A wh-question does something dramatic: it takes a phrase from deep inside a sentence and hurls it to the front, leaving a gap where it belongs. In Hvað keyptirðu? "what did you buy?", hvað "what" is understood as the object of keyptir "bought" — it started after the verb and moved to slot 1. This page is about the long-distance and constrained versions of that operation: how a wh-phrase can move out of an embedded clause entirely (Hvað heldurðu að hann hafi keypt? "what do you think he bought?"), where such movement is blocked (the islands — relative clauses, adjuncts, coordinate structures), where a resumptive pronoun can rescue an otherwise-illegal extraction, and — the point only Icelandic can make vivid — how the moved wh-phrase carries its case across the whole distance, so you can see the gap's case sitting on the fronted word. (Basic wh-question formation is on questions/wh-questions; the case-survives-displacement theme runs through complex/case-preservation. We build on both and link out rather than re-derive.)
Long-distance wh-movement: extracting from an embedded clause
The headline fact is that a wh-phrase need not come from the main clause — it can be the argument of a verb two clauses down and still front to the top of the sentence. In Hvað heldurðu að hann hafi keypt?, hvað is the object of keypt "bought," buried inside the embedded að-clause, yet it surfaces at the very front of the whole question. The movement crosses the clause boundary, and the verb halda "think" leaves a finite að-complement (with the subjunctive hafi, the mood of an embedded report).
Hvað heldurðu að hann hafi keypt?
What do you think he bought? — long-distance extraction: hvað is the object of the embedded 'keypt' but fronts to slot 1 of the whole sentence; halda takes að + subjunctive hafi. The clitic heldurðu (heldur + þú) shows V2.
Hvar sagði Anna að hún hefði fundið lyklana?
Where did Anna say she had found the keys? — hvar ('where') extracts from the embedded clause (modifying 'fundið'); segja að + past subjunctive hefði. The wh-phrase has crossed a clause boundary.
Hvenær heldurðu að þau komi heim?
When do you think they'll come home? — hvenær ('when') extracted from the embedded að-clause; komi is subjunctive under halda.
This is the same operation as a short question, just stretched: a wh-phrase fronts and a gap is left behind, but here the gap is inside a subordinate clause. Icelandic does this freely out of ordinary finite að-complements — more freely than careful English, which often prefers to hedge such long extractions. (Be careful with one common claim, though: it is the mainland Scandinavian languages — Swedish, Danish, Norwegian — that are famous for liberal extraction out of islands like relative clauses; Icelandic is actually more conservative there, as we will see below. The two are alike in free complement-clause extraction, not in island extraction.)
The litmus of Icelandic: the wh-phrase carries the gap's case
Now the point that makes Icelandic uniquely illuminating, and the distinguishing insight of this page. When a wh-phrase moves, it leaves a gap — and the case the wh-phrase wears is the case assigned at the gap, by the verb or preposition that governs that position, even when the gap is clauses away. The fronted word is, in effect, carrying a receipt for the case it was given downstream. English cannot show this (its wh-words barely inflect), but Icelandic spells it out on the surface.
The cleanest demonstration uses a dative-governing verb. Hjálpa "help" takes a dative object (ég hjálpa honum "I help him," dative honum). Question that object across a clause boundary, and the wh-word surfaces as hverjum — the dative of hver "who" — because that is the case hjálpa assigns at the gap, transmitted all the way up:
Hverjum sagðirðu að ég ætti að hjálpa?
Who(m) did you say I should help? — hverjum is DATIVE because the gap is the object of hjálpa, which governs the dative — and that case is carried across the clause boundary to the fronted wh-word. Not nominative hver.
Hverjum heldurðu að hún treysti best?
Who(m) do you think she trusts most? — treysta ('trust') also governs the DATIVE, so the extracted wh-word is dative hverjum, the gap's case shown on the front of the sentence.
Contrast a verb that assigns plain accusative: þekkja "know (a person)" governs the accusative (ég þekki hann "I know him"). Question its object long-distance and the wh-word is hvern, the accusative of hver:
Hvern heldurðu að hún hafi hitt á fundinum?
Who(m) do you think she met at the meeting? — hvern is ACCUSATIVE because the gap is the object of hitta ('meet'), an accusative-governing verb; the accusative is transmitted to the fronted wh-phrase.
Hvern sagði Jón að lögreglan væri að leita að? — eða réttara: Að hverjum sagði Jón að lögreglan væri að leita?
Who(m) did Jón say the police were looking for? — leita að governs the dative via the preposition, so the careful form pied-pipes the preposition: Að hverjum ... (dative hverjum after að). The gap's prepositional case shows on the front.
So the rule is exact: read the gap's governor — the verb or preposition that would have taken the wh-phrase as its argument — and put the wh-word in the case that governor assigns. hjálpa, treysta → dative hverjum; þekkja, hitta → accusative hvern; a subject gap → nominative hver. The case is a long-distance dependency, fixed downstream and worn upstream.
Hver heldurðu að hafi gert þetta?
Who do you think did this? — SUBJECT gap, so nominative hver; the gap is the subject of 'gert', a nominative position. Compare the object cases above: the case always tracks the gap's role.
Islands: where extraction is blocked
Movement is free in principle but not everywhere: certain syntactic configurations are islands — structures you cannot extract out of. The wh-phrase is trapped inside; trying to front it from there yields ungrammaticality. The major islands hold robustly in Icelandic. (This is the point where Icelandic and mainland Scandinavian part ways: Swedish, Danish, and Norwegian are famous for tolerating some extraction out of relative clauses and adjuncts, but Icelandic is comparatively strict about these islands — so don't assume an extraction that works in Swedish will work here.) Three classic islands:
Complex-NP island: no extraction from a relative clause
You cannot extract a wh-phrase out of a relative clause modifying a noun (the "Complex Noun Phrase Constraint"). The relative clause is sealed.
❌ Hvað þekkirðu manninn sem keypti?
Island violation — trying to extract hvað ('what') out of the relative clause 'sem keypti [gap]' (modifying manninn 'the man'). A relative clause is a Complex-NP island; you cannot front out of it.
✅ Hvað keypti maðurinn (sem þú þekkir)?
What did the man (whom you know) buy? — ask about the object directly in its own clause; don't try to extract across the relative-clause boundary.
Adjunct island: no extraction from an adverbial clause
You cannot extract from inside an adjunct — an adverbial clause of reason, time, condition (af því að "because," þegar "when," ef "if"). Adjuncts are islands.
❌ Hvað fór hún heim af því að hún keypti?
Island violation — extracting hvað out of the reason-adjunct 'af því að hún keypti [gap]' ('because she bought ...'). Adjunct clauses are islands; the wh-phrase can't escape.
✅ Hún fór heim af því að hún keypti eitthvað. Hvað keypti hún?
She went home because she bought something. What did she buy? — keep the extraction inside its own clause; split into two sentences rather than extract from the adjunct.
Coordinate-structure island: no extraction from one conjunct
You cannot extract one element out of a coordinate structure (X og Y), pulling a wh-phrase out of just one conjunct.
❌ Hvað keypti hún brauð og?
Island violation — extracting hvað from one half of the coordination 'brauð og [gap]' ('bread and ...'). The Coordinate Structure Constraint blocks pulling one conjunct out.
✅ Hvað keypti hún, annað en brauð?
What did she buy, besides bread? — rephrase to avoid extracting out of a conjunct.
The wh-island: extraction across another wh is degraded
Extracting a wh-phrase across an already-filled wh-position (an embedded question) is heavily degraded — a wh-island violation. The embedded hvort "whether" or hvar "where" occupies the lower front position, and a second wh trying to cross it struggles.
?? Hvað spurði hún hvort hann hefði keypt?
Wh-island — extracting hvað across the embedded question 'hvort hann hefði keypt [gap]' (whether he had bought ...). The embedded wh hvort blocks the path; this is strongly degraded.
✅ Hún spurði hvort hann hefði keypt eitthvað. Hvað var það?
She asked whether he had bought anything. What was it? — don't extract across the embedded wh; ask separately.
Resumptive pronouns: rescuing an island
Here Icelandic shows a striking option that English largely lacks in standard speech. When extraction would cross an island, you can sometimes leave a pronoun in the gap position instead of a silent gap — a resumptive pronoun — and the structure becomes acceptable. The pronoun "holds the place," so nothing has truly moved out of the island; the wh-phrase is linked to a real pronoun inside. This is most natural with relativisation (using sem) into an island, especially a possessor or an object inside a relative clause, and the resumptive pronoun carries the case the gap position demands.
Þetta er maðurinn sem ég veit ekki hvar hann býr.
This is the man that I don't know where he lives. — the relative crosses an embedded wh-island ('hvar [gap] býr'), rescued by the resumptive pronoun hann ('he') holding the subject slot inside. A silent gap would be illegal here.
Hún er stúlkan sem allir muna eftir deginum þegar hún kom.
She's the girl that everyone remembers the day when she came. — relativisation into an adjunct/complex-NP island, rescued by the resumptive hún inside the adjunct clause.
Þarna er bókin sem ég þekki manninn sem skrifaði hana.
There's the book that I know the man who wrote it. — relativisation across a Complex-NP island (relative-within-relative), rescued by the resumptive hana ('it', accusative, the case 'skrifa' assigns its object) inside.
Notice the case on the resumptive: in the last example hana is accusative, because skrifa "write" assigns accusative to its object, and the resumptive sits in that object position — so it wears that position's case directly (no long-distance transmission needed, since the pronoun is right there in the gap). Resumptives are more at home in relatives than in information questions (a wh-question with a resumptive, ?Hvern þekkirðu manninn sem skrifaði hann?, is marginal), and they range from fully natural to colloquial-rescue depending on the island; treat them as the escape hatch when a silent gap is impossible.
English vs Icelandic: more liberal, and case-overt
Two systematic differences. First, Icelandic is freer than English with long extraction out of ordinary complement clauses — the að-clause under halda/segja/telja is no barrier, and English speakers tend to be over-cautious, declining extractions that Icelandic permits, and unaware of the resumptive escape hatch. (Note the often-mangled comparison: it is mainland Scandinavian, not Icelandic, that is liberal about extracting out of the islands themselves — relative clauses and adjuncts. Icelandic is conservative there; its extra freedom is in complement-clause extraction, not island-busting.) Second, and the headline contrast: Icelandic marks case on the moved wh-phrase, English does not. English "who/whom" is largely collapsed in speech, and "what/which/where" never inflect, so the long-distance case dependency is invisible in English. The English learner's signature error is exactly that — failing to case-mark the extracted wh-phrase, producing nominative *Hver sagðirðu að ég ætti að hjálpa? where the gap's governor hjálpa demands the dative Hverjum. The fix is the diagnostic question that recurs through Icelandic syntax: look at the gap, find what governs it, and let the case ride up to the front.
Common Mistakes
❌ Hver sagðirðu að ég ætti að hjálpa?
Case error — the gap is the object of hjálpa, which governs the DATIVE, so the fronted wh-phrase must be dative hverjum, not nominative hver. The case rides up from the gap.
✅ Hverjum sagðirðu að ég ætti að hjálpa?
Who(m) did you say I should help? — dative hverjum, the case hjálpa assigns at the gap, carried to the front.
The signature English-speaker error: leaving the extracted wh-phrase in the default nominative instead of the case its gap's governor assigns.
❌ Hvern heldurðu að hafi gert þetta?
Case error — the gap is the SUBJECT of 'gert' (a nominative position), so the wh-word must be nominative hver, not accusative hvern. Always read the gap's role.
✅ Hver heldurðu að hafi gert þetta?
Who do you think did this? — subject gap → nominative hver.
Case tracks the gap's role: a subject gap takes nominative even though other long-distance questions take dative/accusative. Don't default to an oblique case either.
❌ Hvað þekkirðu manninn sem keypti?
Island violation — you can't extract out of a relative clause (Complex-NP island). The wh-phrase is trapped inside 'sem keypti [gap]'.
✅ Hvað keypti maðurinn sem þú þekkir? / Þetta er það sem maðurinn keypti.
What did the man you know buy? / This is what the man bought. — keep the question inside the clause, or relativise instead of extracting.
Relative clauses are islands; rephrase so the wh-phrase is not pulled out of one.
❌ Hvað fórstu heim af því að þú keyptir?
Adjunct island — you can't extract hvað out of the reason-clause 'af því að þú keyptir [gap]'. Adjuncts seal off their contents.
✅ Þú fórst heim af því að þú keyptir eitthvað — hvað var það?
You went home because you bought something — what was it? — split into two clauses rather than extract from the adjunct.
Adjunct (reason/time/condition) clauses are islands; do not extract from inside them.
❌ Þetta er maðurinn sem ég veit ekki hvar býr.
Island without rescue — relativising across the wh-island 'hvar [gap] býr' needs a resumptive pronoun to hold the subject slot; a silent gap is illegal here.
✅ Þetta er maðurinn sem ég veit ekki hvar hann býr.
This is the man that I don't know where he lives. — the resumptive hann rescues the island; nothing illegally left it.
Some island-crossing relatives are only saved by a resumptive pronoun in the gap; omitting it leaves an illegal silent gap inside the island.
Key Takeaways
- Wh-movement is unbounded in principle: a wh-phrase can extract from an embedded clause (Hvað heldurðu að hann hafi keypt?), even several clauses down, fronting to the top and leaving a gap. Embedded clauses under halda/segja/telja take að
- subjunctive.
- The Icelandic litmus: the fronted wh-phrase wears the case of its gap, fixed by whatever verb or preposition governs the extraction site, across any distance — hjálpa/treysta (dative) → hverjum, þekkja/hitta (accusative) → hvern, subject gap → nominative hver, under a preposition pied-pipe it (Að hverjum …). Always ask "what governs the gap?"
- Extraction is blocked by islands: relative clauses (Complex-NP), adjuncts (reason/time/condition), coordinate structures, and across an embedded wh (wh-island). Don't force these — rephrase or split.
- Icelandic offers resumptive pronouns to rescue island-crossing relatives (sem … hann býr, sem … skrifaði hana): a real pronoun in the gap holds the position so nothing illegally leaves the island, and it carries the slot's case. Most natural in relative clauses, marginal in plain wh-questions.
- English vs Icelandic: Icelandic is freer than English with long extraction out of complement clauses (so English speakers are over-cautious) — but, unlike mainland Scandinavian, it is conservative about island extraction, so don't over-generalise the "Scandinavian is liberal" reputation to Icelandic. And Icelandic marks case on the moved wh-phrase where English shows nothing — the signature error is failing to case-mark the extracted wh (*Hver … að hjálpa for Hverjum … að hjálpa).
Now practice Icelandic
Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.
Start learning Icelandic→Related Topics
- Wh-Questions: hvað, hver, hvar, hvenær, af hverjuA2 — The Icelandic question words — hvað, hver, hvar/hvert/hvaðan, hvenær, hvernig, af hverju/hvers vegna/hví, hve/hversu — and their syntax: the wh-word fronts, the finite verb takes second position (V2), prepositions front or strand, and the frozen idiom Hvernig hefurðu það?
- Relative Clauses with semA2 — How relative clauses work in Icelandic — the invariant sem follows its head noun, the relativised role leaves a GAP whose case is recovered from inside the clause, prepositions STRAND at the end (húsið sem ég bý í), and possessive/oblique relatives often need a RESUMPTIVE pronoun (maðurinn sem bíllinn hans bilaði) where English uses 'whose'.
- Case Preservation and Quirky Case in DepthC2 — The single most-cited fact in Icelandic syntax: a lexically case-marked argument KEEPS that case across every syntactic operation — passive, raising, control, and ECM. The passive of a dative-object verb produces a DATIVE SUBJECT (Honum var hjálpað 'he was helped'; Bílnum var stolið 'the car was stolen'), and raising carries a quirky dative up unchanged (Honum virðist leiðast). This preservation is the clinching proof that some case is lexical, not structural — a property found in almost no other well-studied language, and the crown jewel of the field.