The Devil’s Advocate Question: Asking What Liars Never Rehearse

There are two ways to be caught out by a question. The first is that you do not know the answer. The second, more useful for an interviewer, is that you have not prepared one.

Liars prepare. It is the single most reliable thing about them: given time before an interview, a person intending to deceive will rehearse an account of the expected questions. And rehearsal produces fluency, which is why prepared lies often sound better than the truth — the liar delivers a smooth, ordered narrative while the honest person hesitates, corrects themselves and remembers things out of sequence.

Which means asking the expected questions measures preparation rather than honesty.

The unanticipated questions paradigm asks something else. If a pair of suspects claim to have had lunch together in a particular restaurant, the anticipated questions are about what they ate and what they discussed, and both will have covered those. The unanticipated ones are spatial and temporal: where was your table in relation to the door, who sat facing which way, what was to your left, draw the layout. Genuine episodic memory contains spatial and perceptual information as a matter of course; a rehearsed narrative does not, because nobody rehearses the floor plan.

In studies using this design, pairs of truth-tellers show high agreement on the unanticipated questions and pairs of liars show much lower agreement, while both show high agreement on the anticipated ones. The diagnostic is the gap between the two, within the same interview.

The devil's advocate technique, tested by Leal, Vrij, Mann and Fisher in 2010, applies the same logic to opinions rather than events, which is important because opinions have no verifiable facts to check.

The method is two questions. First, the opinion-eliciting question: what are your reasons for holding this view. Then the devil's advocate question: what would someone who disagrees with you say.

For someone genuinely holding the opinion, the first answer comes readily and at length, because a real attitude sits inside a rich associative network of reasons and experiences. The second answer is shorter and comes more slowly, because arguing against your own position requires effortful construction.

For someone asserting an opinion they do not hold, the pattern reverses. Their supporting arguments are constructed and thin; the opposing arguments — which are their actual views — come easily and in more detail.

So the signal is the ratio, and it is measurable: length, latency, and richness of the pro answer relative to the con answer.

The applications are wherever a stated belief is the thing at issue: visa and asylum interviews, security vetting, screening for extremist intent, insurance investigation, and — considerably more mundanely — assessing whether a candidate or a counterpart actually holds the commitment they are professing.

The general principle behind both techniques is the one worth keeping. Do not ask the question they came prepared to answer.