Results and Discussion
----------------------
STUDY 1
<br><br>
**Subjects**
<br>
We collected data from 138 subjects in 13 sessions consisting of at most 12 subjects per session. The end of the data collection was determined according to the method described in the Method & Procedure node. After collecting data from 110 subjects, we continued running additional sessions until we met the required sample size after exclusions (the exclusion criteria are described in the Data node). Subjects were recruited from our laboratory subject pool mostly consisting of students of Prague universities. They were compensated 100 CZK (approx. $ 5) for participation. All subjects were white and without any vision problems. Twenty subjects were excluded because they: (i) indicated that they knew the tested hypothesis (*n* = 11), (ii) heard about the study design (*n* = 2), (iii) knew the verbal overshadowing effect (*n* = 5), and/or (iv) because some technical problems occurred during the administration (*n* = 5, please note that multiple reasons might have applied to one participant). In the final sample, the mean age of subjects was 21.7 (*sd* = 1.7, range: 18-25) and there were 69% females. Subjects were randomly divided into the experimental (*n* = 68) and control group (*n* = 50).
<br><br>
**Results**
<br>
*Recognition accuracy*
<br>
The target face was correctly identified by 53% of subjects in the Face Verbalization condition and by 46% of subjects in the Control condition, χ<sup>2</sup>(1) = 0.56, *p* = .46, *r* = .07. Thus, our data do not confirm the previously found effect. In fact, subjects in the Face Verbalization condition correctly identified the target face slightly more often than subjects in the Control condition. Misidentifications comprised of 53% of errors in the Face Verbalization condition and 67% of errors in the Control condition (the remaining errors were misses), χ<sup>2</sup>(1) = 1.11, *p* = .29, *r* = .14. No effect was found in the original study as well.
<br><br>
*Confidence*
<br>
Subjects were more confident when they correctly identified the target face, *F*(1, 114) = 7.24, *p* = .008, ω<sup>2</sup> = .06. No significant effect of condition, *F*(1, 114) = 2.53, *p* = .11, ω<sup>2</sup> = .02, or interaction between condition and answering correctly, *F*(1, 114) = 2.29, *p* = .13, ω<sup>2</sup> = .02, was found. Mean confidence ratings are presented in the *Table 1*. The results of analysis of confidence ratings were generally the same as in the original study.
<br>
<br>
*Table 1*
<br>
Mean confidence in responses
Response Verbalization Control Overall
Correct 5.06 5.09 5.07
Incorrect 4.75 4.11 4.46
Overall 4.91 4.56
<br>
**Discussion**
<br>
The primary hypothesis was not supported by the results in our replication. We found no significant difference between the two groups in correct identification of the target face from the lineup. All other analyses showed the same results as the original study. There was no significant difference in the proportion of misidentifications and misses between the two groups. Confidence ratings differed depending on whether the identification was correct: subjects who answered correctly reported higher confidence in their choice than subjects who answered incorrectly--a result which was in accordance to the original study.
<br><br>
<br>
STUDY 2
<br><br>
**Subjects**
<br>
We collected data from 110 subjects in 16 sessions consisting of at most 12 subjects per session. Recruitment and compensation were the same as in Study 1. All subjects were white and without any vision problems. The data collection was terminated after the 16<sup>th</sup> session, when we checked whether our sample met the criteria stated in the approved protocol after all exclusions (the exclusion criteria are described in the Data node). Seven subjects were excluded because they: (i) indicated that they heard about the study design (*n* = 4), (ii) knew the verbal overshadowing effect (*n* = 2), or (iii) knew the tested hypothesis (*n* = 1); one subject (from the experimental condition) was excluded because she did not provide any description of the robber. In the final sample, the mean age of subjects was 21.8 (*sd* = 1.9, range: 19-25) and there were 77% females. Subjects were randomly divided into the experimental (*n* = 52) and control group (*n* = 50).
<br><br>
**Results**
<br>
*Recognition accuracy*
<br>
The target face was correctly identified by 31% of subjects in the Face Verbalization condition and by 34% of subjects in the Control condition, χ<sup>2</sup>(1) = 0.12, *p* = .73, *r* = -.03. Thus, data from Study 2 also do not confirm the previously found effect. Misidentifications comprised of 53% of errors in the Face Verbalization condition and 61% of errors in the Control condition (the remaining errors were misses), χ<sup>2</sup>(1) = .43, *p* = .51, *r* = .08. No effect was found in the original study as well.
<br><br>
*Confidence*
<br>
Subjects were more confident when they correctly identified the target face, *F*(1, 98) = 19.32, *p* < .001, ω<sup>2</sup> = .16. No significant effect of condition, *F*(1, 98) = .02, *p* = .88, ω<sup>2</sup> = .00, or interaction between condition and answering correctly, *F*(1, 98) = 1.84, *p* = .18, ω<sup>2</sup> = .01, was found. Mean confidence ratings are presented in *Table 2*. The results of analysis of confidence ratings were generally the same as in Study 1, as well as in the original study.
<br>
<br>
*Table 2*
<br>
Mean confidence in responses
Response Verbalization Control Overall
Correct 5.12 5.53 5.33
Incorrect 4.33 4.03 4.19
Overall 4.58 4.54
<br>
**Discussion**
<br>
The procedure of Study 2 was amended to directly replicate the original Experiment 1 of Schooler and Engstler-Schooler (1990) study, in which the verbal overshadowing effect was found to be the most robust. However, the results of our replication did not support the primary hypothesis. We found no significant difference between the two groups in correct identification of the target face from the lineup. All other analyses showed the same results as the original study. There was no significant difference in the proportion of misidentifications and misses between the two groups. Confidence ratings differed depending on whether the identification was correct: subjects who answered correctly reported higher confidence in their choice than subjects who answered incorrectly – a result which was in accordance with the original study.
<br><br>
<br>
(You can download the complete dataset for both studies in the File section of the Data node, lab logs under the Lab Log node, R script for analysis under the Analysis plan node, and outputs of the data analyses from R in the File section of this node.)
