Conference Venue: Holiday Inn, Stratford-upon-Avon British Psychological Society
From: 14 Apr 2010 To: 16 Apr 2010
Symposia/Symposium
Facial identification in an applied setting
Allan McNeill
Glasgow Caledonian University
Establishing identity via a face or facial image is
commonplace. For example, within the criminal justice system it is vital to
establish that the suspect in the dock is the person who committed the crime
and very often this is achieved by comparing the face or facial image of the
accused with that of the culprit. Furthermore, this task is performed routinely
by: (1) police officers, attempting to match a suspect to a culprit; (2)
passport officers, controlling entry at national borders; and (3) surveillance
experts, when tracking suspects. However, it is now well known that
establishing identity by this method is highly error prone. In this symposium
six eminent UK based researchers in this field will present their most recent
findings in this area from an applied perspective, with the focus being on
improving the accuracy of facial identification in real world settings.
Paper 1: When is memory for faces contagious?
Charity Brown & Kate Muir, University of Leeds
Objectives: Misinformation introduced by a co-witness can
impair eyewitness reports. The present experiment examines the influence of
co-witness misinformation on the quality of a participant’s subsequent face
description.
Design: Exposure to co-witness misinformation (discussing
the same face vs. different faces) and the participant’s recall criterion in a
subsequent face description task (standard vs. forced description instructions)
were manipulated in a between-subjects design. Post-hoc analysis of co-witness
discussions further categorised participants as adopting a dominant or
non-dominant role in the discussion.
Methods: 40 of participants viewed a picture of
the same face, or two different faces, but were told they had seen the same
face. They then discussed the face’s appearance. Subsequently, participants
individually described their target face. Half received ‘standard’ description
instructions, permitting participants to establish their own recall criterion,
and half received ‘forced’ instructions, designed to lower participants recall
criterion and encourage erroneous recall.
Results: Participants discussing different faces were
exposed to a greater amount of misinformation than participants discussing the
same face; however, subsequent individual face descriptions contained
equivalent amounts of misinformation across both conditions. Forced compared to
standard instructions elicited more information overall, but equivalent amounts
of misinformation. Non-dominant discussants included significantly more
misinformation in their individual face descriptions than dominant discussants.
Conclusions: Increased susceptibility to misinformation about
a face is not inevitable even when exposure to misinformation is prevalent and
a lenient retrieval criterion is adopted. Social factors such as the presence
of a dominant discussant may play an important role.
Paper 2: Identification on the street: A survey of police
use of street identification and other identification procedures
J.P. Davis, University of Greenwich, T. Valentine,
Goldsmiths University of London, A. Memon, Royal Holloway, University of London
& A. Roberts, University of Warwick
Objectives: Following a crime report, the police in
England and Wales may use a street identification (showup) to obtain evidence
when there is insufficient evidence to arrest. This may involve encouraging a
witness to view an individual, or to conduct an area search of the vicinity.
The aim of this research was to survey police use of street identifications,
comparing outcomes with alternative visual procedures and to extrapolate
national estimates of use.
Design: Case data involving identification procedures were
obtained from four police forces. These included street identifications,
familiar suspect identifications, CCTV evidence, ‘mug shot’ images, facial
composites and video lineups.
Methods: Three types of data were collected. Firstly,
diary studies of robbery squad officers provided details of 701 cases involving
an identification procedure and their outcomes. Secondly, data of 37 cases in
which positive street identifications were followed by a video lineup of the
same suspect. Finally, data of 80 cases in which an area search was conducted
with or without a witness present in a police vehicle.
Results: A street identification was attempted in
approximately 25 per cent of robberies and was the most frequently used
identification procedure. Approximately, 12 per cent resulted in a suspect
identification. The majority (66 per cent) were later charged or cautioned,
particularly if a witness identified the same suspect in a subsequent video
lineup.
Conclusions: The police rely on street identification in
the investigation of street robbery. Despite a modest success rate, it is more
likely than any alternative procedure to result in a suspect being charged.
Paper 3: The psychology of face construction: giving
evolution a helping hand
Charlie Frowd, University of Central Lancashire; Melanie
Pitchford, University of Lancaster; Vicki Bruce, Newcastle University; Sam
Jackson, University of Central Lancashire; Gemma Hepton, University of Central
Lancashire; Maria Greenall, University of Central Lancashire; Alex H. McIntyre,
University of Stirling; Peter J.B. Hancock, University of Stirling
Objectives: To evaluate improvements to a
recognition-based facial composite system using face construction procedures
which mirror those of policework.
Design: Eyewitnesses traditionally construct a composite
of an offender’s face by the selection of individual facial features: hair,
eyes, nose, mouth, etc. Research suggests, however, that this method generally
produces unrecognisable images when realistic procedures are followed. We have
been developing a new system called EvoFIT that works by the repeated selection
and breeding of complete faces. Here, two techniques were investigated with the
potential of improving system performance. The first blurred the external parts
of the face, to help a user focus on the important central region. The second
manipulated an evolved face using a set of psychologically-useful scales, to
allow a user’s face recognition ability to improve the overall likeness; these
so-called ‘holistic’ tools change the perceived age, masculinity, honesty, etc.
Methods: Participants looked at an unfamiliar face and
two days later constructed a single composite. They used either EvoFIT, with or
without blurring and holistic tools, or a typical ‘feature’ system. Further
participants attempted to name the composites.
Results: A benefit was found for each technique, and
performance was maximal when both were used together: EvoFIT composites were
correctly named 25 per cent on average compared to 5 per cent for composites
constructed using the ‘feature’ system.
Conclusions: It is now possible to produce a fairly
recognisable image from a fairly weak memory of a face, the norm for real
witnesses. A plausible model to account for the enhancements is discussed.
Paper 4: An exploration of visual behaviour in eyewitness
identification tests
Heather D. Flowe, University of Leicester, School of
Psychology
Objectives: The role of internal (eyes, nose and mouth)
and external features (hair-line, cheek and jawline) in face recognition was
examined in eyewitness identification tests using eye tracking.
Design: Paricipants attempted to identify previously
studied faces from either a simultaneous (test faces presented in an array),
sequential (test faces presented one at a time), or showup (a suspect face
alone was presented) test.
Methods: A total of 36 undergraduates participated for
course credit. All had normal or corrected-to-normal visual acuity. Feature
dwell time was determined for the eyes, nose, mouth, and face regions by summing
up the total length of time spent fixating within the respective region during
a test trial. Dwell time data were square root transformed before submitting
them to inferential statistical analysis A feature distribution score was also
computed for every feature region by dividing feature dwell time for the
respective region by total face dwell time. The dependent measures were
analysed using the general linear model.
Results: Results indicated that foils were analysed for a
shorter period of time in the simultaneous compared to the sequential
condition, whereas a positively identified face was analysed for a comparable
period of time across lineup procedures. In simultaneous lineups and showups, a
greater proportion of time was spent analysing internal features of the test
faces compared to sequential lineups.
Conclusions: Different decision processes across
eyewitness identification tests are inferred based on the results.
Paper 5: Matching to CCTV: The advantage of 3D imagery.
Peter J.B. Hancock & Natasha Tetlow, Department of
Psychology, University of Stirling
Matching different images of unfamiliar faces is
surprisingly hard, with success rates of only around 70 per cent even in
seemingly perfect conditions, using high quality video stills and photographs.
We investigated the task of matching a photograph to a video clip, taken in a
crowded public location, with the task being to spot the target. The photograph
was available for inspection for 30 seconds and then remained present on an
adjacent screen during the video presentation. In Experiment 1 we compared a
normal photograph with an interactive 3D image; in experiment 2 we replaced the
single photograph with an array of stills at different orientations, and added
a target absent condition for each target. In both experiments, the interactive
3D display resulted in a significant improvement in matching performance.
Paper 6: The Lockerbie bomb: An evaluation of the
eyewitness identification of Abdelbaset al Megrahi
Tim Valentine, Goldsmiths University of London.
Purpose: To evaluate the eyewitness evidence in the case,
with respect to the research literature on eyewitness memory.
Background: Eyewitness identification provided key
evidence in the conviction of Abdelbaset al Megrahi for the Lockerbie bomb.
Maltese shopkeeper Tony Gauci identified him as a man who purchased items of
clothing from his shop. The prosecution case was that clothing purchased by Mr
al Megrahi was traced to a suitcase which contained the bomb that destroyed Pan
Am flight 103.
Methods: 20 statements by Tony Gauci, together with
supporting documents, press reports and statements by investigating officers
were studied.
Conclusions: The effects on human memory of the following
factors are relevant to an evaluation of the evidence: Delay, repeated
questioning, leading information, repeated identification procedures, and
feedback on eyewitness identification. It is not possible to conclude whether
the identification was correct or mistaken. It is concluded that circumstances
and procedures which have been shown by research to be associated with in
increased risk of mistaken eyewitness identification were present in the case.
