Super-Recognisers: Reliable or a Recognisable Risk?
The Metropolitan Police’s flagship ‘Super-Recogniser Unit’ (SRU) – a group of ‘super-recogniser’ (SR) officers who have unusually high abilities to see, remember and identify faces – has recently attracted much press coverage. These officers are deployed to large public events such as Notting Hill Carnival and music concerts to identify known offenders who pose a risk to event security. Additionally, super-recognisers trawl through databases of CCTV footage from unsolved crimes to identify suspects and potentially link together offences which may have been committed by the same person. The SRU has been lauded as such an exemplary success that other police forces, both inside the UK and internationally, are consulting the Metropolitan Police on how to replicate the SRU’s functions and results.
When a new means for crime detection becomes available there is often an understandable urge to bring such methods in to use. However, proper care must be taken to ensure that a new form of evidence is reliable before its use becomes systemic. With the potential for the Metropolitan Police’s SRU being replicated in other police forces, there is a need to ensure that there are universally followed guidelines and methodologies for super-recognisers and the use of evidence which they produce.
The Science
The science behind the abilities of super-recognisers is still in its infancy. The reasons for, and extent of, the capacities for recognition that SRs possess are still unclear. It is currently estimated that between one and two percent of the population would fall in to the category of a ‘super-recogniser’. In a study on four super-recognisers in the Met, the officers took part in three tests assessing both familiar and unfamiliar face-matching abilities against a control group. In the first test the SR officers achieved correct results of 95.8% compared to 81.3% in the control; in the second test the SRs’ results were 90.3% compared to the control of 73.6%; the results of the third test were 93% correct by the SRs vs 73% correct by the control group (Robertson D. et al ‘Face Recognition by Metropolitan Police Super-Recognisers’ (2016) PLOS ONE). Similar results were achieved in testing by Anna Bobak’s team at Bournemouth University whereby participants were shown a real CCTV image of a person’s face, and asked to find the same person in a collection of 10 similar images. Those classified as a super-recogniser correctly identified the person 93 per cent of the time, compared to people of standard recognition capabilities who scored 73 per cent on average (Bobak et al ‘Super-recognisers in Action: Evidence from Face-matching and Face Memory Tasks’. (2016) ACP).
Within the classification of ‘super-recogniser’ the capabilities for recall of faces and individuals differs considerably. Variation in ability is not uniform across different forms of identification either. For example: super-recogniser A may achieve high scores in Test 1: recognising individuals on CCTV who are known to them; and low scores in Test 2: linking together two images of a person who is not known to them. Super-recogniser B may achieve opposite scores: low scores in Test 1 and high scores in Test 2. Both individuals are nonetheless labelled as super-recognisers.
The well-documented and well-studied ‘cross-race effect’ – the phenomenon whereby people much more easily and accurately recognise and identify members of their own race – has a further effect on the reliability of identification by SRs. A study at the University of Greenwich (Davis, J.P. et al (2013). ‘Facial identification from CCTV: investigating predictors of exceptional performance amongst police officers’. EAPL) has indicated that the cross-race effect may not hold true to the same extent for SRs as with the general population. The extent to which this is the case, however, is not entirely clear. The study was not able to clarify whether this effect is due to SRs not being affected like the general population by the cross-race effect or whether the police officers who were tested were more able to overcome this difficulty through more extensive experience of dealing with people across races than the general population. If the latter case is true and SRs who have more experience of operating in ethnically diverse communities are more able to overcome the cross-race effect than officers who operate in less diverse communities, then the problems caused by potential variability of capabilities between different SRs is exacerbated and the potential unreliability of their evidence becomes increasingly unclear.
The studies referred to above demonstrate that the abilities of SRs are not all the same. If the evidence of an SR is to be sufficiently reliable so as to be used at trial, the defence must be able to assess the identification by reference to the SR’s demonstrated capabilities. In order to achieve this, an SR officer ought to undergo examination to establish in which circumstances they excel and those which they do not; the results of such testing ought to be disclosed to the accused and their legal representatives to ensure that they are able to properly assess the strength of the evidence in question. Simply knowing that in general an SR may be able to make a correct identification in certain conditions more often than a non-SR would be insufficient for the purposes of scrutinising the reliability of an SR’s evidence. Such a system would be in line with the use of other forms of identification evidence; for example when disclosed to the jury, DNA or fingerprinting evidence is provided alongside scientific data about the margin for error in the testing procedures used. Providing a jury with the specifics of an individual SR’s competences would enable an SR’s identification evidence to be subject to sufficient scrutiny to establish its reliability or lack thereof.
Investigative Use of Super-Recognisers
Investigative bodies must be aware that, absent corroboration, evidence of a super-recogniser will be subject to significant scrutiny by the defence, the trial judge, and the jury. Beyond this, evidence of a super-recogniser may fail the basic test for reliability and be excluded from the trial.
The use of SRs can undoubtedly be of significant value to investigating authorities and has already proved to be so, with a quarter of the near 10,000 identifications made by the Met being done so by the six-person SRU. An SR may well be more capable of making an identification than a person of average faculties, yet nobody is infallible; safeguards exist to prevent potentially unreliable investigative methods creating unreliable and misleading evidence. Code D of the Police and Criminal Evidence Act 1984 (PACE) governs identification processes used by the police. Paragraphs 3.34-3.37 of Code D apply to the identification of individuals by police officers viewing video or still images and were intended to provide:
“[a] safeguard of measuring the recognition [by the police officer] against an objective standard of assessment. Only by such means can there be any assurance that the officer is not merely asserting that which he wishes, however subconsciously, to achieve, namely the recognition of the guilty participant.” (Smith (Dean) and others [2008] EWCA Crim 1342)
These rules provide a tried and tested framework for ensuring that where a police officer makes or attempts to make an identification of an individual by viewing film or images that they do so in isolation so as to prevent mistakes and avoid collusion with other officers and that they make a full contemporaneous record which, amongst other information, must include: whether the officer knew or was given information concerning the name or identity of any suspect; what the person has been told before the viewing about the offence, the person(s) depicted in the images or the offender and by whom; whether the viewing was alone or with others and if with others, the reason for it; whether or not the person claims to recognise any individual seen, and if they do: (i) the reason, (ii) the words of recognition, (iii) any expressions of doubt, (iv) what features of the image or the individual triggered the recognition.
Contemporaneous record-keeping is essential to ensure that the use of SRs is done so openly and fairly and is capable of providing an accused individual the fullest opportunity to defend the case against them. Without PACE-compliant records, and potentially without further corroborating evidence, the testimony of the SR may be vulnerable to challenge by the defence who could request that the judge exclude the evidence in question per s.78 PACE (R v Deakin [2012] EWCA Crim 2637).
Super-Recogniser Evidence at Trial
Should a case go to trial where there has been the involvement of an SR at the investigative stage (and assuming identification is a live and contested issue in the case), one of two situations will arise at trial. The first such situation is that during the investigation the SR will have been used as an intelligence gathering tool (rather than an evidence gathering one) and sufficient corroborative evidence will have been obtained to render the contributions of the SR unnecessary. The second situation is where the evidence of identification by an SR is required by virtue of a lack of sufficient corroborative evidence; it is this second situation which requires great care and close scrutiny.
The evidence of an SR, may seem compelling in the eyes of a jury, however, apparently compelling testimony is not necessarily accurate or reliable. In relation to identification evidence, the risk of a compelling but incorrect witness is mitigated by the guidance directed to judges as set out in the leading and long-standing case of Turnbull such that:
‘where the case against the accused depends wholly or substantially on the correctness of one or more identifications which the defence allege to be mistaken, the judge should warn the jury of the special need for caution before convicting the accused in reliance on the correctness of the identification. In addition he should instruct them as to the reason for the need for such a warning and should make some reference to the possibility that a mistaken witness can be a convincing one and that a number of such witnesses can all be mistaken’ (R v Turnbull (1976) 63 Cr App R 132)
What is important to stress from the Turnbull judgment is that the more central an identification is to the prosecution’s case, the more intense the scrutiny of the identifier’s evidence and its reliability must be. If an identification is made in poor conditions, the unreliability that such conditions would present can be lessened by corroborative evidence. This need not mean corroboration in the legalistic sense, but only evidence which can make the jury sure that there was not in fact a mistaken identification. As such, what must be encouraged is where an SR is used in the course of an investigation, further evidence is gathered to establish the accused individual’s participation in the alleged offence.
Despite the above average faculties of an SR, the jury must been given the same Turnbull direction by the judge as they would if the identifier in question were a lay witness with average facial recognition abilities. To do otherwise than to give such a direction, where applicable, would be to risk removing the potential for error in the identification from the minds of the jury and implicitly assert that SRs are not fallible as the general public.
Notwithstanding the SRs capabilities, where the quality of an identification is poor (for example by being a mere ‘fleeting glance’ or made in ‘difficult conditions’) the case ought to be withdrawn from the jury, per Turnbull, unless there is other evidence to support the correctness of the identification. This is submitted to be the case as an SR, while more able to make an identification than the average person on the street, is not capable of making a fleeting glance last longer or of removing the obscuring effect of darkness or heavy rain from their vision.
Further to the matter of the reliability an SR’s evidence is its presentation to the jury. Reference was made above to the need to ensure that the jury are aware of the fallibility and potential unreliability of any witness, including SRs; continuing in this vein, great care must be taken to ensure that the evidence of an SR does not trespass upon the role of the jury in deciding upon the matter of identification.
Where images used to identify the defendant are sufficiently clear to allow the jury to make a positive identification in court there will be no need for the evidence of the SR. There are then two situations where the jury can be invited to identify the defendant on the basis of images with the assistance of an SR’s evidence: (i) where the image is identified by a witness who knows the defendant sufficiently well to recognise him, or (ii) where the identification is based on the opinion evidence of an individual who did not know the defendant but has acquired specialist knowledge by spending substantial time analysing images from the scene, or is a facial mapper(Attorney-General’s Reference (No. 2 of 2002) [2003] 1 Cr.App.R 21, CA). The evidence of an SR will normally be able to be admitted through either of these gateways depending on the facts).
The risk of admission of an SR’s evidence through the latter gateway is that an SR will be able to provide opinion evidence on identification similar to that of a facial mapper without being subject to the same professional requirements or levels of scrutiny as a conventional expert. Where CCTV images are of a poor quality and an SR is able to make an identification that a lay witness is not, the witness will be offering an opinion on the identification of the individual in question without recourse to a scientifically verifiable methodology. Such a situation means that the jury is not able to scrutinise the evidence of an SR in such a situation as well as they would a facial mapper; the risk of this lack of scrutiny is that the evidence of the SR is accepted too readily and an SR is effectively usurping the jury’s role in determining the issue of identification. The minimum means to prevent such a risk is a judicial warning to the jury on the risk of mistaken identification, even by an SR, and which highlights the care needed to be taken before convicting on the basis of an identification which the jury is not sufficiently able to scrutinise themselves. Such a direction would be in keeping with the jurisprudence on warning the jury of the risk of convicting on the basis of comparing an image presented in court with the defendant in the dock (see R v Downey [1995] 1 Cr.App.R. 547 CA, R v Blenkinsop [1995] 1 Cr.App.R. 7, CA).
Conclusion
The Court of Appeal was founded in 1907 as a result of the Adolf Beck case – a miscarriage of justice which resulted from incorrect identifications by no less than fifteen witnesses. Much has been done to limit the mischief caused by identification evidence in the intervening century. As a form of evidence, however, identification evidence is still vulnerable to inaccuracy.
Super-recognisers can prove pivotal in making the all-important breakthrough in a case, but the benefits they bring to investigators must not be allowed to disguise their potential unreliability. Rigorous scrutiny is essential to prevent inaccurate or unreliable evidence from being put before a jury. Evidence of a super-recogniser ought to be subject to perhaps closer inspection and scrutiny due to the risks that are posed by any potential view of an identification by a super-recogniser being more authoritative and reliable than an identification by an average witness. Adherence to proper procedures and being ever-aware of SRs’ fallibility will go a long way in militating against the risks that this new evidence-gathering method poses.
This article was originally published in Criminal Law and Justice Weekly, and can be found here behind a paywall.