Embedded Questions and Long-Distance Dependencies

Danish does two things with questions that, once mastered, separate a fluent speaker from a near-native one. First, it embeds questions under verbs of saying, knowing, and wondering with strict subordinate word order. Second — and this is where Danish quietly outruns English — it permits the questioned element to be pulled out of a deeply nested clause, including relative clauses, which English grammar treats as syntactic walls. This page assumes you already control basic indirect questions and topicalization; here we go to the structural floor.

Embedded hv-questions: subordinate order is non-negotiable

When a question becomes the object of a verb, it stops being a main clause. It loses inversion, and any sentence adverb (like ikke, altid, også) slots in before the finite verb. This is the heart of the Danish subordinate-clause schema: in a subordinate clause the adverb precedes the verb, the mirror image of the main clause.

Compare the direct question, where the finite verb sits in second position right after the hv-word, with the embedded version, where the subject intervenes:

Notice that in the embedded clause the verb bor lands at the end, after the subject hun. An English speaker's instinct is to keep the question's own shape, producing the classic error of writing hvor bor hun inside the jeg ved ikke frame. Watch what the sentence adverb does — it goes before the verb, which no main clause ever allows:

For yes/no questions there is no hv-word to borrow, so Danish uses the complementiser om ('whether/if'). Subordinate order still rules:

Long-distance dependencies: the questioned word travels

Here is the leap to C2. A hv-word can be questioned in a clause far below the verb that introduces it. The wh-element appears at the very front of the whole sentence, but it is understood as belonging to the embedded clause:

The word hvad is the object of sagde, buried two clauses down, yet it surfaces at the top. English does exactly the same thing here ("What do you think he said?"), so this case feels safe. The same mechanism lets you topicalize a non-question phrase out of an embedded clause — fronting it for emphasis:

The pattern is: [fronted phrase] + [main verb] + [subject] + , + [subordinate clause with a gap]. The gap is the empty slot where the fronted phrase 'belongs'.

The at-trace effect: drop the complementiser under extraction

Danish frequently — and under extraction, preferentially — omits at when something has been pulled out of the subordinate clause. Where it differs sharply from English is subject extraction. English shows the that-trace effect: "Who do you think left?" is fine but "Who do you think that left?" is degraded. Danish does the mirror image — the anti-that-trace effect: when a subject is extracted, Danish requires a special subject-marker, der, in the clause the subject came out of:

This der is obligatory and is one of the most common things advanced learners forget — *Hvem tror du kommer i aften (bare gap, no marker) is wrong. Note that der is not a replacement for at: it fills the embedded subject slot and can sit alongside an at further up (...du påstod, at der havde lånt bogen). It is just that in the short form tror du there is no overt at to begin with. For non-subject extraction, no subject-marker is needed and at is simply optional, usually dropped:

The Scandinavian extraction frontier: out of relative clauses

This is the feature that genuinely has no English equivalent. Mainland Scandinavian — Danish, Swedish, Norwegian — allows extraction out of relative clauses, a domain that is a hard island in English and German. The result sounds perfectly natural to a Dane and ungrammatical when translated literally:

In the first sentence, det is the object of prøvet, which sits inside the relative clause der har prøvet, which is itself inside the clause (jeg) kender mange. English cannot do this — "that, I know many who have tried" is broken English. Danish threads the dependency straight through the clause boundary. One caveat carries the whole construction: the gap has to be an object inside the relative clause. Extract a subject and even Danish balks — *Det er en bog, som jeg kender mange, der har læst is bad, because der already fills the subject slot of har læst and there is no object gap left for en bog to bind.

The standard analysis is that these are not 'true' restrictive relatives semantically but presentational or loosely-attached clauses, which is why the island constraint relaxes. You do not need the theory to use them — you need to recognise that your English instinct to block these is wrong for Danish. Listen for them; natives produce them constantly.

A few more naturally occurring extractions, the kind you hear in conversation:

Common Mistakes

Key Takeaways

• Embedded hv-questions take subordinate order: no inversion, and sentence adverbs land before the finite verb. Yes/no embeds with om, not hvis.
• A questioned or topicalised phrase can be extracted long-distance from an embedded clause, surfacing at the front while leaving a gap below.
• Under non-subject extraction, at is optional; when a subject crosses a clause boundary, insert the obligatory subject-marker der (the anti-that-trace effect — der is not a swap for at).
• Danish — uniquely among the Germanic languages you likely know — extracts out of relative clauses, but only when the gap is an object inside them; subject extraction stays blocked. Don't import the English island ban wholesale; the danger at this level is being too restrictive.