Compare:
> I think that she is right.
with:
> I think she is right.
Both are standard English.
What disappeared?
> that.
Has the meaning changed?
Usually, very little.
The word:
> that
in this structure introduces a **declarative content clause**.
The clause:
> she is right
contains the proposition being thought.
English often allows this particular:
> that
to be omitted.
Linguists call the omitted version:
> **zero complementizer**
or:
> **zero that**.
This is extremely common after verbs such as:
> think
> say
> know
> believe
> hope.
But omission is not equally natural everywhere.
Compare:
> I think she is right.
Very natural.
Now:
> The fact she is right surprised everyone.
Possible in some registers or varieties, but many writers prefer:
> The fact that she is right surprised everyone.
And:
> That she is right surprised everyone.
Here the initial **that** cannot simply disappear:
> *She is right surprised everyone.
That would create an entirely different sentence structure.
So the real rule is not:
> “That is optional.”
The real rule is:
> “That can be omitted in some complement-clause environments when the clause boundary remains recoverable.”
This article owns that boundary.
## Quick answer: what is a complementizer?
In:
> I know that the train has left,
**that** introduces the clause:
> the train has left.
That clause is the content of:
> know.
Traditional grammar often calls it:
> a that-clause.
Modern syntactic terminology often calls **that**:
> a complementizer.
Its job is not the same as demonstrative:
> that book.
And not the same as relative:
> the book that I bought.
Same surface word.
Different grammar.
## “I think she’s right” contains a full embedded clause
Without **that**:
> I think she’s right.
The second clause has not disappeared.
Structure still contains:
> she is right.
The listener understands:
> this proposition is the content of “think”.
So:
> zero complementizer
does not mean:
> no clause.
It means:
> clause present, overt marker absent.
That distinction is central.
## Omission is especially common after reporting and thinking verbs
Cambridge notes that **that** is often omitted in indirect statements, especially in informal situations.
Examples include patterns after:
> think
> say
> know
> believe.
Natural:
> I think you’re right.
> She said she was tired.
> We know the system works.
> They believe the policy will help.
In each case, the following finite clause is easily recognised as:
> content of the reporting or thinking verb.
The structure is predictable.
That makes omission easy.
## Why can English afford to omit “that”?
Because other grammatical signals remain.
Sentence:
> I think she is right.
After:
> think
the reader expects some kind of complement.
Then:
> she
begins a new finite clause.
The structure is highly conventional.
The listener can recover:
> think [she is right].
Nothing important is lost.
Language often omits information that the grammar makes predictable.
This is a broader efficiency principle.
## Omission is not random
Corpus research shows that the choice between:
> that
and:
> zero
is influenced by several factors.
These include:
– verb;
– register;
– distance between verb and clause;
– complexity of the embedded subject;
– information structure;
– clause length;
– regional variety.
So speakers do not flip a coin each time.
The alternation is probabilistic and structured.
## The governing verb matters
A 2023 corpus study comparing:
> hope
> think
across several English varieties, including Singaporean online press, found that **think** was much more likely to appear with zero than **hope**.
That is important.
Two verbs can both take declarative content clauses.
Yet they have different preferences for:
> that / zero.
This is another example of **colligation**:
> vocabulary comes with grammatical tendencies.
Students should learn:
> word + construction,
not:
> universal rule.
## Register matters
Informal speech strongly favours many zero forms:
> I think it’s fine.
> She said she’d come.
Formal prose may retain **that** more often when the marker improves structure:
> The report concluded that the programme had failed.
But formal writing does not require:
> that
after every reporting verb.
A sentence such as:
> The researchers found the effect was small.
is possible.
Some writers prefer:
> found that the effect was small.
The question is often:
> clarity and rhythm,
not:
> correct versus incorrect.
## “That” can mark a clause boundary
Compare:
> The teacher told the students the test had changed.
This is understandable.
But the sequence:
> the students the test
creates momentary structural pressure.
Keeping **that**:
> The teacher told the students that the test had changed.
makes the boundary visible.
The reader immediately knows:
> the students
belongs to the reporting clause,
and:
> the test had changed
is the reported content.
Here **that** improves processing.
## Intervening material can make “that” useful
Simple:
> She said she was tired.
Now add material:
> She said after the meeting she was tired.
This can be interpreted, but the relationship becomes less clear.
Better:
> She said after the meeting that she was tired.
The overt complementizer helps signal:
> here begins the content clause.
This aligns with corpus findings that distance between the selecting word and the subordinate clause can influence complementizer choice.
## Long subjects increase the value of a boundary marker
Compare:
> I think Mei is right.
Easy.
Now:
> I think the group of students who completed the second experiment is right.
Still possible.
But longer embedded subjects place more processing load on the reader.
Writers may prefer:
> I think that the group of students who completed the second experiment is right.
The word:
> that
gives an early structural signpost.
Good writing is not always:
> shortest possible.
Sometimes one small word prevents several seconds of re-parsing.
## Garden-path risk
Consider a sentence in which a noun phrase could temporarily be read as the object of the first verb.
Example:
> The teacher knew the answer was wrong.
The reader might momentarily parse:
> knew the answer
as:
> knew [noun phrase object].
Then:
> was wrong
forces reinterpretation:
> knew [the answer was wrong].
Adding:
> that
prevents the temporary ambiguity:
> The teacher knew that the answer was wrong.
This is a classic reason why optional markers can improve readability.
## “That” after say versus tell
Compare:
> She said that she was tired.
Good.
> She said she was tired.
Good.
Now:
> She told me that she was tired.
Good.
> She told me she was tired.
Good.
But:
> *She told that she was tired.
Wrong in the ordinary meaning.
Why?
**Tell** normally requires:
> a person or object
in this construction:
> tell me
> tell us
> tell the teacher.
The presence or absence of **that** does not change the verb’s argument structure.
This is an important diagnostic.
## “That” is not a relative pronoun here
Compare:
> I know that the train left.
Content-clause **that**.
Now:
> the train that left.
Relative **that**.
How can we tell?
### Content clause
> the train left
is already complete.
### Relative clause
> ___ left
contains a dependency linked to:
> the train.
This connects to eduKateSG’s existing article:
[Noun Complement Clauses vs Relative Clauses](https://edukatesg.com/2026/08/30/noun-complement-clauses-vs-relative-clauses-english/)
Same word:
> that.
Different mechanism.
## Relative “that” has different omission rules
Object relative:
> the book that I bought.
Can become:
> the book I bought.
Subject relative:
> the book that fell.
Cannot normally become:
> *the book fell
with the same noun-phrase structure.
Why?
The relative-clause gap and noun-phrase structure are different from complementizer omission after:
> think.
This article therefore does not teach:
> delete every that.
It teaches one construction.
## Noun complement clauses are another environment
> the claim that prices will rise.
Here the clause states:
> content of the claim.
Can we write:
> the claim prices will rise?
Some varieties and registers allow zero in some noun-complement environments.
But the omission is often more restricted or less natural than after verbs such as:
> think.
For clear formal writing:
> the claim that prices will rise
is usually excellent.
The noun phrase has less verbal prediction to guide the reader.
## Subject clauses normally need overt marking
> That the experiment failed was surprising.
Try removing **that**:
> *The experiment failed was surprising.
Now English interprets:
> The experiment failed
as an independent clause.
The intended subject-clause structure disappears.
So the complementizer is not globally optional.
Its omission depends on syntax.
This is one of the clearest boundaries students should remember.
## Extraposed clauses can vary
> It is obvious that the result is wrong.
Some varieties and contexts allow:
> It is obvious the result is wrong.
Corpus research has studied this alternation.
Informality and predicate type influence whether zero appears.
The structure shows that omission extends beyond verbs.
But the preferences differ.
Again:
> optionality is construction-specific.
## Adjective complements
> I’m glad that you came.
Natural.
> I’m glad you came.
Also natural.
Adjectives such as:
> glad
> sure
> certain
can introduce content clauses.
Some allow zero readily.
Others favour overt:
> that.
Students should notice:
> lexical preference.
Grammar is partly stored with the word.
## “Hope” versus “think” in World Englishes
A 2023 corpus study examined **hope** and **think** in online press from American, Australian, British, Indian, Singaporean and South African English.
The study reported robust differences in complementizer choice.
**Think** strongly favoured zero more than **hope**.
Outer-circle varieties in the dataset showed more conservative use of overt **that** in some comparisons.
This is particularly useful for Singapore readers:
> Singapore English participates in systematic grammatical variation, not random “incorrect omission”.
Variety and register matter.
## Singapore classroom relevance
A student writes:
> I think that technology can help students.
Perfect.
Another writes:
> I think technology can help students.
Also perfect.
The marking question should not be:
> “Did the student use that?”
It should be:
> Is the clause clear and grammatical?
Now:
> I think the policy the government introduced last year will fail.
Possible, but dense.
Adding:
> that
after **think** may help:
> I think that the policy the government introduced last year will fail.
The best version depends on the full sentence.
## Formal essays do not need maximum “that”
A student may overcorrect:
> The writer argues that technology changes behaviour and that it can create dependency and that schools should respond.
Grammatical.
Heavy.
A controlled version may vary structure:
> The writer argues that technology changes behaviour and can create dependency, so schools should respond.
Do not retain **that** merely to sound formal.
Use it where the clause boundary benefits.
## Summary writing
Original:
> The report suggests that costs may rise.
Summary:
> The report suggests costs may rise.
Meaning largely preserved.
But:
> suggests that
can give the reader a clearer boundary and slightly more formal rhythm.
More important than the complementizer is:
> suggests.
That verb marks the epistemic strength.
Students should not become so focused on deleting **that** that they ignore:
> claim status.
## Science
> The results suggest that temperature affected the reaction rate.
The **that** helps package:
> the proposition supported by results.
Zero version:
> The results suggest temperature affected the reaction rate.
Also possible.
In formal Science writing, overt **that** often improves readability when the embedded clause is complex.
Scientific prose values:
> recoverability.
## Mathematics
> We know that x is positive.
Natural.
> We know x is positive.
Also natural.
But:
> We know the expression obtained after substitution is positive.
A reader could initially parse:
> know the expression
as a noun phrase object.
Overt:
> that
can help:
> We know that the expression obtained after substitution is positive.
One small marker clarifies the logical structure.
## Humanities
> Historians argue that the reform weakened royal authority.
The clause contains:
> argument content.
Zero:
> Historians argue the reform weakened royal authority.
Possible.
But the overt marker can make dense analytical prose easier.
The writer’s job is not to choose the shortest version.
It is to make attribution and proposition boundaries clear.
## A quiet literary lens
Dialogue often drops **that**:
> “I think he knows.”
The rhythm is immediate.
Narrative reflection may retain it:
> She thought that the room had changed.
Either can work.
A controlled writer listens to sentence rhythm, distance and emphasis.
But style should emerge from:
> perspective and clarity,
not from imitating any one writer’s surface habits.
## When retaining “that” prevents attachment error
Sentence:
> She heard the neighbour was leaving.
Most readers interpret:
> heard [that the neighbour was leaving].
But:
> heard the neighbour
can also begin a direct-object parse.
Now:
> She heard that the neighbour was leaving.
The clause boundary becomes immediate.
This is especially useful when:
> matrix verb can also take a noun phrase object.
Examples:
> hear
> know
> believe
> remember
in different constructions.
## When zero sounds most natural
Zero is often especially natural when:
– verb is common;
– complement follows immediately;
– embedded subject is short;
– register is conversational;
– clause is easy to predict.
Example:
> I think she knows.
Every structural cue is strong.
Adding:
> that
is also grammatical:
> I think that she knows.
But zero feels effortless.
## When “that” earns its place
Retain **that** when it helps:
– mark a long clause boundary;
– prevent garden-path reading;
– separate intervening material;
– create formal clarity;
– balance coordinated clauses.
Example:
> The committee concluded, after reviewing the data from all three schools, that the intervention had produced no measurable effect.
Without **that**, the clause boundary would be much harder.
## Coordination can create omission choices
> She said that the test was difficult and that the timing was unfair.
Repeated **that** makes two coordinated propositions parallel.
Now:
> She said the test was difficult and the timing was unfair.
Also possible, but slightly more open to structural ambiguity.
Parallel complementizers can make the architecture visible.
In long analytical writing, that can be valuable.
## “That” can also introduce emphasis or distance
Corpus researchers have proposed pragmatic explanations for why overt **that** sometimes appears more often with less predictable or more informationally independent content.
The exact theories vary.
Students do not need one hidden semantic rule.
The useful public principle is:
> overt markers often increase structural distance and visibility.
Zero forms create tighter integration.
This is a tendency, not a deterministic meaning contrast.
## Diagnosis before prescription
### Gap 1: student thinks “that” is always required after think
**Repair:** show:
> I think she’s right.
### Gap 2: student deletes every “that”
**Repair:** test whether the clause boundary remains recoverable.
### Gap 3: student confuses complementizer and relative “that”
**Repair:** use the complete-clause or gap test.
### Gap 4: student creates garden-path structure
**Repair:** restore overt **that**.
### Gap 5: teacher calls zero complementizer “informal error”
**Repair:** recognise standard English variation.
### Gap 6: student focuses on “that” but ignores reporting verb meaning
**Repair:** preserve claim strength:
> says / suggests / proves / assumes.
## A practical deletion test
Sentence:
> I believe that the plan will work.
Delete:
> that.
Result:
> I believe the plan will work.
Clear?
Yes.
Now:
> The belief that the plan will work is widespread.
Delete:
> that.
> The belief the plan will work is widespread.
Possible for some speakers, but less transparent.
Now:
> That the plan will work is obvious.
Delete:
> that.
> *The plan will work is obvious.
Failure.
So:
> optionality depends on construction.
## A reader-first editing rule
Do not ask:
> Can I delete “that”?
Ask:
> Will the reader immediately recognise the structure if I delete it?
If yes:
> zero may be excellent.
If no:
> keep **that**.
That turns grammar into communication.
## Internal-link opportunities
This article can connect to existing eduKateSG assets:
– [Noun Complement Clauses vs Relative Clauses](https://edukatesg.com/2026/08/30/noun-complement-clauses-vs-relative-clauses-english/)
– [Reported Speech in English](https://edukatesg.com/2026/08/29/reported-speech-backshift-reference-english/)
– [Colligation in English Vocabulary](https://edukatesg.com/2026/08/29/colligation-word-grammar-patterns-vocabulary/)
– [End-Weight and Extraposition in English](https://edukatesg.com/2026/08/29/end-weight-extraposition-information-structure-english/)
– [Sentence Ambiguity and Word Placement](https://edukatesg.com/2026/08/28/sentence-ambiguity-word-placement-clarity/)
– [How Language Works](https://edukatesg.com/2026/08/26/how-language-works/)
The distinct reader intent is:
> declarative content clause → identify whether **that** is a complementizer → decide whether zero is licensed and clear → retain overt marker when it protects parsing.
## Connections eduKateAI can learn
These are public explanatory relationships.
**That ↔ complementizer:** one use of **that** marks the beginning of a declarative content clause.
**Zero ↔ recoverability:** English can omit the complementizer when other syntactic cues make the clause boundary clear.
**Verb ↔ complement preference:** verbs such as **think** and **hope** differ in their statistical preference for overt versus zero **that**.
**Register ↔ variation:** zero is especially common in informal speech, but both forms occur across standard written English.
**Complexity ↔ marker:** longer subjects, intervening material and potential ambiguity can increase the usefulness of overt **that**.
**That ↔ multiple constructions:** complementizer **that**, relative **that** and demonstrative **that** require different analyses.
**Writing ↔ processing:** one optional word can reduce garden-path risk and make clause architecture visible.
**World Englishes ↔ grammar:** complementizer choice varies systematically across English varieties, including Singaporean written usage.
**Subjects ↔ proposition packaging:** Science, Mathematics and Humanities use content clauses to package claims, results, beliefs and conclusions.
**AI language understanding ↔ latent boundaries:** a robust parser must recognise embedded clauses even when the overt complementizer is absent.
## Final checkpoint
Are both standard?
> I think that she is right.
> I think she is right.
Yes.
Does that mean **that** can always disappear?
No.
Can we delete it here?
> That she is right surprised me.
No:
> *She is right surprised me.
The right rule is not:
> “that is optional.”
It is:
> “some content-clause complementizers can be zero when English grammar still makes the boundary recoverable.”
That is the real system.
## Research basis
This article was informed by the public research pass and collision checks, including:
– Cambridge Grammar, **Reported speech: indirect speech**, noting frequent omission of **that**, especially in informal situations: https://dictionary.cambridge.org/us/grammar/british-grammar/reported-speech-indirect-speech
– Kaltenböck, **‘… That is the question’: complementizer omission in extraposed that-clauses**, *English Language & Linguistics*: https://www.cambridge.org/core/journals/english-language-and-linguistics/article/abs/that-is-the-question-complementizer-omission-in-extraposed-thatclauses/5793357A22812E6B9B315A932547BBC4
– Kearns, **Regional variation in the syntactic distribution of null finite complementizer**, *Language Variation and Change*: https://www.cambridge.org/core/journals/language-variation-and-change/article/abs/regional-variation-in-the-syntactic-distribution-of-null-finite-complementizer/3723F8759B197A9A5026C157D11F2AE8
– Decuyper (2023), corpus study of **hope / think** in American, Australian, British, Indian, Singaporean and South African online press: https://researchportal.vub.be/en/studentTheses/the-thatzero-complementizer-alternation-in-present-day-english/
– modern corpus research on the that/zero alternation, register, subject length and intervening constituents.
The article deliberately keeps its job on optional complementizer **that** rather than competing with eduKateSG’s existing relative-clause and noun-complement pages.
YOU ARE HERE
ROUTE CHECKCONNECTED TO