In linguistics, wh-movement (also known as wh-fronting or wh-extraction or long-distance dependency) concerns special rules of syntax, observed in many languages around the world, involving the placement of interrogative words. The special interrogatives, whatever the language, are known within linguistics as wh-words because most interrogative words in the English language start with a wh-; for example, who(m), whose, what, which, etc. Wh-words are used to form questions, and can also occur in relative clauses. In languages with wh-movement, sentences or clauses with a wh-word show a special word order that has the wh-word (or phrase containing the wh-word) appear at the front of the sentence or clause (Who do you think about?) instead of in a more canonical position later in the sentence (I think about you) The opposite is called wh in situ.
Wh-movement often results in a discontinuity, and in that regard, it is one of (at least) four widely acknowledged discontinuity types, the others being topicalization, scrambling, and extraposition. Wh-movement is found in many languages around the world, and of the various discontinuity types, wh-movement is the one that has been studied the most.
Historically, the name wh-movement stems from early generative grammar (1960s and 1970s) and was a reference to the transformational analysis of the day in which the wh-expression appeared in its canonical position at deep structure and then moved leftward out of that position to land in its derived position at the front of the sentence/clause at surface structure.
Many modern theories of syntax do not acknowledge movement in the transformative sense, but the term wh-movement (also called wh-fronting or wh-extraction) survives and is widely used to denote the observed phenomenon, even by theories that do not acknowledge movement.
The following sentence pairs illustrate wh-movement. Each a-sentence has the canonical word order of a declarative sentence in English, and each b-sentence has experienced wh-movement, whereby the wh-word has been fronted in order to form a question. The relevant words are bolded:
- a. Tom has been reading Tesnière.
- b. Who has Tom been reading? – The direct object corresponding to Tesnière has been wh-fronted as the wh-word who.
- a. She should stop talking about syntax.
- b. What should she stop talking about? – The object of the preposition corresponding to syntax has been wh-fronted as the wh-word what.
- a. They want to visit us tomorrow.
- b. When do they want to visit us? – The adjunct corresponding to tomorrow has been wh-fronted as the wh-word when.
- a. She is happy.
- b. What is she? – The predicative adjective corresponding to happy has been fronted as the wh-word what.
These examples illustrate that wh-fronting occurs when a constituent is questioned that appears to the right of the finite verb in the corresponding declarative sentence. Consider in this regard that when the subject is questioned, there is no obvious reason to assume that wh-fronting has occurred because the default position of the subject is clause-initial:
- a. Fred is working hard.
- b. Who is working hard? – The subject corresponding to Fred already appears at the front of the sentence, so there is no reason to assume that who has been fronted.
Despite the fact that such data provide no obvious reason to assume movement, some theories of syntax maintain a movement analysis in the interest of remaining consistent. They assume that the wh-subject has in fact moved up the syntactic hierarchy, although this movement is not apparent from the actual linear order of the words.
Wh-expressions without wh-movementEdit
Wh-movement typically occurs to form questions in English. There are, however, at least three kinds of questions in which wh-movement does not occur (aside from when the question word serves as the subject and so is already fronted): 1) echo questions (to confirm what you thought you heard), 2) quiz questions, and 3) multiple questions, when there is already one wh-word at the front:
- You bought what!? – Echo question
- George Orwell was born in which country? – Quiz question
- Who bought what? – Multiple wh-expressions
While wh-movement is the rule (and these three cases are the exceptions to the rule) in English, other languages may leave wh-expressions in situ (in base position) more often. In French for instance, wh-movement is often optional in certain matrix clauses.
In subordinate clausesEdit
The examples in the previous section have wh-movement occurring in main clauses (in order to form a question). Wh-movement is not restricted to occurring in main clauses. It frequently appears in subordinate clauses, although its behavior in subordinate clauses differs in a key respect, viz. word order. The following two subsections consider wh-movement in indirect questions and relative clauses.
In indirect questionsEdit
In English, wh-movement occurs to form a question in both main and subordinate clauses. When the question is expressed with a main clause, it is a direct question. When the question is expressed with a subordinate clause, however, it is an indirect question. While wh-fronting occurs in both direct and indirect questions, there is a key word order difference that distinguishes between the two. This difference is illustrated with the following data:
- a. Fred will ask Jill to leave.
- b. Who1 will2 Fred ask to leave? – Direct question
- c. I wonder who1 Fred2 will3 ask to leave. – Indirect question
- a. Sam likes to get news about hurricanes.
- b. What1 does2 Sam like to get news about? – Direct question; do-support introduced
- c. They asked what1 Sam2 likes3 to get news about. – Indirect question
- a. Larry stayed home due to the weather.
- b. Why1 did2 Larry stay home? – Direct question; do-support introduced
- c. Nobody knows why1 Larry2 stayed3 home. – Indirect question
The subscripts indicate a central word order difference across direct and indirect questions. Wh-fronting in main clauses typically results in V2 word order in English, meaning the finite verb appears in second position, as marked by the 2-subscript in the b-sentences. In indirect questions, however, V3 word order typically obtains, as marked by the 3-subscript in the c-sentences. Despite this systematic word order difference across direct and indirect questions, wh-fronting within the clause is occurring in both cases. Note as well that do-support is often needed in order to enable wh-fronting. Wh-fronting in main clauses is often reliant on subject-auxiliary inversion.
In relative clausesEdit
The examples above all involve interrogative clauses (questions). Wh-movement also occurs in relative clauses, however, which cannot be interpreted as questions. Many relative pronouns in English have the same form as the corresponding interrogative words (which, who, where, etc.). Relative clauses are subordinate clauses, so the characteristic V3 word order seen in indirect questions occurs:
- a. I read Fred's paper.
- b. Fred's paper, which1 I2 read3 – Wh-fronting in relative clause
- c. *Fred's paper, which1 did2 I read – Wh-fronting impossible with V2 word order in subordinate clause
- a. John likes the governor.
- b. the governor who1 John2 likes3 – Wh-fronting in relative clause
- c. *the governor who1 does2 John like – Wh-fronting impossible with V2 word order in subordinate clause
- a. Fred reads the paper in the coffee shop.
- b. the coffee shop where1 Fred2 reads3 the paper – Wh-fronting in relative clause
- c. *the coffee shop where1 does2 Fred read the paper – Wh-fronting impossible in subordinate clause with V2 word order
The relative pronouns have fronted in the subordinate clauses of the b-examples, just like they are fronted in the indirect questions in the previous sections. The characteristic V3 word order is obligatory. If the V2 word of main clauses occurs, the sentence is bad, as the c-examples demonstrate.
Many instances of wh-fronting involve pied-piping. Pied-piping occurs when a fronted wh-word (or otherwise focused word) pulls an entire encompassing phrase to the front of the clause with it, i.e. it "pied-pipes" the other words of the phrase with it to the front of the clause (see the Pied Piper of Hamelin). The following two subsections consider both obligatory and optional pied-piping.
Pied-piping is sometimes obligatory. That is, in order for a wh-expression to be fronted, an entire encompassing phrase must be fronted with it. The relevant phrase of pied-piping is underlined in the following examples:
- a. Susan is reading Fred's novel.
- b. Whose novel is Susan reading? – Pied-piping of novel
- c. *Whose is Susan reading novel? – Sentence is bad because pied-piping has not occurred.
- a. The music is very loud.
- b. How loud is the music? – Pied-piping of loud
- c. *How is the music loud? – Sentence is bad because pied-piping has not occurred.
These examples illustrate that pied-piping is often necessary when the wh-word is inside a noun phrase (NP) or adjective phrase (AP). Pied-piping is motivated in part by the barriers and islands to extraction (see below). When the wh-word appears underneath a blocking category or in an island, the entire encompassing phrase must be fronted. Pied-piping was first identified by John R. Ross in his 1967 dissertation.
There are cases where pied-piping can be optional. In English, this occurs most notably with prepositional phrases (PPs). The wh-word is the object of a preposition. A formal register will pied-pipe the preposition, whereas more colloquial English prefers to leave the preposition in situ, e.g.
- a. She revealed her secret to Tom.
- b. To whom did she reveal her secret? – Pied-piping of preposition associated with a formal register
- c. Who did she reveal her secret to? – Pied-piping absent in colloquial, everyday English
- a. He is hiding behind the red door.
- b. Behind which door is he hiding? – Pied-piping of preposition associated with a formal register.
- c. Which door is he hiding behind? – Pied-piping absent in colloquial, everyday English
The c-examples are cases of preposition stranding, which is possible in English, but not allowed in many languages that are related to English. For instance, preposition stranding is largely absent from many of the other Germanic languages and it may be completely absent from the Romance languages. Prescriptive grammars often claim that preposition stranding should be avoided in English as well; however, in certain contexts pied-piping of prepositions in English may make a sentence feel artificial or stilted.
In many cases, a wh-expression can occur at the front of a sentence regardless of how far away its canonical location is, e.g.
- a. Who does Mary like __?
- b. Who does Bob know that Mary likes __?
- c. Who does Carl believe that Bob knows that Mary likes __?
The wh-word who is the direct object of the verb likes in each of these sentences. There appears to be no limit on the distance that can separate the fronted expression from its canonical position. In more technical terms, we can say that the dependency relation between the gap (the canonical, empty position) and its filler (the wh-expression) is unbounded in the sense that there is no upper bound on how deeply embedded within the given sentence the gap may appear.
However, there are cases in which this is not possible. Certain kinds of phrases do not seem to allow a gap. The phrases from which a wh-word cannot be extracted are referred to as extraction islands or simply islands. The following subsections briefly consider seven types of islands: 1) adjunct islands, 2) wh-islands, 3) subject islands, 4) left branch islands, 5) coordinate structure islands, 6) complex NP islands, and 7) non-bridge islands. These islands types were all originally identified in Ross' seminal dissertation. The islands in the examples that follow are underlined in the a-sentences.
An adjunct island is a type of island formed from an adjunct clause. Wh-movement is not possible out of an adjunct clause. Adjunct clauses include clauses introduced by because, if, and when, as well as relative clauses. Some examples include:
- a. You went home because you needed to do that?
- b. *What did you go home because you needed to do __? – The attempt to extract out of an adjunct clause fails.
- a. Alex likes the woman who wears tight sweaters?
- b. *What does Alex like the woman who wears __? – The attempt to extract out of an adjunct clause fails.
Wh-movement fails in the b-sentences because the gap appears in an adjunct clause.
A wh-island is created by an embedded sentence which is introduced by a wh-word. Wh-islands are weaker than adjunct islands since extraction is often quite awkward, but they are not necessarily considered to be ungrammatical by all speakers.
- a. John wonders where Eric went to buy a gift?
- b. ??What does John wonder where Eric went to buy __? – The attempt to extract out of a wh-island is at best strongly marginal.
- a. Susan asked why Sam was waiting for Fred.
- b. *Who did Susan ask why Sam was waiting for __? – The attempt to extract out of a wh-island fails.
The b-sentences are strongly marginal/unacceptable because one has attempted to extract an expression out of a wh-island.
Wh-movement is not (or hardly) possible out of subjects, at least not in English. This is particularly true for subject clauses, and to a somewhat lesser extent out of subject phrases, e.g.
- a. That John went home is likely.
- b. *Who is that __ went home likely? – Wh-extraction out of a subject clause fails.
- a. The story about Susan was funny.
- b. ??Who was the story about __ funny? – Wh-extraction out of subject phrase is strongly marginal.
The important insight here is that wh-extraction out of object clauses and phrases is quite possible. There is therefore an asymmetry across subjects and objects with respect to wh-movement.
Left branch islandsEdit
Modifiers that would appear on a left branch under a noun (i.e. they precede the noun that they modify) cannot be extracted. The relevant constraint is known as the Left Branch Condition, and Ross (1967) is again credited with having discovered it. The left branch constraint captures the fact that possessive determiners and attributive adjectives in English and many related languages necessarily pied-pipe the entire noun phrase when they are fronted, e.g.
- a. Susan likes Fred's account.
- b. *Whose does Susan like __ account? – Attempt to extract from a left branch under a noun fails.
- c. Whose account does Susan like __? – Extraction succeeds if the entire NP is pied piped.
- a. He bought an expensive boat.
- b. *How expensive did he buy a __ boat? – Attempt to extract from a left branch under a noun fails.
- c. How expensive a boat did he buy? – Extraction succeeds if the entire NP is pied piped.
Extraction fails in the b-sentences because the extracted expression corresponds to a left-branch modifier of a noun. Left branch islands are cross-linguistically variable. While they exist in English, they are absent from many other languages, most notably, from the Slavic languages.
Coordinate structure islandsEdit
In coordination, extraction out of a conjunct of a coordinate structure is possible only if this extraction affects all the conjuncts of the coordinate structure equally. The relevant constraint is known as the coordinate structure constraint. Extraction must extract the same syntactic expression out of each of the conjuncts simultaneously. This sort of extraction is said to occur across the board (ATB-extraction), e.g.
- a. Sam ate [beans] and [broccoli].
- b. *What did Sam eat [beans] and [__]? – Extraction fails because it affects just one conjunct.
- a. Sam ate [beans] and [broccoli].
- b. *What did Sam eat [__] and [broccoli]? – Extraction fails because it affects just one conjunct.
- a. Sam [gave a guitar to me] and [loaned a trumpet to you].
- b. What did Sam [give __ to me] and [loan __ to you]? – Extraction succeeds because it occurs equally out of both conjuncts (ATB-extraction).
- a. He is [waiting for you] and [trying to call you].
- b. Who is he [waiting for __] and [trying to call __]? – Extraction succeeds because it occurs equally out off both conjuncts (ATB-extraction).
Wh-extraction out of a conjunct of a coordinate structure is only possible if it can be interpreted as occurring equally out all the conjuncts simultaneously, that is, if it occurs across the board.
Complex noun phrase islandsEdit
Extraction is difficult from out of a noun phrase. The relevant constraint is known as the complex NP constraint, and comes in two varieties, the first banning extraction from the clausal complement of a noun, and the second banning extraction from a relative clause modifying a noun:
Sentential complement to a noun:
- a. You heard the claim that Fred solved the second problem.
- b. ??What did you hear the claim that Fred solved __? – Attempt to extract out of a complex NP fails.
- a. She likes the possibility that she might get a new phone for X-mas.
- b. ??What does she like the possibility that she might get __ for X-mas? – Attempt to extract out of a complex NP fails.
- a. They hired someone who speaks a Balkan language.
- b. *What Balkan language did they hire someone who speaks __?
Extraction out of object that-clauses serving as complements to verbs may show island-like behavior if the matrix verb is a non-bridge verb (Erteschik-Shir 1973). Non-bridge verbs include manner-of-speaking verbs, such as whisper or shout, e.g.
- a. She thinks that he died in his sleep.
- b. How does she think that he died __? – Extraction out of object clause easily possible with matrix bridge verb.
- a. She whispered that he had died in his sleep.
- b. *How did she whisper that he had died __? – Extraction across a non-bridge verb is impossible.
In other languagesEdit
Wh-movement is also found in many other languages around the world. Most European languages also place wh-words at the beginning of a clause. Furthermore, many of the facts illustrated above are also valid for other languages. The systematic difference in word order across main wh-clauses and subordinate wh-clauses shows up in other languages in varying forms. The islands to wh-extraction are also present in other languages, but there will be some variation. The following example illustrates wh-movement of an object in Spanish:
a. Juan compró carne. John bought meat. 'John bought meat.' b. ¿Qué compró Juan? what bought John 'What did John buy?'
The following examples illustrates wh-movement of an object in German:
a. Er liest Tesnière jeden Abend. He reads Tesnière every evening. 'He reads Tesnière every evening.' b. Wen liest er jeden Abend? who reads he every evening 'Who does he read every evening?'
The following examples illustrate wh-movement an object in French:
a. Ils ont vu Pierre they have seen Peter 'They saw Peter.' b. Qui est-ce qu' ils ont vu? Who is it that they have seen 'Who did they see?' c. Qui ont ils vu? Who have they seen 'Who did they see?'
All the examples are quite similar to the English examples and demonstrate that wh-movement is a general phenomenon in numerous languages. As stated however, the behavior of wh-movement can vary, depending on the individual language in question.
Wh-movement typically results in a discontinuity: the "moved" constituent ends up in a position that is separated from its canonical position by material that syntactically dominates the canonical position, which means there seems to be a discontinuous constituent and a long distance dependency present. Such discontinuities challenge any theory of syntax, and any theory of syntax is going to have a component that can address these discontinuities. In this regard, theories of syntax tend to explain discontinuities in one of two ways, either via movement or via feature passing.
Theories that posit movement have a long and established tradition that reaches back to early Generative Grammar (1960s and 1970s). They assume that the displaced constituent (e.g. the wh-expression) is first generated in its canonical position at some level or point in the structure generating process below the surface. This expression is then moved or copied out of this base position and placed in its surface position where it actually appears in speech. Movement is indicated in tree structures using one of a variety of means (e.g. a trace t, movement arrows, strikeouts, lighter font shade, etc.).
The alternative to the movement approach to wh-movement and discontinuities in general is feature passing. This approach rejects the notion that movement in any sense has occurred. The wh-expression is base generated in its surface position, and instead of movement, information passing (i.e. feature passing) occurs up or down the syntactic hierarchy to and from the position of the gap.
- Accounts of wh-fronting appear in many textbooks on syntax and grammar, e.g. Stockwell (1977:35ff.), Baker (1978:119ff.), Riemsdijk and Williams (1986:19ff.), Borsley (1988:188ff.), Radford (1997:267ff.), Roberts (1999:35ff.), Tallerman (2005:217ff.), Carnie (2013:357ff.).
- For early accounts of question formation and wh-movement, see, for instance, Ross (1967/86:18ff.), Bach (1974:129), Culicover (1976:73f.), Stockwell (1977:172f.), Baker (1978:121f.).
- Concerning wh-in-situ questions, see Radford (1997:267f.)
- Concerning wh-movement in French, see Bošković (2002).
- Concerning the key word order difference across direct and indirect questions, see for instance Roberts (1997:37) and Groß and Osborne (2009:74ff.), and Carnie (2013:367).
- See Carnie (2013:369ff.) for an analysis of relative clauses in terms of wh-movement.
- See Ross' (1967/86:121ff.) original account of pied-piping. For further analyses of pied-piping, see for instance Riemsdijk and Williams (1986:28ff.) and Radford (1997:276ff).
- Concerning preposition stranding in wh-questions in English, see Roberts (1997:212f) and Radford (1999:278ff.).
- For general accounts of island phenomena, see for instance Riemsdijk and Williams (1986:23ff), Roberts (1997:186ff.), Borsley (1999:205ff.), and Carnie (2013:374ff.).
- For accounts of wh-islands, see Roberts (1997:195f.), O'Grady (2005:118ff.).
- Concerning subject islands, see Ross (1967/86:148f.), Culicover (1976:282ff.), Borsley (1999:206), Radford (1997:281).
- Ross (1967/86) gives his left branch condition on page 127: "No NP which is the leftmost constituent of a larger NP can be reordered out of this NP by a transformational rule".
- Concerning the lack of left branch islands in Slavic languages, see Ross (1967/86:145), Grosu (1973), Roberts (1997:189).
- Concerning the coordinate structure constraint, see Ross (1967/86:97ff.), Bach (1974:210), Culicover (1976:281ff.), Roberts (1997:188).
- The term across the board is from Williams (1978). See also Roberts (1997:188), Borsley (1999:207).
- Concerning the complex NP constraint, see for instance Ross (1967/86:272ff.), Culicover (1976:280f.), Baker (1978:200ff.), Borsley (1999:206f.)
- For an example of the movement/copying approach, see Radford (2004:153ff.).
- Bach, E. 1974. Syntactic theory. New York: Holt, Rinehart and Winston, Inc.
- Baker, C. 1978. Introduction to generative-transformational syntax. Englewood Cliffs, NJ: Prentice-Hall.
- Borsley, R. 1999. Syntactic theory: A unified approach. London: Arnold.
- Bošković 2002. On multiple wh-fronting. Linguistic Inquiry 33, 351-384.
- Carnie, A. Syntax: A generative introduction. Malden, MA: Wiley-Blackwell.
- Chomsky, N. 1977. On wh-movement. In Culicover, P. W., Wasow, Thomas, and Akmajian, Adrian (eds), Formal Syntax, New York.
- Culicover, P. 1976. Syntax. New York: Academic Press.
- Crystal, D. 1997. A dictionary of linguistics and phonectis. Oxford, UK: Oxford Publishers.
- Erteschik-Shir, N. 1973. On the nature of island constraints. Ph. D. thesis, Massachusetts Institute of Technology, Cambridge, MA.
- Groß, T. and T. Osborne 2009. Toward a practical dependency grammar theory of discontinuities. SKY Journal of Linguistics 22, 43-90.
- Jurafsky, D. and J. Martin. 2008. Speech and language processing: An introduction to natural language processing, computational linguistics, and speech recognition. Delhi, India: Pearson Education.
- Grosu, A. 1973. On the Left Branch Condition. Linguistic Inquiry.
- O'Grady, W. 2005. Syntactic carpentry: An emergentist approach. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers.
- Parra-Guinaldo, V. 2013. A generativist approach to renewal in the left periphery: The reanalysis of ‘whether’. Saarbrücken: Lambert.
- Radford, A. 2004. English syntax: An introduction. Cambridge, UK: Cambridge University Press.
- Riemsdijk, H. van and E. Williams. 1986. Introduction to the theory of grammar. Cambridge, MA: The MIT Press.
- Roberts, I. 1997. Comparative syntax. London: Arnold.
- Ross, J. 1967. Constraints on variables in syntax. Doctoral dissertation, Massachusetts Institute of Technology.
- Ross, J. 1986. Infinite syntax. (Originally presented as the author's thesis from 1967). Norwood, NJ: Infinite syntax!
- Stockwell, R. 1977. Foundations of syntactic theory. Englewood Cliffs, NJ: Prentice-Hall, Inc.
- Tallerman, M. 2005. Understanding syntax. 2nd edition. Malta: Hodder Arnold.
- Williams, E. 1978. Across the board rule application. Linguistic Inquiry 9, 31-43.
- Lexicon of Linguistics:Wh-movement
- Lexicon of Linguistics:Pied piping
- Lexicon of Linguistics:Island
- Lexicon of Linguistics:Wh-island
- Lexicon of Linguistics:Left Branch Condition
- Lexicon of Linguistics:Wh-in-situ
- Stanford Linguistics 222B: Foundations of Syntactic Theory II Filler-Gap Dependencies (Unbounded Dependency Constructions) has a good archive of essential papers in the field