Enhancing identification of witness composites

Perception, Cognition and Neuroscience (PCN) Research Group

This programme of research is investigating procedures for combining and enhancing the effectiveness of composites from eyewitnesses.

Facial composites have an important role to play in modern policing. They are constructed by witnesses and victims of crime when the identity of a perpetrator is unknown. There are various systems available to create a likeness of the face—feature, sketch and holistic (Frowd et al., 2005)—with considerable research dedicated to methods that create the most identifiable likeness (e.g., Fodarella et al., 2021; Frowd et al., 2012b, 2015; Richardson et al., 2020; Skelton et al., 2019; see Frowd, 2021 for a review).

Research also considers facilitating identification of composites once the face has been created by an eyewitness. There are a range of procedures for facial enhancement, such as by presenting the face as a dynamic caricature (Frowd et al., 2007, 2012a), as a stretched image (Frowd et al., 2013, 2014) or by concealing potentially inaccurate features such as using a hat and sunglasses (Brown et al., 2019).

The current project considers these techniques, as part of replication, but also to investigate which methods could be combined (e.g., dynamic caricature and stretch, as the underlying mechanism for each method is different). The research also investigates the potential benefit gained from combining composites from different witnesses, to create an average (or ‘morphed’) composite (Frowd et al., 2012a; McIntyre et al., 2016; Valentine et al., 2010).

An example of varying appearances within facial composites.
An example of varying appearances within facial composites.

Primary goals and objectives

  • Enhancing composites from witnesses and victims of crime
  • Combining composites from different eyewitnesses
  • Best practice for achieving identification


The project has provided valuable insight into the various post-production techniques available for policing as part of achieving best evidence.

Related research projects