What is the scope for video technicians or analysts to offer their opinion on the identification of suspects observed in CCTV and other images based solely on their observation of such images and their subsequent comparison to known images of suspects? Should opinions on identification be allowed at all? If so, should there be limitations on the opinions that can be offered in these circumstances? This article focuses on a recent UK case but will address this topic more broadly.
In R. v. Garrington [2025] EWCA Crim 52 (Court of Appeal (Criminal Division)), the appellant (A) was convicted of murder and sentenced to life imprisonment. The prosecution established that while A was not directly involved in the murder and other crimes of violence that were committed during a home invasion stemming from an unpaid drug debt, A played a vital supporting role. A key component of the prosecution’s case against A was identification evidence stemming from CCTV video images and the comparison of those images to custody photos of A.
Trial Evidence
As part of its case, the prosecution relied upon an image-based sequence of events and a video timeline prepared by a police video technician/analyst (his specific title was not indicated in the ruling and for ease of reference I will refer to him as the “analyst”). The analyst had spent more than 700 hours reviewing CCTV video images from over 300 cameras. From the CCTV images, he identified several individuals that he believed were involved in relevant events before and after the murder at the residence was committed. He then compared images of those questioned individuals to known custody images of the defendants. One of the suspects was labelled P3 and as a result of the comparison work conducted by the analyst, A was “identified” as P3. A denied that he was P3.
P3 was observed in CCTV images in a relevant outdoor location prior to the crimes being committed. The analyst also stated that he could identify P3 throughout various other segments of the CCTV images. The analyst compared those questioned P3 images with custody images of A that were recorded when he was arrested three months later. The CCTV images of P3 were taken from a higher position than the custody images, which were front facial images, and thus the views of the person(s) of interest were not directly comparable. The questioned images showed a light skinned older male whose general appearance, including his hair colour, style, and length, and “heavy age lines on the face”, were reported by the analyst to be “consistent” with the appearance of A. The neck of P3 was not visible in the CCTV images as it was covered by clothing and so no comparison could be made regarding the tattoos that were visible on A’s neck in the custody photos. The analyst reported that he observed no inconsistent features that would cause him to exclude P3 from being A.
In cross examination, the analyst said, “I am satisfied that there is sufficient there for me to be able to say that there is (sic) consistent features that I can’t exclude him from being one and the same.” The analyst acknowledged that the quality of the CCTV images precluded him from being able to conclusively identify A as P3, though according to the appellate judgment he had said that the jury could conclude that A was P3. There is no explanation for that apparent contradiction in the judgment.
There was other evidence to link A to some of the locations shown in the CCTV images but the analyst’s opinion evidence about the linkage between A and P3 was central to unifying those disparate images and events. The trial judge was satisfied that based upon the entirety of the prosecution’s evidence, including the analyst’s evidence, a reasonable jury, properly instructed, could conclude that A was P3, though they were certainly not bound to so conclude. The jury convicted A of murder and other crimes.
The Appeal
On appeal, A argued that the analyst’s opinion of consistency did not amount to positive identification and could not therefore satisfy the criminal standard of proof. Indeed, the case against A substantially depended upon the identification of A as P3. The Court of Appeal stated that the prosecution was not required to establish that the analyst’s evidence constituted proof on the criminal standard. That is not the law. Rather, the issue is whether the prosecution’s evidence, taken as a whole, entitled a reasonable jury to find A guilty of one or more counts. The analyst’s evidence was merely one of the factors to consider. This is a correct statement of the law – the prosecution is not required to prove individual pieces of evidence beyond a reasonable doubt.
In Attorney General’s Reference No. 2 of 2002 [2002] EWCA Crim 2373, the Court of Appeal set out several instances in which photographic evidence could be used as evidence of identification. One of those was:
(iii) where a witness who does not know the defendant spends substantial time viewing and analysing photographic images from the scene, thereby acquiring special knowledge which the jury does not have, he can give evidence of identification based on a comparison between those images and a reasonably contemporary photograph of the defendant, provided that the images and the photograph are available to the jury (Clare and Peach); …
Applying this longstanding approach to identification evidence, the Court of Appeal in the present case noted that:
54. There was no dispute that by virtue of the substantial time he had spent in looking at and assessing the CCTV footage together with contemporary photographs of the defendants, Mr. Jenkins had gained special knowledge and skills to be able to (a) explain what the CCTV showed to have taken place, (b) distinguish different profiles of persons and (c) identify those persons. Although Mr. Hannam suggested to Mr. Jenkins, and he accepted, that the quality of the CCTV footage for the evening of 25th February 2022 was insufficient to enable him to identify the appellant as P3, that was not suggested for the footage at around 1pm that day.
…
56. Mr. Jenkins’ opinion was that there were a number of points of consistency between the images of P3 and of the appellant and no inconsistency between them. In our judgment, he put forward those points of consistency as positive, and not simply neutral, indicators of the appellant’s identification. The defence did not suggest any contra-indications. Although the witness’ report did not positively say that he could identify the appellant as P3, on a fair reading, that was the overall effect of his evidence. That was the purpose of it being given in the trial. No other purpose has been suggested.
56. No authority was cited to us to the effect that an opinion of the kind given by Mr. Jenkins should not be treated as identification evidence and admissible in accordance with Attorney General’s Reference No 2 of 2002. As a matter of principle we see no justification for taking that approach.
The Court noted that the issue of identification was for the jury to assess. This was not a case where the witness said that he was sure that A was P3. Even if he had been so conclusive, it would still be a matter for the jury to decide. The Court was satisfied that the prosecution had adduced sufficient evidence for the jury to consider. It was therefore appropriate for the trial judge to have left the case with the jury and the conviction was found to be safe.
Commentary
This case piqued my interest because it illustrates what I have long viewed as an incorrect approach to video identification in English law. Based on the Attorney General’s Reference case, as applied in the present case (Garrington), the courts expressly authorize a person to express an opinion on identification based solely on time spent conducting analysis of the images without any corresponding requirement that the witness have any training or expertise in image analysis, the forensic process of image comparison, or in facial recognition or body morphology. Nonetheless, this has been an accepted position in English law for some time. I believe that this approach is incorrect and may lead to juries reaching conclusions on identification premised on a misunderstanding of the evidence presented by the prosecution. There are three issues on this topic that merit some discussion.
First, the Attorney General’s Reference approval of “evidence of identification based on a comparison” is intended for a witness who does not know the defendant and “spends substantial time viewing and analysing photographic images from the scene”. It imposes no requirement that the person doing the comparison have any technical image knowledge or skill, nor any training in the comparison process. This paves the way for an unqualified witness to make observations that may not be valid. For example, a witness may identify what appears to be a scar or mark on a suspect’s face when in fact what is being identified is a compression artifact. Similarly, a witness may rely upon a predictive or bi-directional frame for comparison purposes rather than an intra-frame image captured by the camera. Without a requirement for any technical knowledge, the comparison opinion that is offered may be unwittingly erroneous. The witness may also rely on an image that hasn’t been properly adjusted for the correct aspect ratio. Further, image comparison is a forensic process that must be conducted in a certain way to mitigate bias and other factors that can impact the validity of opinion evidence.
This type of evidence is typically led by the prosecution and therefore the only way to combat this absence of a requirement for technical knowledge in current practice is during cross examination. Herein lies the further problem. Defence counsel are not trained in video forensics nor are they expected to be so trained. They would therefore need to have the services of an imagery expert to assist them and it is unlikely that Legal Aid would routinely provide funding for such an expert. That leaves defence counsel in the unenviable position of having a potential issue to raise but no effective way to do so.
Second, the comparison of questioned items to known items is a forensic process with a scientific basis. We see comparisons being conducted routinely with fingerprints, bullets and shell casings, tool marks, footwear impressions, tire marks, handwriting, etc. Image comparison is simply another application of this longstanding scientific process. The problem is that, with the exception of images, these comparisons would be undertaken by experts who would follow established forensic protocols and would testify in the capacity of an expert witness. Their work would be subjected to peer review and quality assurance measures. The Attorney General’s Reference effectively says that no such safeguards are needed when it comes to image comparison. Respectfully, this misses the mark and can result in evidence being placed before a jury that may result in misinterpretation and misapplication. Worse, it may result in a miscarriage of justice.
The Garrington case illustrates the misinterpretation and misapplication of the analyst’s evidence quite well. In this case, the analyst noted that P3 appeared to be a light skinned older male and commented on his general appearance, including his hair colour, style, and length, and heavy age lines on his face. The analyst stated that those observations were consistent with those that he made of A. There was no evidence of any discussion of height, weight, build or any other features that might be useful for comparison purposes, nor was there any discussion of class, unique, or identifying characteristics. The analyst opined that these noted features were “consistent” with the appearance of A in his custody photo where all that was visible was A’s face. There was no ability to offer even a general commentary on apparent height, weight or build as none of that information was available to the analyst. To be clear, the analyst did not say that A was P3, but he did say that the jury could conclude that A was P3 based on these consistencies.
Consistency is a very problematic word in this context. A literal reading shows that the analyst is merely saying that he noted features in P3 and A that are similar. On its face, that is not objectionable. The problem lies with what is not said by the analyst and what the jury believes that they hear. First, what the analyst did not apparently say is that the similarities that he pointed out are likely consistent with many other people (likely thousands of people in the UK). That is a very important qualification and puts the consistency comment in a much different and diffused light. Instead, he apparently used a sample size of a few people, namely the identified suspects. Second, the jury is unlikely to interpret this consistency in the non-definitive way it is being used. Rather, they are more likely to consider the consistency to be an identification. That is a natural and dangerous interpretation. The Court of Appeal in Garrington observed this analytical quagmire as well. A reminder – the Court said: “Although the witness’ report did not positively say that he could identify the appellant as P3, on a fair reading, that was the overall effect of his evidence. That was the purpose of it being given in the trial. No other purpose has been suggested.” By not qualifying the evidence to state that the consistencies noted would also apply to many other people, the word “consistent” takes on a stronger meaning than is warranted.
The third issue that arises in this situation is that a lay witness (the analyst), who was not qualified as an imagery expert, is permitted to engage in a forensic process and give evidence that a true expert would not be permitted to do. The Forensic Science Regulator (UK) in FSR Regulatory Notice 01/2019 (Image Enhancement and Image Comparison: Provision of Opinion) directed that image analysis experts are not permitted to express an evaluative opinion on the comparison they have undertaken unless they are also experts in the subject matter being compared. For example, an image analysis expert would not be permitted to give an opinion as to whether two vehicles were the same unless the expert was also an expert in that type of vehicle. Though there are diverging views as to whether this position is entirely sound, it nonetheless has merit. While an expert can validly point out similar and dissimilar features within image content, the final evaluative opinion by an expert who is not an expert in the subject matter being compared may be of little forensic value but may be given greater weight than it should be by the jury. What is paradoxical is that in Garrington (and cases like it), a lay witness was entitled to do what an expert witness would not be permitted to do. The safeguards that were put in place by the FSR to prevent this problem are not applicable as long as the witness who gives the evidence is not an expert witness. This is very problematic.
Recommendations
There is a viable solution to this problem that will ensure that the court receives evidence that meets minimums standards and also provides safeguards for the trier of fact. Addressing this problem requires that there be requirements for competence and restrictions on the type of evidence that is provided. As to requirements for competence, when a lay witness is engaged in the analysis and comparison of images, as occurred in Garrington, there must be minimum prescribed standards for training and proven competence in image processing and analysis. The gateway for providing evidence of identification based on image analysis cannot simply be time spent examining the images. An untrained, unqualified person is likely to offer an opinion that may be incomplete, misleading, or incorrect. As to restricting the type of evidence offered, a lay witness should not be permitted to offer an evaluative opinion (e.g., consistency) but rather should only be permitted to note observations. This type of evidence is still of value to the trier of fact but allows them to assess the significance of the observations. This would also be in keeping with the spirit of motivation behind the FSR directive for imagery experts.
The goal of these recommendations is not to remove the ability of lay witnesses to make observations about images but rather to ensure that such witnesses are qualified to do so. The trier of fact should get the best quality assistance available and if that is coming from lay witnesses, such witnesses should be proven to be competent to provide it. To avoid errors in interpretation and the conflation of similarities with identification, confining lay witnesses to observations only and excluding evaluations, the trier of fact will receive more properly measured evidence.