Must a bound pronoun refer to one named person?

No; its interpretation can vary with a quantified expression.

Is every occurrence of their bound?

No; interpretation depends on the sentence and context.

AI Glossary / Natural language and speech / Advanced

Bound pronoun

Definition

A bound pronoun is a pronoun interpreted as a variable whose value depends on a binding expression, often a quantified noun phrase. In “Every reviewer checked their own report,” the bound reading of their varies with the reviewer being considered. It does not pick out one fixed person for all reviewers. Here, bound concerns semantic dependence, not physical attachment to another word.

What it means

Ordinary reference can connect a pronoun to an individual already established in a conversation. For example, in “Lina arrived. She sat down,” she can refer to Lina. Both expressions concern the same person.
A quantified expression such as every reviewer works differently. It ranges over the relevant reviewers instead of naming just one. On the bound reading, the pronoun follows that variation. When considering Ada, their means Ada's; when considering Bo, it means Bo's. Treating the pronoun as a permanent link to one person would lose this dependency and could change the sentence's meaning.

How it works

An interpretation identifies the quantifier, its scope and the pronoun's dependency within that scope. A logical representation can use the same variable for the quantified participant and the pronoun's contribution. For the reviewer sentence, one representation is: for every relevant reviewer x, x checked the report belonging to x.
This differs from saying that every reviewer checked a report belonging to a fixed individual c. In that representation, the reviewer variable changes while c remains constant. Both descriptions can be grammatically expressed, but they describe different conditions.
Not every pronoun near a quantifier must receive a bound reading. Sentence structure and context constrain interpretation, and some sentences permit more than one reading. A useful analysis therefore specifies the reading being discussed rather than using word proximity alone as proof of binding. Detailed binding theories address additional syntactic restrictions beyond this introductory distinction.

Three practical examples

A semantic parser processes “Each participant uploaded their own answer.” On the bound reading, the uploaded answer is associated with the participant currently quantified over, rather than one shared owner.
A question-answering system reads a passage saying that every author revised their manuscript. It must preserve the person-to-manuscript dependency if the intended reading is distributive and bound.
A language-understanding evaluation contrasts a varying possessive interpretation with a fixed named owner. The contrast checks whether a model preserves the logical relationship, not merely whether it finds a nearby name for the pronoun.

Think of it this way

Think of a form filled out once for each person on a list. A field marked “this person's report” changes as you move to the next row. A field containing “Ada's report” remains the same in every row. The first illustrates a bound dependency; the second illustrates a fixed reference.

Worked example

Consider three reviewers: Ada, Bo and Cy. Each has a distinct report, called report A, report B and report C respectively. Assume these are the only relevant reviewers and that the checking events listed below are complete for the example.
In situation one, Ada checks report A, Bo checks report B and Cy checks report C. Under the bound reading of “Every reviewer checked their own report,” test one condition per reviewer: Ada checked Ada's report; Bo checked Bo's report; Cy checked Cy's report. All three conditions are satisfied, so the statement is true in this situation.
Now compare the different statement “Every reviewer checked Ada's report.” Ada satisfies that condition, but Bo and Cy do not. They checked their own distinct reports, and no other checking events are included. The fixed-owner statement is false.
In situation two, all three reviewers check only report A. The fixed-owner statement is now true because Ada, Bo and Cy each checked Ada's report. But the bound own-report statement is false: Bo has not checked report B, and Cy has not checked report C.
The varying dependency can hold while the fixed-reference statement fails, and the fixed-reference statement can hold while the varying dependency fails. Replacing their with one person's name is therefore not a generally meaning-preserving operation.
The word own clarifies the intended dependency. The two statements are not presented as interchangeable readings of every pronoun-containing sentence.

Illustrative case study

In a fictional document-understanding project, a system processes the worked reviewer statement after Ada has been mentioned prominently in the preceding text. Its reference component links their to Ada and produces the summary “All reviewers checked Ada's report.”
A test using situation one reveals the error. The original bound statement is satisfied, but the generated summary is false. The system preserved the word report and the reviewer group while losing the ownership dependency.
The team adds paired tests using both situations. It represents the owner as dependent on the quantified reviewer in the intended reading and checks the resulting relations. It also keeps ambiguous inputs available for separate analysis instead of forcing every possessive pronoun into either a named-person link or a bound-variable interpretation without evidence.

Common misconceptions and mistakes

A bound pronoun is not simply any pronoun that appears after a noun. Nor is binding identical to selecting the closest preceding name. A quantified expression can supply a variable dependency without denoting one fixed individual.
The term should also be distinguished from a pronoun realized as a clitic or bound morpheme. Those descriptions concern how a form attaches to another word; the meaning discussed here concerns variable interpretation. Singular they can participate in the worked dependency, and its form alone does not determine whether a particular occurrence is bound or referential.

Business use

The distinction matters when systems interpret statements about each person, every item or all members of a group. Test whether extraction and summarization preserve relationships that vary across those participants. A fluent paraphrase can still change who owns an object or performs an action if it replaces a dependency with a fixed reference.

Check your understanding

Why is the own-report statement true in situation one?

Answer: Each reviewer checks the report associated with that reviewer.

Why is the fixed-owner statement false there?

Answer: Bo and Cy do not check Ada's report.

What information did the fictional summary lose?

Answer: The dependency between each reviewer and that reviewer's report.

Related terms

Pronoun
Antecedent
Anaphora
Semantic parsing
quantifier
bound variable
scope
coreference
truth conditions

Learn next

Compare referential and bound-variable readings, then examine how quantifier scope and sentence structure constrain interpretation.

Sources

Jurafsky and Martin, Speech and Language Processing, chapter 24, section 24.1.1: pronouns in quantified contexts and the distinction between bound interpretation and fixed reference. https://web.stanford.edu/~jurafsky/slp3/24.pdf The reviewer situations and fictional summarization case are original teaching illustrations.

Previous: Bottom-up design Next: Boundary leakage