Four Nations Cognitive Research – An International Collaboration to Produce Actionable Banknote Design Guidelines

By Jane Raymond and Daniel Dodgson (University of Birmingham); Lewis Mundy-Gill and Ronan McClintock (Bank of England); Kristen Wilson, John Mascioli and Douglas Crane (Federal Reserve Board); Matthew Tsikrikas, Amanda Evans and Vanessa Lussini (Reserve Bank of Australia); Matthew Malarek and Badr Omrane (Bank of Canada).

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.

Introduction

Banknotes are designed to provide a safe, secure, and accessible payment system for the public, making it critical that their needs are properly identified and met. As with most consumer products, this is best achieved by adopting a user-experience perspective.

Although banknote users have many requirements related to ease of use, portability, and durability, their most important consideration is trust. Trust in banknotes is first and foremost based on the perception of authenticity, a psychological state that is central to widespread acceptance of banknotes in the marketplace. Indeed, the consumer’s requirement that all users be able to effortlessly authenticate a banknote is fundamental to confident banknote use.

Central banks around the world have invested substantial resources into supporting this basic consumer need by taking measures to prevent counterfeit banknote production and to make banknotes that are especially difficult, if not impossible, to mimic.

Features providing public cues to authenticity are ‘level-1’ (overt) security solutions such as high-resolution portraits, patches with micro-optics that produce illusions of movement or depth, areas printed with colour shifting ink, windows, watermarks, or foils that produce changing images when the banknote is moved.

Critically, level-1 security features are specifically designed to provide a unique, easily recognised sensory cue that is difficult, if not impossible, to mimic.

Although the intended use of level-1 security features is clear, their real-world effectiveness depends solely on how users perceive and use the critical authenticity cue. For example, a shiny foil that makes one image flip into another is not fit for purpose if users cannot tell the difference between the first and second image.

Similarly, if consumers simply register ‘shiny’ as the key cue to authenticity rather than ‘image change’, they will be easily fooled by counterfeiters who can readily mimic this simple quality of the security feature. Clearly, the sensory and cognitive processes of the user’s brain are key factors that ultimately determine the effectiveness and safety of public aids to authentication.

Creating intuitive banknotes that are easy to use, and yet complex to counterfeit, is the ultimate paradigm of banknote design and development. To deter counterfeiting, central banks need to consider a combination of factors including technology integration, technology maturity, user perception, and design aesthetics.

Turning to human perception analysis

To aid in these challenging decisions, central banks have turned to human perception analysis as a means of evaluating the relative security merit of a new security solution. Lab-based methodologies for ranking security features have been critical to making informed decisions and providing quantitative results. However, these methodologies do not provide a complete understanding of why particular features are performing better, or why some counterfeits are more deceptive than others.

Whereas these investigations offer a better understanding of the psychological processes behind the discrimination of counterfeit notes, further analysis is required to gain a full understanding of the environmental and cognitive factors that influence authentication behaviours. Advancing knowledge also assists associated strategies for security feature development, selection, and integration.

In the past 20 years, significant advances have been made in the understanding of the human brain, its sensory systems, and its capacity to make the split-second decisions required to authenticate a banknote in a typical setting. Such developments have been paralleled by innovations in level-1 security features and print technologies.

However, bringing psychological and materials science together to advance banknote usability has been relatively slow. Public usability research using approaches from cognitive psychology began in the 1990s and the banknote industry’s interest in the importance of perceptual and psychological factors in banknote authentication has steadily increased since then. Yet, banknote usability research has been largely piecemeal, short term, and typically directed at narrow problems and single currencies.

What has been needed to make real advances in this area is a strong theoretical basis and an in-depth understanding of how human cognition is used to authenticate objects in the broader sense.

To address the issue, the Four Nations came together in 2018 to fund a large three-year research programme, led by Prof Jane Raymond at the University of Birmingham. The aims of the project were to develop a theoretical model of the human cognitive processes involved in authentication, to test the model by conducting a series of banknote authentication studies, and then to develop principles and guidelines for better banknote design.

Unprecedented collaboration

Through this unprecedented investment and collaboration, the Four Nations conducted several studies across four different countries to better understand the underlying mechanisms and socio-geographic factors that underpin human authentication behaviours.

The body of work extended beyond banknote discrimination tasks, to include socio-geographic habits and perception of local and foreigner banknotes. The targeted outcomes of the project were to provide the Four Nations with an actionable analysis of human authentication behaviour, and concrete guidelines that will enable informed decision making on banknote design, security feature selection, and quality policies.

Drawing from knowledge of human perception, cognition and decision-making, Prof Raymond along with Dr Daniel Dodgson developed SIMBA, a model of the cognitive processes that underpin human banknote authentication. This is described in Box 1.

The ‘CogNote’ project

They then constructed a programmatic series of human banknote authentication studies, collectively referred to as the CogNote Project. The project used a wide range of approaches including observation, authentication accuracy measures, brain electrophysiology, eye movements, and computer vision to better understand the user-experience during authentication.

The studies were conducted at four different sites around the world, one in each of the four nations, and have involved over 450 participants. Two denominations of banknotes from each country were studied and critically, counterfeits (a mix of forensically recovered and custom-made notes) were matched for quality by experts. This was essential to allow cross-currency comparisons of authentication performance and is a key feature of the CogNote Project.

Having amassed a large quantity of data, the scientists engaged in detailed analyses producing a series of technical reports, culminating in a design-focused summary for internal consumption by the banks. Here are some the highlights from those reports.

Cognitive research on banknote authentication

A driving principle behind the entire CogNote research program was to study authentication under the cognitive conditions akin to those experienced by public users, insofar as possible, without relinquishing the necessary experimental controls required to produce useable data.

This was achieved in three ways.

First, all studies required participants to denominate and authenticate each of a series of banknotes, with each note presented one at a time.

Second, these judgements were required with limited time to consider each note.

Third, banknotes were always presented with some amount of movement and some light transmission so that level-1 security features could be used, even when presentation was digital and very brief.

The same basic procedure was used in all the authentication studies of the CogNote Project and is described in Box 2.

Authentication with very brief viewing

A key feature of the SIMBA model is that the cognitive and perceptual processes that occur in the first half second of viewing a banknote should determine whether a counterfeit will be detected.

The model hypothesises that during the first brief glance, rapidly generated brain signals (sensory data) are matched within the brain to predicted and stored neural data that represent a prototypical genuine banknote. If there is a mismatch between the predicted and physically present banknote, then an alerting or ‘suspicion’ signal is automatically generated. According to the model, this neural signal activates conscious awareness of the discrepancy and initiates search (looking and touching) for more information.

A first critical test of the model was to determine whether accurate authentication is humanly possible with a half second’s glance at a banknote.

Across a series of studies, CogNote results clearly showed that adult humans are indeed accurate at detecting counterfeits when they are permitted only a half second or less to view each note. Although for some banknotes performance improved slightly when longer viewing time (several seconds) was allowed, for other banknotes protracted viewing did not boost performance, suggesting that all the available, useful information had been harvested in the first half second. These data are summarised in Figure 2.

Figure 2: Average counterfeit sensitivity (d’) for the four different currencies (differently coloured dots) when the banknotes were presented for increasing lengths of time. Higher d’ scores mean better performance. During Fast Sensory Analysis, counterfeit sensitivity drastically improves with each small increase in viewing time. Sensitivity is good for all currencies with only half a second to view each note; it improves considerably for some but not all currencies when additional seconds are provided and Slow Inspection Processes can be engaged.

Confirmation of rapid perception of counterfeits was found in electrophysiological studies that measured brain activity (as in EEG) as people judged genuine and counterfeit banknotes - see Figure 3. 

Figure 3. Heat map showing the average level of electrical activity in the brain (obtained using EEG) when viewing counterfeit (top two rows) or genuine (bottom two rows) banknotes. Red indicates greater activity. Each column shows brain events at different times after the banknote is presented. Left: 50- 100 ms after presentation; centre: 450-600 ms after; right: 600-900 ms after.

These studies showed that the brain produces a clear signal identifying a banknote as counterfeit after only a tenth of a second. This signal remains present in the EEG recorded brain activity for about a half second, at which point it is combined with other high-level brain signals to determine overt authentication judgements.

These exciting findings confirm an important prediction from the SIMBA model, namely that the first half second plays a critical role in authentication. The implication is that if counterfeit is not detected during this split-second interval, the user is unlikely to engage in further cognitive analysis of the banknote and will most likely accept it. Clearly, level-1 security features should aim to offer information valuable to the user within this brief time window.

Although somewhat difficult to see in the left column of Figure 3, the unexpected appearance of a counterfeit banknote produced reliably greater activation than an expected genuine note in the first tenth of a second. By the half-second point (centre and right columns), counterfeit notes, especially those eventually rejected, produced substantially more activity. This shows that the brain gains reliable information about the presence of counterfeit within the first half second of viewing and then combines this with other information to determine an authentication decision.

Additional verification

Additional verification of this possibility was obtained in a large eye-movement study conducted on over 160 people as part of the CogNote Project. This study detailed how people scan banknotes with their eyes to extract information needed to make banknote authentication decisions.

The study revealed that about a half second after the banknote was presented, the interval needed to make two or three eye movements, gaze position predicted authentication accuracy. Specifically, if the second or third eye fixation (ie. when the eye is still and vision is possible) was directed at the primary security, performance was more likely to be accurate than if the eyes were directed at other design elements, especially the portrait, during this interval.

This finding is consistent with SIMBA because it shows that the specific type of information collected during the initial, fast stage determines authentication judgements. The eye movement results also show that this ‘early’ information is used to direct subsequent, strategic eye movements to areas able to provide confirmatory evidence.

For example, the study showed that when counterfeit notes were viewed, participants tended to correctly identify them if they had fixated the primary security feature both at the beginning of the viewing interval (early stage) and again in the latter half of the first second (late stage). However, if both early and late fixations were directed elsewhere, counterfeits were more likely to be missed.

Implications for banknote design

Not only do these findings underscore the importance of security features, they also support the notion that slow, strategic scrutiny is used to follow-up on rapidly acquired ‘early’ information. Such findings have clear implications for banknote and security feature design.

Currently the CogNote Project is engaged in further efforts to understand how banknote design elements guide eye movements and attention, how perceptual and cognitive limitations imposed by advancing age affect banknote authentication, and the role of tactility.

Working with tools from computer vision, the research is also aiming to provide banknote designers with development tools useful for assessing design effectiveness in terms of user experience. By adopting a user experience perspective, advances from the CogNote Project will allow banknote designers to take full advantage of innovations in security features yet, at the same time, create banknotes that meet the user’s need for effortless banknote authentication.

Box 1: A model of human banknote authentication

Based on decades of research on human cognitive processes, Raymond & Dodgson developed the Suspicion Initiated Model of Bank Authentication (SIMBA) to explain the cognitive mechanisms that underpin authentication of banknotes in situations where counterfeit is believed to be scarce and there is a strong a priori assumption of authenticity.

In their view, authentication in such contexts does not begin as a conscious, motivated inspection of the banknote, but rather is initiated only when automatic and almost certainly sub-conscious processes detect a fault in a banknote.

SIMBA is outlined in Figure 1. According to this model, even before a banknote appears in a transaction, the brain prepares a prediction, or expectation, of its appearance. When the banknote is present, the brain rapidly (in less than a half second) analyses the look and tactile feel of the note, matching the complex array of sensory data to an equally complex array of stored data about how the banknote should look and feel.

The stored data is based on previous encounters with banknotes, which have all, or nearly all, been genuine. If the currently considered banknote is also genuine, then the match is near perfect, and the user carries on with the transaction, allowing their cognition to engage with other matters at hand (eg. computing change).

However, if the note is counterfeit, then a low-level sensory mismatch will occur. This generates a sub-conscious, alerting signal called ‘suspicion’. It initiates a cascade of cognitive processes that are directed at finding more information about the banknote. If more mismatches are found, then conscious feelings of suspicion may be generated, an emotion that motivates even greater scrutiny, including more eye movements and more tactile exploration.

When sufficient evidence is accumulated in favour or against a counterfeit judgement, overt, conscious action is taken to reject or accept the banknote, respectively. The whole process occurs in less than a second or two.

SIMBA makes several predictions. First, it predicts that authentication can be accurate with less than a half second of seeing a banknote. Second, it predicts the presence of very fast, low-level signal in the brains’ electrical activity within the first half second of seeing a banknote. Third, it predicts that counterfeit detection requires strategically seeking information from areas on the banknote, such as security features, that are most likely to confirm or refute mismatches between current sensory information and information in memory.

Figure 1: SIMBA, Suspicion-Initiated Model of Banknote Authentication. Expectation refers to the brain’s prediction of the banknote’s likely appearance. Time “0” is when the banknote becomes present. Fast Sensory Analysis occurs without conscious awareness and involves matching expectation to current sensory information. Mismatches initiate ‘suspicion’ signals, leading to slow, strategic, and conscious inspection. When sufficient evidence is garnered, a decision (accept/reject) is rendered.

Box 2: How are banknote authentication studies conducted?

Typically (eg. Raymond & Jones, 2018), participants judge the authenticity and denomination of a series of randomly mixed genuine and counterfeit banknotes, considering each note one at a time. The proportion of counterfeit notes judged to be ‘fake’ is mathematically combined with the proportion of genuine notes misjudged as ‘fake’ to compute a value known as d’ (pronounced ‘d-prime’). d’ indexes an individuals’ counterfeit sensitivity without bias from individual tendencies either to label many notes as ‘fake’ or to avoid labelling any notes as ‘fake’. When d’ = 0, the user has been guessing every time; d’ = 1.25 means the user has good sensitivity to counterfeit; d’ > 2.0 means they have very good counterfeit sensitivity.

A critical determinant of authentication performance in perception studies is the ‘quality’ of the counterfeit. At one extreme, very poor counterfeits, eg. a child’s crayon drawing of a banknote, would be readily detected, resulting in very high d’ values. On the other extreme, a masterpiece reproduction might be very difficult to identify, producing very low d’ values.

If counterfeit sensitivity for different currencies is to be compared, each currency’s counterfeit needs be of similar quality in the authentication study for those comparisons to make any sense. In the CogNote Project studies, counterfeit was expertly assessed as matching in quality across currencies, ensuring that cross-currency comparisons were valid.