Wh-Movement

How wh-movement works

Take the English sentence Mary bought a book. To question the object, English does three things at once: it replaces "a book" with the wh-word "what", fronts the wh-word, and inserts an auxiliary in second position (T-to-C movement, also called subject-auxiliary inversion):

     Mary bought a book.
   → Mary bought what?     (echo question, wh in situ)
   → What did Mary buy ___?  (wh-fronted, gap at object position)

The blank in the third line is the gap — the position where "what" is interpreted as the object of "buy". Generative grammar treats the gap as a silent trace (in earlier theories) or an unpronounced copy (in Minimalism, Chomsky 1995) of the moved phrase. The trace lets semantic interpretation reconstruct the argument structure: the verb "buy" sees "what" in its object slot at LF, even though the surface string puts it elsewhere.

The mechanics are usually stated as follows. The complementiser head C of an interrogative clause carries a [+wh] or [+Q] feature. This feature must be checked, and checking is done by attracting a phrase that bears matching wh-features into the specifier of CP. The movement target is therefore fixed: the left edge of the clause. The movement is also successive cyclic — over long distances it stops at every intermediate CP, leaving a chain of traces.

English wh-fronting vs Mandarin wh-in-situ

The most striking typological split is between languages that visibly move wh-words and those that do not. Compare:

	English (wh-fronting)	Mandarin (wh-in-situ)	Japanese (wh-in-situ + ka)
What did Mary buy?	What did Mary buy ___?	Mary mai-le shenme?	Mary-ga nani-o kaimashita ka?
Surface position of wh	Clause-initial	Object position	Object position
Question marker	Aux-fronting (did)	None (intonation)	Final particle ka
Where is the gap?	After "buy"	None (no movement)	None (no movement)
Echo question form	Mary bought what?	Same as ordinary question	Same as ordinary question
Multiple wh	One fronts: Who saw what?	Both stay: Shei kanjian-le shei?	Both stay (with ka)
Long-distance scope	Marked by movement	Marked by particle/intonation; covert movement at LF	Marked by ka placement

C.-T. James Huang's 1982 MIT dissertation showed that Mandarin in-situ wh-phrases obey the same locality constraints as English moved wh-phrases — they cannot scope out of complex NPs or adjunct clauses, the classic island effects. He concluded that wh-movement applies covertly at LF in Mandarin, with the same syntactic operation, just unpronounced. The covert-movement analysis dominated the field for two decades; later work (Reinhart 1998, Cheng 2009) defends in-situ binding via choice functions instead, and the question is still debated.

Worked example: "What does John know that Mary bought?"

Long-distance wh-movement crosses clause boundaries. Consider the sentence:

What does John know that Mary bought ___ ?

Here "what" is interpreted as the object of "bought" in the embedded clause, but it sits at the front of the matrix clause. Standard generative theory derives this in stops:

1. Base order: John knows that Mary bought what.
2. Move "what" to the embedded SpecCP: John knows [_CP what that Mary bought ___].
3. Move "what" from the embedded SpecCP to the matrix SpecCP: [_CP What does John know [_CP ___ that Mary bought ___]].

The trace chain at every intermediate CP is required by the Subjacency principle (Chomsky 1973) and later by phase theory (Chomsky 2000, 2001). Without successive-cyclic stops, the wh-phrase would have to leap over multiple bounding nodes in a single step, which is forbidden. Some Romance and Germanic languages (German, Romani, Bavarian) overtly mark each step with an intermediate "wh-copying" or "partial wh-movement" — visible evidence for the cyclic analysis.

Pied-piping and preposition stranding

When a wh-word is embedded inside a larger phrase, the language must decide whether to move just the wh-word or to drag along its containing phrase. Compare:

• Who did you give the book to ___? — preposition stranding (English, Norwegian, Swedish).
• To whom did you give the book ___? — pied-piping the preposition (English formal register, German, Russian).
• Whose book did you read ___? — pied-piping the entire possessive DP.
• *Whose did you read ___ book? — ungrammatical: cannot extract just the possessor and strand the noun.

John Robert "Haj" Ross's 1967 MIT dissertation Constraints on Variables in Syntax coined the term "pied-piping", after the legend of the Pied Piper of Hamelin: the wh-word leads a chain of accompanying material out of its base position. The Left Branch Condition (Ross 1967) explains why English forbids extracting just "whose" from inside a noun phrase, while many Slavic languages permit it (Polish allows Czyją czytałeś książkę?, "Whose did you read book?").

Multiple wh-fronting

While English fronts only one wh-phrase per clause, several languages front them all. Bulgarian is the textbook case:

Koj kogo vizhda?
who-NOM whom-ACC sees
"Who sees whom?"

Both wh-phrases land at the left edge, in a fixed order (subject before object). Željko Bošković (2002) analyses Bulgarian as moving every wh-phrase into stacked specifiers of CP, while English-type languages move just one and leave the rest in their base positions. Romanian, Russian, Polish, Czech, and Serbo-Croatian all show degrees of multiple wh-fronting, with subtle differences in superiority effects.

Wh-movement vs topicalisation, focus, and relativisation

	Wh-movement	Topicalisation	Focus movement	Relativisation
Example	What did Mary buy?	This book, Mary bought.	It was THIS BOOK Mary bought.	the book that Mary bought
Trigger feature	[+wh]/[+Q]	[+topic]	[+focus]	[+rel]
Landing site	SpecCP	SpecTopP (Rizzi 1997)	SpecFocP	SpecCP
Leaves a gap?	Yes	Yes	Yes	Yes (or resumptive)
Subject to islands?	Yes	Yes	Yes	Yes
Pied-piping?	Optional/required	Optional	Optional	Optional

Luigi Rizzi's 1997 paper "The Fine Structure of the Left Periphery" argued that what looks like a single CP slot is actually a stack of dedicated projections — ForceP, TopicP, FocusP, FinitenessP — each with its own movement-attracting feature. Wh-movement targets the focus position in many languages.

Variants and edge cases

• Echo questions. "Mary bought WHAT?" leaves the wh-word in situ even in English; these are not true interrogatives but requests for repetition.
• Why-questions adjunct status. "Why" behaves differently from argument wh-words: it cannot easily reconstruct, and it shows weaker island sensitivity (Rizzi 1990, Stepanov & Tsai 2008).
• Free relatives. "What Mary bought was expensive" has a moved wh-element introducing a noun-phrase-like clause, not a question.
• D-linked wh-phrases. "Which book", "which student" — D-linked phrases (Pesetsky 1987) escape some superiority effects and some weak islands.
• Wh-the-hell. Aggressively non-D-linked phrases like "what the hell" show stronger island sensitivity (den Dikken & Giannakidou 2002).

Common pitfalls

• Confusing wh-fronting with subject position. Subject wh-questions ("Who left?") look like there is no movement because the wh-word lands where the subject would be anyway, and English suppresses do-support. The standard analysis still posits movement to SpecCP, vacuous on the surface.
• Treating Mandarin as "no syntax for questions". The wh-element does not move overtly, but Huang (1982) and most successors argue covert movement applies at LF. The locality constraints (islands) match.
• Assuming pied-piping is optional everywhere. English allows preposition stranding; most other Indo-European languages do not. Translating "Who did you give it to?" into Russian or German requires pied-piping.
• Ignoring islands. Wh-movement is not free; certain configurations block extraction (complex NP, wh-island, adjunct, subject). Treating wh-movement as unconstrained predicts many ungrammatical sentences.
• Forgetting that the gap is interpreted by the verb. Selectional restrictions hold at the gap site: "What did Mary devour?" requires "what" to be edible, despite its surface position.