How to Separate Facts from Opinions in Spoken Language
Learn to classify spoken claims as checkable facts, judgments, predictions, interpretations, or attributed claims—without confusing classification with truth.
Extract the claim, identify whose claim it is, ask whether outside evidence could settle it, then label its job—without pretending that a “fact claim” has already been proved true.
A speaker says, “The café closes at six, and honestly that’s ridiculous.” One breath. Two information jobs. The closing time is presented as checkable; ridiculous is a judgment. Spoken English mixes these constantly, so stop trying to label whole sentences. Label the claim pieces.
Use this four-step listener method:
- Claim: What exactly is being said?
- Source: Is this the speaker’s claim, or are they reporting someone else?
- Testability: Could outside evidence, records, observation, or measurement settle this claim in principle?
- Label: Fact claim, opinion/judgment, prediction, interpretation—or an attributed claim with another label inside it.
The rule worth protecting from the beginning is simple: classify first, verify later. You are adding a label, not awarding a truth stamp.
First, change “fact” into “fact claim”
British Council’s “Fact or opinion?” activity gives the useful basic distinction: a fact is something that can be proven true or false, while an opinion expresses someone’s ideas or feelings.
For listening, however, one wording change makes the distinction safer: call the first category a fact claim.
Why? Because you often cannot verify the world while somebody is still talking.
“The museum closes at five.”
That is presented as a checkable claim. The opening time could be checked against the museum’s schedule. But the speaker could still be mistaken. Classification tells you what kind of statement you heard; verification tells you whether it is accurate.
This prevents a surprisingly common mental shortcut:
Sounds factual → therefore true.
Nope. A confidently delivered wrong train time is still a fact-type claim. Confidence is not a laboratory coat.
Opinion is about evaluation, preference, or judgment—not just the words “I think”
Cambridge Grammar describes opinions as beliefs or judgments and shows explicit phrases such as in my opinion that can identify whose judgment is being expressed.
Those phrases are useful clues. They are not required.
“In my opinion, the hotel is overpriced.”
Easy: the speaker marks the judgment explicitly.
“The hotel is overpriced.”
Still an evaluation. The speaker removed the opinion badge; the sentence did not magically become a measurement.
Listen especially for evaluative language:
- great, awful, beautiful, boring, unfair
- too expensive, too early, much better
- worth it, a bad idea, the best option
- should, ought to, would be better when they express recommendations
Context matters. Your job is not to memorize an “opinion-word list.” Ask what the claim is doing: describing something that can be externally settled, or evaluating it?
Prediction and interpretation deserve their own labels
The simple fact/opinion split becomes much more useful when you add two middle categories.
Prediction: what the speaker expects to happen
“It’ll probably be busy after lunch.”
The speaker is making a claim about a future state. Probably tells you something about confidence.
Cambridge Grammar’s explanation of modality notes that modal expressions can show a speaker’s view of certainty, possibility, likelihood, and doubt about possible facts. Words such as may, might, must, will, probably can therefore help you hear stance.
They do not tell you whether the prediction will actually come true.
Interpretation: what the speaker thinks the evidence means
“He must be tired—he’s been yawning all morning.”
He’s been yawning reports an observation in this illustrative example. He must be tired is the speaker’s interpretation of that evidence.
“Sales fell last month, so the new price may be putting people off.”
The first part can be treated as a fact claim if the sales data are checkable. The second part interprets what that pattern might mean. May helps signal that the causal explanation is not being presented as certain.
This distinction matters because an interpretation can be thoughtful and evidence-based without becoming an established fact.
Attribution answers a different question: whose claim is this?
“According to Nina, the class moved to Room 4.”
Cambridge Grammar explains that according to means “as reported by” or “as stated by” and points to a source other than the current speaker.
That gives you one important piece of information: Nina owns the reported claim.
It does not automatically prove that the class really moved.
Use a two-layer label:
- Source tag: attributed to Nina
- Underlying claim: fact claim—the room can in principle be checked
Now compare:
“According to Nina, Room 4 is much nicer.”
- Source tag: attributed to Nina
- Underlying claim: opinion/judgment
Same attribution cue. Different claim type. Source tag first; truth stamp much later—if you even need one.
One breath does not mean one label
Real speech happily combines jobs inside the same sentence.
“The train leaves at eight, and that’s ridiculously early.”
- “The train leaves at eight”: fact claim
- “that’s ridiculously early”: judgment
“The café has forty seats, so it’s probably too small for our group.”
- “The café has forty seats”: fact claim
- “it’s probably too small for our group”: interpretation/judgment about suitability, expressed with uncertainty
“According to Maya, the deadline moved to Friday, which is a much better plan.”
- Source: Maya
- “the deadline moved to Friday”: attributed fact claim
- “a much better plan”: judgment
If you force the whole sentence into one box, useful information disappears. Split it where the speaker changes jobs.
Do not let cue words make the decision for you
Language cues are excellent hints. They are lousy automatic classifiers.
“I think” does not always mean personal taste
“I think the meeting starts at nine.”
The speaker may simply be uncertain about a checkable schedule detail. The underlying proposition—what time the meeting starts—is still externally testable.
“I think nine is too early for this meeting.”
Now the speaker is evaluating the time. Same opening phrase; different information job.
Strong certainty does not erase interpretation
“He must be annoyed.”
Must can signal a strong inference from evidence. It does not make the inferred mental state directly observed.
No opinion marker does not create a fact
“That restaurant is terrible.”
There is no I think, in my view, or probably. It is still a judgment.
Your four-step method is stronger than cue hunting because it asks what the claim does, who owns it, and whether evidence could settle it.
Claim Sorter: one label or two?
For each example, decide before opening the answer:
- What is the core claim—or are there two?
- Who is the source?
- Could outside evidence settle it?
- What label belongs on each claim?
Case A
“The train leaves at eight, and that’s ridiculously early.”
Show the labels
Claim 1: train leaves at eight → speaker’s fact claim; schedule evidence could settle it. Claim 2: eight is ridiculously early → judgment. The evaluative adjective is a clue, but the real reason for the label is that “ridiculously early” evaluates the time rather than merely reporting it.
Case B
“I think the meeting starts at nine.”
Show the labels
Source: current speaker. Underlying claim: fact claim about a checkable meeting time. Stance: “I think” signals that the speaker may be uncertain. Do not automatically relabel every sentence after “I think” as personal opinion.
Case C
“It’ll probably be busy after lunch.”
Show the labels
Label: prediction. Probably is a useful clue that the speaker is expressing likelihood rather than reporting a known present fact.
Case D
“He must be tired—he’s been yawning all morning.”
Show the labels
Observation/fact-type claim: he has been yawning. Interpretation: he is tired. Must signals a strong inference here; it does not make tiredness directly observed.
Case E
“According to Nina, the class moved to Room 4.”
Show the labels
Source: Nina. Underlying label: attributed fact claim because the room assignment could be checked. The attribution tells you who reported it, not whether Nina is correct.
Case F
“The café has forty seats, so it’s probably too small for our group.”
Show the labels
Claim 1: forty seats → fact claim. Claim 2: too small for our group → interpretation/judgment about suitability, softened by probably. You need two labels.
What if you label it wrong at first?
Do not mentally restart the whole conversation. Re-label the smallest claim that changed.
“The restaurant is expensive…”
You tentatively hear evaluation. Then the speaker continues:
“…the main courses start at forty euros.”
Now you have more structure. Expensive is the speaker’s evaluation; main courses start at forty euros is presented as checkable evidence for it.
Your repair is simply: Judgment + supporting fact claim.
No drama. Listening is allowed to update itself.
A 60-second real-audio drill
Choose a short, low-stakes English conversation, review, meeting clip, interview, or video.
- Listen without subtitles first.
- Pause after one important claim.
- Say three things aloud: source — testable? — label.
- If the sentence contains two jobs, split them.
- Replay once. Listen specifically for the stance or attribution cue you may have missed.
- Only then check subtitles or a transcript if available.
Do not stop to Google whether the claim is correct. That would train a different skill. For this drill, you are organizing spoken information.
Where FunFluen can help
The method does not require a special tool. But on supported video pages, FunFluen can make the listen-first loop less fiddly: replay a short line, navigate sentence by sentence, adjust speed when needed, or use a listening-first practice pass before checking subtitles when the relevant features and subtitle tracks are available.
Keep the order honest: hear the claim → label it yourself → replay → check the wording. FunFluen does not fact-check the claim and does not automatically decide whether it is opinion, prediction, or interpretation.
The rule worth keeping
When spoken English mixes information, attitude, inference, and attribution, do not ask one giant question—“Is this true?”—too early.
Ask four smaller ones:
What is the claim? Whose claim is it? Could evidence settle it? What job is the claim doing?
That is enough to keep the conversation organised in your head. You are classifying claims, not sentencing them.
For more listening practice built around real video, explore FunFluen’s Media-Based Language Learning hub.