5. Previous research….
Pryke et al (2004)
Higher probability of guilt
with increasing number of
suspect IDs
But performance on body
and voice lineups was poor
6. Previous research….
Sauerland & Sporer (2008)
Face, profile, body, and bag
Some combinations (e.g.,
face + profile) provided
stronger evidence of guilt
than single face IDs
7. Research question: Across multiple lineups, does the
number of suspect IDs predict suspect guilt? (And what
about filler IDs and non-IDs?)
8. Participants
192 participants (384 trials)
• Prolific.co (n = 148)
• Participant pool (n = 46)
Trials were excluded if:
• Attention check answered incorrectly (n = 78; ~20%)
• Participant reported technical difficulties (n = 8; ~2%)
• Participant provided correct, specific information about any of the
lineup members (n = 3; ~1%)
• Programming error (fixed prior to Prolific recruitment) (n= 5; ~1%)
Final sample = 290 trials (~75% of data collected)
13. 3 DVs (each can vary from 0 to 5):
•Number of suspect IDs
•Number of filler IDs
•Number of non-IDs
How well does each of these predict
whether the suspect is guilty?
14. Target present Target absent
Sus ID Filler ID Non ID Sus ID Filler ID Non ID
Target 1 33.33% 17.92% 48.75% 32.45% 21.51% 46.04%
Target 2 37.07% 13.17% 49.76% 11.20% 22.80% 66.00%
Target 3 36.98% 22.26% 40.75% 8.81% 37.29% 53.90%
Target 4 19.15% 23.83% 57.02% 17.58% 25.45% 56.97%
Overall 31.64% 19.58% 48.78% 17.33% 27.28% 55.38%
Decision frequencies
17. Filler IDs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 2 3 4 5
Probabilityofguilt
Number of filler IDs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 to 3 4+
Probabilityofguilt
Number of filler IDs
0
10
20
30
40
50
60
70
80
90
0 1 2 3 4 5
Count
Number of filler IDs
TP
TA
0
10
20
30
40
50
60
70
80
90
0 1 to 3 4+
Count
Number of filler IDs
TP
TA
18. Non-IDs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 2 3 4 5
Probabilityofguilt
Number of non-IDs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 to 3 4+
Probabilityofguilt
Number of non-Ds
0
5
10
15
20
25
30
35
40
45
0 1 2 3 4 5
Count
Number of non-IDs
TP
TA
0
10
20
30
40
50
60
70
80
90
100
0 1 to 3 4+
Count
Number of non-IDs
TP
TA
19. Target present Target absent
Sus ID Filler ID Non ID Sus ID Filler ID Non ID
Target 1 33.33% 17.92% 48.75% 32.45% 21.51% 46.04%
Target 2 37.07% 13.17% 49.76% 11.20% 22.80% 66.00%
Target 3 36.98% 22.26% 40.75% 8.81% 37.29% 53.90%
Target 4 19.15% 23.83% 57.02% 17.58% 25.45% 56.97%
Overall 31.64% 19.58% 48.78% 17.33% 27.28% 55.38%
Decision frequencies
20. Easier lineups “Doppleganger” lineups
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 to 3 4+
Probabilityofguilt
Number of suspect IDs
0
5
10
15
20
25
30
35
40
0 1 to 3 4+
Count
Number of suspect IDs
TP
TA
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 to 3 4+
Probabilityofguilt
Number of suspect IDs
0
5
10
15
20
25
30
35
40
45
50
0 1 to 3 4+
Count
Number of suspect IDs
TP
TA
21. Summary….
• Multiple suspect IDs do seem to increase the probability that the
suspect is guilty
• BUT…. This depends on how easy/difficult the lineup is
• Innocent suspect/Target similarity is an important moderator
• Filler IDs and non-IDs not very informative
22. Discussion points…
Lineup difficulty as a moderator of the informativeness of n/IDs?
(similar to point made by Sauer et al. 2019 about confidence)
What other interesting analyses/questions could we look at?
Any merit in pursuing this further?
@RuthHorry
r.horry@Swansea.ac.uk