Witnesses to and victims of serious crime are often asked by the police to create a visual likeness of an offender using a computerised composite system. The witness provides a description and selects from a large database of examples the facial features that best match their memory of the offender. Good quality facial composites can provide investigative leads. However, laboratory research shows that participants generally produce composites bearing a poor resemblance to the target person, with composites created using traditional UK systems identified only about 20 per cent of the time under favourable conditions. Two simple techniques have been developed by the co-investigator to improve composite quality: introducing a 30 minute delay between the face description and composite construction; asking the witness to attribute personality characteristics to the target's face (eg, masculinity, pleasantness). This project will evaluate the effectiveness of these techniques in improving composite quality with relation to two variables of importance to real-world forensic settings: the delay between seeing a target face and constructing a composite; whether the witness knows (or does not know) to attend to the target's face. The results will indicate when the police should use these techniques with witnesses to improve composite quality.