The Atypicality of Verb-Final Clauses in Japanese Conversation: Toward a Speaker-Centered Characterization of Japanese Clausal Syntax ^†

Abstract

The so-called canonical clause, consisting of case-marked NPs and a final finite verb, has played a central role in discussions of Japanese for the past several decades. The current study explores the nature of such clauses in everyday Japanese conversation. Everyday conversation is considered the most fundamental form of language, and large-scale corpora of Japanese everyday conversation have only become available in recent years, enabling projects like ours. One key finding is that clauses ending with a finite verb are rare, challenging the centrality of the canonical clause in Japanese grammar. Instead, we observe that the verb is usually followed by additional elements that convey pragmatic information. This observation suggests that the canonical clause for Japanese speakers should also include these pragmatic elements. We have observed further that the relatively uncommon examples that do end with a finite verb often involve five frequent semantically light verbs. A preliminary study of one of these verbs, chigau ‘to differ’, reveals that, typically without overt NPs, it functions more like a particle than the verb of a clause. This further calls into question the idea that the canonical clause in Japanese ends with a finite verb.

1. Introduction

We study the Japanese clause in naturally occurring conversation. The current effort is part of our larger project of capturing grammar in conversation (e.g., Ono & Thompson, 1995). In this paper, based on a quantitative investigation, we begin to explore a new understanding of the clause for speakers of Japanese. As conversation is the most fundamental form of language (see Schegloff, 1996, for example), studying conversation is an essential step if your goal is to understand human language in general.1 In particular, we find it critical to describe the nature of the ‘clause’ in conversation as it has commonly been considered a fundamental syntactic unit in the study of language (See Laury et al., 2021). Unless that step is taken, generalizations assuming such a unit are provisional at best. That is, the Japanese clause can be discussed in the context of language universals and linguistic typology only after its structure in conversation is appropriately described. We find that many psycholinguistic, typological, functional, and universal investigations, as well as applied and pedagogical projects involving Japanese, have assumed and continue to assume the clause as the basic unit of the language. Such investigations and projects are problematic because the nature of the Japanese clause has not been established based on the examination of data.

We fully support the recent trend of examining multiple modalities to understand language and interaction; however, exploring these modalities requires a thorough examination of each one. In this investigation, we focus on grammar, particularly the clause, arguably the most central syntactic unit in the study of human language. We hope that the results of our research will produce a better understanding of Japanese syntax, serving as a baseline for all research involving Japanese.

We begin by examining examples of Japanese clauses found in both the traditional and more recent literature.2 We observe that similar examples, specifically case-marked NPs ending with the verb, are present across different theoretical perspectives. However, the new conversation corpus (Koiso et al., 2022) exhibits examples that are very different. Most notably, as Maynard reported in 1993, the majority of clauses do not end with the verb. Instead, verbs are typically followed by elements such as honorific forms and interactional particles, which reflect the relationships between speakers. Given their frequency in conversation, these utterances are considered typical clauses for Japanese speakers, and we propose that they represent the fundamental structure of the Japanese clause.3 Furthermore, the less common type of clauses that end with the verb—characteristic of literature—appear to be more like short fixed expressions rather than syntactic units, such as clauses. These observations challenge the traditional view of Japanese as a verb-final language.

2. Typical Clauses in the Literature

The clause is one of the most, if not the most, celebrated syntactic units in the study of language. It is often defined as

“grammatical units that consist of predicates and the phrases that

‘go with’ them.”

(Couper-Kuhlen & Selting, 2018, p. 361; originally in Thompson & Couper-Kuhlen, 2005, p. 484)

Such definitions are most commonly accompanied by examples such as the following:

		S	V	O
(1)	Vivian:	we	need	some more wine	over he:re. (Thompson, 2021, p. 17)

This is a typical example in that the two NPs, the subject and direct object, along with the verb, represent a ‘basic’ transitive clause in English, exhibiting its SVO word order.4

This manner of characterizing the clause appears to have been adopted in the study of Japanese. For example, the Japanese clause has been described in general linguistics in North America using examples such as the following:

	S		O	V
(2)	Taro:-ga		ringo-o	tabeta
	Taro		apple	ate
	‘Taro ate an apple.’					(Bergmann et al., 2007, p. 229)

Again, this is a transitive clause with two NPs with the verb, which might look similar to the English example in (1). Obviously, the word order in English is different and the example exhibits its well-known SOV order. The two NPs Taro ‘Taro’ and ringo ‘apple’ are followed by particles ga and o. These particles are generally understood to mark the grammatical relations of the subject and the direct object, thus marking the eater and the eatee of the eating action in the example (Shibatani, 1990). Note also that the verb tabeta ‘ate’ is tensed (i.e., finite form) for the past.

Similar examples are found in the literature focused on Japanese. Examples (3) and (4) are both taken from introductory (Japanese) linguistics textbooks, written in English and Japanese. They are both found at the beginning of the syntax section; they illustrate the ‘basic’ or, alternatively, ‘canonical’ clause of the language:

		S				O	V
(3)	a.	Taroo-ga	Kinoo		ookii  hanbaagaa-o          itutu		tabeta.
		Taro-Nom		yesterday	big      hamburger-Acc     five		ate
		“Taro ate five big hamburgers yesterday.” (Tsujimura, 2014, Chapter 4 Syntax)

S

(4)

masao wa

ikkagetsu

mae

ni ie

kara

gakkoo

made

Masao TOP

one.month

before

at home

from

school

to

V

ichiroo

to

jitensha

de

kyoosoo

shita

Ichiro

with

bicycle

by

race

did

‘Masao raced from home to school by bicycle with Ichiro a month ago’

(Ueyama, 1991, 3 How to form a sentence 3.1 The ending is the key, p. 62)

Example (3) is similar to (2) and is a transitive clause with the two NPs ‘Taro’ and ‘five big hamburgers’. These NPs are again marked with ga and o, respectively. Example (4) is an intransitive clause with the argument NP masao ‘Masao’ and several oblique NPs.5 The NP masao ‘Masao’ is marked with the so-called topic particle wa (Hinds et al., 1987), which is known to mark the discourse-level function of the NP. Both (3) and (4) again end with the tensed verb (i.e., finite form) and are understood to represent a ‘basic’ Japanese clause. In fact, the author of the textbook that (4) is taken from highlights the finality of the verb in the Japanese clause.6 Please note that these examples are all constructed by the authors and sound a little unnatural to our ears. They sound written, stiff, or distinctly like example sentences used in linguistics. Some might suggest they represent written language. However, they are apparently intended to represent spoken language. The author of the textbook that (3) is taken from, for instance, states, “…the primary goal of this book is to examine spoken Japanese…” (Tsujimura, 2014, preface, xi).

Examples slightly different from those above are used to describe the clause in usage-based approaches to Japanese, as in the following:

				V
(5)	…	sore ga	kanojo	tsurete kuru.
		that SUB	girlfriend	take	come
‘That person will bring (his) girlfriend.’ (Ono & Suzuki, 1992, p. 430)

(6)	[TYC 21]
								V
	Kanji:	mae	nakamura	ga	na::	hige		nobashiteta.
		before	Nakamura	SP	FP	mustache	was.growing
		“Nakamura was growing a mustache before.”
		(Hayashi, 2003, 2.2 Grammar of conversational Japanese, p. 17)

The above examples are both found in the discussion of the basic clause structure of Japanese and taken from conversation data. As you may notice, unlike what we saw in the constructed examples in (2)–(3), what is considered the direct object, kanojo ‘girlfriend’ and hige ‘mustache’ in (5) and (6), respectively, are not marked by the case marker, a well-known phenomenon for spoken Japanese (e.g., Ono & Fujii, 2000), demonstrating that case markers are, in fact, not used in distinguishing the grammatical relations.7 Other than that, however, these examples look similar to those in (2)–(4) in that they all end with a tensed verb (i.e., finite form). Thus, it seems fair to say that, regardless of the theoretical and methodological orientations, clauses with full NPs ending with a tensed verb have generally been understood to represent the basic clause type, often called the canonical clause, in Japanese.

The current contribution continues the spirit of works by usage-based linguists that have been uncovering the mechanisms of Japanese grammar based on the study of conversation. Relatively recent works in this series of studies explored areas of grammar such as the noun phrase (Ono & Thompson, 2020), clause (Ono et al., 2020; Laury et al., 2021), negation (Ono & Thompson, 2017), pseudo-cleft (Mori, 2014; Kaneyasu, 2019; Ono & Suzuki, 2023), and clause combining (Laury & Ono, 2014). The current effort owes much to an earlier study by Ono and Thompson (1995) on syntax based on conversation data.

Here, we focus on a specific feature of the so-called canonical clause by asking whether clauses ending with the verb in finite form are, in fact, typical of what Japanese speakers produce.

3. What Conversation Tells Us About the Japanese ‘Clause’

3.1. Data

We examine a newly published corpus, the Corpus of Everyday Japanese Conversation (Koiso et al., 2022, hereafter CEJC). CEJC is the first sizable corpus involving the audio, video, and transcripts of 577 Japanese conversations representing everyday situations recorded in the greater Tokyo area. It involves 862 speakers across genders and generations, 200 h of talk, 2.4 million words, and 830,023 verbs. It is relatively large, equipped with a variety of search and sampling options, making it possible to examine spoken Japanese used in context on a meaningful scale for the first time in the study of Japanese.

3.2. Procedures and Quantitative Findings

To gain a general understanding of the role of the canonical Japanese clause in conversation, we first explore its quantitative characteristics. We focus on verbal predicates to see if they end the independent clause in finite form, a key feature of the canonical clause, illustrated in examples (2)–(6) above. Based on the findings of the quantitative investigation, we then examine individual examples in context to gain an understanding of clauses in conversation data.

CEJC marks the end of what they call the hatsuwa tan’i ‘utterance unit’ (see Den et al., 2010), equivalent to the independent clause, whose ending is defined grammatically, prosodically, and/or pragmatically.8 As a first step, we examine the frequency of the hatsuwa tan’i ‘utterance unit’, roughly the independent clause, ending with a finite form of verbs in the entire corpus.9 

Table 1. Ending of hatsuwa tan’i (‘utterance units’).

	N	%
Finite verb ending	24,019	4.6
Other ending	501,115	95.4
Total	525,134	100

presents the results.

The table shows there is a total of 255,134 hatsuwa tan’i (‘utterance units’) in CEJC, but ending hatsuwa tan’i (‘utterance units’) with a finite verb form in conversation is rare (4.6%). Other types of endings are much more common. This finding is interesting, though it is not a direct measure of how independent clauses end in conversation, as the hatsuwa tan’i ‘utterance unit’ can be but may not always be independent clauses.

There is no easy way to search predicates in CEJC; it is thus not possible to obtain the frequency of verbal predicates ending in finite form in the corpus. To overcome this difficulty, we randomly selected 500 hatsuwa tan’i ‘utterance units’.10 We then manually identified examples with verbal predicates, resulting in a total of 397 independent clauses that have a verbal predicate. The rest (103 cases) are those involving adjectival predicates, nominal predicates, etc. We then determined the frequency of how those 397 clauses end, specifically whether they end with a finite verb or not.

Table 2. Ending of independent clauses with verbal predicates.

	N	%
Finite verb ending	92	23.2
Other endings	305	76.8
Total	397	100

presents the results.

The table shows that ending the independent clause with a finite verb form is a minority pattern (23.2%).11 That is, one defining characteristic of the canonical clause in the literature is not typical in conversation, the most fundamental form of language, or is not commonly employed by Japanese speakers. This is not an isolated finding. An early study (Maynard, 1993, p. 120) and a more recent study (Koda, 2015, p. 125) showed similar figures. Our study replicates their findings with naturally occurring conversation. The current finding further questions the idea that examples like (2)–(6) represent the basic Japanese clause because they are not what speakers commonly produce in everyday conversation.

3.3. Other Endings as the New Canonical Ending

Figures in Table 2 lead us to ask what the majority pattern (76.8%) of verbal predicates in independent clauses is like. The following are typical examples, where we find more materials following the predicate:

				V
(7)	kinoo	mo	satsumaimo	tabeta	yone
	yesterday	also	sweet.potato	ate	FP/right
	‘(We) had sweet potatoes yesterday, right?’					(T014_014a_3:57)

							V
(8)	fuyuusoo	mitaina	hitotachi	no	naka	de	fueteiku	no	kana	mitaina
	wealthy:class	like	people	of	inside	in	increase:go	NOM	FP/wonder	FP/like
	‘(It) is like (health consciousness) might increase among people (who) are like (the) wealthy class’
	(T001_003_54:18)

In (7), the verb tabeta ‘ate’ is followed by the final particle yone, which roughly means ‘right’. In (8), the verb fueteiku ‘increase’ is followed by a combination of the nominalizer no and the final particles kana ‘(I) wonder’ and mitaina ‘like’. Note that these additional elements are produced immediately after the verbal predicate without a pause or prosodic break, transcribed and checked by multiple transcribers. Functionally, they provide epistemic and interactional, alternatively subjective and intersubjective, information, as has been widely discussed (e.g., Cook, 1992; Tanaka, 2000; Morita, 2005; Fujii, 2006; Hayano, 2011; Matsumoto, 2018; Yokomori, 2023). In this paper, for lack of a better term, we use the cover term ‘pragmatic elements’ (PEs) for the set of materials that end the clause. Our point is that, unlike the canonical clause in the literature, illustrated in examples (2)–(6) above, a clear majority (76.8%) of independent clauses with verbal predicates in conversation do not end with the verb in finite form; PEs follow it. This finding is not surprising to those who are familiar with spoken Japanese. Numerous studies have focused on the use of individual PEs in conversation (e.g., Cook, 1992; Tanaka, 2000; Morita, 2005; Fujii, 2006; Hayano, 2011; Matsumoto, 2018; Yokomori, 2023). The relevance of PEs to the syntax of Japanese, however, has not been highlighted.

We earlier pointed out that (2)–(4) above sound written, stiff, or distinctly like example sentences used in linguistics. Compared to those examples, (7) and (8) sound much more natural, which is not surprising; they are what Japanese speakers produce in conversation. The English translation of the latter set of examples includes several elements in parentheses, which mark those referents that might be intended but are not expressed overtly in Japanese, a phenomenon widely discussed in the literature under the label ‘zero anaphora’ (Clancy, 1980; Hinds, 1982; Fry, 2003; Ono & Suzuki, 2020). The lack of overt expressions and the resulting abundant use of parentheses might look exotic to those who are used to examining languages like English and German, but let us underscore that ‘zero anaphora’ is not only observed in exceptional circumstances. Instead, the lack of overt expression is more or less the norm in conversational Japanese, with occasional overt expressions of some referents (as in (7)), and that seems to be part of the reason (7) and (8) sound more natural than (2)–(4).12

Based on the above findings that Japanese speakers most typically do not end the independent clause with the verb in finite form but continue with PEs, we think a configuration like the following is a rough but more realistic representation of the basic Japanese clause:

(9)	Canonical Japanese clause in conversation
	(arguments)	verb	PEs

In this proposal, we suggest that PEs are built into the end of the clause with an assigned slot; they are part of the grammar of the basic Japanese clause, the most frequent type in conversation. Not having the PEs is a clear minority; the arguments are in parentheses because they are often not expressed. Some of the conversation-based studies take a view similar to ours. Morita (2005), for instance, examined ‘particles’, examples of PEs, in conversation and said ‘particles often function as mandatory elements in conversational grammar’ (29). Further, we find almost all utterances in Takagi’s (2001, p. 9) L1 acquisition data to end with PEs, which shows that children grow up learning and using Japanese with PEs as part of utterances. The configuration given in (9) thus captures the reality of the Japanese clause both in adult and child language.13

The above representation, in fact, questions the characterization of Japanese as a verb-final language. This discrepancy appears to stem from the methods used in early research that established the verb-finality of Japanese. In such methods, examples of languages like English served as the base to construct examples of other languages. Since languages like English do not grammatically encode (inter)subjective concerns, the translated Japanese examples ended up not having those PEs. The typical lack of context for constructed examples (often single sentences) undoubtedly promoted a disregard for (inter)subjective concerns even further. Not surprisingly, however, different languages grammaticize differently, a fact that early researchers did not recognize, which we highlight here to promote a better understanding and representation of the Japanese clause.14

Some people who heard our claim above questioned if PEs are actually part of the verb or VP. The data do not support that. For example, in (7), the final particle yone ‘right’ is not an inflectional ending of the verb tabeta ‘ate’, as evidenced in its separate representation from the verb. In fact, yone does not appear to function just with the verb but with the whole utterance as ‘(We) had sweet potatoes yesterday, right?’. Similarly, in (8), the PEs involving no kana mitaina ‘(It) is like’ is not an inflectional ending of the verb fueteiku ‘increase’. Again, it does not function just with the verb but with the whole utterance as ‘(It) is like (health consciousness) might increase among people (who) are like (the) wealthy class’. PEs are thus separate from the verb and are required to be represented independently from other elements in the clause. Treating the PEs as part of the verb complex only helps fit Japanese into the structure of some dominant research languages, like English, which do not encode pragmatic information as part of the clause in the way Japanese does. PEs as independent elements in the clause better describe Japanese and provide an excellent data point for our efforts to capture the structure and use of human language in general.

Finally, there is extensive research on individual elements of PEs in conversation, although it typically relies on small datasets. These studies highlight the situated and interactional nature of elements in PEs and identify subtle yet essential differences among them (e.g., Cook, 1992; Tanaka, 2000; Morita, 2005; Fujii, 2006; Hayano, 2011; Matsumoto, 2018; Yokomori, 2023). We hope that this line of research will continue, examining newly available corpora, particularly by exploring the use of less-studied elements and focusing on the combinations similar to those observed in (8). In addition, past studies on the grammatical aspects of PEs have mostly used constructed examples. We hope to see future studies examining naturally occurring conversations to focus on this aspect of PEs. Many unanswered questions about PEs require in-depth work, such as the following: What elements are included in PEs? How are the elements ordered and structured? Are there frequent combinations and patterns? What do elements in PEs do individually and collectively? Is there anything general we can say about PEs based on the study of conversation? We aim to participate in this discussion to achieve a broader goal: determining the role PEs play in the grammar of Japanese.

3.4. Clauses Ending with the Finite Verb (Atypical Clauses in Conversation)

We should remember that Table 2 includes some examples that end with the finite verb (23.2% of the examples), which is the ending of the canonical clause in the literature. It is a minority type, but that is what speakers produce. It seems reasonable to hypothesize that canonical clauses similar to (2)–(6) account for a certain amount of conversation. In examining those examples ending with the finite verb, atypical in conversation, we noticed an interesting quantitative skewing; we found that a large portion of the 92 examples involve a specific set of verbs, as shown below.

In

Table 3. Verbs in finite form.

iu	‘to say’	10
omou	‘to think’	10
chigau	‘to differ’	7		43%
wakaru	‘to understand’	7
suru/yaru	‘to do’	6
other verbs		52		57%
Total		92		100%

, 5 verbs—iu ‘to say’; omou ‘to think’; chigau ‘to differ’; wakaru ‘to understand’; and suru/yaru ‘to do’—account for 40 out of 92 (43%) verbs in finite form ending the clause, close to half of this type. Several random samples yielded a similar set of verbs appearing in the same environment.15 One may point out that these are all frequent verbs regardless, and that their high frequency in ending the clause is simply because of their general high frequency. Interestingly, however, we found these same verbs account for only about 10% of the 830,023 verb tokens in the entire CEJC. So their high frequency does not seem to account for why they appear in a finite form clause finally.

Our preliminary examination of these verbs in finite form ending the clause suggests they are not clauses but are more like particles dedicated to particular discourse functions.16 That is, though those verbs are in finite form and thus classified as ending the clause for our study, they might not be clauses after all. In the remainder of this paper, we focus on one of the five frequent verbs, chigau ‘to differ’, in the table, as our first step toward fully describing the nature of the clause-final verbs. Please note that this is a preliminary study; all five verbs, including chigau, in their finite form ending the clause require thorough investigation.

4. Focus on chigau (‘to differ’)

Interestingly, all seven examples of chigau ‘to differ’ in Table 3 are in the present tense form, as shown below:

(10)	Do you have a date? (T009_006_14:50)
	M:	deeto	ka:?
		date	Q
		‘Is (it a) date?/Do (you have a) date?’
	T:	n chigau
		differ
		‘No.’ ‘?(It) differs.’

To M’s question ‘Is (it a) date?’, T responds by saying n chigau ‘No’, where the initial n appears to mark that he heard the question.17 Functionally, the verb chigau does not quite give the meaning ‘(It) differs’, which is its lexical meaning.18 Instead, it gives a negative response to the question, reflected in our translation ‘No’. The utterance ends in the present form of the verb, but it does not look at all like the canonical clause discussed in the literature. It is one word; no arguments occur with it. Thus, syntax probably does not play a role in its use. Instead, it seems reasonable to suggest that chigau functions as a response particle ‘no’. When we examined 60 randomly sampled chigau, the present form ending the clause, we found 53 of them (88%) to be without arguments, further suggesting that most uses of clause-ending chigau are not syntactic.

The following example shows the use of chigau as a particle even more clearly, where M, a teacher, first introduces a particular genre of music called fooku songu ‘folk song’, popular in Japan in the 1960s and 1970s:

(11)	folk song vs. folk dance (T013_013_16:44)
	M:	fooku	songu
		folk	song
		‘Folk song’
	T:	fooku	dansu	toka	no	kooyuu	yatsu	desu	ka
		folk	dance	like	GEN	this.type	one	COP	Q
		‘Is (it) this type like folk dance?’
	M:	chiga chau chau
		‘no no no’

T, M’s student, does not recognize the genre and asks if it is a type of folk dance, a cover term for dances considered as part of the tradition or custom of particular people. T immediately denies it as chiga chau chau ‘no no no’ where chigau is repeated three times within the same intonation contour, all reduced from chigau.19 The repetition, the reduction of the form, and the non-overt expression of argument NPs all suggest that chigau has lost some of its verbiness and become more like a particle. To be more precise, the repetition suggests that chigau is losing its verb status; it is repeated, though only required once in the clause. The reduction indicates that it is losing its lexical status. Finally, as suggested earlier, the lack of argument NPs indicates the lack of syntactic behavior. That is, we find chigau ‘to differ’, the present tense form, occurring at the end of the clause, behaving like a particle. Again, we found that among the 60 randomly selected instances of chigau in the present tense, discussed above, 14 of them are a part of repetitions (23%). Further, 19 of them (32%) are in reduced forms as chiga, chau, cha, etc. Thus, these verb-final examples (atypical clauses in conversation) are not at all like the canonical clause discussed in the literature; they are more like particles.20 These findings suggest that the canonical clause in the literature, which represents the minority type of independent clauses with verbal predicates in the data (23%; 92 out of 397 clauses), may play an even less significant role, further calling into question its status in the grammar of Japanese.

5. Summary and Conclusions

We have thus seen that Japanese clauses with the verbal predicate most typical in conversation include pragmatic elements (PEs) at the end, along with occasional overt arguments, which we represented in (9) above.

The commonly held view that Japanese is a verb-final language is problematic from the perspective of what speakers produce in conversation, the fundamental form of language (Schegloff, 1996). Only a minority of examples show the verb in finite form ending the clause. Examinations of constructed examples without context, taking the structure of languages like English as a base form, have led to an outcome where syntactic characteristics specific to Japanese that encode (inter)subjective functions have been ignored in the discussion of its syntax.

We found a minority of clauses to end with finite verbs, which one might think are examples of the canonical clause. Such clauses have a clear preference of employing five specific verbs: iu ‘to say’; omou ’to think’; chigau ‘to differ’; wakaru ‘to understand’; and suru/yaru ‘to do’. We further found that examples involving such verbs are not very much like canonical clauses as traditionally understood; many of them are not verbs serving as predicates taking arguments, but they instead look and function like particles. This further underscores our general discovery that clauses ending with the finite verb, the so-called canonical clause, are not what Japanese speakers commonly produce in conversation.

Our findings provide a new understanding of Japanese grammar, a grammar based on conversation, and offer a glimpse into the lexicon of speakers, where different verbs and inflectional forms are associated with specific uses. We have thus seen that everyday interactions, represented in sizable corpora (e.g., CEJC), help us explore what ‘grammar’, ‘clause’, and ‘verb’ might be for actual speakers.