Abstract
We investigated whether acutely induced anxiety modifies the ability to match photographed faces. Establishing the extent to which anxiety affects face-matching accuracy is important because of the relevance of face-matching performance to critical security-related applications. Participants (N = 28) completed the Glasgow Face Matching Test twice, once during a 20-min inhalation of medical air and once during a similar inhalation of air enriched with 7.5% CO2, which is a validated method for inducing acute anxiety. Anxiety degraded performance, but only with respect to hits, not false alarms. This finding provides further support for the dissociation between the ability to accurately identify a genuine match between faces and the ability to identify the lack of a match. Problems with the accuracy of facial identification are not resolved even when viewers are presented with a good photographic image of a face, and identification inaccuracy may be heightened when viewers are experiencing acute anxiety.
Eyewitness memory is well known to be fallible. Because of its critical role in legal settings, there is extensive research aiming to establish the effect of various factors on eyewitness memory (Lindsay, Mansour, Kalmet, Bertrand, & Melsom, 2011). One key factor is stress: Witnesses are often asked to recall information from stressful situations, and yet the effects of stress are complex. A range of studies has shown rather inconsistent effects of stress. However, Deffenbacher, Bornstein, Penrod, and McGorty (2004) have provided a meta-analytic review demonstrating that elevated stress (during encoding) reduces subsequent accuracy in recognizing targets in lineups. This effect holds only for lineups in which the target is present; manipulations of stress do not affect accuracy when the target is absent.
The large literature on eyewitness testimony is focused almost entirely on memory. However, in the modern day, face matching is an equally important issue. People are commonly required to provide photographic evidence of their identity not only in high-security situations, such as border crossings, but also in daily life, for example, when buying age-restricted goods such as tobacco or alcohol. Face matching has been much less studied than face memory, but the growing literature shows consistently that people are surprisingly poor at matching two images of an unfamiliar face (Bruce et al., 1999; Hancock, Bruce, & Burton, 2000). Furthermore, in forensic situations in which viewers match faces shown on closed-circuit television (CCTV) to photographs of suspects, accuracy is high when the faces are familiar, but poor when the faces are unfamiliar (Bruce, Henderson, Newman, & Burton, 2001; Burton, Wilson, Cowan, & Bruce, 1999).
In this study, we asked whether anxiety affects viewers’ ability to match faces. The answer to this question is important for theoretical and practical reasons. For theory, it is important to establish whether the detrimental effect of anxiety on face memory has its locus in memory processes per se or face encoding itself. In a matching task, with face images continually present, one is not required to retain visual information over long intervals, and so it may be possible to use such a task to establish the effects of anxiety on different components of task performance. More practically, it is important to establish the limits of face-matching accuracy. In many critical security-related applications, viewers are asked to match people to photographs (e.g., to scan a crowd to search for someone matching the photograph of a known suspect). The extent to which stress affects performance on such tasks is currently completely unknown.
In previous research on eyewitness memory, stress has been induced in a variety of setting-specific ways. For example, in a recent study by Valentine and Mesout (2009), visitors to the London Dungeon were frightened in the Horror Labyrinth, and heart rate monitors provided an objective measure of stress. In the study reported here, we induced anxiety directly, by having participants inhale air enriched with 7.5% carbon dioxide (CO2). This has been shown to modify threat processing (Garner, Attwood, Baldwin, James, & Munafò, 2011), but has not been used previously to assess the effects of anxiety on performance of naturalistic tasks relevant to security monitoring, such as face matching.
One important aspect of our study is that we examined hits and false alarms separately. The meta-analysis by Deffenbacher et al. (2004) showed differential effects of stress on these two components of face memory, which is consistent with work on recognition memory suggesting that these two types of errors are dissociated (Vokey & Read, 1992). This is a slightly puzzling effect, as one might have predicted that distinctive faces would be both easy to remember if previously seen and easy to reject if previously unseen. Nevertheless, many studies have replicated the dissociation (Hancock et al., 2000). Following Vokey and Read (1992), explanations for the dissociation have typically invoked dual-process accounts in which at least one of these two components relies on the memorability of the stimuli. However, it has been demonstrated more recently that hits and false alarms are also dissociated in face-matching tasks with unfamiliar faces, which do not require long-term memory (Megreya & Burton, 2007). This finding challenges theoretical explanations of unfamiliar-face recognition, but is also important practically because it suggests that the ability to accurately identify a genuine match between two representations of an unfamiliar face is unrelated to the ability to identify a lack of a match between two representations of different unfamiliar faces.
Although it is now quite well established that hits and false positives are dissociated in tasks involving unfamiliar faces, the underlying mechanisms for this dissociation remain unclear. When studying the effects of novel manipulations on performance, it is therefore important to ask whether such manipulations have their effect across both components of task performance or are specific to one or the other component. We have noted that previous studies of stress and face recognition have produced mixed results (Lindsay et al., 2011). It is possible that one reason for this inconsistency is that most studies have examined overall accuracy, without breaking down performance into these components. In our study, we therefore explicitly examined the effects of stress separately on hits and false alarms.
To address these issues, we explored whether acute anxiety modifies performance on the Glasgow Face Matching Test (GFMT; Burton, White, & McNeill, 2010). This is a standardized measure of face matching, requiring viewers to make same/different judgments on pairs of face images. The test is self-paced and does not require long-term retention of the faces in memory. Each image pair is presented until the participant makes a judgment, and instructions emphasize accuracy. Any effects of anxiety on performance are therefore not attributable to long-term memory requirements or to response deadlines.
Method
We recruited 28 healthy volunteers (50% male, 50% female), who were 20 years old on average (range = 18–25, SD = 1.7). Exclusion criteria included recent (past 8 weeks) use of medication (not including local treatments such as topical creams, aspirin or paracetamol, or contraceptive pills), pregnancy, history of asthma or respiratory illness, high blood pressure (> 140/90 mm Hg), cardiovascular disease, migraines, current or lifetime history of psychiatric illness (including family history of panic attacks), smoking, being under- or overweight (body mass index less than 18 or greater than 28), current or past drug or alcohol dependence, and recent use of illicit drugs (urine screen) or alcohol (breath test). The University of Bristol Faculty of Science Research Ethics Committee approved all study procedures. All participants provided informed consent and completed a screening procedure prior to testing.
Participants completed the full GFMT twice, once during a 20-min inhalation of medical-grade air (the air condition) and once during a similar inhalation of air enriched with 7.5% CO2 (21% oxygen, 71.5% nitrogen; the CO2 condition). Testing took place during a single session, and there was a 30-min rest period between the two blocks. This rest period was sufficient time for the effects of inhaling CO2-enriched air to dissipate, as these effects persist only during active inhalation and disappear within a few minutes afterward.
The gases were administered to participants through an oro-nasal face mask with the order of the conditions (i.e., CO2 vs. air) counterbalanced across participants, in a within-subjects, single-blind, crossover design. (For safety reasons, the experimenter was not blind to the gas being administered.) Participants were told prior to enrollment that they would inhale CO2-enriched air, which produces feelings of anxiety in most people, at some point during the experiment, but not whether they would inhale this first or second.
The GFMT requires participants to make 168 two-alternative forced-choice judgments of pairs of different black-and-white photographs. On each trial, two photographs are presented together, and participants indicate whether or not the faces in the pair are of the same individual; in 84 trials, the individuals presented are the same, and in the other 84, they are not. All images are approximately the same size and are front facing; there are minor differences in features such as luminance, gaze direction, and hairstyle. A detailed description of the task and stimulus pairs is provided elsewhere (Burton et al., 2010).
We obtained subjective ratings of anxiety, using the State-Trait Anxiety Inventory state subscale (STAI-S; Spielberger, 1983), as well as measures of physiological arousal (systolic blood pressure and heart rate). These measures were taken immediately after the inhalation of the gas in each condition. In the case of the STAI-S, participants were asked to rate the peak anxiety experienced during the inhalation.
Results
Table 1 summarizes the results for the measures of anxiety, arousal, and performance on the GFMT. Subjective ratings of anxiety were higher in the CO2 condition compared with the air condition, t(26) = 11.99, p < .001. Similar effects were observed for systolic blood pressure, t(26) = 5.31, p < .001, and heart rate, t(26) = 6.70, p < .001. Data on these measures were missing for 1 participant.
Effects of Inhaling 7.5% Carbon Dioxide on Face-Matching Performance, Anxiety, and Arousal
Note: For each of the two conditions, the table reports mean values, with standard deviations in parentheses.
Critically, accuracy on the face-matching task was reduced in the CO2 condition, compared with the air condition, t(27) = 2.88, p = .008. This effect was driven by a reduction in hits, t(27) = 2.55, p = .017; there was no evidence of an effect on false alarms, t(27) = 0.76, p = .45. Thus, our results suggest that face-matching performance is impaired under conditions of elevated anxiety and arousal. However, this effect is restricted to hits, as there is no effect on false alarms.
Discussion
Our results indicate that anxiety degrades performance in a face-matching task, but only with respect to hits, not false alarms. This finding is of theoretical interest because it demonstrates that the detrimental effects of anxiety on face-matching performance can occur in the absence of any demands on memory. It also provides further support for the dissociation between the ability to accurately identify a genuine face match (i.e., anxiety lowers hit rates) and the ability to accurately identify a lack of a match (i.e., anxiety does not affect false alarm rates). Practically, this finding suggests that problems with the accuracy of face matching are not resolved even when viewers are presented with a good photographic image of an unfamiliar individual, and that inaccuracy of facial identification may be heightened under conditions of acute anxiety. Elevated anxiety and arousal during initial processing and encoding of faces may have important effects on the performance of individuals whose job it is to identify faces on CCTV and on the reliability of eyewitness testimony. Our data also provide further support for the dissociation between eyewitnesses’ performance when the target is present in a lineup and when the target is absent.
Monitoring of individuals through covert surveillance, CCTV, and other means for security purposes is increasingly common in modern society, and police forces routinely use still images from such sources to identify suspects. This is typically justified as necessary in order to combat international terrorism and tackle crime. However, such monitoring requires human input, and errors inevitably occur, with serious consequences. Often, the people making judgments of identity do so under conditions of considerable pressure and associated anxiety (Rossmo, 2008). The fallibility of eyewitness testimony is typically ascribed to imperfect memory, and attempts to improve eyewitness performance focus on improving memory retrieval (e.g., by attempting to replicate the conditions of initial encoding). Our results suggest that this is only part of the problem, as identification of unfamiliar faces is also poor in a matching task, in which there is no need for long-term memory retention. Attempts to improve performance may therefore require a focus on encoding itself, as well as the more traditional focus on memory.
Footnotes
Acknowledgements
The authors are grateful to Matthew Hotton, Francesca Seymour, Sarah Shenton, and Cara Struthers for assistance with data collection.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.
Funding
A. S. Attwood and M. R. Munafò are members of the UK Centre for Tobacco Control Studies, a UK Clinical Research Collaboration Public Health Research Centre of Excellence. The authors gratefully acknowledge the funding that the UK Centre for Tobacco Control Studies receives from the British Heart Foundation, Cancer Research UK, the Economic and Social Research Council, the Medical Research Council, and the National Institute for Health Research, under the auspices of the UK Clinical Research Collaboration.