Mova0.6 pre-release

Old Javanese: the red and white stripes of Majapahit 23. Word order, topic, focus and ellipsis

Old Javanese allows a predicate to precede its arguments and allows an argument or a setting expression to precede the predicate. The order helps readers follow what a clause is about, what it adds, and how it connects to the surrounding text. Voice, particles, and the form of noun phrases help determine which readings an order permits. In verse, the same words must also fit the metre.

Clause order variation in prose and verse

A verbal predicate can open a clause, with an overt argument following it. An argument can also come before the verb. This alternation occurs even with an intransitive verb, where switching the order leaves the participant’s role unchanged. A postverbal argument can identify a participant already under discussion or introduce one. A preverbal argument can continue an established topic, select a new topic, or carry contrastive focus. The surrounding passage determines which reading is appropriate.

FormGlossTranslation
tĕka saṅ ratu (constructed)arrive HON ruler‘The ruler arrived.’
saṅ ratu tĕka (constructed)HON ruler arrive‘The ruler arrived.’

In the glosses, HON marks an honorific nominal marker, DET a determiner, PTCL a particle, 3.HON an honorific third-person pronoun in the reading illustrated, and UV an undergoer-voice form. Examples are presented in romanization.

With a transitive predicate, establish the voice and the form of each nominal expression before assigning participant roles. The first noun phrase can be the undergoer. An actor expressed in a phrase headed by deniṅ ‘by’ has a different grammatical form from an unmarked noun phrase. Moving the undergoer in the following pair leaves both roles intact.

FormGlossTranslation
tinulis deniṅ wiku ikaṅ surat (constructed)write.UV by monk DET letter‘The letter was written by the monk.’
ikaṅ surat tinulis deniṅ wiku (constructed)DET letter write.UV by monk‘The letter was written by the monk.’

Preposing the letter makes it available as the point of departure. A context that contrasts the letter with another object can instead make the same phrase focal. The undergoer-voice form and the actor phrase support the same role assignment in either context. A translation that uses the same English order for both clauses can preserve their propositional content while leaving the difference in presentation implicit.

An initial particle or an initial setting expression can precede the predicate as well. Thus a clause beginning with a place expression followed by a verb and an argument still has the predicate before that argument. Conversely, a constituent within a noun phrase can precede a verb because the entire noun phrase does. Determine the boundaries of the constituents before comparing orders. A sequence of nouns, particles, and verbs can contain a nominal dependent or a clause boundary where a list of word classes alone would suggest a single clause.

The same diagnostics apply to prose and verse. Prose contexts often provide several successive clauses in which to track the position and expression of a recurrent participant. In a verse text, first determine whether the words on either side of a line ending belong to the same construction. Then compare the resulting clause with the orders available outside that line. A line-initial noun phrase may occupy the middle of its clause, and a clause can start within a line.

Preposed topics and frame-setting expressions

A topic identifies the entity or matter about which the rest of the clause adds information. An initial topic can itself be a syntactic argument, as saṅ ratu is in saṅ ratu tĕka ‘the ruler arrived’ in an account already following the ruler. A passage may also move from one established participant to another and name the new topic at the start of the clause. The relevant clues are the previous mention of the referent, the continuity or change of subject matter, and what the next clause says about it. Initial position alone leaves these possibilities open.

A frame-setting expression specifies the time, place, or circumstances within which a clause or a stretch of text holds. Its relation is to the event or assertion as a whole. In a dated inscription, the opening specification can date the record or the act it documents. The act and its participants follow within that temporal frame. A date formula at the opening of an inscription has this wider function even when it stands apart from the syntax of the following clause (Brandes 1913, Taji inscription, opening date formula).

FormGlossTranslation
riṅ wana tĕka saṅ wiku (constructed)in forest arrive HON monk‘In the forest, the monk arrived.’
tĕka saṅ wiku riṅ wana (constructed)arrive HON monk in forest‘The monk arrived in the forest.’

The initial place phrase in the first example readily supplies a setting for the arrival. If the next clause describes another event in the same setting, the frame can extend to it. In the second example, the phrase readily specifies the location of the arrival itself. Either order can carry a contrast between locations if the passage supplies alternatives. A following change of time or place gives evidence that an earlier frame has ended. Continued narration without a new expression leaves the scope of the earlier frame to be inferred from the events described.

More than one expression can precede a predicate. In the constructed pattern riṅ wana saṅ wiku tĕka (in forest HON monk arrive), ‘In the forest, the monk arrived,’ riṅ wana can set the frame while saṅ wiku supplies the topic within it. Each expression has its own relation to the clause. A place expression can also be the new information itself, as in an answer specifying where an event occurred. The distinction between a setting and a focused location follows from the question under discussion, the scope of the phrase, and the way the passage proceeds.

A repeated name or full noun phrase does not have a single discourse function. It can reestablish a participant after an interval, identify one where several are available, or contrast two participants. Likewise, a referent can become a topic on its first mention if the clause establishes it as what the passage will discuss. The presence of an honorific or a determiner contributes to the form and interpretation of the phrase. Topic status depends on what the passage does with the referent.

Focus and identificational constructions

Focus supplies the part of an assertion that resolves a question, corrects a previous statement, or selects among relevant alternatives. It can fall on a participant, a place or time, a predicate, or a whole event. An answer to ‘Who arrived?’ places focus on the person who arrived. An answer to ‘What happened?’ can introduce the arrival and its participant together. The same predicate-first order can serve both purposes.

FormGlossTranslation
sapa ta saṅ tĕka? (constructed)who PTCL HON arrive‘Who is the one who arrived?’
tĕka saṅ wiku (constructed)arrive HON monk‘The monk arrived.’
saṅ wiku ta saṅ tĕka (constructed)HON monk PTCL HON arrive‘It was the monk who arrived.’

The second example can answer the question with an ordinary verbal clause. The third singles out saṅ wiku and identifies that person with saṅ tĕka, ‘the one who arrived’. The arrival is treated as given in this exchange, and the identity of the person supplies the answer. The identifying expression and the nominal expression containing the verb together support the cleft-like translation.

Old Javanese can identify one nominal expression with another without an overt copula. When one side is a headless nominal clause such as saṅ tĕka, its internal predicate describes the participant being identified. This creates two distinct relations. Within saṅ tĕka, arriving is predicated of the person referred to. In the larger construction, that person is identified as the monk. An ordinary preverbal argument followed by its verbal predicate has only the first type of predication. Thus saṅ wiku tĕka ‘the monk arrived’ (constructed, HON monk arrive) can have focus on the monk in context without taking the structure of an identificational clause.

The particle ta occurs both in identificational environments and with an ordinary verbal predicate. Compare tĕka ta saṅ wiku (arrive PTCL HON monk), ‘The monk arrived’ (constructed), with the identificational answer above. In the verbal clause, ta follows tĕka. Its presence there supplies no identifying expression of the kind found in the answer. Analyze the particle together with its position, the forms it accompanies, and the discourse context. In particular, the adjacent word need not be the sole focus of the clause.

An identificational reading is strongest when the passage independently establishes an event or description and asks, implies, or corrects the identity of the participant. In a passage with no such background, the same nominal material can introduce a person by description. The relation between the nominal expressions can also be appositional. To distinguish the two, ask whether the clause asserts the identity or whether both expressions jointly name one participant in a separate assertion. The location of the main predicate and the role of any following material help resolve the difference.

Dislocation, apposition and afterthought

A dislocated expression stands at a clause edge and refers to a participant represented in the clause itself. In left dislocation, the expression comes first and sets up the referent. The clause can resume it with a pronoun. In right dislocation, the clause gives the predication first and the expression at the right edge then names or clarifies a participant. An overt resumptive expression gives direct evidence for dislocation. With an unexpressed argument, both an integrated reading and a dislocated reading may be possible.

FormGlossTranslation
saṅ ratu, tĕka ta sira (constructed)HON ruler arrive PTCL 3.HON‘The ruler, he arrived.’
tĕka ta sira, saṅ ratu (constructed)arrive PTCL 3.HON HON ruler‘He arrived, the ruler.’
tĕka saṅ wiku, guru niṅ ratu (constructed)arrive HON monk teacher of.DET ruler‘The monk arrived, the ruler’s teacher.’

In the first example, sira represents the ruler inside the clause, and the initial noun phrase names that referent outside it. Compare the integrated order saṅ ratu tĕka, in which the noun phrase itself supplies the argument. In the second example, the clause can be interpreted through sira before the final noun phrase arrives. The final phrase can clarify which person the pronoun denotes. This reading depends on the two expressions having the same referent. A second expression referring to someone else needs another analysis.

Apposition puts two descriptions of the same referent together. In the third example, ‘the ruler’s teacher’ describes the monk. The two descriptions can be read together as the expression for the person who arrived. In a suitable context, the second can instead be an afterthought added after the arrival has been asserted. An afterthought supplies a clarification, correction, or further description after the preceding clause is already interpretable. Apposition describes the relation between the two nominals. Afterthought describes the way the second is presented. The two descriptions can therefore apply to the same passage.

The commas in these constructed examples indicate the intended readings. In an edition, a comma is evidence for the editor’s division of the text. The analysis also needs evidence from the construction. Check whether both expressions refer to the same person, whether one expression already fills the argument position, and whether a following clause treats the last phrase as the beginning of a new unit. In a passage that admits several divisions, a postverbal noun phrase can be an ordinary argument or the beginning of a new predication. An edge position or a printed pause leaves that choice open.

Omission of arguments and predicates

Unexpressed arguments

Old Javanese permits a clause to leave a participant unexpressed. The verb can supply information about voice and valency, while a preceding clause, the situation in dialogue, or a larger stretch of narrative supplies the referent. This is the basis for reading many clauses with an unexpressed subject or other argument. There is no need for an overt pronoun in each clause that continues to describe the same participant. Where a clause can have a generic or unspecified participant, its interpretation may require no particular antecedent.

FormGlossTranslation
tĕka saṅ wiku. lumaku ta riṅ wana. (constructed)arrive HON monk. walk PTCL in forest‘The monk arrived. Then he walked in the forest.’
tinulis ikaṅ surat. (constructed)write.UV DET letter‘The letter was written.’

In the first example, the monk is the available referent for the walker in the second clause. English supplies ‘he’ in the translation. The Old Javanese clause expresses the walking and its setting. In a longer passage, a different participant could supply that argument if an intervening expression or change of scene established the referent. A nearby noun phrase is a candidate antecedent when its role and the event make it plausible.

In the second example, the writing has an actor, but the clause gives no actor phrase. The interpreter can leave that participant unspecified or recover a particular writer from context if one is established. The omission of an actor in an undergoer-voice clause thus differs from a sequence in which the identity of an unexpressed subject is readily recovered. Both have an unexpressed participant. Only the first example supplies a specific antecedent in the text shown.

An expressed actor can likewise occur with an unexpressed undergoer where the predicate and context support that reading. For example, after a question about whether the ruler saw a particular letter, tumon saṅ ratu (see.ACT HON ruler), ‘The ruler saw it’ (constructed), can leave the seen object implicit. ACT marks the actor-oriented verb form. The object in the English translation follows from the stated context. With a different preceding question, the referent could change. With insufficient context, a definite English object may be more specific than the passage warrants.

When identifying an omission, distinguish a participant licensed by the predicate from a phrase that could simply have been added. A verb of motion can be used without an expressed destination, for example. An analysis can record the movement and its participant even if the narrative supplies no destination. For each possible gap, first determine the predicate’s construction and the roles of the overt expressions. Then consider whether the context identifies a participant, leaves one unspecified, or supplies too little information to decide. This order of analysis is also useful when more than one participant is unexpressed.

Unexpressed predicates

A short reply can consist of the constituent that answers a question. The predicate is recoverable because the question supplies it. In the following exchange, the answer names the person and leaves the arrival understood.

FormGlossTranslation
sapa ta saṅ tĕka? (constructed)who PTCL HON arrive‘Who is the one who arrived?’
saṅ wiku (constructed)HON monk‘The monk.’

Parallel constructions can also prompt a reading in which an earlier predicate supplies the interpretation of later material. The evidence is strongest when the later segment presents a new term in the same relation as a term in the first clause, and the text makes a repeated event or property available. A phrase at the end of a clause can instead be coordinated with an earlier phrase or added in apposition. That reading requires only one predication. Identify what is coordinated or contrasted before deciding that a second predicate is understood.

A nominal expression can itself function as a predicate, and a place expression can predicate a location. For instance, saṅ wiku riṅ wana (HON monk in forest), ‘The monk is in the forest’ (constructed), permits a locative predication. An English translation with ‘is’ does not provide evidence for an omitted Old Javanese verb. A preceding clause about walking could instead support a reading in which the later words give the walker and the setting of another instance of walking. The distinction depends on whether the earlier predicate is needed and whether the later expression already makes a complete assertion in context.

A missing word in a witness raises a separate question. A break in an inscription, a manuscript gap, or a reading that depends on editorial restoration provides evidence about the preservation of a passage. Grammatical omission is interpretable from the text as transmitted. Where both damage and a grammatical omission are plausible, the form of the predicate, the syntax of nearby clauses, and any surviving parallel passages constrain the reconstruction. A recoverable meaning alone cannot establish whether a particular word once stood in the text.

Distinguishing discourse effects from metrical constraints

In metrical verse, the placement of a word affects which syllables occur at particular positions in the line. The syllable sequence and its quantities can therefore favor one order among those the grammar permits. The metrical organization exemplified by the Wṛttasañcaya gives a basis for examining a line alongside the syntax of its clause (Kern 1875, stanza 1).

Begin with the transmitted or edited wording and determine the clause boundaries independently of the line endings. Identify the predicate, the constituents it governs, and any phrases that set a wider frame. A verb near the end of a line may have an argument in the next line. A noun phrase at the beginning of a line may continue an earlier clause. Either situation changes what counts as preposing or postponement in the syntax.

Next, examine the metrical positions occupied by the relevant words. Apply the syllabification and the reading appropriate to the witness, including any sound alternation needed for the line. A reordering that preserves the syntax and meaning may still alter the quantities at the required positions. If a plausible alternative order breaks the pattern, the metre provides a reason for the transmitted order. If multiple orders fit, metre alone leaves their selection open. The fact that a line scans establishes that the order fits. Establishing that metre favored that order requires comparison with plausible alternatives.

Then look for contextual evidence that the order has a discourse function. A preceding question can identify the focus of an answer. An explicit contrast can explain why a participant is named before the event. Repetition across neighboring clauses can mark a shift from one topic to another. A resuming pronoun can support dislocation even where the metre also favors the placement. In prose, comparable constructions help establish which orders the grammar permits. A comparison is strongest when it holds predicate type, voice, and constituent structure constant.

Discourse and metre can favor the same placement. A preposed noun phrase can select the next person to be discussed and occupy a position suited to its syllables. Meter can also limit the available choices in a clause whose topic or focus is clear from other evidence. Where the context supports several readings and the metrical pattern supports the transmitted order, describe the order and the scansion first. Assign a narrower discourse function only where the context or the construction provides evidence for it.

How Mova handles information structure and ellipsis

Old Javanese (kaw) has no level-2 parser model. When tools/mova-parse receives a span assigned to kaw, the # parse comment reports the missing model. The token rows have _ in LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL, with Parse=None in MISC. Mova holds no annotated Old Javanese treebank, so it has no measured treebank baseline for the language.

Language assignment precedes parsing. tools/langseg distinguishes English and French in Latin script, Hebrew-script varieties, and Hebrew or Aramaic in scholarly transliteration. Its Latin-script language identification uses lexicons and a character model. Adding a Latin-script language requires a lexicon and a Lang code. The described coverage does not specifically identify Old Javanese, and scripts outside the recognizer’s coverage are assigned other. An Old Javanese edition in romanization therefore has no dedicated kaw recognition through the stated coverage.

The en crate provides the parser and tagger baseline described in What every language gets. After en xpos-fill, en train trains on a language’s Universal Dependencies training split, and en parse-eval measures UAS and LAS on the development split. For Old Javanese, the first implementation steps concern what a tokenizer, tagger, and treebank baseline require.

Squid reads the grammatical points in this chapter as claims. Formalizable claims enter its queue for implementation. Mova does not yet implement the chapter’s analyses of clause order, topic, focus, dislocation, or ellipsis for kaw. Its frame, pragmatics, and world components, including prag and world, read English trees and have no Old Javanese parse to read.

Sources for this chapter

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/jv/kaw/23-information-structure/