Participants who chose the correct perp will be coded as 1 in a 'correct' variable, all others as 0. People who said the perp was not present will be coded as 1 in a notpresent variable, all others coded as 0. Our interest here is not on statistical significance directly, but more on magnitude of effect. In our previous investigation we found no effect of verbalizing after 10 minutes of waiting (-1% accuracy, *p* = .833), but an effect at 15 minutes (-8%, *p* = .004). We therefore expect the magnitude of the verbal overshadowing effect to vascillate somewhere between 0% and 8% across the 11, 12.5, and 14-minute delays. This will help us identify at what timing would such a 'medium term memory' be most susceptible to interference in between 10-15 minutes. To stay consistent with the precision of our estimates with our previous investigation, we will aim for 1100 participants per timing. Our primary analysis will be done using the following code: by timing: regress correct i.verbalize, vce(robust) Furthermore, we expect in all three conditions that verbalizing to have a significant effect on judgments of syaing the perp was 'not present'. This will be done using the following code: by timing: regress notpresent i.verbalize, vce(robust) Finally, we will also test the probability of selecting correctly given that someone chose anybody. Our previous investigation showed that there was no effect at 10 minutes but an effect at 15 minutes. This will be done with the following code: by timing: regress correct i.verbalize if notpresent == 0, vce(robust)