Limited effects of action on ensemble processing.

Children begin to tell lies around age two and a half (Evans and Lee ; Leduc et al. ). Despite the limited ability to control over semantic leakage, their lies are hard to detect (Evans et al. ). Accurately detecting children's deception is critical for both moral and legal contexts (Domagalski et al. ; O'Connor et al. ). Recent work shows using machine learning methods can detect children's deception through their facial expressions (Bruer et al. ; Zanette et al. ). However, these analyses did not account for the dynamic nature of facial expressions over time. The present study seeks to identify the most effective statistical approach for detecting deception by incorporating new time‐dependent statistics, with implications for improving analytical methods and informing moral education and legal practices. Since the 1970s, psychologists have explored whether verbal and non‐verbal cues, such as linguistic cues (e.g., pitch of voice), non‐verbal cues (e.g., facial expressions), and content‐based factors (e.g., quantity of details), can be used to detect deception (Feldman et al., ; Vrij et al., , ). Though using these cues to detect deception in adults has some limitations (Vrij et al., 2010, ), facial expressions may be particularly valuable for detecting deception in young children due to their developing facial muscle control and cognitive abilities (Bruer et al. ). Young children's facial muscle control is not yet fully developed (Ekman et al. ; Feldman et al., ), making them more likely to display spontaneous facial expressions when lying (Bruer et al. ). Moreover, successful deception requires children to coordinate their facial expressions with their verbal statement and the corresponding emotion display rules (Zanette et al. ). However, these emotion display rules are still developing in early childhood, before children reach primary school age (Cole, ; Gnepp and Hess ; Saarni, ), making it difficult for them to regulate facial expressions to fit these emotion display rules. For these reasons, children's spontaneous facial expressions may serve as valuable nonverbal cues for detecting deception. Nonetheless, adults often struggle to detect children's deception, as facial expressions can be fleeting and too subtle for even professionals to interpret reliably (Evans et al. ). Thus, computerized fine‐grained facial expression analysis may be necessary to improve the detection of children's deception. Lewis and his colleagues () conducted the first study analyzing children's facial expressions during lie‐telling, coding four facial expressions: smiling, gaze aversion, sober mouth, and relaxed interest mouth. They found that truth‐tellers smiled less intensively and had a more relaxed mouth position than deceivers. Building on this pioneering work, Talwar et al. () also examined children's facial expressions during both lie‐telling and lie maintenance. Consistent with Lewis et al. (), their results indicated that deceivers were more likely to exhibit big smiles and less relaxed mouth expressions than truth‐tellers.