Anti-agreement effects in translation occur when the agreement morphology normally expected with a subject is reduced, altered or replaced when that subject undergoes extraction, such as wh-movement, relativisation or focus movement. Classic descriptions include Berber and Northern Italian patterns in which an extracted subject no longer triggers ordinary subject agreement and the verb instead appears in a special participial or default form. The missing agreement is not evidence that the participant stopped being the subject.
People searching for anti-agreement effect, antiagreement, subject extraction, wh subject agreement, relative clause agreement, Berber anti-agreement are solving an extraction-and-role problem. A translator or parser that treats normal agreement as the primary proof of subjecthood can misclassify the extracted subject as an object or impersonal participant. Conversely, copying a default-looking source form literally can produce wrong target person/number agreement because the target language may require ordinary agreement even though the source suppresses it under extraction.
This article owns the anti-agreement precision layer for eduKateSG. It does not replace Grammatical Gender and Agreement, Relative Clauses, Questions and Interrogatives, Information Structure or Sentence Structure. Its narrow task is to preserve subject identity under extraction, distinguish true anti-agreement from ordinary default agreement or agreement failure, handle wh-questions, relatives and focus constructions, and rebuild a target where agreement may need to reappear even though the source deliberately suppresses it.
Quick Read
First reconstruct the ordinary non-extracted clause. Identify SUBJECT and its person, number and gender. Then compare the extracted construction. If ordinary subject agreement disappears specifically under subject extraction, record ANTI-AGREEMENT. Translate the grammatical role, not the reduced source agreement form.
One-Sentence Answer
Translate anti-agreement by keeping the extracted participant’s subject role intact even when source morphology suppresses the agreement that would normally reveal it.
Why This Precision Layer Matters
Anti-agreement is a diagnostic trap. The extracted subject may look less subject-like morphologically at precisely the point where English or another target requires overt subject agreement. A word-by-word system can therefore produce default third-singular verbs, wrong pronouns or object-like relative clauses. The remedy is to reconstruct the base clause before translating the extracted version.
What Anti-Agreement Is
The first step is to reconstruct the ordinary clause before reading the extracted form. Anti-agreement refers to the disappearance or alteration of regular agreement with an argument, classically a subject, when that argument undergoes A-bar extraction.
The main translation risk is assuming reduced agreement means the participant is no longer the subject. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to compare extracted and non-extracted versions of the same predicate. For example, ordinary subject agreement can vanish when the subject is wh-fronted Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve subject role across both forms. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Subject Extraction
This distinction becomes visible once subject role and agreement morphology are separated. The best-known anti-agreement effects are triggered by extracting the grammatical subject in questions, relatives or focus constructions.
The main translation risk is treating extraction as a role-changing operation. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to reconstruct the canonical clause before movement. For example, which woman saw Mohand? still has which woman as subject Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, keep target subject syntax. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Wh-Subject Questions
A fluent target can still fail if default source morphology is mistaken for real subject features. A wh-subject can trigger special morphology where a non-wh subject would trigger normal agreement.
The main translation risk is using source default morphology to choose target third-singular agreement. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to identify wh phrase features independently. For example, which girls came? may use a special source form with reduced number agreement Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, restore target plural where target grammar requires it. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Object Wh-Questions
The source agreement pattern is evidence about extraction, not a target-language template. Non-subject extraction often does not trigger the same anti-agreement effect, providing an important diagnostic.
The main translation risk is applying the anti-agreement rule to every wh-question. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to compare subject versus object extraction. For example, who did Lina see? extracts object and can retain subject agreement Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, limit the rule to the source-triggering configuration. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Relative Clauses
The safest method is to recover the base predicate frame before choosing target agreement. Subject relative clauses can show anti-agreement while object relatives keep ordinary agreement.
The main translation risk is attaching the relative head to the wrong grammatical role. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to expand the relative clause into a full declarative. For example, the woman who saw Lina has WOMAN as extracted subject Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, rebuild target relative syntax from the full clause. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Focus Movement
The first step is to reconstruct the ordinary clause before reading the extracted form. Focused subjects can trigger the same or related anti-agreement morphology in some languages.
The main translation risk is translating the special form as a change in tense or aspect. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to identify the information-structure trigger. For example, it was LINA who came can use extraction-related morphology Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve focus and subject role separately. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Default Agreement
This distinction becomes visible once subject role and agreement morphology are separated. Anti-agreement may surface as default person, number or gender morphology rather than complete absence of marking.
The main translation risk is reading default morphology as the actual features of the extracted subject. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to store subject features from the extracted phrase rather than the verb form. For example, plural extracted subject can occur with default singular morphology Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, generate target agreement from the subject. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Participial or Special Verb Forms
A fluent target can still fail if default source morphology is mistaken for real subject features. Some languages use a dedicated participial or invariant form under extracted-subject configurations.
The main translation risk is translating the form as a nonfinite clause simply because it resembles a participle. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to identify whether the construction is finite at clause level despite special morphology. For example, Berber-style anti-agreement forms can appear in finite wh or relative structures Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, translate clause finiteness correctly. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Berber Anti-Agreement
The source agreement pattern is evidence about extraction, not a target-language template. Berber languages provide classic examples where subject extraction blocks ordinary agreement and triggers a special participial form.
The main translation risk is importing one Berber language’s exact paradigm into all varieties. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to use the specific source grammar and extraction pattern. For example, a wh-subject may trigger participial morphology while a preverbal subject triggers full agreement Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve local grammar. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Northern Italian Patterns
The safest method is to recover the base predicate frame before choosing target agreement. Northern Italian dialects such as Fiorentino and Trentino have been discussed for subject anti-agreement under extraction.
The main translation risk is assuming the special default form signals an impersonal sentence. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to compare the extracted subject with the normal clitic and agreement paradigm. For example, a wh-fronted plural subject can be associated with default agreement Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, restore target subject number. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Agreement Loss vs Subject Loss
The first step is to reconstruct the ordinary clause before reading the extracted form. Morphological agreement can disappear while syntactic and semantic subjecthood remains unchanged.
The main translation risk is reassigning subject to object or oblique because the agreement cue vanished. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to use predicate argument structure and extraction position. For example, the person asked in who came? remains the mover and subject Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve event role. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Agreement Loss vs Impersonal
This distinction becomes visible once subject role and agreement morphology are separated. Impersonal constructions may also show default agreement but lack an ordinary referential subject, unlike anti-agreement structures.
The main translation risk is classifying every default verb as impersonal. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to check whether an extracted DP supplies the subject role. For example, who arrived? has an explicit semantic subject despite default morphology Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, do not add an impersonal one or they. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Agreement Loss vs Pro-Drop
A fluent target can still fail if default source morphology is mistaken for real subject features. Null subjects can leave agreement morphology as a subject cue; anti-agreement instead alters agreement when the subject is overtly extracted.
The main translation risk is assuming absent subject NP and reduced agreement are the same phenomenon. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to ask whether the subject is overtly represented in an extracted position. For example, a wh-subject is not missing; it is displaced Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, translate the extracted phrase as subject. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Person Features
The source agreement pattern is evidence about extraction, not a target-language template. Extracted first- or second-person subjects can interact with anti-agreement differently from third person depending on the system.
The main translation risk is taking default source morphology as the speaker or hearer features. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to recover person from the extracted DP or pronoun. For example, which of us came? contains first-person group reference even if source verb is invariant Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve person in target. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Number Features
The safest method is to recover the base predicate frame before choosing target agreement. Plural subjects remain plural even when source verbal number agreement disappears.
The main translation risk is using singular target agreement because the source verb looks singular or default. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to recover number from the extracted nominal. For example, which students arrived? requires plural semantic reference Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, use target agreement rules. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Gender Features
The first step is to reconstruct the ordinary clause before reading the extracted form. Source anti-agreement can neutralise gender morphology that ordinary agreement would otherwise show.
The main translation risk is inventing masculine or feminine reference from the default form. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to take gender only from the extracted phrase and context. For example, default masculine form does not necessarily make the wh-subject male Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, avoid unsupported gender. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Extraction Chains
This distinction becomes visible once subject role and agreement morphology are separated. A-bar movement creates a relation between the pronounced extracted phrase and a lower syntactic position or gap.
The main translation risk is translating the extracted phrase twice as both fronted NP and lower pronoun. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to build one dependency chain. For example, who_i did __i come? contains one participant Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, represent one subject in target. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Resumptive Pronouns
A fluent target can still fail if default source morphology is mistaken for real subject features. Some languages use resumptive pronouns in relatives or questions instead of or alongside gaps, potentially interacting with agreement.
The main translation risk is duplicating extracted subject and resumptive pronoun as two people. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to link the resumptive element to the extracted phrase. For example, the woman who she came-type source structure can encode one referent under its grammar Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, use natural target relative strategy. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Complementizer Interaction
The source agreement pattern is evidence about extraction, not a target-language template. Anti-agreement can correlate with special complementizers or clause-linking morphology in extracted-subject constructions.
The main translation risk is translating the complementizer as a lexical conjunction with extra meaning. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to treat it as part of extraction architecture. For example, subject relatives can select a different complementizer from object relatives Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, choose target relative or question structure. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Anti-Agreement and Anti-Locality
The safest method is to recover the base predicate frame before choosing target agreement. Recent syntactic work discusses anti-agreement alongside constraints on short subject extraction paths.
The main translation risk is turning a theoretical explanation into reader-facing lexical meaning. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to keep formal analysis separate from observable translation facts. For example, anti-agreement can serve as evidence in analyses of subject extraction Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, translate the sentence rather than the theory. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Tense Preservation
The first step is to reconstruct the ordinary clause before reading the extracted form. Special anti-agreement morphology does not necessarily remove tense from the clause.
The main translation risk is backshifting or making the target nonfinite because the source verb looks participial. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to identify tense elsewhere in the paradigm. For example, a subject question can still be past despite special agreement form Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve tense independently. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Aspect Preservation
This distinction becomes visible once subject role and agreement morphology are separated. Aspectual distinctions can remain active even when subject agreement is neutralised.
The main translation risk is dropping progressive or perfective meaning with the agreement morphology. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to decode TAM and agreement separately. For example, a default agreement form can still be perfective Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, restore target aspect. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Negation
A fluent target can still fail if default source morphology is mistaken for real subject features. Negative subject-extraction constructions preserve both negation and anti-agreement effects.
The main translation risk is letting unusual morphology obscure negation scope. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to bracket NOT over the predicate independently of extraction. For example, which students did not come? still has plural subject and negated coming Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve all layers. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Questions in Translation
The source agreement pattern is evidence about extraction, not a target-language template. Target languages may form wh-questions without the source extraction strategy.
The main translation risk is copying source word order and anti-agreement when target grammar uses in-situ wh or ordinary agreement. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to translate question function after recovering roles. For example, source wh-fronting can become target wh-in-situ Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, let target syntax handle extraction. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Relative Clauses in Translation
The safest method is to recover the base predicate frame before choosing target agreement. Target relative clauses may require overt agreement even where source subject relatives suppress it.
The main translation risk is preserving source default morphology in a target that demands full agreement. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to generate agreement from target relative-head features. For example, a plural relative head can require plural target verb Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, restore target grammar. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Focus Translation
The first step is to reconstruct the ordinary clause before reading the extracted form. Subject focus can be expressed through clefts, particles, stress or word order rather than extraction.
The main translation risk is dropping focus while fixing agreement. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to record FOCUS separately from SUBJECT ROLE. For example, only Lina came differs from Lina came Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, preserve information structure independently. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Parser and MT Failure
This distinction becomes visible once subject role and agreement morphology are separated. Automatic systems often rely heavily on agreement to identify subjects.
The main translation risk is misparsing the extracted subject as object once agreement disappears. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to force base-clause reconstruction before generation. For example, who saw Lina? should expand to X saw Lina Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, translate from the recovered predicate frame. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
AI Default-Form Bias
A fluent target can still fail if default source morphology is mistaken for real subject features. Models can copy the source default-looking singular form into the target.
The main translation risk is producing singular agreement with a plural extracted subject. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to derive target agreement from the extracted phrase. For example, which girls came requires plural semantics Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, regenerate target morphology. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
AI Role Reversal
The source agreement pattern is evidence about extraction, not a target-language template. When the source verb no longer agrees with the subject, models may prefer the nearest overt noun as actor.
The main translation risk is reversing who did what. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to expand every extracted clause into a canonical declarative with names. For example, which woman saw Mohand? becomes WOMAN saw MOHAND Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, translate from canonical roles. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
AI Over-Generalisation
The safest method is to recover the base predicate frame before choosing target agreement. After learning one anti-agreement pattern, a model can suppress agreement in object extraction or unrelated constructions.
The main translation risk is applying a local rule everywhere. Anti-agreement is deceptive because the construction can deliberately remove the very morphology a translator normally uses to identify the subject. A precision review therefore treats agreement loss, extraction, information structure and event roles as separate layers.
A practical diagnostic is to state the exact structural trigger and language variety. For example, subject extraction may trigger anti-agreement while object extraction does not Once the subject and extraction chain are secure, the target can use its normal wh-question, relative-clause or focus syntax, including ordinary agreement if the target requires it. The special source morphology may have no visible target counterpart.
For recovery, limit the repair to licensed contexts. Repair the base clause first, then regenerate the extracted target construction. Do not copy default person, number or gender morphology into the target unless the target grammar independently requires it.
Worked Examples
Worked Example 1: Wh-Subject with Default Verb
Source problem: A plural wh-subject triggers a default singular-looking source verb.
Diagnosis: The wh phrase is still the plural grammatical subject; source morphology is anti-agreement.
Target strategy: Use ordinary target plural subject agreement if required.
Verification: Do not infer singular from the source verb form.
Worked Example 2: Object Wh-Question
Source problem: A wh-object is extracted while the overt subject keeps normal agreement.
Diagnosis: The anti-agreement trigger is absent because the subject was not extracted.
Target strategy: Translate ordinary subject agreement and object wh-role.
Verification: Do not suppress agreement merely because the clause is interrogative.
Worked Example 3: Subject Relative
Source problem: A relative head corresponds to the subject of the relative clause and triggers a special source form.
Diagnosis: The relative head fills the lower subject role.
Target strategy: Use the target’s normal subject-relative construction, including agreement if required.
Verification: The head must remain actor or experiencer of the relative predicate.
Worked Example 4: AI Role Reversal
Source problem: A system sees default verb morphology and a nearby object noun, then treats that noun as subject.
Diagnosis: Agreement was an unreliable cue under extraction.
Target strategy: Expand to a canonical clause with named roles before regeneration.
Verification: Who did what must match the source despite the anti-agreement form.
Practical Diagnostic Workflow
Build an EXTRACTION CARD: BASE CLAUSE, BASE SUBJECT, SUBJECT FEATURES, EXTRACTED PHRASE, EXTRACTION TYPE, ORDINARY AGREEMENT FORM, EXTRACTED FORM, TAM, COMPLEMENTIZER, LOWER GAP or RESUMPTIVE. This shows exactly what changed and what did not.
Run the de-extraction test. Replace the wh, relative or focused phrase with an ordinary noun phrase in its canonical position. Recover the normal agreement paradigm. Then reapply extraction and identify which morphology disappears. Translate from the preserved subject features rather than the reduced form.
Verification Method
From the target, reconstruct the base clause and ask who is subject, who is object and what person, number and gender features the subject has. Compare tense, aspect and focus/question/relative function separately. The target is correct if source anti-agreement disappears without losing source roles.
Recovery Method
If agreement was copied incorrectly, restore the extracted subject’s features from the noun phrase, not the source verb. If roles were reversed, return to the base clause and regenerate the question or relative. If the target became nonfinite, restore clause TAM before choosing final syntax.
Practice Sequence
Create a normal clause “the girls saw Omar.” Derive a subject question, object question, subject relative and focused subject. Imagine a source language where only subject extraction triggers default agreement. Translate each into English and explain which source agreement difference should disappear in the target.
Quality-Assurance Checklist
- Reconstruct the base clause.
- Identify whether subject or non-subject is extracted.
- Preserve subject role under extraction.
- Recover subject person independently.
- Recover subject number independently.
- Recover gender independently.
- Do not treat default agreement as actual subject features.
- Keep anti-agreement distinct from impersonals.
- Keep anti-agreement distinct from pro-drop.
- Track gaps and resumptives as one dependency.
- Preserve tense and aspect.
- Preserve negation.
- Preserve focus, question and relative function.
- Audit AI default-form bias.
- Audit AI role reversal and over-generalisation.
Frequently Asked Questions
What is an anti-agreement effect?
A reduction, disappearance or alteration of expected agreement morphology when an argument—classically a subject—is extracted.
Does anti-agreement mean there is no subject?
No. The extracted phrase remains the semantic and syntactic subject even if ordinary agreement disappears.
Where does anti-agreement occur?
Classic discussions include Berber languages and Northern Italian dialects, among others.
Is every wh-question anti-agreement?
No. The effect is language- and construction-specific and often distinguishes subject extraction from object extraction.
Why can the source verb look singular with a plural subject?
The extraction construction can require a default or invariant form that does not reflect the subject’s ordinary agreement features.
How should the target language agree?
According to the target’s own grammar and the recovered subject features, not by mechanically copying source default morphology.
Is a participial-looking anti-agreement form nonfinite?
Not necessarily. It can occur in a finite clause under a special extraction paradigm.
Why does AI fail?
Agreement is a strong parsing cue, so when it disappears models may misidentify the subject or copy default morphology into the target.
Internal Routes Through the eduKate Translation Architecture
Route ordinary person, number and gender agreement to Grammatical Gender and Agreement: https://edukatesg.com/2026/09/18/translate-grammatical-gender-agreement-reference-gender-number-person-differently/. Route relative-clause mechanics to Relative Clauses Across Languages: https://edukatesg.com/2026/09/18/translate-precisely-relative-clauses-restrictive-supplementary-head-roles-attachment-resumptive/. Route question structure to the broader Sentence Structure owner: https://edukatesg.com/2026/09/18/translate-sentence-structure-word-order-grammar-meaning/. Route focus effects to Information Structure and Emphasis: https://edukatesg.com/2026/09/18/translate-precisely-information-structure-emphasis-topic-focus-contrast/.
Final Principle
Anti-agreement removes an agreement cue, not the subject itself. Reconstruct the base clause, preserve who the extracted participant is, and let the target language restore whatever agreement its own grammar requires.