipsis originally proposed by Williams in terms of phonological adjacency, the current analysis provides a straightforward explanation for a wide variety of syntactic, semantic, and phonological properties of this construction discovered in the literature regarding information structure of remnants (Kuno 1976; Sag 1976), postposition/case-drop (Abe and Hoshi 1997), island-insensitivity (Mukai 2003), semantic identity effects (Mukai 2003), and the amelioration of the Left Branch violation (Kim 1997, An 2007). One important implication of the proposed analysis is that gapping is an interface phenomenon: its multi-faceted properties cut across the traditional modules of grammar. I show that our proposed analysis has right architectural characteristics to accommodate this intermodular nature of gapping in a way that previous uni-modular analyses that resort to Right Node Raising (Saito 1987), LF Copy (Abe and Hoshi 1997), PF Deletion (Mukai 2003), or combinations thereof (Kim 1997, 1998) cannot.

Keywords: gapping, Japanese, coordinate ellipsis, dependent ellipsis, phonological adjacency

Argument Ellipsis in Japanese and Turkish

Serkan Şener and Daiko Takahashi
University of Connecticut and Tohoku University

This article compares Japanese and Turkish with respect to the possibility of ellipsis of subjects and objects (or argument ellipsis), with the aim of elucidating what property (or properties) it correlates with. To date, the literature has presented two hypotheses: One correlates argument ellipsis with free word order (or scrambling), and the other associates it with (anti-)agreement. Japanese is not very informative in evaluating those hypotheses because it both allows free word order and lacks agreement. The present study considers Turkish, which is similar to Japanese in permitting scrambling but differs from it in exhibiting agreement between subjects and their predicates. While it has been shown in the literature that subjects and objects may be elided in Japanese, we point out that Turkish only allows ellipsis of objects. At first sight this discrepancy seems to be puzzling especially given that Turkish is similar to Japanese in permitting scrambling. We claim that the absence of subject ellipsis in Turkish is related to the presence of subject-predicate agreement in the language. This is supported by our findings that even in Turkish, subjects may undergo ellipsis in the environment where agreement is absent, such as a certain kind of adjunct clause and the Exceptional Case-Marking construction. Thus, the present article provides an argument for the hypothesis that (anti-)agreement is crucially involved in the licensing of argument ellipsis.

Keywords: argument ellipsis, scrambling, agreement, null arguments, comparative syntax

A Linking Morpheme Analysis of Japanese Compound Accentuation

Junri Shiamda
Center for Advanced Research on Logic and Sensibility (CARLS), Keio University

This paper presents a new analysis of Japanese compound accentuation by virtue of a linking morpheme bearing a [+acc] feature.

The accentuation of non-compound nouns in Japanese is in general unpredictable, but it has been observed that the default pattern is to put an accent on the syllable containing the antepenultimate mora (McCawley 1968, Kubozono 2008). Under Kubozono’s hypothesis that the default accentuation is due to a “floating” accent, this pattern can be captured by the following constraint ranking in Optimality Theory:

(1) Max[+acc] >> NonFinality(Foot) >> Rightmost

On the other hand, the accentuation of compound nouns in Japanese is mostly predictable from the length and the accentuation of the second member of the compound (N2) (Kubozono 1995, 1997, 2008). If N2 in isolation has an accent, it tends to become the compound accent, but when N2 is not accented or has an accent on the final syllable or foot, the compound accent emerges at the boundary of N1 and N2. The exact syllable at the boundary to be accented is the final syllable of N1 when N2 is “short” (i.e. one or two morae), and the initial syllable of N2 when N2 is “long” (i.e. three or four morae).

To account for the observed generalization of the accentuation of compounds, Kubozono (1995, 1997) has developed a pioneering analysis in Optimality Theory. In particular, Kubozono proposes a constraint called Parse(N2), which dictates preservation of the N2 accent as the compound accent.

The present paper illuminates and attempts to resolve two problems of Kubozono’s approach: that it entails an otherwise unmotivated artificial split of a faithfulness constraint on accents, and that the undesirable Parse(N2) is conveniently and randomly reranked to account for apparently exceptional cases.

The main proposal of the paper is that in the input for a compound, a linking morpheme bearing a [+acc] feature sits between N1 and N2 and that this linking morpheme may be fused to one of its adjacent syllables, resulting in a boundary accent. It is shown that the independently-motivated ranking in (1) basically suffices to derive the accentuation of compounds, dispensing with Parse(N2) and the concomitant, undesirable split of the faithfulness constraint. The first problem of Kubozono’s analysis is thus resolved, and as a consequence, a unified account of the accentuation of non-compounds and compounds is attained.

The second problem of Kubozono’s analysis dissolves as Parse(N2) has been eliminated. Nevertheless, an explanation is still called for of apparently exceptional cases. For this problem, the paper suggests that “exceptions” emerge when N2 somehow becomes immune to NonFinality(). For exceptional N2s that are “short”, such a situation arises when they receive default accentuation with a floating accent, and for exceptional N2s that are “long”, when they are themselves compounds, either etymologically or phonologically.

Keywords: accent, compound, linking morpheme, default accentuation, Japanese.

Adverbial Clauses and Nominative/Genitive Conversion in Japanese

Hisako Takahashi
Stony Brook University

The goal of this paper is to provide a novel observation on the Nominative/Genitive Conversion (NGC) in Japanese by closely examining temporal adverbial clauses headed by a postposition –made ‘until’. The NGC is known as a Case alternation phenomenon in which nominative Case of the subject in a sentential modifier of a noun alternates with genitive Case. It has long been assumed in the literature that the presence of a head noun is crucial for licensing the genitive subject in the NGC construction (cf.Harada 1971, 1976, Bedell 1972, Miyagawa 1993, Ochi 2001, among many others), and that the external D head associated with a head noun licenses the genitive subject. On the other hand, Hiraiwa (2002) shows that some adverbial clauses, such as temporal adjunct clauses headed by –made ‘until’, allow the NGC despite the absence of a nominal head. Based on the empirical evidence, Hiraiwa (2002) argues that the NGC is not licensed by D associated with a nominal head, but the C-T-v-V head amalgamate as a reflex of the attributive (rentai) form of the predicate. Given Hiraiwa’s (2002) counterevidence against the D-licensing approach, Maki and Uchibori (2008) claim that a nominal head is simply unpronounced in the adverbial clauses provided by Hiraiwa (2002), but it actually present in the covert form in the structure. Namely, under Maki and Uchibori’s (2002) covert head noun analysis, it is assumed that the NGC is licensed by D, whether a head noun occurs overtly or covertly. In this paper, I argue against Maki and Uchibori’s (2008) covert head noun analysis, demonstrating that a head noun does not exist in the adverbial clause on the basis of three pieces of empirical evidence: (i) NGC with unergatives: the genitive subject can co-occur with an unergative verb when the temporal adverbial clause headed by –made ‘until’ contains an overt head noun –toki ‘time’, but once the head noun is unpronounced, the genitive subject turns out to be ungrammatical. (ii) Long/short distance readings of -made ‘until’: the temporal adverbial clause containing an overt head noun –toki allows both long-distance and short-distance readings of the postposition –made, while the –made adjunct clause lacking the head noun –toki has only the short-distance reading. (iii) Scope ambiguity: Since the first observation by Miyagawa (1993), it is well known that the genitive subject shows a scope ambiguity between a head noun and a QNP. This kind of scope ambiguity is also observed in the temporal adverbial clause headed by –made when the clause contains a head noun, but once the head noun is unpronounced, the scope ambiguity disappears. Based on those three arguments, I conclude that there is no covert head noun in temporal adverbial clauses headed by –made. The present study further suggests that none of the precious analyses of NGC can provide a sufficient explanation to the fact that the temporal adverbial clause never allows the NGC when it contains an unergative verb, although such unergative verb can occur in the other NGC constructions.

Keywords: nominative/genitive conversion, Japanese, temporal adverbial clauses, overt/covert head nouns, unergativity

Disjunction and Alternative Conditionals in Korean

Jiwon Yun
Cornell University

This paper investigates the conjunctive interpretation of a seemingly disjunctive expression in Korean: the particle -na connects two or more noun phrases to yield a disjunctive interpretation as in A-na B ‘A or B’, but if -na is repeated after the last disjunct, the sentence receives a conjunctive reading instead as in A-na B-na ‘A and B’. I propose that na-conjunctive nominals are actually adjunct clauses with pro-drop, whose underlying structure is essentially the same as what I call alternative conditionals. The claim is based on a number of similarities between the two constructions, na-conjunction and alternative conditional, specifically the similarities i) in their forms in that the element -na is repeated after each conjunct or each conditional clause, and ii) in the semantic representation in that both constructions involve a conjunctive reading.

The alternative conditional approach to na-conjunction also explains several properties of na-conjunction that distinguish it from ordinary conjunction, such as the incompatibility with case markers, exhaustivity, and distributivity. Furthermore, the proposed analysis can be extended to provide a novel account for another puzzling construction, namely wh-na, in which an interrogative expression is interpreted as a universal quantifier when the particle -na is attached to it. I adopt the view of Hamblin (1973) that a wh-word denotes a set of entities and propose that wh-na is an instance of na-conjunction, and accordingly, alternative conditional. It accounts for the source of the universal reading and distributive nature of wh-na.

Keywords: disjunction, alternative conditionals, wh-quantification, universal, Korean