Proceedings of the 5th Workshop on Altaic Formal Linguistics (WAFL5)
, ed R. Shibagaki and R. Vermeulen 2009
Identity Avoidance in Korean and Turkish
The consonant insertions in reduplication of Korean and Turkish are noteworthy in that they are similar but still distinct. The findings show that Korean reduplication is similar to Turkish emphatic reduplication in the sense that epenthetic consonants tend to differ from base consonants. In the reduplicated form of VCVC-CVCVC in Korean, the inserted consonant tends to be non-identical from either of the base consonants. In the reduplicated form of CVC-CVCV... in Turkish, the inserted consonant tends to be dissimilar from either of the first two base consonants. Both Korean and Turkish reduplication show the identity avoidance effects in that the epenthetic consonant in the reduplicant tends to be distinct from the base consonants. However, the choice of epenthetic consonants in Turkish reduplication operates at the segmental level, whereas the choice of inserted consonants in Korean reduplication operates at the featural level. The second C in the CVC prefix of Turkish emphatic reduplication is only taken from the fixed set of four consonants, whereas the inserted C in the Korean total reduplication is not taken from a fixed set of consonants; rather it is chosen by making reference to the features of neighboring consonants. On the other hand, the reduplication phenomena from Korean and Turkish are similar in that the inserted consonant is chosen from a set of consonants in each language, three most frequent consonants in Korean, in the case of the dictionary data, and four consonants in Turkish. However, in the reduplication experiment with nonce base forms, it was found that Korean speakers have a wider range of choices for epenthetic consonants, whereas Turkish speakers have a more limited choice in epenthetic consonants. Lastly, there is a clearly preferred consonant as an inserted C in the Turkish emphatic reduplication, whereas there is no clearly preferred segment as an inserted C in the Korean reduplication.
Keywords: consonant insertion, reduplication, Korean, Turkish, identity avoidance
Deontic Modality in Turkish: Pragmatic and Semantic Constrains
Zeynep Erk Emeksiz
Cross linguistic findings on the forms of modality reveal that epistemic and deontic modalities are very often encoded by the same forms as in the case of Turkish. The deontic modal affix-mAlI in Turkish is polysemous in deontic sense since it may be used to mark obligation, suggestion and desire. The source of polysemy is pragmatic one, and the interpretation of this mood marker depends on the schema evoked by the context. However, sentences with demoted agent subjects are monosemic having only obligation meaning. Similar to many other languages, -mAlI becomes ambiguous between deontic and epistemic interpretations when it occurs in non verbal sentences. This study suggests a feature based analysis to explain the meaning convergence from deontic to epistemic mood. Futurity is considered to be the basic feature dominating the deontic interpretation in Turkish.
Keywords: Turkish, deontic modality, agency, force, futurity
Nai Is Not a Truth Functional Negation
Without any verification, it is widely assumed that nai, sentential negation in Modern Japanese, is truth functional negation ¬. Observing that ‘negative’ sentences instantiated with nai cannot be described in terms of ¬, however, this paper casts doubt on this assumption. In particular, this paper observes such cases with nai through occurrences of hotondo, an approximation modifier similar to almost. When hotondo modifies the matrix predicate, it is accommodated only with those sentences whose matrix predicates have the scalar structures that identify degrees with the maximal or the minimal degrees on the associated scales. This paper calls those predicates that have such scalar structures ‘total predicates’. Nevertheless, hotondo modification can be compatible with ‘negative’ sentences with nai, regardless of the scalar structure that the matrix predicates have. Furthermore, in terms of ¬, the meaning of a ‘negative’ sentence with hotondo has a meaning that cannot be derived from its nonnegative counterpart. Given these, I claim that nai is neither truth functional negation ¬ nor an item derivable from ¬ such as an expression similar to ‘never/always not’. Alternatively, this paper proposes that nai is a total adjective. In fact, nai is morphologically an adjective since it exhibits adjectival inflection. In addition, it can be the predicate of an NP subject, and, just like a prenominal adjective, it can prenominally modify an NP. Furthermore, it exhibits several properties that a total predicate has. It is reported that, if it is used as the predicate of an NP subject, it has a similar meaning to ‘absent/missing’. Then, I claim that this meaning gives apparent negative meanings to sentences with nai: assuming that nai forms a constituent with the matrix predicate to which it attaches, a sentence with nai means that the property denoted by the matrix predicate is absent/missing in the evaluation world.
Keywords: nai, hotondo, truth functional negation, scalar structure, total adjective
On Definiteness and Specificity in Turkish
This paper proposes an alternative account to the widely discussed phenomenon of the nature of definiteness and specificity of Noun Phrases (NPs) in Turkish with respect to its relation to accusative case marking. In the literature the presence of accusative case marker in Turkish has been considered as an indicator of definiteness as well as specificity, which can appear only on nouns that have been mentioned in the previous discourse (Erguvanlı (1984), Dede (1986), Tura (1986), Enç (1991) and Kelepir (2001), among others). Based on Turkish data where accusative case can appear on NPs which are not necessarily available in the discourse, this paper argues that whether any given NP is definite or specific can be accounted for by the Givenness Hierarchy (GH) proposed by Gundel, Hedberg and Zacharski (GHZ, 1993) and later developed by Hedberg (1996) and GHZ (2001, 2004). The argumentation will be that for an NP to be definite, it only requires to be uniquely identifiable by the hearer rather than being obligatorily familiar or given previously in the discourse. On the other hand, partitive indefinite expressions are argued to be referential in that partitive constructions in Turkish do not necessarily denote specific entities.
Keywords: Definiteness, specificity, partitivity, Givenness Hierarchy, unique identifiability.
Differential Object Marking in (Khalkha-)Mongolian
In this paper, I will deal with the phenomenon of Differential Object Marking (DOM) in Mongolian. In this language some direct objects are overtly case marked and others not. In other languages that also exhibit this phenomenon different semantic and pragmatic factors have been identified which trigger it. In this paper I will try to give answers to the following two questions: (i) how relevant are these factors for DOM in Mongolian and (ii) do other factors play a role as well. The discussion is primarily based on my own native speaker intuition. Some results of an empirical questionnaire, which was constructed to clarify some of these questions, will also be discussed.
Keywords: Differential Object Marking, case, case alternation, Mongolian, Altaic
Agreement at the CP Level: Clause Types and ‘Person' Restriction on the Subject
Japanese and Korean (J/K) have been considered as non-agreement languages unlike English and other European languages (cf. Kuroda 1988, Fukui 1986). This paper, however, will show that J/K do exhibit rather extensive agreement processes between the subject and the predicate, once matrix phenomena are taken into consideration, such as Imperatives and V olitionals. W e will resort to Rizzi’ s (1997) recent proposal concerning on the layered CP system to account for these phenomena. That is, the sentential Force marked at the CP projection, such as Imp(erative), Vol(itional), requires a particular predicate form, which in turn requires (or agree with) a particular type of a subject, [+Addressee], [+Speaker], respectively. That is, agreement is observed between the subject, which can be null, and the predicate, which marks a particular sentential force, represented at the CP level.
Then, we will extend this analysis to two other null subject cases. One is the 1st person deletion phenomenon, which is allowed only at the matrix subject. It will be argued that it is different from an empty pronominal bound by a ‘null topic’, which is found not only at the matrix subject but also at the object or at the embedded subject. The other case is PRO in infinitives, whose interpretation is also tied with the structure of CP, as pointed out in Borer (1989). Our analysis of PRO provides an account for an interesting (and novel) fact of matrix arbitrary PRO in J/K. That is, PROarb is possible only when some Spec-CP blocks the control from the above. Thus, we can maintain our claim that a null subject is allowed only when Force of the CP requires a particular predicate form that agrees with a particular person of subject.
If the analysis proposed is on the right track, we seem to come up with a picture quite different from what has been assumed in the GB framework with respect to null subject phenomena in general. Null subject phenomena are relevant to what a head of the CP level specifies and even pro of null subject languages (NSLs) may be analyzed in a similar way, if the predicate (or Infl) is supposed to communicate with CP, as often has been assumed. Then, a null subject is not special to NSLs but is to be observed in more prevalent environments where the function of a CP is more apparent, namely, at the matrix level. The facts from Japanese and Korean show that a null subject (with a particular person feature) is possible when the predicate form clearly indicates the Force or the function of CP; hence, Agreement at the CP level.
Keywords: Imperatives, Layered CP system, Force, Null Subject, PRO
Two Indefinite Articles in Uzbek
Klaus von Heusinger & Udo Klein
In this paper we provide converging evidence both from a corpus study as well as from an online questionnaire that modern Uzbek has two indefinite articles 'bir' and 'bitta'. The corpus tudy revealed a significant difference in usage between 'bir' and 'bitta', to the effect that 'bir' was strongly preferred in the author’s narrative, whereas 'bitta' was preferred in direct speech. On the basis of this we hypothesise that 'bitta' has spread in modern colloquial Usbek to the expense of bir. This hypothesis was confirmed by the evaluation of an online questionnaire, which indicates that in certain contexts of use 'bitta' is judged more acceptable than 'bir'. Moreover, the acceptability judgements of sentences containing 'bitta' in predicative constructions show that theuse of 'bitta' is in the process of spreading to the last stage of development into an indefinite article. The development of 'bitta' to the expense of 'bir' may be the reason for the emergence of semantic constrasts between the use of 'bir' and 'bitta'.
Keywords: indefinite, article, Uzbek, numeral, classifier
On Right Node Raising
The aim of this study is to offer evidence from Right Node Raising structures in Turkish for Right Node Raising (RNR) as (PF-)Ellipsis (Wexler and Culicover 1980; Levine 1985, 2001; Kayne 1994) rather than RNR as ATB movement (Ross 1967; Bresnan 1974; Hankamer 1971; Hudson 1976) or RNR as Multiple Dominance (Phillips 1996; Wilder 1999, 2008). The evidence for the PF-analysis of RNR comes from (i) agreement facts in that a shared verb agrees only with the subject of the second conjunct, (ii) RNR cases where the shared element is a verbal element, which would require adjunction of a head to a phrase under the ATB-analysis, (iii) occurrence of verbs but not DOs as the shared element in complement clauses, which would be illicit under the ATB-analysis since rightward adjunction to complement clauses is illicit in Turkish and (iv) occurrence of shared DOs but not verbs from complement clauses in ‘matrix clause’-final position. All these properties are problematic for the ATB movement analysis, and property (i) is problematic for the Multiple Dominance analysis. The study also shows that RNR structures in Turkish cannot be accounted for in terms of LCA as in Wilder (1999, 2008). RNR in Turkish gives us the following generalization:
1. Right Edge Generalization (REG) In the configuration
[[A....X...] Conj. [B...X...]] X must be rightmost element in the conjunct B or the whole structure – including &P- before X can be deleted in A.
I argue that REG needs to be satisfied for the identical element in the first conjunct to be deleted.
Keywords: RNR, backward gapping, Turkish, PF-deletion, linearization
Non-Simultaneous Transfers, Case Domain Fusion and the Light Verb Construction
This paper deals with the light verb construction in Japanese. The light verb construction involves the light verb suru 'do' and an argument-taking noun called a verbal noun which is marked by the accusative case particle. Since Grimshaw and Mester (1988), the light verb construction has drawn much attention in recent literature mainly due to a paradoxical PF-LF mismatch in argument linking it exhibits. Given that the light verb suru 'do' is semantically empty, the arguments in the light verb construction are assigned θ-roles by a verbal noun. These θ-markings take place within the nominal phrase complement of the light verb suru 'do', which is headed by a verbal noun, given the locality condition of θ-marking, which requires that θ-marking should take place within the maximal projection of a θ-role assigner. This suggests that the arguments should be inside the nominal phrase complement. The arguments, however, may not be marked by the genitive case particle -no, i.e. the case marking of the nominal system. This suggests that the arguments may appear outside the nominal phrase complement. It has been widely assumed that while θ-marking is an LF- phenomenon, Case marking is a PF-phenomenon; the light verb construction exhibits the PF- LF mismatch. This paper proposes that the nominal phrase complement of the light verb suru 'do' functions only as an LF-phase but not as a PF-phase. I also propose "case domain fusion," arguing that when more than one case domain overlaps, "case domain fusion" must take place, where the notion of "case domain" is regulated by the Phase Impenetrability Condition. It is shown that our proposed analysis straightforwardly accounts for the paradoxical PF-LF mismatch in the light verb construction. It is also shown that various properties of the light verb construction like Case marking of an external argument, the ergativity constraint, the distribution of indeterminate pronouns, and resistance of "bare" verbal nouns to movement operations follow from our analysis. If our proposed analysis is on the right track, it presents evidence for the view of non-simultaneous Transfers, where a syntactic object can be transferred to a single interface (either only to the sensory-motor (S- M) interface or only to the conceptual-intentional (C-I) interface) during a derivation.
Keywords: Light Verb, PF-LF Mismatch, Non-Simultaneous Transfers, Phase, Case Domain
Standards of Comparison in Japanese
Based on the characteristics of degree comparisons, this paper surveys several sets of facts in clausal comparatives in Japanese. Specifically, an important issue in Japanese comparatives is whether a description of degree is usable as a standard of comparison. I discuss predicative comparatives, where the predicate of the comparative clause is a gradable adjective, arguing that they are available as long as derived degrees denote the same sorts of degrees. I then analyze amount comparatives in Japanese, concluding that comparative subdeletion constructions are readily derivable in Japanese. By laying out the examples of clausal comparatives, I conclude that degree comparisons are an option available in Japanese.
Keywords: Comparative Subdeletion, Amount Comparatives, Degree Comparisons, Standard of Comparison
Children’s Comprehension of Postverbal Objects in Turkish
F. Nihan Ketrez
Young children’s inability to assign wide scope to indefinite objects with respect to negation is attributed to the fact that the scope relationship is not directly mapped at the syntactic hierarchical structure (Lidz & Musolino, 2002): Direct objects occur in a lower position than negation in syntax. This proposal predicts that children should not have any difficulty with scope assignment when an object is in a higher position. Based on data from Turkish post verbal indefinite objects and preverbal accusative-marked objects, which are higher than negation, this study challenges the original c-command account and proposes a reformulation. Under the present account, children’s difficulty is attributed to their tendency to assign
narrow scope to objects that they treat as focus field items regardless of their location in the hierarchy. Because the focus field and everything that is associated with it are lower than negation under the clause structure adopted, children’s indefinites cannot scope over negation. In addition to its implications for the child grammar, the present study also shows that postverbal accusative marked indefinite objects are not different from their preverbal counterparts as far as the scope properties are concerned. What determines an object’s scope properties is the accusative case rather than the linear ordering of the constituents.
Keywords: Scope of indefinites, c-command, focus, Turkish, postverbal area
Subject Raising in Japanese
This paper presents a set of fresh data that allow us to evaluate the position of subjects in Japanese. It is shown that subject raising, which moves subjects to TP, is in most cases instantiated in Japanese, due to the pervasive requirement that a clause has at least one nominative argument. Nevertheless, we also argue that subjects are not raised if they are either assigned oblique marking or included in idiomatic sequences. This suggests that in Japanese, the EPP requirement of T is not ubiquitous, but rather, is determined depending on whether the clause has any nominative expression that serves as a full-fledged argument in the syntax.
Keywords: subject raising, focusing, clausal idiom, bakari, Japanese
Specificity and Partitivity in Some Altaic Languages
Jaklin Kornfilt and Klaus von Heusinger
The main goal of this paper is to show that partitivity is not an instance of specificity, but is an independent operation. We offer here a comparative study of some Altaic languages (three Turkic languages, with additional brief mention of a fourth, and of Standard Mongolian), which all exhibit Differential Object Marking. Facts based on DOM were used previously in the literature to support the claim that partitivity is an instance of specificity, or even a stronger claim, namely that specificity equals partitivity. Our own comparative Altaic data show that the case suffix for accusative is sensitive to specificity, however only when it is not triggered by formal reasons. We present instances of partitive constructions with non-specific subset expressions, of two types: 1. Direct object partitives without accusative marking, interpreted as non-specific; 2. Similar partitives without lexical head nouns and with agreement (or other) markers in need of overt accusative; those can be interpreted as non- specific. The data show that the case suffix expresses specificity, and not partitivity. We further show that the agreement marker can function as a nominal head of a nominal phrase, when a lexical noun is absent to fill the head position. We characterize this semantically as promoting non-nominal phrases to the status of referential ones.
The languages we have studied here differ (among other properties) with respect to classifiers—in particular, with respect to whether they have [+human] classifiers or not. In addition, one language (Kirghiz) substitutes the agreement marker in its function as a filler of the partitive’s nominal head by a different marker: a morpheme expressing a set. Here, the agreement marker is used to express specificity, given that its presence is not required for formal reasons.
Specificity is different from partitivity. It expresses a functional dependency of the indefinite expression from another referential expression (and not the subset relation). The formal reconstruction of this view of specificity states that a specific NPi signals that the associated index i is linked by a salient function (or relation) to another index j from the very sentence φ. One can instantiate this salient function in different ways: For epistemic specific indefinite the function is to be translated into “j has cognitive contact with i” or “j knows / can identify i”.
Keywords: Specificity; Partitivity; Differential Object Marking; Turkic languages; Referential anchoring
Datives in Japanese
Richard Larson & Naomi Harada
Datives in JPN (and other V-final lgs) raise serious questions for syntactic projection. ENG PP datives like (1a) are typically assigned layered VPs with theme above goal (1b). Inverting head order in vP and VP for the equivalent sentence in head-final JPN yields structure (2a), in which theme precedes goal. But the natural order is (2b), where goal precedes theme. This situation obtains not only with JPN possessor datives, but also with simple directional datives. We expect order (3a) under projection counterpart to English, but the unmarked order is (3b).
(1) a. John gave a book to Mary
b. [vP John [v’ v [VP a book [V’ give [PP to Mary]]]]]
(2) a. [vP Taroo-ga [v’ [VP hon-o [V’ [PP Hanako-ni ] ageta]] v ]]
Taroo-NOM book-ACC Hanako-DAT gave
b. Taroo-ga Hanako-ni hon-o ageta ‘Taroo gave a book to Hanako’
(3) a. Taroo-ga nimotu-o Tokyo-ni okutta (marked order)
Taroo-NOM package-ACC Tokyo-DAT sent
‘Taroo sent a package to Tokyo’
b. Taroo-ga Tokyo-ni nimotu-o okutta (unmarked order)
Dative order in (2)/(3), is in fact part of a bigger puzzle with -ni. In ENG transitive and intransitive locatives (4a)/(5a), theme projects higher than location (4b)/(5b), hence with parallel projection for JPN we expect theme to precede loc (4c)/(5c). But in both cases loc must precede theme. Similar points hold for JPN possessives (6).
(4) a. John put a book on the cart.
b. [vP John [v’ v [VP a book [V’ put [PP on the cart]]]]]
c. [vP Taroo-ga [v’ [VP hon-o [V’ [PP nidai-ni] oita]] v ]]
d. Taroo-ga nidai-ni hon-o oita (unmarked order)
(5) a. John is in Tokyo.
b. [VP John [V’ is [PP in Tokyo]]
c. [VP Taroo-ga [V’ [PP Tokyo-ni] iru]
d. Tokyo-ni John-ga iru. (unmarked order)
(6) Taroo-ni hon-ga aru
Taroo-DAT book-NOM be ‘Taroo has a book’ (unmarked order)
We argue that these data reflect no deep differences in projection between ENG and JPN, but rather a specific fact about -ni. In all the cases above, ni-phrases occur higher than a structurally-case-marked DP - either DP-o or DP-ga. We propose this is an agreement phenomenon. Adopting Pesetsky & Torrego’s (2004) theory of feature valuation, we propose that: (i) ni is not a case-marker, nor a case-probe; and (ii) ni-phrases bear an unvalued, uninterpreted case feature and can obtain case by agreement via a structural case probe (v,T). If DP-ni occurs below a structural case probe, but above an element valued for structural case. DP-ni can agree with the probe before search terminates, and be valued (7a). But if DP-ni occurs below such a probe, and below the element valued for structural case, search will terminate before DP-ni is reached, and its case feature will be unvalued, yielding a crash (7b).
(7) a. Probe ... DP-ni AGREE DP–NOM/ACCSEARCH TERMINATES
b. Probe ... DP–NOM/ACCSEARCH TERMINATES *DP-ni
We propose that Japanese ni-phrases merge in the low positions found in ENG and expected under θ-theory, but that they A-scramble to the edge of vP/VP, achieving the proper configuration for agreement. Thus in datives and transitive locatives, ni-phrases occur between the accusative probe v and o-marked object. In intransitive locatives and possessives, they occur between the nominative probe T and ga-marked subject.
Keywords: dative Case, double object constructions, concordial Case, Case agreement, Japanese
On Accusative Wh-Adjuncts in Japanese
This paper discusses the properties of the Japanese Accusative wh-adjunct “nani-o (what- Acc)” in sentences such as “Kare-wa nani-o sawai-dei-ru no? (Lit. What is he making a noise?/ Meaning. Why is he making a noise?)” Despite the usual assumption that it is identical in meaning with the standard reason adjunct “naze (why)”, I point out that it has some peculiar properties that are not shared by “naze”. For example, questions with an Accusative wh-adjunct have an animacy restriction on its subject, and such questions have a special connotation that the speaker is surprised at or disapproval of the subject’s unexpected behaviour. Also, Accusative wh-adjuncts are incompatible with sluicing, unlike other wh- adjuncts (Ochi 1999). I gave an account of these facts by claiming that Accusative wh- adjuncts are licensed in a higher position than standard reason adjuncts such as “naze”. I illustrate an analysis where “nani-o” is base-generated in the functional projection FP, which is related to speaker’s illocutionary force. Based on Fox and Lasnik’s (2003) Parallelism account of sluicing, I claim that an Accusative wh-adjunct does not license sluicing because it is not parallel in position with the reason adjunct in the antecedent clause. By clarifying the syntactic positions of different types of wh-adjuncts, I aim to contribute to the typological study of adjuncts. I also point out that Accusative wh-adjuncts share some other properties with English aggressively non-D-linked expressions (i.e. “wh-the-hell”). I review Den Dikken and Giannakidou’s (2001, 2002) analysis of “wh-the-hell” as a polarity item and
show that their diagnostics of polarity items also apply to the Japanese Accusative wh- adjunct “nani-o”. First, when “nani-o” co-occurs with a modal, negative answers are forced. Second, when embedded under a veridical predicate such as “know”, the Accusative wh- adjunct is degraded. Third, a pair-list reading is impossible in multiple wh-questions involving an Accusative wh-adjunct. These data indicate that Accusative wh-adjuncts might be best analyzed as a kind of polarity item.
Minimalism and Information Structure: A Case of Ellipsis in Japanese
This paper is an attempt to provide a case study to bridge between “narrow syntax” (Chomsky 1995, 2004, etc) and functional approach to information structure (Kuno 1978, 1995, etc). It has been observed that ellipsis and word order variations in Japanese are governed by the principles of discourse. For instance, [VP object adjunct V] is a “marked” word order in Japanese, and the adjunct, the immediately preverbal element in this case, necessarily carries new information focus. Hence, the ellipsis of the immediately preverbal element alone is not permitted even when a proper antecedent clause is given in the parallel structure. On the other hand, with the corresponding “unmarked” order [VP adjunct object V], it is not always the case that the immediately preverbal element is the new information focus: hence, either the object (the immediately preverbal element) alone or the adjunct alone can be deleted, given an appropriate antecedent clause. However, the specific computational mechanism to compare and evaluate “marked” and “unmarked” structures has not been very explicit in the traditional discourse/functional approaches.
The present paper argues that reference set computation in the sense of Reinhart (2006) is crucially involved here, which provides an explicit mechanism, as an interface strategy, to compare “marked” and “unmarked” structures. This system neatly accounts for the fact that a “marked” structure is allowed only when it gives a focus interpretation which the comparable “unmarked” structure cannot.
Narrow syntax is governed by the principle of computational efficiency (‘economy of derivation’) and thus derivations proceed in the most economical way. Each syntactic movement (internal Merge) must be motivated, typically by feature checking. In this minimalist picture, the theoretical characteristics of so-called optional movements (among which is Japanese scrambling) has be a locus of controversy. The implication in this study is that Japanese scrambling (at least, the VP-internal adjunct-object permutation) is driven by focus/interpretation reason (deeply related to conceptual-intentional system), and thus not a “free” syntactic operation in the technical sense. In the strongly derivational perspective of syntax, “Minimal Ling Condition” of derivation (computational efficiency) is integrated in each derivational step and no comparison of competing derivations is allowed. This study, however, gives a support of Reinhart’s idea that, at least under restricted conditions, comparison of derivations is possible.
Keywords: information structure, ellipsis, reference set computation, interface strategy, scrambling
Subject Positions in Old Japanese
This paper investigates the syntactic difference between ga and no, which in old Japanese (Manyoushuu, 645-759 A.D.), mark subject or possessive. Available data suggests: (a) ga marks a familiar person subject (or possessive) -pronouns (a/wa (I), na (you)), sisters, brothers, and lovers are ga marked-, and no marks the rest; (b) ga marks ta (who), and no marks the remaining wh-expressions; (c) ga marks nominalized subjects without the complementizer; and (d) DP-no is possible in tsutsu (while) clauses. Among these, this paper concentrates on the subject cases. From (a) and (b), I assume that DP in DP-ga is associated with a kind of person feature. From (c), since nominalized subjects without the complementizer are not a regular DP, I assume they are subjects with a certain feature. From (a) to (c) I assume that no behaves like a nominative marker, and that DP-no is a subject without an extra feature. On the other hand, DP-ga is subject with an extra feature. (d) may suggest a structural difference between DP-ga and DP-no: the subordinate tsutsu clauses may be small, which accept DP-no subjects and exclude DP-ga subjects. In other words, DP-ga is structurally higher than DP-no. Combining the features and the structure, it is possible to assume that DP-no is a subject without any extra feature that remains in situ, whereas DP-ga is a subject with an extra feature and, as such, it has to raise. This possibility is acceptable if the VP Internal Subject Hypothesis is true in both old Japanese as well as modern Japanese, hence DP-no remains in situ (Spec, vP). On the other hand, DP in DP-ga is associated with certain features (in (a) and (b), with the person feature, and in (c), with a kind of feature). Under C-T probe, those features can be C related, and inherited to T. As a result, DP-ga has to raise to Spec, TP. It is also possible to assume that the features may be EPP related, which are also C property, and inherited to T under C-T probe. Furthermore, the analysis in this paper suggests that Spec, TP is optional, when DP-no remains in Spec, vP. This result is supported by recent studies on the optionality of Spec, TP in other languages, including modern Japanese.
Keywords: old Japanese, subject, ga, no, the person feature.
The Comparative Syntax of Double Object Constructions in Japanese, Korean, and Turkish
Andrew Simpson, Heeju Hwang, and Canan Ipek
Among verb-final languages which permit scrambling/apparent free word order, there have been a number of different claims about the syntactic structure of ditransitive constructions and whether a unique underlying/neutral word order can be identified. With regard to German, there is a consensus of opinion that the ordering Indirect Object > Direct Object (IO>DO) is basic, following detailed investigations of markedness restrictions governing the re-ordering of IO>DO as DO>IO. Concerning Turkish and Persian, however, there are reports that DO>IO is the basic, underlying order, based on patterns of anaphor construal, weak crossover, prosody, and various other syntactic properties. Finally, a dual position has been proposed in work on Japanese carried out by Shigeru Miyagawa (Miyagawa 1997). On the basis of binding and floating quantifier patterns it is argued that both IO>DO and DO>IO orders can be base-generated as underlying orders. Given such different conclusions about the structure of ditransitives, it is not clear whether there is any kind of universal hierarchical ordering of direct and indirect objects in verb-final (and other) languages, or if languages are open to parametrization in the structuring of double object constructions and the establishment of a neutral linear ordering of goal and theme arguments. The present paper seeks to add a further perspective on this universalist-particularist debate with a comparative
study of ditransitives in Japanese, Korean and Turkish. The paper reports the first results of a larger ongoing project on ditransitives in the three languages and attempts to establish basic patterns relating to (a) the effect of (in)definiteness and (in)animacy on goal/theme ordering, (b) binding and reconstruction phenomena, (c) patterns with floating quantifiers, and (d) relative quantifier scope. The paper also considers a recent interesting analysis in Miyagawa and Tsujioka (2004), based on patterns in Japanese, that there may be two distinct goal positions present with many ditransitive verbs (a goal and a non-recipient end-point location), and investigates the potential consequences of such a possibility in Korean and Turkish.
Keywords: ditransitives, quantifier scope, PF movement, reconstruction
Invisible Degree Nominals in Japanese Clausal Comparatives
Comparative sentences in Japanese can be descriptively divided into two types, phrasal and clausal comparatives, and there are two competing views about their underlying structures. searchers all agree that phrasal comparatives are underlyingly possible in Japanese, but whether underlyingly clausal comparatives exist or not has been debated. This paper offers three pieces of empirical evidence for the claim that all comparative sentences in Japanese are underlyingly phrasal, and the clauses in clausal comparatives are actually relative clauses rather than complement clauses. We present syntactic, morphological and semantic data. Specifically, we point out that Japanese comparative clauses share the following properties with relative clauses: (i) optional genitive marking on the subjects is licensed (nominative/genitive conversion), whose distribution is generally confined to relative clauses; (ii) the copula takes the adnominal form, which is one of the few predicates that retain a distinct adnominal form in the present day Japanese; and (iii) the stative readings of certain change-of-state predicates are licensed, which is likewise only possible in relative clauses. The paper further claims that superficially clausal comparatives involve a silent degree nominal as the head noun of the relative clause, which accounts for the gradient judgment patterns of Japanese clausal comparatives. We present English data as well as Japanese ones to contrast the two languages. Concretely, simple clausal comparatives with the same scalar predicate in the matrix and embedded clauses are ruled out in Japanese unlike in English. Also, both predicative and attributive subcomparatives, which are comparatives with different scalar predicates in the two clauses, are utterly ungrammatical in Japanese, while only the latter is unavailable in English. On the other hand, amount subcomparatives are marginally possible in Japanese, which are perfectly grammatical in English. This difference between two languages suggests that an equal treatment of clausal comparatives in these languages is not viable. Instead, it is shown that positing a silent degree noun that denotes the 'relevant' degree as the head of the comparative phrase straightforwardly accounts for the Japanese data. Possible typological consequences of this analysis are discussed at the end. In particular, it has been established that English allows underlyingly clausal comparatives, although the availability of underlyingly phrasal comparatives has been questioned. Therefore, the analysis of this paper commits to the view that there are at least two types of languages, namely, those that allow underlyingly clausal comparatives and those that do not. However, this claim is not implausible at all given the existence of languages such as Hindi-Urdu and Korean, where only surface phrasal comparatives are grammatical. The paper further suggests the possibility that the availability of clausal comparatives is correlated with the presence of overt wh- movement. Justifications for this speculation include the fact that genuine clausal comparatives in certain languages involve a wh-related morphology and their semantics is generally considered involving an operator movement similar to wh-movement.
Keywords: comparative, degree noun, phrasal comparative, clausal comparative, Japanese
Two Types of Resultatives in Japanese
In his semantically grounded typology of resultative constructions, Washio (1997, 1999) classifies resultatives into two broad types, strong resultatives and weak resultatives. Strong resultatives are the ones in which the meaning of the verb and the meaning of the adjective are completely independent of each other, while weak resultatives involve adjectives whose meanings are closely related to the meaning of the verbs. Elaborating the notion of affectedness (Simpson 1983, Jackendoff 1990, Goldberg 1995 etc), Washio further makes finer sub-divisions of each type of resultatives.
In this paper, I provide support to the sub-divisions of weak resultatives, which I call wipe-resultatives and break-resultatives, by presenting additional evidence from Japanese. In the literature it is argued that at least in English resultative constructions, different types of verbs in terms of telicity combine with different types of adjectives with respect to a scale (Wechsler 2005). Wechsler observes that in English, atelic durational verbs permit a closed maximal-endpoint scale adjective as a resultative predicate, whereas telic verbs do not seem to impose the same restriction. In Japanese, wipe-resultative constructions seem to require a maximal-endpoint adjective, while break-resultative constructions do not have the same restriction. Second, when modified by mata ‘again’, break-resultatives yield both repetitive reading and restitutive reading, while wipe-resultatives yield only repetitive reading. Von Stechow (1996) and Beck and Johnson (2004) argue that the ambiguity between a repetitive reading and a restitutive reading of again can be explained if again adjoins to a different syntactic position. Assuming this, the fact that a break-resultative but not a wipe-resultative is ambiguous between the two readings implies that the two types of resultatives involve different structures. Under Washio’s semantic approach, it is not clear how the differences between break-resultatives and wipe-resultatives are accounted for in a principled manner.
I present a syntactic analysis that explains the differences between the two types of resultatives. Ramchand (2008) argues that the semantic components are directly represented in the syntax structure. Elaborating Ramchand, Son and Svenonius (2008) propose that the syntactic heads responsible for the meaning of resultatives involve ResP and PredP. Assuming ‘Exhaustive Lexicalization’ (Fàbregas 2007), Son and Svenonius further argue that the cross-linguistic differences with regard to the availability of resultatives depend on whether a language contains a lexical item that licenses ResP and PredP. Following Son and Svenonius, I propose that in Japanese the verb lexicalizes Proc, Res, and Pred in a break- resultative construction. While in a wipe-resultative construction, the verb lexicalizes both Proc and Pred but not Res. Since the Res head which denotes a “telos” is absent in the wipe- resultative construction, the result state meaning is created by the bound path complement. Thus, the wipe-resultative construction requires a closed maximal-endpoint scale adjective, which contains an endpoint, to be a resultative predicate. With respect to mata modification, unlike break-resultatives, wipe-resultatives lack a restitutive reading. In my analysis, there is no ResP to which mata can adjoin in the wipe-resultative constructions. The wipe-resultative with mata modifier therefore contains no restitutive reading.
Keywords: Resultative constructions, Japanese, weak resultatives, again modification, and the first phase syntax.
Verbal Nouns of Different Sizes in the Light Verb Construction in Japanese
Kaori Takamine & Naoyuki Yamato
The aim of this work is a reanalysis of the light verb construction in Japanese, translating Grimshaw & Mester’s (1988) Argument Transfer analysis to the current first phase syntax (Ramchand 2008) approach. We re-examine with a broader range of data syntactic diagnostics that Grimshaw & Mester use to distinguish the heavy verb suru ‘do’ from the light verb suru and show that the heavy vs. light distinction does not capture the facts, but rather that verbal nouns in the light verb construction show three-way split behavious, which is unexpected under Grimshaw & Mester’s approach, with respect to syntactic operations such as topicalization, scrambling and relativization. Furthermore, our data also shows that verbal nouns in the target construction must be classified into object-denoting nominals and event-denoting nominals with regard to the aspectual properties of them. Adopting Ramchand’s first phase syntax approach, according to which the vP is decomposed into fine- grained functional projections, each of which corresponding to a sub-event component, we argue that the distinction between the object-denoting verbal nouns and the event-denoting verbal nouns is accounted for by different positions the verbal nouns occupy, namely that the undergoer-type of verbal nouns generated in Spec VP (Ramchand’s Spec ProcP) behave as object-denoting nominals while the rhematic verbal nouns are eventive nominals co- describing the event together with the head verb. With regard to the three-way split syntactic behaviours of the verbal nouns, we argue, adopting Déchaine & Wiltschko’s (2002) decomposition of pronouns, that the differences are attributed to different sizes of the verbal nouns: (i) nouns need the D-projection in order to undergo topicalization or scrambling; and (ii) assuming Kayne’s (1994) analysis of N-final relative clauses, relativization involves remnant movement which is constrained in a way that Collins & Sabel (2007) state, i.e., it is possible only if the gap contains Φ-features.
Keywords: first phase syntax, Japanese, light verb, pronoun decomposition, remnant movement
Causativization and Aspectual Composition
Sergei Tatevosov & Mikhail Ivanov
We address a puzzle of the aspectual composition (AС) in indirect causatives in Tuba dialect of Altai. In Tuba, both English-type AC and Russian-type AC are attested. The former obtains in aspectless clauses where the cumulativity/quantization status of incremental arguments (in the sense of Krifka 1989, 1992, 1998) determines that of verbal predicates, cf. examples like He ate two apples in five minutes and He ate soup for five minutes in English. The latter occurs in perfective clauses whereby the aspectual semantics restricts the interpretation of an incremental argument, cf. Russian examples like On s”jel sup za pjat’ minut ‘He ate-PFV the || *∅ soup in five minutes’.
The puzzling characteristic of Tuba is that in the indirect causative configuration (one involving causativization of transitive verbs), English-type AC is only available despite the presence of perfectivizing morphology. Our account for this puzzle consists of the three essential ingredients: semantics for uninflected vPs, semantics for the perfectivizing operator, semantics for the causative.
Assuming the event semantics framework, we propose that depending on the interpretation of incremental DP arguments, uninflected vPs in Tuba denote either cumulative or quantized event predicates. Elaborating on Krifka 1992, we suggest that the perfective morpheme in Tuba is an identity function with a presupposition that an event predicate it
modifies is quantized. Merging with vP, the functional head hosting the perfective morpheme ‘filters out’ non-quantized event predicates generated at the vP level. Following much recent work on causativization, we suggest that the causative morpheme merges as a v head and that with transitive verbs, this yields a configuration consisting of two vPs. Semantically, the causative morpheme introduces a causing event and its participant. The crucial effect the causative v has on its complement vP is existential binding of the event variable in the semantic representation of the latter, making it inaccessible for further semantic operations. Therefore, in causative configurations the perfective operator only takes scope over the ‘outer’ causing event. Being existentially bound, events denoted by the inner vP escape the scope of the perfective: even if that vP originally denotes a non-quantized event predicate, due to existential binding this cannot lead to the failure of the presupposition associated with the perfective functional head. In effect, the perfective only checks properties of event predicates available locally – those denoted by its complement vP, but not those associated with more deeply embedded eventive heads. This explains why causativization replaces the Slavic-type AC with the English-type AC in perfective clauses in Tuba.
Evidence from Tuba, where Russian-type AC originates at the level of functional structure projected above vP, while at the vP-level English-type AC only exists, can shed a new light on Russian-type AC in general. A theoretically attractive cross-linguistic hypothesis that emerges at this point is that in all languages with Russian-type AC, it emerges at later stages of syntactic derivation, when aspectual functional structure is projected. At the vP level, languages universally have English-type AC.
Keywords: aspectulal composition, perfectivity, causative, syntax-semantics interface
Person Restriction in CP-Domain in Japanese
This paper argues for assuming at least three independent heads, C1, C2, and C3, in the CP-system in Japanese. Two of them, C1 and C2, are morphologically realized with modal forms, Epistemic- modals (henceforth E-modals) and Utterance-modals (henceforth U-modals). The right-most- peripheral head, C3, is realized with Sentence Final Particles (henceforth SFPs). We propose a syntactic structure (1) in terms of the following three points: (i) embedding tests, (ii) person restriction on matrix subjects, and (iii) the EPP-satisfaction on C.
(1) [CP3 [CP2 [CP1 [TP [v*P [VP V] v* ] T] C1E-modal ] C2 U-modal ] C3SFP]
This provides evidence for Split CP hypothesis in a sense of Rizzi (1997) by means of Japanese modal forms and their person restriction phenomena on subject NPs. We claim that distribution of Japanese modals reveals a rich discourse-interface system of CP-domain.
We explore properties of Japanese modality intentionally focusing our attention on their forms and syntactic positions. First, we divide Japanese modal forms into G(enuine)-modals and Q(uasi)-modals and we treat the G-modals only in this paper, because the formal discrimination between the G-modals and the Q-modals plays a significant role in distilling the genuine characteristics of Japanese modality from diverse modal expressions. Furthermore, the G-modal forms are further divided into two types, E(pistemic)-modals and U(tterance)-modals.
Traditional Japanese grammarians observed that Japanese has an agreement-like phenomenon, that is, person restriction on subject NPs (Nitta 1991) as in (2). (2) shows that the 2nd person subject sounds awkward in matrix clauses, but this awkwardness is removed when the (a)- sentence is embedded.
(2) a *Kimi-wa gakko-e iku-daroo.
You-top school-to go-will
‘You’ll probably go to school.’
b. [ Kimi-wa gakko-eiku-daroo]-ga boku-wa ik-anai.
[ You-cont school-to go-will]-though I-cont go-neg
‘You’ll probably go to school, but I won’t.’ (Nitta 1991)
We further explore the following syntactic properties of Japanese modal forms: (i)embedding in Mikami’s (1973) C-type conjunction clauses, (ii)co-occurrence with Topic –wa. The results of our investigations are summarized as Table 1 given in (3).
(3) Table 1: Syntactic properties of the two types of Japanese modals
We will show that split CP system gives a unified account for all of the properties of Japanese modals given in Table 1 Finally, we claim Sentence Final Particles should not be treated as a typical U-modal form, but as an independent head, that is, C3, in the CP-system.
Keywords: CP-system, person restriction, modal forms, Utterance-modals, Genuine-modals
Topics, Contrast, and Contrastive Topics in Japanese
The standard characterisation of contrastive topics in Japanese is that they are marked with the so-called topic marker wa, bear an emphatic stress and optionally move to clause-initial position (e.g. Kuno 1973, Saito 1985, Hoji 1985). This paper presents new arguments that there is in fact no such optionality: contrastive topics must appear in clause-initial position and those contrastive wa-phrases (i.e., wa-marked phrases bearing an emphatic stress) that appear in-situ do not function as contrastive topics. As a consequence, the particle wa should not be considered a marker for topic. Concentrating mainly on objects, rather than subjects, I present four arguments for this proposal.
First, examining particular discourse contexts that require a contrastive topic on independent grounds, I show that a contrastive wa-phrase that is to function as a contrastive topic must appear in clause-initial position and cannot remain in-situ. Second, there are interpretive differences between contrastive wa-phrases in-situ and those moved to clause- initial position. They both implicate contrast, but only the latter are interpreted in addition as what the rest of the sentence is about, i.e., as ‘topic’ in the sense of Reinhart (1981). Moreover, in cases they contain a numeral quantifier, those in-situ allow for a non-specific as well as specific reading, while those displaced to clause-initial position can receive only a specific reading. Third, I examine three different contexts in which a contrastive wa-phrase appears in-situ. In these contexts, the wa-phrase cannot optionally move to clause-initial position. Thus, contrastive wa-phrases in clause-initial position and those in-situ are licensed in distinct discourse contexts. The final argument concerns a prediction with respect to the syntactic distribution of topics that derive from independent considerations at the interface between syntax and information structure. It is demonstrated that contrastive wa-phrases in clause-initial position show the predicted behaviour, while those in-situ do not.
Thus, contrastive topics in Japanese must occupy clause-initial position and do not have the option to remain in-situ. This is in fact the syntactic distribution generally assumed for topics that are not contrastive (‘thematic’ wa-phrases in Kuno’s (1973) terminology). The current proposal therefore has the desirable consequence that it provides a straightforward and uniform account of the syntactic distribution of topics in general in Japanese. Moreover, it provides a strong empirical basis for the idea often only suggested in the literature that the particle wa, when used contrastively, does not function as a canonical topic marker (Kuroda 2005, Hara 2006, Oshima 2008).
Keywords: topic, contrast, contrastive topic, Japanese, wa
Mood in Abstract Complementizer
The question whether comparative clauses have a negative operator (Jesperson 1917; Ross 1969; Seuren 1973; Klein 1980; Larson 1988) or not (Stechow 1984; Rullmann 1995; Kennedy 1997) has divided literature into two parties. The debate seems irresolvable because the negative approach provides empirical evidence such as English dialects with overt negative nor and French ne comparatives, whereas the non-negative approach provides more straightforward semantics for Russell’s ambiguity or differential readings. Thus the empirical variations among comparative constructions are captured by neither approach.
However, the current study proposes a novel dichotomy of comparative constructions between ‘rhetorical comparatives’ and regular degree comparatives. Just like rhetorical questions, we argue that rhetorical comparatives are used when a speaker wants to express an illocutionary force of opposite polarity whereas regular degree comparatives merely establish an ordering between two objects. The crucial division is based on the notional moods (à la Portner 1997), i.e. the irrealis for rhetorical subtype and the realis for degree subtype, which in turn predicts systematic differential properties for each comparative type. In particular, we propose that a standard degree (d2) of a rhetorical comparative is posited in a set of possible worlds and hence has attributive and non-specific properties. In contrast, d2 of a degree comparative refers to a real degree in a particular world and yields a referential and specific interpretation.
By positing distinct semantic and pragmatic properties for degree and rhetorical comparatives, the current proposal leads us to predict the systematic asymmetries between comparatives with and without irrealis mood-indicators such as NPIs or modals. It thus becomes clear why the previous negative approaches are claimed to be descriptively adequate for data with NPIs while the non-negative approaches are not, as von Stechow (1984) evaluates comparative theories. Further important theoretical consequences of the current dichotomy analysis are accounting for negativity-related puzzles such as the quantificational variability and the NP- vs. S-comparative asymmetry.
Keywords: rhetorical comparative; degree comparative; notional mood; realis; irrealis