Complementizer agreement in translation occurs when the element introducing an embedded finite clause carries person, number, gender or related agreement features associated with the embedded subject or another licensed nominal. Cambridge’s handbook treatment defines complementizer agreement as agreement between an embedded-clause complementizer and its subject, especially in Frisian and nonstandard Dutch/German varieties; newer work on Jordanian Arabic shows that complementizer agreement can also be obligatory with certain extracted left-peripheral nominals.
People searching for complementizer agreement, complementizer inflection, C agreement, subject agreement on complementizers, Frisian complementizer agreement, Dutch dialect agreement are solving an agreement-locus problem. The translator must identify which nominal controls the features, distinguish C-agreement from verbal agreement and clitics, and recover null or extracted subjects without turning agreement morphology into an extra pronoun.
This article owns the complementizer-agreement precision layer for eduKateSG. It does not replace Polypersonal Agreement, Reported Speech, Clitic Placement or Agreement Resolution. Its narrow task is to preserve the embedded clause’s subject features when grammar realizes them on the left-peripheral linker rather than—or in addition to—the finite verb.
Quick Read
Map four things: EMBEDDED CLAUSE, COMPLEMENTIZER, CONTROLLER NOMINAL and AGREEMENT FEATURES. Then ask whether the controller is overt, null, extracted or focused. Translate the embedded subject once; do not convert agreement morphology into another participant.
One-Sentence Answer
Translate complementizer agreement by preserving the controller’s person, number and gender while allowing the target language to drop the agreement morphology from the clause linker.
Why This Precision Layer Matters
Agreement on C is easy to ignore because English that/if do not inflect. It is also easy to overtranslate because the suffix can resemble a pronoun or clitic. In either direction, the risk is participant duplication or loss. The safe method is to treat C as an agreement locus, not as a new semantic argument.
What Complementizer Agreement Is
The first task is to identify the agreement controller. Complementizer agreement occurs when the element introducing an embedded finite clause carries agreement morphology related to the subject or another nominal in that clause. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is treating the complementizer as a fixed equivalent of that/if and ignoring inflection on it. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to identify the embedded clause boundary, then list the features carried by the complementizer. For example, a linker can encode plural, person or gender features matching an embedded subject. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
The Complementizer Is Not the Subject
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. The agreeing complementizer belongs to the clause’s left periphery and does not itself replace the embedded subject’s semantic role. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is translating agreement morphology as an extra pronoun or actor. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to separate CLAUSE LINKER from EMBEDDED SUBJECT. For example, the target may need only that + subject even when source C carries subject features. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Germanic Complementizer Agreement
A precise translation separates agreement locus from semantic participant count. Frisian and nonstandard varieties of Dutch and German are classic sites for complementizer agreement, especially person/number agreement with an embedded subject. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is assuming agreement on C is a spelling variant of the complementizer with no grammatical content. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to compare different subject persons under the same embedded-clause linker. For example, the complementizer form can vary when I, you or they is the embedded subject. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Jordanian Arabic Left-Periphery Agreement
The target may preserve every feature while deleting the source agreement site entirely. Recent work on Jordanian Arabic documents obligatory agreement inflection on an overt complementizer in certain A-bar dependencies, including extracted subjects and focused nominals. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is treating complementizer suffixes as object pronouns or ordinary clitics without checking the extraction structure. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to identify the left-peripheral nominal and the silent copy it leaves behind. For example, an extracted plural subject can require plural features on the embedded complementizer. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Person Features
Complementizer morphology should be interpreted through the embedded clause, not in isolation. Complementizer agreement can encode first, second or third person, though paradigms are often defective or restricted. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is assuming third-person morphology means the embedded subject is third person when the paradigm lacks other cells. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to read the full CA paradigm before resolving person. For example, a dialect may overtly distinguish only selected persons on the complementizer. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Number Features
The first task is to identify the agreement controller. Singular/plural distinctions are common in complementizer-agreement systems. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is dropping plurality carried by C when the embedded subject is omitted or silent. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to cross-check C with subject/agreement morphology. For example, plural complementizer marking can help recover an embedded plural subject. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Gender Features
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. Some systems or constructions can express gender on complementizers or complementizer-like elements. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is turning grammatical gender into lexical gender description. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to treat gender as agreement feature unless the referent’s sex is independently relevant. For example, feminine plural agreement on C can identify the feature bundle of an extracted nominal. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Defective Paradigms
A precise translation separates agreement locus from semantic participant count. Complementizer-agreement paradigms are often incomplete: some persons/numbers have special forms while others use a default or zero form. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is inventing a complete paradigm by analogy. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to mark unattested cells as unavailable/default. For example, absence of special morphology in one person does not mean agreement is absent from the system. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA vs Verbal Agreement
The target may preserve every feature while deleting the source agreement site entirely. Complementizer agreement and verbal subject agreement can coexist, match or differ. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is translating both agreement sites as duplicate pronouns. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to link both sets of features to one embedded subject unless grammar indicates otherwise. For example, C and the finite verb can both show plural features for the same subject. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA vs Subject Clitics
Complementizer morphology should be interpreted through the embedded clause, not in isolation. Subject clitics can occur near complementizers and may historically or synchronically resemble agreement morphology. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is classifying every form attached to C as either clitic or agreement without evidence. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to test mobility, doubling and paradigm behavior. For example, a fixed inflection on C behaves differently from a mobile subject clitic. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA vs Object Clitics
The first task is to identify the agreement controller. Some complementizers carry pronominal-looking material that can be confused with object clitics. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is assigning a patient/recipient role to agreement morphology. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to check which nominal controls the features. For example, left-peripheral subject agreement should not create a new object. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA and Pro-Drop
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. Complementizer agreement can interact with null subjects because agreement morphology contributes information about an unpronounced subject. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is adding an indefinite subject when source pro is recoverable. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to recover person/number from C plus verb before choosing target pronoun. For example, a null embedded subject can still be identifiable from complementizer morphology. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA and Subject Extraction
A precise translation separates agreement locus from semantic participant count. Extraction of an embedded subject can leave a silent lower copy while C shows its agreement features. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is treating the extracted wh/focus phrase and complementizer agreement as two referents. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to coindex extracted nominal, lower copy and C-features. For example, which boys… C-3PL …went represents one plural subject chain. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA and Wh-Movement
The target may preserve every feature while deleting the source agreement site entirely. Wh-subject extraction can trigger or require complementizer agreement in some systems. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is translating the C-agreement suffix as a separate pronoun after the wh phrase. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to map the wh phrase as controller of C. For example, the wh feature and the agreement feature belong to one dependency. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
CA and Focus Fronting
Complementizer morphology should be interpreted through the embedded clause, not in isolation. Focused nominals can similarly interact with agreement on C in languages where left-peripheral agreement is licensed. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is adding focus meaning to the agreement suffix itself. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to separate FOCUS position from phi-feature agreement. For example, the focus operation licenses agreement, but the suffix need not mean only/even. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Embedded Finite Clauses
The first task is to identify the agreement controller. CA is classically associated with embedded finite clauses introduced by complementizers. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is looking for complementizer agreement in every nonfinite infinitive automatically. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to identify finiteness before expecting the paradigm. For example, that-type finite complements can show CA where to-infinitives do not. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Main Clauses
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. Some languages extend complementizer-like particles or clause-edge agreement into root contexts, but classic CA analyses focus on embedded C. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is calling any clause-initial agreeing particle complementizer agreement. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to check whether the construction truly introduces a subordinate CP or root speech-act layer. For example, root particles need separate analysis unless they share the CA system. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Word Order
A precise translation separates agreement locus from semantic participant count. Agreement on C can be especially informative when embedded word order moves subjects, objects or topics. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is using linear adjacency to decide which noun controls C. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to use syntactic role and feature matching. For example, the closest noun to C need not be the agreement controller. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Adjacency Effects
The target may preserve every feature while deleting the source agreement site entirely. Some CA systems are sensitive to adjacency between the complementizer and subject, while others are not. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is treating an intervening phrase as proof agreement must fail universally. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to consult language-specific distribution. For example, one dialect may restrict CA to C+subject adjacency while another permits interveners. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Clause Left Periphery
Complementizer morphology should be interpreted through the embedded clause, not in isolation. CA provides evidence that agreement can be realised in the CP/left-peripheral domain rather than only on verbs. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is moving the morphology down onto the verb in analysis and losing the construction. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to record agreement locus explicitly. For example, the target may erase locus while preserving the same embedded subject. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Agreement with Objects
The first task is to identify the agreement controller. Recent research shows that in some left-peripheral configurations, complementizer agreement can target extracted objects under restricted conditions. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is assuming C can only ever agree with subjects. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to identify the A-bar chain and intervening subject conditions. For example, an extracted object can control C when structural conditions permit in a language-specific grammar. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Default Agreement
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. A complementizer may use a default feature bundle when no agreeing goal is available or a paradigm cell is missing. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is interpreting default morphology as a real third-person masculine participant. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to label DEFAULT unless an actual controller is present. For example, default 3MSG-like form need not introduce a male singular referent. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Feature Matching
A precise translation separates agreement locus from semantic participant count. When C agreement is present, mismatched person/number/gender can be ungrammatical or meaning-changing. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is normalising a mismatch during translation and missing a source agreement error. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to compare C, verb and subject features. For example, source proofreading may require preserving or flagging an agreement mismatch. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Dialect Translation
The target may preserve every feature while deleting the source agreement site entirely. Complementizer agreement is frequently dialectal or nonstandard in Germanic varieties. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is erasing dialect identity by treating C forms as typos. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to identify whether the target should preserve regional/nonstandard voice. For example, a dialect CA form can signal speaker origin even when English semantics is ordinary. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Sociolinguistic Meaning
Complementizer morphology should be interpreted through the embedded clause, not in isolation. Because CA can index regional varieties, it may contribute social identity beyond syntax. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is over-translating morphology into caricatured dialect spellings. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to separate grammatical meaning from sociolinguistic association. For example, target dialogue may preserve regionality through restrained register rather than invented eye-dialect. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Translation into English
The first task is to identify the agreement controller. English complementizers do not ordinarily inflect for embedded-subject phi-features, so CA morphology usually disappears overtly. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is adding extra English pronouns to translate every source agreement suffix. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to express the subject once through normal English syntax. For example, C+3PL subject agreement can become plain that they…. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Translation into Agreement-Rich Targets
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. A target language may have its own complementizer/clitic/agreement system but with different feature restrictions. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is mapping source C morphology directly to target C morphology by surface similarity. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to transfer embedded subject features and clause type first. For example, both languages may agree in the left periphery but with different persons or distributions. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Reported Speech and Complements
A precise translation separates agreement locus from semantic participant count. Speech/thought predicates create many embedded environments where CA can occur. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is treating C agreement as speaker/addressee marking rather than embedded-subject agreement. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to separate matrix speech participants from embedded clause subject. For example, she said [C-PL they left] still has embedded they as controller. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Legal and Formal Text
The target may preserve every feature while deleting the source agreement site entirely. Agreement on clause linkers can identify an embedded party even when overt nouns are omitted or extracted. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is dropping features and creating antecedent ambiguity. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to reconstruct the embedded nominal chain before formal target drafting. For example, plural CA can support a plural antecedent in a dense subordinate clause. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
AI Locus Confusion
Complementizer morphology should be interpreted through the embedded clause, not in isolation. Models can move complementizer features onto the finite verb or independent pronoun. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is preserving phi-features but attaching them to the wrong syntactic relation. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to record C as agreement locus during analysis. For example, target may omit C-agreement overtly but should not create an extra subject. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
AI Duplicate-Subject Error
The first task is to identify the agreement controller. AI can translate the controller noun and then separately translate C agreement as a pronoun. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is creating two subjects from one agreement chain. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to coindex morphology before generation. For example, which boys… C-PL should not become ‘which boys, they…’ unless target syntax independently requires it. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
AI Feature Loss
This construction belongs to the clause left periphery rather than ordinary lexical argument structure. When the embedded subject is null, models can ignore CA and lose person/number information. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is generating a vague they/one or omitting the subject entirely. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to recover phi-features from C and verb jointly. For example, agreement morphology can be the only overt cue to subject features. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Back-Translation Agreement-Locus Check
A precise translation separates agreement locus from semantic participant count. A final audit should recover which source element carries agreement and which nominal controls it. Complementizer-agreement systems are often dialect-specific and paradigmatically defective, so no universal C-agreement template should be assumed.
The main translation risk is checking only that target subject person/number is correct. Because agreement morphology looks pronominal, a target can accidentally duplicate a subject, or it can ignore the only overt cue to a null/extracted controller. A feature map prevents both errors.
A practical diagnostic is to write CONTROLLER + LOCUS + FEATURES + CLAUSE TYPE. For example, translation succeeds when subject identity survives even if target has no CA morphology. Once controller and features are secure, generate the target embedded clause through its own ordinary complementizer and agreement system.
Worked Examples
Worked Example 1: Dialectal Subject CA
Source problem: An embedded complementizer changes form when the subject changes from singular to plural.
Diagnosis: C carries subject number in addition to the clause link.
Target strategy: English uses plain that plus the plural subject.
Verification: Do not translate the C suffix as an extra they.
Worked Example 2: Null Embedded Subject
Source problem: The subject is omitted but C and verb morphology jointly show first-person plural.
Diagnosis: The controller is pro=we.
Target strategy: English normally requires ‘we’ overtly.
Verification: Recover we from agreement rather than inventing a generic subject.
Worked Example 3: Wh-Subject Extraction
Source problem: A plural wh-subject moves to the left periphery and the embedded complementizer shows plural agreement.
Diagnosis: The extracted wh phrase controls C through its dependency chain.
Target strategy: English asks the wh-question with ordinary complementizer/word order.
Verification: Do not create two plural subjects.
Worked Example 4: Default C Form
Source problem: The complementizer shows a 3MSG-like default in a context with no matching overt controller.
Diagnosis: The morphology may be default rather than a male singular referent.
Target strategy: Translate clause structure without introducing ‘he.’
Verification: A target pronoun should appear only if semantics independently requires one.
Practical Diagnostic Workflow
Create a C-AGREEMENT CARD: MATRIX PREDICATE; EMBEDDED COMPLEMENTIZER; EMBEDDED FINITENESS; CONTROLLER; CONTROLLER STATUS=OVERT/NULL/EXTRACTED/FOCUSED; PERSON; NUMBER; GENDER; VERBAL AGREEMENT; CLITICS; DIALECT. This separates overlapping agreement sites.
Run the controller-substitution test: change the embedded subject person/number in a known grammatical paradigm. If the complementizer changes, record which cells are distinct and which use defaults. Do not fill missing cells by analogy.
Verification Method
Back-read the target and identify one embedded subject with the correct feature bundle. Compare against source C and verb morphology. If the target has two subjects, or if a null source subject became featureless, revise. The target need not preserve C morphology itself.
Recovery Method
If agreement was duplicated as a pronoun, remove the extra referent and link the morphology back to its controller. If features were lost, reconstruct them from C/verb agreement before target generation. If dialect marking was flattened but socially important, compensate through restrained register rather than invented grammar.
Quality-Assurance Checklist
- Identify embedded clause and complementizer.
- Identify agreement controller.
- Separate C agreement from verbal agreement.
- Separate C agreement from clitics.
- Record defective/default paradigm cells.
- Track person, number and gender.
- Check null subjects.
- Check wh/focus extraction.
- Check adjacency and left-periphery restrictions.
- Check dialect/variety.
- Do not translate agreement as an extra participant.
- Do not interpret defaults as real referents.
- Preserve feature information from silent subjects.
- Audit AI locus confusion.
- Audit AI duplicate-subject and feature-loss errors.
Frequently Asked Questions
What is complementizer agreement?
Agreement morphology on the element introducing an embedded clause, commonly controlled by the embedded subject or another licensed nominal.
Where is it found?
Classic descriptions focus on Frisian and nonstandard Dutch/German varieties, with related phenomena documented elsewhere, including Jordanian Arabic.
Is complementizer agreement the same as verbal agreement?
No. Both can index the same subject, but they occur at different syntactic loci.
Does English have ordinary complementizer agreement?
No. English that/if generally do not inflect for the embedded subject’s person/number/gender.
Can C agree with an extracted subject?
Yes in some systems; extracted wh/focus nominals can control or license complementizer agreement.
Can C agreement help recover a null subject?
Yes. In some systems it contributes person/number information when the subject is unpronounced.
Are all complementizer forms complete paradigms?
No. Complementizer agreement paradigms can be defective or use defaults.
Why does AI mistranslate CA?
Agreement suffixes can look like clitic pronouns, causing duplicated participants, while null controllers can be lost if the morphology is ignored.
Where This Fits in the eduKate Translation Architecture
Route multi-argument verbal indexing to Polypersonal Agreement; reported/complement clauses to the Reported Speech and Complement Clauses owners; left-peripheral clitic placement to the live Second-Position Clitics owner; coordinated feature resolution to Agreement Resolution in Coordination.
Final Principle
Agreement can live on the clause linker without creating a new participant. Find the controller, recover its features, and translate the embedded clause once. The target may lose the agreement morphology completely while preserving the meaning perfectly.