We will first confirm the overall VO effect can be replicated using new distractor tasks. This will be done using the following code: regress accurate i.verbalize, vce(robust) where accurate = 1 if they chose the correct option and 0 if they chose any other option. We will also confirm that verbalizing causes people to be more likely to select 'not present'. This will be done with the following code: regress notpresent i.verbalize, vce(robust) where 1 = selecting the 'not present' option and 0 = selecting any other option (including the correct option). We will confirm that verbalizing makes people more confident in their responses when they are incorrect or think the robber is not present but not when they are actually correct. This will be tested with the following code: regress confidence verbalize if whodunnit == 9, vce(robust) regress confidence verbalize if whodunnit != 6 & whodunnit != 9, vce(robust) regress confidence verbalize if whodunnit == 6, vce(robust) Next, we will confirm that people who are more likely to remember someone's face are more susceptible to the verbal overshadowing effect. This will be tested with the following code: gen facerememberverb = facerememberer*verbalize interflex accurate verbalize facerememberer facerememberverb, vce(robust) Next, we will confirm that mental rotation ability moderates the verbal overshadowing effect but only for people who are 'face rememberers'. This will be done with the following code: regress accurate verbalize mrts mrtsverbazlie if facerememberer == 1, vce(robust) regress accurate verbalize mrts mrtsverbazlie if facerememberer != 1, vce(robust) where mrts is the total number of mrts items correct using the standard scoring procedure (each item = 1 if both selections are correct and 0 otherwise). We predict to replicate the fact that there is a non-linear relationship among mental rotation ability and verbal overshadowing such that people low in mental rotation ability are most affected by it but those high are unaffected (only among face-rememberers). This non-linear relationship will be shown via the following code (Hainmuller et al., 2019): interflex accurate verbalize mrts mrtsverbazlie if facerememberer == 1, vce(robust) Finally, we will test our new variable that the better someone is at remembering faces the more affected they will be. This will be done with the following code: regress accurate verbalize faceconf faceconfverb, vce(robust) All these analyses will be done with complete case analysis, meaning we will not drop any participants who fail the exclusion criteria of Alogna et al., 2014. **References** Alogna, V. K., Attaya, M. K., Aucoin, P., Bahník, Š., Birch, S., Birt, A. R., ... & Buswell, K. (2014). Registered replication report: Schooler and engstler-schooler (1990). Perspectives on Psychological Science, 9(5), 556-578. Hainmueller, J., Mummolo, J., & Xu, Y. (2019). How much should we trust estimates from multiplicative interaction models? Simple tools to improve empirical practice. *Political Analysis*, 27(2), 163-192.