Polypersonal agreement in translation occurs when one verbal form indexes features of more than one participant in the event. Instead of a verb agreeing only with a subject, the verb may simultaneously encode person, number, gender, noun class or grammatical role for an agent, patient, recipient or another argument. In languages with rich argument indexing, a single verb or auxiliary can therefore contain information that English would distribute across a subject pronoun, verb, object pronoun and indirect-object phrase. Translating only the lexical verb can silently delete participants.
People searching for polypersonal agreement, argument indexing, subject object agreement, multiple argument agreement, Basque verb agreement, person and number on verbs are solving a participant-recovery problem. The term polypersonal agreement is commonly used when a verb cross-references more than one argument, while broader typological literature also speaks of argument indexing because the exact boundary between agreement, pronominal indexing and clitic-like systems varies across analyses and languages. For translation, the practical question is stable: which participants are grammatically represented on the verbal complex, and which of those participants need independent expression in the target?
This article owns the polypersonal-agreement and argument-indexing precision layer for eduKateSG. It does not replace Grammatical Gender and Agreement, Verb Valency, Direct–Inverse Systems, Ergative–Absolutive Alignment, or Pronouns and Reference. Its narrow job is to recover every participant indexed on the verb, keep subject, object, recipient and beneficiary features attached to the right semantic roles, handle null noun phrases and portmanteau forms, and rebuild a natural target sentence without duplicating or dropping arguments.
Quick Read
Segment the verb into LEXICAL ROOT + PERSON/NUMBER/GENDER/CLASS INDEXES + TAM/VOICE/DERIVATION. For every index ask: WHICH ARGUMENT DOES THIS REFER TO? Then build the event frame before adding target pronouns.
One-Sentence Answer
Translate polypersonal agreement by decoding every indexed argument first, then expressing those participants once—and only once—in the target language.
Why This Precision Layer Matters
Polypersonal systems concentrate participant information inside morphology. The danger is either deletion, when an object or recipient index is ignored, or duplication, when an overt noun phrase and its verbal cross-reference are translated as two separate people. Alignment and voice add another layer: a Basque-style absolutive index does not map neatly to English subject agreement. The stable method is to treat the verbal word as a compressed event graph.
What Polypersonal Agreement Is
The first translation task is structural rather than lexical. A polypersonal system cross-references more than one core or selected participant on the verb or auxiliary, often encoding person, number and sometimes gender/class.
The main translation risk is assuming every agreement marker refers to the subject. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to write one slot for every indexed role before deciding target pronouns. For example, a verb can encode ‘we’, ‘it’ and ‘to him/her’ simultaneously Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, restore the complete participant set before drafting. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Agreement vs Argument Indexing
This becomes easier once morphology and participant roles are separated. Some traditions reserve agreement for feature matching with an overt noun phrase while broader work uses indexing for verbal person marking more generally.
The main translation risk is turning a terminology dispute into a translation error. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to focus on what participant information the form contributes. For example, whether called agreement or indexing, a marker can still identify first-person agent and third-person patient Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, translate the participant rather than the label. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Subject and Object Indexing
A fluent target can still fail if this relation shifts. Many systems index both an agent/subject-like participant and a patient/object-like participant.
The main translation risk is recovering the subject but dropping the object because English agreement usually marks only subjects. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to map the predicate as A→O and assign each exponent. For example, a source form may encode 1SG acting on 3PL with no overt pronouns Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, express both participants if target grammar requires them. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Recipient and Dative Indexing
The source form is evidence, but it is not the target blueprint. Some systems also index a recipient, goal or dative participant in addition to agent and theme.
The main translation risk is mistaking recipient indexing for a second theme object. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to build A-R-T frames before translation. For example, a verbal complex can encode ‘we bring it to him’ Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, keep theme and recipient distinct. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Three-Argument Verbs
The safest method is to analyse function first and surface form second. Ditransitive predicates can create dense templates because agent, theme and recipient may all be represented morphologically.
The main translation risk is using English object order to guess marker roles. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to compare each slot with the source ditransitive paradigm. For example, one index may represent absolutive theme while another represents dative recipient Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, reconstruct transfer semantics before target order. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Person Features
The first translation task is structural rather than lexical. First, second and third person are often central to argument indexing and may interact with hierarchy restrictions.
The main translation risk is converting grammatical person into named referents without context. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to separate person features from discourse identity. For example, 3PL tells you ‘they/them’ but not which group Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, use conservative target reference. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Number Features
This becomes easier once morphology and participant roles are separated. Agreement may encode singular, plural, dual or other number values separately for several participants.
The main translation risk is letting number from one indexed participant spread to another. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to record number per role. For example, 1PL agent + 3SG patient differs from 1SG agent + 3PL patient Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, verify every target pronoun matches its own role. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Gender and Noun Class
A fluent target can still fail if this relation shifts. Indexes may encode grammatical gender or noun class for one or more participants.
The main translation risk is treating grammatical gender as biological sex. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to map gender/class as grammatical information first. For example, a feminine index can constrain reference without proving biological female sex in every system Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, choose target pronouns from context plus grammar. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Head-Marking
The source form is evidence, but it is not the target blueprint. Polypersonal indexing is a strong head-marking strategy: participant relations can be concentrated on the verbal head.
The main translation risk is dropping argument information when morphology disappears in English. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to treat the verb as a compressed clause. For example, one word can equal several English pronouns plus a verb Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, expand only as much as target grammar needs. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Overt NPs Plus Indexing
The safest method is to analyse function first and surface form second. A source can contain both a full noun phrase and verbal marking referring to the same participant.
The main translation risk is translating both as separate participants. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to link overt NP and index into one reference chain. For example, Lina plus 3SG agreement still denotes one Lina Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, represent each participant once. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Null Arguments
The first translation task is structural rather than lexical. Rich indexing often allows one or more noun phrases to be omitted.
The main translation risk is treating omitted arguments as absent from meaning. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to decode verbal indexes before adding target pronouns. For example, a form can mean ‘I saw him’ with neither pronoun overt Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, insert target pronouns where required. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Null vs Unspecified
This becomes easier once morphology and participant roles are separated. Morphology can reveal person/number while exact discourse identity remains unresolved.
The main translation risk is inventing a named referent from partial features. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to separate grammatical recovery from referent identification. For example, 3PL object marking identifies plurality but not which people Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, keep target reference neutral when needed. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Basque-Style Multi-Argument Indexing
A fluent target can still fail if this relation shifts. Basque is a well-known example where finite verbal forms can cross-reference absolutive, ergative and dative participants.
The main translation risk is forcing Basque morphology into an English subject-only model. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to label ABS, ERG and DAT before mapping roles. For example, a bring/give form can encode giver, theme and recipient on the auxiliary Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, rebuild the English event from case-role mapping. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Ergative–Absolutive Indexing
The source form is evidence, but it is not the target blueprint. In ergative systems the absolutive series can track intransitive S and transitive O rather than English-like subjects.
The main translation risk is calling absolutive marking subject agreement in every clause. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to identify clause valency first. For example, ABS may be intransitive subject or transitive object Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, map A/S/O before translating. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Direct–Inverse Interaction
The safest method is to analyse function first and surface form second. Hierarchical languages can combine multi-argument indexes with direct–inverse morphology that signals direction between them.
The main translation risk is decoding person markers correctly but reversing who acts on whom. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to interpret direction morphology after participant features. For example, 1SG and 3SG can both be encoded while inverse/direct tells direction Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, repair direction before pronoun order. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Clitics vs Agreement
The first translation task is structural rather than lexical. Some bound person forms behave more like clitics or incorporated pronouns than canonical agreement, and analyses differ.
The main translation risk is assuming attachment proves one syntactic category. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to check obligatoriness, doubling, position and paradigms. For example, a bound marker can function pronominally while still carrying participant information Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, preserve the referent regardless of theory. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Portmanteau Forms
This becomes easier once morphology and participant roles are separated. One exponent can encode a combination of participant features rather than one transparent slot per argument.
The main translation risk is splitting a fused form into imaginary morphemes. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to consult the paradigm for combination meaning. For example, one form may uniquely signal 1→2 Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, translate the participant combination rather than invented pieces. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Syncretism
A fluent target can still fail if this relation shifts. Different participant combinations can share one surface form.
The main translation risk is assuming one form has one unique reading. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to use context and other morphology to resolve syncretic cells. For example, the same suffix may serve more than one third-person combination Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, preserve ambiguity if the source does. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Zero Marking
The source form is evidence, but it is not the target blueprint. Some argument values are represented by zero morphology while others are overt.
The main translation risk is assuming no visible marker means no participant. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to read the full paradigm including zero cells. For example, 3SG can be zero in a slot where 1SG/2SG are overt Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, recover the zero-indexed participant from grammar/context. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Agreement and Valency
The safest method is to analyse function first and surface form second. Applicatives, causatives, passives and antipassives can change which participants are eligible for indexing.
The main translation risk is copying the base template onto a derived verb. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to compare argument structure before and after derivation. For example, an applicative can add a beneficiary index Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, rebuild the derived frame first. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Passive and Agreement
The first translation task is structural rather than lexical. Voice changes can shift which participant controls an agreement slot.
The main translation risk is keeping the active actor as target subject because its features remain encoded somewhere. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to map source voice independently from feature presence. For example, patient may become the privileged indexed participant under passive Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, choose target subject after voice analysis. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Applicatives and Indexing
This becomes easier once morphology and participant roles are separated. An applied beneficiary or recipient can enter the agreement system.
The main translation risk is dropping the applied participant because no full NP appears. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to compare base and applicative templates. For example, new dative agreement can be the only overt sign of the beneficiary Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, restore it when target semantics requires. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Causatives and Indexing
A fluent target can still fail if this relation shifts. Causatives add causer and causee roles and can create multi-person verbal forms.
The main translation risk is confusing causer, causee and patient. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to draw CAUSER [CAUSEE performs EVENT on PATIENT]. For example, three person values may correspond to three event roles Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, translate the layered event rather than slot order. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Negation and Agreement
The source form is evidence, but it is not the target blueprint. Negative morphology can reorganise the verbal complex without changing participant identity.
The main translation risk is letting a negative auxiliary obscure who each index refers to. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to decode person features independently of polarity. For example, ‘I did not see them’ still has 1SG agent and 3PL patient Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, restore the same participant map under negation. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Narrative Reference
The safest method is to analyse function first and surface form second. Rich indexing can maintain characters across clauses without repeated noun phrases.
The main translation risk is adding names so frequently that target discourse becomes heavy. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to keep a referent ledger and use target pronouns strategically. For example, several verbs may continue the same actor/patient chain Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, preserve continuity without over-explicating. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Instructions
The first translation task is structural rather than lexical. Procedures can encode operator, item and recipient directly in verbal morphology.
The main translation risk is dropping an indexed participant because the instruction still sounds possible. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to translate operational actor and affected participant explicitly. For example, a 2→3 form can mean ‘you attach it’ Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, verify who performs and what receives the action. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Translation into English
This becomes easier once morphology and participant roles are separated. English usually expands multi-argument indexing into independent pronouns, noun phrases and prepositional complements.
The main translation risk is trying to imitate source morphology or leaving participants unexpressed. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to expand the participant graph into ordinary English syntax. For example, one source form may become ‘we will bring it to them’ Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, avoid both under-translation and duplication. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
AI Participant Dropping
A fluent target can still fail if this relation shifts. Models can treat one agreement slot as expendable inflection and omit that participant.
The main translation risk is losing object, recipient or beneficiary information. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to count indexed roles before generation. For example, a three-place source form becoming ‘we bring’ drops two participants Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, restore every selected argument. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
AI Participant Duplication
The source form is evidence, but it is not the target blueprint. Models can translate both a verbal index and its overt noun phrase separately.
The main translation risk is creating extra actors or objects. Because the target can remain grammatical, the error may survive ordinary proofreading. A precise review reconstructs the participant map, discourse relation or event layer that the source grammar encoded instead of asking only whether the same nouns and verbs appear.
A practical diagnostic is to link each index to an overt NP when possible. For example, Lina + 3SG index is one participant, not Lina and she Once the underlying relation is stable, the target language may use independent pronouns, word order, auxiliary verbs, prepositions, adverbs, clause expansion, particles or no visible counterpart at all. Formal resemblance is optional; preservation of the source-supported relation is not.
For recovery, merge duplicated reference before final prose. Repair from the first unstable structural decision, then restore register, rhythm and target-language naturalness.
Worked Examples
Worked Example 1: Subject and Object Both Indexed
Source problem: A source verb encodes first-person singular agent and third-person plural patient with no overt noun phrases.
Diagnosis: The verb supplies grammatical ‘I’ and ‘them’; exact third-person identity still comes from discourse.
Target strategy: English normally needs ‘I saw them.’
Verification: Do not drop the object and do not invent a named group.
Worked Example 2: Three Indexed Arguments
Source problem: A transfer verb indexes first-person plural giver, third-person singular theme and third-person recipient.
Diagnosis: The event contains A, T and R compressed into one verbal complex.
Target strategy: English may expand to ‘We brought it to him/her.’
Verification: Theme and recipient must remain distinct.
Worked Example 3: Basque ABS/ERG/DAT
Source problem: A verbal form cross-references absolutive, ergative and dative participants.
Diagnosis: The absolutive index is not always an English subject; its role depends on valency.
Target strategy: Map A/S/O/DAT first, then generate target subject/object/recipient syntax.
Verification: Do not equate absolutive indexing with subject agreement automatically.
Worked Example 4: Overt NP Plus Index
Source problem: A full noun phrase occurs alongside matching verbal agreement.
Diagnosis: Both expressions represent one participant.
Target strategy: Translate the participant once.
Verification: Count semantic participants, not morphological expressions.
Practical Diagnostic Workflow
Build an ARGUMENT INDEX table with one row per semantic participant and columns for ROLE, PERSON, NUMBER, GENDER/CLASS, OVERT NP?, VERBAL SLOT, CASE/ALIGNMENT and DISCOURSE IDENTITY.
Run a full-clause expansion test: replace every verbal index with an independent noun phrase or pronoun in a rough paraphrase. If the expanded source has more participants than the target, something was dropped; if the target has more, something was duplicated.
Verification Method
From the target alone, reconstruct every source-selected participant and assign person/number features. Compare event direction and valency with the source. The same participants should be recoverable exactly once.
Recovery Method
When an indexed participant is missing, return to the verbal template and restore that role. When one has been duplicated, merge the overt noun phrase and its index into a single reference chain. Only then polish target pronouns.
Practice Sequence
Take SEE, HELP, GIVE and BRING. Imagine paradigms that index subject only, subject+object and subject+object+recipient. Translate each into English, then reverse the exercise by compressing English pronouns into a feature table.
Quality-Assurance Checklist
- Segment lexical root from indexing.
- Identify every indexed argument.
- Attach person and number to the correct role.
- Treat grammatical gender/class cautiously.
- Handle overt NP plus index as one referent.
- Recover null arguments conservatively.
- Check ergative–absolutive alignment.
- Check direct–inverse direction separately.
- Use paradigms for portmanteau and syncretic forms.
- Track valency changes under voice/applicatives/causatives.
- Preserve narrative reference continuity.
- Audit instructions for omitted participants.
- Audit AI participant dropping.
- Audit AI participant duplication.
Frequently Asked Questions
What is polypersonal agreement?
A system where a verb indexes features of more than one participant, such as subject, object and sometimes recipient.
Is it the same as polysynthesis?
No. They often co-occur, but they are distinct properties.
Can one verb encode three participants?
Yes. Some languages cross-reference several arguments on one verb or auxiliary.
Do indexed participants need overt noun phrases?
Not always. Rich indexing can license null arguments.
Is every bound person marker agreement?
Not necessarily; some analyses call certain forms clitics or pronominal indexes.
Why is Basque a common example?
Basque verbal forms can cross-reference absolutive, ergative and dative participants.
How should English translate it?
Usually by expanding the participant information into ordinary subject, object and prepositional pronouns or noun phrases.
Why does AI fail?
It can ignore small agreement markers or translate both the marker and overt noun as separate participants.
Internal Routes Through the eduKate Translation Architecture
Route general agreement to Grammatical Gender and Agreement: https://edukatesg.com/2026/09/18/translate-grammatical-gender-agreement-reference-gender-number-person-differently/. Route participant frames to Verb Valency and Argument Structure: https://edukatesg.com/2026/09/18/translate-precisely-verb-valency-argument-structure-transitivity-complements/. Route case to Grammatical Case: https://edukatesg.com/2026/09/18/translate-grammatical-case-who-did-what-to-whom-language-roles/. Route hierarchy direction to the prepared Direct–Inverse Systems owner. Route transfer frames to the prepared Ditransitive Alignment owner.
Final Principle
A polypersonal verb is a compressed participant map. Decode every indexed person, number and role before translating the lexical event. The target may need several words where the source uses one, but each source participant must appear exactly once.