Case grammar
Case grammar is a theory of linguistic analysis that represents the semantic relationships between a predicate and the participants associated with it. Developed within early generative semantics, the theory treats a clause as an abstract proposition organized around a predicate and a set of semantic cases. These cases identify the roles played by entities in the event, state, or process denoted by the predicate.
The theory was formulated principally by Charles J. Fillmore during the 1960s. Its foundational presentation, “The Case for Case,” appeared in 1968 as a response to limitations in purely configurational accounts of deep structure. Case grammar influenced later work on thematic relations, semantic roles, argument structure, and frame semantics. It also contributed to computational methods that represent who did what to whom without requiring those relationships to coincide with surface word order.
Theoretical foundations
Case grammar distinguishes semantic cases from the morphological case systems found in languages such as Latin, Russian, and Finnish. Morphological case is expressed through inflection or associated grammatical marking, whereas a semantic case represents the conceptual relationship between a participant and a predicate. The two systems interact, but they do not correspond one to one.
In the sentence “The navigator opened the hatch with a lever,” the navigator functions as the intentional initiator of the event, the hatch is the entity affected by the event, and the lever is the means through which the action is performed. An active English clause expresses these relationships through word order and a prepositional phrase. A language with a different grammatical organization can encode the same semantic structure through inflection, particles, incorporation, or contextual omission.
Fillmore represented the basic clause through the schematic relation:
[ S \rightarrow M + P ]
Here, (S) denotes the sentence, (M) denotes the modality component, and (P) denotes the proposition. The proposition contains a verb together with one or more case relationships:
[ P \rightarrow V + C_1 + C_2 + \cdots + C_n ]
The modality component includes such information as tense, mood, negation, and aspectual qualification. The proposition describes the underlying event or state independently of the particular syntactic arrangement used to express it.
This separation allowed active and passive constructions to receive related semantic descriptions. “The mechanic opened the door” and “The door was opened by the mechanic” differ in grammatical organization, but both contain an opening event involving an initiator and an affected entity. Case grammar therefore treats voice as part of the mapping between semantic structure and surface realization rather than as a change in the underlying participant relationships.
Deep cases
The original theory used the term “deep case” for a semantically defined relationship that remains identifiable across syntactic alternations. The inventory changed among Fillmore’s publications because the categories were analytical hypotheses rather than forms directly copied from traditional grammatical case.
Agentive case
The agentive case identifies an animate participant understood as intentionally responsible for an action. In “The engineer lowered the bridge,” the engineer occupies the agentive role because the participant initiates and controls the lowering event.
Agency is not equivalent to grammatical subjecthood. In a passive clause, the agent can occur in an optional phrase while the affected entity becomes the subject. A sentence can also have a grammatical subject without containing an intentional agent, as in “The wind lowered the temperature.”
Instrumental case
The instrumental case identifies the inanimate force or means that produces an event. In “The key opened the cabinet,” the key is interpreted as the immediate means through which the change occurs, even though the sentence does not express the person who used it.
Instrumental participants occupy different syntactic positions according to construction and language. English permits an instrument to appear as the subject when discourse suppresses the human initiator, while other contexts express the same participant through a phrase headed by “with.”
Dative and experiencer functions
Fillmore’s early dative case covered an animate participant affected by a state or event. It included relationships later divided among recipient, beneficiary, and experiencer. In “Mira heard the alarm,” Mira undergoes a perceptual state rather than intentionally producing the sound.
The broad dative category exposed a recurring problem in semantic-role classification. Participants that share an animacy requirement can differ substantially in causal involvement and syntactic behavior. Later theories consequently separated the experiencer of a psychological state from the recipient of a transferred object.
Objective case
The objective case represents the semantically neutral entity whose role is determined primarily by the meaning of the verb. In “The ice melted,” the ice undergoes a change of state. In “The sunlight melted the ice,” the same entity appears as the affected participant even though the clause now contains an external cause.
This category anticipated later concepts such as the patient and theme roles. Its breadth also illustrates why no single inventory of semantic cases became universal. A physically altered object, a displaced object, and the content of a mental state share certain grammatical patterns but differ in their event structure.
Locative case
The locative case identifies the spatial orientation of a proposition. In “The documents remained in the archive,” the archive supplies the place associated with the state described by the predicate.
Later analyses divided spatial relationships according to whether they denote a stable location, a source, a destination, or a path. Case grammar initially captured their common spatial basis while allowing individual verbs to impose more specific restrictions.
Case frames and lexical structure
A case frame specifies the set of semantic cases associated with a predicate. The frame for “open,” for example, permits an affected object and can also contain an intentional initiator or an instrument. The frame for “give” requires a transferred entity and relates that entity to a source and a recipient.
Case frames do not amount to unrestricted lists of participants. They encode constraints imposed by the lexical meaning of the predicate. A perception verb selects an experiencer and a perceived phenomenon, whereas a motion verb organizes a moving entity relative to a trajectory. The acceptability of a clause depends partly on whether its expressed participants can be interpreted within the predicate’s frame.
Fillmore proposed a case hierarchy to account for subject selection. When an agentive participant was present, it received priority as the unmarked subject. In the absence of an agent, an instrumental participant could occupy that position, while an objective participant became available when neither of the preceding roles was expressed. This hierarchy accounted for contrasts among “The technician opened the lock,” “The key opened the lock,” and “The lock opened.”
The hierarchy described grammatical selection rather than a ranking of conceptual importance. Different constructions can reorganize the same case frame to reflect voice, perspective, or information structure. The resulting variation connected case grammar with research on diathesis alternation and the syntax–semantics interface.
Japanese data and the Watanabe analysis
The relation between deep cases and surface particles became a major issue when the theory was applied beyond English. A 1969 analysis by You Watanabe examined Japanese clauses containing the particles ga, o, and ni. The analysis demonstrated that these particles could not be assigned invariant deep-case meanings because their interpretation changed with the predicate, voice construction, and argument omitted from the clause.
Watanabe compared active clauses with their causative and passive counterparts. The comparison showed that a participant’s semantic relationship to the predicate could remain stable while its particle marking changed. Conversely, the same particle could mark participants with different semantic functions. The findings supported a distinction between the semantic case frame and the language-specific rules that realize it through Japanese particles.
The analysis also addressed argument omission. Japanese permits contextually recoverable participants to remain unexpressed, but an omitted argument can still occupy a position in the predicate’s semantic frame. This separation between semantic presence and phonological realization later became relevant to analyses of pro-drop languages and zero anaphora.
Relation to grammatical case
Traditional descriptions use labels such as nominative and accusative for morphosyntactic categories. These categories regulate agreement, government, and the formal marking of noun phrases. They therefore belong to a different level of analysis from deep cases.
A nominative noun phrase can represent an agent in an active construction, an affected entity in a passive construction, or an experiencer with a psychological predicate. An accusative noun phrase likewise receives its semantic interpretation from the predicate and construction rather than from accusative marking alone. Case grammar explains these patterns by assigning semantic roles before the rules that determine surface form.
The distinction does not make morphological case semantically empty. Case marking can constrain interpretation and can preserve participant relationships when word order varies. The central claim is narrower: morphological categories cannot serve as universal names for all underlying semantic relationships.
Later development
John M. Anderson developed an autonomous theory of case in which case relations formed part of the organization of grammatical structure rather than merely supplementing transformational syntax. His 1971 treatment connected case with dependency relations and with the internal structure of predicates.
Fillmore subsequently moved from fixed inventories of deep cases toward frame semantics. In that approach, understanding a word involves access to a structured conceptual frame whose elements receive interpretations specific to the situation represented. A commercial transaction, for example, contains participants associated with goods, payment, acquisition, and transfer, but their labels derive from the transaction frame rather than from a small universal list.
This transition retained the central insight of case grammar while changing its granularity. General roles such as agent and instrument capture recurring patterns across many predicates. Frame-specific roles provide distinctions required by particular conceptual domains. Modern linguistic description uses both levels, depending on whether the analysis concerns broad grammatical generalizations or detailed lexical meaning.
Influence on linguistic theory
Case grammar helped establish semantic roles as a distinct component of grammatical analysis. Later versions of generative grammar recast related constraints through the theta criterion, which associates arguments with thematic roles assigned by predicates. Although theta-role theory differs in formal assumptions, it preserves the view that argument structure cannot be reduced to linear position or morphological marking.
Role and reference grammar organizes participant relationships through generalized macroroles linked to more specific positions in a logical structure. Lexical functional grammar distinguishes grammatical functions from argument structure and semantic representation. Construction grammar assigns meaning to pairings of form and function, allowing a construction to contribute participant relationships not supplied exclusively by the verb.
The terminology inherited from case grammar varies across these frameworks. A participant described as objective in an early case frame can be analyzed elsewhere as a patient, theme, undergoer, or frame element. These differences reflect distinct criteria for grouping participant relationships rather than simple changes in nomenclature.
Computational applications
Case-based representations entered natural language processing through systems that mapped sentences onto predicate–argument structures. Such representations supported text understanding by normalizing active and passive clauses that described the same event. They also provided an intermediate level between syntactic parsing and domain-specific inference.
Walter A. Cook systematized case frames in linguistic description and language pedagogy, emphasizing their use in classifying verbal patterns. His matrix representations linked predicate classes with the semantic configurations licensed by those predicates, thereby making differences in argument structure explicit without treating surface word order as the sole organizing principle.
Contemporary semantic role labeling continues this line of analysis through statistical and neural models. A system identifies a predicate and assigns labels to spans that denote its participants or circumstances. Resources such as FrameNet use frame-specific relations derived from frame semantics, while PropBank employs numbered arguments whose interpretation depends on the predicate.
These computational schemes do not reproduce Fillmore’s original case inventory. They retain its structural premise that clauses encode relations between predicates and semantically differentiated participants. The transition from manually formulated case frames to corpus-based annotation changed the method of representation while preserving the underlying analytical problem.