Polypersonal agreement in translation occurs when a verb indexes grammatical features of more than one argument. Instead of marking only a subject-like participant, a single verb can encode information about the agent, patient, recipient or another object-like participant at the same time. Cross-linguistically, verbal person marking can index both A and P arguments, and rich systems can combine person, number, gender/class, animacy or definiteness with voice and tense morphology.
People searching for polypersonal agreement, multi-argument agreement, subject and object agreement, verbal person marking, multiple argument indexing are solving a morphological argument-tracking problem. A source verb may contain enough information to identify ‘I→you,’ ‘she→them,’ or ‘we→him’ even when no independent pronouns are written. Translation precision therefore requires decoding the complete agreement bundle before English or another target language supplies its own pronouns and word order.
This article owns the polypersonal-agreement precision layer for eduKateSG. It does not replace the broader Grammatical Gender and Agreement, Verb Valency, Direct–Inverse, Ergative Alignment or Word Formation owners. Its narrow task is to preserve multiple-argument indexing: which argument each affix indexes, how alignment and person hierarchies affect the paradigm, how object/recipient indexing interacts with valency, and how to expand a morphologically dense verb without losing or duplicating participants.
Quick Read
Do not translate the verb stem first and treat the rest as decoration. Segment PERSON/NUMBER/CLASS markers, identify which source argument each marker indexes, then build a participant table. Only after the agreement bundle is decoded should the target introduce pronouns, noun phrases and word order.
One-Sentence Answer
Translate polypersonal agreement by decoding every indexed argument on the verb first, then expressing the same participant set through the target language’s pronouns, nouns, agreement and word order.
Why This Precision Layer Matters
A morphologically complex source verb can contain a sentence’s worth of participant information. If one object marker is ignored, the target can lose the recipient. If an agreement affix is mistaken for a free pronoun and doubled, the target creates an extra person. Multi-argument indexing therefore needs a participant audit separate from general morphology.
What Polypersonal Agreement Is
The first task is to map morphology to arguments. Polypersonal agreement refers to verbal morphology that indexes features of more than one argument, commonly an agent/subject plus a patient/object, and in some systems additional arguments. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is translating the verb as though only the subject were grammatically represented. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to list every argument whose person, number, gender/class or definiteness is encoded on the verb. For example, a verb can simultaneously encode I→you, she→him or they→us without independent pronouns. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement vs Full Pronouns
This distinction becomes clear when the same verb stem is compared across person combinations. Agreement affixes are not simply miniature free pronouns, even though they can encode similar person/number information. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is expanding every agreement marker into an emphatic independent pronoun and making the target repetitive. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to separate indexing morphology from overt noun/pronoun expressions. For example, a source can have a full noun plus agreement on the verb without meaning two participants. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Indexing A and P
A precise translation treats the paradigm as a system rather than isolated suffixes. Many languages index both the more agent-like transitive argument A and the more patient-like argument P on the verb. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is reading one affix as tense/aspect because English only expects subject agreement. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to build an A/P feature row before translating. For example, one verb can mark 1SG agent and 2SG patient at the same time. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Indexing S in Intransitives
The target may need several words for information encoded in one source verb. Intransitive verbs can index their single argument S, but its morphology may align with A, P or another pattern depending on the language. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is assuming S always uses the same agreement series as transitive A. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to compare intransitive and transitive paradigms. For example, an ergative alignment can make S pattern with P rather than A. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Accusative Agreement Alignment
Argument features should be recovered before target pronouns are chosen. In accusative-style agreement, S and A tend to use the same agreement pattern while P is indexed differently. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is projecting this familiar pattern onto all languages. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to compare S, A and P agreement slots directly. For example, English-like subject agreement groups S/A, though English usually lacks full P indexing. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Ergative Agreement Alignment
The first task is to map morphology to arguments. Some polypersonal systems align S with P in agreement while A uses a different series. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is treating the P/S index as subject agreement in every clause. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to label alignment before assigning grammatical roles. For example, a single affix series can encode S of intransitives and P of transitives. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Hierarchical Agreement
This distinction becomes clear when the same verb stem is compared across person combinations. Some systems use person hierarchies or direct–inverse morphology so the verb encodes participant combinations without a simple fixed subject-slot/object-slot mapping. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is forcing every affix into an A-prefix/P-suffix template. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to decode the whole paradigm by participant hierarchy. For example, 1→3 and 3→1 can differ through direction markers rather than independent subject/object slots. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Portmanteau Forms
A precise translation treats the paradigm as a system rather than isolated suffixes. A single verbal morpheme can encode a combination of two participant features rather than decomposing neatly into one subject marker plus one object marker. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is splitting an indivisible form into invented affixes. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to treat the form as a paradigm cell when morphology is portmanteau. For example, one ending can mean 1SG acting on 2SG as a whole package. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Subject–Object Prefix/Suffix Systems
The target may need several words for information encoded in one source verb. Other languages do have relatively segmentable slots for multiple arguments, sometimes with one argument before the stem and another after it. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is assuming linear order of affixes equals semantic prominence. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to read person/number features from the paradigm rather than position. For example, object indexing can precede the stem while subject indexing follows it. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Ditransitive Indexing
Argument features should be recovered before target pronouns are chosen. Some systems can index a recipient, indirect object or other secondary object in addition to subject and theme, though the exact maximum and restrictions are language-specific. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is dropping the third participant because the target verb remains grammatical with two arguments. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to map giver, theme and recipient independently before decoding morphology. For example, a verb form can distinguish ‘I give it to you’ from ‘I give it to them’. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Beneficiary and Applied Arguments
The first task is to map morphology to arguments. Applicative constructions can create an additional object-like participant that then becomes eligible for agreement/indexing. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is mistaking applicative agreement for a second theme or object. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to decode valency-changing morphology before agreement. For example, a beneficiary introduced by an applicative can receive object indexing. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Person Features
This distinction becomes clear when the same verb stem is compared across person combinations. First, second and third person are the most common agreement features and can create special interactions when speaker and addressee are both involved. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is normalising first/second-person contrasts into generic pronouns. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to record SPEAKER, ADDRESSEE and THIRD PERSON explicitly. For example, 1→2 and 2→1 forms can be morphologically exceptional. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Number Features
A precise translation treats the paradigm as a system rather than isolated suffixes. Agreement can encode singular, plural, dual or other number distinctions for more than one argument. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is preserving subject number while losing object number. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to write number for every indexed argument. For example, a verb can encode plural agent acting on singular patient. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Gender and Noun Class
The target may need several words for information encoded in one source verb. Third-person indexing can also encode gender or noun class in some languages, especially where agreement systems are rich. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is inventing natural gender from grammatical class or dropping class agreement that disambiguates reference. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to separate lexical noun class from biological gender. For example, an object marker can identify a noun-class category rather than male/female sex. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Animacy Features
Argument features should be recovered before target pronouns are chosen. Animacy can influence agreement choice, especially in hierarchical or differential systems. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is treating animate/inanimate indexing as a lexical adjective meaning alive/dead. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to use the language’s grammatical animacy classification. For example, a grammatically animate noun can trigger a special agreement series. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Definiteness-Conditioned Conjugation
The first task is to map morphology to arguments. Some languages alter verbal agreement or conjugation according to the definiteness/specificity of the object rather than indexing all object features in the same way. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is calling the distinction pure subject agreement or translating morphology as a definite article. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to record object definiteness separately from subject person. For example, Hungarian-type definite/indefinite conjugation shows how object properties can affect the verb. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Optional Object Agreement
This distinction becomes clear when the same verb stem is compared across person combinations. Object indexing can be obligatory in some contexts and optional or conditioned by discourse/reference in others. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is assuming absence of an object marker means no object exists. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to check the language-specific agreement conditions. For example, a full object NP may appear with or without an object marker depending on the system. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Clitic Doubling vs Agreement
A precise translation treats the paradigm as a system rather than isolated suffixes. Object clitics can coexist with full noun phrases and sometimes develop agreement-like properties. The boundary between clitic and agreement is language-specific. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is translating both clitic and noun as separate participants. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to link doubling elements into one reference chain. For example, a doubled object can be represented twice morphosyntactically but once semantically. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement vs Incorporation
The target may need several words for information encoded in one source verb. Polypersonal indexing marks participant features; noun incorporation places lexical noun material inside the verb. They are different mechanisms that can coexist. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is treating a person marker as an incorporated pronoun/noun or vice versa. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to ask whether the morpheme carries lexical noun meaning or grammatical features. For example, 3PL object agreement does not mean the object noun has been incorporated. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement and Pro-Drop
Argument features should be recovered before target pronouns are chosen. Rich agreement can allow independent pronouns or noun phrases to be omitted because the verb already identifies participants. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is adding generic pronouns before decoding the agreement bundle. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to recover omitted arguments from morphology first. For example, a single verb can correspond to ‘I saw you’ with neither English-style pronoun overt. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement and Word Order
The first task is to map morphology to arguments. When the verb robustly indexes multiple arguments, noun phrase order can be freer and serve topic/focus functions. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is assigning subject/object from source word order after ignoring agreement morphology. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to decode the verb before linear order. For example, an object can precede the subject without changing who acts on whom. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement and Reference Tracking
This distinction becomes clear when the same verb stem is compared across person combinations. Multi-argument indexing can maintain participant identity across discourse even when nouns are omitted. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is inserting the nearest noun as antecedent without checking person/number features. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to build a discourse ledger keyed to agreement features. For example, a 3PL object marker can continue a previously introduced plural group. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement and Voice
A precise translation treats the paradigm as a system rather than isolated suffixes. Passive, antipassive, applicative or other voice operations can change which arguments are eligible for indexing. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is copying the active agreement interpretation into a derived voice form. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to decode valency/voice before agreement slots. For example, an applicative object or passive subject can control a different agreement relation. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement and Tense/Aspect Stacking
The target may need several words for information encoded in one source verb. Person markers can appear in the same complex verb as tense, aspect, negation and derivational morphology. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is misidentifying a person marker as TAM or vice versa because the verb is morphologically dense. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to segment the verb by grammatical category. For example, PREFIX-person + STEM + TAM + object suffix can encode several layers at once. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Georgian-Type Multi-Argument Verbs
Argument features should be recovered before target pronouns are chosen. Georgian is a well-known case where verbs can encode features associated with multiple arguments through complex morphology and series-sensitive patterns. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is trying to decode Georgian-like forms using a simple English subject-ending model. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to analyse the full screeve/series and argument structure. For example, one verb form can encode subject and object person relations that require several English words. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Basque-Type Agreement
The first task is to map morphology to arguments. Basque finite auxiliaries can index multiple core arguments, including ergative, absolutive and dative participants in suitable constructions. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is treating the auxiliary as semantically empty and ignoring who is indexed. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to map case roles before decoding auxiliary agreement. For example, a finite form can encode who did something, what/who was affected and to whom. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Bantu Subject and Object Marking
This distinction becomes clear when the same verb stem is compared across person combinations. Many Bantu languages use subject markers and object markers on the verb, with noun-class distinctions particularly visible in third-person forms. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is assuming object markers are always free pronouns or treating noun class as lexical meaning. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to identify agreement/controller noun phrases and class features. For example, a verb can index both a classed subject and object while full NPs are also present. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Agreement in Translation into English
A precise translation treats the paradigm as a system rather than isolated suffixes. English generally expresses participant identity through independent pronouns/nouns and word order rather than rich multi-argument verb morphology. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is trying to preserve the source verb as one English word and dropping encoded participants. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to expand each indexed participant into the target only when needed for grammatical/reference clarity. For example, one source verb may require ‘I gave it to them’ in English. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Avoiding Over-Explicitation
The target may need several words for information encoded in one source verb. Not every indexed argument needs an overt English noun if context or target grammar already supplies it. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is expanding every affix into a repeated full noun and producing unnatural redundancy. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to distinguish required target arguments from recoverable discourse participants. For example, a target pronoun can carry person/number information that took several source morphemes. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
AI Morphological Segmentation
Argument features should be recovered before target pronouns are chosen. Models can mis-segment a complex verb and assign person markers to the wrong argument. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is trusting a fluent English pronoun set without verifying the source paradigm. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to gloss each person/number marker before generation. For example, 1SG-agent plus 2PL-object can be reversed by a model unfamiliar with slot order. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
AI Argument Loss
The first task is to map morphology to arguments. AI can omit one indexed participant when English permits a syntactically complete but semantically reduced clause. Multi-argument indexing varies greatly across languages: some systems are segmentable, others use portmanteau forms, hierarchy-driven paradigms or clitic-like markers. The translator should therefore decode the actual paradigm rather than assume a universal subject-prefix/object-suffix template.
The main translation risk is accepting ‘I gave it’ where source morphology encoded a recipient as well. Because the target sentence can remain grammatical with fewer participants, omitted agreement information is easy to miss. Conversely, literal expansion of every marker can produce duplicate noun phrases. The correct target participant inventory must match the source event rather than the number of overt source words.
A practical diagnostic is to compare source indexed-argument inventory with target participants. For example, a recipient encoded only on the verb must still reach the target meaning. Once the agreement bundle is understood, generate the target from semantic roles and discourse reference. Word order and overt pronouns can change; participant identity cannot.
Worked Examples
Worked Example 1: Subject and Object Indexed Together
Source problem: A source verb marks 1SG agent and 2SG patient with no independent pronouns.
Diagnosis: The morphology itself identifies I as actor and you as affected participant.
Target strategy: English needs a clause such as ‘I saw you’ or the relevant lexical predicate.
Verification: Both first- and second-person roles must survive.
Worked Example 2: Ditransitive Participant Indexing
Source problem: A verb indexes a giver and recipient while the theme is also overt or separately marked.
Diagnosis: The agreement inventory contains more than a subject/object pair.
Target strategy: English may need ‘I gave it to you.’
Verification: Recipient must not disappear simply because English verb morphology does not encode it.
Worked Example 3: Portmanteau Form
Source problem: One indivisible ending encodes 1→2 rather than separate 1st-person and 2nd-person affixes.
Diagnosis: The form represents the participant combination as one paradigm cell.
Target strategy: Translate the whole cell as ‘I [verb] you’ according to the lexical stem.
Verification: Do not invent a morpheme boundary unsupported by the source grammar.
Worked Example 4: AI Reverses Person Markers
Source problem: A model mistakes a 2nd-person object prefix for a 2nd-person subject marker.
Diagnosis: The source event direction has been reversed.
Target strategy: Decode agreement roles from the paradigm, then regenerate the target.
Verification: Actor and patient must match the source person features.
Practical Diagnostic Workflow
Create an AGREEMENT GRID: VERB ROOT; S/A PERSON+NUMBER; P PERSON+NUMBER; RECIPIENT/OTHER OBJECT; GENDER/CLASS; ALIGNMENT; DIRECTION/HIERARCHY; VOICE; TAM. Fill unknown fields explicitly rather than guessing. This turns a dense verb into a stable participant structure.
Then compare a small paradigm: change one participant at a time while holding the verb constant. The morpheme that changes reveals which feature it indexes. Portmanteau systems require whole-cell comparison rather than affix subtraction.
Verification Method
Read the target and list every participant, their semantic role, person and number. Compare with the source agreement grid. The translation passes only when every source-indexed participant is represented semantically and no target participant has been duplicated from both agreement and a full noun phrase.
Recovery Method
When the wrong person has been assigned, remove all target pronouns temporarily. Decode the source agreement paradigm first, rebuild a bare participant graph, then reinsert target pronouns/nouns according to discourse. When a participant was lost, restore its semantic role before polishing the clause.
Quality-Assurance Checklist
- Segment agreement morphology.
- Identify S, A and P indexing.
- Check alignment.
- Check hierarchy/direct–inverse interactions.
- Recognise portmanteau forms.
- Track person and number for every argument.
- Track gender/noun class where relevant.
- Separate agreement from clitic doubling.
- Separate agreement from noun incorporation.
- Check applicative/ditransitive participants.
- Check voice before agreement interpretation.
- Recover omitted pronouns from morphology.
- Do not over-explicitate duplicate NPs.
- Audit AI segmentation.
- Audit AI argument loss and reversal.
Frequently Asked Questions
What is polypersonal agreement?
Verbal agreement or indexing of features from more than one argument, often subject/agent plus object/patient and sometimes additional arguments.
Does every polypersonal language mark the same arguments?
No. Which participants are indexed and under what conditions is language-specific.
Is polypersonal agreement the same as pronoun incorporation?
Not necessarily. Agreement markers are grammatical indexes; noun/pronoun incorporation is a distinct morphological/syntactic phenomenon.
Can one affix encode two participants at once?
Yes. Portmanteau forms can express a person combination as one indivisible paradigm cell.
Can a verb index both A and P?
Yes. WALS documents many languages with verbal person marking of both A and P.
Why can nouns be omitted?
Rich agreement can make participant identity recoverable without an independent pronoun or noun phrase.
Can object definiteness affect the verb?
Yes in some languages; agreement/conjugation can be conditioned by object definiteness or specificity.
Why does AI fail on polypersonal verbs?
Dense morphology is easy to mis-segment, and models may default to target-language expectations that only subjects are indexed.
Where This Fits in the eduKate Translation Architecture
Route broad agreement to the existing Grammatical Gender and Agreement owner; participant frames to Verb Valency and Argument Structure; hierarchy-sensitive paradigms to Direct–Inverse Systems; alignment to the Ergative–Absolutive owner; complex morphology to Word Formation and Morphological Meaning.
Final Principle
A polypersonal verb can contain several people before the sentence contains several words. Decode every indexed participant first; only then let the target language decide how many pronouns, nouns and agreement markers it needs.