CCTV on trial: Matching video images with the defendant in the dock

The experiments reported in this paper investigated simultaneous identity matching of unfamiliar people physically present in person with moving video images typical of that captured by closed circuit television (CCTV). This simulates the decision faced by a jury in court when the identity of somebody caught on CCTV is disputed. Namely, ‘is the defendant in the dock the person depicted in video’? In Experiment 1, the videos depicted medium‐range views of a number of actor ‘culprits’. Experiment 2 used similar quality images taken a year previously, some of which showed the culprits in disguise. Experiment 3 utilised high‐quality close‐up video images. It was consistently found that in both culprit‐present and culprit‐absent videos and in optimal conditions, matching the identity of a person in video can be highly susceptible to error. Copyright © 2008 John Wiley & Sons, Ltd.     

The Effect of Image Pixelation on Unfamiliar‐Face Matching

Low‐resolution, pixelated images from closed‐circuit television (CCTV) can be used to compare the perpetrators of crime with high‐resolution photographs of potential suspects. The current study investigated the accuracy of person identification under these conditions, by comparing high‐resolution and pixelated photographs of unfamiliar faces in a series of matching tasks. Performance decreased gradually with different levels of pixelation and was close to chance with a horizontal image resolution of only 8 pixels per face (Experiment 1). Matching accuracy could be improved by reducing the size of pixelated faces (Experiment 2) or by varying the size of the to‐be‐compared‐with high‐resolution face image (Experiment 3). In addition, pixelation produced effects that appear to be separable from other factors that might affect matching performance, such as changes in face view (Experiment 4). These findings reaffirm that criminal identifications from CCTV must be treated with caution and provide some basic estimates for identification accuracy with different pixelation levels. This study also highlights potential methods for improving performance in this task. Copyright © 2013 John Wiley & Sons, Ltd.     

Improving Discrimination and Face Matching with Caricature

Identification of faces from photographs is a common security measure, but matching unfamiliar faces produces high rates of error. Caricatures of familiar people are highly identifiable because they exaggerate distinctive features. We investigated whether exaggerating unfamiliar faces through caricaturing could also improve face‐matching accuracy. In Experiment 1, face‐matching arrays were caricatured relative to an average by 30%, 50% and 70%. Correct rejection of the target‐absent arrays was improved at all levels. Accurate matches increased at 30%, but at 70%, the transformation was too extreme, and all of the arrays were more likely to be rejected. In Experiment 2, photographic identification (ID) images were caricatured by 30% and 50% and matched to life‐size photographs. Rejection of foils improved, but the ID of matching images was impaired. Modest levels of caricature may improve discrimination in unfamiliar face matching, but at stronger levels, a conservative response bias may inhibit accurate ID. Copyright © 2013 John Wiley & Sons, Ltd.     

Does a Judicial Warning Improve Defendant‐Culprit Matching?

During a trial, jurors may be asked to compare the defendant with video of the culprit. Yet matching unfamiliar people in this manner can be difficult and error‐prone. Judges may warn jurors about the dangers of identification from video images, but the effect of warnings is unknown. Here we report two studies that assessed the effect of warning instructions on defendant‐culprit matching. In Experiment 1, college students read either control, warning, or illustrated warning instructions before deciding if a defendant seen in one video matched the culprit in another video. Experiment 2 compared control and warning instructions in a sample of Amazon Mechanical Turk workers. Matching accuracy in both experiments was poor, and warnings did not consistently improve performance or make participants more cautious. These results do not support the supposition that warning instructions help jurors compare the defendant to images of the culprit.Copyright © 2016 John Wiley & Sons, Ltd.     

Matching Face Images Taken on the Same Day or Months Apart: the Limitations of Photo ID

It is well‐established that matching images of unfamiliar faces is rather error prone. However, there is an important mismatch between face matching in laboratory and realistic settings. All of the currently available face‐matching databases were designed to establish the baseline level of unfamiliar face perception. Therefore, target and test images for each face identity have been taken on the same day, minimizing within‐face variations. In realistic settings, on the other hand, faces do vary, even day to day. This study examined the proficiency of matching images of unfamiliar faces, which were taken on the same day or months apart. In two experiments, same‐day images were matched substantially more accurately and faster than different‐date photographs using the standard 1‐in‐10 and pairwise face‐matching tasks. This suggests that experimental studies on face matching underestimate its difficulty in real‐world situations. Photographs of unfamiliar faces seem to be unreliable proofs of identity, especially if the ID documents do not use very recent images of the holders. Copyright © 2013 John Wiley & Sons, Ltd.     

Checking ID‐cards for the sale of restricted goods: Age decisions bias face decisions

Matching unfamiliar faces is highly error‐prone, and most studies highlight the implications for real‐world ID‐checking. Here we study a particular instance of ID‐checking: proof of age for buying restricted goods such as alcohol. In this case, checkers must establish that an identity document is carried by its legitimate owner (i.e., that the ID photo matches the face of the bearer) and that the ID proves the bearer to be old enough to make the purchase. Across three experiments, using two common forms of photo‐ID (i.e., driving licenses, PASS+ cards) we show that observers produce very high error rates when age requirements are met, but faces mismatch. This bias away from detecting a face mismatch remained evident in experienced cashiers—though to a somewhat attenuated level. We discuss interactions between face matching and other tasks, and the practical consequences of a bias which favours those using photo‐ID with fraudulent intent.     

Biased lineup instructions and face identification from video images

Previous eyewitness memory research has shown that biased lineup instructions reduce identification accuracy, primarily by increasing false-positive identifications in target-absent lineups. Because some attempts at identification do not rely on a witness's memory of the perpetrator but instead involve matching photos to images on surveillance video, the authors investigated the effects of biased instructions on identification accuracy in a matching task. In Experiment 1, biased instructions did not affect the overall accuracy of participants who used video images as an identification aid, but nearly all correct decisions occurred with target-present photo spreads. Both biased and unbiased instructions resulted in high false-positive rates. In Experiment 2, which focused on video-photo matching accuracy with target-absent photo spreads, unbiased instructions led to more correct responses (i.e., fewer false positives). These findings suggest that investigators should not relax precautions against biased instructions when people attempt to match photos to an unfamiliar person recorded on video.     

Not looking yourself: The cost of self‐selecting photographs for identity verification

Photo‐identification is based on the premise that photographs are representative of facial appearance. However, previous studies show that ratings of likeness vary across different photographs of the same face, suggesting that some images capture identity better than others. Two experiments were designed to examine the relationship between likeness judgments and face matching accuracy. In Experiment 1, we compared unfamiliar face matching accuracy for self‐selected and other‐selected high‐likeness images. Surprisingly, images selected by previously unfamiliar viewers – after very limited exposure to a target face – were more accurately matched than self‐selected images chosen by the target identity themselves. Results also revealed extremely low inter‐rater agreement in ratings of likeness across participants, suggesting that perceptions of image resemblance are inherently unstable. In Experiment 2, we test whether the cost of self‐selection can be explained by this general disagreement in likeness judgments between individual raters. We find that averaging across rankings by multiple raters produces image selections that provide superior identification accuracy. However, benefit of other‐selection persisted for single raters, suggesting that inaccurate representations of self interfere with our ability to judge which images faithfully represent our current appearance.     

Performance of typical and superior face recognizers on a novel interactive face matching procedure

Unfamiliar simultaneous face matching is error prone. Reducing incorrect identification decisions will positively benefit forensic and security contexts. The absence of view-independent information in static images likely contributes to the difficulty of unfamiliar face matching. We tested whether a novel interactive viewing procedure that provides the user with 3D structural information as they rotate a facial image to different orientations would improve face matching accuracy. We tested the performance of ‘typical’ (Experiment 1) and ‘superior’ (Experiment 2) face recognizers, comparing their performance using high-quality (Experiment 3) and pixelated (Experiment 4) Facebook profile images. In each trial, participants responded whether two images featured the same person with one of these images being either a static face, a video providing orientation information, or an interactive image. Taken together, the results show that fluid orientation information and interactivity prompt shifts in criterion and support matching performance. Because typical and superior face recognizers both benefited from the structural information provided by the novel viewing procedures, our results point to qualitatively similar reliance on pictorial encoding in these groups. This also suggests that interactive viewing tools can be valuable in assisting face matching in high-performing practitioner groups.     

Infrequent identity mismatches are frequently undetected

The ability to quickly and accurately match faces to photographs bears critically on many domains, from controlling purchase of age-restricted goods to law enforcement and airport security. Despite its pervasiveness and importance, research has shown that face matching is surprisingly error prone. The majority of face-matching research is conducted under idealized conditions (e.g., using photographs of individuals taken on the same day) and with equal proportions of match and mismatch trials, a rate that is likely not observed in everyday face matching. In four experiments, we presented observers with photographs of faces taken an average of 1.5 years apart and tested whether face-matching performance is affected by the prevalence of identity mismatches, comparing conditions of low (10 %) and high (50 %) mismatch prevalence. Like the low-prevalence effect in visual search, we observed inflated miss rates under low-prevalence conditions. This effect persisted when participants were allowed to correct their initial responses (Experiment 2), when they had to verify every decision with a certainty judgment (Experiment 3) and when they were permitted “second looks” at face pairs (Experiment 4). These results suggest that, under realistic viewing conditions, the low-prevalence effect in face matching is a large, persistent source of errors.     

Facial Wipes don't Wash: Facial Image Comparison by Video Superimposition Reduces the Accuracy of Face Matching Decisions

In cases of disputed CCTV identification, expert testimony based on the results of analysis by facial image comparison may be presented to the Jury. However, many of the techniques lack empirical data to support their use. Using a within‐participants design, we compared the accuracy of face‐matching decisions when images were presented using a ‘facial wipe’ technique (where one image is superimposed on another and the display gradually ‘wipes’ between the two) with decisions based on static images. Experiment 1 used high‐quality image pairs; Experiment 2 used disguised target images; and Experiment 3 used degraded target images. Across all three experiments, rather than optimising performance, facial wipes reduced accuracy relative to static presentations. Further, there is evidence that video wipes increase false positives and therefore may increase the likelihood that images of two different people will be incorrectly judged to show the same individual.Copyright © 2016 John Wiley & Sons, Ltd.     

Orangutans (Pongo abelii) and a gorilla (Gorilla gorilla gorilla) match features in familiar and unfamiliar individuals

Great apes can perceive images as representative of corresponding real-life objects. Coupled with the potential advantages of identifying specific members of one’s species and mounting evidence for individual recognition in other non-humans, it seems likely that great apes would have the ability to identify conspecifics in photographs. The ability of four orangutans and a gorilla to match images of individuals of their own and a closely related but unfamiliar species was examined here for the first time. First, the subjects matched photographs of familiar conspecifics taken at various time points in a delayed matching-to-sample procedure (Experiment 1). Second, they matched different photographs of unfamiliar individuals of a different species (Experiment 2) at above chance levels. These results suggest that the subjects matched photographs by matching physical features, not necessarily by recognizing the identity of the individuals depicted. However, they also quickly learned to select photographs of familiar individuals when these photographs were paired with photographs of unfamiliar individuals of their own species (Experiment 3), and three subjects showed transfer to novel images of familiar and unfamiliar individuals. Thus, the findings support the idea that subjects attended to physical features to identify individuals that they could categorize on the basis of familiarity.     

Motivational Incentives Improve Unfamiliar Face Matching Accuracy

Unfamiliar face matching is the process of simultaneously verifying the equivalence of two faces. Accuracy within this task is crucial given its importance and prevalence in security domains. However, research has consistently demonstrated that unfamiliar face matching is error prone, suggesting the need for improvement techniques. Motivational incentives have been shown to reliably enhance performance on a wide array of cognitive tasks. Therefore, the current study examined the effect of motivational incentives on unfamiliar face matching. Experiment 1 showed increased accuracy in mismatch trials with the addition of a motivational incentive. The improvement was associated with an increase in reaction time and bias towards making a mismatch response. This bias, however, did not produce a decrease in accuracy on match trials. Experiment 2 replicated this result with a more ecologically valid methodology. The results suggest motivational incentives to be effective at improving performance on mismatch trials, arguably the most important aspect of real‐world unfamiliar face matching. Copyright © 2013 John Wiley & Sons, Ltd.     

Dominance of texture over shape in facial identity processing is modulated by individual abilities

For face recognition, observers utilize both shape and texture information. Here, we investigated the relative diagnosticity of shape and texture for delayed matching of familiar and unfamiliar faces (Experiment 1) and identifying familiar and newly learned faces (Experiment 2). Within each familiarity condition, pairs of 3D‐captured faces were morphed selectively in either shape or texture in 20% steps, holding the respective other dimension constant. We also assessed participants’ individual face‐processing skills via the Bielefelder Famous Faces Test (BFFT), the Glasgow Face Matching Test, and the Cambridge Face Memory Test (CFMT). Using multilevel model analyses, we examined probabilities of same versus different responses (Experiment 1) and of original identity versus other/unknown identity responses (Experiment 2). Overall, texture was more diagnostic than shape for both delayed matching and identification, particularly so for familiar faces. On top of these overall effects, above‐average BFFT performance was associated with enhanced utilization of texture in both experiments. Furthermore, above‐average CFMT performance coincided with slightly reduced texture dominance in the delayed matching task (Experiment 1) and stronger sensitivity to morph‐based changes overall, that is irrespective of morph type, in the face identification task (Experiment 2). Our findings (1) show the disproportionate importance of texture information for processing familiar face identity and (2) provide further evidence that familiar and unfamiliar face identity perception are mediated by different underlying processes.     

Super-Recognisers show an advantage for other race face identification

The accurate identification of an unfamiliar individual from a face photo is a critical factor in several applied situations (e.g., border control). Despite this, matching faces to photographic ID is highly prone to error. In lieu of effective training measures, which could reduce face matching errors, the selection of “super-recognisers” (SRs) provides the most promising route to combat misidentification or fraud. However, to date, super-recognition has been defined and tested using almost exclusively “own-race” face memory and matching tests. Here, across three studies, we test Caucasian participants' performance on own- and other-race face identification tasks (GFMT, MFMT, CFMT+, EFMT, CFMT-Chinese). Our findings show that compared to controls, high-performing typical recognisers (Studies 1 and 2) and SRs (Study 3) show superior performance on both the own- and other-race tests. These findings suggest that recruiting SRs in ethnically diverse applied settings could be advantageous.     

Feedback training for facial image comparison

People are typically poor at matching the identity of unfamiliar faces from photographs. This observation has broad implications for face matching in operational settings (e.g., border control). Here, we report significant improvements in face matching ability following feedback training. In Experiment 1, we show cumulative improvement in performance on a standard test of face matching ability when participants were provided with trial-by-trial feedback. More important, Experiment 2 shows that training benefits can generalize to novel, widely varying, unfamiliar face images for which no feedback is provided. The transfer effect specifically benefited participants who had performed poorly on an initial screening test. These findings are discussed in the context of existing literature on unfamiliar face matching and perceptual training. Given the reliability of the performance enhancement and its generalization to diverse image sets, we suggest that feedback training may be useful for face matching in occupational settings.     

Face Recognition in Poor-Quality Video; Evidence From Security Surveillance

Security surveillance systems often produce poor-quality video, and this may be problematic in gathering forensic evidence. We examined the ability of subjects to identify target people captured by a commercially available video security device. In Experiment 1, subjects personally familiar with the targets performed very well at identifying them, but subjects unfamiliar with the targets performed very poorly. Police officers with experience in forensic identification performed as poorly as other subjects unfamiliar with the targets. In Experiment 2, we asked how familiar subjects can perform so well. Using the same video device, we edited clips to obscure the head, body, or gait of the targets. Obscuring body or gait produced a small decrement in recognition performance. Obscuring the targets' heads had a dramatic effect on subjects' ability to recognize the targets. These results imply that subjects recognized the targets' faces, even in these poor-quality images.     

In the Dock: Chimeric Image Composites Reduce Identification Accuracy

The aim of presenting chimeric images (formed from opposing halves of a pair of same or different faces) in court settings is to optimise the accuracy of identification decisions based on CCTV evidence. The experiments reported here examined the utility of this technique. Experiment 1 examined the accuracy of face matching with vertically split, aligned chimeric images, misaligned hemi‐faces and full‐face images. Experiment 2 replicated the first experiment but replaced the misaligned images with opposing hemi‐faces separated by a gap. The final experiment used horizontally split faces. All three experiments showed that matching was less accurate with aligned chimeric images than with full‐face images. Furthermore, the pattern of responses obtained with chimeric images differed significantly from full‐face matching and misaligned/separated hemi‐face matching. Chimeric images produced a bias towards same responses even when the face halves were different. The results suggest caution in the use of chimeric images in court. Copyright © 2011 John Wiley & Sons, Ltd.     

Faces retain attention

In the present study, we investigated whether faces have an advantage in retaining attention over other stimulus categories. In three experiments, subjects were asked to focus on a central go/no-go signal before classifying a concurrently presented peripheral line target. In Experiment 1, the go/no-go signal could be superimposed on photographs of upright famous faces, matching inverted faces, or meaningful objects. Experiments 2 and 3 tested upright and inverted unfamiliar faces, printed names, and another class of meaningful objects in an identical design. A fourth experiment provided a replication of Experiment 1, but with a 1,000-msec stimulus onset asynchrony between the onset of the central face/nonface stimuli and the peripheral targets. In all the experiments, the presence of an upright face significantly delayed target response times, in comparison with each of the other stimulus categories. These results suggest a general attentional bias, so that it is particularly difficult to disengage processing resources from faces.     

The basis of the 3/4 view advantage in face recognition

Three-quarter views of faces promote better recognition memory for previously unfamiliar faces than do full-face views. This paper reports experiments which examine the possible basis of the effect, and, in particular, examine whether the effect reflects some ‘canonical’ role for the 3/4 view of a face. Experiment 1 showed no advantage of 3/4 views over full-face views when the task was to decide whether or not each of a series of faces was that of a highly familiar colleague. In Experiment 2 a sequential matching task was used, where subjects had to respond positively if both members of a pair of faces were of the same person. When the faces used were highly familiar to the subjects, there was no evidence of an advantage for a 3/4 view in the matching task. Three-quarter views and full-face views led to equivalent performance, though profiles produced decrements in performance. When the same faces were shown to subjects who were unfamiliar with the faces, 3/4 views did lead to increased speeds in same trials, compared with full-face, though profiles again proved difficult. Thus a 3/4 view advantage appeared only where the faces were unfamiliar, and the task had to be performed at the level of visual matching. It appears that the 3/4 view advantage may be obtained only when the task involves explicit matching between test views and remembered target photographs, rather than reflecting any more fundamental properties of the representations used to recognize highly familiar faces.