Syntax

Principles and representations governing structured combinations of linguistic elements. Experiments must distinguish building or interpreting a structure from holding its elements in memory, detecting errors and selecting responses.

A grammatical constituent is a unit that participates as a component of a larger grammatical construction. Constituency describes hierarchical grouping, so adjacency alone does not establish that an arbitrary sequence of words forms one unit. Different levels of an analysis can contain constituents of different sizes.^[ref]

In The red kite rose, a conventional analysis groups the red kite as a noun phrase and assigns it the subject relation. Replacing the whole group with it gives It rose, providing one structural diagnostic. The example separates the number of written words from the number of units occupying that grammatical position.

A useful analysis combines several diagnostics and states the construction under consideration. Context can license omissions or exceptional patterns, so one substitution does not prove every proposed grouping. Grammatical constituents provide the units of a theory of syntax; judgments and controlled comparisons supply evidence for choosing among theories.

In a traditional subject-predicate analysis, a predicate is the portion of a clause that says something about its subject. It can include a verb together with other material, rather than being identical to a single verb form.^[ref]

For The sample changed color, the subject is the sample and the predicate is changed color. In The sample is red, is red expresses the predication, although the central property is contributed by the adjective red. The examples show why “predicate” and “action word” are not interchangeable classifications.

An explicit model can distinguish the grammatical expression from its interpretation. A formal semantics might represent the property red by a set and ask whether the sample belongs to it, while the syntax describes how is red combines with the subject. The linguistic unit and the mathematical representation are related analytical choices.

A syntactic head is the element treated as central to a construction under a specified structural analysis. It can determine grammatical properties or which other elements may occur. Headedness therefore concerns relationships within a construction, beyond which word seems most informative in the conversation.^[ref]

In a conventional noun-headed analysis of the small boxes, boxes heads the noun phrase. The head's plural grammatical number helps explain the grammatical agreement in The small boxes are ready. Changing small to heavy changes the description without changing that number requirement.

In a dependency representation, a head also provides the attachment point for syntactic dependencies. A reproducible analysis states its conventions: different grammatical frameworks can choose different heads for some constructions. The success of a representation depends on the patterns it explains and predicts, not solely on whether its links form a mechanically valid rooted tree.

Passive voice is a grammatical construction that changes how an event's participants are expressed, typically placing an affected participant in the subject relation while omitting the actor or expressing it differently. It belongs to grammatical voice, not simply to a choice of slow or indirect wording.^[ref]

Compare The mechanic repaired the pump with The pump was repaired by the mechanic. Both can describe the same repair, but the pump changes from object to subject; the mechanic is expressed in a separate phrase. Omitting that phrase gives The pump was repaired without specifying who performed the repair.

This is useful for separating thematic roles from grammatical relationships. A representation that treats every subject as the actor will misdescribe the passive example. The Universal Dependencies label nsubj:pass preserves subject status while recording the passive construction, making the analysis inspectable in a text corpus.

A clause is a grammatical unit organized around a predicate and its associated subject, which may be expressed or understood. A written sentence can contain one such unit or combine several. Identifying clauses is a structural analysis rather than a matter of counting punctuation marks.^[ref]

In the original example The bird landed because the wind dropped, the bird and landed form one subject-predicate pairing, while the wind and dropped form another. The latter is introduced as a subordinate clause giving a reason; the first supplies the main assertion. Both occur within a single written sentence.

Separating the units supports explicit accounts of syntax and interpretation. The labels should still follow the language and construction being analyzed: an omitted subject is not automatically evidence that no clause is present. A proposed analysis must state which elements are expressed and which relationships it treats as implicit.

A main clause is a clause that functions as an independent grammatical unit in the construction being analyzed, contrasted with an embedded subordinate clause. It can be accompanied by subordinate material within a larger sentence. Its status concerns the syntax of that construction, not which information seems most surprising.^[ref]

In the original example The pump stopped although the tank was full, the pump stopped supplies the main clause while although the tank was full contributes subordinate material. Moving that material to the beginning gives Although the tank was full, the pump stopped; the main clause is now later in the sentence. Position alone therefore does not identify it.

The example also separates a clause's own subject and predicate from its relationship to surrounding material. Both parts have internal grammatical structure. Annotating those structures and the embedding relation supports reproducible linguistic annotation, while the speaker's beliefs about the two statements remain a separate interpretive question.

A subordinate clause is a clause embedded within a larger grammatical construction, where it supplies part of the structure or interpretation. It contrasts with a main clause in that construction. Subordination describes a relationship among units, rather than merely identifying whichever words occur after a comma.^[ref]

In the original example The sensor stopped because the battery failed, because the battery failed supplies a reason for the main assertion. In I know that the sensor stopped, the embedded material instead supplies what is known. Both contain their own subject-predicate organization, while contributing differently to the larger expression.

A structural analysis can vary the embedded content while preserving its role in the surrounding syntax. Comparing that pattern across examples helps distinguish an embedded grammatical constituent from an unrelated second assertion. The analysis must also keep the grammatical embedding separate from whether the speaker actually believes the embedded content.

A noun phrase is a grammatical unit conventionally analyzed as headed by a noun, potentially together with dependents that modify or specify it. A single noun or pronoun can also form a minimal noun phrase under this analysis.^[ref]

In The small bird landed, the words the small bird form a noun phrase containing the head bird, the adjective small and a determiner. The whole unit occupies the subject position. Replacing it with it gives It landed, illustrating that a multiword expression can occupy a position also available to one word.

Such substitution is useful evidence for a grammatical constituent, but one test alone does not settle every structural question. A clear syntax states how the proposed head and dependents combine and tests the same account across further constructions. Counting spaces therefore measures a different property from counting noun phrases.

A syntactic dependency is a typed relation between an expression and its syntactic head under an explicit grammatical representation. In basic Universal Dependencies, the words form a rooted tree: one word attaches to a notional ROOT, and each other word has one basic head. An enhanced representation can instead form a more general graph.^[ref]

For the original example The small birds fly, number the words one through four. In head-to-dependent notation, the links are 4→3 for the subject, 3→1 for the determiner, and 3→2 for the adjective; 0→4 identifies the sentence head. The determiner link spans a nonadjacent word, showing why dependency is not merely adjacency.

The direction of an arrow and the meaning of each relation must be stated. The project also notes that some dependency labels encode relationships without asserting a simple head-subordinate linguistic analysis. A mechanically valid tree therefore remains a representation to justify, rather than a proof of its own grammatical interpretation.

An object is a grammatical relation assigned to certain complements of a verb. In a typical active construction it identifies a core participant other than the subject, often an entity affected by the event. This is a grammatical classification rather than a requirement that the referent be a physical object.^[ref]

In Mira opened the gate, the gate is an object. In The gate was opened by Mira, the same participant is instead the subject of a passive voice construction. The event can remain closely related while the grammatical relationships change.

An explicit annotation scheme also needs boundaries among complements. Universal Dependencies distinguishes obj from iobj and from other dependent relations; a traditional label “object” does not automatically determine the appropriate code. This makes syntactic dependency criteria part of the analysis needed to compare annotated text corpus resources.

A subject is a grammatical relation identified through language-specific structural tests, such as word order and grammatical agreement. It often corresponds to the actor in a simple active clause, but that semantic association is not its definition. A noun phrase can be a subject without describing an intentional action.^[ref]

Compare A gust broke the branch and The branch was broken by a gust. The first has a gust as its subject; in the passive voice version, the branch has that relation. The participants' involvement in the event has not simply exchanged places. Separating grammatical relations from thematic roles explains the contrast.

In Universal Dependencies, a nominal passive subject can receive nsubj:pass, explicitly preserving its subject status. Such a representation supports a testable structural account and prevents a procedure from assigning “subject” solely to whichever expression appears to name a doer.

Word order concerns the linear arrangement of linguistic elements. A basic clause-order description commonly records the relative positions of the subject, verb and object in an ordinary declarative clause. Such a description summarizes one construction type; it does not specify every ordering permitted by a language.^[ref]

In the original example Mira follows Noor, the English subject-verb-object arrangement identifies who follows whom. Noor follows Mira uses the same words with the participants reversed. A toy subject-object-verb grammar could instead express the first relation as Mira Noor follows. Here ordering and the participants' thematic roles are separate analytical dimensions. Comparing ordinary clauses with questions or embedded clauses helps determine whether a proposed pattern generalizes. Comparative syntax investigates those relationships across languages.

Comparative syntax investigates structural similarities and differences across languages or varieties. It asks which proposed grammatical relationships generalize, which vary, and whether apparently different constructions realize a shared organization. Comparable meanings, contexts and data-collection methods are needed before a difference can be assigned to grammar.^[ref]

For an original toy example, suppose one grammar expresses “A acts on B” with order A V B, while another uses A B V. Reversing the participants changes the intended meaning, whereas changing the position of V follows the grammatical template. This separates word order from the participants' thematic roles. Actual comparison examines more than one translated sentence: questions, embedded clauses and other environments can distinguish a superficial ordering similarity from a general structural account.

A description of structural demands in a sentence, such as a noncanonical dependency. Its operational definition should be explicit; longer strings, unfamiliar words and greater working-memory load do not define syntax by themselves.

An impairment in grammatical structure in production or comprehension. It is not equivalent to reduced speech rate, dysarthria or all forms of aphasia; task design must separate grammatical dependencies from lexical and memory demands.

An embedded clause serving as a complement, such as “that the toy is in the box” in “she says that the toy is in the box”. The full report can be true even when the embedded proposition is false, making such constructions relevant to distinguishing a representation from reality.^[ref]