The study examines the assessment of the non-verbal behavior of interrogators to expose the sincerity of statements in a police interrogation and critically rethinks the stereotypes of «body language». The relevance is due to the risk of investigative errors due to reliance on unreliable behavioral markers, reinforced by the confirmation effect and the «Othello error». The aim was to systematically analyze the diagnostic value of specific non-verbal indicators and verify the accuracy of police officers in distinguishing true and false testimony on controlled material. Materials and methods included a two-sample experiment: 60 active employees acted as expert appraisers, 120 students staged sincere and insincere versions in a standardized scenario, interrogations were video-recorded from two angles. The police evaluated 20 fragments on a confidence scale and noted indicators (looking away, adapter gestures, hesitation pauses, illustrative gestures, tone change). Independent assistants carried out objective coding. Statistical processing included t-criteria for comparing groups, signal detection analysis (sensitivity, specificity), Pearson correlations between confidence, length of service and accuracy, as well as ANOVA by length of service groups. The results showed no significant differences in looking away, which disavows one of the main stereotypes. Statistically significant but moderate effects were revealed for adapter gestures and hesitation pauses, and illustrative gestures more often accompanied sincere statements; the change in pitch approached significance. The practical suitability of indicators is limited due to overlapping distributions. The overall accuracy of the police was 53.85% with a sensitivity of 48.17% and a specificity of 59.53%, which reflects a «shift towards the truth». There was no correlation between confidence and accuracy, as well as the dependence of accuracy on length of service, which indicates the inefficiency of intuitive learning by practice. The discussion interprets the low accuracy as a consequence of the low signal-to-noise ratio in non-verbal lie detection and cognitive distortions of the observer. Non-verbal behavior cannot serve as a sufficient basis for conclusions about guilt; its role is the generation of verifiable hypotheses. A transition to information-oriented, verifiable interrogation tactics, standardized coding and feedback, training to reduce bias, as well as prioritization of objective evidence and verification protocols are recommended.