18. Complementation and Quotation
A complement supplies an argument of a predicate. In Old Japanese it can present words, the content of a thought, or an event that someone sees, hears or knows about. The same predicate can take more than one kind of complement. The analysis depends on the form at the end of the embedded expression, any particle that follows it, and the relation of the expression to the surrounding clause.
Examples use romanization with the Old Japanese consonant p. Hyphens divide the forms discussed in the gloss. The cited verse excerpts give the form in Japanese editorial spelling as well as in romanization.
Predicate classes and the complements they license
A complement normally precedes its selecting predicate. A nominal complement bears the case or other marking appropriate to its function in that predicate's clause. A quotation generally ends in to, with the verb of speaking or thinking outside it. Material belonging to the surrounding clause can intervene between a complement and its selecting predicate. Identify the end of the complement and its relation to the surrounding predicate before assigning the preceding words to one clause or the other.
| Predicate use | Complement pattern | Interpretation |
|---|---|---|
| Speaking, as with ip-u ‘say’ | Quoted material followed by to | The words or content represented as spoken |
| Thinking, as with omop-u ‘think’ | Represented thought followed by to | A judgment, supposition or intention attributed to the thinker |
| Hearing, as with kik-u ‘hear’ | An expression in to when what is heard is a report | The content learned or heard from speech |
| Seeing or hearing an occurrence, as with mi-ru ‘see’ or kik-u ‘hear’ | A nominal expression with an adnominal clause, with wo when marked as an object | The occurrence or entity perceived |
| Knowing, as with sir-u ‘know’ | A nominal expression, with wo when marked as an object | An entity, matter or occurrence known |
The table gives construction types, with representative predicates. The relevant sense of the predicate matters. Hearing a bird call and hearing that a bird has called involve different relations between what is heard and what happened. The same lexical predicate can also take an ordinary noun phrase.
| Form | Gloss | Translation |
|---|---|---|
| tori wo kik-u (constructed) | bird OBJ hear-FIN | ‘Someone hears a bird.’ |
| tori no nak-u wo kik-u (constructed) | bird GEN cry-ADN OBJ hear-FIN | ‘Someone hears a bird calling.’ |
| tori nak-u to kik-u (constructed) | bird cry-FIN QUOT hear-FIN | ‘Someone hears that a bird is calling.’ |
In the second example, tori no belongs with ‘cry’, and object wo follows the expression for the calling. In the third, to marks the content of the report. The report can concern an audible event even when the hearer learns about it from another speaker. The distinction is between the two constructions, so the English choice between hear a bird call and hear that a bird calls should follow the interpretation of the Old Japanese.
A verb of speaking can also occur with a noun for a word or matter, and it can be used when the content is left to the context. The same applies to a thought predicate when what is thought is already established. The lexical possibility of taking a complement gives a place to look for one. An identified complement still needs a form and an interpretation supported by the clause.
Nominalized and non-finite complements
An adnominal predicate can modify an overt noun, or an adnominal expression can occupy a nominal position with the understood content supplied by the context. Both patterns can occur as objects. When wo follows the entire expression, its scope is the object of the higher predicate. A noun or particle inside that expression can belong to the embedded predicate instead.
| Pattern | Where the object marker attaches | Interpretation |
|---|---|---|
| Adnominal clause + noun + wo | After the noun and its modifier | The entity or matter denoted by the noun, as specified by the clause |
| Adnominal expression + wo | After the adnominal expression | An understood entity, event or circumstance that the expression identifies |
| Form | Gloss | Translation |
|---|---|---|
| tuki no id-uru koto wo sir-u (constructed) | moon GEN emerge-ADN matter OBJ know-FIN | ‘Someone knows about the rising of the moon.’ |
| tuki no id-uru wo mi-ru (constructed) | moon GEN emerge-ADN OBJ see-FIN | ‘Someone sees the moon coming out.’ |
In the first example koto supplies a nominal head with the meaning ‘matter’ or ‘affair’. A clause modifying such a noun can give the content of the matter. The noun still occupies a nominal position and takes the case marking of that position. In the second example the adnominal expression itself supplies the object. An English translation with the fact that or the one that may make the interpretation explicit. Neither translation establishes an additional Old Japanese word in the expression.
A particle immediately after an adnominal form can also have a function at the boundary between clauses. In particular, a sequence ending in wo requires attention to the next predicate. If the predicate takes that expression as its object, an object analysis accounts for the relation. Where the preceding expression sets a circumstance or supplies a contrasting proposition for the continuation, the connective use of wo is relevant. The sequence of forms alone leaves that choice open. Likewise, the no after tuki in the examples above is internal to the embedded clause. Its position and its relation to ‘emerge’ distinguish it from a nominal head at the end of the complement.
Inflection at the complement boundary
The adnominal form is easy to identify when the verb gives it a distinct ending. For verbs whose conclusive and adnominal forms coincide, the following noun or particle and the expression's function supply the evidence. The continuative form has a different distribution. It can join a following element in a verbal complex or link predicates in a sequence.
| Verb | Continuative | Conclusive | Adnominal |
|---|---|---|---|
| id- ‘emerge’ | ide | idu | iduru |
| nak- ‘cry’ | naki | naku | naku |
| mat- ‘wait’ | mati | matu | matu |
Thus iduru in the object example is recognizably adnominal. The ending in naku in the hearing examples cannot by itself distinguish an adnominal object from the predicate of represented speech. A continuative form can retain its own objects or other dependents. Whether it heads a dependent clause, participates in a complex predicate, or continues into a succeeding clause depends on the whole construction. The conclusive and adnominal forms in the table are inflectional labels. In particular, adnominal inflection can occur in a clause without an overt noun, including under the conditions that govern focus.
Quotative marking with speech and thought predicates
Quotative to follows the material represented as spoken or thought. The governing speech or thought predicate follows the quotation in the ordinary order. The material before to can be as small as a designation or as large as a clause. A quoted clause can have its own subject, time reference, modality, interrogative marking or imperative. Interpret those features within the quotation before assigning a force to the utterance as a whole.
| Form | Gloss | Translation |
|---|---|---|
| “ware pa yuk-amu” to ip-u (constructed) | 1SG TOP go-PROSP QUOT say-FIN | ‘Someone says, “I will go.”’ |
| “ta ka k-uru” to ip-u (constructed) | who FOC come-ADN QUOT say-FIN | ‘Someone says, “Who is coming?”’ |
| “mat-e” to ip-u (constructed) | wait-IMP QUOT say-FIN | ‘Someone says, “Wait.”’ |
| tuki id-u to omop-u (constructed) | moon emerge-FIN QUOT think-FIN | ‘Someone thinks that the moon is coming out.’ |
The question contains interrogative ta ‘who’ with focus ka, and the embedded predicate takes the adnominal form kuru. The imperative in the next example comes from the words attributed to the speaker. The outer predicate in both examples reports an utterance. In the thought example, English indirect discourse is natural. To can accompany a representation of someone's words or the formulation of a thought, so the particle leaves open how closely the wording corresponds to an utterance. In the first example, the interpretation of ware starts from the speaker represented in the quote. In a report rendered indirectly, the translator may shift person or other expressions to suit the reporting context.
The same boundary marking is available for a projected act that gives the aim of another action. The fragment in the following verse expresses an intention, and the continuation describes waiting for the moon. The waiting is a step toward departure by boat. The context therefore supports a purpose interpretation of the expression ending in to. (Keichū 1690, book 1, poem 8)
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 「船乗りせむと」 | puna-nori se-mu to | boat-boarding do.IRR-PROSP QUOT | ‘Intending to board the boat’ |
In a different verse, to follows a clause with the negative modal zi. The wider passage presents the speaker's boast about having no equal in enduring hardship. The negative modal belongs to the represented content. (Keichū 1690, book 5, poem 892)
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 「人はあらじと」 | pito pa ara-zi to | person TOP exist.IRR-NEG.MOD QUOT | ‘That there would be no one [else]’ |
A purpose interpretation depends on the action and its projected outcome. An intention reported under ‘think’ can contain the same prospective marking and serve as that predicate's content. A prospective form before to therefore supports an interpretation of the projected act, while the outer predicate and its context establish the act's relation to the surrounding clause. The analysis also has to distinguish quotative to from nominal uses of to, such as marking a companion. The meaning of the expression before the particle and the requirements of the following predicate matter when the preceding material is short.
Modern quotation marks are editorial. In the episode of the heavenly rock cave, the narrative contains speech exchanges that translations can set out as dialogue. The dialogue provides context for recognizing a speech event. To identify an Old Japanese quotation boundary, use the wording and particles of the reading being analyzed. This distinction matters particularly when consulting translations of the Kojiki and Nihongi alongside a commentary on the Japanese text. (Chamberlain 1882, § “The Door of the Heavenly Rock-Dwelling”; Aston 1896, vol. 1, book I, § on the rock cave; Motoori 1822, § on the heavenly rock cave)
Complement subjects and the interpretation of omitted arguments
An overt embedded subject belongs to the embedded predicate even if the higher clause has no expressed subject. In an adnominal clause, subject marking with no or ga can have the form of genitive marking. The reading depends on what the marked phrase combines with. A noun that supplies the participant of an embedded predicate has a subject interpretation. A noun that determines another noun in the same phrase can have a possessive interpretation. In an object such as tuki no iduru wo, these dependencies occur at different levels. The moon is the subject of ‘emerge’, and the whole expression in wo is the object of ‘see’.
| Form | Gloss | Translation |
|---|---|---|
| kimi ga k-uru wo mi-ru (constructed) | lord GEN come-ADN OBJ see-FIN | ‘Someone sees you coming.’ |
| ko ni “mat-e” to ip-u (constructed) | child DAT wait-IMP QUOT say-FIN | ‘Someone says to the child, “Wait.”’ |
In the first example, kimi ‘lord, you’ supplies the one who comes. In the second, ko ni supplies the addressee of ‘say’. The command is understood as addressed to the child, while its own subject is unexpressed. The two relations can involve the same person. The dative belongs to the reporting predicate, and the interpretation of the command comes from the speech situation. In a quoted declarative a first-person expression instead ordinarily picks out the represented speaker, and a third-person participant can be identified within the quoted content or from the discourse.
Subjects and objects can be left unexpressed when their reference can be recovered. In an intention about what the thinker proposes to do, the thinker is a natural interpretation of an omitted actor. In a report of an independently observable event, the surrounding discourse can supply a different actor. Parallel clauses and continued topics can support a common referent. A change of participants, an overt addressee or the lexical requirements of an embedded predicate can favor a different one. Some passages leave the referent uncertain.
In the following verse fragment, kimi is understood as the object of ‘wait’. The wider verse supplies the speaker as the one who waits and longs. A translation can recover both participants even though the fragment expresses only the person awaited. The to expression contributes the speaker's purpose or preoccupation in the context of longing. (Keichū 1690, book 4, poem 488)
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 「君待つと」 | kimi mat-u to | you wait.for-FIN QUOT | ‘As I wait for you’ |
Recovering a referent in a translation and identifying an overt dependent in the Old Japanese are separate tasks. An English subject supplied to make the translation idiomatic can represent the higher actor, the embedded actor or a participant inferred from the surrounding discourse. Agreement between the English subjects therefore gives no additional evidence about the Old Japanese construction.
Control, raising and the limits of structural diagnostics
A control analysis treats an unexpressed embedded subject as dependent for its reference on an argument of the higher predicate. In subject control that argument is the higher subject. In object control it is a higher object or other selected participant, such as the addressee of a directive. Raising instead treats the apparent higher argument as receiving its semantic role from the embedded predicate. These analyses make different claims about the higher predicate and the distribution of arguments. They require more evidence than a translation with an infinitive or the recovery of a shared participant.
| Form | Gloss | Translation |
|---|---|---|
| yuk-amu to omop-u (constructed) | go-PROSP QUOT think-FIN | ‘Someone intends to go.’ |
| tuki id-u to omop-u (constructed) | moon emerge-FIN QUOT think-FIN | ‘Someone thinks that the moon is coming out.’ |
In the first example the reading as an intention favors identifying the one who goes with the thinker. The embedded verb has prospective marking and is followed by to. The same thought predicate can take content with its own overt subject, as in the second example. This establishes a distinction between the intention and judgment readings. A claim of obligatory control would require evidence that a construction with the relevant reading fixes the reference of the omitted actor, even where the discourse makes a different actor available. The common interpretation of the intention example alone supports a statement about how the actor is understood in that context.
A directive likewise gives a strong semantic reason to identify the person expected to act with the addressee. In the quotative directive above, the command supplies this interpretation. Structural object control would require evidence that the higher participant enters a grammatical dependency with the embedded subject. An independently formed quotation can give the same interpretation through its own speech roles. An overt addressee next to ‘say’ and a translated infinitive such as ‘told the child to wait’ are compatible with either reasoning about the English rendering. The Old Japanese marking in the example supports the analysis as a reported command.
For raising, first establish the relation of the potentially raised phrase to the embedded predicate. Then ask whether the higher predicate gives that phrase an independent role and whether case or placement establishes its membership in the higher clause. An initial noun phrase can also be the subject of the embedded clause or a topic interpreted over a wider stretch of discourse. In the example with the rising moon and ‘see’, wo is at the edge of an adnominal expression, and the moon is interpreted within that expression. English sees the moon come out gives no independent basis for moving the moon into the higher clause in the Old Japanese analysis. An epistemic reading of a higher expression can motivate a raising hypothesis, but the reading alone cannot locate the noun phrase.
Useful comparisons preserve the same construction and reading while varying the disputed property. A switch from an intention to a judgment may change the construction being tested. An attested overt embedded subject establishes its availability in that use. A case marker on the whole embedded expression helps identify its external function. A case marker internal to it helps identify an internal relation. Where the inflection is syncretic, the subject is omitted and there is no decisive case marking, the result can be a secure interpretation of who does what with the structural relation still open. The written record also offers limited evidence for contrasts that would require speakers' judgments about possible but unattested alternatives.
Complex predicates and the boundary with clause chaining
Two verbal forms in succession can describe parts of a single predication or link events in a sequence. An inflecting auxiliary can follow a form selected by that auxiliary. A lexical verb in the continuative can also combine with another verb in a compound, and a clause in the continuative or in -te can link to what follows. In a complement construction, one predicate selects an expression as an argument. In clause chaining the earlier predication instead advances the account or supplies a circumstance, such as manner or prior action, for the later one.
| Form | Gloss | Translation |
|---|---|---|
| mi-tamap-u (constructed) | see.CONT-HON-FIN | ‘The honored person sees.’ |
| tori naki-watar-u (constructed) | bird cry.CONT-cross-FIN | ‘A bird calls as it goes across.’ |
| tori naki-te watar-u (constructed) | bird cry.CONT-LINK cross-FIN | ‘A bird calls and goes across.’ |
In the honorific example, tamapu in its auxiliary use marks regard for the subject of ‘see’. The continuative mi combines with it, and the finite ending occurs on the auxiliary. When tamapu has its lexical sense of giving, its own participants matter to the analysis. The form of the second verb must therefore be read together with its use in the construction.
In naki-wataru, the continuative naki and the following verb can express a single extended action in which motion and calling are combined. In nakite wataru, -te expressly links the first predication to the second. The context can make the two actions simultaneous or put the calling before the crossing. Both examples can share a subject. Subject sharing is thus compatible with a complex predicate and with a clause chain. Final inflection on the last verb also occurs in both configurations.
“Serial verbs” can describe close combinations of verbs provisionally. A more specific analysis asks whether each verb retains a distinct lexical predication, how the participants are distributed, where inflection and any joining particle occur, and whether a dependent can be assigned to one predicate or has scope over the whole combination. A second verb with a stable honorific or aspectual function and a fixed attachment form supports a complex predicate or an auxiliary construction. Distinct temporal settings or independently established subjects support two clauses. Shared time, shared subject and uninterrupted spelling can occur under either analysis. A bare continuative with an independently interpretable first predication deserves particular attention because overt -te marking is only one way to continue into a following clause.
A sequence can contain both complementation and clause chaining. A quotation in to can be selected by a speech verb whose own continuative form links the report to a later event. Conversely, a to expression of aim can modify an action such as waiting, as in the departure verse. In each case the relation of the expression in to to the immediately governing predicate must be established before relating that predicate to the rest of the sentence. (Keichū 1690, book 1, poem 8)
How Mova handles proposed complement and quotation analyses
Mova has no level-2 parser model for Old Japanese (ojp) and holds no annotated Old Japanese treebank. No treebank baseline has been measured. The multilingual parser tools/mova-parse uses language spans from tools/langseg and needs the corresponding language model to parse a span. The script and lexicon coverage of tools/langseg does not identify Old Japanese. Scripts outside its supported coverage are classified as other.
When tools/mova-parse encounters tokens assigned to ojp without a model, it reports the missing model in the # parse comment. It leaves LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL as _ and gives the tokens Parse=None in MISC. Old Japanese tokenization, tagging, lemmatization and parsing are unavailable. The frames and pragmatics components read English trees and have no Old Japanese trees to read.
Squid reads the grammatical points in this chapter as claims. Formalizable points, including conditions on complement boundaries, quotative marking and verbal attachment, wait in its queue. The chapter's rules are unimplemented. The first points needed for implementation are those that support a tokenizer, a tagger and a Universal Dependencies treebank baseline.
Sources for this chapter
- Chamberlain, Basil Hall. 1882. A Translation of the “Ko-ji-ki,” or “Records of Ancient Matters.” Transactions of the Asiatic Society of Japan 10, supplement. — public domain; not held in the library, to verify
- Aston, William George. 1896. Nihongi: Chronicles of Japan from the Earliest Times to A.D. 697. 2 vols. Supplement 1 to Transactions and Proceedings of the Japan Society, London. London: Kegan Paul, Trench, Trübner & Co. — public domain; not held in the library, to verify
- Keichū. 1690. 萬葉代匠記 [Man’yō daishōki, commentary on the Man’yōshū]. First manuscript recension, shokō-bon. — public domain; not held in the library, to verify
- Motoori, Norinaga. 1822. 古事記傳 [Kojiki-den, commentary on the Kojiki]. 44 fascicles. Woodblock edition issued in installments from 1790 to 1822. — public domain; not held in the library, to verify
This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/ja/ojp/18-complementation-and-quotation/