We analyzed three components of the data, as required in the protocol. This included: (1) computing a Chi-square comparing percentage of correct identifications in the experimental and control conditions; (2) computing a Chi-square comparing the ratio of the percentage selecting the wrong face (misidentification) to the percentage indicating “not present” across the experimental and control conditions; and (3) computing a 2 (Condition) x 2 (correct vs. incorrect/miss) ANOVA on confidence ratings. With regard to correct identifications, results indicated that there was not a significant difference between those participants who described the suspect, relative to those who did not describe the suspect, ᵡ² (1) = .00, p = 1.00. In fact, participants in both conditions correctly identified the target suspect 41.8% of the time. In examining the ratio of the percentage of participants selecting the wrong faces, relative to the percentage of those selecting “not present,” results indicated a marginally significant effect between those who described the suspect and those who did not, ᵡ² (1) = 3.09, p = .079. These results suggest that those who describe the suspect trend toward being less likely to select the wrong answer (34.4%), relative to selecting “not present” (65.6%). However, participants who do not give a description trend toward being more likely to select the wrong answer (56.3%), relative to selecting “not present” (43.8%). Finally, our analysis of confidence ratings indicated that there was a marginally significant effect on confidence depending on whether the participant selected the right suspect, F (1,106) = 3.86, p =.052. Namely, participants who were correct (M = 4.67, SD = 1.01) also appeared to be more confident, compared to incorrect participants (M = 4.20, SD = 1.38). There was no main effect for those who described the suspect versus those that did not, F (1,106) =.50, p =.482, and no interaction between accuracy of decision and description condition, F (1,106) = 1.54, p =.214. ***Study 2*** We analyzed three components of the data, as required in the protocol. This included: (1) computing a Chi-square comparing percentage of correct identifications in the experimental and control conditions; (2) computing a Chi-square comparing the ratio of the percentage selecting the wrong face (misidentification) to the percentage indicating “not present” across the experimental and control conditions; and (3) computing a 2 (Condition) x 2 (correct vs. incorrect/miss) ANOVA on confidence ratings. With regard to correct identifications, results indicated that there was a significant difference between those participants who described the suspect, relative to those who did not describe the suspect, ᵡ² (1) = 6.42, p = .011. Specifically, in line with the original verbal overshadowing study, participants who gave a verbal description were significantly less likely to correctly identify the perpetrator (.22 vs. .46). In examining the ratio of the percentage of participants selecting the wrong faces, relative to the percentage of those selecting “not present,” results indicated a marginally significant effect between those who described the suspect and those who did not, ᵡ² (1) = 3.56, p = .059. These results suggest that those who describe the suspect trend toward being less likely to select the wrong answer (25.6%), relative to selecting “not present” (74.4%). However, participants who do not give a description are approximately equally likely to select a wrong answer (48.1%), relative to selecting “not present” (51.9%). Finally, our analysis of confidence ratings indicated that there was a significant effect on confidence depending on whether the participant selected the right suspect, F (1,96) = 6.592, p =.012. Namely, participants who were correct (M = 4.88, SD = 1.30) also appeared to be more confident, compared to incorrect participants (M = 4.14, SD = 1.31). There was no main effect for those who described the suspect versus those that did not, F (1,96) =.054, p =.817, and no interaction between accuracy of decision and description condition on confidence, F (1,96) = .310, p =.579.