Most question words sit at the front of a single clause: Wat doen jy? ("What are you doing?"). But Afrikaans, like English, lets a question word reach much deeper — it can be lifted out of an embedded clause and planted at the very front of the sentence, with the rest of the question stretched out behind it. Wat dink jy het hy gedoen? — "What do you think he did?" — has wat belonging logically to the lowest clause (hy het [wat] gedoen) but appearing at the top. This is a long-distance dependency: the fronted element and the empty position it came from (the gap) can be separated by one, two, or more clause boundaries. This page maps how far the dependency can stretch and, just as importantly, where it slams into a wall. Basic single-clause questions are covered on wh-questions; here we cross clause boundaries.
The basic pattern: extraction across one boundary
Start with an embedded statement: Jy dink [hy het dit gedoen] — "You think he did it." To question the object of the lowest clause, the question word travels all the way to the front of the whole sentence, and a gap is left behind where it would otherwise stand.
Wat dink jy het hy gedoen?
What do you think he did?
Wat sê jy het hy gedoen?
What do you say he did?
Wie glo jy het die ruit gebreek?
Who do you believe broke the window?
Read the structure of Wat dink jy het hy gedoen?: the matrix clause is jy dink ("you think"), the embedded clause is hy het [gap] gedoen, and wat has been extracted from that gap up to the front. The main verb dink still sits in second position after wat (V2 holds in the matrix clause), and the embedded clause that follows shows up in main-clause order (het hy gedoen — verb before subject) rather than the verb-final order you would expect of a subordinate clause. This last point is the distinctively Afrikaans signature of long extraction.
Bridge verbs: the verbs that let extraction pass through
Extraction does not work through every embedding verb. The clause in the middle has to be a bridge — a verb that reports thought, speech, or belief and lets material pass up through it. The reliable bridges in Afrikaans are the same semantic class as in English: dink (think), sê (say), glo (believe), meen (reckon), vermoed (suspect), hoop (hope), dink/voel (feel), beweer (claim).
Waar meen jy het hulle die geld weggesteek?
Where do you reckon they hid the money?
Vir wie het sy gesê sy koop die geskenk?
Who did she say she's buying the present for?
Hoe vermoed die polisie het die inbreker ingekom?
How do the police suspect the burglar got in?
Notice that the bridge can itself be in the past while the extraction still reaches the lower clause: Vir wie het sy gesê... — the auxiliary het and participle gesê form the matrix, and the lower clause sy koop die geskenk vir [gap] supplies the gap for vir wie. The dependency simply spans the whole structure.
Pied-piping a preposition out of the embedded clause
When the extracted element is the object of a preposition down in the embedded clause, the *waar-*compound (for things) is fronted whole, carrying its preposition to the top of the sentence across the boundary.
Waaroor dink jy praat hulle?
What do you think they're talking about?
Waarmee glo jy het hy die slot oopgemaak?
What do you think he opened the lock with?
Vir wie het jy gehoor wag hulle?
Who did you hear they're waiting for?
Waaroor in the first sentence belongs to praat oor [gap] in the lowest clause, yet it surfaces at the very front, the whole *waar-*compound pied-piped across *dink jy. This mirrors the people-vs-things split you already know from single-clause fronting; long distance does not change *what moves, only how far.
Long relativisation: gaps far from the head noun
Extraction is not limited to questions. A relative clause can also have its gap sunk deep inside an embedded clause, far from the head noun, with wat doing the long-distance binding.
Die plan wat ek dink ons moet volg, is die goedkoopste.
The plan that I think we should follow is the cheapest.
Daar is dinge wat sy sê sy nooit sal vergewe nie.
There are things she says she will never forgive.
In Die plan wat ek dink ons moet volg, the head is die plan, but the gap is the object of volg two clauses down (ons moet [gap] volg). Wat binds it across ek dink. The same bridge-verb requirement applies: you can relativise across dink, sê, glo, but not across an island.
Where it stops: island constraints
Here is the advanced payoff, and the reason long-distance dependencies are a C2 topic: Afrikaans's famously free fronting is not unlimited. Certain configurations are syntactic islands — you cannot extract anything out of them, no matter how natural the meaning. This is where the apparent freedom reveals a hard grammatical boundary.
Relative-clause island. You cannot pull a question word out of a relative clause. From Jy ken die man [wat die motor gesteel het], you cannot question the object of the relative.
❌ Wat ken jy die man wat gesteel het?
Incorrect — extraction out of a relative clause is blocked (a relative-clause island).
✅ Wat het die man gesteel wat jy ken?
What did the man you know steal? (the question word belongs to the main clause)
Adjunct island. You cannot extract out of a reason or time adjunct clause (omdat..., toe..., terwyl...).
❌ Wat was sy kwaad omdat jy gebreek het?
Incorrect — extraction out of an omdat-adjunct is blocked (an adjunct island).
✅ Sy was kwaad omdat jy iets gebreek het — wat was dit?
She was angry because you broke something — what was it? (asked as a separate clause)
Wh-island. You cannot extract across an already-questioned clause — once a lower clause is itself introduced by a question word, it is sealed.
❌ Wat wonder jy wie gekoop het?
Incorrect — the lower clause is already a wh-clause (wie...), so it is a wh-island.
✅ Jy wonder wie wat gekoop het?
You're wondering who bought what? (echo/multiple question, nothing extracted)
The pattern is consistent: extraction passes freely through a complement clause under a bridge verb, but it cannot escape a relative clause, an adjunct clause, or a wh-clause. These island effects are not Afrikaans quirks — they are among the deepest cross-linguistic constraints on movement, and Afrikaans obeys them as tightly as English does.
A subtlety: the dat-complement and the "missing" complementiser
When the bridge clause would normally carry dat ("that"), long extraction typically forces it to drop. Jy dink dat hy dit gedoen het ("you think that he did it") loses the dat under extraction: Wat dink jy het hy gedoen? — not Wat dink jy dat het hy gedoen? The fronted question word and the retained dat compete for the same edge, and the question wins.
Wie hoop jy kom môre kuier?
Who do you hope is coming to visit tomorrow?
Wat dink jy gaan volgende gebeur?
What do you think is going to happen next?
This dropped-complementiser pattern, paired with the main-clause word order in the lower clause (het hy gedoen, gaan volgende gebeur), is the fingerprint of a genuine long-distance dependency in Afrikaans — and distinguishes it from a flat single question.
Common mistakes
❌ Wat dink jy dat hy gedoen het?
Incorrect — under long extraction the dat drops and the lower clause takes main-clause order: het hy gedoen.
✅ Wat dink jy het hy gedoen?
What do you think he did?
❌ Waaroor dink jy hulle praat?
Incorrect — the lower clause should surface in main-clause order under extraction: praat hulle.
✅ Waaroor dink jy praat hulle?
What do you think they're talking about?
❌ Wat ken jy die man wat gekoop het?
Incorrect — you cannot extract out of a relative clause; this is an island.
✅ Wat het die man gekoop wat jy ken?
What did the man you know buy?
❌ Hoekom is sy kwaad omdat jy gedoen het?
Incorrect — you cannot extract out of an omdat-adjunct; it is an island.
✅ Sy is kwaad omdat jy iets gedoen het — wat het jy gedoen?
She's angry because you did something — what did you do?
Key takeaways
- A long-distance dependency fronts a question word or relativiser from inside an embedded clause, leaving a gap below; the fronted item answers the lowest verb, not the bridge.
- Extraction passes only through bridge verbs of thought and speech — dink, sê, glo, meen, vermoed — and the lower clause then surfaces in main-clause order (het hy gedoen), with dat dropped.
- The same machinery feeds long relativisation: die plan wat ek dink ons moet volg binds wat across a clause boundary.
- The apparent freedom hits hard limits at islands: you cannot extract out of a relative clause, an adjunct clause (omdat, toe, terwyl), or a wh-clause. Rephrase instead.
- These island effects are universal movement constraints, not Afrikaans idiosyncrasies — the language obeys them as strictly as English.
Now practice Afrikaans
Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.
Start learning Afrikaans→Related Topics
- Question Words: wie, wat, waar, wanneer, hoekom, hoeA1 — How to ask open questions in Afrikaans with wie, wat, waar, wanneer, hoekom/waarom, hoe, watter and hoeveel — question word first, verb second, no 'do'.
- Topicalisation and Focus FrontingB2 — Afrikaans fronts almost any constituent to the first slot for topic or contrast — forcing V2 inversion — and uses the dit is ... wat cleft to spotlight a focus, where English leans on stress alone.
- Extraposition and Heavy ClausesC1 — Why heavy subordinate clauses move to the right of the verb bracket in Afrikaans — the rule that explains the real shape of complex sentences.
- Subordinate Clauses: Verb to the EndA2 — In an Afrikaans subordinate clause the finite verb moves to the very end — the single biggest word-order adjustment English speakers have to make.
- Raising and Control ConstructionsC2 — Why blyk and skyn behave differently from probeer and weier — the raising/control distinction, and the tests that reveal it.