Before the accusatory interrogation begins, the Reid method calls for a screening interview. The Behavior Analysis Interview is a structured set of questions — what do you think should happen to the person who did this, would you be willing to take a polygraph, do you think the person had a good reason — asked in a non-accusatory manner, while the investigator observes both the content of the answers and the suspect's behavior.
The claimed indicators are specific. Truthful suspects are said to sit upright, lean forward, maintain eye contact, answer directly and promptly, and to be helpful and forthright. Deceptive suspects are said to be evasive, to shift and cross their limbs, to grooming and self-touch, to avoid eye contact, to be less helpful, and to give qualified or delayed answers.
Aldert Vrij, Samantha Mann and Ronald Fisher tested this in 2006. They set up scenarios in which participants either committed a mock crime or did not, then interviewed them with the BAI protocol and coded their behavior against the manual's predictions.
The results ran against the manual on the key indicators. Liars showed fewer movements than truth-tellers, not more. Liars were more likely to be forthright rather than evasive. On several criteria the direction of the effect was the opposite of what the technique predicts.
The explanation is that the manual's model of a liar is a person suffering from guilt and anxiety, and that is not what a liar is. A liar under interview is performing a cognitive task under strategic pressure. They are constructing, monitoring the listener, and controlling their own behavior — and the control is the point, because they know the stereotype too. Deliberate behavioral control produces stillness, not fidgeting. Chapter 108's cognitive load approach is built on the correct model, and it works.
Meanwhile the innocent suspect has the emotion the manual is looking for. Paul Ekman named this the Othello error, after the scene in which Desdemona's terror at not being believed is read by Othello as evidence of the guilt he already suspects. Fear of disbelief and fear of exposure produce the same autonomic signature, and the interviewer cannot distinguish them, because there is nothing there to distinguish.
The consequence is not merely that the screen is inaccurate. It is what an inaccurate screen does to the process that follows.
The BAI's function is to decide who gets interrogated. Once the investigator has concluded, on the basis of a near-chance instrument, that this person is guilty, everything after that is evaluated by someone who is no longer neutral. Confirmation bias does the rest: hesitations become evasions, inconsistencies become lies, a request for a lawyer becomes consciousness of guilt, and every subsequent statement is read as corroboration of a conclusion reached before any evidence arrived.
Kassin has demonstrated this experimentally — investigators who believe a suspect is guilty ask more guilt-presumptive questions, use more pressure, and elicit behavior from innocent suspects that other observers then rate as suspicious.
The error at the screening stage does not stay at the screening stage. It selects who enters the machine in chapter 91.
Cross-ref: Part 2 nonverbal leakage limits