31. Reference and discourse
Middle Korean can track a participant through a name, a description, a pronoun, a demonstrative or an unexpressed argument. The interpretation depends on the predicate, the participants already under discussion and the point of view from which the passage is told. These same factors help readers connect clauses and recognize a change of speaker. The preface to the Hunminjeongeum eonhae supplies examples of these processes in fifteenth-century prose (text) 1459, § preface).
Introducing and tracking discourse participants
A referent is the person, group, object, event or other entity that an expression picks out in context. A discourse participant is a referent that can be tracked as the text proceeds. Introduction makes a referent available for subsequent mention. Coreference holds when two mentions pick out the same referent. Anaphora is the interpretation of a form or an unexpressed position through something already available in the discourse. An anaphoric expression may pick out an earlier participant, a part of one, or the situation described in a whole passage.
A lexical noun phrase can introduce an individual, a group or a kind. Its first occurrence in a passage can also retrieve an entity already identifiable by name, by shared knowledge or by its relation to the setting. A title for a known person and a description such as ‘the country's speech’ can therefore open a discourse with identifiable referents. Conversely, a noun phrase repeated later can pick out a new individual of the same kind. The identification comes from the phrase and its context together.
The royal preface begins by putting the country's speech and written characters into relation. It then brings the people affected by that relation into the account. Its referents have different statuses. The country's speech is identifiable through its relation to the country. The people are presented as a group, while the later reference to what an individual wishes to express is understood member by member. The speaker becomes explicit when the preface turns to the response to the problem (text) 1459, § preface).
The attested forms below are short excerpts. Side dots are omitted and word division is editorial. In the romanization, ă represents ㆍ, ŏ represents ㅓ, ŭ represents ㅡ, and ăi represents ᆡ. Glosses use NOM for nominative, ACC for accusative, GEN for genitive, DEM for demonstrative, REFL for reflexive, DECL for declarative and LINK for a clause-linking ending. A period joins gloss labels on one form, and a hyphen marks an internal boundary shown in the gloss.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| “나랏말ᄊᆞ미” | narat-malssămi | country.GEN-speech.NOM | ‘the country's speech’ |
| “제 ᄠᅳ들” | ce ttŭtŭl | REFL.GEN intention.ACC | ‘one's own intention’ |
| “사ᄅᆞᆷ마다” | sarăm-mada | person-each | ‘each person’ |
The first phrase identifies speech through an associated entity. The second makes the possessor depend on the participant whose expression is at issue. The third distributes the intended benefit over people individually. Its domain is supplied by the passage, and its distributive meaning does not require a separate antecedent for every individual (text) 1459, § preface).
Case marking contributes the role of a mention in its own clause. A participant may be an object when first brought into a narrative and a subject when its subsequent action is described. An explicit topic can likewise establish a participant for several clauses whose subjects or objects vary. Keep the discourse topic, the grammatical subject and the referent of a particular noun phrase distinct. The connection among them is part of the interpretation.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 사ᄅᆞᄆᆞᆯ 보니라. 그 사ᄅᆞ미 오니라. | sarămăl ponira. kŭ sarămi onira. | person.ACC see-DECL. DEM person.NOM come-DECL | ‘Someone saw a person. That person came.’ (constructed) |
Here the demonstrative phrase takes up the person first mentioned as the object of ‘see’. The seer is also a participant, although the first clause leaves that argument unexpressed. The second sentence gives the referent taken up by the demonstrative a different local role, subject of ‘come’. If the passage had already established two visible people, the demonstrative could also be interpreted by pointing or by the narrator's attention to one of them. The coreference in the translation is the reading supported by this two-sentence context.
A chain of mentions can include several kinds of expression. A full description can establish the participant, a later demonstrative phrase can retrieve it, and an unexpressed argument can continue it when the role is recoverable. A subsequent full description can resume the same chain after an interruption. To decide whether that later description resumes a participant or introduces another one, check whether the intervening discourse has made another candidate available and whether the predicate and event sequence suit the proposed referent.
Reference also operates at different levels of detail. A collective description can be taken up as a whole or narrowed to its members. A possessive expression may then pick out something belonging to each member in turn. In the preface, the reference to an individual's own intention belongs with the people encountering the difficulty. The later expression ‘each person’ states the intended reach of the new writing in distributive terms. Both passages concern people, but a possessive dependency within a description and distribution over a class are different relations (text) 1459, § preface).
Zero anaphora and possible antecedents
A zero argument is an understood argument for which the clause has no overt noun phrase or pronoun. Middle Korean allows zero subjects and objects where the discourse, situation or construction supplies their interpretation. More than one argument of a predicate can be unexpressed. Zero anaphora is the particular case in which an unexpressed argument is understood through an earlier mention or situation in the discourse. An understood speaker, addressee or indefinite agent may instead receive an interpretation from the speech situation or the predicate in context.
Identify the role before assigning a referent. A predicate of seeing supplies a role for a perceiver and a role for what is perceived. A predicate of coming supplies a role for the one who comes. An unexpressed participant in the first predicate might therefore compete with an overt noun phrase when the reader interprets the subject of the second. In a clause whose predicate has an ordinary intransitive reading, the absence of an object phrase gives no reason to posit another participant merely because a preceding clause had one.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 사ᄅᆞ미 오니라. 가니라. | sarămi onira. kanira. | person.NOM come-DECL. go-DECL | ‘A person came. The person left.’ (constructed) |
| 사ᄅᆞᄆᆞᆯ 보니라. 가니라. | sarămăl ponira. kanira. | person.ACC see-DECL. go-DECL | ‘Someone saw a person. Then someone left.’ (constructed) |
The first sequence favors continuation of the overt subject, since the two clauses describe movements and the fragment makes only one person salient. In the second sequence, either the seer or the person seen can leave. The second English translation preserves the uncertainty about who left. A longer passage might settle it by describing whose route is being followed or whose movements motivate the next event. The change from object in one clause to understood subject in another is grammatically possible.
When an antecedent is available, proximity is one cue among several. The closest phrase can belong inside a larger description, refer to an entity of the wrong kind, or name a participant incidental to the current action. A previous topic can remain available across an intervening mention, especially when the following predicate returns to that topic's actions or circumstances. If two candidates fit the predicate and the sequence equally well, the passage can leave the reference unresolved.
A clause boundary affects the evidence for continuity. In a tightly linked chain, a single participant may do several things, and the shared interpretation can be supported by the order of the actions or by their participation in one event. A causal, concessive or temporal link can also connect two differently predicated situations with different subjects. At a sentence or paragraph transition, the previous topic may remain in force if the next unit develops it. A new scene, a move into speech or the introduction of a different participant can change the preferred interpretation.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 사ᄅᆞ미 오며 보니라. | sarămi omyŏ ponira. | person.NOM come-LINK see-DECL | ‘A person came and saw [something].’ (constructed) |
In this reading, ‘a person’ supplies the subject of both predicates and the object of seeing is understood from a broader situation. The linked form alone leaves the object unspecified. If the same string occurred in a discourse with a salient observer, that observer could supply the subject of ‘see’ given a context that licensed the shift. The English brackets make the missing object visible in translation. They do not represent a written form.
In the preface, the people are introduced overtly in the account of the difficulty. The desire to speak and the inability to express one's intention are understood against that introduction. The reflexive possessive “제” in “제 ᄠᅳ들” helps keep the possessor aligned with the person whose difficulty is being described. The content desired and the intention to be expressed can be closely related in meaning even though the two phrases occupy different positions in the argument (text) 1459, § preface).
Speech creates another source of understood arguments. An omitted first person can refer to the person currently speaking, and an omitted second person can refer to the addressee if the turn supplies those roles. A directive makes the addressee a natural candidate for its understood actor, subject to the possibility that the addressee is to cause or permit another participant to act. In narration about ordinary or recurring conduct, an unexpressed participant may instead be generic. Thus, in deciding whether a clause contains zero anaphora, look for a recoverable antecedent as well as a plausible interpretation.
Person, social relations and predicate honorification can help narrow the candidate set. In a suitable context, subject honorification favors a subject eligible for that treatment. Several participants can have the same eligibility, and the narrator or quoted speaker can use honorification from a different social standpoint. Speech level is evidence about the relation in which the utterance is made, while the subject's identity is a separate question. Read both alongside lexical meaning and the current speaker's perspective.
The written text may support several kinds of partial result. It can establish the role but leave the referent indefinite, support one antecedent without ruling out another, or establish that successive clauses share a participant whose identity is unknown. A translation can preserve those distinctions by using an indefinite expression, by indicating a preferred reading in a note, or by specifying only the shared role. In each case the source clause can retain its zero argument in the analysis.
Demonstratives, repeated names and shifts of viewpoint
Demonstratives can direct attention to something in the speech situation or retrieve something accessible in the discourse. The Middle Korean demonstrative series has three basic members. Their ordinary spatial values provide a starting point. For discourse reference, accessibility through a previous mention, the hearer's perspective or a represented scene can matter more than physical distance.
| Form | Romanization | Gloss | Value in context |
|---|---|---|---|
| 이 | i | PROX.DEM | ‘this’, near the speaker or presented as immediately available |
| 그 | kŭ | MED.DEM | ‘that’, associated with the hearer's side or available through the context |
| 뎌 | tyŏ | DIST.DEM | ‘that over there’, presented as away from both interlocutors |
The forms in the table are the basic demonstrative bases. In an adnominal phrase the following noun identifies the kind of referent to look for. A case-marked expression based on a demonstrative can itself function as an argument. If the demonstrative retrieves a situation, the antecedent may span multiple propositions. A following noun such as ‘cause’, where present, tells the reader how the text construes that material.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| “이런 젼ᄎᆞ로” | irŏn cyŏnchă-ro | such cause-INSTR | ‘for this reason’ |
| “내 이ᄅᆞᆯ” | nae i-răl | 1SG.NOM PROX.DEM-ACC | ‘I … this’ |
The first phrase looks back to the situation at the opening of the preface and presents it as the cause of the following difficulty. In the second, the speaker takes the situation just described as grounds for action. The excerpt ends within the construction containing ‘this’, so the English rendering is also a fragment. The two demonstrative expressions have overlapping context, but their scope can differ. The first retrieves the mismatch introduced at the opening. The second occurs after the account has included its consequences for the people (text) 1459, § preface).
The available antecedent for a demonstrative can be a proposition, a difficulty or a course of events. A difference in scope can change the meaning even when adjacent demonstratives are translated by the same English word. To determine scope, use the relation asserted by the clause that contains the demonstrative. A statement of a reason can retrieve the material that explains the result. A statement of a response can retrieve the problem to which the speaker responds, including consequences described after its introduction.
Repeated names and descriptions have their own uses. They can reestablish a referent after competing participants have intervened, make a change of actor explicit or renew a title appropriate to the scene. The identity of two titles can also be inferred if the context presents them as roles of one person. Conversely, identical titles can denote different people, and a repeated common noun can introduce a new member of the same class. Determine identity from the available individuals, the events attributed to them and any wording that distinguishes or equates them.
A shift of viewpoint changes the center from which an expression is interpreted. In an account given in the narrator's voice, place and distance may be organized around the narrator or a scene the narrator presents. Within a character's speech, first and second person are interpreted from the character's speech event. A narrator can also describe an event from the position of a participant without transferring the voice of the passage to that person. Evidence for whose location or concern matters can thus differ from evidence for who is speaking.
First-person reference can itself mark a change in how an account is organized. The preface first describes the relation between speech and writing and the trouble it causes the people. With “내”, the royal speaker enters as an explicit participant and supplies the perspective for the ensuing response. The change is from a description of circumstances to an account of what the speaker does about them, within the same prefatory voice (text) 1459, § preface).
A new viewpoint can also alter the choice of description while leaving reference intact. A person introduced by office in the narration can be addressed by an appropriate title and can refer to themself in the first person inside a quotation. The identity across these mentions rests on the speech frame and the scene. Inside the quotation, a demonstrative can point to something available to the characters, including an entity that the narrator has yet to describe. In that case the reader can identify its role in the speech before learning its identity.
Discourse relations in prose and verse
A discourse relation describes how one stretch of text contributes to the interpretation of another. The units can be clauses, sentences or larger passages. The same passage can participate in several relations at different scales. A clause can state a cause for its immediate neighbor while the combined material provides background for a subsequent action. The relation between whole stretches of text and the grammatical dependency between individual clauses should be identified separately.
Temporal succession is supported when successive units describe successive events and the text gives the earlier event as the setting for the next. Order of mention can guide that reading. A later clause can also explain a prior event by supplying an earlier circumstance, and a second description can present another aspect of the same event. Temporal adverbs, the meanings of the predicates and explicit changes in time or place can help distinguish these readings. When the chronology remains open, an analysis can describe the units as related without imposing an order of events beyond what the wording supports.
Cause and result concern the relation between circumstances or events. A reason offered for a decision concerns the actor's grounds for acting. A purpose presents a desired outcome from an actor's perspective, and that outcome can remain unrealized. Concession presents an outcome that holds despite a circumstance that would ordinarily favor a different outcome. Contrast can instead set differing descriptions or alternatives alongside each other without invoking such an expectation. These distinctions affect how a reader groups the propositions and how the passage can be translated.
| Span of the preface | Relation supported by the wording | Scope |
|---|---|---|
| The difference between the country's speech and the writing in use, followed by the people's difficulty | Cause and result | The initial situation explains the difficulty described next. |
| The existence of something a person wishes to say, alongside difficulty expressing it | Concession | The wish and its content supply the circumstance against which the difficulty is understood. |
| The difficulty followed by the speaker's creation of letters | Grounds for action | The speaker responds to the problem developed in the preceding passage. |
| The account of the new letters followed by the desired ease of learning and everyday use | Purpose | The desired benefit supplies the aim of the speaker's action. |
The preface develops these relations across changes in grammatical subject and scale. The causal link connects the opening situation to the group's difficulty. The concession is internal to the description of that difficulty. The speaker's response then takes up the larger problem, and the desired benefit explains the aim of the response. The purpose concerns what the letters are intended to let people do, so translating the last relation as accomplished result would change the assertion (text) 1459, § preface).
Elaboration and exemplification preserve an earlier center of attention while developing it. An elaboration can expand a statement by identifying parts, restating it at another level of detail or specifying its consequences. An example supplies a particular instance of a more general description. A text can shift from a general referent to its parts or instances without switching to an unrelated topic. The movement from a group to people considered one at a time in the preface is an instance of a change in level of description within a continuing concern (text) 1459, § preface).
In verse, use lineation as evidence for grouping, together with clause endings, the recurrence of wording and the distribution of predicates across lines. A line may complete a clause, carry only part of one, or pair with another line as a parallel unit. Matching phrases can invite comparison between two participants or present repeated aspects of one participant. Names, demonstratives and the compatibility of each action with the available referents help settle which reading applies. A refrain can resume a concern at a larger scale even when the immediately preceding line follows a different participant.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 사ᄅᆞ미 오며 / 사ᄅᆞ미 가니라 | sarămi omyŏ / sarămi kanira | person.NOM come-LINK / person.NOM go-DECL | ‘A person comes, / a person goes.’ (constructed) |
The slash represents a line boundary in the constructed example. The parallel wording supports an interpretation that compares coming and going. The first predicate has a linking ending, so the boundary between the lines also falls inside a linked sequence. In a scene with one salient traveler, the two mentions could be of that traveler. A scene describing movement in two directions can supply distinct participants. The translation leaves that choice to the larger context.
Explicit linkage can narrow the relation without resolving every detail. A causal expression can identify which material supplies an explanation while leaving open whether the explanation concerns the event itself or someone's grounds for treating it a certain way. In a paired passage, the wording can establish contrast while leaving open which property is most important to the comparison. Select the more specific reading when the passage identifies the relevant proposition, intention or feature.
Connectives and markers of transitions
Middle Korean can signal relations through a predicate ending, an independent phrase or a combination of the two. An ending links the clause containing its predicate to another unit. A phrase can point back to a proposition or situation and tell the reader how to use it in interpreting what follows. The scopes therefore need to be determined separately when both devices occur at the same transition.
The opening of the preface gives a case in which both occur. The causal ending at the end of the first stretch projects a continuation. The phrase ‘for this reason’ begins the next stretch and retrieves the preceding circumstances as the explanation. The ending and the phrase support the same causal relation from their respective clauses. The phrase can retrieve the larger situation even though the ending attaches to one predicate (text) 1459, § preface).
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| “아니ᄒᆞᆯᄊᆡ” | ani-hă-lssăi | NEG-do-CAUSE | ‘because [it] does not do so’ |
| “이런 젼ᄎᆞ로” | irŏn cyŏnchă-ro | such cause-INSTR | ‘for this reason’ |
| “이셔도” | isyŏdo | exist.CONC | ‘even though [there] is’ |
| “ᄆᆞᄎᆞᆷ내” | măchămnae | ultimately | ‘in the end’ |
These are fragments from the preface. NEG marks negation, CAUSE the causal connective, INSTR the instrumental expression and CONC the concessive form. The causal ending attaches to the negative predicate in the opening. The concessive form on ‘exist’ subordinates the circumstance that the person has something to say. The adverb ‘in the end’ situates the inability to express the intention as the outcome within that account (text) 1459, § preface).
A connective can have local scope even when the clause containing it does work at a larger scale. The concessive form in the difficulty passage links the availability of something to say with the failure to express it. The entire difficulty passage in turn provides the grounds for the speaker's response. Reading the concessive form as an indication of the larger transition would join the wrong units. Read the linked predicates first, then consider how their combined assertion relates to the surrounding passage.
An independent phrase can serve two functions at once. A demonstrative expression can refer to an earlier situation within the syntax of its clause and make the reference a bridge to a new part of the text. In the preface, the demonstrative object after the explicit first person helps supply the grounds for the speaker's action. Accordingly, the phrase does discourse work through an ordinary referential expression in the clause (text) 1459, § preface).
Transitions can also be inferred from the shape of a passage. A new time or place expression can reset the scene. A full mention following a sequence of understood arguments can signal a return to a participant or a transfer of attention. A change from description to speech can introduce a new center of person reference. Each cue has a local interpretation as well. A time adverb can place one event within a continuing scene, and a full noun phrase can identify an argument while the larger topic continues.
Use the ordering of clauses and their semantic fit when the text leaves a relation implicit. For instance, an action followed by a purpose clause and an action followed by a report of its result have different consequences for what is asserted. When an English translation needs an explicit connective for readability, choose one compatible with the source sequence. If both succession and cause remain plausible, a translation can preserve the sequence while leaving the causal reading available.
Page and editorial divisions can help a reader locate stretches that are interpreted together. Punctuation in a modern presentation can also embody an editor's reading of the transition. A sentence break after a causal form, for example, can coexist with the causal dependency that the form establishes across the break. Clause morphology and the lexical scope of the neighboring expressions provide evidence independently of the placement of the break.
Dialogue, quotations and shifts in speaker
A reported exchange contains at least a reporting context and the speech it presents. The reporting context identifies a speaker, addressee or occasion to the extent that the text makes them recoverable. The reported turn can establish its own person, time and place references. For each turn, identify the speech event relevant to the expression under analysis. The same first-person form can refer to different individuals in adjacent turns, and a speaker in the reported event can be the person just described in the third person by the narrator.
Direct speech represents wording from the speaker's position. First person inside it ordinarily picks out that speaker and second person the person addressed in the represented event. Expressions for place and time may likewise be interpreted from within that event. In a report that integrates the content into the narrator's clause, the narrator can instead describe the participants and circumstances by reference to the reporting context. The form of the speech predicate, the syntax of the reported material and the behavior of person expressions all contribute to the interpretation.
| Form | Romanization | Gloss | Translation |
|---|---|---|---|
| 사ᄅᆞ미 닐오ᄃᆡ, “내 가노라.” | sarămi nilodăi, “nae kanora.” | person.NOM say-LINK 1SG.NOM go-DECL | ‘A person said, “I am leaving.”’ (constructed) |
The subject of the reporting predicate and the first person inside the quotation refer here to the same person. The quotation marks are editorial aids in this example. The reporting predicate and its linked form establish a frame for speech, and the clause within the frame is presented as that person's utterance. A separately reported reply with the same “내” would assign the form to the speaker of that reply, given a context with distinct speakers.
The bounds of a turn can be recoverable from the grammar and sequence even when a presentation supplies little punctuation. A reporting predicate can open the turn. A new act of speaking, an explicit mention of another speaker or a return to narrated action can help identify its end. A vocative or an answer licensed by an earlier question can be evidence for the direction of the exchange. A stretch of speech with several sentences can maintain the same speaker through those sentences. Count shifts by evidence of a change in the speech event, taking sentence divisions as one cue.
When a reporting subject is zero, seek its antecedent in the same way as for another unexpressed argument, with the extra evidence of turn structure. Two participants may supply plausible speakers. A question typically makes a reply from its addressee relevant, while a follow-up, self-answer or extended turn can continue the questioner's speech. The wording of the reply may identify its source through first person, address or knowledge available to one participant. If multiple assignments still fit, the turn can remain attributed only to a possible speaker. Assigning a speaker just because turns often alternate would add information to the text.
A change in address also has to be distinguished from a change in speaker. A speaker can address successive participants within one turn or move from a specific individual to a wider audience. A vocative can identify the current addressee while a separate expression identifies the subject of the clause. Speech-level and honorific forms can support an interpretation of the relation among participants. Their force depends on which expression honors which participant and on the standpoint of the narrator or speaker who uses them.
Embedded reports can create more than one active level of speech. If a character recounts what another person said, the innermost first person is interpreted with respect to the innermost represented speaker when the wording is presented directly. On return to the outer turn, first person again takes the outer speaker as its reference. Establish the extent of the embedded material before linking its person forms to mentions outside it. A demonstrative within the inner turn may still retrieve something mentioned outside it if that thing is accessible within the represented exchange.
The preface illustrates a first-person voice without a sequence of alternating turns. Its initial description presents the circumstances of the people. The explicit first person then identifies the royal speaker as the one responding to those circumstances. The later distributive reference to ‘each person’ concerns the people for whom the letters are intended. The speaker, the people discussed and the beneficiaries are roles that can be recovered independently of the number of overt personal expressions in the passage (text) 1459, § preface).
Translations can distinguish attribution from wording by making clear whose utterance is being rendered and where it begins and ends. When the source leaves the speaker implicit, an English pronoun can reflect an assignment established by the discourse. Where attribution remains open, paragraphing or an impersonal report can preserve that uncertainty. The same care applies to events referred to by ‘this’ or ‘that’ inside a turn, since the speaker can be pointing within the scene or referring back to what another speaker has said.
How Mova handles reference and discourse
tools/mova-parse assigns languages to sentences and spans at level 1 through tools/langseg. Hangul falls under other in tools/langseg. Level 2 selects a parser model for each span's language. Middle Korean (okm) has no level-2 model. The parser reports a missing model in the # parse comment and leaves the affected tokens with _ in LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL and Parse=None in MISC. Middle Korean tokenization, tagging, lemmatization and parsing are unavailable.
Mova holds no annotated Middle Korean treebank, so a Middle Korean treebank baseline has yet to be measured. The parser and tagger baseline specified in What every language gets uses en xpos-fill followed by en train on a language's Universal Dependencies training split. en parse-eval measures unlabeled and labeled attachment on the development split. The first grammar points needed for okm are those that support a tokenizer, a tagger and a dependency treebank baseline.
prag and world read English trees. Their use of another language requires parsed input, which is unavailable for okm. The analyses in this chapter, including the assignment of antecedents, the interpretation of discourse connectives and the attribution of quoted speech, currently have the status of manual claims. Squid reads the manual and makes a claim for each grammatical point. Formalizable claims enter its implementation queue. All of this chapter's rules await implementation.
Sources for this chapter
- Hunminjeongeum eonhae (Vernacular Translation of the Correct Sounds for the Instruction of the People). Prefatory text in Worin seokbo, vol. 1, Joseon royal edition, 1459. A vernacular rendering of the alphabet’s statement and letter descriptions, also a dated witness to Middle Korean prose. — public domain; not held in the library, to verify
This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/ko/okm/31-reference-and-discourse/