Read the full question rather than its summary
A television graphic may shorten the wording to fit the screen. Locate the original questionnaire where possible. If it asks respondents to agree with two linked propositions, one answer may not reveal their view on each proposition independently. Full Fact’s polling guide describes this problem as compound questioning.
Work through an illustrative example
Imagine a question asking whether a new streaming app is more affordable and should replace broadcast television. A respondent may think it is cheaper while wanting both options to remain. A single agree-or-disagree answer does not cleanly measure those two views separately. These are invented statements used to explain wording, not findings about any actual app or audience.
Check the surrounding questions
Full Fact also notes that question order and wording can influence answers. Read the options offered and any preceding information before interpreting the percentage. Do not invent an unasked response such as preferred another candidate, wanted a different service or supported a wider policy from a narrower question about satisfaction.
Describe what the data can support
A careful summary keeps close to the actual proposition and acknowledges when the wording combines ideas. That is different from declaring the entire survey fraudulent. If the headline claims a separate belief the questionnaire did not measure, explain that mismatch with the original text. The useful outcome is a narrower, accurate interpretation rather than a stronger conclusion selected because it fits a preferred story.
Questions, answered
Can one answer to a compound question settle both views independently?
Not necessarily. Read how the propositions were linked.
Should the original question be checked when a headline paraphrases it?
Yes. Wording and answer options determine what was measured.