Loaded question

“Have you stopped...?”

The back gate was found open one morning and the dog got out. At dinner, Priya turns to her brother: “So, have you stopped leaving the gate open at night?”

“Yes” means he used to; “no” means he still does. Either answer admits that he has left it open, which nobody has shown. This is the clear-cut case: a yes-or-no question whose two answers share the same hidden premise.

A “where” question

A parent to a teenager: “Where did you put my car keys after you borrowed them?”

This isn’t a yes-or-no question, but it works the same way. Any answer that names a place (“on the counter”) admits borrowing the keys, and “nowhere” can sound like “I lost them.” Patrick Hurley and Lori Watson give an example of the same shape and point out that either kind of answer leads to a conclusion the questioner never supported. If the teenager didn’t borrow the keys, the honest reply is to reject the question, not to answer it.

A “why” question in a meeting

The office moved to a new filing system last month. At the team meeting, a manager who opposed the change asks: “Why has the new system slowed everyone down so much?”

This one sounds like a respectable request for analysis, and in a meeting people tend to start answering it: “Maybe the folder names are confusing...” But whether the new system has slowed anyone down is the very point under dispute, and nobody has measured it. A “why” question presupposes that the thing to be explained actually happened, so offering explanations concedes it. The fair question comes first: “Has the new system slowed us down?”

A loaded question isn’t an argument on its face, but Hurley and Watson describe it as carrying an implicit one, which appears once an answer is given:

• Question Q presupposes P.
• You gave a direct answer to Q.
• Therefore you’ve admitted P.

Hans Hansen notes that Aristotle placed the fault in the conclusion drawn from the answer rather than in the question alone, and that some modern authors deny it is a fallacy at all, since asking a question isn’t making an argument. The implicit argument above is what makes it an error in reasoning: the answer is used as evidence for something the answer never addressed.

• The yes-or-no trap: both answers share a presupposition. “Are you still wasting money on that gym membership?”
• The “where”, “when” or “why” question: any answer that fills the blank grants the premise. “When did you decide to stop caring about the garden?”
• Two questions asked as one: in the traditional sense of complex question, which Hurley and Watson follow, two questions are presented as a single question and a single answer is applied to both. “Did you finish the report and send it to the client?” A “no” doesn’t say which part failed. This isn’t always loaded; it becomes a problem when the answer to one part is used as an answer to the other.
• Loaded wording: the presupposition sits in a description rather than in the grammar. Bradley Dowden’s example is a reporter asking whether a policy maker will “continue your policy of wasting taxpayer’s money”, which presumes the disputed claim that the money is being wasted.
• Polls and surveys: Dowden notes that loaded questions are common in “push polls”, which are designed to spread a claim rather than measure opinion. A survey item that restricts the answers (only “good”, “very good” or “excellent”) is a different move, the loaded menu described under False dilemma; a loaded question presupposes a premise.

Names. Logic textbooks usually call this the fallacy of complex question or many questions (Latin plurium interrogationum, “of many questions”). “Many questions” goes back to Aristotle, for whom the fallacy was two questions asked as if they were one. Douglas Walton separates the two properties the traditional name runs together: a question’s complexity (it contains more than one proposition) and its loadedness (in his words, it has a presupposition “that is unwelcome or objectionable” to the person answering). A question can be complex without being loaded, and loadedness is the part that does the damage, which is why this page uses the name “loaded question”.

Is it a kind of begging the question? Sources disagree. Dowden’s Internet Encyclopedia of Philosophy entry says the complex question “is a form of Begging the Question”, since the question assumes a controversial claim without argument. Hurley and Watson list the two side by side as separate fallacies of presumption, and Hansen treats many questions as a fallacy of its own. The family resemblance is real either way: in both, a point that needs support is treated as already settled. See Begging the question.

Every question takes something for granted, and most of the time that’s harmless.

• The presupposition is already established or agreed. “When did you quit smoking?” is a fair question to someone who has told you they quit.
• The person asked can reject the presupposition freely. Walton points out that asking a loaded question need not be fallacious in itself. The trouble starts when the answerer is pressed for a direct answer and that answer is then treated as an admission.
• The presupposition is stated openly as a supposition. “If someone did leave the gate open, who was last out there?” makes the assumption visible and leaves room to deny it.
• The question suggests an answer but assumes nothing disputed. Hurley and Watson distinguish a leading question, one in which the answer is suggested by the question (an issue in courtroom questioning), from a complex question, and say leading questions don’t commit a logical fallacy.

The test: does the question take for granted something the other person hasn’t accepted, and would they look evasive for saying so?

The presupposition is on the record

Last week, Dev told the team he’d stopped using the old invoice template. Today his manager asks: “Since you switched templates, have the clients stopped sending invoices back with questions?”

Out of context, “since you switched” plants a premise. But Dev asserted it himself, so the question assumes nothing he hasn’t already accepted, and any answer about the clients concedes nothing new. This is the already established or agreed condition above.

A pointed question from a doctor

A patient says her knee has been aching. The doctor asks: “Does it hurt more when you go down stairs than when you go up?”

The question suggests a specific pattern, and a patient who hasn’t noticed one might feel pushed toward an answer. But its only presupposition, that the knee hurts, is what the patient reported, and “I haven’t noticed a difference” is a perfectly good answer that costs her nothing. This is the suggests an answer but assumes nothing disputed condition: a leading question, not a loaded one.

Questions don’t look like claims, so what they take for granted gets less scrutiny than an open assertion would. The presupposition arrives in the background while attention goes to the thing being asked.

The format adds pressure. Hansen’s standard example is a context where a yes-or-no answer must be given, and in conversation, refusing to answer the question as asked tends to look like dodging it. Hurley and Watson note that people who don’t spot the trap may answer innocently and be caught by a conclusion supported by no evidence, or be led into supplying the evidence themselves.

• Split the question. Hurley and Watson’s advice: break it into its component questions and answer each separately. “Did I leave the gate open? No. So there’s nothing to stop.”
• Name the presupposition. “That assumes the new system slowed us down. Has anyone checked?”
• When you’re asking, ask about the presupposition first. It gets you the same information without the trap.
• Don’t read a refusal as an answer. Someone who rejects a loaded question hasn’t admitted anything; treating the refusal as a sign of guilt is close to the move described under Kafka trap. Walton adds that sometimes the better course is not to answer a loaded question directly at all.