Wh-in-situ and wh-fronting in translation are two major ways languages form content questions. In wh-fronting, a question word moves to the clause edge, as in standard English What did you buy?. In wh-in-situ, the wh-expression remains where the answer would normally appear, as in a structure equivalent to You bought what?. Current Cambridge research published in 2025 shows that Colloquial Singapore English can productively use both true wh-movement and true wh-in-situ, making the distinction especially relevant for Singapore English translation and dialogue.
People searching for wh-in-situ, wh-fronting, wh-movement, question word order, wh-scope, question particles, Colloquial Singapore English questions are solving a scope-and-syntax problem. A wh-word’s surface position does not by itself tell you whether the question is neutral, echoic, embedded or matrix-level. Languages can also combine in-situ wh-elements with question particles, case marking and rich agreement.
This article owns the wh-in-situ / wh-fronting precision layer for eduKateSG. It does not replace the existing Multiple Wh-Fronting and Superiority owner, Preposition Stranding, or general Question/Interrogative owner. Its narrow task is to preserve wh-role, interrogative scope, question particles, target variety and the distinction between neutral in-situ grammar and marked echo/repair questions.
Quick Read
For every wh-expression record ROLE, BASE POSITION, SURFACE POSITION, SCOPE CLAUSE, QUESTION TYPE, PARTICLE and REGISTER/VARIETY. Then rebuild the target question using its own fronting or in-situ strategy.
One-Sentence Answer
Translate wh-questions by preserving the information gap and scope first, then moving—or not moving—the wh-expression according to the target language and variety.
Why This Precision Layer Matters
A neutral source question can become an incredulous echo question if in-situ order is copied into standard English. A Colloquial Singapore English question can lose its local grammar if automatically fronted. Embedded scope can drift if a wh-word is moved across clause boundaries. Translation therefore needs an explicit question map.
What Wh-in-Situ Is
The first task is to identify the wh-expression’s semantic role and scope. Wh-in-situ leaves a question word such as who, what, where, when or how in the ordinary argument or adjunct position where its answer would appear. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is assuming a wh-word must move to the front because English usually fronts it. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to reconstruct the declarative base position of the wh-expression. For example, a source equivalent of ‘you bought what?’ can be an ordinary neutral wh-question in a wh-in-situ language. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
What Wh-Fronting Is
This construction should be analysed from the information gap outward rather than from word order alone. Wh-fronting places the interrogative expression near the clause edge, often sentence-initially, as in standard English ‘What did you buy?’ Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is treating fronting as semantic emphasis in every language. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to separate question formation from focus/emphasis. For example, fronting can be the ordinary grammatical strategy. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Movement vs Surface Position
A precise translation separates target question syntax from source wh-position. A wh-expression’s visible position and the theoretical analysis of scope are not identical questions. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is translating theoretical LF/movement assumptions as if they were lexical meaning. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to use surface syntax plus interpreted scope for translation. For example, two languages can express the same question with different visible positions. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Question Scope
The target language can move a wh-word without changing what is being asked. Wh-scope identifies the clause or proposition over which the question word takes interrogative force. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is letting an embedded wh-word take matrix scope or vice versa during restructuring. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to draw the clause boundaries and mark where the question is interpreted. For example, ‘I wonder [who came]’ differs from ‘Who do you wonder [came]?’. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Matrix Questions
Register and language variety determine whether in-situ order is neutral or marked. In a matrix wh-question, the interrogative force applies to the main clause. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is embedding the question accidentally because the wh-word remains low. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to check clause type before target word order. For example, in-situ position does not make the question subordinate. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Embedded Questions
The first task is to identify the wh-expression’s semantic role and scope. Wh-words inside embedded interrogatives can remain in situ or move according to the language. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is fronting an embedded wh into the matrix clause and changing who is asking what. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify the selecting verb and embedded CP boundary. For example, ‘I know what she bought’ is an embedded wh-clause, not a matrix question. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Wh-Question Particles
This construction should be analysed from the information gap outward rather than from word order alone. Many languages use question particles or sentence-final markers alongside wh-in-situ. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is translating the particle as an extra wh-word or lexical question verb. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify particle as clause-level interrogative marking. For example, a final particle can signal question force while what remains object-position. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Sentence-Final Particles
A precise translation separates target question syntax from source wh-position. Colloquial varieties, including Colloquial Singapore English, can combine wh-in-situ with sentence-final particles. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is dropping the particle’s interactional force or interpreting it as part of the noun phrase. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to separate interrogative syntax from discourse particle meaning. For example, ‘You buy what ah?’ can have both wh-in-situ and particle-based stance. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Wh-in-Situ in Colloquial Singapore English
The target language can move a wh-word without changing what is being asked. Current research shows Colloquial Singapore English can use true wh-in-situ as well as wh-fronting. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is correcting all in-situ questions into standard fronted English without considering register. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify whether the target is standard English, colloquial Singapore English or another variety. For example, ‘You want what?’ can be a genuine CSE structure, not merely learner error. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Wh-Optionality
Register and language variety determine whether in-situ order is neutral or marked. Some languages or varieties allow both fronting and in-situ strategies with differences in discourse conditions, syntax or register. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is treating both forms as free stylistic variants with identical use. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to compare question type, focus and context. For example, a variety may allow both ‘What you want?’ and ‘You want what?’ under different conditions. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Echo Questions
The first task is to identify the wh-expression’s semantic role and scope. English itself allows wh-in-situ readily in echo questions such as ‘You bought WHAT?’ Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is assuming every in-situ English wh-question is echo/repair. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to check whether the source variety has neutral in-situ questions. For example, Mandarin or CSE in-situ can be neutral where standard English in-situ sounds echoic. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Clarification Questions
This construction should be analysed from the information gap outward rather than from word order alone. Wh-in-situ can mark repair or clarification in languages that otherwise front wh-elements. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is normalising it into a neutral question and losing surprise/repair force. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify conversational trigger. For example, ‘She met who?’ after hearing an unclear name is different from neutral ‘Who did she meet?’. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Multiple Wh-Questions
A precise translation separates target question syntax from source wh-position. When more than one wh-expression appears, languages vary in whether all, one or none move visibly. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is moving all wh-words because one language does so. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify the language’s multiple-wh strategy. For example, English often fronts one wh while another remains in situ. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Superiority Effects
The target language can move a wh-word without changing what is being asked. Some wh-fronting languages restrict which wh-expression can move in multiple questions; this is distinct from simple in-situ syntax. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is choosing target wh-order solely from source linear order. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to apply the target’s wh-movement constraints. For example, ‘Who bought what?’ differs from awkward/incompatible alternatives in specific contexts. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Wh-Islands
Register and language variety determine whether in-situ order is neutral or marked. Extraction across embedded interrogative boundaries is restricted in many languages. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is fronting a wh-expression out of an island and creating ungrammatical target syntax. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to mark embedded question boundaries before movement. For example, target may need a paraphrase rather than direct extraction. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Adjunct Wh-Words
The first task is to identify the wh-expression’s semantic role and scope. Why, how, when and where can behave differently from argument wh-words. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is assuming all wh-items share one movement/in-situ distribution. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to classify argument versus adjunct. For example, why often has special scope and positional restrictions. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Who/What Arguments
This construction should be analysed from the information gap outward rather than from word order alone. Subject and object wh-expressions can have different surface behavior and auxiliary patterns in English. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is inserting do-support in subject questions or omitting it in object questions. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify grammatical role before target formation. For example, ‘Who came?’ versus ‘What did she buy?’. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Subject Wh-Questions
A precise translation separates target question syntax from source wh-position. English subject wh-questions generally do not use ordinary do-support because the wh-word itself is subject. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is copying object-question syntax and producing ‘Who did come?’ in neutral context. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to label WH=SUBJECT before generation. For example, source in-situ subject questions can require major restructuring into English. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Object Wh-Questions
The target language can move a wh-word without changing what is being asked. English object wh-questions typically front the wh-word and use auxiliary inversion/do-support when needed. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is leaving target object wh in situ in formal standard English unintentionally. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to apply target question syntax after preserving object role. For example, ‘What did Lina buy?’ maps from many in-situ source orders. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Prepositional Wh-Phrases
Register and language variety determine whether in-situ order is neutral or marked. Wh-elements governed by adpositions can strand or pied-pipe in English, while other languages keep adposition+wh together or case-mark the wh-word. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is moving the wh-word and losing the adpositional relation. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to reconstruct governor + adposition + wh complement. For example, source ‘talk with who?’ can map to ‘Who did you talk with/to?’ or ‘With whom…?’. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Case-Marked Wh-Words
The first task is to identify the wh-expression’s semantic role and scope. Case-rich languages can show wh-role through morphology even when the wh-word stays in place. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is ignoring case because target English will front the wh-word. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to decode source role before movement. For example, same wh stem can take nominative, accusative, dative or other case. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Wh-Agreement
This construction should be analysed from the information gap outward rather than from word order alone. Question words can trigger or interact with agreement on verbs, complementizers or particles. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is translating agreement morphology as extra pronouns. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to link features to the wh-expression/controller. For example, a plural wh-subject can control plural agreement even when extracted or in situ. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Wh-Indefinite Syncretism
A precise translation separates target question syntax from source wh-position. In many languages, forms meaning who/what can also function as indefinites under particles, quantifiers or polarity environments. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is translating every wh-shaped form as a question. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to check clause force and licensing particles. For example, the same form can mean who? in one construction and someone/anyone in another. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Free-Choice Wh-Derivatives
The target language can move a wh-word without changing what is being asked. Wh-bases can form whoever/whatever/any-type expressions. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is mistaking free-choice forms for direct questions. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify morphology/particles that change quantification. For example, whoever came is not a question. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Focus and Wh
Register and language variety determine whether in-situ order is neutral or marked. Wh-elements are intrinsically information-seeking and often interact with focus prosody or syntax. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is adding contrastive focus markers simply because the wh-word is prominent. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to separate interrogative focus from corrective contrast. For example, neutral questions need not sound accusatory. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Prosody
The first task is to identify the wh-expression’s semantic role and scope. In-situ questions can rely heavily on intonation and particles to signal question force. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is flattening prosody in subtitle/script translation and making an in-situ clause look declarative. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to use punctuation/particles/word order appropriate to target. For example, ‘You saw who?’ may need rising or marked intonation in standard English. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Register and Variety
This construction should be analysed from the information gap outward rather than from word order alone. Wh-in-situ can be neutral in one language or variety and highly marked in another. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is copying source position into target without register check. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to classify target variety first. For example, Mandarin neutral in-situ cannot simply become neutral standard-English in-situ. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Translation into Standard English
A precise translation separates target question syntax from source wh-position. Standard English usually fronts ordinary wh-phrases in matrix questions. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is preserving source in-situ position and making a neutral question sound incredulous or echoic. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to move the wh-expression while preserving role/scope. For example, source SVO-wh often becomes Wh-Aux-S-V. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Translation into CSE or Other In-Situ Varieties
The target language can move a wh-word without changing what is being asked. When the target intentionally uses Colloquial Singapore English or another in-situ variety, fronting everything can erase local syntax and voice. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is standardising a genuine variety unnecessarily. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to preserve licensed in-situ forms according to target norms. For example, CSE dialogue may naturally retain low wh-position. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Translation into Wh-Fronting Languages
Register and language variety determine whether in-situ order is neutral or marked. A wh-in-situ source can require overt movement in the target. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is moving the wh-item but leaving case/adposition/auxiliary syntax unresolved. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to rebuild the full target interrogative. For example, movement can trigger inversion, do-support and preposition choices. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Translation into Wh-in-Situ Languages
The first task is to identify the wh-expression’s semantic role and scope. A fronted English question can map to ordinary declarative-like word order plus wh-word and/or particle. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is fronting the wh-expression because it is fronted in English. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to place it in the target argument/adjunct position. For example, English ‘What did you buy?’ can become ‘you bought what?’ structurally. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Subtitles and Dialogue
This construction should be analysed from the information gap outward rather than from word order alone. In-situ forms can convey surprise, repair, local variety or conversational rhythm. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is normalising them into formal questions and flattening character voice. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to decide whether markedness is grammatical or pragmatic. For example, ‘You did what?’ is more expressive than neutral ‘What did you do?’. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Legal and Interview Questions
A precise translation separates target question syntax from source wh-position. Formal interviews require clear scope and role assignment, especially with embedded questions. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is preserving ambiguous in-situ forms where target legal English needs explicit structure. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to front/restructure for unambiguous target scope. For example, ‘You said he paid who?’ may require careful parsing before formalisation. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
AI Standardisation Bias
The target language can move a wh-word without changing what is being asked. Models often rewrite all in-situ wh-questions into standard English fronting, erasing variety or echo force. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is treating every nonstandard-looking form as an error. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to identify source variety and discourse function before normalisation. For example, CSE wh-in-situ can be grammatical and meaningful. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
AI Scope Drift
Register and language variety determine whether in-situ order is neutral or marked. During restructuring, AI can move a wh-word across clause boundaries and change which clause is questioned. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is preserving wh vocabulary while changing question scope. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to mark clause boundaries before movement. For example, embedded who must stay semantically embedded unless source asks a matrix extraction question. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
AI Particle Loss
The first task is to identify the wh-expression’s semantic role and scope. Question particles can be dropped because the wh-word already looks interrogative to the model. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is losing register, stance or clause-force cues. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to record particle function separately. For example, a final particle can add pragmatic force beyond wh semantics. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Back-Translation Question Check
This construction should be analysed from the information gap outward rather than from word order alone. A final audit should reconstruct wh-role, scope, clause type and register from the target. Wh-systems vary sharply across languages and even within English varieties, so surface order should never be normalised before clause scope and discourse function are known.
The main translation risk is checking only that a question word remains. Because the same wh-word survives, a scope or register error can be hard to notice. The safest workflow labels role, clause and question type before target movement/inversion is applied.
A practical diagnostic is to write WH-ROLE + SCOPE + CLAUSE + STRATEGY + REGISTER. For example, translation succeeds when the same information gap is asked in the same discourse mode. Once the question map is stable, apply the target’s wh-placement, auxiliary, particle, case and adposition rules.
Worked Examples
Worked Example 1: Neutral Wh-in-Situ
Source problem: Source language asks a neutral object question with S-V-WH order.
Diagnosis: The wh-word is object and has matrix scope.
Target strategy: Standard English: ‘What did you buy?’
Verification: Do not preserve ‘You bought what?’ unless the target intends echo/surprise or an in-situ variety.
Worked Example 2: CSE Wh-in-Situ
Source problem: A Colloquial Singapore English speaker asks a neutral/variety-licensed in-situ question.
Diagnosis: The low wh-position is part of target variety grammar.
Target strategy: Preserve CSE form if localisation target is CSE.
Verification: Do not ‘correct’ the variety into standard English automatically.
Worked Example 3: Embedded Wh
Source problem: Source means ‘I wonder who came.’
Diagnosis: Who has scope only inside the embedded question.
Target strategy: Keep it embedded in target syntax.
Verification: Do not produce a matrix question ‘Who do I wonder came?’
Worked Example 4: Wh + Particle
Source problem: Source wh-word stays in place and a final particle marks interrogative/discourse force.
Diagnosis: Wh-role and particle function are separate.
Target strategy: Use target question syntax plus an appropriate discourse cue if needed.
Verification: Do not translate the particle as another wh-word.
Practical Diagnostic Workflow
Create a WH CARD: WH FORM; ARGUMENT/ADJUNCT ROLE; BASE POSITION; SURFACE POSITION; MATRIX/EMBEDDED SCOPE; ECHO/NEUTRAL/REPAIR; QUESTION PARTICLE; ADPOSITION/CASE; TARGET VARIETY. This prevents standardisation from preceding analysis.
Run the answer-slot test: replace the wh-expression with a plausible answer. The resulting declarative shows the base grammatical role. Then restore the wh-form using the target language’s question strategy.
Verification Method
Back-read the target and answer it. The expected answer type and clause must match the source. Check whether the target sounds neutral, echoic, formal, colloquial or regional in the intended variety. Scope and register should both align.
Recovery Method
If scope drifted, return the wh-expression to its source clause semantically before rebuilding target syntax. If standardisation erased a meaningful variety, restore licensed in-situ grammar. If a neutral source became an echo question, front the wh-word or adjust particles/prosody according to target norms.
Quality-Assurance Checklist
- Identify wh grammatical role.
- Identify base position.
- Identify surface position.
- Mark matrix vs embedded scope.
- Distinguish neutral from echo/repair questions.
- Check subject vs object question syntax.
- Check adjunct wh behavior.
- Check question particles.
- Check case and adposition dependencies.
- Distinguish wh-indefinites/free-choice forms.
- Respect target language variety.
- Check multiple-wh interactions without duplicating existing owner.
- Preserve embedded interrogatives.
- Audit AI standardisation bias.
- Audit AI scope drift and particle loss.
Frequently Asked Questions
What is wh-in-situ?
A question strategy where the wh-expression stays in its ordinary argument or adjunct position rather than moving to the clause edge.
What is wh-fronting?
A strategy where the wh-expression appears at or near the front of the interrogative clause.
Are all English in-situ wh-questions echo questions?
No. Standard English strongly associates many in-situ forms with echo/repair, but varieties such as Colloquial Singapore English can license true wh-in-situ more broadly.
Does Mandarin use wh-in-situ?
Yes; Mandarin is a classic wh-in-situ language.
Can one language use both strategies?
Yes. Colloquial Singapore English is a well-studied case where wh-fronting and wh-in-situ coexist.
Why do embedded questions matter?
Moving a wh-expression across the wrong clause boundary changes which proposition is being questioned.
Do question particles matter if there is already a wh-word?
Yes. They can mark clause force, discourse stance or variety-specific interaction.
Why does AI mistranslate wh-in-situ?
Models often standardise toward familiar fronted English or move wh-words without preserving scope and local variety.
Where This Fits in the eduKate Translation Architecture
Route multiple wh-movement interactions to the live Multiple Wh-Fronting and Superiority owner; adposition movement to Preposition Stranding and Pied-Piping; complementizer/extraction agreement to Complementizer Agreement; broad clause reordering to the Sentence Structure owner.
Final Principle
Question meaning lives in the information gap and its scope, not in one universal word order. Preserve what is being asked, where the question takes scope and what register the source uses—then let the target decide whether the wh-word moves.