Something visible in Japanese

A number of languages of the world show a striking change in word order when what is modified is an indefinite pronoun: the modifier necessarily appears postnominally (e.g. something visible vs. *visible something). English and French are examples of this ordering pattern. Japanese, a head-final language with exclusively prenominal modifiers, nevertheless exhibits the same order as English and French when indeterminate phrases (indefinite pronouns) are modified. I argue that although the modifier in such structures may appear to be a postnominal modifier of the preceding indeterminate phrase, it is actually a prenominal modifier of the following light noun n. I further propose that the English/French-type construction has the same underlying structure as the Japanese-type construction and that apparent differences follow from a difference in the composition of so-called indefinite pronouns: an indeterminate system versus a light-noun system.


Introduction
A number of languages of the world show a striking change in word order when what is modified is an indefinite pronoun: adjectives necessarily appear postnominally and cannot appear prenominally (see Emonds 1985;Abney 1987;Hulk & Verheugd 1992;Kishimoto 2000;Larson & Marušič 2004;Leu 2005;Cardinaletti & Giusti 2006;Roehrs 2008;Marušič & Žaucer 2009;Cinque 2010, among others). Examples (1a) and (1b) show that the English adjective interesting must precede the ordinary noun book. On the other hand, examples (1c) and (1d) show that the same adjective must follow the indefinite pronoun everything.
In this article, for convenience, I will call syntactic constructions of the type seen in (1d), where an adjective follows an indefinite pronoun, indefinite-pronoun structures. In the literature, while indefinite-pronoun structures have often been studied in headinitial languages, they have not been described for Japanese, a head-final language with exclusively prenominal modifiers. Surprisingly, Japanese too exhibits the order in which a modifier follows an indeterminate phrase (the term I will use instead of indefinite pronoun). 2 Examples (3a)-(7a) illustrate this with five different existential indeterminate phrases (existential quantifiers). 3 (3) 'Someone' a. Dareka himasoona hito-o yoboo. someone free.looking ln-acc invite 'Let's invite someone who looks free.' b. Himasoona dareka-o yoboo. free.looking someone-acc invite 'Let's invite someone who looks free.' (4) 'Something' a. Nanika karai no-o tabe-tai. something spicy ln-acc eat-want 'I want to eat something spicy.' 2 In this article, I will eventually demonstrate that the "indefinite pronouns" in (3a)-(7a) are not syntactically nominal but rather QPs. Although it has been customary in the literature to call them indefinite pronouns (Takahashi 2002;Shimoyama 2008;Kuroda 2013, among others), I will use the term indeterminate phrase to refer to a phrase consisting of an indeterminate and a quantificational particle. 3 Prenominal modification (the (b) examples) sounds better if the indeterminate phrase is interpreted as specific (see Nishigauchi 1990 andHirose &Suzuki 2009). I suggest that when an indeterminate phrase is interpreted as specific and/or is directly modified, as in the (b) examples, it is a full-fledged DP, unlike the indeterminate phrase that appears in the (a) examples, which is a QP and is not directly modified. See Section 4 for relevant discussion. For some reason that I do not understand, doreka (5b) does not readily accept a prenominal modifier. It might be due to its D-linked specificity.
Likewise, replacing the indeterminate phrases in (3a)-(7a) with ordinary noun phrases, as in (9), again results in ungrammaticality. The exceptional "postnominal" word order is only allowed with indeterminate phrases.
(9) a. Nanika karai no-o tabe-tai. something spicy ln-acc eat-want 'I want to eat something spicy.' b. *Tabemono/raamen karai no-o tabe-tai. food/ramen spicy ln-acc eat-want (Intended: 'I want to eat spicy food/ramen.') Two main questions about indefinite-pronoun structures that need to be addressed are: (10) a. Why do modifiers appear to the right of indefinite pronouns/indeterminate phrases, while they are prenominal with ordinary nouns? b. What is the structure of the indefinite-pronoun structures?
It is important to observe that all the data above on indefinite-pronoun structures in Japanese involve a light noun at the right periphery-no 'thing', hito 'person', tokoro 'place', toki 'time', etc.-which is obligatory. Thus, the key to understanding the wordorder permutation is the syntax of light nouns. As I will lay out in the next section, some recent studies on indefinite-pronoun structures have tried to show that the apparently postnominal modifier is underlyingly a prenominal modifier of an invisible nominal (see Leu 2005;Roehrs 2008). I will argue that in Japanese, too, a modifier that follows an indeterminate phrase is actually a prenominal modifier of the following light noun. (11) A modifier that follows an indeterminate phrase in indefinite-pronoun structures in Japanese does not modify it but modifies prenominally the light noun that appears at the end of the entire noun phrase.
In this sense, Japanese provides visible evidence for a structure invisible in other languages, and our investigation reveals the existence of a universal light-noun structure in indefinite-pronoun structures. Furthermore, I will show that indeterminate phrases in Japanese, when they occur in certain structures (including indefinite-pronoun structures), are not "nouns" syntactically. I will demonstrate that Japanese-type and English/French-type indefinite-pronoun structures have the same underlying structure and that syntactic differences follow from a difference in the composition of so-called indefinite pronouns: an indeterminate system versus a light-noun system.

Overview of previous approaches
Let us briefly review previous analyses proposed in the literature. Abney (1987) proposed that the postnominal position of the adjective in English indefinite-pronoun structures is derived from N raising to D: According to him, indefinite pronouns have some special property that obligatorily drives this head raising of an indefinite pronoun. On the other hand, ordinary nouns cannot raise to D. On the surface, the approach seems attractive, given that most (if not all) indefinite pronouns in English morphologically consist of a determiner every/some/any and a general noun one/body/thing/place. However, Kishimoto (2000) argues against the raising-to-D analysis, because English does not have N-to-D raising, unlike Romance languages (see Longobardi 1994;Cinque 1995). Instead, Kishimoto (2000) proposes that what he calls a light noun raises only up to the Num head, passing an adjective:  Leu (2005), however, who argues that they do not explain (i) the morphology -s exhibited by modifiers of (neuter) bare nouns in Swiss German and (ii) the appearance of de in French. The postnominal adjective in indefinitepronoun structures in Swiss German in (14a)-(14c), just like a standard prenominal adjective, carries the inflection -s, but it is missing on the predicative adjective in (14d). This indicates that the apparent postnominal adjective in in (14a)-(14c) is indeed a prenominal attributive adjective rather than a predicative adjective. Given the data above, Leu (2005) argues that there must be something invisible that the adjective modifies prenominally.
this is funny-S 'This is funny.' Similarly, in French, whereas adjectives can be postnominal or prenominal for ordinary nouns, as shown in (15a) and (15b), in indefinite-pronoun structures they must follow the indefinite pronoun, as shown in (15a). Leu (2005) claims that the appearance of de in the indefinite-pronoun structure in (15c) constitutes evidence for an invisible nominal, because de introduces nominal projections, as shown in (15d).  Given that Kishimoto's analysis does posit an underlying prenominal structure, Leu's objection to it is not completely convincing. Furthermore, he leaves unanswered (see Leu 2005: 154) why the empty category ec cannot be lexically realized. Larson & Marušič (2004) criticize the raising analyses of English indefinite-pronoun structures, arguing that the adjectives that appear with indefinite pronouns in English show all the hallmarks of postnominal adjectives. Their counterarguments to a raising analysis include: the impossibility of stacking of adjectives (17), differences in measure-adjective inflections (18), the ungrammaticality of exclusively prenominal adjectives (19), and semantic differences. (20) Their arguments against the prenominal analysis of indefinite-pronoun structures, however, seem to apply only to English, because Leu (2005) shows that there is much evidence for the prenominal status of adjectives in indefinite-pronoun structures in other languages. Furthermore, Roehrs (2008) presents evidence that adjectives in German indefinite-pronoun structures are prenominal. Take adjective inflections. Notice that the adjective in the indefinite-pronoun structure in example (21d) behaves like the prenominal adjective in example (21a). Both of the adjectives in (21a)   the exhibit is important for the sentencing 'The exhibit is important for the sentencing.'

d. etwas
Wichtig*(es) something important STRONG 'something important' Thus, Roehrs (2008) proposes a binominal analysis of indefinite-pronoun structures, as illustrated in (22) with the French example from (15c). All the prenominal properties discussed above reduce to the posited NP structure, independent of the indefinite pronoun. Roehrs assumes that the light noun chose is located at the IPR (="Indefinite Pronoun Restrictor") head position and de at the Mod (="Modifier") head position. Agr NP e N Furthermore, he argues that the peculiarities of English indefinite-pronoun structures seen in (17)-(19) are nicely captured by allowing AP to replace the AgrP layer in (22), which explains the postnominal behaviors of adjectives. 4 The success of his proposal depends on the extent to which the posited various functional heads are justified and on the extent to which the structure in (22) does not overgenerate ill-formed noun phrases generally.
In the sections that follow, I argue that Japanese indefinite-pronoun structures have a light-noun structure (nP) and a modifier is not a postnominal modifier of an indeterminate phrase (QP) but rather a prenominal modifier of a light noun. Importantly, the rightmost nominal category is silent in English/French, but it is obligatorily overt in Japanese (and optionally so in Danish and German). Thus, Japanese provides visible evidence for the invisible lower nominal element that has been proposed by Leu (2005) and Roehrs (2008). As I will demonstrate, the apparent differences between Japanese and other languages stem from distinct mechanisms that compose so-called indefinite pronouns. The present detailed study contributes to understanding the principles and parameters involved in indefinite-pronoun structures, because there has not been extensive work on these phenomena in head-final languages.

Proposal: Something (in)visible
In this section, I will argue that indefinite-pronoun structures in Japanese, like those in English/French, have a light-noun structure, but while the light noun appears together with a determiner-like element some/quelque in the French/English construction, the light noun and the indefinite pronoun are separated by a modifier in Japanese.
This necessitates articulation of the internal structure of "indefinite pronouns" in Japanese. Elaborating on the idea that so-called indefinite pronouns in Japanese are decomposed into indeterminates and particles (see Kuroda 1965), it will be shown that they are distinct from light nouns and that both are present in the syntax of indefinitepronoun structures.

QP and light nouns
I propose that there is a functional projection n-light noun-within a noun-phrase structure (see Marantz 1997;Borer 2005a;b;Hiraiwa 2012;2016;Belder & van Craenenbroeck 2015). Assuming that a root category is not a noun before it is merged with n, it is reasonable to think that restrictive modifiers are located in the nP domain. The tree in (24) gives the underlying structure of the indefinite-pronoun structures in (3a)-(7a). I propose that an indeterminate phrase QP is merged in the specifier of some functional phrase XP. 5 (24) 'something spicy' in Japanese Due to the lack of a root category in the complement of n, n is stranded. Assuming that the category-setting functional categories n and v are affixal and hence cannot be stranded (cf. Lasnik 1981;Abels 2001, among others), Japanese saves the structure by inserting a semantically appropriate light noun-no, hito, etc.-much like do support in English saves a stranded affix on T. The n head in Japanese indefinite-pronoun structures is always occupied by an overt element, as shown in (25a)-(25b), because the language does not have a null light noun or pro-form. This is evidenced by example (26b) (see Hiraiwa 2016). If there were a null light noun or pro-form, (26b) should be as grammatical as (26c If a root category occupies the complement position of n (e.g. √DOG in (23)), then, nP becomes "NP" (e.g. dog) in a traditional sense. In this case, n is not stranded, and it is morphologically realized as ∅.
As the examples in (27) show, not only light nouns but also ordinary nouns can appear after the adjectives in indefinite-pronoun structures.
something spicy ln/dish-nom eat-want 'I want to eat whichever spicy thing/dish.' d. Dokoka tooi tokoro/mati-e iki-tai. someplace far.away ln/town-to go-want 'I want to go to a place/town far away.' e. Ituka himana toki/hi-ni aimasyoo. sometime free ln/day-on meet 'Let's meet in your free time/on a day when you are free.' Example (27b) with the ordinary noun is structurally represented as in (28).

(28)
The structure of (27b (24) and (28) that so-called existential indefinite pronouns in Japanese are not built on light nouns. In fact, (33) shows how existential quantifiers in English and Japanese are composed: in English they contain the determiner some and what Kishimoto (2000) calls a light noun (or what Leu 2005 calls a restrictor), one/body/thing/place/time, while those in Japanese have nothing to do with light nouns but rather are built with the quantificational particle ka and an indeterminate (or informally, a wh-word), dare/nani/dore/doko/itu, as shown in (34). This is known as an indeterminate system (Kuroda 1965;Haspelmath 1997). 8 The difference between English and Japanese is stated in (35).

(33) Existential indefinite pronouns in English and Japanese
English Japanese  (27), light nouns are exactly what we find at the right periphery of indefinite-pronoun structures in Japanese, and they have nothing morphologically in common with the Japanese indefinite pronouns in (33), but they do resemble the light nouns found in the English indefinite pronouns: (36) Light nouns in English and Japanese (Hiraiwa 2012;2016) English Japanese one/body hito thing mono/koto/no place tokoro time toki Thus, Japanese indefinite-pronoun structures provide visible evidence for the existence of a lower nominal element that an apparently postnominal adjective modifies prenominally, as Leu (2005) and Roehrs (2008) proposed.

Evidence for prenominal adjectives
If the proposed analysis is correct, Japanese provides a direct argument for the presence of a lower nominal element in indefinite-pronoun structures. It follows that a modifier that appears to the right of an indeterminate phrase is in fact a prenominal modifier of this overt light-noun element. A couple of pieces of evidence provide support for this conclusion. First of all, as I have just shown, the very existence of the overt light noun provides the strongest evidence that the modifier in question is syntactically prenominal to it.
Second, there is a class of adjectives/predicates that can only be used with first-person subjects in the present tense (see Kuroda 1973;Nitta 1991;Tenny 2006;Nishigauchi 2014). The adjectival predicate -tai in example (37a) shows this person restriction, and hence the sentence is ungrammatical with second-and third-person subjects. This restriction is lifted, however, when -tai is used attributively, as shown in the relative-clause example in (37b). Finally, in indefinite-pronoun structures in Japanese, adjectives always appear in an attributive/adnominal form. As example (39a) shows, the conclusive form da is required when the adjective is used predicatively (see Yamakido 2005). This contrasts with the attributive form na when the adjective is used attributively, as shown in (39b). Given this contrast, example (39c) shows that the adjective in indefinite-pronoun structures is an attributive prenominal modifier of the following light noun.

Constituency
I have argued that an indeterminate phrase in a Japanese indefinite-pronoun structure is merged in the specifier of XP (see (24)).
There are at least three constituency test frames in Japanese: a pseudocleft frame, a genitive-case-marking frame, and a postposition frame. Let us examine the results of each test below.
Cleft constructions in Japanese allow for multiple foci, but crucially, pseudocleft constructions do not (see (40)). Thus, any phrase that appears in the focus position in pseudocleft constructions must form a syntactic constituent (see Hiraiwa & Ishihara 2012).
(40) a. Boku-ga e i e j tabeta no-wa sushi-o i Ginza-de j datta. 1sg-nom ate ln-top sushi-acc Ginza-in cop '(Lit.) It was sushi, in Ginza, that I ate.' b. *Boku-ga e i e j tabeta mono-wa sushi i Ginza j datta.
1sg-nom ate ln-top sushi Ginza cop '(Lit.) What I ate was sushi, at Ginza.' Now, the grammaticality of (41) shows that the QP and the rest of the structure form a single constituent. To conclude, our evidence indicates that indefinite-pronoun structures in Japanese have a light-noun structure and the light noun n and the indefinite pronoun (QP) are separated by a modifier, which directly modifies the former.

Pronouns and linkers
Now important questions to ask are (i) why the postadjectival light noun n is not realized in French and Italian, while it is obligatorily overt in Japanese, and (ii) what the nature of the element de in French (and similarly di in Italian) is, while a corresponding element is absent in Japanese. These are the differences that I will address in the final section of this article. They reduce to a difference in how indefinite pronouns are composed.

n movement in languages without an indeterminate system
Crucially, indefinite pronouns in French and Italian are not based on an indeterminate system (cf. (33) and (36)). That is, in languages without an indeterminate system like Japanese's, indefinite pronouns are built from a determiner and a light noun (Kishimoto 2000;Leu 2005). For example, the light nouns thing in English and chose in French appear as part of the indefinite pronouns something and quelque chose. Assuming that the basic structure proposed in the preceding section for Japanese indefinite-pronoun structures is universal, it follows that a light noun itself undergoes movement leftward in indefinite-pronoun structures in English and French, leaving the original copy unpronounced for standard reasons. The derivation of the French example (2b) is represented in (44) (I will come back to the nature of X shortly, in the next subsection). The different types of indefinite-pronoun structures reduce to what comes in the specifier of XP. As I have proposed, in Japanese-type indefinite-pronoun structures, an indeterminate phrase QP appears there. Recall that so-called indefinite pronouns in Japanese are based on an indeterminate system. Thus, an indeterminate phrase consists of an indeterminate and a quantificational particle, independent of a light noun. In contrast, in French-type indefinite-pronoun structures, an indefinite pronoun consists of a light noun and a determiner D. Assuming a theory of head movement like that of Matushansky (2006), I propose that in languages without an indeterminate system, the light noun n moves to the specifier of XP and forms an indefinite pronoun with D. 9,10 To summarize, indefinite-pronoun structures come at least in two kinds.
(45) Two types of indefinite-pronoun structures a. An indefinite pronoun is externally merged in the specifier of XP as QP in languages whose indefinite pronouns are based on indeterminate pronouns (Japanese). b. A light noun n is internally merged in the specifier of XP in languages whose indefinite pronouns are based on light nouns (English/French).
(46) Japanese type English/French type a. Leu (2005) characterizes de in French as a partitive, but Roehrs (2008) observes that this is not quite right: in English and other languages, partitives are overt but never appear in indefinite-pronoun structures.

X as linker
(47) a. three kilos of wood b. something (*of) new In light of the analysis I have proposed in this article, the most natural position for de (and di in Italian) is X. I suggest that this head X is a linker and is required by the light noun that is moved to its specifier (Baker & Collins 2006;Larson & Yamakido 2008;Larson 2009;Hiraiwa 2012). This immediately raises a question: why doesn't a linker overtly appear in Japanese indefinite-pronoun structures? The most important idea shared by the literature on linkers is that a linker is Case-related (see Baker & Collins 2006;Larson & Yamakido 2008;Larson 2009). From this viewpoint, it follows that a linker is missing in (48) because the indeterminate phrase in Japanese in indefinite-pronoun structures is not a "noun" but a QP, hence does not require case marking (see Hirose &Suzuki 2009 andHiraiwa 2015;2017). This is indeed what we are assuming in (34). Importantly, the absence of case marking on indeterminate phrases is never due to case drop. Case drop in Japanese is colloquial and is a main-clause phenomenon, hence it is excluded in embedded clauses like the one in (51a). By contrast, bare indeterminate phrases are perfectly fine in embedded clauses, as shown in (51b) In fact, these indeterminate phrases behave similarly to caseless numerals and quantifiers.
Ken-top fruit-acc two-cl/all ate 'Ken ate two fruits/all the fruits.' Given this, the absence of case marking on indeterminate phrases in indefinite-pronoun structures supports the conclusion. The ungrammaticality of (54c) and (54d)  d. *Boku-wa karai no/mono nanika-o tabe-tai. 12 An anonymous reviewer asks an interesting question here. If a numeral-classifier combination such as huta-tu 'two-cl' is a QP, one may expect it to appear in the same position as an indeterminate phrase. But first of all, the numeral-classifier combination is not a QP. Our definition of QP is a phrase that is headed by a quantificational particle. Numeral-classifier constituents lack a quantificational particle. Rather, they consist of a numeral and a classifier, forming possibly a #P (see Watanabe 2006). Thus, they are not predicted to pattern with QPs. Still, a sentence like the following seems fine at first sight.
(i) [ #P San-nin] himasoona gakusee-o sagasiteimasu. three-cl free.looking person-acc looking.for 'I am looking for three free-looking students.' But a pseudocleft test shows that the numeral-classifier combination san-nin is actually floated out of the entire DP.
(ii) ?*Sagasiteiru hito-wa [ #P san-nin] himasoona gakusee desu. looking.for person-top three-cl free.looking student cop 'What I am looking for are three free-looking students.' Interestingly, it is possible to turn the numeral-classifier combination into a QP, by adding nan 'what' and a quantificational particle ka (see Hiraiwa 2018). And in that case, the sentence passes the pseudocleft test.
(iii) Sagasiteiru hito-wa [ QP nan-nin-ka] himasoona gakusee desu. looking.for person-top what-cl-KA free.looking student cop 'What I am looking for are some free-looking students.' Thus, it seems that a QP is syntactically required in indefinite-pronoun structures, but I do not have any deeper explanation for this constraint.
In conclusion, we have firm evidence that QP is not a nominal category, and this explains why a linker does not appear in indefinite-pronoun structures in Japanese. As we have seen in (50), however, existential indeterminate phrases such as dareka 'someone' and nanika 'something' are structurally ambiguous between DP and QP. Thus, the remaining question to answer is why a full-fledged DP cannot be merged in the specifier of LkP. The answer is straightforward. If it were a DP, its case would be left unvalued because the entire indefinite-pronoun structure (=DP) would be closer to a higher probe and would receive case valuation instead. In the next, final subsection, I will show that this line of reasoning is indeed correct.

Size matters: Varieties of indeterminate phrases
So far, we have only looked at existential indeterminate phrases (e.g. dare-ka 'someone', nani-ka 'something', etc.). But they are not the only kind of indeterminate phrases that can occur in indefinite-pronoun structures. In Japanese, there are two other kinds, NPIs and free-choice items, illustrated in (55) and (56). NPIs in Japanese are formed by combining an indeterminate with the particle -mo, instead of -ka, while free-choice items take the particle -demo. 13 (55) a. Recall that the proposed analysis of indefinite-pronoun structures has a visible n, which is obligatorily case marked. Thus, it is predicted that an indeterminate phrase in the specifier of LkP cannot be a full-fledged DP. The above data confirm this. QPs, unlike DPs, are nonnominal and hence resist prenominal modification and case marking. Thus, only QPs (caseless existential quantifiers, NPIs, and free-choice items) are grammatical in indefinite-pronoun structures, and full-fledged DPs (universal quantifiers and wh-words) are excluded. This is summarized in (65).
(65) QP and DP structure modification case marking indefinite-pronoun structure DP full-sized ok ok * QP downsized * * ok To summarize the discussion in Section 4, QPs in indefinite-pronoun structures in Japanese are not nominals, and this explains why no linkers appear on QPs. On the other hand, the appearance of linkers in French and Italian (de and di) shows that the indefinite pronouns in these languages are nominals that incorporate light nouns.

Conclusion
In this article, I have shown that Japanese indefinite-pronoun structures have light-noun syntax-a QP and a light noun. Thus, the present study supports Leu's (2005) analysis of indefinite-pronoun structures, providing visible evidence for a syntactic head that is invisible in some other languages. To the extent that the proposed analysis is right, what are called "indefinite pronouns" are not syntactically uniform but come in (at least) two varieties, QP and DP; only the former is allowed in indefinite-pronoun structures in Japanese. I have mostly focused on the syntax of indefinite-pronoun structures in Japanese. Obviously, it still remains to be seen whether the syntax of indefinite-pronoun structures in all the other languages can be unified. I have suggested a formal way to understand parametric differences between Japanese and English/French indefinite-pronoun structures. Unification, ultimately, will require collaborative research by linguists working on various languages. I leave that for the future.