Projection principle
The Projection Principle is a core constraint in generative linguistics, formulated by Noam Chomsky in 1981, stipulating that representations at each syntactic level—including D-structure, S-structure, and Logical Form (LF)—must be projected from the lexicon such that they observe the subcategorization properties of lexical items.[1] This ensures that the thematic structure and argument requirements specified in lexical entries, such as a verb's need for complements or specifiers, are categorially represented throughout the syntactic derivation.[2]
Within Chomsky's Government and Binding (GB) theory, the Projection Principle functions as a bridge between the lexicon and syntax, interacting with modules like Theta Theory—which governs the assignment of thematic roles—and X-bar theory, which regulates phrase structure projections.[3] For example, in passive constructions like "The book was read t by John," the trace (t) in the object position preserves the verb "read"'s subcategorization for an internal argument at all levels, preventing ill-formed structures where lexical properties are violated.[1] It also underpins the requirement for traces or empty categories in movement operations, such as wh-movement in questions, to maintain theta-role assignments.[2]
The principle is closely related to but distinct from the Extended Projection Principle (EPP), which additionally requires that every clause project a subject in the specifier of Tense Phrase (TP).[3] As a foundational element of GB theory, the Projection Principle has influenced subsequent frameworks, including the Minimalist Program, where its effects are often derived from broader constraints like Full Interpretation and the uniformity of theta-role assignment, ensuring economical mapping of lexical items to syntactic structures.[4]
Introduction
Core Definition
The Projection Principle is a fundamental constraint in generative syntax, as formulated by Noam Chomsky, that requires the lexical properties of syntactic heads—specifically, their subcategorization frames (c-selection) and theta-marking requirements (s-selection)—to be preserved at every level of syntactic representation, including D-structure, S-structure, and Logical Form (LF).[5] This ensures that the argument structure specified in the lexicon for a head, such as a verb's requirement for a direct object or complement, is faithfully reflected in all syntactic derivations without alteration or omission.[5]
Chomsky's 1981 formulation in Lectures on Government and Binding provides the precise statement of the principle:
- If \alpha selects \beta in \gamma as a lexical property, then \alpha selects \beta in \gamma at some level L_i.
- If \alpha selects \beta in \gamma at level L_i, then \alpha selects \beta in \gamma at level L_j.[5]
This bidirectional condition guarantees that selectional relations originating from the lexicon project into the syntax and remain invariant across transformationally derived levels, preventing rules from introducing spurious arguments or deleting obligatory ones.[5] For instance, if a verb lexically selects a noun phrase complement, that complement must be represented (possibly via traces or empty categories) at each syntactic level to satisfy the principle.[5]
The Projection Principle thus enforces a direct linkage between lexical entries and syntactic structures, constraining transformational operations to operate within the bounds of lexical specifications rather than freely altering them.[5] It distinguishes the initial projection of heads and their arguments from lexicon to D-structure—where the basic phrase structure is built—from subsequent movement rules, which may displace elements but cannot violate the preserved selectional properties.[5]
The Projection Principle operates alongside the Theta-Criterion as a complementary constraint, ensuring proper assignment of thematic roles to arguments without over- or under-generation.[5]
Historical Origins
The Projection Principle was formally introduced by Noam Chomsky in his 1981 monograph Lectures on Government and Binding, where it served as a foundational constraint within the Government and Binding (GB) framework of generative grammar.[5] This principle addressed longstanding challenges in syntactic theory by ensuring that the subcategorization properties of lexical items, particularly verbs, remained invariant across different levels of syntactic representation.[5]
Its development was motivated by limitations in the earlier Standard Theory of syntax, as articulated in Chomsky's 1965 Aspects of the Theory of Syntax, where deep structure was posited as the primary locus for semantic interpretation and argument structure, but transformations to surface structure often disrupted the preservation of these elements.[6] In Standard Theory, lexical insertion occurred at deep structure to satisfy subcategorization frames, yet unrestricted transformations risked violating these frames, leading to inadequate accounts of phenomena like passivization and dative movement.[6] The Projection Principle thus provided a mechanism to enforce uniformity, preventing such violations by projecting lexical properties throughout the derivation.[5]
A key influence stemmed from Jerrold J. Katz and Paul M. Postal's 1964 work An Integrated Theory of Linguistic Descriptions, which argued for meaning-preserving transformations and emphasized the role of deep structure in mediating lexical insertion and subcategorization to maintain semantic integrity.[7] Katz and Postal's framework highlighted how syntactic structures must project semantic readings without alteration, laying groundwork for later constraints on syntactic operations.[7] This idea resonated in the Projection Principle's insistence on structural preservation, adapting it from semantic to syntactic domains.[5]
In the early 1980s linguistic context, the principle responded to advancements in trace theory, which posited empty categories left by movement operations, necessitating uniform constraints to ensure traces inherited and satisfied subcategorization requirements at all syntactic levels, including D-structure, S-structure, and Logical Form.[5] As part of the GB theory, it complemented emerging modules like theta-theory by safeguarding the projection of argument roles across derivations.[5]
Key Components
Subcategorization Frame Preservation
Subcategorization constitutes the lexical specification of the complements that a syntactic head requires, dictating the category and number of arguments it selects, such as a transitive verb mandating an NP object or a ditransitive verb requiring both an NP and a PP.[8] This specification forms part of the head's lexical entry and governs the syntactic structure in which it appears.[9]
The Projection Principle enforces the preservation of these subcategorization frames across all levels of syntactic representation, from D-structure to S-structure and Logical Form.[10] Formulated by Chomsky (1981), it stipulates that if a head α selects a complement β as a lexical property, then α selects β at every syntactic level, and conversely, selection at one level implies selection at all others.[11] At D-structure, the complete subcategorization frame is thus realized, with all required complements appearing as immediate constituents (sisters) of the head.[10]
Transformational rules, including movement operations like wh-movement or passivization, are constrained by this principle such that required arguments cannot be eliminated; instead, an empty category (typically a trace) must occupy the original position to uphold the head's subcategorization requirements.[9] This ensures that the argument structure remains intact, preventing ill-formed derivations where lexical properties are violated.[10]
A representative example involves the ditransitive verb give, which subcategorizes for both a direct object NP and an indirect object PP, as in the D-structure "John gave [NP the book] [PP to Mary]."[10] If the indirect object undergoes wh-movement to form "To whom did John give the book?", the PP position in the verb phrase must contain a trace (t) to satisfy give's subcategorization frame, yielding the structure where the trace co-indexed with "to whom" projects the required PP complement.[10] Similarly, movement of the direct object, as in "What did John give to Mary?", leaves a trace in the NP slot, preserving both arguments.[10]
The Projection Principle operates in tandem with X-bar theory to regulate this preservation, as heads project phrasal categories (e.g., VP, NP) that locally accommodate their subcategorization frames within the intermediate and maximal projections.[1] Under X-bar structure, the head's selectional demands are satisfied by positioning complements in the complement position of the head's bar-level projection, thereby maintaining hierarchical consistency and lexical integrity throughout the derivation.[1] These subcategorized positions serve as the loci for theta-role assignment to arguments.[8]
Theta-Role Projection
The Projection Principle integrates with Theta Theory by mandating that theta-roles, as specified in the lexical entries of predicates, must be represented in the syntax at every level of derivation, from D-structure to Logical Form (LF). This ensures that the semantic argument structure of a lexical head, such as a verb's theta-grid listing roles like agent and patient, is preserved throughout the syntactic computation. Chomsky (1981) formulates the principle to guarantee that representations at each syntactic level observe the theta-marking properties of lexical items, preventing any alteration or omission of these roles during transformations.[5]
Central to this integration is the Theta Criterion, which stipulates that each argument must bear exactly one theta-role, and each theta-role must be assigned to exactly one argument. As articulated by Chomsky (1981), this biunique mapping links syntax to semantics, with the Projection Principle enforcing its projection across levels.[5][12] Violations of the Theta Criterion, such as over- or under-assignment of roles, result in ungrammaticality, as the principle prohibits introducing or eliminating arguments post-lexically. For instance, in the sentence "John hit the ball," the verb "hit" assigns an agent role to "John" and a patient role to "the ball" at D-structure, and these roles must remain satisfied at S-structure and LF.[5][12]
Theta-role preservation occurs through syntactic chains and traces during movement operations, ensuring that displaced arguments retain their original thematic associations. When an argument moves, such as in passive constructions like "The ball was hit by John," a trace in the base position links back to the moved NP via a chain, maintaining the patient theta-role assignment from the verb. Haegeman (1994) explains that this mechanism upholds the Projection Principle at LF, where full interpretation requires all theta-grids to be saturated without residue. Thus, no theta-role can be lost or added after lexical insertion, as any such deviation would violate the principle's structure-preserving requirements.[12]
While theta-roles are inherently semantic, the Projection Principle ensures their syntactic visibility by projecting them alongside subcategorization frames, the syntactic counterpart to theta-assignment. This distinction highlights that theta-roles concern thematic interpretation, whereas subcategorization specifies structural complements, yet both must align across levels to yield grammatical sentences. Chomsky (1981) emphasizes this linkage to maintain the lexicon-syntax interface's integrity.[5]
Extended Projection Principle
The Extended Projection Principle (EPP), introduced by Noam Chomsky in 1982, stipulates that every clause must contain a subject, irrespective of whether the predicate assigns a thematic role to that position. This formulation addresses the empirical fact that finite clauses in many languages require an overt or covert element in the subject position, preventing constructions without subjects in non-pro-drop languages. The principle was developed within the Government and Binding theory to explain phenomena such as the obligatory presence of subjects in English, where subjectless sentences like "*Rains" are ungrammatical.[13]
Building on the Projection Principle, which preserves lexical properties like subcategorization and theta-role assignment throughout derivations, the EPP specifically enforces the projection of a subject into the specifier position of the inflectional phrase (Spec-IP) in finite clauses. This extension ensures that clausal structure universally projects a filled subject position, linking lexical projections to higher functional layers without altering the core conservation of argument structure.
The EPP operates through a requirement on the Infl (or Tense) head, which demands that a determiner phrase (DP) occupy Spec-IP, often achieved via movement or insertion. When no suitable argument is available for raising—such as in impersonal constructions—non-thematic expletives fulfill this role, as in English "There is a unicorn in the garden" or "It appears to be cold today." In pro-drop languages, such as Italian or Spanish, the EPP is satisfied by a null pronominal pro in the subject position, licensed by the verb's rich agreement features that identify its phi-properties. This mechanism blocks EPP violations in such languages, allowing subject omission without compromising clausal integrity.[14][13]
Locality of Selection
The Locality of Selection constraint in generative syntax stipulates that selectional relations—encompassing both subcategorization for syntactic categories and theta-role assignment to arguments—must be satisfied exclusively by immediate complements within the phrasal projection of the selecting head, without crossing phrasal boundaries.[15] This ensures that lexical properties dictating complement requirements are realized in a strictly local domain, aligning with the head-driven nature of phrase structure formation.[12]
In the X-bar theoretic framework, formalization of this locality requires that heads select only their sister constituents as complements, as non-local selection would violate the hierarchical organization of projections.[15] For example, a transitive verb like destroy subcategorizes for an NP complement, which must appear as a sister to the V-head within VP, as in Poirot destroyed the manuscript, where the NP the manuscript directly satisfies the verb's selectional requirements without intervening layers.[12] Similarly, ditransitive verbs such as give select two immediate complements (e.g., a theme NP and a goal PP or NP) as sisters to the head, ruling out configurations where selection skips structural positions.[15]
This principle bears a direct relation to the Projection Principle by enforcing the local realization of projected lexical properties at each stage of the syntactic derivation, from D-structure through transformations, thereby preserving subcategorization frames across levels.[16] Without such locality, the Projection Principle could permit derivations where theta-roles or category requirements are satisfied non-locally, undermining the lexicon-to-syntax mapping.[15]
Among its key implications, Locality of Selection prevents overgeneration by blocking ill-formed structures, such as a verb attempting to select an indirect object without its obligatory direct object counterpart—for instance, John gave the park is ungrammatical because give requires both a theme (e.g., NP) and a goal (e.g., DP) as local sisters, disallowing the PP the park to be selected in isolation.[12] This constraint thus maintains grammatical restrictiveness, ensuring that syntactic representations reflect the precise argument structure specified in lexical entries.[15]
Theoretical Applications
Role in Syntactic Representations
In Government and Binding (GB) theory, the Projection Principle ensures that the subcategorization and theta-marking properties of lexical items are represented at every syntactic level, including D-structure, S-structure, and Logical Form (LF), thereby maintaining representational consistency across the derivation. This principle, as formulated by Chomsky, stipulates that no transformation can eliminate required arguments or alter their lexical projections, with violations rendering a derivation ill-formed.[17][18]
At D-structure, the Projection Principle requires the complete projection of a head's arguments directly from its lexical entry, positioning them in base-generated sites that satisfy subcategorization frames and the theta-criterion. This level thus embodies the full lexical realization of complements and specifiers within the hierarchical structure provided by X-bar theory. No movements have occurred, so argument positions directly reflect the verb's or noun's requirements without mediation.[17][18]
At S-structure, following optional transformations like NP-movement or wh-movement, the Projection Principle preserves lexical properties through the co-indexation of traces in the original argument positions. These traces, governed by the Empty Category Principle, stand in for the moved elements, ensuring that theta-roles remain assigned and subcategorization is satisfied despite surface displacements. This preservation allows derivations to proceed to Phonetic Form only if all projected positions are accounted for.[18]
At LF, the Projection Principle enforces a terminal check on theta-satisfaction, scope ambiguities resolved by quantifier raising, and the absence of any unprojected or deleted arguments, blocking illicit structures that fail to uphold lexical integrity. It further interacts with Binding Theory by delimiting the structural positions available for anaphora and pronouns, as traces in projected sites influence c-command and antecedent accessibility in binding domains.[18]
Examples in Verb Phrases
The Projection Principle ensures that the subcategorization requirements of a verb, such as the need for a direct object in transitive constructions, are satisfied at every level of syntactic representation, including deep structure (D-structure), surface structure (S-structure), and logical form (LF).[16] A basic illustration appears in the transitive verb phrase where the verb eat projects its argument structure, requiring both a subject and an object to fulfill its lexical properties. For instance, the sentence John eats an apple is grammatical because the subject John receives the agent theta-role and the object an apple receives the theme theta-role, preserving the verb's transitive frame at all syntactic levels.[12] Omitting the object, as in the ungrammatical John eats, violates the Projection Principle by failing to project the required complement position.[12]
This preservation extends to cases involving movement operations, where traces left behind maintain the original argument structure. Consider the wh-question What did John eat?, derived from the underlying structure akin to John ate what. Here, what moves from the object position to the specifier of CP, leaving a trace (t) in the VP complement position. The trace ensures that the subcategorization frame of eat—requiring a theme argument—and its theta-role assignment remain intact at D-structure and S-structure.[12] Without the trace, the structure would incorrectly represent eat as intransitive, contravening the Projection Principle.[16]
A more complex case involves verbs like give, which subcategorizes for three arguments: an agent subject, a theme direct object, and a goal indirect object or prepositional phrase. The double-object construction John gave Mary a book satisfies this by projecting both Mary (goal) and a book (theme) within the VP, with theta-roles assigned accordingly.[12] In contrast, John gave a book is ungrammatical because it omits the goal argument, failing to project the full subcategorization frame required by give at all syntactic levels.[12] This enforcement highlights how the principle prohibits partial projections that would alter the verb's lexical properties.
To visualize this, consider a simplified tree diagram for the VP in John eats an apple, where the verb head projects the phrase structure according to X-bar theory, ensuring head-complement relations:
VP
├── [NP](/page/NP) (John, subject, external argument)
└── V'
├── [V](/page/V.) (eat)
└── [NP](/page/NP) (an apple, complement, internal argument)
VP
├── [NP](/page/NP) (John, subject, external argument)
└── V'
├── [V](/page/V.) (eat)
└── [NP](/page/NP) (an apple, complement, internal argument)
This structure reflects the Projection Principle by embedding the required complement under V', preserving the transitive frame from the lexicon.[12] In movement cases like What did John eat?, the object NP position hosts a trace (t) in the same V' complement slot, maintaining the projection.[16]
Developments and Criticisms
Evolution in Minimalist Program
In the Minimalist Program introduced by Chomsky in 1995, the Projection Principle is reinterpreted as an emergent property of the core syntactic operation Merge, which constructs hierarchical structures directly from lexical items and their associated features, thereby generalizing the projection of subcategorization and theta-role properties across the derivation without invoking distinct levels like D-structure.[19] This shift eliminates the need for the principle as an independent constraint, as Merge—operating recursively via external (combining new elements) and internal (removal and reattachment, akin to movement) applications—ensures that lexical requirements persist through structure-building, with heads determining the label and properties of the resulting syntactic object {γ, {α, β}}.[19]
Central to this evolution is bare phrase structure theory, which derives phrase structure relationally from the properties of merging elements rather than predefined X-bar templates, reducing projection effects to minimalist principles of economy such as last resort (operations apply only when necessary for convergence) and greed (elements move to satisfy their own unchecked features).[19] Lexical items enter the derivation with formal features (e.g., uninterpretable case or phi-features) that must be valued and deleted via Agree or movement, enforcing the consistent projection of argument structure as a byproduct of computational efficiency rather than a stipulated rule.[19] For example, a verb's selectional properties project via Merge to form a vP, where theta-roles are assigned to complements and specifiers within the structure.
Subsequent integration with phase theory, as developed in Chomsky (2001), further embeds projection within cyclic domains, where phases like vP and CP serve as propositional units in which lexical properties are realized and checked at edges before transfer to the interfaces (PF and LF).[20] In vP phases, theta-role assignment occurs through Merge, with external arguments merging in the specifier and internal arguments as sisters to the verb, while edge features on phase heads trigger movement to escape the phase, ensuring projection constraints are satisfied incrementally to minimize computational load.[20]
By 2025, the Projection Principle operates implicitly in feature-driven syntax, subsumed under the dynamics of Merge and Agree, where violations of subcategorization would crash the derivation at interfaces, rendering the original principle redundant as a discrete postulate in favor of a more streamlined, economy-based system.[21] The Extended Projection Principle, for instance, is reformulated as an interpretable EPP-feature on T, compelling specifier occupancy without separate projection machinery.[22]
Challenges and Alternatives
One significant challenge to the Projection Principle arises in non-configurational languages with free word order, such as Warlpiri, where strict subcategorization requirements at all syntactic levels lead to overgeneration of ungrammatical structures unless the principle is restricted to lexical structure alone, decoupling it from phrase structure projections.[23][24] In polysynthetic languages, noun incorporation poses difficulties, as incorporated nominals appear to violate the principle's insistence on projecting lexical argument structures into syntactic representations without intermediate levels, complicating the morphology-syntax interface.[25]
Empirical issues further highlight apparent violations, particularly with clitics, where placement in restructuring constructions seems to bypass subcategorization projections, and light verbs, which exhibit variable argument realization that challenges uniform projection across levels; these are often addressed through feature relativization, allowing context-sensitive adjustments to projection rules.[26][27]
Within the Minimalist Program, the Projection Principle faces critiques for redundancy with mechanisms like Attract and Move (now internal Merge), as selectional requirements are enforced at the numeration stage rather than across derivations, leading to proposals for its elimination in favor of phase impenetrability, which limits access to prior syntactic domains and simplifies locality constraints.[28]
Alternatives appear in other frameworks, such as Lexical Functional Grammar (LFG), where projection occurs via functional structures (f-structures) that map grammatical functions directly from lexical entries to a non-syntactic level, rejecting the Projection Principle's multi-level syntactic enforcement in favor of lexical dissociation from predicate-argument structure.[29] Similarly, in Head-Driven Phrase Structure Grammar (HPSG), schema inheritance handles projection through constraint-based type hierarchies, where head features like valence propagate via schemata (e.g., head-complement schema) without deriving structures from deep lexical projections, relying instead on feature unification for argument realization.[30]