Baseline First: Why Body Language Means Nothing Without Context

The most-cited meta-analysis on veracity judgment pools around two hundred studies and finds average accuracy near 54% — effectively chance on a two-alternative task.¹ The more uncomfortable detail is that trained professionals — police officers, interviewers, auditors — do not reliably outperform students. What training increases with some consistency is confidence, not precision.²

The explanation is not inattention. It is that we are looking at the wrong thing. A second meta-analysis catalogued well over a hundred behavioural cues associated with lying and found effect sizes that were, in the great majority, small: no single cue works as a reliable indicator.³

So the problem is not that a dictionary of gestures exists and you failed to memorise it. The problem is that the dictionary does not exist.

Anyone wanting the full version of this argument should go to Aldert Vrij's Detecting Lies and Deceit, which is the academic reference most popular books on the subject draw from without saying so, and to Timothy Levine's Duped, which explains the accuracy figures rather than merely reporting them. Levine's position is that we are poor lie detectors because we default to believing people, and that this default is adaptive — a society of accurate sceptics would function worse, not better.⁵

What survives when the dictionary falls

What survives is comparison — and comparison requires a reference. The structural error in the popular way of "reading people" is comparing someone's behaviour against an imagined population average. Crossed arms indicate closure relative to what? To what most people do? Most people cross their arms when the room is cold, when the chair has no armrest, and when they are holding a coat.

The only comparison that carries information is against the same person, minutes earlier. That is the baseline: the individual pattern of speech, rhythm, posture and response latency under low pressure. Without it there is no deviation; there is only a gesture.

And here is the precision most content on this subject refuses: a deviation is not evidence of lying. A deviation is evidence that something changed — cognitive load, discomfort, perceived stakes, fatigue, an unpleasant memory. The legitimate inference is "this topic costs this person more than the previous one." Anything beyond that is you filling the gap with what you already wanted to believe.

The four-minute protocol

This is not an interrogation technique. It is the minimum required for observation to be informative rather than decorative.

1. Four deliberate minutes of low pressure. The traffic, the building, how the week went. The goal is not rapport — it is recording the normal. Speech rate, how much the person gestures, whether they look at you or away while thinking, how long they take between question and answer.
2. Track three variables, not twenty. Response latency, change of pace, resting posture. Attempting to observe everything guarantees you observe nothing — and the attention left over is exactly what you need to listen to the content.
3. Ask questions in ascending order of cost. Easiest to most expensive. That way a deviation, when it appears, has a "before" to be compared with.
4. Treat the deviation as a question, not a conclusion. The correct use of a deviation is to return to the topic with different phrasing, not to mentally record "lied". Increasing cognitive demand — asking for detail, for the sequence in reverse, for specificity — discriminates better than watching gestures.⁴

The error the method introduces

Every instrument has a false positive, and anyone teaching a method owes you theirs.

People who start observing baselines begin seeing deviations everywhere, because deviations genuinely are everywhere — people change pace constantly, for banal reasons. The predictable result is generalised suspicion: you start treating honest counterparts as evasive and, worse, they notice. A posture of distrust produces precisely the discomfort behaviours you will then read as confirmation. It is a closed loop and it feeds itself.

Calibrate asymmetrically: a baseline is far better at showing you where to spend more conversation than at deciding who not to trust.

Used to open topics, it is a good tool. Used to close judgments, it is a machine for expensive error.

Why this matters outside an interrogation room

In commercial practice the question is almost never deception. It is something else: the person across the table has an objection they will not state — a budget that does not exist, a decision-maker above them, an incumbent contract they would rather not mention. Nobody is lying. There is simply a relevant piece of information going unsaid, and it is usually the one that decides the deal.

A baseline does not reveal that information. It indicates where it is — which topic the cost went up on. That indication alone is the difference between a proposal made in the dark and a question asked in the right place.