Wh-movement
Wh-movement is a core syntactic operation in generative linguistics, whereby a wh-phrase—such as who, what, where, or when—is displaced from its base-generated position within a clause to a designated clause-peripheral position, typically the specifier of the complementizer phrase (Spec,CP), leaving behind a trace or copy to maintain the sentence's interpretive structure.[1] This movement, first systematically analyzed in the framework of the Extended Standard Theory, unifies the formation of yes/no questions, wh-questions, relative clauses, and certain comparatives under a single transformational rule, subject to locality constraints like subjacency and the Specified Subject Condition.[1] In English, for instance, the declarative "John saw who" transforms into the interrogative "Who did John see?" via overt wh-movement, where the auxiliary "did" inverts to fill the tense position.[2]
Introduced prominently by Noam Chomsky in the 1970s as part of the Move α framework, wh-movement exemplifies A-bar movement, distinguishing it from A-movement (like subject raising) by targeting non-argument positions and enabling scope interpretation at Logical Form (LF).[1] The operation is driven by feature-checking mechanisms, where an uninterpretable [+wh] feature on the complementizer (C) attracts the wh-phrase to satisfy the Extended Projection Principle (EPP).[2] Key evidence for its existence includes island effects, where extraction is blocked across certain syntactic boundaries, such as wh-islands (What did Mary wonder whether John bought?) or complex noun phrases (Who did Mary discuss the claim that John saw?), demonstrating that wh-movement is not mere base-generation but a structure-dependent rule.[2] These constraints, including the Complex NP Constraint and the Empty Category Principle, ensure grammaticality and have been central to debates on locality in minimalist syntax.[1]
Cross-linguistically, wh-movement exhibits significant variation, informing parametric theories of Universal Grammar.[3] In English-like languages, a single wh-phrase undergoes overt movement to Spec,CP, while others remain in situ in multiple wh-questions (e.g., "Who bought what?"), often obeying a superiority condition that prioritizes the structurally highest wh-phrase for extraction.[2] Languages like Japanese and Chinese employ covert wh-movement at LF, keeping wh-phrases in situ at surface structure (e.g., "John-ga nani-o kaimashita ka?" – "What did John buy?"), yet still respecting island constraints, suggesting universal principles underlie the process.[3] In contrast, Slavic languages such as Bulgarian feature multiple wh-fronting, where all wh-phrases move overtly to clause-initial positions (e.g., "Koj kogo vidja?" – "Who sees whom?"), sometimes via tucking-in to multiple specifiers.[3] These parametric differences—whether movement is overt/covert, single/multiple, or optional—highlight wh-movement's role in probing the boundaries of syntactic variation and acquisition.[3]
Fundamentals
Definition and Core Properties
Wh-movement is a syntactic operation within generative grammar in which a wh-phrase, such as those headed by interrogative words like who, what, or where, is displaced from its base-generated position in the clause to the specifier position of the complementizer phrase (CP). This displacement occurs specifically in interrogative and relative clauses and is triggered by an uninterpretable [+wh] feature on the complementizer head (C^0), which attracts the matching interpretable [+wh] feature on the wh-phrase to satisfy the grammar's feature-checking requirements.[4] The operation ensures that the wh-phrase scopes over the clause, forming the necessary syntactic dependency for question interpretation.
Core properties of wh-movement include its role in distinguishing wh-questions from yes/no questions: in the latter, no phrasal movement occurs, but a [+Q] feature on C^0 drives auxiliary inversion or other marking of interrogativity, whereas wh-questions require both the [+Q] feature and wh-phrase fronting. In languages like English, this fronting obligatorily triggers subject-auxiliary inversion, where the finite auxiliary raises to C^0, but the movement itself is a distinct A-bar movement operation. Additionally, wh-movement is obligatory for matrix wh-questions in some languages, such as English, to achieve well-formedness, while it remains optional or entirely in-situ (without overt displacement) in others, like Mandarin Chinese, where licensing occurs through alternative mechanisms such as focus or particle association.[5][6]
The term "wh-movement" was coined by Noam Chomsky in his 1977 paper "On wh-movement," which formalized it as a bounded transformational rule within the extended standard theory, building on the foundational concept of transformations introduced in Syntactic Structures (1957).[5][7] Understanding wh-movement presupposes key concepts in generative syntax, including X-bar phrase structure, where clauses are headed by CP (encompassing the complementizer system) dominating IP (inflection phrase, hosting tense and agreement); theta-roles, assigned to arguments in their base positions within VP or subject position before movement; and case assignment, governed by structural relations such that the trace left by wh-movement inherits nominative or oblique case from its antecedent via antecedent-government.[8] These elements ensure that movement preserves argument structure and licensing while enabling scope and feature satisfaction.
Basic Examples in English
In English, wh-movement is exemplified in the formation of interrogative clauses, where a wh-phrase is displaced to the specifier position of the complementizer phrase (CP). For subject questions, consider the sentence "Who saw John?", in which the wh-phrase "who" originates in the subject position of the inflectional phrase (IP) and moves to Spec-CP, leaving a trace (t) in its base position; unlike yes/no questions such as "Did John see Mary?", subject wh-questions do not trigger auxiliary insertion or subject-auxiliary inversion because the wh-phrase is base-generated sufficiently high in the structure.[1] This results in a structure like [CP Who [C [IP t saw John ]]], highlighting that no additional verbal material intervenes between the fronted wh-phrase and the verb.[1]
Object questions, by contrast, involve extraction from a lower position and necessitate auxiliary inversion for do-support in simple present or past tenses. A canonical example is "What did John see?", where "what" moves from the direct object position within the verb phrase (VP), the auxiliary "did" raises to the complementizer (C) position to check features, and a trace remains in the base site, yielding the representation [CP What [C did [IP John [VP see t ]]] ].[1] This inversion distinguishes object wh-questions from their declarative counterparts like "John saw what," and the trace fills the gap, maintaining argument structure and enabling the interpretation of "what" as the theme of "see."[9]
Wh-movement exhibits varying distances while remaining unbounded in principle, allowing extraction over short spans in matrix clauses or longer paths across embeddings. For instance, short movement appears in adjacent clause questions like the object example above, whereas longer extraction occurs in constructions such as "What do you think John saw?", where "what" originates in the embedded clause, moves successively through intermediate Spec-CP positions (if applicable), and lands in the matrix Spec-CP, with traces marking each step: [CP What [C do [IP you [VP think [CP t [C [IP John [VP saw t ]]]]] ] ].[1] This successive-cyclic process underscores the operation's ability to span multiple clause boundaries without degradation, provided no obstructing factors intervene.[1]
A fundamental property of wh-movement in English is the creation of a gap or trace at the extraction site, which signals the displaced element's original theta-role and ensures grammaticality by satisfying locality and binding conditions.[1] Additionally, English mandates overt fronting of the wh-phrase to Spec-CP in matrix questions, distinguishing it from languages with in-situ wh-elements.[10]
Constraints and Phenomena
Pied-Piping
Pied-piping is a syntactic phenomenon in wh-movement where a wh-phrase embedded within a larger constituent, such as a prepositional phrase (PP) or noun phrase (NP), triggers the fronting of the entire containing phrase rather than the wh-phrase alone. This process preserves the internal structure and linear order of the pied-piped constituent. The term "pied-piping" was coined by Ross (1967) in his seminal dissertation, drawing an analogy to the Pied Piper of Hamelin who leads others along with him.[11]
In English, pied-piping can be obligatory or optional depending on the type of containing phrase. Obligatory pied-piping occurs with possessive constructions, where the wh-possessor cannot be extracted independently without stranding the possessed noun, which would violate syntactic well-formedness. For instance, the sentence "Whose book did you read?" requires movement of the full NP "whose book," as the stranded variant "*Whose did you read book?" is ungrammatical in standard English. This is exemplified more complexly in recursive possessives like "[Whose guardian's employer] did you meet?", where the entire DP must front to satisfy the wh-feature requirements. Similarly, certain quantifiers or modifiers within NPs can trigger obligatory pied-piping to maintain phrase integrity.[12][13]
Optional pied-piping is common with prepositional phrases, allowing speakers to choose between fronting the entire PP or stranding the preposition. In questions like "With whom did you speak?", the PP "with whom" can pied-pipe to the front, but the stranding alternative "Whom did you speak with?" is also grammatical, particularly in informal contexts. Pied-piping is preferred in formal registers for such cases, as in "To which city did she travel?" versus "Which city did she travel to?". Examples with other prepositions, such as "From whose perspective did you view the issue?", further illustrate this optionality, where stranding remains acceptable.[12][14]
The mechanism underlying pied-piping involves feature percolation, where the [+wh] feature of the embedded wh-phrase spreads upward to the head of the containing phrase, making the larger constituent attractable by the movement operation. This percolation, as proposed in early generative models, ensures that the wh-feature can drive displacement of the host phrase without violating linearization constraints or case assignment requirements within the structure. For example, in possessive pied-piping, the genitive case on the wh-possessor percolates to the NP, preventing isolated extraction that would leave the possessed noun without proper licensing. This process is widespread in languages exhibiting wh-movement, serving to resolve potential syntactic violations that would arise from partial extraction.[14]
Extraction Islands
Extraction islands, also known as syntactic islands, refer to specific structural domains in a sentence from which wh-extraction is ungrammatical, imposing locality constraints on wh-movement despite its generally long-distance nature.[15] These constraints were first systematically identified by Ross in his 1967 dissertation, where he described islands as configurations that block movement transformations, later formalized under principles like subjacency in subsequent work.[16] Islands arise because wh-movement cannot cross certain bounding nodes, such as NPs or CPs, leading to degraded or impossible sentences when extraction attempts to violate these domains.[17]
One major type is the adjunct island, where extraction from an adverbial or adjunct clause is prohibited. For example, the sentence "*Which movie did Blanche have dinner before Rose saw?" is ungrammatical because the wh-phrase attempts to extract from the adjunct "before Rose saw [which movie]."[15] In contrast, extraction from a non-adjunct clause, such as "Which movie did Rose see?" remains grammatical.[15]
Wh-islands block extraction from embedded wh-questions or clauses containing a wh-element. A classic example is "*What do you wonder who saw?" where the wh-phrase "what" cannot be extracted across the embedded wh-question "who saw what."[16] This constraint highlights the sensitivity of wh-movement to intervening wh-elements in specifier positions.[15]
Subject islands prevent extraction from the subject position of a clause. For instance, "*Who has [a comment about] annoyed Dorothy?" is ungrammatical due to the attempt to extract from the subject NP "a comment about who."[15] This type often overlaps with complex NP variants, as in "*Who did you see the man that kissed?" where extraction targets the subject of the relative clause within the NP.[16]
Left-branch islands restrict extraction of the leftmost element from a noun phrase. An example is "*Whose did you buy [ ] book?" which fails because "whose" is the left branch of the DP "whose book."[16] Ross (1967) identified this as a specific prohibition on partial extraction within DPs.[15]
Coordinate structure islands forbid extraction from one conjunct in a coordinated phrase. For example, "*What did John buy beer and?" is ungrammatical, as "what" cannot be pulled from the second conjunct "and what."[16] This constraint, noted by Ross (1967), preserves the integrity of coordinated elements.[15]
Complex NP islands block extraction from a clause embedded within a complex noun phrase, such as a relative clause modifying a noun. The sentence "*Which book does Dorothy like the author who wrote?" is ungrammatical because extraction targets the relative clause inside the NP "the author who wrote which book."[15] Ross (1967) exemplified this with cases like "*Which book did John meet a child who read?"[16]
Non-bridge-verb islands arise with factive or non-bridge verbs like "regret," whose complements resist extraction. For instance, "*What does John regret that he ate?" is ungrammatical, unlike extractions under bridge verbs like "think."[17] These islands stem from the semantic or selectional properties of such verbs blocking long-distance dependencies.[17]
Ross's 1967 work remains foundational, cataloging these islands as empirical constraints on unbounded movement rules.[18] Notably, island effects show partial amelioration in echo questions, where violations become more acceptable, as echo wh-questions often bypass standard island sensitivity.[19]
Embedded and Complex Structures
Wh-Movement in Subordinate Clauses
Wh-movement in subordinate clauses occurs within embedded contexts such as indirect questions and relative clauses, where the wh-phrase targets the specifier position of the embedded CP, distinct from the auxiliary inversion observed in matrix wh-questions.[20][1] In these structures, the movement is subject to the same core principles as in main clauses, including successive-cyclic application through intermediate CPs, but without the requirement for I-to-C movement that triggers subject-auxiliary inversion in root questions.[1] For instance, the matrix question "Which wildebeest will the lions devour?" involves inversion, whereas the embedded counterpart "They wonder which wildebeest the lions will devour" places the wh-phrase in the Spec-CP of the subordinate clause without altering the auxiliary position.[20]
In indirect questions, wh-movement fronts the interrogative element to the embedded Spec-CP under verbs like "wonder" or "ask," forming structures such as "I wonder who John saw," where the wh-phrase originates from the object position and leaves a trace.[21] This contrasts with matrix clauses by lacking inversion, as in the ungrammatical "*I wonder who did John see," ensuring the subordinate clause retains declarative word order except for wh-fronting.[20] The embedded CP is often headed by a null complementizer equivalent to "that," though overt "that" appears in some dialects or historical varieties, such as Middle English examples like "he wiste wel… what that he wolde answere".[20] Complementizer deletion is possible in certain contexts, allowing forms like "I asked what he wanted" without an overt linker, but this does not affect the wh-movement itself.[21]
Relative clauses involve wh-movement of a relative pronoun or operator to the Spec-CP of the embedded clause, modifying a head noun, as in "the people who moved in next door," where "who" moves from the subject position.[20] Unlike indirect questions, relative clauses permit an optional wh-pronoun, alternating with "that" or null complementizers, yielding equivalents like "the people that you saw" or "the people you saw," all involving movement to Spec-CP with a silent head.[21] This optionality distinguishes relatives from indirect questions, where wh-fronting is obligatory, and no inversion occurs, maintaining the embedded clause's internal structure.[12] In object-relative constructions, such as "the book which you can read," the movement is straightforward from the object gap, paralleling indirect questions but integrated as a nominal modifier.[1]
These embedded movements highlight the bounded nature of wh-displacement within the subordinate CP, adhering to principles like Subjacency, which limits crossing of cyclic boundaries such as additional S or NP nodes, though long-distance dependencies are possible through successive cycles.[1] Overall, wh-movement in subordinate clauses underscores the uniformity of A'-movement operations across clause types while adapting to embedding constraints that preserve the host clause's declarative properties.[20]
Multiple Wh-Questions
Multiple wh-questions are interrogative constructions containing more than one wh-phrase, such as "Who saw what?", where these phrases inquire about different arguments or elements in the clause. In English, one wh-phrase typically undergoes overt movement to the specifier position of the complementizer phrase (Spec-CP), while the remaining wh-phrases remain in situ in their base-generated argument positions. This pattern reflects a general constraint on wh-movement in single wh-fronting languages, ensuring that only the highest-ranked wh-phrase is extracted to the clause-initial position.[22][23]
The ordering of wh-phrases in multiple wh-questions is governed by the superiority condition, a hierarchical principle that prioritizes structurally higher positions for movement. In English, the subject wh-phrase must precede the object wh-phrase; thus, "Who saw what?" is grammatical, whereas "*What did who see?" is ill-formed or marginal. This condition, originally formulated as applying to transformations where no element dominates another in a certain way, enforces a subject-over-object hierarchy to avoid crossing dependencies during successive-cyclic movement. Superiority effects extend to languages with multiple wh-fronting, such as Bulgarian, where all wh-phrases are extracted clause-initially but must maintain the superior order, with the subject wh-phrase appearing highest in the cluster.[24][25][26]
Violations of the superiority condition can sometimes yield pair-list interpretations, where the answer pairs each wh-phrase with a corresponding element (e.g., "John saw Mary, and Bill saw Sue" for "Who saw who?"), particularly in contexts allowing functional application or when the violating order aligns with discourse focus. However, such readings are less robust than in superiority-obeying structures and often require specific pragmatic licensing. Exceptions to strict superiority arise when one wh-phrase is an adjunct rather than an argument; for instance, an object wh-phrase may front over an adjunct wh-phrase without violation, as the hierarchy places arguments above adjuncts. Additionally, superiority effects can be ameliorated through focus marking or in echo-question contexts, improving acceptability by reinterpreting the structure as non-interrogative or discourse-driven.[27][28][29]
Syntactic Representation
Syntax Trees and Headedness
In syntactic representations of wh-movement within generative grammar, the process is typically illustrated through tree diagrams that depict the base structure and subsequent displacement of the wh-phrase to the specifier position of CP (Spec-CP). For the English question "What did John see?", the underlying structure originates with the wh-phrase in object position within the VP, as part of an IP (or TP in later frameworks). The movement involves the wh-phrase displacing to Spec-CP, leaving a trace (t) in its original site, while the auxiliary "did" undergoes head movement to C. This is represented in bracketed notation as:
[CP What_i [C' did [IP John [VP see t_i ]]]]
[CP What_i [C' did [IP John [VP see t_i ]]]]
Such trees highlight the hierarchical phrase structure governed by X-bar theory, where phrases are labeled as XP (maximal projection), X' (intermediate), and X^0 (head), ensuring uniform branching and headedness.[20]
A key aspect revealed by these trees is that wh-movement constitutes A-bar movement, targeting a non-argument position (Spec-CP) rather than an A-position like subject or object slots, distinguishing it from movements such as subject raising. The landing site, Spec-CP, can involve either substitution (filling an empty specifier) or adjunction (adjoining to the phrase), depending on the framework, but both serve to establish the necessary syntactic dependencies for question formation. In minimalist terms, this movement often satisfies wh-feature checking, where the [+wh] feature on the wh-phrase agrees with and checks the corresponding feature on C, rendering the derivation convergent.[20]
Headedness plays a crucial role in these representations, parameterized in X-bar theory as the head directionality parameter, which determines whether heads precede (head-initial, as in English) or follow (head-final) their complements and specifiers. In head-initial languages like English, the complementizer C appears to the right of Spec-CP in the tree (e.g., [CP Spec C']), facilitating leftward wh-movement to the specifier. In contrast, head-final languages reverse this order (e.g., [CP C' [C head [Spec]]]), potentially altering the direction or mechanics of movement while preserving the core Spec-head relation for feature checking. This parametric variation accounts for cross-linguistic differences in phrase structure without altering the universal principles of movement.[30]
In Government and Binding (GB) theory, extraction islands are formally modeled as structural barriers to movement, visualized in syntax trees to demonstrate violations of locality constraints on traces left by wh-movement. The Subjacency Condition, a core principle in this framework, prohibits a moved wh-phrase from crossing more than one bounding node—typically NP and S (later refined to IP and DP)—in a single step of the derivation, ensuring successive-cyclic movement through intermediate positions like Spec-CP.[8] This condition accounts for island effects by showing how improper paths in trees block extraction, as detailed in Chomsky's seminal work.[31]
A classic wh-island violation illustrates this barrier-crossing in tree terms. Consider the ungrammatical sentence "*What do you wonder [CP who [IP t_who saw t_what]]?" Here, the wh-phrase "what" originates inside the embedded IP and targets the matrix Spec-CP. The syntax tree reveals the movement path crossing two bounding nodes: the embedded S/IP (containing the trace) and the matrix S/IP (dominating the embedded CP), without an intermediate landing site to satisfy subjacency.[8] Schematically, the relevant structure is:
[CP What [IP you [VP wonder [CP who [IP t_who [VP saw t_what ]]]]]]
[CP What [IP you [VP wonder [CP who [IP t_who [VP saw t_what ]]]]]]
The trace t_what is improperly bounded, as the derivation cannot cycle through the embedded Spec-CP (already occupied by "who"), leading to deviance.[31]
Subjacency similarly constrains extraction from complex NPs and adjuncts, as trees expose multiple barriers. For a complex NP island, the ungrammatical "*Who did you read [NP the book [CP that t_who wrote the book]]?" shows the wh-phrase "who" moving from within the relative clause CP, crossing the bounding NP ("the book") and the embedded S/IP. The tree structure highlights the NP as an impermeable barrier L-marked only by its head, preventing the trace from being governed across it:
[CP Who [IP you [VP read [NP the book [CP that [IP t_who [VP wrote the book ]]]]]]]
[CP Who [IP you [VP read [NP the book [CP that [IP t_who [VP wrote the book ]]]]]]]
In adjunct islands, such as "*Why did you leave [PP because [IP John [VP saw t_why ]]]?", the tree depicts the adjunct PP as containing an S/IP bounding node, with movement crossing both the adjunct's S/IP and the matrix S/IP, again violating the single-bounding-node limit.[8] These configurations underscore how trees formalize islands as derivational impossibilities under subjacency.[31]
Chomsky's GB theory (1981) integrates these island phenomena through bounding nodes and the Empty Category Principle (ECP), where trees reveal not only subjacency violations but also improper government of traces—requiring antecedents to govern traces via head-government, antecedent-government, or theta-government.[8] For instance, in island extractions, the trace lacks a local governor due to intervening maximal projections, as the tree paths show blocked c-command relations.[31]
Coordinate islands further exemplify formal barriers in GB trees, where extraction from one conjunct is blocked. In "*What did John buy apples and t_what?", the coordinate structure is often represented as an &P with conjuncts (NP "apples" and the trace) as sisters under a shared specifier, preventing proper government of the trace across the conjunction node, which acts as a barrier akin to an adjunct.[8] The tree illustrates:
[VP John [V' buy [&P [NP apples] and [NP t_what ]]]]
[VP John [V' buy [&P [NP apples] and [NP t_what ]]]]
This shared structure ensures the trace cannot be antecedent-governed from the matrix Spec-CP without violating ECP locality, formalizing the Coordinate Structure Constraint within GB's government-based model.[31]
Cross-Linguistic Variation
Languages with Wh-Movement
Wh-movement is a prominent feature in many Indo-European languages, where interrogative elements are displaced to the left periphery of the clause, often targeting the specifier position of the complementizer phrase (Spec-CP). Similar to English, languages such as German, French, and Bulgarian exhibit this fronting, though with variations influenced by clause type and verb positioning. These languages share core properties like sensitivity to extraction constraints but diverge in phenomena such as verb-second (V2) effects and multiple extractions.[32]
In German, wh-movement in matrix questions interacts with the V2 constraint, requiring the finite verb to raise to the C head while the wh-phrase occupies Spec-CP. For instance, the question "Wen hast du gesehen?" ('Whom have you seen?') derives from underlying subject-verb-object order by fronting the wh-object and positioning the auxiliary verb second.[33] In embedded wh-questions, the wh-phrase fronts to the initial position of the subordinate clause (Spec-CP), with the finite verb appearing in final position, as in "Ich weiß nicht, wen du gesehen hast" ("I don't know whom you have seen"), where "wen" is fronted within the embedded clause. This distinction highlights German's sensitivity to matrix-embedded asymmetries in wh-displacement.[34]
French employs wh-movement to the clause-initial position, accompanied by stylistic inversion in subject wh-questions, where the subject follows the verb. The example "Qui as-tu vu?" ('Who have you seen?') illustrates this, with the subject wh-phrase fronted and the auxiliary verb inverted before the lexical verb and subject.[35] This inversion is obligatory in formal registers for subject questions but optional or absent in object ones, underscoring register-based variation.[36]
Bulgarian permits multiple wh-movement, allowing all wh-phrases in a question to front to the left periphery, often forming a cluster. In "Koj kogo vidja?" ('Who whom sees?'), both wh-elements are extracted, with the higher one in Spec-CP and the lower potentially in an adjacent position, preserving hierarchical order.[37] Recent cartographic analyses in the 2020s propose that this involves multiple specifiers or projections in the CP domain, attracting wh-phrases stepwise from closest to farthest, as evidenced in studies of Slavic left-periphery structure.[38] This multiple extraction contrasts with single wh-fronting in other languages but aligns with Bulgarian's pro-drop and topic-prominent traits.[22]
Across these languages, wh-movement converges on Spec-CP as the landing site, facilitating scope and focus marking akin to English.[32] Pied-piping operates similarly, where a containing phrase fronts with the wh-element, as in German prepositional wh-questions like "Mit wem hast du gesprochen?" ('With whom did you speak?'). Extraction islands, such as relative clauses or coordinate structures, constrain movement with variable strictness—stronger in French than in Bulgarian, where multiple whs may evade some barriers.[15] These shared and divergent patterns underscore wh-movement's role in encoding interrogative force cross-linguistically.[39]
Languages with Partial or No Wh-Movement
In languages with partial or no wh-movement, question formation relies on in-situ wh-phrases, scope markers, or prosodic cues rather than obligatory fronting to Spec-CP, leading to distinct interpretive mechanisms and often weaker island constraints compared to full-movement languages.[40] These strategies highlight cross-linguistic variation in how interrogative scope is established without overt displacement to the clause periphery.[41]
Mandarin Chinese exemplifies a language with no overt wh-movement, where wh-phrases remain in their base-generated positions to form questions, as in Ni xihuan shei? ('Who do you like?'), avoiding Spec-CP displacement entirely.[42] Matrix scope for these in-situ wh-elements is achieved through unselective binding by a covert question operator, which binds multiple wh-phrases simultaneously without requiring movement, thus permitting wide scope interpretations even from embedded positions.[42] In multiple wh-questions, superiority-like intervention effects arise, where a wh-phrase closer to the probe is preferentially attracted, mirroring constraints in movement languages but resolved covertly.[43] Post-2015 analyses have revisited these patterns, proposing unified accounts where island insensitivity for non-causal wh-phrases stems from binding at varying adjunction heights rather than late PF-movement, reinforcing the role of semantic modification over syntactic relocation.[41]
Languages like Japanese and Korean feature pure wh-in-situ strategies, with no overt movement and scope marked by dedicated question particles, such as Japanese -ka or Korean -ni/-nunci, as in Japanese Dare-ga nani-o tabeta no ka? ('Who ate what?').[44] In these systems, interrogative scope is disambiguated prosodically through rising intonation contours or pitch compression aligning with the wh-domain, particularly in multiple questions where phrasing boundaries reflect scope relations.[44] Experimental evidence from South Kyeongsang Korean confirms that wh-intonation forms a high plateau for matrix scope, contrasting with embedded readings marked by particles like -na.[44]
Partial movement strategies also appear in Hungarian, where wh-phrases undergo focus-driven displacement to embedded CPs, paired with a scope-marking expletive like mi ('what') in the matrix clause, as in Mit gondolsz, mi történt? ('What do you think happened?').[45] This wh-scope-marker construction handles multiple wh-phrases indirectly via dependencies, differing from full movement by licensing partial raising under focus projections and obligatory complementizers, without direct chaining to the matrix periphery.[45]
Across these languages, the absence of Spec-CP movement results in weaker or absent island effects for most wh-phrases, as binding or prosodic licensing bypasses extraction constraints that constrain fronting languages; causal wh-adverbs remain sensitive, however, due to higher adjunction requirements.[41]
Theoretical Frameworks
In the early stages of transformational generative grammar, wh-movement was treated as an unbounded rule in interpretive semantic models, where question formation involved deletion of a wh-element from its base position, subject to initial constraints identified by Ross (1967) in his dissertation on extraction islands. These constraints, such as complex NP islands and coordinate structure islands, limited extractions and were initially viewed as performance limitations or transderivational rules, but they laid the groundwork for integrating locality into core grammar.
Chomsky's 1973 work in the Extended Standard Theory (EST) marked a pivotal shift, introducing Move Alpha as a general operation that unified wh-movement with other displacements, replacing language-specific rules with universal conditions on transformations. In this framework, wh-movement was formalized as an instance of A-bar movement to the specifier of COMP (later CP), driven by the need to interpret questions, with Ross's island constraints reinterpreted through the Subjacency Condition, which prohibits movement across more than one bounding node (e.g., NP or S) in a single step. Superiority effects, where a subject wh-phrase must precede an object wh-phrase in multiple questions (e.g., *Who do you think what saw?), were explained as arising from the structural hierarchy, favoring movement of the structurally higher wh-element first to satisfy hierarchical ordering.
Within the Government and Binding (GB) theory outlined by Chomsky (1981), wh-movement evolved into a successive-cyclic process involving intermediate steps through specifier positions of intermediate CPs, ensuring compliance with locality constraints like Subjacency and the Empty Category Principle (ECP). The ECP required traces of wh-movement to be properly governed, either by an antecedent or a local head (e.g., V or Infl), preventing ungoverned traces in island contexts like subject islands (e.g., *Who do you think that pictures of t is on sale?). This successive cyclicity, further elaborated in Chomsky (1986), posited that wh-phrases move step-by-step to evade barriers, with islands arising as violations of bounding nodes or government domains.
Central to the GB analysis was the [+wh] feature on the complementizer C, which attracts the wh-phrase to its specifier for feature checking and proper binding of the trace, ensuring the chain formed by the moved wh and its traces is antecedent-governed. Ross's original constraints were thus fully incorporated into the modular core grammar via these principles, transforming descriptive island phenomena into parametric effects of universal bounding and government conditions. This approach emphasized wh-movement as a structure-dependent, cyclic A-bar operation, distinguishing it from A-movement while accounting for long-distance dependencies through chain formation and locality.
Alternative and Contemporary Approaches
In the Minimalist Program, initiated by Chomsky in 1995, wh-movement is driven by an edge feature on phase heads such as vP and CP, which attracts wh-phrases to the phase edge for feature checking via the Agree operation prior to overt movement.[46] This framework reduces wh-movement to general principles of economy and locality, where Agree establishes a relation between a probe (e.g., C with a Q-feature) and a goal (the wh-phrase), followed by internal Merge if the edge feature requires pied-piping. The Phase Impenetrability Condition further constrains extraction by rendering the complement of a phase (e.g., TP within CP) inaccessible for operations outside that phase, explaining island effects as failures to escape phase edges in successive-cyclic movement.[47] For multiple wh-questions, multiple Agree relations with the phase head allow clustering at the edge without violating locality, as seen in languages like Bulgarian where wh-phrases front successively.[48]
Alternative approaches challenge the transformational nature of wh-movement. Base-generation posits that wh-phrases are merged directly in Spec-CP, with no displacement from an argument position, accounting for apparent movement effects through interpretive licensing rather than syntactic operations; this view gains support in analyses of scrambling and long-distance dependencies where reconstruction is absent.[49] PF-movement theories treat certain reconstruction effects in wh-movement as arising from phonological adjustment at the syntax-PF interface, as in analyses of English and Japanese data.[50] In Construction Grammar, wh-questions emerge from form-meaning pairings stored as constructions, obviating discrete movement rules in favor of holistic patterns shaped by usage and frequency, as in Goldberg's framework where interrogative schemas integrate wh-forms without transformational derivations.
Contemporary developments refine these ideas within cartographic and constraint-based models. Rizzi's 1997 cartographic approach splits the CP domain into a fine-grained hierarchy (e.g., ForceP, TopicP, FocusP, FinP), positioning wh-movement to a dedicated FocusP or InterrogativeP for scope and illocutionary force, allowing precise mapping of left-peripheral phenomena across languages.[51] Optimality Theory applies ranked constraints to wh-movement, where faithfulness to base positions (e.g., Stay) conflicts with scope requirements (e.g., Wh-Scope), yielding surface forms like partial movement as optimal resolutions rather than parametric choices; Grimshaw and Samek-Lodovici's work illustrates this for functional projections in questions.[52]
Recent 2020s research addresses gaps in minimalist derivations, particularly labeling challenges in wh-subject movement, as explored in analyses that permit movement from Spec,TP to Spec,CP despite apparent blocks.[53] For in-situ wh in Mandarin, empirical studies confirm covert movement at LF to satisfy intervention effects and scope, with eye-tracking revealing processing akin to overt dependencies in wh-ex-situ languages, supporting Agree-based licensing without surface displacement.[54]