Jump to ContentJump to Main Navigation
Show Summary Details
More options …


An Interdisciplinary Journal of the Language Sciences

Editor-in-Chief: Gast, Volker

IMPACT FACTOR 2018: 1.066

CiteScore 2018: 0.97

SCImago Journal Rank (SJR) 2018: 0.384
Source Normalized Impact per Paper (SNIP) 2018: 1.409

See all formats and pricing
More options …
Volume 57, Issue 5


Japanese subject markers in linguistic change: A quantitative analysis of data spanning 90 years and its theoretical implications

Satoshi Nambu
  • Corresponding author
  • School of Languages, Literatures, Cultures & Linguistics, Monash University, Menzies Building, 20 Chancellors Walk, Clayton 3800, Australia
  • Email
  • Other articles by this author:
  • De Gruyter OnlineGoogle Scholar
Published Online: 2019-05-31 | DOI: https://doi.org/10.1515/ling-2019-0018


Based on a statistical analysis of a corpus data from the period of 1915–2005, this article discusses two variants for a subject marker in Japanese, and argues that it is a case of linguistic change in progress. While representing effects of three linguistic factors on the use of the variants, the chronological observation of each factor revealed that this phenomenon demonstrates the Constant Rate Effect. The quantitative data also provides firm evidence for effects of other independent diachronic changes on the current phenomenon, pushing the change further by shrinking the linguistic environment for the dying-out variant. Dissecting the relationship between those diachronic changes and the current phenomenon in a quantitative manner, the findings of the study reveal that the two competing hypotheses in theoretical syntax properly capture the essential syntactic properties of the phenomenon in the contemporary Japanese and the discrepancy of the two accounts is attributable to their data reflecting a different stage of its ongoing change.

Keywords: Japanese; syntax; quantitative study; language change; constant rate effect

1 Introduction

This article discusses two subject markers in Japanese as a case of linguistic variation, whereby a subject in certain embedded clauses can be marked either by the nominative case particle ga or the genitive case particle no, as shown below.


Case particles ga/no as a subject marker


“Ken ate the sushi that Naomi bought.”

In the field of variationist sociolinguistics (Tagliamonte 2012), many cases of linguistic variation have been identified as a reflex of language change in progress in which one variant (typically as an innovative form) diffuses while the other variant (as a conservative form) diminishes (Weinreich et al. 1968; Chambers 2002). In the same vein, this article argues that the two case particles as linguistic variants for a subject marker represent a state of an ongoing change, providing quantitative evidence for the change in which one variant is dying out in return for the expansion of the other. As will be overviewed in Section 2, the competition between the two particles is not a short-tem phenomenon but is rather part of the longitudinal change in the distribution of grammatical functions of the case particles. In the current study, a quantitative analysis of the data from two speech corpora, which spans 90 years of utterance (1915–2005), reveals that the long-standing change has not yet reached its end and is observable from a data with a sufficient period of time (Sections 3 and 4). As for how the case particles undergo the change in grammatical environments, the data serves a case study of the Constant Rate Effect (Kroch 1989, Kroch 2001), which argues that a syntactic change progresses in every context at the same rate over time. In addition to the quantitative properties of the change, this article discusses how other linguistic changes contribute to a further progress of the change in question, narrowing the linguistic environment for the dying-out variant (Section 5). The findings of the study reveal that the two competing hypotheses in theoretical syntax properly capture the essential syntactic features of the phenomenon in the contemporary Japanese and the discrepancy of the two accounts can be treated as a reflex of its ongoing transition of the syntactic properties (Section 6).

2 Background

Syntactic properties of subjects with the particles ga and no, particularly regarding where they can(not) occur, have been delved into in the large body of literature in theoretical syntax (e.g. Harada 1971; Watanabe 1996; Ochi 2001; Hiraiwa 2005; Miyagawa 2011; among others; see Maki and Uchibori 2008 for an overview), and various syntactic restrictions on the use of the particle no have been found as given below.


Subject in a main clause


“Naomi bought sushi.”


An embedded clause subject with complementizer -to (Inoue 1976)


“(I) thought that Naomi swam.”


An embedded clause subject with an accusative object (Harada 1971)


“a day when Naomi bought sushi”

The examples in (2) indicate that no is inferior as a subject marker in comparison to ga. Among the previous studies, Harada (1971) and Shibatani (1975) point out that the acceptability of the inferior variant no differs in terms of speakers’ generation, indicating that the older generation tends to be more generous to the use of no, and suggest the existence of an ongoing change in the variation. In the field of Japanese historical linguistics, it is known that the distribution of particles ga and no in grammatical function, including subject marker, has been changing in the long period of time (Yamada 1936; Frellesvig 2010; Yamada 2010; Kinsui 2011). Table 1 presents a summary of the historical transition of particles ga and no adapted from Frellesvig (2010).

Table 1:

Transition of ga and no in grammatical function (Frellesvig 2010: 368)

As given in the categories for row in Table 1, the two particles have a variety of functions other than the embedded subject marker (represented in the second row). Note that no in the second row has the symbol (+) in the latest period of time; it indicating that no is a less preferred choice than ga. As for the chronological change, Table 1 indicates that the grammatical functions of the two particles overlapped extensively in the earlier period, but the distribution has been shifting toward a split, suggesting that this longitudinal change is heading to their complementary distribution. Given this historical context, the variants ga and no as an embedded subject marker can be recognized as the last stand in the course of the longitudinal change, reinforcing the view that the variants is a reflex of an ongoing linguistic change, whereby the use of no is ceasing to exist.

Notice that the third row in Table 1 indicates that a main clause subject used to be zero marked for a while, which was taken by the particle ga at least in the fourteenth century. Since the emergence of ga as a main clause subject marker precedes the change in the embedded subject markers ga and no, at least in a recognizable way as given in Table 1, its relevance to the current phenomenon has been taken up in Japanese historical linguistics (Yamada 2010; Kinsui 2011; and references therein). Prior to the emergence, both particles ga and no were considered as genitive markers (adnominal marking in Table 1), as given in (3).


“my life”

(Man’yoshu, the 7-8th century, cited in Frellesvig 2010: 89)

Given the fact that the use of ga as a genitive marker has disappeared while ga started being used as a subject marker in a main clause, as indicated in Table 1, Frellesvig (2010) argues that the grammatical function of ga has shifted from genitive to nominative, in contrast to no that remained as a genitive marker. During the time period of the shift, the change in the variants ga and no as embedded subject markers has already started. Yamada (2010) provides quantitative evidence for the change in the direction of the decrease in the use of no as an embedded subject marker, comparing the variable uses of ga and no between the texts of the same story from different periods of time: Heike Monogatari (thirteenth–fourteenth century) and Amakusaban Heike Monogatari (1592), as given in Table 2. The different rate of the uses of ga and no between the two texts in Table 2 suggests a transition from no to ga in 200–300 years. Due to a plausible difference in the criteria of data extraction, we cannot compare the numbers here directly with the quantitative data in the current study, but the ongoing change presented in later section suggest that the observed change in the sixteenth century continues up to date.

Table 2:

A transition of embedded subject markers (Yamada 2010).

3 Data

3.1 Corpus

Aarts et al. (2013) obtained data of English modal uses from Diachronic Corpus of Present-day Spoken English, which consists of two corpora (ICE-GB and London-Lund Corpus) ranging from 1958 to 1992, and presents their quantitative change during the period. In a similar vein, this study used two speech corpora, which share similar properties and are chronologically consecutive. One is Okada Collection (hereafter OC) (Kanazawa and Aizawa 2015; Aizawa and Kanazawa 2016), which stores speeches by politicians and other distinguished people recorded in the first twentieth century (years of utterance: 1915–53). The other is the Minutes of Japanese Diet (hereafter MJD), which stores records of Diet members’ speeches from every meeting in the Diet in the latter twentieth century (1947 up to the present) (Matsuda 2004, Matsuda 2008) 1. Spanning a relatively long period of time, the two corpora provide us with an ideal dataset to investigate the language change in progress. Moreover, both corpora primarily consist of monologue speeches in public by those distinguished figures, which can be categorized as formal speech, and thus, these properties minimize stylistic differences in speech between the corpora.

3.2 Data extraction

The target of data extraction is the use of ga and no as embedded subject markers, defining the set of environments where the variable could possibly occur in (i.e. principle of accountability in variationist sociolinguistics, Tagliamonte 2012). For instance, as delimiting a type of subordinate clause, embedded clauses containing a ga/no subject were extracted when the clauses accompany a head noun. The examples from the OC data given in (4) represent a use of ga/no subject in a relative clause with the head noun koufuku ‘happiness’ and a clause with the nominalizer no as the head noun.


Relative clause


“the happiness that a person desires”

(u: Motojiro Makino, 1925) 2


Embedded clause with the nominalizer no


“(the thing) that we cover (the cost)”

(u: Yoshiharu Tazawa, 1936)

From the MJD corpus, 100 tokens of the variants were extracted from each of 76 speakers in that the data spreads throughout the period of time in equal proportion. As for the OC corpus, due to its small size (relative to the MJD corpus), the uses of the variants by all speakers in the corpus were extracted, yielding the data of the uses by 78 speakers. After manually searched in the corpora, 7,600 tokens were obtained from the MJD corpus (1947–2005), and 724 tokens from the OC corpus (1915–44), yielding the total data spanning 90 years (1915–2005). Table 3 shows the distribution of ga and no in each data. As expected, the general tendency in the two datasets indicates that no is less frequent than ga as an embedded subject marker.

Table 3:

Frequency of ga and no as variants for an embedded subject marker.

4 Quantitative analysis

4.1 Ongoing change

Figure 1 demonstrates a transition of the rate of the use of no over ga from 1915 to 2005, represented by a logistic regression line with a 95% confidence interval for the data. 3 As predicted from the historical path of the two particles in Table 1 in Section 2, the regression line represents about a 30% decrease in the no use in the period of 90 years. The graph uses speech year on the x-axis as timeline, but replacing it with speaker’s birth year produces a similar pattern, since they are highly correlated in the data (r = 0.93, p < 0.001).

Quantitative shift in the use of ga/no as variants.
Figure 1:

Quantitative shift in the use of ga/no as variants.

4.2 Statistical approach

4.2.1 Logistic regression model: Factors and factor levels

In order to confirm that the change observed in Figure 1 is not due to some effect of other linguistic factors, a logistic regression model was constructed in the R environment (R Core Team 2016). Table 4 is a summary of the factors and factor levels used for the analysis, as introduced below. As for the data used for the analysis, linguistic environments that contain a value yielding only one variant were excluded from the statistical analysis (i.e. knockout factor in Tagliamonte 2012), resulting in 6,264 tokens for this analysis.

Table 4:

Factors and factor levels for the logistic regression model.

Speech year was employed as a time scale to examine the change in the variation. As mentioned in Section 4.1, this factor has a high correlation with speakers’ birth year, and thus, using speakers’ birth year yields a similar result.

Adjacency with two levels adjacent and non-adjacent was configured in order to take into account the effect of a linear distance between a subject with ga/no and its predicate. The existence of interveners between the subject and its predicate degrades the acceptability of no (Harada 1971; Miyagawa 2011; Nambu and Nakatani 2014), as the examples from the OC data given in (5).


Adjacent environment


“the part that the government discusses”

(u: Shigenobu Okuma, 1916)


Non-adjacent environment


“the fact that the citizens equally acknowlege”

(u: Joichiro Yamaji, 1931)

Predicate type with three levels was employed for the analysis as a relative scale of stativity, following Nambu (2007) and Kim (2009) that reports that a stative predicate has more uses of no as a subject maker. 4 The relative stativity is simplified for this study, as a nominal property of the embedded clause, represented by a set of features +/−N and +/−V (Chomsky 1970), yielding the order of stativity as adjective/nominal adjective 5 [ +N, +V]> verb [−N, +V].

Head noun was configured with two levels: formal noun and content noun. As suggested in the case of Predicate type, nominal properties of the embedded clause has a large impact on the use of no subject (Ochi 2001; Kikuta 2002; Whitman 2006; Kinsui 2011; Miyagawa 2011), which will be discussed in Section 5, and therefore, it is predicted that formal nouns, which functions as a grammatical placeholder, are relatively less preferred for the use of no in comparison to content nouns.

4.2.2 Results

Table 5 shows the coefficient estimate, the standard error, the resulting z-value and the corresponding p-value of the computed model, adopting no as the application value. In Table 5, the factor Speech year shows a significant effect in the direction of decreasing no as an increase of one speech year, given the minus coefficient estimate as log odds (−0.026), which takes into account the effects of other factors in the model (cf. Tagliamonte and Baayen 2012). 6 In other words, the ongoing change in the ga/no variation was attested in the data even after excluding the effects of other linguistic factors.

Table 5:

Summary of the logistic regression model (N = 6,264).

Table 5 also provides values for each factor level relative to a reference level as a baseline of each factor. 7 The factor Adjacency has adjacent as a reference level, and the results indicate that the non-adjacent environment has fewer uses of no with a statistical significance, as argued in the previous studies (Harada 1971; Miyagawa 2011; Nambu and Nakatani 2014). The effects of the factors Predicate type and Head noun reflect the nominal properties of the embedded clause: more nominal properties have more uses of the no subject. Predicate type has adjective as a reference level, and the coefficient estimate of verb indicates that the use of no with verbs is less frequent than with adjectives with a statistical significance, but nominal adjective is not significantly different from adjective, as predicted in the order of stativity adjective/nominal adjective>verb in Section 4.2.1. Head noun indicates that content nouns as a head of an embedded clause are likely to have more uses of the no subject in comparison to formal nouns as the reference level.

4.3 Three linguistic factors and the constant rate effect

Let us now focus on a chronological shift regarding the three linguistic factors introduced in Section 4.2. Figure 2 exhibits the shift as the rate of no over ga as the variants in terms of the three factors. As given in Table 5 in Section 4.2.2, the subcategories (factor levels) of each factor represent different degrees of the no use. More noteworthy is the fact that almost all of the subcategories shift in parallel with respect to the chronological change. This parallel shift between different linguistic environments calls our attention to the Constant Rate Effect (CRE, Kroch 1989; Pintzuk 1991; Santorini 1992; also see Matsuda 2003; Nambu and Matsuda 2007), which originally refers to quantitative data of diachronic changes in English syntax and argues that a syntactic change proceeds in equal proportion across diverse linguistic environments with different degrees of a variable use. The CRE implies that a syntactic change does not apply categorically to linguistic environments, which could cause a disappearance of one variable in some environments without any change in others; instead, a change applies equally in relative frequency. The data in this study presents such a pattern in an ongoing change and successfully demonstrates the CRE, which is originally from English data.

Quantitative shift in the use of ga/no in each linguistic environment.
Figure 2:

Quantitative shift in the use of ga/no in each linguistic environment.

One large exception of the CRE in Figure 2 is the nominal adjective as predicate type (NomAdj in Figure 2). Its rate shows a large decline as a S-shaped curve (Denison 2003; Blythe and Croft 2012), crossing the other lines. However, instead of implicating a drastic quantitative change within the category, this idiosyncratic behavior is in fact due to a mixture of different types in terms of style for the category, and reflects a categorical effect of an independent linguistic change. There are two types of nominal adjective in Japanese in the time period of the data: colloquial and literary forms. In addition to a relatively small number of tokens for nominal adjectives in the OC data (the older corpus), most of the nominal adjectives are the literary form (14/16), which only appeared with the no subject in the data. In contrast, the MJD data, reflecting the contemporary Japanese, has a zero use of the literary form. 8 Thus, it can be interpreted that the no subject as a conservative form of the change co-occurred more frequently with the archaic literary form of nominal adjective in the data, and the disappearance of the form caused the drastic decline of the no subject in nominal adjective in Figure 2. In other words, the idiosyncratic behavior of nominal adjective seen in Figure 2 reflects the independent linguistic change across two linguistic forms for nominal adjective, instead of a quantitative change in one form, and the change in nominal adjective resulted in a disappearance of a linguistic environment where the no subject was favorably used. Therefore, nominal adjective in Figure 2 is not inconsistent with the CRE as a quantitative attribute of syntactic change. In Section 5, I introduce this type of effects of other linguistic changes that bring such a “qualitative” change to the variation due to a disappearance of a linguistic environment for a variable use, which leads to a sudden quantitative shift as observed in nominal adjective in Figure 2.

5 Effects of other linguistic changes

The quantitative analysis in Section 4 provides evidence for the ongoing change in the variation, whereby the use of no as an embedded subject marker has been decreasing over time, and also that the change progresses in equal proportion across different linguistic environments, demonstrating the CRE. This section discusses how the change has been progressing in a “qualitative” sense, as in the case of nominal adjective observed in Section 4.3, addressing a question about effects of other linguistic changes on the variation in that other changes pushed the decrease in the use of the no subject further.

For the purpose, I introduce two linguistic environments found in the OC data (the older corpus) that produce a higher frequency of no than ga as variants. 9 Given the fact that the two environments have already disappeared in the contemporary Japanese due to some other linguistic changes, I argue that their disappearance directly brought about a further decrease in the no use.

5.1 Change in verbal morphology

One linguistic environment that has a high frequency of the no use in the OC data is the adnominal form of verbs and auxiliary verbs. From a historical perspective, the adnominal form used to be distinguishable from the conclusive, as given in (6), but they have merged due to a long-term linguistic change and the difference is no longer attestable in verbal morphology in the contemporary Japanese. 10

Table 6:

Morphological forms of verbs and ga/no in the OC data.


“the thing called diary which men are said to keep”

(Tosa Nikki, the 10th century, adapted from Frellesvig 2010: 355)

The change in verbal morphology has a long trajectory, which began at least in the twelfth century (Frellesvig 2010; Takayama 2011). The OC data (the older corpus) contains some uses of the adnominal form, as exemplified in (7), but the MJD data, which reflects the contemporary Japanese, does not contain any of the uses. Table 6 is the frequency of ga and no occurred with the adnominal and conclusive form of verbs in the OC data. It indicates that the adnominal form has the use of no more frequently than ga (X2 = 35.84, df = 1, p < 0.001), and thus, the disappearance of the adnominal form directly brought about a further decrease in the use of the no subject.


“an invention that the old man has accomplished”

(u: Eiichi Shibusawa, 1922)


“the workplace where you all have”

(u: Hideki Toujou, 1940/41)

As discussed in Section 4.3, effects of other linguistic changes on the chronological status of the variation is independent from the CRE, and therefore, taking those effects into account refines the data to demonstrate the CRE more accurately. For this reason, I present a graph for predicate again in Figure 3 (without nominal adjective), including the new category Verb’ that represents verbs without the adnominal form in comparison to the original data under Verb. Compared to Verb that includes the adnominal form, the chronological shift of Verb’ is more parallel to Adjective, and thus, this case study shows that excluding effects of other linguistic changes contributes to elucidating the quantitative attribute of syntactic change as the CRE describes.

Refining the shift in the use of ga/no (Verb = data including the attributive form, Verb’ = data without the attributive form).
Figure 3:

Refining the shift in the use of ga/no (Verb = data including the attributive form, Verb’ = data without the attributive form).

5.2 Emergence of nominalizer no

In historical syntax, it is known that the adnominal form of verbs, introduced in Section 5.1, functioned as a marker of embeddedness of a clause with a null/zero head of the clause, as shown in (8) (Yoshimura and Nishina 2008; Aoki 2010). Due to the adnominal and conclusive merger, such an embedded clause required an alternative marker of embeddedness, and consequently, nominalizer no came out in Japanese grammar, at least in the early seventeenth century (Aoki 2010; Frellesvig 2010; Kinsui 2011). In the contemporary Japanese, it is obligatory to put a nominzalizer for an embedded clause such as (9).


Embedded clause with a zero head (φ)


“hearing that there was a person who was crying terribly”

(Ise Monogatari, the 10th century, adapted from Frellesvig 2010: 363)


Embedded clause with nominalizer no in the contemporary Japanese


“the time when Ken woke up”

Here again, the OC data contains the uses of ga/no subject in embedded clauses without a nominalizer, as opposed to the MJD data that has none of the uses. Examples of the attested uses in the OC data are given in (10), and the frequency in the OC data is given in Table 7. The result of Fisher’s exact test shows a significant difference in the ratio of the no use between the categories with and without a nominalizer (p < 0.01), suggesting a high frequency of the no subject in a clause without a nominalizer. Thus, the disappearance of the environment caused a further decrease in the use of the no subject in the course of the linguistic change.

Table 7:

Nominalizer φ/no and ga/no in the OC data.


“even reached to the fact that the term of session comes to its end”

(u: Senjuro Hayashi, 1937)


“the fact that the enemy has not come”

(u: Seigo Nakano, 1942)

6 Theoretical implication

In addition to the CRE, providing empirical evidence for the language change bears another aspect of theoretical implication, particularly rationalizing the existence of two competing hypotheses for the contemporary Japanese in theoretical syntax: the C- and D-licensing hypotheses (Hiraiwa 2005; Miyagawa 2011). The major distinction between the two hypotheses is that the former presumes an equivalent syntactic structure for subjects with the nominative ga and the genitive no, while the latter considers their structures different.

As given in Figure 4, the D-licensing hypothesis assumes that the nominative structure contains a CP level (Figure 4(a)) but the genitive does not (Figure 4(b)), which is to capture the restricted features of the subject with the genitive no in that it occurs only with a predicate with a stative interpretation (Miyagawa 2011:1277), as exemplified in (11).

Structures under the D-licensing hypothesis (Miyagawa 2011).
Figure 4:

Structures under the D-licensing hypothesis (Miyagawa 2011).


‘He is wearing the shirt that sustained a stain.’


‘Please show me the shirt that was suddenly stained.’ (Miyagawa 2011:1279)

The examples in (11) represent that the use of no for the subject is not allowed with the predicate when the adverb totsuzen ‘suddenly’ is added to the clause and changes the interpretation of the predicate from stative to eventive. Reflecting this property, the syntactic structure for the genitive subject given in Figure 4(b) does not include the CP level, which, according to its syntactic theory, leads to the “defectiveness” of T that only permits a predicate with a stative interpretation.

On the contrary, the C-licensing hypothesis proposed by Hiraiwa (2005) assigns the same structure, which contains a CP level, to both the nominative and genitive structures, but adds the nominal feature [+N] in C as a condition for the case alternation. Figure 5 represents the nominative/genitive structure under the assumption of Hiraiwa’s C-licensing analysis. Contrary to the D-licensing hypothesis, the C-licensing hypothesis predicts that the predicate with the genitive subject is adnominal, as a reflex of the nominal feature [+N]. This assumption derives from the fact that a nominal adjective, such as suki-na ‘like’ in the example (12) given below, represents its adnominal form in the embedded clause where the genitive subject can occur.

Structure under the C-licensing hypothesis (Hiraiwa 2005).
Figure 5:

Structure under the C-licensing hypothesis (Hiraiwa 2005).


‘The music that John likes is the Blues.’

(Hiraiwa 2005:112)

In this article, I do not delve into the detail of their case licensing mechanisms based on their theoretical assumptions; however, from a perspective of the ongoing change verified in this article, I bring into focus their empirical data that the two hypotheses are founded on, arguing that their difference attributes to syntactic properties of the contemporary Japanese that lie in a different stage of the linguistic change in progress.

As presented in Sections 4 and 5, the findings of the corpus-based study reveal that the use of the genitive subject has been decreasing over time from a quantitative perspective, accompanying with the qualitative change as narrowing the linguistic environment for the no subject. Supporting the perspective of this section, let us briefly overview Whitman (2006), which discusses the phenomenon in terms of diachronic syntax. Whitman (2006) argues that no subjects originate in clauses headed by a complementizer in C with the nominal feature [N], such as adnominal clauses in which the [N]CP is morphologically realized as the adnominal form of predicate. Whitman (2006) discusses that the merger of the adnominal and conclusive forms resulted in the loss of the cue for [N]CP (and further caused the loss of the feature [N] in C), and then, although no subjects continue to appear in environments that used to be [N]CP contexts, they have been disappearing, as observed in Section 4.

In the light of the discussion in Whitman (2006), the adnominal form, which the C-licensing hypothesis is based on, as given in (12), used to be an essential cue for the use of the no subject in the history of Japanese, but, due to the morphological merger, at least on the surface level, it is fading away from the central role of the phenomenon, accompanied by a further decrease in the use of no subject. Therefore, we can interpret the C-licensing hypothesis in a way that it frames one aspect of the contemporary Japanese, particularly in the early stage of the ongoing change. As the other side of the ongoing change, the data featured by the D-licensing hypothesis emerged from a further progress of the change with more restrictions on the use of the genitive subject; the break-up of the nominative- and genitive-subject structures in the D-licensing hypothesis indicate the direction of the further syntactic change in the contemporary Japanese.

Thus, the perspective of the ongoing change incorporating the view of its diachronic status rationalizes the existence of the two hypotheses in theoretical syntax; both of the hypotheses properly capture the essential syntactic features of the phenomenon but are founded on data from a different stage of the ongoing change. Although this article does not argue for a certain syntactic structure along with some specific case licensing mechanism for the phenomenon, it suggests that empirically substantial data aligning with diachronic syntax can bring out a new perspective on syntactic properties of the contemporary Japanese that has been delved into in theoretical syntax.

7 Conclusion

Benefiting from the data spanning 90 years, this corpus-based study provides evidence for a quantitative change in the use of the variants ga and no as a subject marker in Japanese. The findings from the statistical analysis also confirmed the effects of other linguistic factors, and the chronological observation of each factor revealed that this phenomenon exemplifies the CRE. The corpus data also provides substantial evidence for the effects of other independent diachronic changes on the current phenomenon, pushing the change further by shrinking linguistic environments for the dying-out variant. The intriguing fact is that those observed effects are attributed to the changes in nominal properties of the clause in question, which is one of the essential aspects to understand its syntactic components of the contemporary Japanese. Theoretical aspects of the phenomenon can be shed light on further by scrutinizing more data from a corpus and delving into a diachronic shift of relevant linguistic features; however, we leave this to further research.

The abbreviations used in the glosses and graphs are




adjectival copula






complementizer phrase






determinate phrase


emphatic topic












nominal feature






noun phrase












tense phrase




verb phrase


little verb


little verb phrase


  • Aarts, Bas, Joanne Close & Sean Wallis. 2013. Choices over time: Methodological issues in investigating current change. In Bas Aarts, Joanne Close, Geoffrey Leech & Sean Wallis (eds.), The verb phrase in English: Investigating recent language change with corpora, 14–45. Cambridge: Cambridge University Press. Google Scholar

  • Aizawa, Masao & Hiroyuki Kanazawa. 2016. SP-ban Enzetsu Rekoodo ga Hiraku Nihongo Kenkyu [Exploring 78rpm record archives for Japanese language research]. Tokyo: Kasama Shoin. Google Scholar

  • Aoki, Hirofumi. 2010. Juntaiku [Adnominal clause]. In Yoshiyuki Takayama & Hirofumi Aoki (eds.), Gaidobukku Nihongo Bunpooshi, 105–117. [Guidebook to the history of Japanese grammar]. Tokyo: Hitsuji Shobo. Google Scholar

  • Blythe, Richard A. & William Croft. 2012. S-curves and the mechanisms of propagation in language change. Language 88(2). 269–304. Web of ScienceCrossrefGoogle Scholar

  • Chomsky, Noam. 1970. Remarks on nominalization. In Roderick A. Jacobs & Peter S. Rosenbaum (eds.), Readings in English transformational grammar, 184–221. Boston: Ginn. Google Scholar

  • Denison, David. 2003. Log(ist)ic and simplistic S-curves. In Raymond Hickey (ed.), Motives for language change, 54–70. Cambridge: Cambridge University Press. Google Scholar

  • Frellesvig, Bjarke. 2010. A history of Japanese language. Cambridge: Cambridge University Press. Google Scholar

  • Harada, Shin-Ichi. 1971. Ga-no conversion and idiolectal variations in Japanese. Gengo Kenkyu 60. 25–38. Google Scholar

  • Hiraiwa, Ken. 2005. Dimensions of symmetry in syntax: Agreement and clausal architecture. Cambridge, MA: MIT dissertation. Google Scholar

  • Inoue, Kazuko. 1976. Henkei Bunpoo to Nihongo [Transformation Grammar and Japanese]. Tokyo: Taishukan. Google Scholar

  • Iwasaki, Shoichi. 2002. Japanese. Amsterdam/Philadelphia: John Benjamins. Google Scholar

  • Jack, Chambers. 2002. Patterns of variation including change. In J. K. Chambers & Natalie Schilling (eds.), The handbook of language variation and change, 297–323. Malden & Oxford: Wiley-Blackwell. Google Scholar

  • Kanazawa, Hiroyuki & Masao Aizawa. 2015. Taishoo Shoowa Zenki, Seiji/Jitsugyoo/Bunka, Enzetsu/Kouenshu: SP-ban rekoodo mojika shiryoo [Taisho and the early Showa-era, politics/business/culture, speech/lecture: a transcribed material of 78rpm record]. Tokyo: Nichigai Associates. Google Scholar

  • Kikuta, Chiharu. 2002. Ga-no Kootai Genshoo no Hihaseiteki Bunseki: Jutsugo Rentaikei no Meishisei [Ga-no conversion in the new light: A non-derivational, mixed category analysis]. Doshisha Studies in English 74. 93–136. Google Scholar

  • Kim, Eunju. 2009. Gendaigo no rentai shuushokusetsu niokeru joshi “no”. [The particle ‘no’ in the modificational construction in Modern Japanese]. Nihongo Kagaku 25. 23–42. Tokyo: Kokuritsu Kokugo Kenkyujo. Google Scholar

  • Kinsui, Satoshi. 2011. Toogoron [Syntax]. In Satoshi Kinsui, Yoshiyuki Takayama, Tomohide Kinuhata & Tomoko Okazaki (eds.), Bunpooshi, 77–166. [History of grammar]. Tokyo: Iwanami Shoten. Google Scholar

  • Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1. 199–244. CrossrefGoogle Scholar

  • Kroch, Anthony. 2001. Syntactic change. In Mark Baltin & Chris Collins (eds.), The handbook of contemporary syntactic theory, 699–729. Malden: Blackwell Publishers. Google Scholar

  • Maki, Hideki & Asako Uchibori. 2008. Ga/no conversion. In Shigeru Miyagawa & Hajime Saito (eds.), The Oxford handbook of Japanese linguistics, 192–216. Oxford: Oxford University Press. Google Scholar

  • Matsuda, Kenjiro. 2003. Constant rate hypothesis, age-grading, and the apparent time construct. University of Pennsylvania Working Papers in Linguistics 9(2). 123–134. Google Scholar

  • Matsuda, Kenjiro. 2004. Gengo shiryoo toshiteno kokkai kaigiroku kensaku shisutemu [The online full-text database of the Minutes of the Diet: Its potentials and limitations]. Theoretical and Applied Linguistics at Kobe Shoin 7. 55–82. Google Scholar

  • Matsuda, Kenjiro. 2008. Kokkaikaigiroku o Tsukatta Nihongo Kenkyu [Research on Japanese using the minutes of Japanese diet]. Tokyo: Hitsuji Shobo. Google Scholar

  • Miyagawa, Shigeru. 2011. Genitive subjects in Altaic and specification of phase. Lingua 121. 1265–1282. Web of ScienceCrossrefGoogle Scholar

  • Nambu, Satoshi. 2007. Teiryooteki bunseki ni motozuku gano kootai saikoo [Reconsideration of Ga/no conversion based on a quantitative analysis]. Gengo Kenkyu 131. 115–149. Google Scholar

  • Nambu, Satoshi & Kenjiro Matsuda 2007. Change and variation in ga/no conversion in Tokyo Japanese. In Joseph C. Salmons & Shannon Dubenion-Smith (eds.), Historical Linguistics 2005: Selected papers from the 17th international conference on historical linguistics, 119–131. Amsterdam/Philadelphia: John Benjamins. Google Scholar

  • Nambu, Satoshi & Kentaro Nakatani. 2014. An experimental study on adjacency and nominative/genitive alternation in Japanese. In Shigeto Kawahara & Mika Igarashi (eds.), Proceedings of FAJL 7: Formal approaches to Japanese linguistics, 131–142. Cambridge: MITWPL. Google Scholar

  • Ochi, Masao. 2001. Move F and ga/no conversion in Japanese. Journal of East Asian Linguistics 10. 247–286. CrossrefGoogle Scholar

  • Pintzuk, Susan. 1991. Phrase structures in competition: Variation and change in Old English word order. Philadelphia, PA: University of Pennsylvania dissertation. Google Scholar

  • R Development Core Team. 2016. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. http://www.R-project.org/

  • Santorini, Beatrice. 1992. Variation and change in Yiddish subordinate clause word order. Natural Language & Linguistic Theory 10. 595–640. CrossrefGoogle Scholar

  • Shibatani, Masayoshi. 1975. Perceptual strategies and the phenomena of particles conversion in Japanese. In Robin E. Grossman, L. James San & Timothy J. Vance (eds.), Papers from the Parasession on Functionalism, 469–480. Chicago: Chicago Linguistic Society. Google Scholar

  • Tagliamonte, Sali. 2012. Variationist sociolinguistics: Change, observation, interpretation. Malden & Oxford: Wiley-Blackwell. Google Scholar

  • Tagliamonte, Sali & R. Harald Baayen. 2012. Models, forests, and trees of York English: Was/were variation as a case study for statistical practice. Language Variation and Change 24. 135–178. CrossrefWeb of ScienceGoogle Scholar

  • Takayama, Yoshiyuki. 2011. Jutsubu no koozoo [Structure of predicate]. In Satoshi Kinsui, Yoshiyuki Takayama, Tomohide Kinuhata & Tomoko Okazaki (eds.), Bunpooshi, 19–75. [History of Grammar]. Tokyo: Iwanami Shoten. Google Scholar

  • Watanabe, Akira. 1996. Nominative-genitive conversion and agreement in Japanese: A crosslinguistic perspective. Journal of East Asian Linguistics 5. 373–410. CrossrefGoogle Scholar

  • Weinreich, Uriel, William Labov & Marvin Herzog. 1968. Empirical foundations for a theory of language change. In Winfred P. Lehmann & Yakov Malkiel (eds.), Directions for historical linguistics, 95–195. Austin: University of Texas Press. Google Scholar

  • Whitman, John. 2006. The attrition of genitive subjects in Korean and Japanese. Paper presented at Syracuse/Cornell Workshop on the Internal Syntax of Nominalized Clauses, 28 February. Google Scholar

  • Yamada, Masahiro. 2010. Kakuzyoshi “ga” no Tsuujiteki Kenkyu [A diachronic study on the case particle “ga”]. Tokyo: Hitsuji Shobo. Google Scholar

  • Yamada, Masaki. 1936. Edokotoba no Kenkyu: Ukiyoburo/Ukiyodoko no Gohoo [A study on Japanese in the Edo period: Word usage in Ukiyoburo/Ukiyodoko]. Tokyo: Musashino Shobo. Google Scholar

  • Yoshimura, Noriko & Akira Nishina. 2008. Gano kootai to karadaimeishi nituite [Gano conversion and empty pro]. Paper presented at The Sixth International Conference on Practical Linguistics of Japanese. Google Scholar


  • 1

    The MJD corpus is publically available from their website (http://kokkai.ndl.go.jp/). 

  • 3

    Data that do not have speech year information (223 tokens) are not included in this analysis. 

  • 4

    Tokens with copula as a factor level of the factor Predicate type were excluded in the statistical analysis, because there were only two uses of the no subject in the MJD data and zero in the OC data. 

  • 5

    Nominal adjectives describe a state of an entity, like adjective but are morphologically like nouns that require a copula to be a predicate (Iwasaki 2002: 62).


    ‘a beautiful person’


  • 6

    In addition to this analysis, statistical significances of the effects of the three linguistic factors in the model were confirmed by a likelihood-ratio chi-squared test (all three shows p < 0.001). 

  • 7

    The effects of these factors were also investigated within each corpus data individually, confirming their consistent effects between the two data. 

  • 8

    This change regarding style may be linked to the drastic change in the society; the OC data (the older corpus) were recorded in the time period of Pre-World War II, in contrast to the MJD data recorded in Post-World War II. 

  • 9

    The two environments were not used for the statistical analysis in Section 4, since one environment was only found in the OC data and the other was not included in the original data extraction but extracted independently for the discussion in this section. 

  • 10

    According to Frellesvig (2010: 100), some verbal conjugation classes, such as quadrigrade verbs, have already shown the identical form for the adnominal and conclusive functions in Old Japanese in the ninth century. However, the data given in Table 6 contain all of the conjugation classes, since it aims to claim that the adnominal form that is different from the conclusive form is the cue to trigger the use of the no subject (but not to compare the uses of no between the two forms only in the conjugation classes that allow both of the forms). 

About the article

Published Online: 2019-05-31

Published in Print: 2019-09-25

Citation Information: Linguistics, Volume 57, Issue 5, Pages 1217–1238, ISSN (Online) 1613-396X, ISSN (Print) 0024-3949, DOI: https://doi.org/10.1515/ling-2019-0018.

Export Citation

© 2019 Walter de Gruyter GmbH, Berlin/Boston.Get Permission

Comments (0)

Please log in or register to comment.
Log in