A long dependency cannot end in every possible embedded position. Syntacticians call a phrase or clause that resists extraction an island (academic). Hausa is especially revealing: gap-based wh-questions and focus fronting are sensitive to islands, while some relative-clause dependencies can cross the same boundary when a resumptive pronoun—or in restricted cases a recoverable zero pronoun—represents the relative head inside the island.
All unlabelled examples are (neutral Nigerian Standard Hausa). Ordinary boko omits tone and vowel length. Island judgments are among the most theory-sensitive judgments in grammar: processing load, dialect, animacy, and context all matter. The practical production rule is firm, however: keep wh-expressions local to embedded questions and relative clauses; use short clauses or explicit resumption instead of inventing an English-style extraction. Review the unrestricted baseline on Long-Distance Dependencies before applying the restrictions here.
A bridge complement is not automatically an island
First compare a neutral report with a productive long question:
Ka ce ma’aikatan sun ga Bello a tashar mota.
You said that the employees saw Bello at the bus station. (neutral S–TAM–V–O in the embedded clause)
Wa ka ce ma’aikatan suka gani a tashar mota?
Whom did you say the employees saw at the bus station? (long object question across a bridge complement)
The verb ce “say” can introduce a proposition through which the wh-dependency reaches the object gap after gani. Lower relative suka identifies the clause containing that gap. This well-formed bridge pattern establishes the baseline: distance alone is not the problem.
Embedded questions keep their wh-scope local
An embedded ko question has its own open value:
Ban san ko wa ya rubuta littafin ba.
I do not know who wrote the book.
Amina ta tambaya ko me yaran suka saya.
Amina asked what the children bought.
The wh-expression stays inside the ko domain, and the entire sentence may remain declarative. Trying to pull a new matrix wh-expression out of that already interrogative domain is not the ordinary Hausa pattern:
❌ Wane littafi ka san ko wa ya rubuta?
Unacceptable for the intended reading ‘Which book do you know who wrote?’; it extracts across an embedded wh-question.
Adding an object pronoun does not generally rescue a wh-question across this boundary. Hausa resumption is not a universal repair device for every construction. The clearest repair is to ask two questions or keep one wh-expression local:
✅ Ka san ko wa ya rubuta littafin? Wane littafi ne?
Do you know who wrote the book? Which book is it?
This is longer on paper but much clearer in speech. It gives each question one scope domain.
Relative clauses block wh-extraction
A relative clause already connects its head with a role inside the clause:
Na ga yarinyar da ta san Musa.
I saw the girl who knows Musa.
Attempting to ask for Musa by extracting a matrix wh-expression from inside that relative clause is unacceptable in the standard gap-based analysis:
❌ Wa ka ga yarinyar da ta san shi?
Unacceptable for the intended reading ‘Whom did you see the girl who knows?’; the wh-expression is extracted from a relative clause, and shi does not rescue it.
Ask locally or split the proposition:
✅ Ka ga yarinyar? Ta san wa?
Did you see the girl? Whom does she know? (two local questions)
The repair also prevents a scope ambiguity: the listener no longer has to decide whether wa is the object of ga or san.
Relative dependencies have a special resumptive option
Here Hausa does not treat all fronted dependencies alike. In advanced descriptions, a relative head may be related to a resumptive pronoun inside a relative-clause island:
Ga mutumin da ka ga yarinyar da ta san shi.
Here is the man such that you saw the girl who knows him. (marked long relative with resumptive shi)
The outer head mutumin is represented by shi inside the inner relative clause. This is (academic/marked) syntax used here to show the grammatical contrast; it is not the recommended everyday production model. A natural speaker usually divides such a heavily nested message:
Ga mutumin. Ka ga yarinyar da ta san shi.
Here is the man. You saw the girl who knows him. (clear everyday alternative)
Descriptions also report long relative dependencies involving a recoverable zero object when the intended referent is nonhuman. That pattern interacts with Hausa’s independent ability to omit a contextually recoverable nonhuman object. Because an inaudible resumptive and a true extraction gap look identical on the page, analyses differ over their formal representation.
Resumption depends on grammatical role
Possessors and complements of non-locative prepositions require overt resumptive material even in shorter dependencies. A bare preposition is not enough:
Ga mutumin da Amina ta yi magana da shi.
Here is the man whom Amina spoke with.
For direct objects, short focus and wh-fronting strongly favor a gap:
Wa Amina ta gani?
Whom did Amina see?
Descriptions of Hausa strongly prefer a gap in short direct-object wh-questions, though they do not categorically exclude every short resumptive. A form such as Wa Amina ta gan shi? is marginal and is not the reliable learner model. Inside an island-crossing relative dependency, by contrast, an overt human object pronoun can be the form that licenses the interpretation.
Adjuncts, subjects, and non-bridge complements
Advanced analyses also identify adjunct clauses, subject clauses, and complements of non-bridge predicates as extraction-resistant domains. Exact judgments become less stable when semantic plausibility and discourse improve, so this guide does not manufacture categorical example pairs where Nigerian Standard speakers may disagree. The production strategy is to retain the needed phrase locally:
Amina ta tafi bayan da ta gama rahoton.
Amina left after she finished the report.
If the speaker wants to ask about the report, the clear form is a separate local question:
Amina ta tafi bayan ta gama wane rahoto?
Amina left after finishing which report? (in-situ clarification; informal and context-dependent)
For an unmarked information-seeking question, divide the clauses: Amina ta gama wane rahoto? Yaushe ta tafi? This avoids forcing fronted extraction from an adjunct. The (informal) in-situ version is suitable as an echo or clarification question, not a blanket alternative to canonical fronted wh-syntax.
Negation does not dissolve an island
Negation can occur in either the matrix or embedded clause, but it does not change an embedded question or relative clause into an extraction corridor.
Ban san ko wa bai rubuta rahoton ba.
I do not know who did not write the report.
The outer ban ... ba negates knowing; lower bai ... ba negates writing. Ko wa remains local to the embedded question. Do not move rahoto outside and expect the two negative frames to identify its role.
Likewise, a polar question can contain an embedded question without extracting from it:
Ka san ko wa ya rubuta rahoton?
Do you know who wrote the report?
The whole sentence asks whether the addressee knows. The identity of the writer remains the open value in the embedded domain.
A short discourse
—Ban san ko wa ya rubuta wannan saƙon ba. —Ka tambayi Amina; ta ga mutumin da ya kawo shi. —Wa ta gani? —Kada mu cire tambaya daga cikin wannan doguwar jumla. Mu tambaye ta kai tsaye: ‘Kin ga wa?’ —Haka ya fi bayyana.
—I do not know who wrote this message. —Ask Amina; she saw the man who brought it. —Whom did she see? —Let’s not extract a question from inside that long sentence. Let’s ask her directly: ‘Whom did you see?’ —That is clearer.
The discourse keeps the embedded ko wa local and then replaces a potentially tangled dependency with a direct question. The relative da ya kawo shi retains shi as the ordinary object “it,” referring to the message; it is not a resumptive for mutumin.
Common Mistakes
1. Extracting from an embedded ko-question
❌ Wace mota ka san ko wa ya gyara?
Unacceptable intended extraction from an embedded wh-question.
✅ Ka san ko wa ya gyara motar?
Do you know who repaired the car?
2. Assuming a pronoun rescues a wh-island violation
❌ Wa ka san ko yaran suka gan shi?
Still unacceptable for the intended extraction: correctly formed suka gan shi does not make wh-extraction out of the embedded question grammatical.
✅ Ka san ko wa yaran suka gani?
Do you know whom the children saw?
3. Extracting a question from a relative clause
❌ Wa ka ga yaron da ya kira shi?
Unacceptable for the intended long reading out of the relative clause.
✅ Ka ga yaron? Ya kira wa?
Did you see the boy? Whom did he call?
4. Copying English preposition stranding
❌ Ga matar da Amina ta yi magana game da.
Incorrect bare preposition stranding.
✅ Ga matar da Amina ta yi magana game da ita.
Here is the woman whom Amina spoke about.
5. Omitting subject marking inside the embedded domain
❌ Ban san ko wa rubuta rahoton ba.
Incorrect—the embedded finite predicate lacks subject–TAM marking.
✅ Ban san ko wa ya rubuta rahoton ba.
I do not know who wrote the report.
Key Takeaways
- Ordinary bridge complements permit long dependencies, so linear distance itself is not an island.
- Embedded wh-questions and relative clauses block ordinary wh/focus extraction from inside them.
- An overt resumptive pronoun does not generally rescue a wh-question across an island.
- Some relative dependencies can cross islands through overt or recoverable resumption; this is a construction-specific, advanced exception.
- When judgments are heavy or variable, keep the wh-expression local or split the message into short, complete Hausa clauses.
Related Topics
- Long-Distance DependenciesC1 — Interpret Hausa wh-words, focused phrases, and relative heads whose grammatical role is realized inside an embedded clause.
- Coordination and ExtractionC1 — Front a whole Hausa coordinate phrase or build an across-the-board dependency without extracting from only one conjunct.
- Wh-Scope in Embedded QuestionsC1 — Keep Hausa wh-expressions inside the question they belong to, using ko and relative TAM to distinguish embedded information gaps from matrix questions and focused assertions.
- Object Relative ClausesA2 — Build Hausa relatives whose linked noun head is understood as the missing object of the modifying clause.
- Resumptive Pronouns in RelativesB1 — Know when a Hausa relative leaves a gap and when a pronoun or possessive suffix must preserve the head noun's grammatical relation.
- Multiple Wh-ExpressionsA2 — Ask about two unknowns in one Hausa question by fronting one wh-expression and leaving the other in its natural clause position.