In Part 1 of this series, the Four Nations Group described their research to determine how public users benefit from the considerable investment by Central Banks in banknote design and complex anti-counterfeit security features if they barely have time to notice them. The group’s findings show that accurate authentication of a banknote is well within human capability even when a banknote is seen for less than a second. Moreover, accurate authentication, especially when time is short, appears to be highly dependent on banknote design.
In this article, the authors delve into the details of their research on consumer behavior when authenticating banknotes under time pressure and their eye movements during this process. You’ll also learn about the computer-based image analysis researchers applied to test and quantify these factors.
Authentication under time pressure
Using systematic protocols and sophisticated eye movement measurement techniques, the CogNote project made two important discoveries that confirm SIMBA’s predictions. First, the studies showed that under a range of circumstances authentication performance on eight different banknotes was good to very good with as little as a half of second of viewing the banknote. Across all currencies, denominations and participants, counterfeit sensitivity indexed by d’  was 1.25 with a half second of viewing a banknote and improved only modestly (to 1.40) when viewing time was extended to 3 seconds. This performance improvement with longer viewing was statistically significant for only two of the four currencies studied. Finding good counterfeit sensitivity with less than a second of viewing not only explains why people glance so briefly at banknotes during a typical transaction, it underscores the necessity of providing high quality anti-counterfeit information that is accessible with a brief glance at a banknote.
Eye movements during authentication
The second important finding was that authentication performance was clearly linked to eye movements made during that crucial first second of viewing a banknote. Data from the CogNote Project showed that during this period, people typically make two to three eye movements, allowing them to fixate different locations on the banknote. Not only do different people scan banknotes differently, they also vary their scan patterns as they authenticate successive banknotes. Nevertheless, using averaging and other statistical processes, a typical picture of scanning eye movements emerged that could be linked to authentication performance. On average, gaze was most likely to be directed at the portrait initially and then to the security feature. As the scanning of the note progressed over the first second or two, the probability of fixating the portrait went down and the probability of fixating the primary security feature went up. This effect is shown in Figure 1 where the special case of judging counterfeit is examined. Looking behaviour directed at the portrait of the counterfeit note was the same regardless of whether the counterfeit was detected or missed, suggesting that information provided by the portrait did not contribute to authentication judgements at that time. However, eye movements directed at the security feature occurred with a high probability when counterfeit were correctly detected than when they were missed. This shows that information in the security feature was important for counterfeit detection, even during initial eye movements.
Figure 1. The proportion of eye fixations directed at the portrait (left) or primary security feature (right) when counterfeit banknotes were randomly presented within a series of genuine banknotes. Values are plotted separately for each successive fixation. Black lines show values when participants correctly rejected counterfeit (Cft correct); red line show values when counterfeit were incorrectly judged to be genuine (Cft incorrect). Vertical lines the indicate ±95% confidence limit for each point. These data clearly show that correct rejection of counterfeit is associated with looking directly at the security feature, showing that the value of security features on banknotes depends on their capacity to attract attention.
Another way to consider the role of eye movement in authentication behaviour is to ask whether counterfeit sensitivity (using responses to both genuine and counterfeit notes) depends on where the first, second, or third fixation is directed. Again, averaging across all banknotes and participants, CogNote found that when the primary security feature was fixated with the first or second eye movement, counterfeit sensitivity was significantly better than if the first or second eye movement was directed at the portrait or another area of the banknote. These findings are shown Figure 2. They robustly show that rapidly drawing the eyes to the primary security feature should be is a primary banknote design objective.
What was particularly interesting from these studies is that the design of the different banknotes under consideration appeared to play an important role in determining where people looked on their initial fixations. The close link between scene elements and eye movements has already been well established by basic psychological science researchers, but the CogNote project was able to analyse how specific variations in banknote design determined both looking behaviour and authentication performance. Basic research has repeatedly shown that eye movements and attention are initially attracted by visual patterns that are critical for object recognition, such as edges, contours and areas of high colour or tonal contrast. These areas of high perceptual salience are automatically prioritised for processing by the brain and attract the eye and attention at the expense of other less salient areas. This suggests that if banknote design aims to draw the eye immediately to a security feature on a banknote, this area should be higher in perceptual salience than other more easily mimicked areas of the banknote.
Figure 2. Counterfeit sensitivity measured when the first, second, or third fixation was directed at the primary security feature (SF), the portrait, or any other area of the banknote. Fixating the security feature in either the first or second fixation ultimately resulted in better counterfeit detection than when the portrait or other areas were fixated at this stage.
Computer-based image analysis
To test these ideas and to quantify perceptual salience on banknotes with different designs, the CogNote Project turned to image analysis techniques used in artificial intelligence (AI). Computer-based object recognition, a key component of AI, often uses image analysis techniques based on human sensory capacity. These techniques employ computer algorithms that identify and quantify sources of information in visual patterns that are especially salient to the human brain for the purpose of object recognition and scene analysis. Using these types of algorithms, the CogNote Project analysed several different banknote designs, producing salience maps of each banknote. These maps were then correlated with eye movement heat maps. When initial but not later fixations are considered, AI generated salience maps correlated closely with eye movement heatmaps, supporting the contention that eye movements used to scan a banknote are initially driven by the sensory data provided by the banknote’s design.
In this way, the CogNote Project successfully connected eye movement patterns to banknote design. Having already linked authentication performance to eye movements, the inference can be straightforwardly made that banknote design determines authentication performance. By engaging in a large systematic series of studies that were based on a well-founded theoretical model, the CogNote Project has been able to provide conclusive evidence that security features provide useful authentication information to users, even when they have only a moment to glance at them. The project also offers the clear guideline that for security features to be effective, they need to be perceptually obvious and rapidly attract the user’s attention.
MORE ABOUT THE AUTHORS
The Four Nations group was founded in 1978 and consists of five central bank members – the Board of Governors of the Federal Reserve System, Bank of Canada, Reserve Bank of Australia, Bank of England, and Banco de México. The group is a forum in which its members share information and experiences on banknote development, issuance and distribution, counterfeit deterrence, and relevant technical studies.
 Counterfeit sensitivity is conventionally measured using d’, a statistic that combines the probability of detecting a counterfeit when it is indeed present with the probability of erroneously judging a genuine note to be counterfeit. This statistic is used to avoid response biases. A value of 0.0 means guessing; values exceeding 1.25 show good sensitivity and values exceeding 1.5 show very good sensitivity.
2 Raymond J. E. & Jones, S. P. (2019). Strategic eye movements are used to support object authentication. Scientific Reports. 9:2424 | https://doi.org/10.1038/s41598-019-38824-z
The Four Nations Group was founded in 1978 and consists of five central bank members – the Board of Governors of the Federal Reserve System, Bank of Canada, Reserve Bank of Australia, Bank of England and Banco de México. The group is a forum in which its members share information and experiences on banknote development, issuance and distribution, counterfeit deterrence, and relevant technical studies.