Research
My research focuses on the scientific validity of facial identification methods and biometric systems, particularly how they perform in real-world conditions.
This work examines how decisions are made using these systems, where results become unreliable, and what this means for their use in practice.
Research themes
Validity of facial identification methods
Human and algorithmic decision-making
Effects of real-world variability (image quality, ageing, pose)
Evaluation and interpretation of evidence
Impact
My research underpins my advisory and training work, providing an evidence base for evaluating systems, improving decision-making, and understanding risk.
Much of my work focuses on identifying where methods appear valid under controlled conditions, but fail when applied in real-world settings.
My research has been published in peer-reviewed journals in forensic science, psychology, and biometrics, and has contributed to international guidance and practice.
Publications
-
This research examined masked face identification, showing that both examiner teams and algorithms outperform individuals, and that combining these approaches can improve performance in practice.
The study tested forensic face examiners, teams of examiners and state of the art algorithms in the comparison of masked to unmasked faces, in light of changing behaviours and increased mask wearing during the COVID-19 pandemic. Examiner teams and algorithms provided the highest levels of performance on the task highlighting the potential benefits of combining these two types of expertise in forensic contexts.
Published: 2025
-
This research examined the impact of ageing on facial comparison, showing that while examiners outperform novices in identification tasks, their ability to estimate age is limited, highlighting boundaries of expertise.
The first study to empirically test the effects of facial ageing on the performance of facial image comparison. The data revealed that individual forensic facial examiners were still more accurate than novices even with a large age gap between the compared faces. Teams of examiners further outperformed individual examiners. However, examiners were no better than novices at estimating the age of a face, highlighting limitations in the expertise of forensic face examiners.
Published: 2024
-
This research compared different forms of expertise in facial identification, showing that humans and algorithms make distinct types of errors and that combining complementary strengths can improve overall performance.
A landmark study comparing different types of expertise in facial identification. Innately high performing super recognisers, trained forensic facial examiners and artificial deep neural networks were compared across a range of facial comparison tests. The results highlighted stark differences in perception and decision making between the different types of expertise, and in particular the fact that although all groups made errors, they made different types of errors. The study recommended that certain types of expertise are better suited to different roles and combining expertise from different sources provides an additive performance gain.
Published: 2023
-
This research used eye-tracking to examine how forensic facial examiners view faces, showing that performance differences are not explained by how faces are visually examined, but by how information is interpreted.
Eye-tracking technology was used to analyse the viewing strategies of forensic facial examiners, fingerprint experts and novices on a challenging facial comparison task. The results revealed markedly similar viewing strategies across the three groups, indicating the performance advantage of forensic facial examiners is not derived from qualitative differences in how examiners view faces compared to untrained people.
Published: 2022
-
This research examined international training practices in forensic facial comparison, identifying significant variability and limited use of evidence-based approaches in the development of expertise.
A survey of 24 international agencies that conduct forensic facial examiner training revealed that in general forensic face examiners are trained for long periods of time and that one-to-one mentoring of examiners was commonplace. However, the survey also found variability in the types of topics covered in training and low inclusion of evidence-based training practices.
Published: 2021
-
This chapter provides a critical overview of forensic facial comparison practice in the UK, examining its development, limitations, and the challenges raised by casework over the past 30 years.
Published in Forensic Face Matching: Research & Practice, this chapter provides an overview of forensic facial comparison in the UK in the past 30 years, including past and current criticisms of the field and case studies that have challenged and tested facial comparison procedures.
Published: 2021
-
This research identified gaps in the empirical validation of training, raising questions about how expertise is developed and assessed in practice.
Despite being commercially available, prior to this publication, short facial identification training courses had not been test and empirically validated as to their effectiveness. This study tested both online and in person facial comparison training courses ranging in duration from one hour to three days. The results found no improvements for courses less than three days, and only limited inconsistent improvements for the in person three day course. The study recommended evidenced-based practices to improve facial comparison performance.
Published: 2019
-
This work highlights the importance of aligning selection and training of personnel with real-world tasks, rather than relying on assumptions about ability.
Super recognisers are individuals who are purported to have innate superior ability in identifying faces and there has been growing interest in using these high performing individuals in policing and other security applications. This commentary recommended that tests used to recruit super recognisers must reflect the role and tasks they will undertake, and that the design of such tests requires collaboration and cooperation between researchers and applied users.
Published: 2019
-
These findings demonstrate that photo anthropometry is not reliable for identification in practice, and influenced international guidance advising against its use.
Photo anthropometry is a quantitative technique that has been used in forensic facial image comparison, based on the ratio of distances between facial features. This study tested the technique on images of the same person taken at various resolutions of CCTV and different camera angles. The results found that variations in pixel resolution, camera angle and locating facial landmarks introduced greater variation into the measurements of the same persons face than were found between different persons. This research influenced international good practice in recommending that photo anthropometry not be used in forensic facial image comparison
Year published: 2011