1 Introduction

This paper discusses the structural properties and presumptive origin of focus particle constructions with clause-final nominalized predicates, henceforth “focus concord constructions”, which are a sub-group of the so-called dependent-concluding (kakari-musubi) constructions in older Japanese, from the viewpoint of grammaticalization. The term “dependent-concluding”1 is based on the notion that in strictly head-final Japanese, a non-clause-final constituent always relates to, or “depends on” a following constituent, such as an argument of a predicate “depending on” a predicate. The special thing in the case of some of these constructions is that not only the “dependent” constituent is marked but also the “concluding” predicate, which is otherwise not the case in Japanese. As in ex. (1), the depending-concluding constructions involve a focused constituent with a specific particle, in this case the interrogative particle ka, that corresponds to a specific nominalized predicate.

    1. (1)
    1. Man’yōshū 420
    1. oyodure=ka
    2.     lie=QUE
    1. wa=ga
    2. I=GEN
    1. kiki-t.uru,
    2. hear-PFV.ANP
    1. tapakoto=ka
    2. nonsense=QUE
    1. wa=ga
    2. I=GEN
    1. kiki-t.uru=mo.
    2. hear-PFV.ANP=EXC
    1. ‘Is it lies that I’ve heard, is it nonsense that I’ve heard?’

In (1), oyodure ‘lie’ and tapakoto ‘nonsense’ are objects marked with the question particle ka. They correspond to the clause-final predicates (both kiki-t- ‘has heard’) in adnominal form, that is, with the ending –uru.

Frequently used in older Japanese, these constructions rapidly declined from the 12th century on and are not preserved in Modern Standard Japanese. The focus of this paper within the historically documented language periods is on Old Japanese. Some of the problems involved in exploring this topic are rather complex, unresolved, and even controversial. This paper aims to offer a broad and balanced view on the topic, especially by reflecting the Japanese research literature, which tends to get a short treatment in English publications. Furthermore, I try to give at least some room to all major extant hypotheses irrespective of their theoretical background, which is usually not the case. The common tendency is for descriptive analyses to ignore analyses motivated by specific syntactic theories, and vice versa.

Japanese is historically documented relatively well from the 8th century on, with two exceptions, namely the 9th century, which is generally not well documented, and a period from from the 12th to the 16th century, where most written language ceased to reflect contemporary spoken language reliably. There are two distinct core areas in the research literature on focus concord constructions, namely explanation of the origin and the structure of the focus concord constructions in the earliest period, Old Japanese (CE 6th~8th; henceforth, “OJ”) on the one hand, and their use and meaning in the classical period (CE 10th~11th) on the other hand. There is also a limited amount of literature describing the decline and demise of the focus concord constructions, which is not discussed in this paper.

In section 2, I will present the elements contributing to, or historically related to, the construction, namely the nominalized predicate forms that crucially characterize the focus concord constructions, the focus particles, and thirdly, case marking. I will also discuss the notions of focus and topic. Section 3 will give a brief overview of the constructions as they present themselves at the oldest stage, namely Old Japanese; section 4 presents the main competing analyses of these constructions that have been proposed, including accounts of how they may have emerged, before section 5 concludes the paper with a proposal on the presumably most likely scenario of their evolution.

2 Elements contributing to the interpretation of focus concord constructions

Focus concord constructions contributed to information structuring at historical stages of Japanese starting from the earliest documented times to around the 14th to 15th century with the last remnants of the constructions still in use up to around the 18th century. Some of the same constructions are also documented in Ryukyuan, a language family closely related to Japanese, leading to the common assumption that their genesis predates the split between mainland Japanese and Ryukyuan. This paper will focus on Japanese except in a few places where we refer to comparative evidence between these languages.

The major means for information structuring in Japanese are particles, word order, and prosody, whereby we do not know much about the latter historically. Going back to the 18th century scholar Motoori Norinaga2, traditional Japanese linguists (e.g. Morishige 1959; 1971; Funagi 1987; Ōno 1993; Sakakura 1993; Kawabata 1994; Fiala 2000) have perceived all information structuring particles, sometimes including even zero marking on topics, and the predicate forms to which they correspond as a system of kakari-musubi (lit. ‘dependent-concluding’). Our interest here is only a sub-part of this system, namely those constructions with an overt particle and a nominalized predicate form. Therefore we find the use of the Japanese term, which covers a much wider range of constructions than those discussed in this paper, somewhat misleading, and instead use the English term focus concord construction that was chosen for this special issue. Indeed, among the dependent-concluding constructions, it is only the focus concord constructions that raise questions of broader theoretical and descriptive interest, and have been the subject of research in general linguistics.

In this section, I will proceed as follows. Section 2.1 briefly discusses the notions of topic and focus adopted in this paper. Section 2.2 introduces case marking in historical Japanese. The particles and the nominalizing verb forms will be briefly introduced in sections 2.4 and 2.3, respectively, before we will deal with the whole constructions in section 3.

2.1 Topic and focus

Before entering the description, a clarification of the concepts of topic and focus is needed. Topic has been traditionally associated with the notions “givenness” or “old information”. However, currently, going back to Reinhart (1982) and Lambrecht (1994), topic is almost universally defined in terms of “aboutness”: “The topic of a sentence is the thing which the proposition expressed by the sentence IS ABOUT” (Lambrecht 1994: 117).3 The rest of the proposition, then, is relevant to and increases the knowledge about the sentence topic. As an example, in a sentence like (2), without further context or special intonation, we might assume that ‘the children’ is the topic of the sentence.

(2) Lambrecht (1994: 120)
  The children went to school.

In contrast, there is not a single widely accepted concept of focus. Furthermore, most of the literature agrees that focus has different sub-types. Generally, three major avenues of thinking can be distinguished. First, most broadly, focus can be conceptualized as the part of a proposition that complements the topic, that is, “what is predicated about the topic” (Gundel & Fretheim 2004: 176; cf. also van Valin 2001: 184). Secondly, focus can be defined as the part of a sentence where the assertion differs from the presupposition (Lambrecht 1994: 213). This is not necessarily everything outside the topic, but that part of the proposition that is “unpredictable or non-recoverable for the addressee at the time of the utterance” (Lambrecht 1994: 218; cf. also Luraghi & Parodi 2008: 112–113). When the sentence contains no topic, it can be the whole sentence. In a third line of thinking based on the work of Rooth (e.g. Rooth 1996), focus “indicates the presence of alternatives that are relevant for the interpretation of linguistic expressions” (Krifka & Musan 2012: 7). While such alternatives are always available, “focus especially stresses and points out the existence of particular alternatives” (Krifka & Musan 2012: 7). The second concept of focus can also be termed as “informational” focus and the third as “identificational” or “contrastive” focus (cf. Gundel & Fretheim 2004: 181; van der Wal 2016: 262). The notions of informational focus and alternative/contrastive focus can be illustrated with the following example:

(3) A: Who made all this great food?
  B: BILL made the CURRY.

Bill is an informational focus in direct response to the question word who and a contrastive focus, indicating a choice among alternatives, while curry has only a contrastive focus (Gundlach & Fretheim 2004: 182). Note that in general a sentence can have multiple foci (Krifka & Musan 2012: 20).

Even within the concept of focus in terms of alternatives, one can make further distinctions between sub-types, such as exhaustive focus vs. non-exhaustive focus. Exhaustive focus “indicates that the focus denotation is the only one that leads to a true proposition, or rather more generally: that the focus denotation is the logically strongest denotation that does so.” (Krifka & Musan 2012: 21). Cleft constructions in English are typically associated with this type of focus, while in-situ focus is not (cf. Krifka & Musan 2012: 7).

For the purposes of this paper, I adopt a broad notion of focus: A sentence constituent qualifies for the status of focus both if it is informational focus in the sense of Lambert, that is, contains information that is unpredictable or non-recoverable for the addressee, and if it indicates alternatives in the sense of Rooth. The reason is that in this paper we primarily deal with formal expressions of information status. These expressions often have more than one denotation in terms of information structure; e.g. English accent can both express informational and contrastive focus. We want to cast our net wide here and allow for ambiguity in the expression of different types of focus. In contrast, I will not assume here that all elements of a sentence have to be either topical or focal, that is, everything outside of topic is focus, as many authors do.

Lastly, we need to point out the special status of “contrastive topic”. The highlighted constituent in (4) would be a contrastive topic.

(4) Gundlach & Fretheim (2004: 300)
  A: What do you think I should take to the camping trip?
  B: That COAT I bought you, I think you should take.

Coat invokes a contrast to other items that are commonly taken to camping trips. At the same time, it is a topic for the sentence. In current linguistics, contrastive topics are often not considered as a sub-type of topic but either as a special case of contrastive focus (cf. Gundel 1998: 300; Umbach 2001; Gundel & Fretheim 2004: 183; Titov 2013: 452),4 or as combining topic and focus properties (Krifka & Musan 2012). This is also due to the increasing prevalence of Rooth’s concept of focus as evoking alternatives. Evoking a contrast falls squarely within evoking alternatives. Therefore, even authors who do not classify contrastive topics outright as foci admit their focal nature. Thus, Krifka & Musan (2012: 30) suggest that, “[contrastive topics] consist of an aboutness topic that contains a focus, which is doing what focus always does, namely indicating alternatives.”5 Despite the difference in classification, this stance is actually not very different from the stance of authors who regard contrastive topic as a type of contrastive focus. Thus, Umbach (2001: 4) suggests that “A contrastive focus may occur in any position in a sentence. If, however, it occurs in the topic part, it represents a contrastive topic…”6 Figure 1 may serve to represent the understanding of topic and focus that we have come to espouse here.7

Figure 1
Figure 1

The notions of topic and focus.

Figure 1 is meant to express that contrastive foci are a subset of foci. Contrastive topics in turn are a subset of contrastive foci, but have properties of topics (specifically, they are found in topic position, which is at the start of the clause). Note that Figure 1 should not be misunderstood as representing the information structure of a clause. I do not commit to the view here that all elements of a sentence have to belong to either the topic or the focus part of the sentence, in a topic-comment type of structure (e.g. the Romanist tradition as described in van Valin 2001: 184; or Dufter & Jacob 2009).

2.2 Core case marking in Old Japanese

Case marking is relevant to the focus concord constructions because (1) information-structuring particles were partially in complementary distribution with case marking particles and (2) because the ordering between constituents marked by case vis-à-vis constituents marked with focus-structuring particles, as will be seen in section 4, is generally believed to provide hints about the syntactic structure of the focus concord constructions.

Old Japanese had unmarked nominatives. Accusative was optionally marked by the particle wo, to which traditionally a function of ‘appeal’ (yobikake) (Kamada 1979: 179) and more recently (non-obligatory) marking of specific objects (Frellesvig et al. 2015) have been attributed. It is clear that OJ wo was multi-functional, also marking perlative (a spatial case for the notion “through”), for example. But opinions are split as to whether one should acknowledge “accusative case” as one of these functions (e.g. Suda 2010 against; and Kinsui 2011: 99 in favor of an accusative function). The particle ni marked datives and a wide range of oblique arguments and adverbials.

In addition to the core case marking, OJ had two genitive (adnominal) particles, ga and no, which are particularly relevant for the study of focus concord constructions. Their distributions in OJ have been described in detail by Nomura (1993a; b). While their basic function was marking adnominal relations between nouns, they could also mark the adnominal relation between a nominalized predicate and its first argument if it was an external argument. This could be any predicate marked by one of the nominalizing inflectional forms listed in 2.3, including those in focus concord constructions. In (5), the particle ga appears twice, marking the relationship between the predicate k- ‘come’ and the first argument a ‘me’ in the first clause, and a possessive relationship between two nouns in the second clause.

    1. (5)
    1. Man’yōshū 3589
    1. […]
    1. Ikoma~yama
    2. PN-mountain
    1. koye-te=so
    2. cross-GER.FOC
    1. a=ga
    2. I.GEN
    1. k.uru.
    2. come.ANP
    1. Imwo=ga
    2. girl=GEN
    1. me=wo
    2. eye=ACC
    1. pori.
    2. desire
    1. ‘It is across Mt. Ikoma that I have come, wanting to see my wife.’

The first clause in (5) is a focus concord construction. The author marks the adverbial phrase, ‘crossing Mt. Ikoma’ as focus, emphasizing that he took the more direct but more arduous route to Nara out of desire to see his wife as quickly as possible. The question with examples like this is of course, to which extent they do not actually mark an adverbal (nominative – verb) rather than an adnominal relationship. Based on the verb morphology, one can still classify them as adnominal. According to the traditional descriptions, clear examples of main clause nominative uses of the genitive particles only develop a few centuries later in Early Middle Japanese (cf. e.g. Frellesvig 2010; Yamada 2010).

For the purpose of the analysis of focus concord constructions, it is the accusative wo and the genitives ga and no, and the constituents marked by them, which are of particular relevance.

2.3 The nominalizing predicate forms

Focus concord constructions are characterized by the nominalizing form of the predicate. OJ has two predicate forms that clearly nominalize clauses, only one of which is involved in focus concord constructions.8 For a third predicate form which is involved in the focus concord constructions, it is contentious whether it is also nominalizing. The clearly nominalizing forms are (i) the nominal form -(ura)ku for cleft nominalizations, and (ii) the adnominal forms –(ur)u “non-past adnominal”, –si “past adnominal” for verbs, and –ki “non-past adnominal” for adjectives there. In contrast, the clause-nominalizing function of the presuppositional-conditional forms, namely non-past presuppositional –(ur)e and its past presuppositional counterpart –sika is contentious. Some discussion will follow below. First, ip-an.aku and kwop.uraku in (6) are examples for (i) the nominal form -(ura)ku.

    1. (6)
    1. Man’yōshū 2725
    1. Mitu=no
    2. PN=GEN
    1. panipu=no
    2. clay=GEN
    1. iro=ni
    2. color=DAT
    1. ide-te
    2. appear-GER
    1. ip-an.aku=nomwi=so
    2. say-NEG.NMZ=LIM=FOC
    1. wa=ga
    2. I=GEN
    1. kwop.uraku=pa.
    2. love.NMZ=TOP
    1. ‘I only keep it to myself till it appears in my face like the red clay of Mitsu, the fact that I love [you].’

Second, the adnominal inflection (ii) had three basic functions in OJ, namely (a) forming relative (adnominal) clauses, (b) forming complement clauses, and (c) forming standalone clauses. The most interesting among these for the history of the focus concord constructions is the standalone one, as will become clear from the following sections.9 From what we know, this was not yet a common use in OJ. Yamauchi (2003: 122–127) presents about fifteen unambiguous occurrences in the Man’yōshū that do clearly not depend on the preceding or following clause for their nominalization. (7) is an illustration.

    1. (7)
    1. Man’yōshū 4273
    1. Ametuti=to
    2. world=COM
    1. api.sakaye-m.u=to
    2. RCP.prosper-FUT.FNP=CPL
    1. opomiya=wo
    2. imperial.palace=ACC
    1. tukapye~matur.e=ba
    2. serve~HML.CON=TOP
    1. taputo.ku
    2. precious.ADV
    1. uresi.ki.
    2. joyful.ANP
    1. ‘Since I serve the imperial court in order to prosper together with the whole world, this is such a precious and joyful thing.’

As I tried to indicate in the translation, in those OJ examples that end on an adnominal form, it is always possible to also construe a nominalizing function. Japanese philology interprets the standalone use as an emotive expression (cf. Kojima et al. 1975: 355). It is therefore sometimes labeled as kantaiku, ‘emotive nominalized clause’. For example, the utterance in (7) can be understood as a direct expression of the author’s joy and gratitude.

The allomorph –u of the non-past adnominal form of verbs suffixed to consonant stem verbs, the class of verbs with the highest type frequency, is in writing segmentally identical to the finite non-past (also –u), e.g. ip.u ‘say’ or kazas.u ‘decorate’ are ambiguous between being finite or adnominal without context. Note though, that they may have developed from different morphological segments (cf. e.g. Unger 2000: 672, 675), and had different accents (cf. e.g. Yamaguchi 1986: 333; Uwano 2006: 8, 17).

Third, there are clauses with predicates ending on non-past and past presuppositional –(ur)e and –sika; e.g. (8) and (9), both for –(ur)e.

    1. (8)
    1. Man’yōshū 471
    1. Ipye
    2. house
    1. sakari
    2. leave
    1. imas.u
    2. go(HON).ANP
    1. wagimwo=wo
    2. wife=ACC
    1. todome-kane
    2. hold.back-NPO
    1. yama~gakusi-t.ure
    2. mountain~hide-PFV.PRE
    1. kokorodo=mo
    2. heart=FOC
    1. na.si.
    2. not.be.FNP
    1. ‘Since I was unable to hold back my wife who left home [= died] and hid her in the mountain [= buried her], I have lost my spirits.’
    1. (9)
    1. Man’yōshū 1266
    1. Opo~pune=wo
    2. big~boat=ACC
    1. arumi=ni
    2. rough.sea=LOC
    1. kogi~’de
    2. row~leave
    1. ya~pune
    2. many~ships
    1. tak.e
    2. steer.PRE
    1. wa=ga
    2. I=GEN
    1. mi.si
    2. see.ANP
    1. kwora=ga
    2. girl=GEN
    1. mami=pa
    2. eye=TOP
    1. siru.si=mo.
    2. clearly.visible.FNP-EMP
    1. ‘Although I have steered many ships, rowing big boats out to the rough sea, I still see the eyes of the girl I met clearly in front of me [i.e., I cannot forget her].’

The presuppositional endings are only considered nominalized by some scholars (e.g. Schaffar 2003; Yanagida & Whitman 2009).10 One can make a good case in favor of this hypothesis from the fact that they share the case distributions of the nominalized clauses on –(u)ru and –(ur)aku. Especially, the first argument of the verb is marked as a genitive (no or ga) if it is an external argument, and has zero marking if it is an internal argument (cf. Yanagida & Whitman 2009: 103).

Furthermore, they are the only other verb forms used in focus concord constructions besides the clearly nominalized ones, and like the nominalized verb forms in the other focus concord constructions mark the backgrounded part of the sentence (the predicate) in contrast to the highlighted part. Moreover, it seems quite likely that –(ur)e etymologically contains the same nominal suffix as the adnominal form –(ur)u, which is variously reconstructed as *-ra, *-l, *-rua, *-or etc. (cf. Robbeets 2015: 339).11

(10) cf. Unger (2000: 664)  
  A.    attributive [-(ur)u] predicative12 + *ra + *u
  B.    subjunctive [presuppositional; -(ur)e]: predicative + *ra + *Ci

Ishida (1939a; b) presented a detailed analysis of the presuppositional form and its relationship with the focus particle koso in OJ, including diachrony within OJ, and came to the following conclusions that have been supported in Japanese linguistics to the present (cf. e.g. Ōno 1956b; Iwai 1970: 511; Ōno 1993: 101–102, 115; Katsumata 2007: 1; Kinsui 2011: 158): 1) this form on its own had no finite function; 2) it served for the expression of a presupposition13; 3) it marked established facts (note that presuppositions are not necessarily established facts; so there is only partial overlap); 4) the clause marked by this form was open to both concessive and causal interpretation. Thus, in short, there has been general agreement in Japanese linguistics that the basic function of these forms was to mark some state of affairs as a given fact and form a subordinate clause in a complex clause construction, in which this state of affairs serves as the presupposition, namely either a causal clause, as in (8), or a concessive clause, as in (9). Note that in (8) the truth of ‘I was unable to hold my wife who died and hid her’ and in (9) the truth of ‘I have steered many ships, rowing big boats out to the rough sea’ is presupposed. Secondarily, as a new development, these forms also had clause-final use, as will be seen in section 3.3.14

Finally, there is the negated future –(a)zi, which is ambiguous between being finite and nominalized, and is also occasionally found in focus concord constructions with adnominal verb forms.

Overall, then, the forms involved in focus concord constructions were the adnominal forms –(ur)u “non-past adnominal” and –si “past adnominal” and the corresponding forms for adjectives, the non-past presuppositional and past presuppositional –(ur)e and –sika, and the corresponding forms for adjectives, and the negated future –(a)zi. All of these forms can be considered as nominalizing, or the presuppositional forms may have been merely adverbial, but they have in common that they mark the predicate as non-asserted and backgrounded, thus additionally shifting the focus of the sentence from the predicate as the natural locus of focus to the constituent marked by the focus particle.

2.4 Information-structuring particles

As mentioned in section 2, the particles participating in the focus concord constructions are part of a larger group of particles that marked information structure and have been labeled as kakari (dependent) particles in traditional Japanese grammar. Table 1 shows these particles as a group and their basic features and functions,15 namely a semantic label, whether they are used in focus concord constructions, that is concord with a non-finite predicate form, are used clause-finally, in copular function, and lastly, concerning their origin, whether they are possibly of deictic origin, and if they are shared with Old Okinawan focus concord constructions (cf. Serafim & Shinzato 2011; 2013). A “no” in brackets means that we do find this distribution but the function is distinct, so that we may be dealing with homonymy rather than polysemy. The “yes” with a question mark will be discussed below.

Table 1

Information-structuring (so-called kakari-musubi ‘dependent-concluding’) particles in OJ.

Particle semantics predicate form clause-final use copular function deictic origin also in Old Okinawan
So/zo16 focus adnominal yes yes yes yes
ka doubt/question adnominal yes (yes)17 yes? yes
koso contrast presuppositional no18 no yes? yes
ya question adnominal yes no no no
namo/namu focus adnominal no no no no
mo scalar focus (finite) (no) no no no
si focus (finite) no no no no
pa topic (finite) (no) no no no

More details on those particles participating in focus concord constructions will be provided in section 3. However, I will not discuss their deictic origin and the correspondence to Okinawan there, so these issues will be briefly discussed in this paragraph. As Serafim & Shinzato (2011) and Shinzato & Serafim (2013) show in detail, cognates of so/zo “focus”, ka “question/doubt”, and koso “contrast” have also participated in focus concord constructions in Okinawan. (11) from a 16th~17th century Old Okinawan text is an example with the “doubt” particle gya (also ga, ka), which is presumably a cognate of the Japanese “question/doubt” particle ka. The final predicate ‘compare’ is in an adnominal form, like in OJ focus concord constructions.

    1. (11)
    1. Omoro Sōshi 16: 1144 (Shinzato & Serafim 2013: 50; glosses adjusted)
    1. KacɨriN=wa
    2. PN=TOP
    1. nawu=nyi=gya
    2. what=DAT=QUE
    1. tatuyir.u.
    2. compare=ANP
    1. ‘To what is it that Katsuren compares?’

OJ so/zo “focus” has a counterpart in Okinawan do “emphatic assertion” that also corresponds to an adnominal form in the predicate (cf. Shinzato & Serafim 2013: 105–106), and koso “contrast” has a counterpart in Okinawan su (labeled by Shinzato & Serafim (2013: 111) as “strong conviction”, which corresponds to a presuppositional verb form, as in (12), an example from the same 16th~17th century Old Okinawan text.

    1. (12)
    1. Omoro Sōshi 13: 899 (Shinzato & Serafim 2013: 111; glosses adjusted)
    1. Facɨ-nyisya=sɨ
    2. first-north.wind=FOC
    1. mac-y[i-y]u-tar.i.
    2. await-PRG-PST.PRE
    1. ‘It was the first North Wind in the offing itself that we awaited.’

Ya “question” has an Okinawan cognate yi that does not participate in focus concord constructions (Shinzato & Serafim 2013: 56), while namo/namu “focus” does not have a cognate at all (Shinzato & Serafim 2013: 164–165).

Deictic origins, that is, original notions such as “this” and “that”, have been claimed in traditional scholarship for all focus particles in focus concord construction that are also found in Okinawan. The same claim is put forward by Serafim and Shinzato, as well as by Quinn (1997), although the latter does not include koso “contrast” in his analysis. So ‘that’ has been a common medial deictic marker throughout Japanese language history, and its deictic origin is the most widely accepted.

The case with ka “question” and koso “contrast” is more complicated. Ka’s deictic origin has been supported in the traditional research literature (e.g. Sakakura 1993; Ōno 1993), and it is attested as a distal deictic marker in OJ, however only rarely. Thus, the question is how to reconcile its frequent occurrence as a focus particle with its rarity as a deictic marker. Quinn (1997: 65–66) tries to explain it with the nature of the OJ texts (poetry). Serafim & Shinzato (2005) and Shinzato & Serafim (2013: 156) reconstruct koso “contrast” as the proto-Japanese proximal deictic *ko (also found throughout Japanese documented language history), enhanced by the nominalizing ending *swo ‘one, thing’. Yanagida (2016: 160) suggests that *ko is the proximal deictic, but *so was the focus particle so, that is, koso emerged when so was already established as a particle (also NKD5 2001: 817). In terms of an overall system of correspondences between deictics and final verb forms, Shinzato & Serafim (2013: 279–285) point out that the proximal deictic ko in koso “contrast” iconically is in concord with the realis (presuppositional) verb form, while distal ka frequently corresponds to predicates ending on irrealis suffixes in adnominal form19. This can be illustrated by (13) and (14). (13) has a constituent with koso corresponding to a presuppositional (Shinzato & Serafim: “realis”) verb form that presents the proposition as a fact, and (14) has a constituent with ka corresponding to a predicate with future (Shinzato & Serafim: “realis”) –(a)m-.

    1. (13)
    1. Man’yōshū 145
    1. Pito=koso
    2. person=CFC
    1. sir-an.e,
    2. know-NEG.PRE
    1. matu=pa
    2. pine=TOP
    1. sir.u=ram.u.
    2. know.FNP=EPI.FNP
    1. ‘It is humans who don’t know it, but the pine will know it.’
    1. (14)
    1. Man’yōshū 3891
    1. … idure=no
    2. … when=GEN
    1. toki=ka
    2. time=QUE
    1. a=ga
    2. I=GEN
    1. kopwi-zar-am.u.
    2. love-NEG-FUT-ANP
    1. ‘…, when would I not love [you]?’

These are entirely reasonable reconstructions, especially in view of our knowledge about cross-linguistic grammaticalization paths from demonstrative to focus (e.g. Heine & Kuteva 2002: 111–112; details are found in Serafim & Shinzato 2005: 38–39). Nevertheless, their status is as reconstructions rather than historical facts.

In principle, every theory or approach that tries to describe the focus concord constructions should also be able to explain (1) why those information structure particles that participated in them did, and why those that didn’t participate in these constructions didn’t, although some of them had similar meanings; (2) why the particles correlated with nominalized predicates; and (3) why among a number of alternative nominalized verb forms it was precisely the adnominal and the presuppositional ones that participated in the focus concord constructions. However, in practice, especially the first and the third question are hardly ever asked outside the traditional Japanese research literature, and even the second is sometimes neglected. To the extent that an approach to the interpretation of focus concord constructions is explanatory, it should be able to give at least as good an account of these features as any traditional Japanese approach. As we will see in parts of section 4, some analyses of the focus concord constructions are better positioned to respond to these questions than others.

In conclusion, then, five information-structuring particles participated in the focus concord construction, three of which, so/zo, ka and koso seem to form a historical core, and two which (ya, namo/namu) may be analogical extensions. There were other information-structuring particles that did not participate in the construction although some of them had similar meanings.

3 Focus concord constructions – a family of constructions

As already seen, the Japanese focus concord constructions are characterized by a clause constituent marked with specific information-structure particles, and a specific nominalized verb form on the predicate. Whitman (1997: 162) summarized their structural properties as follows.

(15) A. The focus particle designates the scope-bearing [i.e., focussed] constituent in a scoped (focus, interrogative) construction
  B. The focus-marked element is contained in a clause whose predicate takes a nominal ending.
  C. The nominalized predicate indicates the scope of the focus-marked element.
  D. There are locality restrictions on the relationship between the focus-marked element and the nominalized predicate.

As already indicated in section 2, other than these commonalities, we find considerable differences, though. Two classifications are particularly plausible.

First, based on their functional and structural characteristics, the focus concord constructions can be divided into three groups, namely

  1. focus concord constructions with the particles ka (exx. (1),(16)) or so/zo (exx. (21),(22)), which can have copular function, and are clearly focal;

  2. focus concord constructions with ya (ex. (27)) and namu (exx. (28), (29)), which are semantically similar to ka and so/zo, respectively, but have no copula-like function, and also don’t clearly mark a constituent focus; and

  3. focus concord constructions with koso (exx. (30), (31)), which mainly participate in the formation of subordinate (adverbial) clauses and normally correspond to a different verb form.

Secondly, based on their semantics, one could classify focus concord constructions based on whether they are used for interrogatives, for assertions, or for contrastive subordinate constructions.

  1. focus concord constructions with question particles ka and ya for interrogatives,

  2. focus concord constructions with focus particles so/zo and namo for assertions, and again,

  3. focus concord constructions with the focus particle koso, typically marking a contrastive topic in a subordinate clause

In this paper, the focus is on the structures and their origins, so we will privilege the first classification, and introduce the constructions in turn in the following three sub-sections. The description is based on general descriptions of focus concord constructions in the Japanese linguistic literature, in particular Kōji (1988), Fiala (2000), Nomura (2002b), and Kinsui (2011). It goes without saying that the accuracy of the description is subject to the general limitations of semantic analysis of historical written language.

3.1 Focus concord constructions with ka and zo

Constructions with ka and with zo are the “prototypical” focus concord constructions. Many more descriptions and analyses are available about them than about the other focus concord constructions in OJ and EMJ. Ōno (1993: 19–26) put emphasis on the fact that ka and zo are the only focus concord construction particles that can both mark question words and trigger a different inflectional form on the predicate if suffixed to the predicate. Fiala (2000: 324) has labeled them accordingly as “strong” focus concord construction particles vs. the “weak” ones discussed in the following sections 3.2 and 3.3. Only these particles share the properties in OJ that, (a) the constituent marked by these particles are clearly focused, (b) the particles also have quasi copular function, and (c) if used clause-finally, ka and zo attach to adnominal, and not to finite verb forms.

The particle ka is found in OJ in two functions: to mark questions when used clause-internally and in clause-final position to mark exclamations (cf. Kondō 2014: 70–71). Nomura (2001: 115) suggests an overall underlying meaning of rendering constituents as indefinite, which in context led to an interrogative interpretation. As Nomura (1995) shows, clause-internal OJ ka marks with overwhelming frequency adverbial constituents, including adverbial clauses and other adverbial adjuncts. The biggest single type of ka-marked constituents are reason clauses, which make up about a third of all occurrences of ka in focus concord constructions. (16) is typical in showing ka marking a reason clause, which alone makes up about a third of the occurrences of ka that can be related to a specific sentence-final verb form. The sentence-final form is always adnominal if the predicate has morphologically distinguishable endings for adnominal and finite forms, as in (16).

    1. (16)
    1. Man’yōshū 276
    1. Imwo=mo
    2. lover=FOC
    1. wa~re=mo
    2. I~PRN.FOC
    1. pitotu=nar.e=ka=mo
    2. one=COP.CON=QUE=FOC
    1. mikawa=nar.u
    2. PN=COP.ANP
    1. putami=no
    2. PN=GEN
    1. miti=yu
    2. way=ABL
    1. wakare-kane-t.uru.
    2. part-NPO-PFV.ANP
    1. ‘Is it because my lover and me are one, that we cannot part from the way on Futami (‘Two-Look’) in Mikawa?’

(17) shows the capacity of ka with copular force in sentence-final position. Note that this is not a focus concord construction.

    1. (17)
    1. Man’yōshū 892
    1. Kaku=bakari
    2. like.this=LIM
    1. sube
    2. means
    1. na.ki
    2. not.be.ANP
    1. mono=ka,
    2. thing=QUE
    1. yo=no
    2. world=GEN
    1. naka=no
    2. middle=GEN
    1. miti.
    2. way
    1. ‘Is it so hopeless, the way of the world?’

(17) shows ka marking on arguments of the predicate (in this case objects), which according to Nomura (1995: 18–19), accounts for roughly a fifth of all occurrences of this particle in focus concord constructions.

    1. (18)
    1. Man’yōshū 420
    1. oyodure=ka
    2. …lie=QUE
    1. wa=ga
    2. I=GEN
    1. kiki-t.uru,
    2. hear-PFV.ANP
    1. tapakoto=ka
    2. nonsense=QUE
    1. wa=ga
    2. I=GEN
    1. kiki-t.uru=mo.
    2. hear-PFV.ANP=EXC
    1. ‘…is it lies that I’ve heard, is it nonsense that I’ve heard?’

According to Nomura (2002b: 30, 33) OJ ka basically denotes a doubt, uncertainty, or a self-directed question, while the other interrogative particle ya (see below), basically denotes an other-directed question. It is still rarely found on wh-words, mostly in rhetorical questions such as (19)(cf. Kōji 1988: 734).

    1. (19)
    1. Man’yōshū 3891
    1. … idure=no
    2. … when=GEN
    1. toki=ka
    2. time=QUE
    1. a=ga
    2. I=GEN
    1. kopwi-zar-am.u.
    2. love-NEG-FUT-ANP
    1. ‘…, when would I not love [you]?’

However, these semantic differences blurred in EMJ, as the distinction between the two markers became mainly a syntactic one: ya extended its functional domain to become the general interrogative marker while ka came to be used exclusively on wh-words (cf. Kinsui 2011: 162).

So/zo simply marks a focus and its syntactic behavior in OJ is very similar to ka, both in marking the focus of the sentence if there is one, and in having copular function, as shown in (20).

    1. (20)
    1. Man’yōshū 3254
    1. … yamato=no
    2. PN=GEN
    1. kuni=pa
    2. country=TOP
    1. koto~dama=no
    2. word~spirit=GEN
    1. tasuk.uru
    2. save.ANP
    1. kuni=zo.
    2. country=ASS
    1. ‘Yamato is a country that is helped by spells.’

On the other hand, so/zo’s distributions with different types of phrases differ from ka’s. It does not mark adverbial clauses as frequently as ka, and instead more often marks arguments, including a relatively frequent pattern in which so/zo follows the object marker wo (cf. Kōji 1988: 675–690), as in (21).

    1. (21)
    1. Man’yōshū 3286
    1. … taka~tama=wo
    2.      bamboo~ball=ACC
    1. sizini
    2. densely
    1. nuki~tare
    2. pierce~drop
    1. ame~tuti=no
    2. heaven~earth=GEN
    1. kamwi=wo=so
    2. deity=ACC=FOC
    1. wa=ga
    2. I=GEN
    1. nom.u.
    2. pray.ANP
    1. ‘I pray to the gods of heaven and earth putting as much bamboo beads [on the prayer string] as possible.’

Like ka, so/zo can also mark wh-words, although this pattern is not frequent in OJ.

One puzzling thing about so/zo and ka is that these particles can occasionally appear in a position where their status as focus is questioned (cf. Kinsui 2011: 161). In the case of so/zo, a constituent with so/zo can precede a topic (pa)-marked constituent, as in (22).

    1. (22)
    1. Man’yōshū 3698
    1. ama~zakar.u
    2. heaven~be.separate.ANP
    1. pina=ni=mo
    2. PN=LOC=FOC
    1. tuki=ga
    2. moon=GEN
    1. ter-er.e=do=mo,
    2. shine-RES.CON=AVS=FOC
    1. imo=so
    2. wife=FOC
    1. topo.ku=pa
    2. far.ADV=TOP
    1. wakare~ki-ni-ker.u.
    2. part~come.PFV.PRT.ANP
    1. ‘Even in faraway Tsushima the moons shines, but it is from my wife that I have come to be separated by great distance.’

Watanabe (2007) has argued quite convincingly that in these cases so/zo is a contrastive topic, which is a sub-type of focus, as seen above. Additionally, there are six cases in the Man’yōshū in which so is even added to a constituent already marked with pa, as in (23).

    1. (23)
    1. Man’yōshū 2933
    1. Api.omop-azu
    2. RCP.feel-NEG
    1. kimi=pa
    2. you=TOP
    1. imas.e=do
    2. be(HON)=AVS
    1. kata~kopwi=ni
    2. half~love=ADV
    1. ware=pa=so
    2. I=TOP=FOC
    1. kop.uru.
    2. love.ANP
    1. kimi=ga
    2. you=GEN
    1. sugata=ni.
    2. apparition=DAT
    1. ‘You are not reciprocating my love, while it is me who has one-sided feelings.’

As (23) shows, in these cases as well, the constituent is obviously a contrastive topic, the contrast being between ‘you’ in the preceding clause, and ‘me’ in the main clause (marked by so). Since pa alone can already mark contrastive topics, so apparently adds to the meaning of the sentence by putting a focus on one of the two contrastive topics.

Ka can also indicate a contrast, when it is used in alternative questions, as in (24).

    1. (24)
    1. Man’yōshū 3587
    1. kimi=ga
    2. you=GEN
    1. me=wo
    2. eye=ACC
    1. kepu=ka
    2. today=QUE
    1. asu=ka=to
    2. tomorrow=QUE=QUO
    1. ipapi-te
    2. forbear-GER
    1. mat-am.u.
    2. wait-FUT.ANP
    1. ‘…I will wait for you patiently, [wondering] whether it will be today or tomorrow.’

Ka is generally described as being a true focus on the constituent of the clause that it marks rather than on the whole clause, in contrast to ya (next section) (cf. e.g. Kondō 1987: 265). Nevertheless, when ka is used in yes/no-questions, it is sometimes unclear whether the constituent to which it is added actually constitutes the focus of the question, (cf. (25); Nomura 2005: 42–43; 2011: 92–93). Instead, it is also possible to consider the clause as a whole as focused, as the alternative translation indicates.

    1. (25)
    1. Man’yōshū 59
    1. wa=ga
    2. I=GEN
    1. se=no
    2. husband-GEN
    1. kimi=ga
    2. lord=GEN
    1. pitori=ka
    2. alone=QUE
    1. ne=ram.u.
    2. sleep-SPP.ANP
    1. ‘Will you, my lord and husband, sleep alone?’, or ‘Is it alone that you, my lord and husband will sleep?’

3.2 Focus concord constructions with ya and namo/namu

If the original focus concord construction particles ka and so/zo can be labeled as “strong” focus concord construction particles, then ya and namo/namu – which are also not found in Old Okinawan focus concord constructions – are their “weak” counterparts (cf. Ōno 1993: 335–340; or Fiala 2000: 324). For one thing, ya and namo/namu were unable to mark question words, and had no copular function. Also, sentence-final ya was suffixed to final and not to adnominal verb forms, as in (26).

    1. (26)
    1. Man’yōshū 1976
    1. pototogisu
    2. cuckoo
    1. naki-te
    2. cry-GER
    1. sa.watar.u.
    2. PRF.pass.FNP
    1. Kimi=pa
    2. you=GEN
    1. kiki-t.u=ya.
    2. hear-PFV.FNP=QUE
    1. ‘…the cuckoo has passed crying. Have you heard it?’

Ya was used in yes-/no-questions as in (27), including rhetorical questions, still competing with ka in this function in OJ.

    1. (27)
    1. Man’yōshū 74
    1. Miyosino=no
    2. PN=GEN
    1. yama=no
    2. mountain=GEN
    1. arasi=no
    2. storm=GEN
    1. samu.kyeku=ni
    2. cold.NMZ=ADV
    1. pata=ya
    2. perhaps=QUE
    1. koyopwi=mo
    2. tonight=FOC
    1. a=ga
    2. I=GEN
    1. pitori
    2. alone
    1. ne-m.u.
    2. sleep-FUT.ANP
    1. ‘Will I have to sleep alone again tonight in the cold of the storm blowing from Mt. Yoshino?’

In contrast to ka, ya is often presented as focusing the whole clause and not the particular constituent to which it is suffixed. Kinsui (2011: 162) writes that ya was independent of focus and even “avoided” the sentence focus/question word if there was one in a sentence. Kondō (1987: 264–265) similarly claims that with ya, “from its character, the scope of interrogation is the whole clause, and it doesn’t matter where ya appears.”

In OJ, ya was still much less frequent than ka in non-rhetorical questions.20 Later in EMJ, ka retreated to the domain of question words, and ya generalized to cover all other questions. Thus, Nomura (2002b: 32) considers focus concord constructions with so/zo and ka as the original focus concord construction, and ya as a newcomer.

Something similar holds for namo (EMJ namu), which is relatively rare in OJ, with only one occurrence in the Man’yōshū, but more occurrences in the somewhat later Senmyō (8th century imperial edicts) prose. Since poetry is conservative, namo’s virtual absence from poetry is also taken as an indicator of its novelty. Similar to ya encroaching on ka, Nomura (2002b: 35) assumes that namo/namu infiltrated on so/zo. (28) is a typical example from the Senmyō, where it often accentuates the contents of something thought or said.

    1. (28)
    1. Senmyō 48
    1. Ame=no
    2. heaven=GEN
    1. sita=no
    2. below=GEN
    1. maturigoto=pa
    2. government=TOP
    1. tapirakye.ku
    2. smooth.ADV
    1. yasurakye.ku
    2. peaceful.ADV
    1. tukape~matur.u=be.si=to=namo
    2. serve~conduct.FNP=DEO.FNP=QUO=FOC
    1. omoposimes.u.
    2. think[hon].ANP
    1. ‘[The emperor] deigns to think that the government of the world should be conducted smoothly and peacefully.’

The literature suggests that namo/namu marks a “mild” or “polite” emphasis of the speaker’s thoughts towards the hearer/reader, and is often used in contexts of explanation or persuasion (cf. e.g. Imuta 1976; Oda 2003), a function that is obviously less compatible with poetry than with prose. In the above example, the emperor speaks to his subjects, explaining how he thinks government should be conducted. The style is very formal and contains honorific language, expressing the nobility of the speaker.

It has furthermore been observed that it may never have been a strict focus on a specific constituent, but rather an emphasis on the whole clause in which it appears. Furthermore, it is unique in not being able to appear sentence-finally, and it is prone to be repeated in sentences that consist of two or more clauses. Oda (2003: 103) found that in a large corpus of EMJ literature, in 31% percent of subordinate clauses with namu, the following main clause also contains namu, as in ex (29).

    1. (29)
    1. Yamato monogatari 3 (265)
    1. [poem]=tote=namu
    2. [poem]=QUO=FOC
    1. yari~tamafe-ker.e=ba,
    2. send~give-PRT.CON=TOP
    1. itoni
    2. really
    1. na.ku
    2. not.be.ADV
    1. mede.te
    2. be.lovable=GER
    1. noti=made=namu
    2. later=LMT.FOC
    1. katari-ker.u.
    2. tell-PRT.ANP
    1. ‘Since he sent the poem saying [….], [Toshiko] was very much touched, and kept telling [everyone] about it until much later.’

In summary, ya and namo/namu were two focus particles that largely corresponded to ka and so/zo, respectively, in their meaning, but had some features that set them apart from the latter particles. The question particle ya, in contrast to ka, was not used with question words. Instead, it is frequently found in rhetorical questions. Namu was stylistically restricted, and seems to have emerged later historically than the other particles.

3.3 Focus concord constructions with koso

The focus concord constructions with koso are the most puzzling of all, since they don’t fit into the system of the other four focus concord constructions that seem to complement each other in the domain of declaratives and interrogatives. In the Man’yōshū, koso is used overwhelmingly clause-internally (161 out of 169 examples), and where it is found clause-finally, usually a missing main clause can be inferred (cf. Kōji 1988: 141–151). Otherwise, koso differs from so/zo and namo/namu primarily formally, namely (a) in its association with the predicate form –(UR)e or its past equivalent –sika, which are apparently nominalizing verb forms with primarily adverbial function, and (b) in being originally used in subordinate clauses built on this predicate form. (30) shows the non-past form, and (31) the past form.

    1. (30)
    1. Man’yōshū 145
    1. Pito=koso
    2. person=CFC
    1. sir-an.e,
    2. know-NEG.PRE
    1. matu=pa
    2. pine=TOP
    1. sir.u=ram.u.
    2. know.FNP=EPI.FNP
    1. ‘It is humans who don’t know it, but the pine will know it.’
    1. (31)
    1. Man’yōshū 1843
    1. Kinopu=koso
    2. yesterday=CFC
    1. tosi=pa
    2. year=TOP
    1. pate.sika
    2. end-PRP
    1. paru~kasumi
    2. spring~mist
    1. kasuga=no
    2. PN=GEN
    1. yama=ni=pa
    2. mountain=LOC=TOP
    1. haya
    2. early
    1. tati-ni-keri.
    2. rise-PFV.PRT
    1. ‘While it is just yesterday that the year has ended, [today] the spring mist is [already] rising on Mount Kasuga.’

In this manner, koso contributed to the formation of concessive and causal constructions. The generally assumed historical priority of the subordinate construction in which the conditional form is unmarked (cf. Ishida 1939; Ōno 1956b; Iwai 1970: 511; Ōno 1993: 115; Katsumata 2007; Kinsui 2011: 158), as in (30) and (31), is supported by the fact that we find it frequently in OJ but it disappears in EMJ, mainly in favor of non-subordinate main clause constructions, and subordinate constructions marked by additional particles.

With respect to its semantics, going back to 18th century scholars, koso has been identified as a particle that indicates a contrast and a “strong emphasis”. Okamura (1995: 174) cites the 1773 Ayuishō as stating that “[koso] chooses from different items and takes and [makes us] look at only one thing.” Likewise, more modern scholars like Ōno (1956c; d; 1993: 105–106), Fiala (2000: 358), Handō (2003: 22–23) and Oda (2003: 104) suggest that koso primarily indicated a contrast, or a selective choice between alternatives. Ōno (1993: 105) writes that “koso selects one among different things and shows the result to us as “This is it!”” Handō (2014: 232) states that koso forms “an absolute focus in an emphatic clause structure.” Vovin (2009: 1202) labels koso as “especially strong emphasis on a preceding word or phrase”.

As for the contrastive meaning, if we assume that the original use of clauses with presuppositional form was indeed concessive, which implies some contrast by definition, this could be either the original meaning of koso or a feature of the context that was imputed on koso via hypoanalysis. The contrastive use was eventually extended to mark a general emphasis on a constituent or clause (cf. Handō 2014: 232; “pure formalization” according to Fiala).21

In (32), no second clause follows the clause with the koso-constituent, but the meaning is nevertheless concessive, with a contrasting proposition implied.

    1. (32)
    1. Man’yōshū 1951
    1. Uretaki=ya
    2. annoying=EXC
    1. siko~pototogisu.
    2. bad-cuckoo
    1. Ima=koso=ba
    2. now=CFC=TOP
    1. kowe=no
    2. voice=GEN
    1. kar.u=gani
    2. get.hoarse.ANP=LIM
    1. ki
    2. come
    1. naki~toyome-m.e.
    2. cry-resound-FUT.PRE
    1. ‘Bad, bad cuckoo! You should come and sing now [and not before] so loud that your voice gets hoarse, [but you actually don’t].’

According to Kōji (1988: 741–751), among the 161 clause-internal koso-marked constituents in the Man’yōshū, 75 are found in concessive clauses (elliptic and non-elliptic), as in (30), (31), and (13) (Kōji 1988: 741–745), 8 in causal clauses (no example provided here) (Kōji 1988: 745–746), and 50 in main clauses without apparent concessive meanings (Kōji 1988: 746–749), as in (33) and (34).

    1. (33)
    1. Man’yōshū 1584
    1. Medurasi=to
    2. adorable=QUO
    1. a=ga
    2. I=GEN
    1. omop.u
    2. long.for.ANP
    1. kimi=pa
    2. lord=TOP
    1. aki~yama=no
    2. fall~moutain=GEN
    1. patu-momiti~ba=ni
    2. first-maple~leave=DAT
    1. ni.te=koso
    2. resemble.GER=CFC
    1. ari-ker.e.
    2. be-PRT.PRE
    1. ‘It is the first red leaves of the mountain in fall that you, who I adore, resemble.’
    1. (34)
    1. Man’yōshū 3493
    1. Oso~paya=mo
    2. late~early=FOC
    1. na=wo=koso
    2. you=ACC=CFC
    1. mat-am.e.
    2. wait-FUT.PRE
    1. Muka-tu-wo=no
    2. distant-GEN-hill=GEN
    1. sipi=no
    2. oak=DAT
    1. koyade=no
    2. twig=GEN
    1. api=pa
    2. meet=TOP
    1. tagap.azi
    2. miss.NFU
    1. ‘Early or late, it is for you that I will wait. We will meet without fail, like the twigs of the oak trees on the hills over there.’

There is one more not uncommon pattern, with 25 occurrences, in which koso marks a whole subordinate clause ending on the presuppositional form, as in (35) (Kōji 1988: 749–751). The main clause then also ends on the presuppositonal form.

    1. (35)
    1. Man’yōshū 118
    1. Nageki-tutu
    2. cry-DUR
    1. masurawo=no
    2. strong man=GEN
    1. kwo=no
    2. son=GEN
    1. kwop.ure=koso
    2. love.PRE=CFC
    1. wa=ga
    2. I=GEN
    1. yup.u
    2. tie.ANP
    1. kami=no
    2. hair=GEN
    1. piti-te
    2. get.wet-GER
    1. nure-ker.e.
    2. get.wet-PRT.PRE
    1. ‘It is because you strong man are longing for me with deep sighs, that my tied up hair has got wet and untied.’

This pattern can be classified as a subtype of main clause use (i.e., koso marking a constituent of the main clause) without the implication of concessivity. The striking fact is that the sentence now consists of two clauses with predicates in the presuppositional form, that is, two propositions in which the predicate is presenting given information. The most plausible interpretation is that koso is marking the causal relationship between the clauses as such as the focus (cf. also Kōji 1988: 749; and the modern Japanese translation in Kojima et al. 1971: 128).

Overall, it seems likely that koso originally occurred in subordinate clauses presenting presupposed information, and marked the informationally most salient constituent of the clause, often implying a contrast, because a concessive clause as such has a contrastive implication. This explains some of the oddities of koso: since the concessive clauses cannot contain question words it does not mark question words, and because of the status of these clauses as a whole as presuppositions, the information value of the marked constituent is not as high as that of focused constituents of the main clause. Koso then extended its uses as seen above.

The scenario of subordinate clauses converted into main clauses via ellipsis of the main clause is known in the literature as “insubordination”. Evans (2016: 2) defines it “diachronically as the recruitment of main clause structures from subordinate structures, or synchronically as the independent use of constructions exhibiting prima face characteristics of subordinate clauses”. (36) and (37) would be simple examples from Modern English.

(36) If I only had a brain!
(37) That he could say such a thing!

Insubordination can be divided into 3 steps (cf. Evans 2007: 370):

  1. Occasional ellipsis of the main clause

  2. Conventionalized ellipsis

  3. Reanalysis as main clause structure

The construction in (36), for example, may already be considered to be at stage (iii), that is, “conventionalized main clause use of [a] formally subordinate clause” (Evans 2007: 370). While a main clause can be reconstructed, the meaning of (36) is arguably not that of a full sentence augmented by a main clause as in (38), but that of the structurally unrelated simple clause in (39).

(38) If I only had a brain, I would be smarter.
(39) I wish I had a brain.

Insubordination has also been an extremely prolific phenomenon in recorded Japanese language history (cf. Narrog 2016). This includes concessive constructions such as those with the presuppositional forms in OJ. In Modern Japanese, the insubordinate use of concessive noni and keredo clauses is well-known. Note that noni forms syntactically more tightly integrated ‘although’ clauses, and keredo more loosely integrated ‘but’ clauses. (40) and (41) are constructed examples.

    1. (40)
    1. Kinoo
    2. yesterday
    1. ik-e.ta=noni.
    2. go-POT.PST=CNC
    1. ‘You should have gone yesterday!’ (Lit, ‘Although you could have gone yesterday’)
    1. (41)
    1. Sonna
    2. such
    1. koto=o
    2. thing=ACC
    1. sir-ana.i=keredo.
    2. know-NEG.NPS=AVS
    1. ‘I don’t know about that.’ (Lit., ‘I don’t know that, but…’)

Insubordinate constructions are cross-linguistically often (but by no means always) used for exclamation, as seen in (36), (37), and (40). If we assume that the sentences with presuppositional forms are indeed insubordinate, this may explain the label as “exclamative” that some Western scholars (Frellesvig 2010; Wrona 2010) have tagged on them. This label is quite puzzling as it suggests practically the opposite meaning of the Japanese term izenkei “given existence form” (and my own term “presuppositional”), but it may match the pragmatic effect in standalone use.

The OJ subordinated presuppositional verb forms are neither an exact equivalent of Modern Japanese noni as in (40) nor of keredo as in (41). Modern translations vary between both. However, in the case of the insubordinate construction with implied concessive reading as in (32), Modern Japanese native speakers’ intuitions seem to favor noni (cf. (42)).

    1. (42)
    1. Kojima et al. (1973: 77)
    1. Baka~hototogisu….
    2. stupid~cuckoo
    1. ki~naiki~toyomosu=ga
    2. come~cry~resound=NOM
    1. yo.i=noni.
    2. good.NPS=CONC
    1. ‘Bad cuckoo! You should come and sing [, but you actually don’t].’

Sentences such as (32) are so frequent that they must be considered at least as having reached stage (ii) of the insubordination process. If follows then that those sentences where the concessive meaning is difficult to retrieve, and Modern Japanese translations do not add concessive markers, such as (33) and (34), are the last stage of conventionalization. This would be an equivalent to the English insubordinate clause (36), where the clause is not necessarily understood as a condition anymore, and can be reformulated as a non-conditional clause as in (37). An alternative interpretation would be that the presuppositional form in these sentences does not correspond to a tightly integrated noni ‘although’ as in (32) (Modern Japanese (40)), but merely to a losely integrated keredo ‘but’ (Modern Japanese (41)), and that the implication of a following contrastive proposition is simply lost on modern speakers. This is difficult for us to resolve here, and not really crucial either. The historical fact is that sentences ending on presuppositional forms eventually lost their concessive implication.

The last question that we may want to clarify is whether OJ koso originally indicated a contrastive focus in general or only a contrastive topic. The latter would presuppose some special motivation, because of the close relationship between contrastive topic and contrastive focus. As noted in 2.1, the prevalent marking of contrastive topics in major European languages is focus prosody, which therefore as a means of expression covers both notions. Indeed, the semantic descriptions of the Japanese grammarians also strongly suggest the former: koso has persistently been associated not only with contrast, but also with “strengthening” (tsuyomeru) or “emphasis” (kyōchō), and not with backgrounding a constituent, or with indicating that it is a topic. Nevertheless, we need to make sure that the description actually matches our terminology. Recall that a contrastive topic is a focus in topic, that is, clause-initial position. Otherwise we are dealing with a contrastive focus. Koso in (30), (31), and (32) represents a focus in topic position, and thus is a contrastive topic. There is no shortage of such examples in the Man’yōshū. In contrast, koso in (33) and (34) is not in an apparent topic position. In (33), kimi=pa is the topic, and patu-momiti~ba is the contrastively focused information of the sentence. (34) lacks further context but is most likely a sentence without a topic, since already the first constituent is focused. (43) is another example in which koso is not in topic position.

    1. (43)
    1. Man’yōshū 2055
    1. Amanogapa
    2. galaxy
    1. topo.ki
    2. far.ANP
    1. watari=pa
    2. passage=TOP
    1. na.kyere=domo
    2. not.be.PRE=CNC
    1. kimi=ga
    2. lord=GEN
    1. puna~de=pa
    2. ship~leave=TOP
    1. tosi=koso
    2. year=FOC
    1. mat.e
    2. wait.PRE
    1. ‘The passage across the galaxy is not so long; but it is one year we have to wait for your ship to leave.’

In (43), quite the opposite, the preceding constituent puna~de=pa seems to be not merely a topic, but a contrastive topic, marking the contrast between the passage across the galaxy, which is (according to the author) not long, and the ferry service, which is only scarcely available.

So, while the distinction between contrastive focus and contrastive topic may not be decisive, we may conclude that koso was able to mark a contrastive focus, which was very often, but not exclusively, a contrastive topic.

3.4 Summary

This section has provided a descriptive overview of the Old Japanese focus concord constructions, divided into those with ka and so/zo (3.1), those with ya and namo/namu (3.2), and those with koso (3.3). Foregoing the details, ka and ya marked questions, so/zo and namo/namu a focus on a sentence constituent, mainly in assertions, and koso a contrastive focus, often in subordinate clauses and usually of the subtype of contrastive topic. As we have seen, ka, ya and namo/namu are only associated with focus, while so/zo and koso are also associated with contrastive topic, a function that was also fulfilled by pa. So, there was some functional overlap between the particles in terms of what broader categories of information structure they represented. To illustrate this overlap, we can go back to our representation of the notions of contrast in 2.1, and fill in so/zo, pa and koso into our schema in Figure 1, resulting in Figure 2.

Figure 2
Figure 2

OJ particles for topic and focus with a contrastive meaning.

The idea expressed in Figure 2 is that we have pa as a typical topic particle and so/zo as a typical focus particle. Koso indicates a contrast, at least a contrastive topic, but probably even broader a contrastive focus. The overlap between particles occurred in the area of “contrastive focus”, which is generally an area of overlap between topic and focus: Especially, pa could indicate a contrastive topic, while koso and so/zo could do so as well. So it may seem that there was substantial overlap in the “contrastive topic” function of the three particles, but the actual ambiguity was probably rather limited: pa was the default particle for contrastive topic, and the other ones were used with specific functions and in specific contexts: So/zo is only rarely found in this function, mainly to back up and strengthen the emphasis of the constituent already marked by pa, and koso was mainly used in very specific constructions (concessive clause, or insubordinate clause with concessive implication).

In this context, the question arises why koso in the intuition of most modern Japanese native speaker linguists, and also some non-native linguists is judged as a particularly “strong” emphasis in contrast to other particles (Ōno (1993: 105) compares koso figuratively to a “strong light beam”, for example, and Vovin (2009: 1202) claims that koso is “much stronger than the particles so/[zo] and namo”). It goes without saying that this intuition about koso’s meaning cannot be derived from phonological form of koso itself, but must be sought in other elements of the context. The obvious answer is first that in contrast to pa as a contrastive topic, koso not only highlighted the preceding constituent as a selection from several alternatives, but also was accompanied by a backgrounding (nominalizing) verb form in –(ur)e (and sika). That is, unlike in sentences with pa, not only was the preceding constituent highlighted, but the natural locus of focus in a clause, the concluding predicate, was explicitly backgrounded. However this is merely a general feature of the focus concord constructions, so we additionally need to answer what set koso apart from the other focus particles in focus concord constructions. Here it comes into play that koso was originally part of a sentence construction in which the subordinate clause marked with koso and –(u)re as a whole presented non-asserted/presupposed information. That is, it was originally the lone focused constituent in an otherwise fully backgrounded clause. Thus, the “strong light beam” effect (Ōno (1993: 105). In general, it is reasonable to assume that so/zo and koso added to, or emphasized a contrast as compared to a “default” marking with pa.

4 Competing analyses

In this section I will critically discuss the most important analyses that have been proposed for focus concord constructions in domestic Japanese linguistics and from outside. Most of these analyses are also linked to ideas about the origin of focus concord constructions in proto-Japanese. As of yet, there is still no consensus on the best analysis.

4.1 Focus concord constructions as clefts

Ōno (1956a; 1964; 1993), based on an idea by Tani (1889), proposed in comprehensive form that the majority of the focus concord constructions, namely those on ka, so/zo, namo/namu and ya, are originally inverted pseudo-clefts. In English, pseudo-clefts are cleft sentences with a what-clause, such as “What I miss most, is my cat”. The inversion would be something like “It is my cat what I missed”, which is, of course, not grammatical in English. It would also be a superfluous construction since the regular cleft “It is my cat that I miss most” is already available. In contrast, in Japanese all overt clefts are pseudo-clefts, and their inversion is colloquially possible. So this scenario is in principle possible.

In this scenario, then, the focus concord construction particles are originally emphatic clause-final particles, and the clause-final nominals and nominalized predicates are the topic of the original clause. Ōno (most explicitly in Ōno 1993: 194–202) based his analysis on constructions with the particle so/zo, and actual examples with two nominal predicates, since he considered nominal predicates to be equivalent to nominalized verbal predicates. The source construction in such sentences is [… noun=pa] XP=so/zo. (44) is an example for the presumptive source construction with two nominal predicates before inversion.

    1. (44)
    1. Man’yōshū 4109 (cf. Ōno 1993: 197)
    1. Kurenawi=pa
    2. safflower=TOP
    1. uturop.u
    2. fade.ANP
    1. mono=so.
    2. thing=FOC
    1. ‘The color of safflower fades.’

(45), then, shows an equivalent inverted construction.

    1. (45)
    1. Man’yōshū 2 (cf. Ōno 1993: 197)
    1. Umasi~kuni=so,
    2. beautiful~country=FOC
    1. akidusima
    2. PN
    1. yamato=no
    2. PN=GEN
    1. kuni=pa.
    2. country=TOP
    1. ‘The country of Yamato is a beautiful country.’

This very intuitive hypothesis was adopted into the general linguistic literature through Harris & Campbell (1995: 161–162). Furthermore, clefts in general are the best-known source of focus constructions cross-linguistically (cf. Heine & Kuteva 2002: 95–96). However, in Japanese linguistics, this hypothesis has been met with skepticism, mainly for the reason that the expected source structures with verbs are largely absent in OJ.22 For Ōno’s scenario to work, we would expect the three steps of development as in (46) ((46) is the attested target structure):

(46) (a) [… verb]=pa XP=so >
  (b) XP= so [ … verb]=pa >
  (c) XP= so [ … verb]

First of all, Ōno (1993) apparently chose examples with nominal predicates in the topic position as in (44) and (45) because instances of the crucial source construction (46)(a) are extremely scarce. Kōno (2010: 40) was able to identify two in the whole Man’yōshū (Ōno most likely did not find them, since he doesn’t cite them). Secondly, no single example can be found of the inverted construction (46)(b) with a verbal predicate in sentence-final position and the particle pa. There is no structural reason that would preclude the existence of (46)(b) with sentence-final pa.23 In fact, we find this construction with verbs nominalized with –(ura)ku, as in (47).

    1. (47)
    1. Man’yōshū 2725
    1. Mitu=no
    2. PN=GEN
    1. panipu=no
    2. clay=GEN
    1. iro=ni
    2. color=DAT
    1. ide.te
    2. appear.GER
    1. ip-an.aku=nomwi=so
    2. say-NEG.NMZ=LIM=FOC
    1. wa=ga
    2. I=GEN
    1. kwop.uraku=pa.
    2. love.NMZ=TOP
    1. ‘I only keep it to myself till it appears in my face like the red clay of Mitsu, the fact that I love [you].’

So, since the source structures are largely missing already in OJ, in order for Ōno’s hypothesis to work, we would have to assume that the focus concord constructions in OJ had already conventionalized and developed away from their source constructions to a considerable extent, thus resulting in this gap. But there is no particular grounds for such an assumption, and therefore, it has been claimed that Ōno’s hypothesis lacks empirical support (e.g. Yanagida 1985: 156; Wrona 2008: 199; Kōno 2010: 41).

There are a number of additional counterarguments to the inverted cleft hypothesis. Yanagida (1985: 156–157) shows that the inverted cleft structure has no obvious functional advantage to a non-inverted cleft with zo or with nari as the copula in terms of emphasizing the focused part of the sentence. Whitman (1997: 170) points out that extracting a focused element from within the topic and inserting it in the position of a predicate nominal would constitute a lowering operation. Wrona (2008: 199–201) argues, among others, that right dislocation as a feature of poetic texts, should not have influenced language change. Lastly, Shinzato & Serafim (2013: 246–248, 292) in their study of Okinawan vs. OJ focus concord constructions provide a number of reasons that the inverted pseudo-cleft hypothesis is the only major hypothesis of the origin of focus concord constructions that is incompatible with the Okinawan data.

Whitman (1997) also favors a cleft analysis of the focus concord constructions, but tries to avoid the manifold problems of the inverted pseudo-cleft analysis by positing an in-situ it cleft structure. For him, (49) is the structure of the focus concord construction in (48) on an abstract level of representation, while the surface structure is monoclausal as in (50).

    1. (48)
    1. Whitman (1997: 162; cited from Ise monogatari)
    1. …fafa=namu
    2. mother=FOC
    1. fuzifara=nari-ker.u
    1. ‘…it is mother who was a Fujiwara.’
(49) [[fafa=namu]I [ti fuzifara=nari-ker.u]]
(50) [fafa=namu fuzifara=nari-ker.u]

As support for his analysis, Whitman (1997: 171–172) cites the fact that both pre-modern Japanese focus concord constructions and English it-clefts do not occur in imperative and exclamatory contexts, and that wh-questions both in Japanese and Sinhala (which is also SOV and has focus concord constructions) can be analyzed as it-clefts in situ.

Schaffar (2002, 2003) also favors an analysis in terms of it-cleft. However, he posits three variants, one variant in situ (51), which he considers to be the most common for OJ, and to also hold for ModJ noda (see (54) below), one variant ex situ (52), and one variant with the particle at the end of the sentence ((53); Schaffar provides no actual examples for each construction). Note that *–re- is a nominalizing/complementizing suffix that is contained both in –(ur)u “adnominal present” and –(ur)e “presuppositional”, and *–u is copula-like inflection, added to “presuppositional” –(ur)e to render “adnominal present” –(ur)u. “Op” is an operator and “PredP” the predication phrase.

(51) [PredP ei [CP Opi […NPi- KAKARI…V][C0 –rei]] –u/0]]
(52) [PredP NPi- KAKARI [CP Opi […ti…V][C0 –rei]] –u/0]]
(53) [PredP ei [CP[…V][C0 –rei]] –u/0]- KAKARI

Schaffar assumes change from a pre-OJ original construction, in which according to his scenario (2002: 329) the inflection –re- was not yet a complementizer but a nominal, to the constructions presented in (51) to (53). The peculiarity of Schaffar’s analysis lies in the specific morphological analysis of the nominalizing verb forms, and the fact that he assigns copula status to the final inflection. By doing so, Schaffar unifies the analyses of focus concord constructions and the ModJ no da construction, as in ex. (54), which clearly has a clause-final copula (no desu is the polite form of no da).

    1. (54)
    1. Kuno (1973: 223)
    1. [Watasi=wa]
    2. [I=TOP]
    1. kaze=o
    2. cold=ACC
    1. hii.ta=no=des.u
    2. color=NMZ=COP.NPS
    1. ‘It is that I’ve caught a cold.’

At first sight, this seems syntactically more plausible than Ōno’s (1993) and Harris & Campbell’s (1995) analysis with clause-medial copula in a SOV language.24 On the other hand, while it is clear that at least some of the focus concord construction particles had a copular function, the assignment of copula status to the adnominal non-past inflection is dubious, since there is no independent evidence for copular function in OJ. Assuming a copula in this position may altogether be unnecessary, since as Wrona (2011) showed, at least in the history of no(=da) in Japanese, standalone nominalizing uses preceded the sentence-final nominalizer with copula. Also, the assumption of the presence of the nominalizing morpheme –re even in consonant stem verbs has been criticized (Wrona 2008: 201). Finally, one must add that Schaffar (2002; 2003) is putting forward hypotheses about structures without presenting any empirical evidence.

A general critique that can be leveraged at it-cleft hypotheses is that Japanese, under common analyses, only has the equivalent of pseudo-clefts and not it-clefts (cf. e.g. Kizu 2005: 3; Iwasaki 2013: 251).25

Lastly, Fiala (2000: 401, 417, 443) espouses the idea that focus concord constructions are “information-structural clefts” but not syntactic clefts, and serve to construe “subjective word order” in which a focused constituent is either moved to the front or remains in situ to emphasize the high information value of a particular constituent. While he sympathizes with Ōno’s (1993) intuitive analysis, for him, syntactic movement is inessential (not required) for a focus concord construction. Thus, his view seems to be compatible with both the cleft in-situ and with the particle insertion (4.4) analysis.

4.2 Focus concord constructions as wh-movement

An analysis similar to the cleft analyses in positing initial bi-clausality is that by Watanabe (2001; 2002; and elsewhere) of focus concord constructions as high wh-movement. It is based on an earlier observation by Nomura (1993a), who found the regularity of constituent order between focus particle-marked constituents and no/ga-marked “nominative” constituents in OJ Man’yōshū focus concord constructions (“X” and “Y” stand for some constituent) represented in Table 2.26

Table 2

Frequencies of ordering between akari-constituent and ga/no-marked constituent in OJ, according to Nomura (2002a).

Structure Frequency Structure Frequency
X-ka….Y-ga/no 120 Y-ga/no….X-ka max. 5
X-so….Y-ga/no 49 Y-ga/no….X-ka 1
X-ya….Y-ga/no 40 Y-ga/no….X-ka 0

In essence, the focus-particle marked constituents with very few exceptions precede the genitive and not vice versa. (55) is an example for this ordering with a ka-marked constituent, as provided by Watanabe (2002: 182).

    1. (55)
    1. Man’yōshū 3117
    1. iduku=yu=ka
    2. … where=ABL=QUE
    1. imo=ga
    2. wife=GEN
    1. iri~ki.te
    2. enter~come.GER
    1. yume=ni
    2. dream=DAT
    1. mie-t.uru.
    2. appear-PFV.ANP
    1. ‘…from where is it that my wife came and appeared in my dream?’

Nomura (1993a: 11) also showed that in relation to topic (pa)-marked constituents, the focus particle-marked constituents regularly followed them. Together with the assumption that ga/no in OJ already marked subjects, Watanabe (2002) concluded that the focus-marked constituents occupy a position above the subject but below the topic, that is, a focus position, to which they move in terms of wh-movement. The assumed structure is as in (56).

(56) [TopP Spec Top [FocP Spec Foc [IP Subj VP I]]]

Watanabe’s proposal has been the subject of some controversy. In OJ, the subject of main clauses was unmarked. Going back to the original data, Tonoike (2002: 87–88) pointed out that the number of sentences in the Man’yōshū in which the unmarked subject preceded instead of following the focus constituent was considerable (Tonoike cites over 40 by including 9 sentences with wh-phrases without focus particle while Nomura 1993a: 11 counts only 30). As a counterargument, an anonymous reviewer has suggested that the presumptive unmarked subjects preceding the focus constituent could also be analyzed as unmarked topics, since topic marking with pa was not obligatory in OJ. However, as Aldridge (2018: 11–14) shows, these unmarked subjects are very unlikely to be topics. In addition, there are even more examples in which objects or adverbial constituents preceded the focus constituent. In general, one should ask the question where in clause order the focus-marked constituents would be expected if they were not focus-marked. While the numbers in Table 2 are impressive, a large part of the focused constituents such as the focus-particle-marked adverbial clauses, which were especially frequent with ka, would have been placed before the subject in any case. So, the raw numbers may give a misleading impression. Still, even subtracting those adverbials, there is a regularity that requires attention.

Another issue that has been raised with regard to Watanabe’s hypothesis is his assumption that ga/no in OJ marked a subject that resided in [Spec, T], above objects. That this is a problematic assumption has been pointed out by a number of researchers including Kikuta (2004), Yanagida (2006) and Aldridge (2010). According to these studies, ga/no is still a genitive constituent, maximally residing in the Spec of vP (Wrona 2010 even suggests only VP). Specifically, Aldridge (2010) provides the following arguments that ga/no-marked constituent has a base position inside vP and moves no higher than to [Spec, TP].

  • – the genitive case particles ga or no appear only on subjects of embedded, typically nominalized, clauses and do not mark subjects of finite root clauses. Therefore, there is no reason to assume that they are structural nominative case markers (Aldridge 2010: 551–553).

  • wo-marked objects move to the outer specifier of vP while the genitive subject remains in its base position in [Spec, vP] (Aldridge 2010: 553–554).

  • wh-constituents can follow nominative subjects, scrambled objects, and temporal adverbials, and sometimes follow two or more of these (Aldridge 2010: 554–556) (however, note that the nominative subjects, since they are unmarked, are also amenable to an analysis as unmarked topics). While it is possible to assume multiple topics, the constituents preceding the wh-constituent do not observe their original order in the TP-domain as would be expected under Minimality.

  • – in a clause with two focused constituents, the one marked by ka occupies the lower position (Aldridge 2010: 557–559).

In conclusion, Aldridge (2010: 562) proposed a low (short) wh-movement analysis as in (57).

(57) [CP Cwh [TP [FocP [YP … XPwh··· ka] [vP .. tYP … ]]]]

While the description by Aldridge (2009) seems better supported by the facts than Watanabe’s (2002), a general weakness of approaches in terms of wh-movement remains, namely, they do not explain why adnominal verb forms and nominalized clauses are a required part of the construction. Furthermore, one could bring up the fact that wh-movement is not common in SOV languages, but a number of exceptions like Quechua, Hungarian, Basque or Malayalam are known (cf. e.g. Watanabe 2002: 189; Jayaseelan 2004).

Watanabe (2002, and elsewhere) treats the verb forms like mechanical reflexes of agreement with K-particles, specifically koso agreeing with the presuppositional inflection and the other particles agreeing with the adnominal, which could in principle involve any verb form except the expected finite form. However, in a criticism of the wh-movement approach, Wrona (2010: 186–187) puts even the assumption that OJ focus concord constructions involve agreement into question: “A predicate in the Adnominal or Exclamatory form on its own was in some cases sufficient to signal that the sentence-type was a question or an exclamation (non-declarative), independently of K-particles […]. Thus the Adnominal and Exclamatory form were independently motivated and did not agree with the K-particle.” One could add that there are also a few examples in OJ where clause-internal focus particles do not accompany adnominal but rather finite inflection on the predicate (cf. Kawabata 1994: 5). In (58), the predicate kinu ends finite –u instead of adnominal –uru.

    1. (58)
    1. Man’yōshū 4401
    1. Kara~koromu
    2. Chinese~coat
    1. suswo=ni
    2. hem=DAT
    1. tori~tuki
    2. take~attach
    1. nak.u
    2. cry=ANP
    1. kwora=wo
    2. child=ACC
    1. oki-te=so
    2. put=GER=FOC
    1. ki-n.u=ya.
    2. come-PFV.FNP=EXC
    1. ‘I have come [here] leaving behind my child that was crying, clinging to the hems of my coat.’

There are only six examples of this kind, and most of them are dialectal, as (58). However, according to Wrona (2010: 187), it was only later, in the Heian period that an “agreement” relation obtained: “Later the K-particles were reanalyzed as agreeing with the Adnominal or Exclamatory verb morphology, but this was a post-OJ development.”

Overall, the wh-movement hypothesis has attracted more attention and has been discussed more controversially in the English-language literature than any other hypothesis about the Japanese focus concord constructions in the past 20 years. Its strength is its appeal to quasi-universal syntactic mechanisms, while critics have pointed to potential inconsistencies with the OJ data, and the lack of motivation for the characteristic adnominal verb forms.27

4.3 Focus concord constructions as merger of two juxtaposed clauses

Nomura took his own findings about word order regularities in focus concord construction clauses in a different direction. His scenario is mainly based on an OJ synchronic analysis of focus concord constructions with ka. Kōno (2010) provides the same type of analysis in some detail for focus concord constructions with so/zo. Nomura suggested that the final predicate in a focus concord construction originated from a “kantaiku”, that is, an exclamatory clause with a nominal predicate used mainly to describe states-of-affairs before the speaker’s/writer’s eyes as a compressed unit, without inserting an explicit modal judgment of the speaker. This bareness of expression presumably conveys the immediacy of the speaker’s affect (cf. (7) in section 2.3 for an example; Nomura 1995: 9–10; 2002b: 13). The idea, then, is that an adjunct (“comment”) clause with focus particles, usually for the purpose of giving some reason or other background for the state-of-affairs described in the nominalized clause, was added before it. This results in sentences such as (59), which Nomura (2002b: 16) calls a “comment-like juxtaposition of two clauses” (chūshakuteki nibun renchi).28

    1. (59)
    1. Man’yōshū 712
    1. ipap.u
    2. …worship.ANP
    1. sugi
    2. cedar
    1. te~pure.si
    2. hand~touch.APT
    1. tumi=ka
    2. violation=QUE
    1. kimi=ni
    2. you=DAT
    1. api~kata.ki.
    2. meet~difficult.ANP
    1. ‘Is it [because] I broke the rules, touching the sacred cedar with my hands, that it is so difficult to meet you?’

The emergence of such complex sentences presupposes (1) the existence of nominalized clauses (kantaiku), and (2) the status of so/zo and ka as “predicative” (jutsugo kōseiteki) copula-like particles. Both should be uncontroversial, but note that nominalized clauses unrelated to focus particles were infrequent in OJ (cf. 2.3).

Nomura’s (2002b) analysis is based on a diachronic analysis in terms of three stages of development that can be interpreted as stages of grammaticalization. The synchronic occurrences of focus concord constructions (with ka and so/zo) and related structures in OJ are classified in terms of how far they have proceeded along these stages. At the first stage, we have the nominalized clauses in construction with the preceding clauses ending on final so/zo and ka as comments on the nominalized clauses. The sentence (59) is representative of that stage, and is not yet considered to be a focus concord construction by Nomura. According to Nomura (2002b: 27), they turn into focus concord constructions at the next stage, namely when the information-structural weight gets shifted from the main clause to the comment clause, as seems to be the case in (60).29

    1. (60)
    1. Man’yōshū 276
    1. imwo=mo
    2. …lover=FOC
    1. ware=mo
    2. I=FOC
    1. pitotu=nar.e=ka=mo,…
    2. one=COP.PRE=QUE=FOC
    1. putami=no
    2. PN=GEN
    1. miti=yu
    2. way=ABL
    1. wakare-kane-t.uru.
    2. part-NPO-PFV.ANP
    1. ‘…is it because my lover and me are one, that we cannot depart on the way of Putami [place name; literally ‘two views’].’

Nomura’s clause merger hypothesis is supported by the fact that constructions with focus concord particles on adverbial clauses that could be either interpreted as comment clauses or part of a focus concord construction are highly frequent in OJ (cf. section 3.1), and steeply decline in frequency in EMJ, even within the same genre of poetry (cf. Nomura 2002b: 28). So they do seem to be an older type of construction, although surely, frequencies alone are not conclusive evidence. At the third stage, in EMJ, the comment clause-structures like those in (59) start to have finitely marked predicates (Nomura 2002b: 15). This leads to a disassociation between clause nominalization and focus.

Currently Nomura’s hypothesis is the most widely accepted by linguists in Japan, which is also due to the fact that it is based on a long research record on the topic by the author, and a more exhaustive analysis of the empirical data of OJ than any other hypothesis. Nevertheless, in English writing on focus concord constructions, this hypothesis has been largely ignored and Nomura is practically only cited for his 1993 paper on ga and no that includes data on the order of focus-marked constituent vs. genitive subject constituent. The only published critique of Nomura that I am aware of comes from Fiala (2000: 392–393) and Yanagida (2016: 146–148). Fiala argues from an information-structural point of view and suggests that (1) the “comment-like juxtaposition of two clauses” approach cannot explain why the informational weight of focus concord construction structures is on the focus-marked part rather than the predicate”; and (2) this hypothesis is not good at explaining why focus concord constructions as we know them from the synchronic materials are obviously mostly mono- rather than bi-clausal. With respect to (1), as seen above, Nomura (2002: 27) proposed that at the second stage of development there was a diachronic shift in informational weight. With respect to (2), Nomura (2011: 86), viewed focus concord constructions as bi-clausal in OJ, becoming mono-clausal in EMJ, just as Watanabe (cf. 4.2). This is in contrast to many other authors that analyze focus concord constructions as mono-clausal already in OJ. Yanagida (2016: 146–148) main criticism is that a function of language as fundamental as a question should not be based on a complex construction with a comment clause. Further potential weak points are that Nomura’s hypothesis does not have the intuitive appeal of the inverted-cleft hypothesis and also has no known cross-linguistic parallels like the cleft hypothesis and the focus-movement hypothesis. However, note that juxtaposed clauses are cross-linguistically a frequent source of grammatical structure in general. They are best known as the source of relative clauses30, but also seem to function as complement clauses, purpose clauses, conditional clauses etc. (e.g. Mithun 1988: 354–356; Heine & Kuteva 2007: 224–225; Palancar 2012). To these critiques one could add that nominalized clauses unrelated to focus constructions did exist in Old Japanese, but were infrequent (cf. 2.3). Furthermore, Nomura’s approach works best for the data with ka, while for so/zo, Kōno (2010) struggles to find comment-clause constructions with the same structure in a similar quantity.

Morishige (1959: 226–229; 1971: 159–161), Yanagida (1985: 158; 2016: 148–162) and Quinn (1997) entertain an idea of focus concord constructions similar to Nomura’s, as originating from two juxtaposed clauses, the first of which ended on a sentence-final particle.31 But their idea of the original semantic relationship between the clauses is different. I will label their hypothesis as “added clause hypothesis” here in reference to Yanagida’s (2016) term hosoku ‘addition, supplement’. Yanagida (1985: 158), for example, writes that, “[focus concord constructions on zo] did not result from inversion, but from first presenting the emphasized clause rounded off by zo as one sentence to the hearer, and then continuing the narration.” Thus, in their scenario, the first clause carried focus from the beginning. That is explicitly claimed by Quinn (1997). For him, ka and so/zo were originally demonstrative pronouns and attached to the preceding clause as an afterthought. The following adnominalized clause is also an afterthought. Example (61) illustrates this idea:

    1. (61)
    1. Man’yōshū 870 (cf. Quinn 1997: 79)
    1. kepu
    2. …today
    1. yuki.te
    2. go-GER
    1. asu=pa
    2. tomorrow=TOP
    1. ki-n-am.u=wo
    2. come-PFV-FUT.ANP=CNC
    1. nani=ka
    2. what=QUE
    1. sayar-e.ru#
    2. impede-RES-ANP
    1. ‘…what is it that keeps me [from traveling the Matsura Road], although I could go today and come back tomorrow?’

According to Quinn (1997: 79), nani ka ‘what is it?’ was originally a clause on its own, and ‘that keeps me’ (‘that’s in the way’) was an afterthought to complement this clause. That is, unlike in Nomura’s hypothesis, the phrase with the focus particle was the focus already at the origin of the construction, and the following nominalized clause was known or backgrounded information.

For Yanagida (2016), the first clause with the focus particle is always “new information”, while the second, nominalized clause may be either “new” or “old information”. The second clause simply ‘supplements’ or ‘adds to’ (hosoku) the first clause. Furthermore, he suggests that it was sentences such as (62), in which the focused element of the first clause (‘a tiger’) is identical to the subject of the second clause (‘[someone] roars’) in which the shift from sentence-final particle to sentence-internal focus particle, and thus to a mono-clausal construction probably took place (Yanagida 2016: 153–156).

    1. (62)
    1. ata
    2. enemy
    1. mi-tar.u
    2. see-RES.ANP
    1. tora=ka
    2. tiger=QUE
    1. poy.uru.
    2. roar-ANP
    1. ‘…is it a tiger facing an enemy that roars?’

According to this view, clause integration then progressed from core towards peripheral constituents. The obvious advantage with the “added clause hypothesis” is that the information structure of the source construction can already be that of the focus concord construction (focus preceding presupposition).

Quinn (1997) is criticized by Fiala (2000: 399) and Wrona (2008: 206–207) on various accounts. Both believe that a right dislocation or afterthought is not likely to be the source of a major clause structure. Wrona (2008: 206) understands right dislocation as a “mainly poetic device” but overlooks that is in fact a pervasive feature of conversational language, at least to the extent that we can tell from Modern Japanese (cf. e.g. Ono 2006). Wrona also points out, among others, that ka was very infrequent as a demonstrative pronoun in OJ. This is a critique that could also be leveraged against any other hypothesis suggesting that the particle ka derives from the demonstrative. However, the presumptive status of ka as a common demonstrative does not have to apply to OJ, but to a potentially much earlier historical period at which the demonstrative turned into a final particle, and from there into a particle marking clause-internal constituents. In contrast, Yanagida’s hypothesis has not been subject to direct criticism yet, as far as I am aware of. It shares many of the features of Nomura’s and Quinn’s juxtaposition hypotheses, and would therefore potentially attract the same criticism. Furthermore, its presentation is more detailed than Quinn’s but still much sketchier than Nomura’s, and it lacks the differentiation with respect to the divergent features of constructions with different particles.

In conclusion, the Morishige/Yanagida/Quinn hypothesis seems just as plausible as Nomura’s, and has the advantage that it does not require a reversal in information-structural weight. However, it has not been fleshed out yet with systematic empirical data, or in terms of stages of development, as Nomura’s hypothesis.

4.4 Focus concord constructions as particle insertion

Lastly, there has also been the proposal that focus concord constructions resulted from “insertion” (sōnyū) of a focus particle into a nominalized (kantaiku) clause (e.g. Sakakura 1993; Funagi 2013: 307–308). This is a line of thought whose origins can be traced back to the Yamada (1908).32 Wrona (2008), ignoring the previous literature written in Japanese, independently makes the same proposal in English. For him, clauses nominalized on the adnominal, and the “exclamative” (i.e., presuppositional) form, already had the functions of questions and exclamations, and the focus particles were merely added “to mark the scope and/or type of the question or focus construction” (Wrona 2008: 209). As we have already seen above, especially sentences with a wh-word are often nominalized even without a focus particle, as in (63), provided by Wrona (2008: 218).

    1. (63)
    1. Nan=no
    2. what=GEN
    1. tapakoto
    2. nonsense
    1. ima~sara=ni
    2. now~again=ADV
    1. warapa~goto
    2. child~speech
    1. s.uru,
    2. do.ANP
    1. oi~pito=ni
    2. old~person=DAT
    1. si.te.
    2. do.GER
    1. ‘What nonsense, even now you talk like a child, though you are so old!’

Now, it is not necessary to add a focus particle to this sentence, but in principle it is conceivable that eventually focus particles were added for additional emphasis.

It is Wrona’s idea, then, that neither the adnominal verb form required a focus particle, nor a focus particle required a nominalizing verb form, but the nominalized sentences existed first, and then particle insertion occurred as a choice, that is, independently of the verb form, and not required by the verb. Wrona (2008: 228) thus suggests the original structure as in (64), whereby “K” is the kakari particle and “M” the musubi verb form.

(64) [[NP-K] [VP-M]]

Perhaps the most radical view of focus concord construction as particle insertion has been proffered by Yamada (2004), who considers focus concord constructions to be a purely stylistic and rhetorical, and not a syntactic, phenomenon. For him, it is the sentence-endings that selected particles for rhetorical purposes, rather than vice versa. Putting his extreme stance succinctly, “kakari-musubi did not exist” (Yamada 2004: 30). That is, it did not exist as a syntactic structure but merely as a stylistic feature.

Overall, though, the insertion hypothesis has not been well received in research in Japan (cf. Nomura; Kinsui), especially since it does not convincingly explain the nominalizing verb forms, but simply shifts the burden of their explanation to pre-existing nominal structures. The particle-insertion hypothesis has issues in two more areas. One is the fact that the nominalized structures on which they allegedly piggybacked, except for those with wh-words, are rare in OJ (cf. 2.3), by far less common than the focus concord constructions themselves. Also, they have hardly any flexibility with respect to modality, even as late as in the EMJ of the Heian period. Based on the modal endings found on each clause type, Kondō (2000: 248–260) argues that the focus concord construction clauses are basically verbal, since they allow modal suffixes such as –(a)m-, –kem-, –(a)ma.si, –(a)zi, rasi etc. (65) (ex. (25) above) would be an example with –(a)m- on the final predicate.

    1. (65)
    1. Man’yōshū 59
    1. wa=ga
    2. I=GEN
    1. se=no
    2. husband-GEN
    1. kimi=ga
    2. lord=GEN
    1. pitori=ka
    2. alone=QUE
    1. ne=ram.u.
    2. sleep-SPP.ANP
    1. ‘Will you, my lord and husband, sleep alone?’, or ‘Is it alone that you, my lord and husband will sleep?’

In contrast, the clauses with the adnominal ending but without focus particles do not allow such modal endings. This still holds for the Heian period, when the “pure” adnominal-ending clauses finally increase in frequency. It seems then, that the pure adnominal-ending clauses were less developed (less grammaticalized) in comparison to the focus concord construction clauses, and actually spread later. This needs to be explained.

One more point that seems difficult to account for with the particle-insertion hypothesis is why the focus concord constructions were restricted to specific particles. If it was just a matter of inserting a particle into an extant structure, what would be the constraint that prevents speakers from freely inserting semantically similar focus particles such as si or mo? Furthermore the hypothesis does not explain the differences in features of the focus particles actually used in the focus concord constructions (cf. Table 1). Lastly, the overwhelming consensus in previous research in Japan is that the presuppositional forms do not have an original “exclamative” but a subordinating function (cf. e.g. Ōno 1956b; 1993: 101–102; Kinsui 2011: 158). Wrona (2008), perhaps not being aware of the research literature, does not offer arguments against the consensus assumption or in favor of his own claim.

In summary, the fourth major hypotheses that has been proposed (Sakakura 1993; Wrona 2008; Funagi 2013) assumes that the focus concord constructions were simply an extension of already extant nominalized clause constructions without specific focus, through the insertion of focus particles on focused elements. This raises a number of questions, e.g. where these focus particles came from, why we only find specific focus particles in these constructions, and why the source constructions were not more frequent in OJ.

But the fact that this hypothesis raises a number of questions is shared with the other three hypotheses presented here.

In conclusion to this section, I have presented the four most influential hypotheses about the source of the focus concord construction, and offered evidence and arguments both in favor and against each hypothesis. Since the emergence of the focus concord construction precedes historical documentation, none of the evidence can be considered as conclusive. It does seem, though, that not all hypotheses are equally plausible, and in the final section I will put forward a proposal that should account well for the features of the focus concord constructions and their components.

5 Conclusion

5.1 Summary

In this paper I have first provided an overview of the descriptive features of the focus concord constructions in sections 2 and 3, and then critically discussed their interpretation in section 4. If we want to account exhaustively for the actual morphological features of the focus concord constructions and their constitutive elements, we cannot dismiss some of them as accidental. Furthermore, we want to do so in a diachronically accountable manner. Also, despite the fact that the focus concord constructions are considered to be a cross-linguistic rarity, we would like the analysis to be intuitively accessible, with parallels to developments in other languages. Under these premises, we have seen that none of the analyses presented in section 4 is without flaws. The inverted pseudo-cleft hypothesis has often been acknowledged as particularly intuitive, and would to some extent parallel the best-known path of development for focus constructions, but has been criticized on a wider range of accounts than any other hypothesis (cf. 4.1). Especially, supporting empirical data are lacking in OJ as well as in Okinawan (cf. Shinzato & Serafim 2013: 244–256, 286–287). In its present state it also cannot account for the use of namu (which is not found clause-finally) in the construction and cannot explain why among the nominalizing inflections it is the adnominal one that participates in the construction, and not –(ura)ku, which is the nominalizing verb form commonly used to form clefts in OJ (cf. 2.3). Schaffar’s it-cleft hypothesis suffers from positing implausible morphological structures for the sake of his syntactic analysis, and furthermore does not account for the differences between the particles. The wh-movement approach in its most recent form as short wh-movement offers the most solid syntactic description of the structure of the OJ focus concord constructions currently available. However, it does not account for a number of features of the focus concord constructions, especially the nominalization of the predicate, the copular function of some particles but not others, or why some focus particles participate in focus concord constructions and not others (Table 1). The juxtaposed clause merger account (4.3) lacks attested cross-linguistic parallels in the area of information structure. Both the clause juxtaposition and the particle insertion hypothesis (4.4) are marred by the fact that standalone nominalized clauses that form its basis were rare and modally restricted. The problem is even bigger for the particle insertion hypothesis because it entirely relies on the existence of full-fledged nominalized clauses on a par with regular non-nominalized clauses, while in the juxtaposition hypotheses, the nominalized clauses can be fragmentary additions to the preceding clause. The particle insertion hypothesis does furthermore not account for the differences between the various focus particles, those that are actually used in these constructions, and those that are not.

In this manner, each of the hypotheses on the emergence of the focus concord constructions in Japanese has at least one major shortcoming, and all of them have been subject to considerable criticism. However, there is no particularly plausible fifth alternative in sight either. What I will propose as a solution in the last subsection is a combination of two of the hypotheses that may be especially satisfactory as a diachronic account from the perspective of grammaticalization.

5.2 Proposal

While each of the hypotheses has its flaws, from the perspective of diachronic change in general and grammaticalization in particular, a clause merger hypothesis is especially attractive, because merger or “integration” of two clauses into one is a particularly common grammaticalization strategy (e.g. Heine & Kuteva 2007: 224–254). More specifically, bi-clausal genesis is even the major source of focus constructions attested cross-linguistically (cf. Heine & Kuteva 2002: 95–96, 111–112), and in terms of attested source morphemes (copulas, demonstratives) for focus markers, so/zo and ka also fit the bill. While from a cross-linguistic perspective, clefts (4.1) are more likely candidates than juxtaposition (4.3) in the area of information structure, they lack empirical support in OJ data. The biggest detractor for the juxtaposition hypotheses by Nomura, Quinn, Morishige and Yanagida is a lack of cross-linguistic parallels in the historical linguistic literature. However, the OJ Japanese focus concord constructions differ from the well-known cleft-derived focus constructions in historically well-documented SVO languages in any case. So this lack would not be surprising. Clause juxtaposition would also explain why the nominalized clause must take the adnominal form, not verb base or –(ura)ku, since neither the verb base nor –(ura)ku can form independent clauses. On the other hand, problems for the clause merger approaches arise especially if one assumes that the focus concord constructions in OJ are all bi-clausal, or that all particles participating in the constructions (that is, also including ya and namu) should have the same status in their diachronic development.33 But these are not necessary assumptions.

In the following I will briefly describe a diachronic scenario of the emergence of OJ focus concord constructions that I consider to be most likely, based on clause juxtaposition (4.3). This scenario could fit both the “comment clause” and the “added clause” hypotheses, but I consider the “added clause” hypothesis, which does not require a reversal in information structure, as more likely. In this scenario, the focus concord constructions developed from bi-clausal constructions with so/zo and ka, and had already become largely mono-clausal by OJ. At this stage, particle insertion can account for analogical extension through the new focus concord constructions with ya and namo. The outlier focus concord construction on koso should be considered as intermediate.

To start with, if we take the different features of the focus particles (Table 1) seriously, the constructions with focus particles can be divided into four groups, namely,

  1. a group of focus concord constructions with so/zo and ka, which are original deictics and have copular (predicative) function, and clause-final position, and are also present in Old Okinawan.

  2. focus concord constructions with koso, which share with group (1) the fact that they are also present in Old Okinawan, and may have deictic origin. However, they are not found clause-finally or predicatively.

  3. focus concord constructions with ya and namo, which are not found in Old Okinawan and are neither deictic in origin nor have copula function. There is one differentiation in this group, namely that ya but not namo is also found clause-finally.

  4. Constructions with focus particles that do not involve concord, namely mo (“scalar focus”), si (“emphasis”), and pa in its contrastive topic use. None of the features of the particles in (1) apply to them.

We can now assume that the starting point of the focus concord constructions was a construction with two clauses. The first clause ended in so/zo (emphasis/focus) or ka (question/doubt) with predicative function, and the second clause was nominalized. The first clause typically expressed a discovery by the speaker, or something that s/he wanted to alert the addressee to, while the second clause provided the fact that is the subject of the discovery, or added a background for the discovery. These would be sentences such as (16), here rendered again as (66), for ka, and (67) for so/zo.

    1. (66)
    1. Man’yōshū 276
    1. Imwo=mo
    2. lover=FOC
    1. wa~re=mo
    2. I~PRN.FOC
    1. pitotu=nar.e=ka=mo
    2. one=COP.CON=QUE=FOC
    1. mikawa=nar.u
    2. PN=COP.ANP
    1. putami=no
    2. PN=GEN
    1. miti=yu
    2. way=ABL
    1. wakare-kane-t.uru.
    2. part-NPO-PFV.ANP
    1. ‘Is it because my lover and me are one, that we cannot part from the way on Futami (‘Two-Look’) in Mikawa?’
    1. (67)
    1. Man’yōshū 2945
    1. …kimi=ga
    2. …lord=GEN
    1. tukapi=wo
    2. servant=ACC
    1. mati.si
    2. wait=APT
    1. yo=no
    2. night=GEN
    1. nagori-so.
    2. ember=FOC
    1. Ima=mo
    2. now=FOC
    1. i
    2. sleep
    1. ne-n.u
    2. sleep=PFV=ANP
    1. yo=no
    2. night=GEN
    1. opo.ki.
    2. many= ANP
    1. ‘…it’s the ember of the night when I waited for your messenger; the fact that even now I cannot sleep so many nights.’

Koso, ya, and namu were not yet involved.

The second stage is clause integration. As suggested by Kōno (2010) and Yanagida (2016), this was prone to happen when the focused element was a core argument of the nominalized predicate, as in (62), here repeated as (68).

    1. (68)
    1. ata
    2. enemy
    1. mi-tar.u
    2. see-RES.ANP
    1. tora=ka
    2. tiger=QUE
    1. poy.uru.
    2. roar-ANP
    1. ‘…is it a tiger facing an enemy that roars?’

This then, was the “switch context” from juxtaposed clause construction to focus concord construction.

Once the two clauses had merged, the constructions on koso, and on ya, namo could be formed on analogy. Therefore we do not have to require that they also had copular function or would form independent clauses, as ka and so/zo had to at the stage before the merger. In the case of ya and namo, this is simply possible by inserting these particles in the same or similar position as ka and so/zo. That is, only at this later stage, particle insertion (4.4) operated. The similarity of the original focus concord construction particles in (1) is greater with ya than with namo, since the latter had no clause-final function. In fact, as we have already seen, namo is most likely a more recent development than ya, since namo is not yet found in OJ poetry, poetry being a conservative genre, but only in the later Semmyō. There is little doubt that the extension to both ya and namo took place after the split between mainland Japanese and Okinawan.

The emergence of the construction with koso must have preceded those with ya and namo since we find it in Okinawan. A case can be made that koso is also deictic in origin; so in this sense it exhibits greater similarity to the original constructions than namo and ya. However it is not found predicatively or clause-finally, so we must assume that it does not represent the same first stage as ka and so/zo, but also emerged on analogy with the already established construction, with the specific use in subordinate clauses.

In the view presented here, it would have been entirely conceivable that mo and si, which share the fewest similarities with so/zo and ka34 had also entered the focus concord constructions by analogy if their development and extension had persisted. However, the last member to be added, probably not much earlier than the 8th century, was namo, and no later than from the 10th century, the focus concord constructions started their decline. This decline was, as some scholars have argued, at least partially due to the focus concord constructions’ own success, i.e., their expansion beyond the original core, which resulted in the dilution of their functional motivation and contributed to the proliferation of sentence-final adnominal and presuppositional forms. Had si or mo joined the focus concord constructions, it would have only additionally accelerated the markedness-reversal between final and nominalizing forms, and consequentially the decline of the focus concord constructions.

The question of how focus concord constructions should be analyzed syntactically as synchronic structures in OJ and EMJ is not necessarily dependent on their origin. We lack clear evidence for a bi-clausal analysis already in OJ, except in cases such as (66) and (67), where we have two clauses in a more trivial sense. Therefore, a syntactic description should have room for both a bi-clausal structure for the genesis of focus concord constructions, and a monoclausal structure for focus concord constructions at a mature stage. For a synchronic description in OJ, Aldridge’s (2009; 2018) short wh-movement approach is the one spelt out most solidly.


ABL = ablative; ACC = accusative; ADV = adverbial; ANP = adnominal present; ASS = assertive; AVS = adversative; CAU = causative; CFC = contrastive focus; COM = comitative; CON = conditional; COP = copula; CPL = complementizer; DAT = dative; DEO = deontic; DUR = durative; EMP = emphatic; EPI = epistemic; EXC = exclamative; FNP = finite non-past; FOC = focus; FUT = future; GEN = genitive; GER = gerund; HML = humilitive; HON = honorific; LIM = limitative; LOC = locative; NEG = negation; NMZ = nominalization; NPO = negative potential; PFV = perfective; PN = proper noun; PRE = presuppositional; PRG = progressive; PRN = pronominalizer; PRT = (remote) preterite; PST = past tense; QUE = interrogative; QUO = quotative; RCP = reciprocal; SPP = suppositive; TOP = topic; EMJ = Early Middle Japanese; ModJ = Modern Japanese; OJ = Old Japanese


  1. An alternative translation for kakari-musubi would be ‘relating-concluding’. Note, though, that in its literal physical sense kakar- means something like ‘hang on’ or ‘cling’ to something, that is, an asymmetric relationship between two entities, and not a symmetric or reciprocal relationship, as ‘relate’ may imply in everyday language. [^]
  2. For explanations of the original Edo period concept and its integration into modern language studies by Yamada Yoshio, see Funagi (1987; 2013). [^]
  3. Cf. also Gundel & Fretheim (2004: 176): “Topic is what the sentence is about”, or Luraghi & Parodi (2008: 186): “The topic or theme is what is being talked about in a sentence”, or Radford et al. (2009: 391): “Broadly speaking, the topic of a sentence (or utterance of a sentence) is what the sentence is about”. [^]
  4. Alternatively, Steedman (2000) uses the term “theme focus”. [^]
  5. Even Erteschik-Shir (2007), who continues to classify contrastive topics as topics, essentially shares this analysis. In the following example, ‘John’ is an example for a contrastive topic: [^]
    (i) Erteschik-Shir (2007: 48)
      B: Tell me about your brothers John and Bill.
      A: JOHN is the smart one.
    [^] His analysis goes as follows: [^]
    (ii) Erteschik-Shir (2007: 48)
      [{Johnfoc, Bill}top]top [is the smart one]foc
    [^] That is, for Erteschik-Shir, John is a focus embedded in a topical set that serves as a topic. [^]
  6. Furthermore supporting the lack of distinguishability of contrastive topic from contrastive focus and focus in general is recent phonological research that shows that the prosodic features of contrastive topic in major European languages does not differ substantially from prosodic focus (cf. van Hoof 2003; Repp 2016). [^]
  7. Note that there are even more competing concepts of contrastive focus and topic. One idea that has floated is that “contrast” is an information structure notion on par with “topic” and “focus” and not a subset of them (cf. Molnár 2002; Fiedler 2009). [^]
  8. In OJ, and later stages of Japanese, the verb base ending on –i, and in the case of adjectives, the bare stem, can (zero) derive nouns from verbs, but it is not used for forming nominalized clauses. (i) below is an example for the noun kazasi ‘decoration’, derived from the verb kazas- ‘decorate’. [^]
      1. (iii)
      1. Man’yōshū 820
        1. Ume=no
        2. plum=GEN
        1. pana…
        2. flower
        1. kazasi=ni
        2. decoration=DAT
        1. si-te.na.
        2. make-PFV.DES
        1. ‘Let us make hair decoration with the plum blossoms.’
  9. From what we know from the grammaticalization literature, the diachronic ordering of these three functions is most likely nominalization > relativization > complementation (cf. Heine & Kuteva 2002; Deutscher 2009), but this is a question of limited relevance for our concerns here. [^]
  10. The most valid counterargument against this classification is that the presence of a genitive subject may simply be due to the fact that these embedded clauses are nonfinite and therefore lack nominative case (E. Aldridge, personal communication). [^]
  11. Note that Russell (2006: 198) suggests that this suffix was not nominal but “stative”, but there is no research at all that would show an aspectual contrast between the adnominalizing and the finite verb forms, as stative vs. non-stative. Previous research has practically uniformly considered the additional component of –(ur)u vs. the finite –(r)u as adnominalizing and a close equivalent to Modern Japanese no(=da) and koto, and not an equivalent to –te iru. [^]
  12. The “predicative” in turn, is reconstructed by Unger (2000) as the verb stem plus a suffix *-u. [^]
  13. Cf. Huang (2012: 245) “A presupposition is a proposition whose truth is taken for granted in the utterance of a sentence. […]. E.g. the uttering of the sentence John regrets that he became temperamental presupposes that John became temperamental.” [^]
  14. Frellesvig (2010: 55) labels the presuppositional form as “exclamatory”, which is puzzling, since this suggests practically the opposite of the traditional Japanese labeling (‘given existence form’). He also does not provide arguments why he rejects the established view in domestic Japanese linguistics going back to Ishida (1939). However, in section 3.3 I provide a scenario in which Frellesvig’s label makes sense as an advanced stage in the development of this form. [^]
  15. Note that the particles included in kakari-musubi (focus concord constructions) vary somewhat from scholar to scholar, especially concerning those particles or forms that do not accompany a special predicate form. [^]
  16. Zo is in OJ still more frequently found in the form so. However, although it is generally assumed that so is the older form, we lack proof for this assumption (cf. Vovin 2009: 1186). [^]
  17. The copular function of ka is less clear in its usage outside of the focus concord constructions, e.g. clause-finally. An anonymous reviewer suggested a unified analysis in which ka accompanies a null copula after nominal predicates. [^]
  18. Eight clause-final examples of koso can be found in the Man’yōshū, but they are generally interpreted as cases of inversion or ellipsis of the following clause (cf. Kōji 1988: 751). [^]
  19. Shinzato & Serafim (2013: 285) refer primarily to “future” –(a)m-. Besides that, the “suppositive” ram- can also be often found on predicates in sentences with ka. Note, though, that since –(a)m- and ram- have consonant stems, their adnominal form is indistinguishable in writing from the finite form (cf. previous section 2.3). [^]
  20. Kōji (1988) counts 444 instances of ya vs. 379 of ka in the Man’yōshū. However, only 220 of those are non-rhetorical questions vs. 353 in the case of ka. [^]
  21. As noted in section 2.1, in the research literature “contrastive topic” is generally not considered as a sub-type of topic but rather as a sub-type of focus (cf. Gundel 1998; Gundel & Fretheim 2004; Titov 2013), or as a combination of topic and focus (Krifka & Musan 2012). Thus, if a contrastive topic marker generalizes to mark more generally focus (as claimed, for example, by Fiala) this does not mean a shift from topic to focus, but an extension from a specific type of focus to focus in general. [^]
  22. Also note that in Sinhala, which has focus concord constructions that have been taken as the closest cross-linguistic equivalents to the Japanese ones, the origins are not clefts even though the constructions developed the features of clefts (Slade 2011: 249–250; 2018: 5–8). [^]
  23. In the case of constructions with ka, 10 instances can be found of the presumptive non-inverted source construction with a verbal predicate, and no instance of the inverted construction and the sentence-final verbal predicate marked by pa (cf. Nomura 1995: 2; 2002b: 19). [^]
  24. Note, though, that, as an anonymous reviewer suggests, it is not implausible to assume a displacement of elements and deviation from basic word order in focusing constructions. [^]
  25. The Japanese construction consists of a free relative clause followed by the focused element and a copula, and is therefore structurally fairly parallel to the English wh-(pseudo-)cleft. The following is an example of a Modern Japanese cleft. [^]
      1. (i)
      1. Watasi=ga
      2. I=NOM
      1. kat.ta=no=wa
      2. buy.PST=NMZ=TOP
      1. Nissan=da.
      2. PN=COP
      1. ‘It’s a Nissan that I bought’ (lit.: ‘What I bought is a Nissan.’)
    [^] Note that the semantic differences between the different types of clefts are negligible and mainly concern pragmatic weightings of their constituents and (cf. e.g. Collins 2006). [^]
  26. Note that Ikawa (1998) proposed a movement analysis earlier, but in terms of movement to subject position (Spec of AGRP), competing with the nominative subject there. This hypothesis did not have much influence on later research. [^]
  27. Note, though, that Mitrović (2014) has made an attempt to explain the adnominal verb form by basically adopting the syntactic analysis by Aldridge (2010) but analyzing the musubi (concluding) part of the construction as a Free Relative: “The kakari component results from movement of a segment contained within the vP to [Spec, kP] […] the adnominal marker is an exponent of the (semantically nominal) Top0, hence movement of the remnant vP material […] to its specifier position results in the pronunciation of the specifier and head as an adnominally marked verbal element […]” (Mitrović 2014: 25). [^]
  28. Clause juxtaposition means that two clauses are adjacent and semantically related to each other, but there is no formal marking of this relationship. Juxtaposed clauses may be in a paratactic or in a hypotactic relationship (cf. e.g. Matthews 2014(3): 208; Cristofaro 1998; Palancar 2012). [^]
  29. Note that this is the semantic interpretation of Nomura (2002b). I cannot see independent evidence to confirm the presumptive shift in informational-structural weight. [^]
  30. Incidentally, one author, Mitrović (2014), actually analyzes the Old Japanese focus concord construction as a relative clause construction (cf. 4.2). [^]
  31. Morishige refers to kakarimusubi (cf. section 1) in general; not only those ending on a nominal predicate form. [^]
  32. Also, the development of focus concord construction in Sinhala could be similar, in this case, since the Sinhala construction apparently started out as a nominalized clause, with focusing elements being added later (cf. Slade 2018: 21–24). [^]
  33. In fact, this is the case with Yanagida (2016), who unlike Nomura in his “comment clause” hypothesis does not distinguish between different particles and their properties. [^]
  34. Fiala (2000: 334) has an interesting proposal for one more feature with respect to which si and mo differed from the focus particles participating in FPCNs. According to him, si and mo have a wider scope on information structure level. Koso has the narrowest scope, being confined to a subordinate clause. Zo, ka, ya and namo/namu have main clause scope, while si and mo have scope beyond the sentence; e.g. mo as inclusive focus (“also”) often refers to a discourse entity from a previous sentence. So in this view, since the informational scope of si and mo went beyond the sentence, they did not interact with the predicate of the sentence and its inflection. [^]


I am deeply grateful to the editor in charge (E. Aldridge) and three anonymous reviewers for their invaluable help in improving my manuscript. All remaining errors are my own.

Competing Interests

My research on language change in Japanese has been supported through grant number 16H03411 by the Japanese Society for the Promotion of Science.


Aldridge, Edith. 2009. Short wh-movement in Old Japanese. In Shoishi Iwasaki, Hajime Hoji, Patricia M. Clancy & Sung-Ock, Sohn (eds.), Japanese/Korean Linguistics 17. 549–563. Stanford: CSLI.

Aldridge, Edith. 2018. C-T Inheritance and the left periphery in Old Japanese. Glossa: A Journal of General linguistics 3/1. 1–22. DOI:  http://doi.org/10.5334/gjgl.254

Collins, Peter. 2006. It-clefts and wh-clefts: Prosody and pragmatics. Journal of Pragmatics 38. 1706–1720. DOI:  http://doi.org/10.1016/j.pragma.2005.03.015

Cristofaro, Sonia. 1998. Grammaticalization and clause linkage strategies. In Anna Giacalone, Ramat & Paul J. Hopper (eds.), The limits of grammaticalization, 59–88. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/tsl.37.03cri

Deutscher, Guy. 2009. Nominalization and the origin of subordination. In T. Givón & Masayoshi Shibatani (eds.), Syntactic complexity: Diachrony, acquisition, neuro-cognition, evolution, 199–214. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/tsl.85.08nom

Dufter, Andreas & Daniel Jacob. 2009. Introduction. In Andreas Dufter & Daniel Jacob (eds.), Focus and background in Romance languages, 1–18. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/slcs.112.02duf

Erteschik-Shir, Nomi. 2007. Information structure: The syntax–discourse interface. Oxford: Oxford University Press.

Evans, Nick. 2007. Insubordination and its uses. In Irina Nikolaeva (ed.), Finiteness. Theoretical and empirical foundations, 366–431. Oxford: Oxford University Press.

Evans, Nick & Honoré Watanabe. 2016. The dynamics of insubordination: An overview. In Nick Evans & Honoré Watanabe (eds.), Insubordination, 1–37. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/tsl.115.01eva

Fiala, Karel. 2000. Nihongo no jōhō kōzō to tōgo kōzō [Information structure and syntactic structure of Japanese]. Tokyo: Hitsuji Shobō.

Fiedler, Ines 2009. Contrastive topic marking in Gbe. In The Linguistic Society of Korea (ed.), Current issues in unity and diversity of Languages, 295–308. Seoul: The Linguistic Society of Korea.

Frellesvig, Bjaerke. 2010. A History of the Japanese language. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511778322

Frellesvig, Bjaerke, Stephen W. Horn & Yuko Yanagida. 2015. Differential object marking in Old Japanese. A corpus-based study. In Dag T. T. Haug (ed.), Historical linguistics 2013, 195–211. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/cilt.334.11fre

Funagi, Shuntarō. 1987. Kakarimusubi [Kakarimusubi]. In Akiho Yamaguchi (ed.), Kokubunpō kōza 3, 278–305. Tokyo: Meiji Shoin.

Funagi, Shuntarō. 2013. Kakarimusubi-kō [Thoughts on kakarimusubi]. Tokyo: Benseisha.

Gundel, Jeanette K. 1998. On different kinds of focus. In Peter Bosch & Rob van der Sandt (eds.), Focus: Linguistic, cognitive, and computational perspectives, 293–305. Cambridge: Cambridge University Press.

Gundel, Jeannette K. & Thorstein Fretheim. 2004. Topic and focus. In Laurence Horn & Gregory Ward (eds.), The handbook of pragmatic theory, 174–196. Blackwell.

Handō, Hideaki. 2003. Kakari joshi to kakarimusubi no honshitsu [The true nature of the kakari particles and kakarimusubi]. Tokyo: Shintensha.

Handō, Hideaki. 2014. Koso. In Nihongo Bunpō Gakkai (ed.), Nihongo bunpō jiten, 231–233. Tokyo: Taishūkan.

Harris, Alice C. & Lyle Campbell. 1995. Historical syntax in cross-linguistic perspective. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511620553

Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511613463

Heine, Bernd & Tania Kuteva. 2007. The genesis of grammar. Oxford: Oxford University Press.

Huang, Yan. 2012. The Oxford dictionary of pragmatics. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oxfordhb/9780199697960.001.0001

Ikawa, Hajime. 1998. On kakarimusubi in Old Japanese: A possibility under a perspective of generative grammar. Journal of Japanese Linguistics 16. 1–38. DOI:  http://doi.org/10.1515/jjl-1998-0103

Imuta, Tsunehisa. 1976. Namu no kakarimusubi nit tsuite [About the kakarimusubi with namu]. In Saeki Umetomo, Hakase Kiju, Kinen Kokugogaku & Ronshū Kankōkai (eds.), Kokugogaku ronshū, 339–359. Tokyo: Hyōgensha.

Ishida, Shunjō. 1939. Koso-kere no hongi (1) [The original meaning of koso-kere (1)]. Kokugo to Kokubungaku 162/2. 66–82.

Iwasaki, Shoichi. 2013. Japanese. Revised edition. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/loall.17

Jayaseelan, K.A. 2004. Question movement in some SOV languages and the theory of feature checking. Language and Linguistics 5/1. 5–27.

Kamada, Ryōji. 1979. Joshi wo ni tsuite [About the particle wo]. In Tanabe Hakase, Koki Kinen, Kokugogaku Ronshū & Henshū Iinkai (eds.), Kokugo joshi jodōshi ronshū, 179–191. Tokyo: Ōfūsha.

Katsumata, Takashi. 2007. Jōdai koso~izenkei-setsu no kinō to sono hensen nit tsuite [About the function and the change of clauses with koso~presuppositional form in Old Japanese]. Kokugo Kokubun 76/4. 1–19.

Kawabata, Yoshiaki. 1994. Kakarimusubi no keishiki [The forms of kakarimusubi]. Kokugogaku 176. 1–13.

Kikuta, Chiharu U. 2004. An optimality-theoretic alternative to the apparent wh-movement in Old Japanese. In Miriam Butt & Holloway King (eds.), Proceedings of the LFG04 Conference, 290–306. Stanford: CSLI Publications.

Kinsui, Satoshi. 2011. Tōgoron [Syntax]. In Satoshi Kinsui et al. (ed.), Bunpōshi (Shiriizu Nihongoshi 3), 77–166. Tokyo: Iwanami Shoten.

Kizu, Mika. 2005. Cleft constructions in Japanese syntax. New York: Palgrave Macmillan. DOI:  http://doi.org/10.1057/9780230503618

Kōji, Kazuteru. 1988. Man’yōshū joshi no kenkyū [A study of particles in the Man’yōshū]. Tokyo: Kasama Shoin.

Kondō, Yasuhiro. 1987. Kobun ni okeru gimon hyōgen [Interrogative expressions in older Japanese]. In Akiho Yamaguchi (ed.), Kokubunpō kōza 3. 258–277.

Kondō, Yasuhiro. 2000. Nihongo kijutsu bunpō no riron [The theory of Japanese descriptive grammar]. Tokyo: Hitsuji Shobo.

Kondō, Yasuhiro. 2014. Ka. In Nihongo Bunpō Gakkai (ed.), Nihongo bunpō jiten, 70–71. Tokyo: Taishūkan.

Kōno, Tomoaki. 2010. Zo no kakarimusubi no hasshō ni tsuite [About the emergence of kakarimusubi with zo]. Kokugo Kokubun 79/12. 37–54.

Krifka, Manfred & Renate Musan. 2012. Information structure: Overview and linguistic issues. In Manfred Krifka & Renate Musan (eds.), The expression of information structure, 1–43. Berlin: Mouton de Gruyter. DOI:  http://doi.org/10.1515/9783110261608

Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge, MA: MIT Press.

Luraghi, Silvia & Claudia Parodi. 2008. Key terms in syntax and syntactic theory. London: Continuum.

Matthews, Peter. 2014. The concise Oxford dictionary of linguistics. Third edition; Oxford: Oxford University Press.

Mithun, Marianne. 1988. The grammaticization of coordination. In John Haiman & Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 331–359. Amsterdam: Benjamins.

Mitrović, Moreno. 2014. Deriving and interpreting ‘ka(karimusubi)’ in premodern Japanese. Acta Lingustica Asiatica 4/3. 9–28. DOI:  http://doi.org/10.4312/ala.4.3.9-28

Molnár, Valeria. 2002. Contrast – from a contrastive perspective. In Hilde Hasselgard, Stig Johansson, Bergljot Behrens & Cathrine Fabricius-Hansen (eds.), Information structure in a cross-linguistic perspective, 147–161. Amsterdam/New York: Rodopi.

Morishige, Satoshi. 1959. Nihon bunpō tsūron [Outline of Japanese grammar]. Tokyo: Kazama Shobō.

Morishige, Satoshi. 1971. Nihon bunpō no shomondai [Problems of Japanese grammar]. Tokyo: Kasama Shoin.

Narrog, Heiko. 2016. Insubordination in Japanese diachronically. In Nick Evans & Honoré Watanabe (eds.), Insubordination, 247–281. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/tsl.115.10nar

NKD5: Nihon Kokugo Daijiten Dainihan Henshū Iinkai (eds.). 2001. Nihon kokugo daijiten [Great Dictionary of the Japanese National Language], vol. 5. Tokyo: Shogakkan.

Nomura, Takashi. 1993a. Jōdaigo no no to ga ni tsuite (1) [About no and ga in Old Japanese (1)]. Kokugo-Kokubun 62/2. 1–17.

Nomura, Takashi. 1993b. Jōdaigo no no to ga ni tsuite (2) [About no and ga in Old Japanese (2)]. Kokugo-Kokubun 62/3. 30–49.

Nomura, Takashi. 1995. Ka ni yoru kakarimusubi shiron [A tentative discussion of kakarimusubi with ka]. Kokugo Kokubun 64/9. 1–27.

Nomura, Takashi. 2001. Ka. In Akiho Yamaguchi & Morihide Akimoto (eds.), Nihongo bunpō daijiten, 115–117. Tokyo: Meiji Shoin.

Nomura, Takashi. 2002a. Kakarimusubi no kōbun: Jōdai kara chūko e [Kakarimusubi sentences: From Old Japanese to Late OLd Japanese]. Handout at the Twenty-Sixth Meeting of the Kansai Linguistic Society. Ōsaka: Momoyama Gakuin Daigaku.

Nomura, Takashi. 2002b. Rentaikei ni yoru kakarimusubi no tenkai [The development of kakarimusubi with the adnominal form]. In Hiroto Ueda (ed.), Nihongogaku to gengokyōiku, 11–37. Tokyo: Tōkyō Daigaku Shuppankai.

Nomura, Takashi. 2005. Chūko kakarimusubi no hen’yō [The change of kakarimusubi in Late Old Japanese]. Kokugo to Kokubungaku 2005/11. 36–46.

Nomura, Takashi. 2011. Hanashikotoba no nihonshi [Japanese history of the spoken language]. Tokyo: Yoshikawa Kōbunkan.

Oda, Masaru. 2003. Musubi no nagare kara mita kakarimusubi no hyōgen kachi [The expressive value of kakarimusubi from the point of view of kakarimusubi in non-finite clauses]. Kokugo Kenkyū 66. 94–107.

Okamura, Kazue. 1995. Koso. Kakari-joshi [Koso, kakari-particle]. In Akira Matsumura (ed.), Kotengo, gendaigo joshi jodōshi shōsetsu [Ancient Japanese and Modern Japanese handbook of particles and auxiliaries], 570–576. Tokyo: Gakutōsha.

Ōno, Susumu. 1956a. Nihon koten bunpō (2). Sono 1, kakarimusubi (2) [Japanese grammar 2/1. Kakarimusubi (2)]. Kokubungaku Kaishaku to Kanshō 2/1956. 76–84.

Ōno, Susumu. 1956b. Nihon koten bunpō (3). Sono 1, kakarimusubi (3) [Japanese grammar 3/1. Kakarimusubi (3)]. Kokubungaku Kaishaku to Kanshō 4/1956. 147–154.

Ōno, Susumu. 1956c. Nihon koten bunpō (4). Sono 1, koso no kakarimusubi (4) [Japanese grammar 4/1. Kakarimusubi (4)]. Kokubungaku Kaishaku to Kanshō 5/1956. 90–96.

Ōno, Susumu. 1956d. Nihon koten bunpō (5). Sono 1, koso no kakarimusubi (5) [Japanese grammar 5/1. Kakarimusubi (5)]. Kokubungaku Kaishaku to Kanshō 6/1956. 87–95.

Ōno, Susumu. 1964. Kakarimusubi no kigen wa donna koto ka [What is the origin of kakarimusubi?]. Kokubungaku Kaishaku to Kanshō 10/1964. 96–102.

Ōno, Susumu. 1993. Kakarimusubi-no kenkyuu [Research on kakarimusubi]. Tokyo: Iwanami.

Ono, Tsuyoshi. 2006. An emotively motivated post-predicate constituent order in a ‘strict predicate final’ language. In Satoko Suzuki (ed.), Emotive communication in Japanese, 139–153. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/pbns.151.07ono

Palancar, Enrique L. 2012. Clausal juxtaposition and subordination: New light from Oromi. In Volker Gast & Holger Diessel (eds.), Clause linkage in cross-linguistic perspective: Data-driven approaches to cross-clausal syntax, 37–76. Berlin: Mouton de Gruyter.

Quinn, Charles J. 1997. On the origin of the sentence particles ka and zo. In Ho-min Sohn & John Haig (eds.), Japanese/Korean linguistics 6. 61–89. Stanford: CSLI.

Radford, Andrew et al. 2009. Linguistics: An introduction. Second Edition; Cambridge: Cambridge University Press.

Reinhart, Tanya. 1982. Pragmatics and linguistics: An analysis of sentence topics. Bloomington, IA: Indiana University Linguistics Club.

Robbeets, Martine. 2015. Diachrony of verbal morphology: Japanese and the Transeurasian languages. Berlin: Mouton De Gruyter. DOI:  http://doi.org/10.1515/9783110399943

Rooth, Mats. 1996. Focus. In Sharon Lappin (ed.), The handbook of contemporary semantic theory, 271–297. Oxford: Blackwell. DOI:  http://doi.org/10.1111/b.9780631207498.1997.00013.x

Russell, Kerri L. 2006. A reconstruction and morphophonemic analysis of proto-Japonic verbal morphology. Honolulu, HI: University of Hawai’i Ph.D. Dissertation.

Sakakura, Atsuyoshi. 1993. Nihongo hyōgen no nagare [The flow of expression in Japanese]. Tokyo: Iwanami Shoten.

Schaffar, Wolfram 2002. Kakari Musubi, noda-constructions and how grammaticalization theory meets formal syntax. In Patricia M. Clancy (ed.), Japanese/Korean linguistics 10. 320–333. Stanford: CSLI

Schaffar, Wolfram. 2003. Die Informationsstruktur der japanischen Sprache. Eine Analyse von Nominalisierungen, Fokuspartikeln und Kakari-Musubi-Konstruktionen im Modernen Japanischen, Klassischen Japanischen und in der Sprache von Ryukyu. Münster: LIT Verlag.

Serafim, Leon A. & Rumiko Shinzato. 2005. On the Old Japanese kakari (focus) particle koso: Its origin and structure. Gengo Kenkyu 127. 1–49.

Serafim, Leon A. & Rumiko Shinzato. 2011. On the origins of the Old Japanese kakari particles, ka, zo, and koso, and their Okinawan counterparts: An iconicity-based hypothesis. In Ho-min Sohn, Haruko Minegishi Cook, William O’Grady, Leon Serafim & Sang Yee Cheon (eds.), Japanese/Korean linguistics 19. 449–464. Stanford: CSLI.

Shinzato, Rumiko & Leon A. Serafim. 2013. Synchrony and diachrony of Okinawan kakari musubi in comparative perspective with premodern Japanese. Leiden: Brill. DOI:  http://doi.org/10.1163/9789004219038

Slade, Benjamin. 2018. History of focus-concord constructions and focus-associated particles in Sinhala, with comparison to Dravidian and Japanese. Glossa: A Journal of General Linguistics 3/1. 2. 1–28. DOI:  http://doi.org/10.5334/gjgl.241

Slade, Benjamin Martin. 2011. Formal and philological inquiries into the nature of interrogatives, indefinites, disjunction, and focus in Sinhala and other languages. Urbana, IL: University of Illinois at Urbana-Champaign PhD dissertation.

Steedman, Mark. 2000. Information structure and the syntax-phonology interface. Linguistic Inquiry 31/4. 649–689. DOI:  http://doi.org/10.1162/002438900554505

Suda, Junichi. 2010. Jōdaigo wo no kaku gigi [Doubts about the case feature of wo in Old Japanese]. In Junichi Suda & Junya Niida (eds.), Nihongo keitai no shomondai, 325–335. Tokyo: Hitsuji Shobo.

Tani, Chinari. 1889. Shi no kumitate. gengo kōzōshiki chūkai [The combination of words – comments on language structures]. Daihasshū Gakkai. [Note that this document has not been available to me. It is cited in Ōno (1956a)].

Titov, Elena. 2013. Do contrastive topics exist? Journal of Linguistics 49. 413–454. DOI:  http://doi.org/10.1017/S0022226712000370

Tonoike, Shigeo. 2002. Jōdai Nihongo ni sahō idō wa atta ka? [Was there left movement in Old Japanese?]. Gekkan Gengo 31/3. 86–91.

Umbach, Carla. 2001. Contrast and contrastive topic. In Proceedings of the ESSLLI’01 Workshop on Information Structure, Discourse Structure and Discourse Semantics. http://citeseerx.ist.psu.edu/viewdoc/summary?doi=

Unger, Marshall. 2000. Reconciling comparative and internal reconstruction: The case of Old Japanese /ti, ri, ni/. Language 76/3. 655–683. DOI:  http://doi.org/10.2307/417138

Uwano, Zendo. 2006. Nihongo akusento no saiken [On the reconstruction of Japanese accents]. Gengo Kenky 130. 1–42.

Van Hoof, Hanneke. 2003. The rise in the rise–fall contour: Does it evoke a contrastive topic or a contrastive focus? Linguistics 41/3. 515–563. DOI:  http://doi.org/10.1515/ling.2003.018

Van Valin, Robert D. Jr. 2001. An introduction to syntax. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9781139164320

Vovin, Alexander Y. 2009. A descriptive and comparative grammar of Western Old Japanese. Part 2. Folkestone: Global Oriental.

Watanabe, Akira. 2001. Kodai nihongo o miru [Looking at Ancient Japanese]. Gekkan Gengo 30/7. 90–95.

Watanabe, Akira. 2002. Loss of overt wh-movement in Old Japanese. In David W. Lightfoot (ed.), Syntactic effects of morphological change, 179–195. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/acprof:oso/9780199250691.003.0010

Watanabe, Akira. 2007. Topic-focus articulation in Old Japanese: So/Zo and koso. MIT Working Papers in Linguistics 54. 121–137.

Whitman, John. 1997. Kakarimusubi from a comparative perspective. In Hajime Hoji (ed.), Japanese/Korean linguistics 6. 161–178.

Wrona, Janick. 2008. The Old Japanese complement system: A synchronic and diachronic study. London: Global Oriental. DOI:  http://doi.org/10.1163/9789004213180

Wrona, Janick. 2010. Non-movement and kakari-musubi in Old Japanese. Studies in Language Sciences 9. 177–188. Tokyo: Kuroshio.

Wrona, Janick. 2011. A case of non-derived stand-alone nominalization. Evidence from Japanese. In Yap, Foong-Ha et al. (eds) Nominalization in Asian languages. Diachronic and typological perspectives, 423–443. Amsterdam: Benjamins. DOI:  http://doi.org/10.1075/tsl.96.15wro

Yamada, Akihiro. 2010. Kakujoshi ga no tsūjiteki kenkyū [Diachronic research on the case particle ga]. Tokyo: Hitsuji Shobō.

Yamada, Kenzō. 2004. Kakarimusubi saikō [Rethinking kakarimusubi]. Kokugo Kokubun 73/11. 15–33.

Yamada, Yoshio. 1908. Nihon Bunpōron [Japanese grammar theory]. Tokyo: Hōbunkan

Yamaguchi, Yoshinori. 1986. Kodai nihongo bunpō no seiritsu no kenkyū [Study of the coming into being of Ancient Japanese]. Tokyo: Yūseidō.

Yamauchi, Yōichirō. 2003. Katsuyō to katsuyōkei no tsūjiteki kenkyū [Diachronic research on inflection and inflectional forms]. Tokyo: Seibundō.

Yanagida, Seiji. 1985. Muromachi jidai no kokugo [The national language of the Muromachi period]. Tokyo: Tōkyōdō Shuppan.

Yanagida, Seiji. 2016. Nihongo no rekishi 6. Shukaku joshi “ga” no sennenki [The history of Japanese 6. 1000 years of the nominative particle “ga”]. Tokyo: Musashino Shoin.

Yanagida, Yuko. 2006. Word Order and Clause Structure in Early Old Japanese. Journal of East Asian Linguistics 15. 37–67. DOI:  http://doi.org/10.1007/s10831-005-2165-2

Yanagida, Yuko and John Whitman. 2009. Alignment and word order in Old Japanese. Journal of East Asian Linguistics 18. 101–144. DOI:  http://doi.org/10.1007/s10831-009-9043-2

Text sources

Man’yōshū. Edited by Kojima, Noriyuki, Masatoshi Kinoshita & Akihiro Satake. 4 vols, 1973-1975 (Nihon Koten Bungaku Zenshū 1-4). Tokyo: Shogakukan.

Yamato monogatari. Edited by Shōji Takahashi. 1994 (Shinpen Nihon Koten Bungaku 12). Tokyo: Shogakukan.