Affective priming from briefly presented facial expressions: a three-level and network meta-analysis

OBJECTIVE: Affective priming typically refers to two related phenomena: facilitated processing of targets that match the affective properties of a preceding prime, and systematic biases in subsequent judgments toward the prime's valence. Whether affective priming reliably occurs under very brief, often subliminal, presentations of emotional facial expressions remains debated. To address this issue, we conducted a three-level meta-analysis and a network meta-analysis of affective priming effects elicited by briefly presented (≤ 150 ms) facial expressions. METHODS: A total of 62 studies, comprising 182 effect sizes and involving 4,772 participants, were included. Three-level random-effects models were used to estimate the overall priming effect and to examine moderator influences, and network meta-analysis was applied to compare and rank priming effects across basic emotion categories. RESULTS: Three-level random-effects models revealed a significant overall priming effect (Hedges' g = 0.32). Specifically, happy (g = 0.31), sad (g = 0.31), angry (g = 0.29), and fearful (g = 0.28) facial expressions produced reliable priming effects, whereas disgusted expressions did not. Moderator analyses showed that supraliminal primes elicited larger effects than subliminal primes; nevertheless, subliminal presentations yielded stable and reliable priming effects. Importantly, category-oriented instructions produced significantly stronger priming effects than valence-oriented instructions. Network meta-analysis further demonstrated that happy faces elicited the strongest priming effect (p-score = 0.82). CONCLUSION: Together, these findings provide converging evidence for the robustness of affective priming elicited by briefly presented facial expressions, including under subliminal conditions. The results suggest a relatively consistent processing advantage for happy facial expressions and indicate that brief affective priming is systematically modulated by awareness level, emotion category, and instruction type.