Experiment 2: A preregistered replication testing how face sex biases the evidence for angry and happy expressions

This preregistration covers Experiment 2, a new sample collected with the same task, stimuli and procedure as Experiment 1. It tests hypotheses that were partly generated by Experiment 1. No Experiment 2 data have been collected at the time of registration. **Relationship to earlier registrations and to Experiment 1.** Experiment 1 was preregistered on 12 April 2023, before data collection (Tipples, 2023, PsychArchives, https://doi.org/10.23668/psycharchives.12670). Data were collected on 14 April 2023 at Leeds Beckett University. A second registration of secondary analyses of the same data was made on OSF in October 2026 ([DOI]); it replaced a withdrawn registration (https://doi.org/10.17605/OSF.IO/3AJVQ). For the analysis of new samples, this registration supersedes the October 2026 registration. Experiment 1 (51 White women before exclusions) showed the following: 1. *Choices (registered analysis).* Neutral male faces were categorised as angry more often than neutral female faces. The posterior probability of this direction was about .97. The registered interval criteria were narrowly not met. Sensitivity (the slope of the psychometric function) differed little between face categories. 2. *Response times (registered analysis).* At 100% morphs, angry-male and happy-female faces were categorised faster (a crossover). Registered H4--H6 were met. 3. *Diffusion model (registered and exploratory).* Male faces added angry-leaning evidence (drift) for neutral and angry expressions, and essentially none for happy expressions. This was the case in both the registered model and an exploratory model with participant-level non-decision time. 4. *Starting point (exploratory, unresolved).* A face-sex effect on starting point was small and uncertain in the registered model (single non-decision time, which was fixed by the sample's fastest response) and clear in the participant-level non-decision-time model. Whether face sex shifts the starting point therefore depends on the model specification in Experiment 1. 5. *Face race (exploratory).* Neutral Black faces were categorised as happy more often than neutral White faces, and drift for neutral Black faces was more "happy"-leaning. Experiment 2 tests findings 1--3 as replications and findings 3--5 as new confirmatory hypotheses, with the analysis specification fixed in advance. That specification includes the participant-level non-decision-time model identified in Experiment 1.