Impact of EFIT6 construction on line-up identification accuracy
Updated: Aug 3, 2020
Gesmina Tsourrai and Josh P Davis
University of Greenwich
We would like to thank participants who contributed to this project online or in the University of Greenwich labs in Spring 2020.
If the police have no suspect, witnesses to a crime may be asked to create a facial composite of the offender from memory. The methods for creating facial composites have evolved over the years, from police artists drawing sketches, to PhotoFIT, where witnesses manually selected and assembled individual facial features into a whole from very large libraries, to computer systems in which a composite face is constructed on a screen (for a review see Davis, Gibson, & Solomon, 2017). The most recently designed computer systems draw on algorithms designed to match genetic principles (EFIT-V, EFIT6). The witness is presented with a series of arrays of computer-generated realistic faces. From each array, the witness is asked to select the face(s) that best matches their memory of the suspect, the genetic algorithm ‘remembers’ their selection and adjusts the faces the witness sees, while later in the process witnesses can adjust facial characteristics (e.g. age, perceived facial characteristics), and individual facial features until they are satisfied the final image is the best they can produce.
A video of the EFIT-V creation process (the immediate predecessor of EFIT6) can be found here (https://www.jove.com/t/53298/holistic-facial-composite-creation-subsequent-video-line-up)
Note – facial composites are not facial photos. They are ‘type likenesses’ and are designed with the aim that with publication of information about the crime (date, place, type) someone familiar might recognise the suspect depicted. The key psychological factor is that the person creating the composite – the first witness – will always be unfamiliar with the suspect – the person identifying the suspect will be familiar. Unfamiliar and familiar faces may be processed differently in human memory (e.g. see Johnston & Edmonds, 2009) and recently introduced facial composite systems are designed to take account of this difference.
Finally, if a potential suspect is identified by someone familiar, the original witness who created the composite may be asked to view that suspect in a line-up (identity parade) to see if they can identify them. Early research on this topic found that creating a composite had a negative impact on line-up identification (e.g. Wells, Charman, & Olson, 2005), particularly in children (Davis, Thorniley, Gibson, & Solomon, 2016). In other words, making a composite of a suspect meant a witness would be less likely to identify that suspect later. Other research has found it may have a positive impact on subsequent suspect identification (e.g. Davis, Gibson, & Solomon, 2014), while the most recent large-scale research in realistic conditions suggests virtually no impact (Pike, Brace, Turner, & Vredeveldt, 2019a, 2019b).. Some of these differences in study outcome may be due to the use of different research protocols and of different composite systems – more positive outcomes may be associated with more modern facial composite systems. Nevertheless, it is important for police to be aware of whether a witnesses’ memory for a suspect will be impacted by creating facial composites with the most up to date systems.
The current research, the final year project of Gesmina Tsourrai’s BSc Psychology degree at the University of Greenwich, was designed to evaluate the impact of creating an EFIT6 (https://visionmetric.com/efit6/), a very recently introduced facial composite system marketed by VisionMetric – on video line-up identification. EFIT6 (or its predecessor EFIT-V) is used by most police forces in the UK and by police forces in over 30 countries (see Davis, Maigut, Jolliffe, Gibson, & Solomon, 2015 for a video depicting EFIT-V construction – EFIT’s predecessor using a procedure very similar to the current research).
1. All participants (n = 194) viewed a video of an actor (the ‘suspect’) for about 1 minute
2. Participants were allocated to an EFIT6-creator group or to a Non-EFIT6-creator group.
3. Approximately two days later, EFIT6-creator group members completed a questionnaire asking for a description of the suspect, and then created an EFIT6 with the assistance of operator – Gesmina Tsourrai – in a laboratory.
At approximately the same time, the Non-EFIT6-creator group completed the description questionnaire – but did not create an EFIT6.
Examples of EFIT6 type likenesses created from memory by participants in the research
4. At least another 2 days later, all participants attempted to identify the suspect from a randomly ordered nine-person video line-up. The suspect was always present in the line-up in this study to match conditions in which the police have arrested the actual offender. Participants, however, were given the normal police warning that “the person you saw in the first video may or may not be present in the line-up”. They also had to watch all nine 15s line-up videos twice (the other eight were of foils) as this is standard police procedure in England and Wales (see Davis et al., 2015 for a video depicting video-up line-up requirements).
Note – all participants were also tested on the Cambridge Face Memory Test: Extended (Russell, Duchaine, & Nakayama, 2009) to measure their face recognition ability, and all EFIT6s were rated for similarity to a real photo of the target actor by two groups of participants. One group did not contribute to any element of the design described above. The second group were the participants who had created the EFIT6s. The aim was to see if people with better face recognition ability create better EFIT6s. This component of the research is ongoing and is not reported here. Sorry to those participants who contributed and are keen to find out more. Watch this space in 2021.
However, we can report the line-up outcome results. Based on the results of the Cambridge Face Memory Test: Extended we also divided participants into two groups (super-recognisers, controls).
Key result 1: Overall, super-recognisers were significantly more likely than controls to correctly identify the target actor from the video line-up at least four days after seeing that actor in a video.
Key result 2: There were no significant differences in line-up accuracy between EFIT6-creator group and Non-EFIT6-creator group participants. In other words, consistent with recent research articles, we cannot find any evidence that creating an EFIT6 has a positive, or a negative impact on the ability of a witness to identify a suspect in a line-up.
There were limitations with the research, including that all witnesses were aware they were not viewing a crime and that the consequences of making a correct or an incorrect identification decision were minimal to all involved. In addition, Gesmina, the operator, was relatively inexperienced in using the EFIT6 system. An experienced police operator might have organised the procedure somewhat differently. Finally, sometimes witnesses create composites after longer delays than those reported here; and memory for faces, regardless of ability, is adversely impacted by delay (e.g., Davis, Bretfelean, Belanova, & Thompson, 2020).
Once again, we would like to thank everyone who contributed to this research. We would also like to thank VisionMetric for the use of the EFIT6 system. Note – Josh Davis is currently engaged in an Innovate UK funded collaboration with VisionMetric called E2ID (details here: https://gtr.ukri.org/projects?ref=105042) and here
Davis, J. P., Bretfelean, D., Belanova, E., & Thompson, T. (2020). Super-recognisers: face recognition performance after variable delay intervals. Applied Cognitive Psychology, DOI:10.1002/acp.3712 (read pre-print here)
Davis, J. P., Gibson, S, & Solomon, C. (2014). The positive influence of creating a holistic facial composite on video lineup identification. Applied Cognitive Psychology, 28, 634–639. DOI: 10.1002/acp.3045 (read here)
Davis, J. P., Gibson, S., & Solomon, C. (2017). Holistic facial composite systems: Implementation and evaluation, in M. Bindemann and A. M. Megreya (Eds.), Face Processing: Systems, Disorders and Cultural Differences. Nova Science Publishers. (read here)
Davis, J. P., Maigut, A. C., Jolliffe, D., Gibson, S, & Solomon, C. (2015). Holistic facial composite creation and subsequent video line-up eyewitness identification paradigm. Journal of Visualized Experiments, e53298. DOI:10.3791/53298 (read here)
Davis, J. P., Thorniley, S., Gibson, S, & Solomon, C. (2016). Holistic facial composite construction and subsequent lineup identification accuracy: Comparing adults and children. Journal of Psychology: Interdisciplinary and Applied, 150, 102-118. doi: 10.1080/00223980.2015.1009867 (read here)
Johnston, R. A., & Edmonds, A. J. (2009). Familiar and unfamiliar face recognition: a review. Memory, 17(5), 577-596. doi: 10.1080/09658210902976969.
Pike, G. E., Brace, N. A., Turner, J., & Vredeveldt, A. (2019a). The effect of facial composite construction on eyewitness identification accuracy in an ecologically valid paradigm. Criminal Justice and Behavior, 46(2), 319-336. https://doi.org/10.1177/0093854818811376 (read here)
Pike, G., Brace, N., Turner, J., Ness, H., & Vredeveldt, A. (2019b). Advances in facial composite technology, utilizing holistic construction, do not lead to an increase in eyewitness misidentifications compared to older feature-based systems. Frontiers in Psychology, 10, 1962. https://doi.org/10.3389/fpsyg.2019.01962 (read here)
Russell, R., Duchaine, B., & Nakayama, K. (2009). Super-recognizers: People with extraordinary face recognition ability. Psychonomic Bulletin & Review, 16, 252-257. doi:10.3758/PBR.16.2.252 (read here)
Wells, G. L., Charman, S. D., & Olson, E. A. (2005). Building face composites can harm lineup identification performance. Journal of Experimental Psychology: Applied, 11, 147–156. doi:10.1023/a:1025750605807 (read here)