However, it should be noted that the amount of time participants observed the culprit while he was handing out the papers before they were instructed to remember him may have varied and was not directly controlled by the experiment. The participants were then asked to retrieve features of the culprit, which they would generally use to describe him to another person, and were asked not to talk about him with the other class attendees.
Participants then had their classroom sessions as scheduled. At the end of their session, the examiner entered the room again, but this time without the culprit, and handed out the questionnaire to the participants. The experiment was conducted in this way to resemble a possible real-life theft without having to stage a theft. As in a theft, the culprit might be seen up close, but the victim often only realizes the theft after the perpetrator has left or is leaving and while the victim is looking e.
The creation of a similar situation was attempted, with the participants being able to view the culprit in proximity while distributing the sheet protector, yet they only learned that they had to remember the culprit once he is not as close anymore. In the open questions, answers would count as correct if at least part of the answers were correct. For example, if eye color was stated as blue-green and the true color was blue, this would be correct.
If only dark hair was stated instead of brown, this would be counted as a missing value. The reason, therefore, is that when selecting pictures to conduct a mugbook search later and all culprits with blue or green eyes are included, the picture would be included yet dark hair would not be precise enough to include or exclude pictures. Age and height were not rated as correct or incorrect, as it is hard to determine what estimation can still be counted as correct and what not Meissner et al.
In these categories, accuracy was measured as the difference between estimated and true value. If a person did not estimate a certain age but a range instead e. Data were collapsed over all the culprits collectively as the design of the experiment would not allow identifying differences between culprits or classes, as each culprit was present in only one class. According to the age of the potential culprit, the mean difference of estimate and true age was 2.
When the distribution of these estimates are examined, Plots of the estimates can be found in Figure 1. Figure 1. The left boxplot shows the distribution of the difference of estimated age and true age of the culprit. Numbers represent the differences in years. The right boxplot shows the distribution of the difference of estimated height and true height of the culprit.
Numbers represent the differences in centimeters. Assumptions of the ethnicity of the culprit were correct for 65 participants Hair color was remembered correctly by 44 participants To determine the relationship of how well people thought they could remember the culprit and their true features, the relationship between the estimation of the own memory accuracy and the actual person descriptor accuracy was measured. Overall, Similar results were obtained for estimations of ethnicity, hair color, and eye color, which were analyzed with the Chi-Squared test to see if the results in accuracy depended on the memory.
Results from experiment 1 support the findings by Kuehn that witnesses would provide a good description about more general culprit features but were less accurate in specific features such as hair or eye color. Also, our results are similar to the numbers found by Fahsing et al.
The question nevertheless remains whether a good general description and descriptive features provide a valid basis for decisions about which culprit pictures are included in the subsequent mugbook search and which not. And if only ethnicity is considered, then However, especially in large cities or populations, where many thousands or up to a million mug-shots might be stored Lindsay et al.
These results seem rather daunting. It appears to be ideal to use a matching approach with caution and consider a sequencing approach when attributes are considered with a high error rate.
A possible explanation for this might be that people are simply not good at estimating their own memory capability, yet it is also possible that these differences were found because people were accurate in their estimation of their memory, however, the estimations were wrong from the beginning. The latter could indicate that participants were overall inaccurate in making judgments about culprit traits directly, which would indices a general cognitive difficulty in judging person traits rather than a subsequent memory problem.
Future studies thus might be interested in finding out if people were able to encode attributes in the first place and forgot them subsequently or if they were not able to encode them correctly meaning it would not be a problem of their own memory and therefore the judgment of the own memory accuracy might be right but the attribute all together was wrong. While experiment 1 aimed to simulate the preparatory phase of a mugbook search in an experimental setting, experiment 2 aimed to experimentally simulate the mugbook search itself as the second phase of a mugbook investigation.
In experiment 2, witnesses encountered a staged crime and were subsequently asked to identify the potential culprit in an experimentally controlled mugbook search.
The experiment included an independent sample of participants. Two participants did not understand the questionnaire correctly and had to be excluded, as they incorrectly answered the questions about a simulated crime they had seen prior to participating in the experiment. The remaining participants consisted of 82 female Additionally, there were 11 participants 9. All participants were fluent in German as the whole experiment was conducted in German. Regarding the highest level of education, two participants were junior high school graduates 1.
Nine 7. Participants were shown one of two videos depicting a non-violent theft of a handbag. Both movies were identical except for the perpetrator. The movie lasted 3 m 15 s with the theft taking place in the final seconds of the movie and the perpetrator being seen for 45 s overall with a clear view of his face for approximately half the time.
The video was made with cuts in a movie like manner and the culprit was seen from different distances varying from 1 to 20 m approximately.
The only people seen in the movie were the victim, a witness, and the perpetrator. Two culprits were chosen to minimize the chance that the results are affected by distinct features of one specific culprit Badham et al. A mugbook was created, including a total of pictures. Most faces depicted in the mugbook were Caucasian males, however to increase the size of the mugbook, a small amount of pictures depicted Arabic or Latin faces.
All pictures used were in color, depicting males with neutral emotional expression in frontal view. All pictures were edited to have a white-gray background and cropped to have the head sizes matched. The picture of the culprit was added to the other pictures creating two separate mugbooks, each consisting of pictures. The picture of the culprit matched the other pictures in the mugbook search. Instead of creating an extra culprit absent mugbook, the mugbook containing the picture of the culprit not seen in the movie was used when an absent mugbook was tested.
The order of the pictures, including the culprit was randomized. The questionnaire contained short demographic questions, a slightly adapted version of the Self-Administered Interview SAI Gabbert et al. The SAI offers a tool for witnesses to recall their memories of an incident by themselves, following specific instructions and questions. The SAI supports people to give a complete description of everything the witness has seen in the crime and follows up with more specific questions about the perpetrator and other details of the crime Gabbert et al.
The SAI was used to support the witness to provide as much information as possible without the necessity of an interaction between subject and examiner, which could affect the results.
The participants were randomly assigned in a 2 culprit A vs. As mugbook searches tend to show less correct identifications due to the large number of pictures seen Meissner et al. The second 60 participants were randomly assigned to the different conditions in the aforementioned 2. Participants were told they would participate in a study about daily activities.
They were tested individually, and the session lasted approximately 1 h. Participants viewed a short movie and were told to watch the movie attentively without knowing a crime was about to happen. After the movie, they were instructed to memorize the theft they had witnessed and were told that they would be questioned about what happened later. They were given a 10 min distractor task i. Following the distractor task, participants were asked demographic questions about the culprit followed by the SAI and some additional questions to the crime.
In the end, they were asked how likely they were to recognize the perpetrator in a following mugbook search if he was present. Subsequently, they completed the mugbook search.
The search included either the culprit present or culprit absent mugbook. Pictures were presented in a grouped procedure resulting in 15 pictures per page, and 20 pages presented one after the other. They were not allowed to go back to previously viewed pages, but they were allowed to choose up to 15 pictures of possible culprits, which would be shown again later on a final screen.
Finally, they were presented with an eight-person line-up, with the culprit included and they could again choose if the culprit was present and indicate the person or not. After testing, participants were fully debriefed.
A visual depiction of the Experimental procedure can be seen in Figure 2. Figure 2. Experimental procedure left panel : After a movie of a staged crime was shown, participants were given a distractor task for exactly 10 min. Mugbook procedure right panel : Participants looked through 20 pages with each 15 pictures and were able to choose a person if they think they recognized the culprit.
All people selected on the 20 pages were shown on the last page where participants could make a definitive selection or reject the mugbook search. Statement analysis of the SAI was conducted by two independent coders who were blind to the outcome of the mugbook search.
Statements were analyzed using a scoring template. The template was generated according to the information presented in the movie. Information about the culprit was collected concerning facial characteristics and additional characteristics. For each item every individual characteristic of this item was scored with two points if mentioned correctly.
Two points were given for a correct item to also allow partly correct statements, which were given one point e. Incorrect statements about an item were given zero points. Additionally, a category was created for statements that did not contain falsely reported additional features not previously accounted for in other items. If no false additional features were reported, participants received two points in this category. If only minor false additions were made things that should not affect the recognition greatly; e.
Items were included into categories, and points were added together accordingly. Finally, scores were made for features mentioned about clothing, general characteristics e. Since experiment 1 and 2 included different ratings strategies, we would like to give our reasoning for this decision. The different rating strategies included in experiment 1 vs. In experiment 1, the goal was to assess if the suspect would be included in the mugbook search, meaning that a correct answer even on the most general level would include the suspect.
More precisely, the scientific interest was primarily in falsely remembered attributes where the suspect would be excluded from the lineup. Therefore as long as the answer was not wrong, the accuracy of the description is sufficient. In experiment 2, we wanted to assess how accurately the suspects were remembered, meaning we had to differentiate participants who give answers on a different level of accuracy.
As the goal was to assess if a better accuracy of remembering would offer a better recognition, we had to differentiate e. However, both of these descriptions would lead to the inclusion of the suspect in the mugbook in experiment 1. The results showed that Incorrect decisions were on the one hand false person identifications, as in A resume of these results can be found in Table 2.
Demographic differences and differences in decision rates between the results for each culprit were marginal. However, it has to be mentioned that the proportion of female participants in the experiment was not equally distributed between both culprits. Yet, this did not affect the results as analyses were only made for culprits combined.
A summary table of the ICC can be found in Table 3. Table 3. Results of the ICC quantification for the statement analysis in study 2. To calculate the relationship between written statements of a witness and the accuracy of the witness in the mugbook search, the total amount of features mentioned was analyzed for correct and incorrect mugbook searches as seen in Figure 3. Therefore, a one-tailed independent t -test assuming equal variances was used to analyze differences in means between the groups.
Figure 3. Distribution of the number of features mentioned for correct and incorrect mugbook decisions. Frequencies represent the number of witnesses that mentioned the number of features. Additionally, to analyze the relationship between total features mentioned and correctness in the mugbook search a point-biserial correlation was conducted. These results show that people who were able to state more information about the culprit in the previous questionnaire also tend to be better in the recognition task later on.
Additionally, statements were analyzed concerning external and internal facial features. Descriptive statistics concerning the amount of external and internal facial features mentioned by participants can be seen in Table 4. Table 4. Descriptive statistics for external and internal facial features of culprits reported by participants.
To test if the amount of correct statements about external facial features is related to the accuracy in a following mugbook task, the number of correct statements were analyzed depending on correctness of the later mugbook search.
Therefore, a one-tailed Mann-Whitney- U test was conducted. Total distributions of predictive confidence levels can be seen in Table 5. The goal of experiment 2 was to assess if the quality of the culprit description given by the witness is positively correlated with the ability to correctly identify the culprit in a subsequent mugbook search.
Here, the participants who mentioned more features of a culprit in total tended to be more accurate in their decision when trying to identify the culprit in a mugbook search. These results are in line with early assumptions made by the United States and German supreme court, which assumed that the quality of a person description was an indicator in the evaluation of the accuracy of person identifications in criminal trials Sporer and Cutler, However, correct statements and descriptions alone should not be considered as the sole indicator for identification accuracy.
First, the effect size in our study was only small to medium. Second, if results are considered individually, the six participants who mentioned the most features all made an incorrect mugbook decision.
Third, in this experiment only correct statements were counted and not the total amount of statements. In real life, the accuracy of statements cannot be verified prior to the mugbook search and sometimes even after the search Fahsing et al.
Yet, our results can nonetheless be seen as optimistic as the ability to give a correct verbal description of a culprit seems to be related to the ability to correctly identify the culprit in a mugbook task. To analyze which specific facial features were more important for the identification task and the witness, two separate hypotheses were tested. First, witnesses were expected to describe significantly more external than internal features, and our participants indeed mentioned a larger number of external compared to internal facial features.
This is in line with previous research showing that internal features were rarely mentioned in eyewitness descriptions Lindsay et al. Second, a positive correlation between the number of correctly remembered external facial features with identification accuracy was expected.
Participants who either correctly rejected a culprit absent in the mugbook or who identified the culprit in a culprit present mugbook, tended to give more external facial descriptors compared to those who made a false identification or missed the culprit.
This supports the assumption that external features play an important role in the recognition of unfamiliar faces Ellis et al. However, a problem with these features is that while they might be easily expressible, they might not always be useful to differentiate between faces Meissner et al. External features such as hair color or length may be changed rather easily, or they can be easily disguised, e.
These simple changes in the external features can lead to impaired recognition accuracies Chan and Ryan, Especially in mugbook searches where pictures might be rather old, there is a high probability that some certain external face features might have changed in potential culprits.
The final goal of experiment 2 was to assess if predictive confidence will show a positive correlation with actual performance in mugbook searches. We, however, did not find evidence for this proposed association here. There was no relation between predictive confidence and identification accuracy in the mugbook search, even though they were made after a brief delay and not immediately after viewing the culprit.
These findings differ from line-up studies, which did find a relation between predictive confidence and identification accuracy, when confidence ratings were made after a brief delay Nelson and Dunlosky, ; Nguyen et al. One reason for this might be the way how witnesses form their accuracy assumption concerning a later identification task.
According to Leippe et al. While intrinsic sources and self-credibility are person-dependent Leippe et al. For the witness, it might be difficult to estimate how challenging a mugbook task is, and therefore, they cannot properly estimate how accurate they will be in a following mugbook search. The aim of the experiments reported here was to analyze if there were any differences between correct and incorrect mugbook searches when looking at prior culprit descriptions and information in the preparatory phase.
Especially we wanted to assess if a positive outcome can be predicted after an initial statement and description about the culprit is given by the witness. Experiment 1 showed that there is a high risk that the real culprit is not included in the mugbook if a matching approach i. Yet some matching has to be done due to the large number of pictures in certain population settings Lindsay et al.
The challenge, therefore, is to analyze which culprit description items should be considered or disregarded in the preparation for the mugbook search. However, results from previous studies show that height estimates were more accurate compared to age estimates Fahsing et al. It seems that the decision, which items should be used to minimize the mugbook size, cannot be answered in general but is more a case-to-case question.
Yet how this decision is made is challenged by the fact that results in our experiment did not show a correlation between subjective memory strength estimations and accuracy in descriptions. This suggests that attributes, which are remembered with higher certainty, should not automatically be given a higher value. This highlights the notion that information that could affect identification accuracy should be carefully taken into account.
First, persons are better at estimating someone of their own age group Voelkle et al. Second, for height estimates, people tend to use eye height scaling Twedt et al. This means that if the eye-height of the witness during the crime was about the same height as the culprit during the crime, a smaller margin of error can be chosen when height estimates are analyzed. Also, height estimations seem more accurate if they are done in the same body position as at the time of encoding the height Twedt et al.
The results of Twedt and colleagues also show that ideally a matching approach presentation of the pictures are limited to those that match the description is only used if really necessary; it seems overall better to use a sequencing approach pictures are sorted that pictures which better match the description are presented earlier for as many attributes as possible.
Especially, as hair color for example is easily changeable and eye color was not remembered correctly by many participants, it seems suboptimal to eliminate pictures according to eye or hair color.
The use of these easily changeable and not often accurately remembered attributes should therefore be limited to a use in a sorting procedure. Interestingly in Experiment 1, there was no correlation between self-estimation of how well a witness could remember the culprit and the accuracy of the person description. If witnesses were able to analyze how well they remembered features that are visible but not those that have to be estimated prior to memorizing them, we might conclude that people did not forget these attributes but instead that they had difficulties with making accurate estimates from the beginning.
Furthermore, body scaling or eye height scaling is used to making estimations Twedt et al. In our experiments, however, all participants were seated when they viewed the culprit, which did not allow them to use their own eye height to estimate the height of the culprit. In experiment 2, an experimentally simulated mugbook search revealed that only 30 of 90 participants selected the culprit, and only 20 participants correctly identified the culprit in the final decision.
This might seem to be a satisfactory identification rate, as pictures had to be viewed, and the possibility of finding the right person by chance is 1— However, a critical notion concerns the fact that from total mugbook searches, 48 wrong identifications were made.
We have to note that the calculation of a diagnostic ratio was not performed here due to the fact that in mugbook searches, the difference between misses and false alarms is not as clear as in line-ups. In line-ups an innocent person is usually presented in the culprit absent line-up condition, as compared to real line-ups, where there is also one known culprit placed among innocent foils Lindsay et al.
Calculations in this case are usually made by considering foil identification in culprit present conditions simply as misses and false alarms as identification of the innocent culprit Wells and Penrod, This allows a clear distinction between false alarms and misses.
In mugbook searches. However, the culprit is not known and every foil has to be considered to be an innocent person. For this reason, no diagnostic ratio was created in our study. Yet, the high number of false identifications supports the assumption that mugbook searches should not be used as an identification tool but rather as an investigative tool Lindsay et al.
A correct identification in the mugbook is thus not proof that the culprit has committed the crime but rather that this person should be further investigated. The question remains why such a large number of false identifications were found in our study. One possible reason is that due to the design of the mugbook search, people are likely to use a relative judgment, where they choose the person who looks the most like the culprit, and they do not try to recognize the culprit specifically Lindsay and Bellinger, To increase the chance of finding the culprit amongst all of the pictures, witnesses were allowed to choose more than one culprit in a first identification attempt.
The grouped mugbook procedure should prevent the use of relative judgment McAllister et al. However, once participants are allowed to choose more than one culprit, they could be tempted to select different pictures of people who look similar, eliminating the advantages of the grouped procedure. Pictures that were chosen initially were later presented simultaneously on the last page.
It is possible that the number of false identifications could be reduced by presenting the last pictures sequentially. Sequential presentations generally seem to be superior for line-up identification and reduce the amount of false identifications Wells and Lindsay, Nevertheless, as the process of a mugbook search should be considered rather as an investigative tool instead of an identification process Lindsay et al.
One of the main motivations for our experiment was to assess if there are certain indicators that can be used to estimate if a mugbook search is advised or not. Here, accuracy in a mugbook search was positively correlated with the accuracy of a verbal statement about the culprit and the amount of external facial features correctly mentioned.
There are certain aspects that could not be addressed in our experiments. Concerning experiment 1, findings can only be generalized for culprits who present the same attributes as those used in this sample. While the participant sample in experiment 2 seemed to be more diverse overall, college students were represented above average. This might lead to the fact that eyewitness abilities are overestimated, as college students tend to be more accurate as eyewitnesses Wells et al. Second, there are numerous estimator variables that may influence recognition, such as racial bias, stress, weapon focus, exposure duration, disguise, and presentation delay Wells et al.
A third limitation of our study is that the culprits did not significantly change the features of their face between movie and picture. In real-life mugbooks, it is possible that certain external facial features, such as hair color or length, may have changed drastically, especially if the culprit was registered for a previous offense long ago.
The same question can be raised if the culprit has a very distinctive internal facial feature such as a very recognizable nose or a birthmark. In that case, it might be possible that mentioning of this internal feature will show a better correlation compared to external features.
Fourth, the Self Administrated Interview offers a tool to get as much information about the crime witnessed Gabbert et al. While this tool offers the possibility to test the participants in a standardized way and minimizing any possible effect the experimenter would have when asking questions, it might not obtain all relevant information.
Although the SAI uses techniques from the Cognitive interview to generate an elaborate statement about the crime Gabbert et al. Our data provide valuable insights in the accuracy of person descriptions in the preparation for and for performing of actual mugbook searches. Experiment 1 showed that the use of too many culprit features to down-scale the mugbook would lead to a high chance of the culprit being eliminated from the mugbook.
For uncertain culprit attributes, a sequencing approach should be prioritized. Taken together, the results of our experiments allow a few recommendations about when a mugbook search might or might not be advisable. Predictive confidence did not show any correlation with the actual identification accuracy. As there is no absolute indicator about the success in the mugbook search, witnesses who are only able to describe very few details about facial features will probably perform worse in the subsequent mugbook search.
Third, there are no absolute indicators that allow prediction on the outcome of a mugbook search. For example, witnesses who make very elaborate statements in the description phase mentioning a multitude of facial descriptors and having promising estimator variable features, can still make a false positive identification.
Fourth, our study shows nevertheless that there are certain circumstances that make a positive outcome of the mugbook search more likely. Specifically, a mugbook search is more advisable if the witness is able to give an elaborated and convincing description of the potential culprit. The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. Short 6. Does the person have any distinguishing features?
Striped shirt 8. Is the person wearing any jewelry? If so, describe it. Gold necklace and earrings 9. Can you describe where the picture was taken? School room or office Teacher Is it possible to identify the interests of the person based upon evidence in the room? Forensics and biology books imply this might be a science teacher. Discussion Have your views on eyewitness testimony changed? Discuss with your classmates.
Slides: 30 Download presentation Eyewitness Lab. You will have 15 seconds to look at the following photograph. Photograph 1. Now turn over your sheet and answer questions 1 -7 for photograph 1.
Photograph 2. Now turn over your sheet and answer questions 1 -7 for photograph 2. Photograph 3. Now turn over your sheet and answer questions 1 -7 for photograph 3. Smash and Grab:. Test Your Awareness. The Phone Joke Test. The Bank Job. Eyewitness experiment with Dr. Watch 60 minutes- Eyewitness Memory Part 1 and 2. Still interested in learning more? Daniel Simons- illusions. Our memories are reconstructed.
Which is the real penny? Authors have about 10 seconds to grab you. You have 10 seconds to name Concepts of. Label the picture You have 60 seconds Convict. I have you have they have This term. Evaluating eyewitness testimony Does gender matter Igor Areh. Sharing Testimonies In The Bible eyewitness testimony quotes. Show video 60 minutes eyewitness segment 13 min.