The deverbal noun suffix pJ *-ra can be reconstructed as a suffix that derived nominal and adnominal forms from verbal adjectives such as in OJ aka- ‘to be red’ → akara ‘red’, usu- ‘to be fine’ → usura ‘fine’, uma- ‘to be tasty’ → umara ‘tasty’, yo₂ - ‘to be good’ →yo₂ra ‘good’, sakasi- ‘to be wise’ → sakasira ‘wisdom’, kanasi- ‘to be sad’ → kanasira ‘sadness’, EMJ be- necessitive → bera nar- necessitive, etc.; see (6a) (Antonov 2007: 102, 111, 128–132, 153, 160, 196, Vovin 2009: 436–440).² Unless OJ tatara 3.4 ‘foot-bellows’ can be derived from OJ tat- B ‘stand’ (< *tata-) and this suffix, there are no indications of deverbal derivation.

Clausal (ad)nominalization makes use of a suffix pJ *-oro reflected as -uru/-ru /-u in Old Japanese and as *-uru /-ru in the Ryukyuan languages, which may go back to a complex form pJ *wo-ra consisting of a copula *wo- and the deverbal noun suffix *-ra (Robbeets 2015: 339–346). This is illustrated by the Old Japanese complement clause in (6b) and the relative clause in (6c).

Although the standard use of this suffix is adnominal, it can be used as a finite form marking independent sentences in both Old Japanese (see (6d)) and the Ryukyuan languages. In such cases, the insubordinated form signals the evaluative nature of the proposition and it may be accompanied by focus particles specifying the exact nature of the speaker’s reaction to such as question, exclamation, confirmation, or explanation, etc., a phenomenon known as kakari-musubi in Japanese.

The examples in (6) suggest that pJ *-ra began as a derivational nominalizer applied to verbal adjective stems to create noun stems: ‘be sad’ → ‘sadness’. In Old Japanese – as in most Transeurasian languages – noun stems could be juxtaposed to other nominals to add supplementary information, thus functioning as property nouns, i. e., nominally encoded adjectives: ‘be red’ → ‘red (oranges)’. Whereas lexical nouns were derived by adding the suffix *-ra directly to verbal stems, clausal nominalizations incorporated the copula *wo- ‘to be’. The nominalized auxiliary *wo-ra fused into a suffix and became the clausal (ad)nominalizer OJ -uru, which marked complement clauses and relative clauses in Old Japanese. The relative clauses with -uru developed one step further to mark syntactically independent sentences which added supplementary information in discourse such as question, exclamation, confirmation, explanation.

The clausal (ad)nominalizer pK *-l can be reconstructed on the basis of K -(u)l, MK -(·u/o)l. The contemporary Korean adnominalizer K -(u)l is usually called “prospective” because reference to the future is the most common meaning today. In Middle Korean, however, MK -(·u/o)l is the default imperfective adnominalizer, in essence time neutral (Martin 2002: 376). While the standard function of MK -(·u/o)l is nominal modification as in (7c), it may also be used for complementation, preceding case suffixes, such as the Middle Korean genitive marker s in (7b). The so-called “modulator” MK -·w^u/o-, which has been derived from an original copula pK *wo- ‘to be’ by Martin (1996: 13, 83, 2006: 222); sometimes appears before MK -(·u/o)l. In case the modified noun is semantically the object of the adnominalized verb, as in Example (7d), the modulator is always added; otherwise, the modulator is not obligatory. There are few examples left of lexical nominalization reflecting pK *-l such as MK kuch- ‘to stop’ → ku·chul ‘cessation’ in (7a), but it is not unlikely that originally, lexical nouns were derived with pK *-l, while clausal nominalization needed the incorporation of the copula pK *wo-, which would be in line with the behavior of the nominalizer pK *-m in Section 3.2.

As a finite marker the suffix pK *-l precedes the interrogative marker MK ·kwo expressing rhetorical, exclamatory or quoted questions (Martin 1992: 667) and it is incorporated in the ending of explicit statement K -uli, MK -(·u/o)·l i (Martin 1992: 856–857) and in the subjunctive attentive ending K -(u)la, MK -(·u/o)·la (Martin 1992: 851, 2002: 378–379), as illustrated in (7e). Whereas the subjunctive attentive is morphologically segmentable into the imperfective adnominalizer and the vocative particle a, which usually follows nouns (e. g., K palk-un tal-a [shine-ADN moon-VOC] ‘Oh shining moon!’), the explicit ending derives from the adnominalizer and a bound noun MK i ‘fact (that); that (which)’.

The examples in (7) suggest that pK *-l began as a derivational nominalizer applied to verb stems to create nouns: ‘to stop’ → ‘cessation’. Noun stems expressing properties could be juxtaposed to other nominals to add supplementary information. Whereas lexical nouns were derived by adding the suffix *-l directly to verbal stems, clausal nominalizations incorporated the copula *wo- ‘to be’. The relative clauses marked with pK *-(wo)-l developed one step further to mark syntactically independent sentences which added supplementary information in discourse such as question, exclamation, confirmation and explanation. As such, the deverbal (ad)nominalizer pK *-l does not only correspond to pJ *-ra in form and function, but the correlations also involve the incorporation of a cognate copula *wo- ‘to be’ in clausal (ad)nominalization and the further development to finite use in particular discourse contexts.

The deverbal noun suffix pTg *-rA can be reconstructed as a suffix that derived nominal and adnominal forms from verb stems such as Even da:l- ‘to be sweet, pleasant, light’ → dalra ‘sweet, tasty’, Even eden- ‘to be windy, to blow (of wind)’ → edenre ‘windy’, Even eman- ‘to snow, fall (of snow)’ → emanra ‘snow, snow-’ and Evenki langa- ‘to break a tooth’ → langara ‘toothless’ (Nedjalkov 1997: 305); see (8a).

The suffix is also used for clausal (ad)nominalization, as illustrated by the Manchu complement clause in (8b) and the relative clause in (8c). In the other Tungusic languages more recent deverbal noun suffixes of the shape pTg *-ri: have replaced the old ones on *-ra in complement and relative clauses. The forms reflecting pTg *-rA are maintained however in the finite paradigms throughout all Tungusic languages, including the Manchu example in (8d).

The examples in (8) suggest that pTg *-ra began as a derivational nominalizer applied to verb stems to create nouns: ‘to snow’ → ‘snow’. In Tungusic – as in most Transeurasian languages – noun stems expressing properties could be juxtaposed to other nominals: ‘to snow’ → ‘snow (man)’. The (ad)nominalizers were then extended to the clause level to mark clausal (ad)nominalization in complement and relative clauses. The relative clauses developed one step further to mark syntactically independent sentences. Initially, these constructions may have added supplementary information, but gradually they generalized into the default indicative ending.

The deverbal noun suffix pMo *-r can be reconstructed as a suffix that derived verb stems from nouns in Mongolic. As illustrated in (9a)–(9c), the Written Mongolian and Middle Mongolian suffix -r, with the epenthetic vowel -u- /-ü- if the stem ended in a consonant, commonly derived deverbal nouns such as MMo. andaqa- ‘swear an oath of friendship’ → andaqar ‘oath of friendship’, WMo. belčige- ‘to pasture, graze (tr.)’ → belčiger ‘pasture, grazing grounds, grass on a pasture’ and WMo. irüge- ‘bless, pray, wish well (tr. /intr.)’ → irüger ‘prayer, blessing’ or deverbal property nouns such as MMo. /WMo. amu- ‘to rest, relax; be relieved (intr.)’ → amur ‘peace, rest; easy’ and WMo. qusu- ‘to scrape, shave (tr.)’ → qusur ‘pointed’ (Poppe 1954: 49; Street 1957: 58.)

As illustrated in (9b), there are relics of clausal nominalization in the final converb in -rA that can be derived from *-r marking a complement clause plus the dative suffix in *-A and in the preparative converb on -run, which is a compound of *-r and the genitive suffix in *-un (Poppe 1954: 59, 98, 180). Examples of finite use in Mongolic are lacking, but I think that the past tense suffix -r in (9c) in Khitan, a sister language of proto-Mongolic, is related here.

The examples in (9) suggest that pMo *-r began as a derivational nominalizer applied to verb stems to create nouns: ‘to pray’ → ‘prayer’, whereby noun stems expressing properties could be juxtaposed to other nominals: ‘to relax’ → ‘easy (work)’. The (ad)nominalizers were then extended to the clause level to mark clausal (ad)nominalization. There is no evidence that these dependent clauses developed one step further to mark syntactically independent sentences in Mongolic proper, but they probably did in the para-Mongolic language of the Khitan. Therefore, I have put the dependency “independent” between brackets in Table 1.

The deverbal noun suffix pTk *-rV has a reflex in Old Turkic as a suffix that derived nouns from verb stems. It is formed with -Ar after most simple consonant stems, -Ur or -Ir after diathetic consonants stems and -yUr or -r after vowel stems. In Robbeets (2015: 158–159), I have proposed that the allomorphs -yUr /-Ur and -Ar derive from suffix strings in which pTk *-rV follows the copular verbs *u- ‘to become’ and *a- ‘to be’, respectively, while the allomorph -Ir would reflect a stem-final -i.

Even if the suffix is not included in Erdal (1991) and is generally regarded in the Turcological literature to be a participle marker (see 10b), it can be understood from derivational pairs such as OTk teg- ‘to reach, be worth (intr.)’ → tegir ‘share, value, price’, OTk. tug- ‘to be born, to rise (of sun) (intr.)’ → tugar ‘sunrise, east’ and OTk. yat- ‘to lie down (intr.)’ → yatar /yatur ‘(something) lying down, invalid’ that it had lexicalized as a deverbal nominalizer in Old Turkic; see (10a). In Old Turkic, the adnominalizer -(A)r, which is known under the label “aorist”, is still productive in relative clauses as in (10b), but it is used more often as a finite predicate as in (10c). For a discussion of the Chuvash and Yakut cognates of this suffix, see Robbeets (2015: 359–361).

The examples in (10) suggest that pTk *-rA began as a derivational nominalizer applied to verb stems to create nouns: ‘to rise’ → ‘sunrise’, whereby noun stems expressing properties could be juxtaposed to other nominals: ‘to lie down’ → ‘(thing) lying down’. The (ad)nominalizers were then extended to the clause level to mark clausal (ad)nominalization. Relative clauses developed one step further to mark syntactically independent sentences with a present continuous meaning.

The deverbal noun suffix pJ *-m can be reconstructed as a suffix that derived nominal and adnominal forms from verbal adjectives. The evidence comes from the accent class 2.5 of disyllabic nouns with a unique low-falling pitch, which is limited to the Kansai dialects. Polivanov (1924: 126) was the first to link the origins of this accent class with the loss of a final consonant pJ *-m. Vovin (1994: 250, 2008: 142–150); identifies the lost consonant as the nominalizer pre-pJ *-m in verbal adjectives often denoting colors, such as in OJ awo- B ‘to be blue/green’ and derived awo 2.5 ‘blue/green (n.)’ (< *awo- ‘be blue/green’ + *-m NML). The vowel alternation in some adjectives such as OJ kura- B ‘to be dark’ ~ OJ kuro₁ 2.5. ‘black’ and OJ sira- B ~ OJ siro₁ 2.5. ‘white’ suggests that the copula *wo- ‘to be’ may be involved in the derivation, yielding *kura-wo-m (thick-COP-NML) and *sira-wo-m (thick-COP-NML) respectively.

Clausal (ad)nominalization makes use of a suffix pJ *-om reflected as -u in Old Japanese and as *-um in the Ryukyuan languages, which may go back to a complex form pJ *wo-m consisting of a copula *wo- and the deverbal noun suffix *-m (Robbeets 2015: 361–365). Although the standard use of WOJ -u and Ryukyuan *-um is finite as in (11d), we find a few relic examples of nonfinite use. As a nominalizer, it is preserved following the negative suffix OJ -(a)z- in constructions, where the negative nominalizer -(a)zu occurs before the converb ni of the defective copula n- or before the converb site of the verb se- ‘to do’ (Vovin 2009: 760–763); see (11b). Adnominal use of -u surfaces in some lexicalizations, such as J izumi ‘(well-)spring’ which derives from *id-u mi₁ (emerge-ADN water) and OJ yo₂s-u-ka (stop-ADN-place) ‘a place to hold’ (MYS VII: 1382; BS 18; Martin 1987: 807). There are also instances where vowel-stem verbs, such as OJ mi₁te- ‘to fill (tr.)’ in (11c), take an adnominal form on -u instead of the standard -uru.

The examples in (11) suggest that before pJ *-ra started to grammaticalize, an earlier cycle of parallel grammaticalization had already taken place in Japanese, involving the derivational nominalizer pJ *-m. Similarly, lexical nouns were derived by adding the suffix *-m directly to verbal stems, whereas clausal (ad)nominalizations incorporated the copula *wo- ‘to be’. These clausal (ad)nominalization ultimately developed to mark syntactically independent sentences.

As in Contemporary Korean, the Middle Korean deverbal noun suffix MK -(·u/o)m was used to derive lexical nouns and to nominalize sentences, but in Middle Korean, the morphology of these two uses was different (Lee and Ramsey 2011: 176–177). Lexical nouns were generally derived by adding the suffix directly to verb stems, while clausal nominalizations incorporated the modulator MK -·w^u/o-, which has been derived from an original copular verb pK *wo-. This distinction is illustrated in Example (12a), but it already showed instability in the fifteenth century and fell into disuse in the sixteenth century, yielding many exceptions to this rule such as (12b).

Except for some expressions such as K wul-um swori (cry-NML voice) ‘a tearful voice’, where the nominal modification can be interpreted as an unmarked genitive case, the suffix is not used as an adnominalizer but it can appear as a marker of finiteness. As illustrated in (12c), the Middle Korean finite -(·u/o)m is always followed by the vocative particle a, which is also incorporated in the subjunctive attentive MK -(·u/o)·la, discussed in Section 2.2. In the documentary style of written contemporary Korean K -(u)m appears in main clauses without the vocative, often expressing an impersonal proposition as in (12d).

The examples in (12) suggest that a cycle of grammatical change involving the derivational nominalizer pK *-m took place in Korean, parallel to the development described in Section 2.2. Similarly, lexical nouns were derived by adding the suffix pK *-m directly to verbal stems, whereas clausal (ad)nominalizations incorporated the copula pK *wo- ‘to be’ and ultimately developed to mark syntactically independent sentences in particular discourse contexts. As such, the deverbal noun suffix pK *-m does not only correspond to pJ *-m in form and function, but the correlations also involve the incorporation of a cognate copula *wo- ‘to be’ in clausal (ad)nominalization as well as the different stages in the development to finite use and the cyclicity of the development.

The deverbal noun suffix pTg *-mA can be reconstructed as a suffix that derived nominal and adnominal forms from verb stems, such as in Evenki girku- ‘to walk’ → girkuma ‘pedestrian’, omngo- ‘forget’ → omngomo ‘forgetful, absent-minded’, tuksa- ‘run’ → tuksama ‘running’, muru- ‘to walk round, return’ → murume ‘round’ (Poppe 1955: 262, Menges 1968: 67, Nedjalkov 1997: 305) and in Udehe sigili- ‘stir’ → sigilime ‘thick soup’, kua- ‘cut, make a frame-work’ → kuaima ‘log’, xui- ‘boil’ → xuili-me ‘boiling’, zegde- ‘burn’ → zegdelime ‘burning’ (Nikolaeva 1999: 113). It is particularly frequent in the derivation of colour nouns and adjectives, such as Evk. bagda- ‘to become white, freeze’ → bagdama ‘white (adj. and n.)’, Evk. koŋno- ‘to be black’ → koŋnomo ‘black (adj. and n.)’ and pTg *pula- ‘to be red’ (in Evk. hularin ‘red’, Even hulal- ‘to become red’, hulańa- ‘red’) → Evk. hulama ‘red (adj. and n.)’, which recalls the color derivations in Japanese in Section 3.1.

In Sibe, a contemporary descendant of Manchu, the corresponding suffix -m is still productive as the citation form of verbs and is used for deriving infinitives as in (13a) as well as for marking independent clauses as in (13d).

As illustrated in (13c), most Tungusic languages use a converb suffix with distinct singular and plural forms, e. g., Evk. -mi /-mil, Nanai -mi /-mari /-meri, Ud. -mi /-mei, Olč -mi /-mari /-meri, Oroč -mi /-mai. This number distinction reflects contraction of an original clausal nominalizer pTg *-mA with the possessive-reflexive suffixes pTg *-wi singular and pTg * -wari plural, respectively (Benzing 1955: 1090; Menges 1968: 212). The presence of the possessive-reflexive marker gives the converb an inherently co-referential function, which means that it can only be used in same-subject constructions. Unlike other converb suffixes, the use of the *-mi (< *-mA-wi) converb as a nominalizer in complement clauses as in (13b) is characteristic of all Tungusic languages, an observation that also supports the proposed origin of the suffix.

The examples in (13) suggest that a cycle of grammatical change involving the derivational (ad)nominalizer pTg *-mA took place in Tungusic, parallel to the development for pTg *-rA described in Section 2.3. The suffix was extended to the clause level, leaving a trace in clausal adverbialization, which can be derived from original nominalization. Although in most Tungusic languages, the suffix either did not undergo insubordination or was replaced by newer items such as pTg *-rA, Sibe leaves a trace of the use of the suffix in independent clauses.

The deverbal noun suffix pMo *-mA alternates with *-m and can be reconstructed as a suffix that derived nominal and adnominal forms from verb stems, such as in MMo. daqa- ‘to follow (tr.)’ → daqama ‘menses’, WMo. jagura- ‘to jam, be arrested in motion’ → jagurma ‘interrupted, incomplete; that which has not reached its end; on the way (n./adj./adv.)’, WMo ulayi- ‘to get red-hot, become red (intr.)’ → ulayima ‘red, red-hot’, MMo. quri- ‘to come together (intr.)’ → qurim ‘feast’, WMo. toqo- ‘to saddle (tr.)’ → toqom ‘saddle cloth’, WMo. toγuri- ‘to go about, circle, surround (tr.)’ → WMo. toγurim ‘approximate, around; environment (n./adj)’ (Poppe 1954: 47–48, 1955: 261–262; Street 1957: 58). There are doublets, such as degerem ~ degerme ‘robbery, robber’ from pMo *degere- ‘to lift (tr.)’ in degerede- ‘to be lifted’ and toγum ~ toγuma ‘sensibleness; good behavior’ from WMo. toVu- ‘to esteem, value’ that indicate that the (ad)nominalizers *-ma and *-m share a common origin.

Janhunen (2010: 166–167) notes that the marker of the preconditional converb -mAA/n in the Central Mongolic languages, e. g., sour-maa/n (study-CONV) ‘only if you study’, can be derived from the deverbal noun suffix -m and the reflexive possessive marker -AA/n. This identification is confirmed by the fact that the final /n follows the pattern of the possessive-reflexive marker, being present in some dialects (as in Chakhar) and absent in others (as in Khalkha). Similar to the Tungusic converb in Section 3.3, the possessive-reflexive marker gives the converb an inherently co-referential function. As such, clausal adverbialization can be derived from clausal nominalization.

In Middle Mongolian texts of the thirteenth and fourteenth century, -m is the common ending for the imperfective present indicative (Poppe 1955: 261; Weiers 1966: 143–150); see (14b). This suffix is used less in Written Mongolian, where it has been replaced by -mUi or -mU.

The examples in (14) suggest that a cycle of grammatical change involving the derivational (ad)nominalizer pMo *-mA ~ *-m has taken place in Mongolic, paralleling the development for pMo *-r described in Section 2.4. The suffix was extended to the clause level, the original clausal nominalization leaving a trace in converbial sentences in the Central Mongolic languages. Ultimately, the suffix developed into an imperfective present marker in independent clauses.

The deverbal noun suffix pTk *-mA alternates with *-m and can be reconstructed as a suffix that derived nominal and adnominal forms from verb stems, such as in OTk. tut- ‘to hold, grasp, seize (tr.)’ → tutma ‘chest, coffer’, OTk. yar- ‘to split (open) (tr.)’ → yarma ‘crack’, OTk. erksin- ‘to have power or authority over (tr.)’ → erksinme ‘dominant’, OTk. tut- ‘to hold, grasp, seize (tr.)’ → tutum ‘handful’, OTk. yar- ‘to split (open) (tr.)’ → yarïm ‘half’, OTk. yil- ‘to catch on to something, to hang, to fasten (tr.)’ → yilim yï ‘creeper plant’, etc. (Erdal 1991: 316–320, 1991: 290–300). Since OTk. -mA and -(X)m have similar functions and produce near doublets such as OTk. örüm ‘something knitted’ ~ örma ‘plaited’, they probably go back to a single origin. There are some rare cases of relative clauses such as (15b) below, in which the noun but ‘leg’ is governed by the -mA form. There are no indications that this suffix has developed finite function.

Comparison of the above developments leads to the reconstruction of two distinct deverbal noun suffixes *-rA and *-mA in the presumed ancestral Transeurasian language, illustrated in Table 1. Derived nouns expressing properties were juxtaposed to other nominals, thus functioning as nominally encoded adjectives. The deverbal noun suffixes were then extended to the clause level to mark clausal nominalization in complement clauses. Relative clauses developed from cases in which a nominalized complement clause stood in apposition to a head noun. This process may already have been initiated in proto-Transeurasian, but the proto-Japanese-Korean branch underwent a common innovation in the sense that it incorporated a common copula *wo- ‘to be’ for the derivation of clausal (ad)nominalization. Relative clauses then developed one step further to mark syntactically independent sentences. Because the Japanese and Korean independent constructions still clearly mark supplementary information in discourse, their development seems to be relatively late in the course of individual linguistic history.

The common Altaic deverbal noun suffix *-rA may have been extended to the clause level and developed to mark independent sentences, before the separation of the Turkic branch. The common Tungusic-Mongolic suffix *-mA, however, separated from the Turkic nominalizer at the stage when complementation was accomplished, but the use as a marker of adverbial and relative clauses had not developed yet. The suffix *-mA specialized for converbial use in the common ancestor of Tungusic and Mongolic and after the split of both families, the native possessive-reflexive markers were added to express co-referential function in same-subject constructions. The Tungusic-Mongolic nominalizer *-mA further developed to mark syntactically independent sentences. Initially, these constructions may have added supplementary information, but gradually they generalized into the default indicative ending. The Turkic reflexes of pTk *-mA came to be used in relative clauses, but not in independent sentences. In this way, the development of finiteness on common deverbal noun suffixes in the Transeurasian languages seems to support the branching of the Transeurasian family proposed in Figure 1 on the basis of other shared innovations (see Robbeets 2005, 2015).

Figure 1:

Assumed branching of the Transeurasian family.

The comparative evidence above thus indicates that the nominalization construction began as a derivational process at the lexical level, was then extended to dependent clauses at the syntactic level, and was eventually – through a pragmatic role in discourse – extended still further to independent clauses. This is a development to which Evans’ (2007: 367) term “insubordination” can be applied because formally dependent clauses become conventionalized as independent clauses, which draw their morphological marking only from the former dependent clause.

On the basis of classical definitions such as Kurylowicz (1965: 52) or Heine and Reh (1984: 15) that describe grammaticalization in terms of an increase of grammatical status and a loss of semantic content, phonetic substance, categorial properties, and syntactic freedom, the processes of insubordination explored above can be characterized as grammaticalization. First, the gradual transition from derivation to inflection reflects an increase of grammatical status, whereby the cline is from deverbal noun affixes over participial affixes to finite affixes. Second, semantic content is gradually lost as one moves on the insubordination cline since nonfinite suffixes may change the meaning of the base, as opposed to finite markers. Third, some relativizers and finite affixes above are phonologically reduced vis-à-vis their deverbal nominal sources. Fourth, insubordination involves loss of categorial properties because it can be viewed as a gradual process of de-nominalization, in line with Malchukov’s (2004: 88– 93) view. Finally, syntactic freedom is lost because the use of a deverbal noun suffix is optional in a sentence, while the use of a finite marker is obligatory.

Since the examples given above do not involve the ellipsis of a matrix verb, they can be subsumed under “direct insubordination”: the nominalized forms are directly reanalyzed as finite forms rather than being part of an original copula construction that became verbalized and then lost its copula. This can be deduced from the observation that the examples above show virtually no trace of an eroding copula such as the final particle in (3b). And even if these nominalized forms appear as complements of finite copular verbs ‘to be’ or ‘to become’ in rare cases, historical and comparative evidence suggests that these copula represent later additions under influence of contact languages and/or to allow for maximum inflectional marking on the insubordinated nominalized form.

The Manchu finite imperfective -mbi illustrated in (16a), for instance, corresponds to the Sibe -m marking independent clauses without copula intervention in (13c) (Gorelova 2002: 232, 286–288, 441–443). This form can be derived from a reflex of the Tungusic nominalizer *-mA and the Manchu predicative copula Ma. bi, which is an obligatory component in nominal predication.

The Manchu imperfective in -mbi has a parallel in the finite past in -hAbi, which alternates with the simplex finite past in -hA and is derivable from the perfective adnominalizer Ma. -hA (< pTg *-gA) and the predicative copula bi (Gorelova 2002: 123, 232, 290–291, 444, 470, 533). These forms are illustrated in (20a) to (20c). It can be argued that the Manchu alternant in -hA reflects the original construction, which developed through direct insubordination from the perfective adnominalizer because in Northern Tungusic languages such as Even and Southern Tungusic languages such as Nanai and Udehe the corresponding finite perfect suffix is not accompanied by a copula; see (20c). This is supported by the observation that there is no alternative past form in **-kAbi, in which the copula would have followed the original allomorph of Ma. -hA. Parallelly, the Sibe -m marking independent clauses without copula intervention probably reflects the original construction. It is not unlikely that the copula constructions in Manchu, a highly siniticized language, developed under Chinese influence. The addition of the copula enables the speaker to mark the insubordinated form with the maximum of inflectional categories, such as the addition of tense marking in (19b).

There are other instances across the Transeurasian languages, such as in Mongolian and Japanese, where historical evidence suggests that copula are added to nominalized forms, only after these forms have directly developed into markers of finiteness. Although in Khalkha and other Central Mongolic languages the temporal-aspectual nominalizers -(e)x future, -deg habitual and -g.AA continuative occur either obligatorily or optionally with a copula yum in independent clauses, the use of a copula following the temporal-aspectual nominalizers seems to be rather rare in Middle Mongolian. Even though the future nominalizer -(e)x in (21c) obligatorily takes a copula in finite position, the Middle Mongolian imperfective nominalizer -KU from which it descends appeared more often than not without a copula in independent clauses; see (21b). Having made a rough count, Rybatzki (p.c.) finds that in Middle Mongolian, nominalizers indicating finite forms occur more frequently without intervention of a copula. He therefore assumes that the use of a copula is a secondary development in Mongolic, which might be influenced by Turkic languages where a similar structure of adnominalizer plus copula (e. g., OTk. -mIš ärti) is rather common and earlier than in Mongolic texts.

Note that the assumption of secondary addition of copula in Mongolic solves Mithun’s (2008: 102) puzzle that – contrary to other nonfinite forms in Khalka, which have directly extended to markers of finiteness – the use of the temporal-aspectual nominalizers as finite forms “does not appear to be the result of direct extension, however. They appear in compound tense formations, as complements of a finite verb ‘be’ or ‘become’”. The observation that these copula constructions may be secondary further contradicts Bisang’s (2016: 24) viewpoint that the use of the copula yum in Example (21c) can be seen as an indication of original clefting, whereby subsequent copula loss would lead to the development of finite future function on the nominalizer. It does not alter Bisang’s basic argumentation, however, that the development of finiteness in these examples is primarily driven by information structure.

It has also been argued that the use of nominalizers as finite forms in Japanese derives from finite copula constructions, in which the copula has been lost (Yap et al. 2004; Yap and Matthews 2008: 6–9; Horie 2008: 176–177; Bisang 2016: 28). Wrona (2011), however, provides historical evidence from Japanese, showing that some finite nominalizations such as the ones marked by -sa have never been embedded by a copula throughout the history of the Japanese language, whereas for others such as nominalizations with no and koto the so-called “stand-alone” type of nominalization is attested before the copula type. There are, for instance, no examples of a finite copula following ko₂to₂ in Old Japanese; only examples of the type illustrated in (21b) are found. Copula constructions as in the contemporary Japanese example in (21c) emerged later in Japanese history, perhaps under Chinese influence.

In sum, direct insubordination is a recurrent development across the Transeurasian languages. A serious number of insubordinated nominalizations in these languages have never been embedded by a copula and even if they have been, historical and comparative evidence frequently suggests that the copula represents a secondary addition. It is not unlikely that copula were added under influence of contact languages and/or to allow for maximum inflectional marking on the insubordinated nominalized form.

In other parts of Asia clausal nominalization in constructions with a copula is, nevertheless, a major source for developing new finite constructions. Many Sinitic languages use focus constructions consisting of a nominalizer plus a copula verb and dropping the copula then paves the way for finite stance constructions. The Mandarin shi … de focus constructions, for instance, consist of a copula shi and a nominalizer de, whereas the finite stance construction appears without the copula (Yap and Matthews 2008: 20). Similar processes are found in the Siberian area, for instance in Yukaghir (Malchukov 2013: 192–195). In Kolyma Yukaghir, the suffix -l is used to derive action nouns, anterior participles and indicative finite forms in subject focus constructions, as illustrated in (22). The intransitive subject ‘I’ in (22b) takes a focus marker -ek, which is also used to mark nominal predicates, thus pointing to an origin as a copula-like form. As such, the example in (22b) can be derived from a cleft-like construction “It is me sitting”.

As demonstrated by several scholars (e. g., Matisoff 1972; Noonan 1997; Bickel 1999; DeLancey 2011), nominalizer plus copula constructions are also the major source of finitization in the Tibeto-Burman family. Finite verb forms in Modern Tibetan such as -pa yin, -pa red, for instance, transparently reflect a copula construction, consisting of a nominalizer -pa followed by an equational copula. As such, constructions become more opaque, the copula may develop into a semantically empty final particle as in Example (3b).

As opposed to the situation in the Transeurasian languages, final particles and focus markers leaving a trace of an eroding copula prevail in languages that develop finiteness from nominalization in construction with a copula. As such, the inherent mechanism for developing finiteness in the Transeurasian family is demonstrably distinct from that used in the Sinitic, Yukaghir or Tibeto-Burman languages.

This does not mean, however, that the frequency and concentration of the process of “direct insubordination” in the Transeurasian languages by itself should be taken as evidence of common ancestorship. Mithun (2008: 102) finds that processes of direct insubordination are undoubtedly more common crosslinguistically than has hitherto been recognized, partly because traditional studies have focused on isolated sentences rather than on longer stretches of discourse. Given the frequency of the process crosslinguistically, sharing “direct insubordination” may be the result of universal principles in linguistic structuring rather than a shared linguistic history. Moreover, as far as the concentration of the process is concerned, direct insubordination seems to be particularly concentrated in the wider Transeurasian contact zone, appearing among others in Uralic languages, Eskimo and Nivkh.

In Nivkh, for instance, the deverbal action noun and infinitive suffix -d’ has developed over participial use into a finite form -d’, as illustrated in Example (23), which illustrates both the nonfinite and finite use of the suffix. Since Nivkh has had close contact with Tungusic languages situated to the east such as Olcha, Oroch, Orok, Nanai, Udehe, Negidal and Evenki, it is not unlikely that the development was induced by language contact.³ Note that Anderson (2006: 25) finds that “the features of the Siberian linguistic macro-area cluster around those of the Northern Tungusic languages” and refers to the Tungusic languages as “vectors of diffusion”. However, rather than being a case of “direct insubordination”, Gruzdeva (2016: 196) attributes the development to the lexicalization of a modally marked form of the copular verb ha- ‘do so’. When the modally marked copula was lexicalized into a modal particle, -d’ was reanalyzed as a finite form.

Even if direct insubordination can diffuse from one language to another as it probably did in the Siberian area, there are indications that the direct insubordination shared across the Transeurasian languages is genealogically motivated. The main indication is that the process of direct insubordination is shared on formally corresponding suffixes, such as the common forms *-rA and *-mA above. For the assumption that globally shared grammaticalization – or any input and output of grammatical change shared on formally corresponding morphemes – is a strong indication of genealogical relatedness, I refer to Robbeets (2013).⁴ Within a borrowing scenario, the global correspondences of *-rA should be explained as follows: first, the recipient language, say Mongolic, borrowed nominalized verbs of the model language, say Turkic. Second, the contact was intensive in such a way that the underlying verb roots were borrowed as well. Third, following a massive borrowing of verb roots, Mongolic speakers saw the relation between the borrowed verb base and the borrowed nominalization and started to apply *-rA productively in their own language. At this stage, Mongolic had globally copied the Turkic suffix, but the borrowing of the outcome of grammatical change was yet to come. Later in the history of Turkic, the nominalizer developed into a marker of finiteness. Fifth, Mongolic speakers, drawing an equivalence between their earlier borrowed nominalizer *-rA and the Turkic adnominalizer, fill the gap by borrowing the finite function. This process has to be repeated pairwise four times between Mongolic and Tungusic, Tungusic and Koreanic and so on, until it reaches the Japonic languages. Needless to say, such a complicated borrowing scenario, where one has to invoke chance in case after case, is highly unlikely.

Moreover, direct insubordination is not restricted to a single set of formally corresponding suffixes like those yielding the reconstruction of pTEA *-rA. In this article, I have also discussed the development for the nominalizer *-mA and in Robbeets (2015), I trace the process on at least five formally corresponding suffixes. As such, the recurrence of indirect insubordination on formally related suffixes is also pointing towards common ancestorship. It is known that languages tend to renew their formal encodings in cyclic processes of grammatical change while trying to maintain their inherited grammatical categories. Therefore, newly inserted grammatical items are expected to develop along shared conceptual pathways to restore old categories (Heath 1998; Aikhenvald 2013). Consequently, genealogically motivated grammatical change is expected to recur in different forms at various points in time, while contact-induced grammatical change is expected to be restricted to a single formal encoding (or to a very limited number of encodings) during the period of contact. Therefore, I take instances of shared grammatical change whereby the specific pathway recurs in more sets of formally related morphemes as an indication of common ancestorship.