Does the “it” always refer to the later clause?

It is commonly treated as dummy or expletive in this construction, rather than an ordinary entity reference; grammatical theories describe the relationship in different ways.

Is extraposition an error?

No. It is a normal grammatical construction.

Does the term apply only to “it” sentences?

No. Broader linguistic discussions also use it for other delayed constituents, including certain relative-clause patterns.

AI Glossary / Natural language and speech / Intermediate

Extraposition

Definition

Extraposition is a grammatical construction in which a clause or phrase appears later in a sentence than in a corresponding more basic arrangement. A common English pattern places a clause near the end and uses nonreferential “it” earlier, as in “It is helpful to label the cables.”

What it means

Compare “To label the cables is helpful” with “It is helpful to label the cables.” Both express a judgment about labeling the cables, but they organize the sentence differently. In the second version, “it” does not refer to a cable or another object mentioned earlier.
This matters for language-processing systems because a familiar pronoun form does not always function as a reference to an entity. A system that automatically links every “it” to a nearby noun can invent a relationship that the sentence does not contain.
The term extraposition also occurs in analyses of other delayed constituents, including relative clauses. This entry concentrates on English clause extraposition with “it,” the pattern especially relevant to recognizing nonreferential pronouns.

How it works

A common subject-clause pattern pairs an earlier “it” with a following finite or infinitival clause. For example, a statement about something being surprising can place the content clause after the main predicate.
The later clause supplies the substantive content, while “it” occupies a grammatical position without naming a separate participant. Analyses and annotation systems differ in how they describe the relationship between these parts.
Universal Dependencies, for example, marks this “it” with its expletive relation and treats the associated clause as a core clausal argument in the relevant construction.
Identifying the pattern requires sentence-level evidence. The mere presence of “it” or a following “that” does not establish extraposition, and mechanically moving clauses around does not always produce an acceptable equivalent.

Three practical examples

A coreference system avoids linking “it” in “It is useful to keep a backup” to an earlier device name.
A grammar lesson compares a clause-first sentence with an extraposed version to explain information arrangement.
A parsing evaluation checks whether a model recognizes a nonreferential pronoun and the later clause that carries the relevant content.

Think of it this way

Think of a short signpost that prepares the reader for a longer explanation appearing later. The signpost helps organize the route but is not another destination. The analogy is limited because grammatical “it” follows language-specific construction rules, rather than freely standing in for any material a writer wishes to postpone.

Worked example

Consider the invented pair “That the sensor restarted surprised the engineer” and “It surprised the engineer that the sensor restarted.”
In both sentences, the surprising content is that the sensor restarted. In the second, the initial “it” does not identify the sensor. A coreference system should therefore not add a sensor - it entity link merely because “sensor” is a nearby singular noun.
Now compare “The sensor restarted. It surprised the engineer.” In an appropriate context, this second sentence can use “it” referentially, for example to refer to the preceding event or situation. The surface words overlap with the extraposition example, but the construction is different.
A reviewer can check the first example by identifying the following content clause and asking what proposition is surprising. The second example requires discourse interpretation rather than the same extraposition analysis.
This exercise demonstrates why pronoun classification and reference resolution should use both grammatical structure and context. A simple rule that all occurrences of “it surprised” are nonreferential would also be too broad.

Illustrative case study

In a fictional project, Willow Text Tools analyzes equipment manuals. Its reference display links “it” in “It is important to disconnect the charger” to the charger, suggesting that the charger itself is important in the intended sense.
An editor marks the construction as nonreferential and checks the clause describing the recommended action. The team adds similarly structured examples to its evaluation set alongside genuinely referential uses of “it.”
The imagined case illustrates a specific analysis error. It does not imply that an extraposition detector alone can determine whether a technical instruction is correct or that all sentences beginning with “it is” have the same structure.

Common misconceptions and mistakes

Extraposition is not ordinary reference to a previously mentioned object. In the pattern discussed here, dummy “it” does not introduce a separate discourse entity.
Not every nonreferential “it” is an example of this construction. Weather expressions and other patterns need their own analysis.
Extraposition is also not the same as an it-cleft, such as “It was Lina who submitted the form,” which has a different focus structure.
Finally, “moving a clause” is a helpful way to compare sentence patterns, not a claim that every grammatical theory represents the construction through a literal movement operation.

Business use

Recognizing extraposition can improve the interpretation of reference chains, grammatical analyses, and explanations generated by writing tools. It helps prevent systems from treating structural pronouns as additional objects or participants.
For practical evaluation, include both nonreferential and referential examples with similar wording. Otherwise a model may appear successful by memorizing a few phrases while failing on context-sensitive cases.
Editing applications should also respect the writer’s intended emphasis. An extraposed version may be natural and clear, but it is not automatically superior to every clause-first alternative. The surrounding discourse and the purpose of the sentence matter.

Check your understanding

In “It is surprising that the package arrived early,” should “it” automatically be linked to “the package” as the same entity?

Answer: No. This is a clause-extraposition pattern: the surprising content is the package’s early arrival, and the initial “it” is not an ordinary reference to the package.

Related terms

Coreference resolution
Constituency
Discourse deixis
Expletive pronoun
pleonastic pronoun
clausal subject
dependency parsing
it-cleft
information structure

Learn next

Learn the distinction between referential and nonreferential pronouns, then compare extraposition with it-clefts and discourse deixis. Inspect a few annotated examples under one grammatical framework before comparing labels across frameworks, because similar sentences can receive different formal descriptions.

Discourse deixis

Sources

Jurafsky and Martin, Speech and Language Processing, nonreferring expressions: https://web.stanford.edu/~jurafsky/slp3/24.pdf. Universal Dependencies, official expletive-relation documentation: https://universaldependencies.org/u/dep/expl.html. Keller, “Towards an Account of Extraposition in HPSG” (1995), original research abstract: https://arxiv.org/abs/cmp-lg/9502027. Independent review consulted the official UD expletive examples and core-clausal-argument rule, and Keller’s primary abstract for broader extraposition scope.

Previous: External validity Next: Extremal Ensemble Learning