Raising
In linguistics, raising refers to a type of syntactic construction in which the subject of an embedded infinitival clause is displaced to the subject position of the main (matrix) clause, without receiving a thematic role from the matrix verb.[1] This movement, often termed subject-to-subject raising, is typically triggered by the need for nominative case assignment, as raising predicates like seem, appear, or be likely do not assign an external θ-role or accusative case to their subjects.[2] For example, in the sentence "John seems to be happy," the NP John originates as the subject of the embedded clause to be happy but raises to the matrix subject position, allowing the construction to satisfy syntactic requirements while preserving the semantic interpretation that John is the agent of happiness.[3] Raising constructions are distinguished from control structures through several diagnostic tests, reflecting their differing syntactic and semantic properties.[3] Unlike control verbs (e.g., want or try), which assign a θ-role to their subjects and involve a null pronoun (PRO) in the embedded clause controlled by the matrix subject, raising verbs permit expletive subjects like it or there without altering meaning, as in "It seems to be raining" or "There appears to be a problem."[2] Additionally, raising preserves the idiomatic integrity of embedded phrases, such as "The cat seems to be out of the bag," whereas control constructions disrupt it (e.g., "*The cat wants to be out of the bag").[3] These tests highlight that raising involves pure syntactic movement from the embedded clause's specifier position to the matrix clause, often across an infinitival complement lacking a full CP layer.[1] The phenomenon has been central to generative syntax since the 1970s, informing theories of movement, case, and θ-role assignment, and it varies cross-linguistically, with parallels in languages like French (via sembler) or restrictions in others due to parameters like the verb raising parameter in Romance languages.[2] Raising also interacts with other processes, such as passivization, where the raised element can shift positions while maintaining truth conditions, as in "Mary is believed to be intelligent" paralleling "It is believed that Mary is intelligent."[1] Overall, raising exemplifies how syntax can decouple surface structure from underlying semantics, a core insight in modern linguistic theory.[3]Overview
Definition
In linguistics, raising refers to a syntactic process in which an argument, typically the subject of an embedded clause, moves to a higher position in the matrix clause within a biclausal structure. This movement results in the raised element appearing to function as the subject or object of the matrix verb, though it receives its thematic role (theta-role) from the embedded predicate rather than the matrix one. The phenomenon was first systematically analyzed in generative grammar as a transformational rule deriving such structures from underlying representations where the argument originates in the subordinate clause.[4] Raising predicates, such as seem, appear, or happen, are distinctive in that they contribute little to no semantic content and fail to theta-mark the surface subject of the matrix clause. Instead, these verbs function primarily as facilitators for the structural raising, allowing the embedded argument to surface in the higher position without assigning it an external thematic role. For instance, in the sentence "John seems to be happy," the noun phrase "John" originates as the subject of the embedded infinitival clause "to be happy" and raises to the matrix subject position, where it receives its agentive or experiencer role solely from the embedded adjective.[5][6] A key diagnostic distinguishing raising verbs from control verbs involves the permissibility of expletive insertion in the matrix clause. Raising constructions allow an expletive like it to occupy the subject position when no argument raises, as in "It seems to rain," demonstrating that the matrix verb does not select for a specific thematic subject. In contrast, control verbs, such as try or want, require a controller that bears a thematic role from the matrix clause and disallow such expletives. This property underscores the non-argument status of the raised element relative to the matrix predicate.[6]Historical Development
The concept of raising constructions originated in the transformational generative grammar framework pioneered by Noam Chomsky during the 1960s, which posited rules that derive surface sentence structures from underlying deep structures through successive transformations. Although Chomsky's "Aspects of the Theory of Syntax" (1965) did not explicitly formalize raising, it established the theoretical apparatus for analyzing embedded clauses and predicate complements, enabling later derivations of movement operations in infinitival constructions.[7] This work shifted focus from structuralist descriptions of immediate constituents to explanatory rules accounting for syntactic relations, with raising emerging as a key example of such derivations. Pre-Chomskyan structuralist linguistics, such as that of Leonard Bloomfield or Otto Jespersen, discussed complementation and copular verbs but lacked the movement-based analysis that defined raising, treating such phenomena primarily as surface-level juxtapositions without deep structure motivations. A pivotal advancement came with Peter Rosenbaum's 1967 monograph, which introduced the notion of subject raising as a transformational rule whereby the subject of an embedded infinitival clause moves to the matrix clause subject position, as in derivations involving verbs like "seem." Joseph Emonds's 1970 dissertation further refined this by classifying raising among "root transformations," which apply only at the sentence periphery and preserve underlying structures, distinguishing them from more local operations.[8] Paul Postal's influential 1974 book "On Raising" provided extensive diagnostics to identify raising verbs, such as their behavior under passivization and lack of semantic selection for subjects, solidifying raising as a distinct category separate from control constructions and challenging earlier unified analyses of infinitivals.[9] In the evolution toward Government and Binding (GB) theory, Chomsky (1981) reanalyzed raising as NP-movement (A-movement), motivated by principles like the Extended Projection Principle (EPP) requiring subjects and the Empty Category Principle governing traces, integrating it into a modular system with theta theory and case theory to explain why raising verbs assign no thematic roles to their apparent subjects. This framework emphasized universal constraints on movement, positioning raising as a canonical instance of structure-preserving relocation. By the 1990s, the Minimalist Program recast raising as feature-driven internal Merge, where an uninterpretable EPP feature on T-heads attracts the embedded subject for checking, reducing it to economy-driven operations without reference to deep vs. surface structures. Chomsky (1995) argued this unification minimizes theoretical apparatus, aligning raising with broader derivational economy principles.Types of Raising Constructions
Raising-to-Subject
Raising-to-subject is a syntactic construction in generative grammar where the subject of an embedded infinitival clause moves to the subject position of the matrix clause, with the matrix verb assigning no theta-role to that subject.[10] This movement results in structures where the apparent subject of the main clause originates deeper in the sentence, as first systematically argued in Postal's analysis of English grammar.[11] Common English raising-to-subject verbs include seem, appear, happen, turn out, tend, prove, fail, and continue.[10] These verbs are characterized by their semantic neutrality; they contribute little independent meaning and serve primarily to license the raising operation without imposing selectional restrictions on the raised subject.[6] For instance, in "The earth seems flat," the deep structure underlying the sentence is "[It] seems [the earth [is] flat]," where "the earth" begins as the subject of the embedded clause and raises to the matrix subject position.[12] Similarly, "John happens to be sick" derives from a structure like "[It] happens [John [is] sick]."[13] A key diagnostic of raising-to-subject is the alternation with expletive constructions, where the raised subject is replaced by an impersonal "it" and the embedded clause becomes finite.[10] Examples include "It seems that the earth is flat" or "It appears to be raining," which preserve the propositional content while avoiding raising.[10] This alternation highlights the optional nature of the movement in certain contexts. Raising-to-subject constructions interact with tense and aspect by preserving the temporal and aspectual features of the embedded clause, as the matrix verb does not intervene in their licensing.[10] For example, "John seems to have left" embeds a perfect aspect ("have left") that is unaffected by the matrix present tense, allowing the construction to convey completed action prior to the seeming.[12] Such patterns can also yield evidential interpretations, as in "This seems to have been common practice," where the perfect suggests indirect evidence.[12] While the core inventory of raising-to-subject verbs is limited, rarer or dialectal variants exist, such as chance in some Southern US English varieties, as in "He chanced to see her," implying happenstance without a controlling agent.[14] These less common verbs extend the construction's productivity in regional dialects but retain the same core properties of subject movement and semantic bleaching.[10]Raising-to-Object
Raising-to-object is a syntactic construction in which the subject of an embedded infinitival clause undergoes A-movement to occupy the object position of the matrix verb, receiving its thematic (theta-) role solely from the embedded predicate rather than the matrix verb.[15] This phenomenon, also known as subject-to-object raising, contrasts with control structures where the matrix verb assigns a theta-role to its object.[15] In English, raising-to-object typically involves verbs that select a non-finite clausal complement without providing an external argument role to the raised element.[9] Common English raising-to-object predicates include verbs of cognition, perception, and declaration (often termed "believe-type" verbs) such as believe, prove, consider, find, think, and suppose, which embed infinitival clauses and allow the embedded subject to surface as the matrix object in the accusative case.[16] These verbs participate in the "accusative and infinitive" (ACI) construction, where the raised subject precedes the infinitival complement.[16] A representative example is "They proved him to be innocent," where the deep structure involves the embedded clause [him [to be] innocent] as the complement of prove, and him raises to object position without receiving a theta-role from prove.[15] In contrast, a non-raising construction like "They wanted him to leave" assigns a theta-role (e.g., theme or goal) to him from the matrix verb want, treating the infinitival as a controlled clause rather than a raised one.[15] Another example is "I consider John to be intelligent," illustrating how the raised object John originates as the subject of the embedded predicate to be intelligent.[15] A key diagnostic for raising-to-object is the passive alternation, which passivizes the matrix verb and promotes the raised object to subject position while preserving the original meaning and theta-role assignment.[15] For instance, "They proved him to be innocent" alternates to "He was proved to be innocent," confirming that him (now he) bears no theta-role from the matrix verb, as the passive form does not alter the embedded semantics.[15] This test distinguishes raising from object control, where passivization (e.g., "He was wanted to leave") yields an infelicitous or semantically shifted result due to the lost theta-role.[15] Similarly, "John is believed to be guilty" exemplifies the nominative and infinitive (NCI) passive counterpart to ACI constructions with believe.[16] Perception verbs such as see and hear exhibit optional raising-to-object behavior, allowing bare infinitival complements where the embedded subject raises to object position, as in "I saw him leave the room," analyzed as movement from the embedded clause without a theta-role from the matrix verb.[9] This optionality contrasts with their gerundive alternates (e.g., "I saw him leaving"), which involve distinct aspectual interpretations rather than raising.[17]Distinguishing Raising from Control
Key Diagnostics
Key diagnostics for identifying raising constructions in syntax rely on empirical tests that probe the thematic and structural properties of predicates, distinguishing them from control constructions. These tests exploit the fact that raising predicates do not assign a thematic role to their surface subject, which originates in the embedded clause, whereas control predicates do assign such a role to their subject or object.[3][18] Common diagnostics include expletive insertion, extraposition alternations, selectional restrictions, and passivization effects, each providing observable evidence for the non-thematic status of the raised argument. The there-insertion test identifies raising-to-subject predicates by checking whether an expletive there can appear as the matrix subject with an existential embedded predicate. For example, There seems to be a problem is grammatical because seem does not select a thematic subject, allowing the expletive to satisfy matrix requirements while the existential be licenses the associate a problem. In contrast, control predicates like try reject this: There tries to be a problem is ungrammatical, as try demands a thematic agent. This test, rooted in the inability of raising verbs to theta-mark their subjects, reliably distinguishes the classes.[18][3] The it-extraposition alternation further diagnoses raising-to-subject by permitting a finite clause with an expletive it in the matrix, as in It seems that John left, where the embedded subject John is not selected by seem. This alternation is impossible with control predicates: It tries that John leaves is ill-formed because try requires a thematic subject. The test highlights the non-argument status of the surface subject in raising, as the matrix predicate tolerates a dummy subject without thematic implications.[3] Selectional restrictions provide a semantic diagnostic, as raising predicates impose no theta-role on their subject, allowing semantically incompatible fillers. For instance, The rock seems to regret leaving is acceptable despite the rock being an implausible agent for regret, since seem selects nothing and the embedded verb governs semantics. Control predicates like try, however, enforce restrictions: The rock tries to regret leaving is anomalous because try demands an agentive subject. This test underscores the embedded origin of the subject in raising.[18][19] For raising-to-object constructions, passivization serves as a key diagnostic, preserving the embedded subject as the matrix surface subject without altering truth conditions. Consider John believes Mary to be intelligent, which passivizes to Mary is believed by John to be intelligent, where Mary raises to matrix subject position. This holds because the matrix verb believe does not theta-mark its apparent object. Control counterparts like John persuades Mary to leave passivize to Mary is persuaded by John to leave, but the thematic role assignment differs, as persuade selects Mary as a goal. The test confirms the non-thematic status of the raised element in object raising.[20] Corpus-based studies validate these diagnostics quantitatively by simulating grammaticality patterns in large datasets, achieving high accuracy in classifying verbs. For example, using unannotated corpora like Gigaword (1.7 billion words) and the Yahoo! Web corpus, regression tree models trained on frequency features from there-insertion patterns (e.g., "there V to be") distinguish raising from control verbs with mean squared errors of 0.134 and 0.110, respectively. Such approaches confirm the robustness of the tests across natural language use.[21]Theoretical Differences
Raising predicates are semantically neutral with respect to their surface subjects, assigning no theta-roles to the raised argument; instead, the embedded predicate provides the thematic interpretation for that argument. This neutrality arises because raising verbs lack an external argument structure, satisfying the theta-criterion solely through the embedded clause's projections. In contrast, control constructions involve verbs that theta-mark their controller, such as the subject in "John promised to leave," where the controller (matrix subject) receives an agent theta-role from the matrix verb "promise", while PRO receives its theta-role solely from the embedded clause. This difference underscores the argumenthood of the controller in control, which bears multiple thematic relations, whereas the raised element in raising receives only one theta-role from the embedded predicate. These structural distinctions have significant implications for binding theory. In raising constructions, the raised subject c-commands and can bind into the embedded clause, as in "John seems to himself to be intelligent," where "John" binds the reflexive "himself" in the embedded position.[22] Controllers in control structures, however, do not exhibit the same binding potential into the embedded clause due to the locality constraints on PRO, which functions as an anaphoric pronominal unable to establish such relations across clause boundaries. Theoretical debates surrounding raising center on whether it involves syntactic movement or alternative base-generation mechanisms. The standard analysis posits A-movement of the embedded subject to the matrix subject position, forming a chain that preserves theta-role assignment from the base position (Postal 1974).[23] Non-movement approaches, such as those employing small clause structures, argue that the surface subject is base-generated in the matrix clause, with the complement forming a tenseless small clause lacking a full CP layer, as proposed in analyses of infinitival complements without vP projection (Wurmbrand 2001; Bresnan 1978).[23] Within the minimalist framework, phase theory further refines these analyses by imposing cyclic constraints on movement via the Phase Impenetrability Condition (PIC), which limits access to completed phases like vP and CP (Chomsky 2001).[24] For raising, defective phases—such as those with a defective C or T—enable successive-cyclic movement through phase edges, allowing the subject to escape the embedded domain without violating locality, unlike the phase-bound restrictions in control where PRO remains local (Chomsky 2008).[24] This integration of phases optimizes derivation efficiency while maintaining the core distinctions in theta-assignment and binding.Syntactic Representations
Constituency Trees
In constituency-based syntactic theories, such as those within the Government and Binding framework, raising constructions are represented through hierarchical trees that depict the deep structure and subsequent transformations via NP-movement. The deep structure for a raising-to-subject verb like "seem" features an embedded infinitival clause (IP) as the complement of the matrix verb, with the raised NP originating in the subject position of the embedded IP, leaving a trace (t) to indicate the movement gap. For instance, in the sentence "John seems to sleep," the deep structure tree positions "John" as the subject of the embedded IP under "seem," as follows:This representation highlights the embedded clause's full clausal status, with the trace co-indexed to the originating NP, preserving argument structure across clauses. Following NP-movement, the surface structure tree shows the NP "John" raised to the matrix subject position, adjoined to the matrix S, while the trace remains in the embedded subject position to satisfy the verb "sleep"'s theta-role requirements. The resulting tree for "John seems to sleep" can be notated in labeled bracketing as: [S [NP John_i] [VP seems [S' to [VP sleep t_i ]]]] This structure captures the expletive-like behavior of "seem," which assigns no external theta-role, allowing the embedded subject to raise without violating the projection principle. The co-indexation (i) between "John" and its trace ensures locality and c-command relations, crucial for binding and reconstruction effects in raising. Constituency trees excel in modeling island effects in raising, such as complex NP islands, where movement out of certain embedded domains is blocked, as the hierarchical layering enforces subjacency constraints that prevent crossing constituency boundaries. For example, the ungrammaticality of "*John seems [the claim that t_i left] to be true" arises because the trace violates islandhood within the NP-headed constituent, a phenomenon constituency representations diagnose through branch domination rather than linear relations. Labeled bracketing further aids accessibility by explicitly marking phrasal nodes (e.g., NP, VP, S), facilitating analysis of adjunct islands or wh-island violations specific to raising paths. These tree structures tie briefly to movement rules in transformational grammar, where raising exemplifies A-movement driven by case and theta-criterion satisfaction.S | VP | \ seem IP | S' / \ to VP | V | sleep / \ NP t_i | John_iS | VP | \ seem IP | S' / \ to VP | V | sleep / \ NP t_i | John_i