In 2016, consumers in the euro area made on average 1.2 cash payments per day (Esselink & Hernandez, 2017). These cash transactions were largely habitual (Van der Horst & Matthijsen, 2013). Upon receiving a banknote – either from a retailer or in a person-to-person transaction – people typically prioritize determining its value. Determining whether the banknote is fake or real is regarded as less important (Klöne, Vrakking, & Zondervan, 2019). Research has shown that Dutch citizens have strong confidence in the authenticity of euro banknotes because the likelihood of receiving a counterfeit is very low (Van der Horst, De Heij, Miedema, & Van der Woude, 2017). For example, in Europe in 2018, the number of counterfeit euro banknotes that were removed from circulation (563,000) constituted only 0.003% of the number of genuine euro banknotes in circulation (22 billion) (European Central Bank (ECB), 2019). Mainly because of this, people tend to not authenticate banknotes, especially when, at first glance, the banknote appears normal (Van der Horst et al., 2017). Indeed, in the study of Klöne et al. (2019), 70% of a sample of Dutch respondents claimed to have never intentionally and consciously authenticated a banknote in the last 5 years.
The relatively high levels of trust exhibited by the general public fuel the need for banknotes of which the authenticity can be easily confirmed – which, in consequence, should boost one’s ability to detect deviants beyond the limits imposed by naïveté. For these reasons, all central banks incorporate various security features in their banknotes to assist various user groups in identifying counterfeits without specialized equipment. Examples are a watermark, a security thread that is imbedded in the paper, optically changing elements, security foils (sometimes including holograms), paper structure and alto-relievo induced by intaglio printing (raised ink). These authentication features appeal to two of our five senses, namely sight and touch (see Fig. 1).
As of yet, we do not have complete knowledge of the factors contributing to counterfeit detectability. In particular, we know little of the respective contribution of visual and haptic perception in the decision-making process. We additionally do not know how much time is needed to ensure that the exploitation of these senses prompts at least a decent detection performance. For instance, would one either feel anomalies within a split second or never at all, or would anomaly detection improve as one accrues more haptic evidence over time? Additionally, we do not yet know how these factors are influenced by expertise. Specifically, might expertise increase the value of evidence accrued beyond the first impression? The present study is aimed at answering these questions.
Two decision-making systems
Before reporting our experiments, we should outline a few theoretical constraints. Prior research has led one to believe that humans have two separate cognitive systems driving decision-making. One of these is fast, automatic and largely non-conscious; a type of processing that has been labeled System-1 or Type-1 processing (Frankish, 2010; Kahneman, 2010). In the present context it could be argued that a typical cash transaction would solely involve Type-1 processing. However, in probing counterfeit detection with a cognitive experiment, we may inherently be unable to assess Type-1 processing: specifically, asking participants whether a given banknote is real or fake is likely to induce atypical levels of distrust. In consequence, authentication would consist of a slower, controlled and conscious decision-making process, which in the literature has been labeled Type-2 processing (Frankish, 2010; Kahneman, 2010). See Klöne et al. (2019) for a discussion of factors driving a more deliberate banknote verification process.
Hence, in assessing the interactions of perceptual modality, time and expertise in the detection of counterfeit banknotes, our conclusions will largely pertain to Type-2 decision-making processes. The extent to which findings may inform us about Type-1 processes will be addressed in the “General discussion” section.
It must be stressed that human authentication has its limits even when invoking Type-2 processes. For example, in an experiment in which genuine and manipulated photographs were presented on a computer screen, it was shown that people have poor ability in identifying whether an image is the original or has been manipulated (Nightingale, Wade, & Watson, 2017). It has been argued that our inability to detect changes is largely driven by the fact that the overall gist of the percept remains unaltered (e.g., Standing, Conezio, & Haber, 1970).
In addition to limitations in perception, we must highlight the “prevalence effect.” Visual search experiments are perceptual tasks that require active scanning of the visual environment for a particular object or feature (the target) among other objects or features (the distractors). In most visual search experiments, targets appear on at least 50% of trials (Wolfe & van Wert, 2010). However, Wolfe and van Wert showed that when targets were rare (1% prevalence) observers made more than four times the number of “miss” errors made when targets were common (50% prevalence). To miss a disproportionate number of targets when these targets are rare, is especially problematic in important everyday contexts such as medical or airport screening (Wolfe & Van Wert, 2010). Wolfe, Horowitz, and Kenner (2005) showed that if observers repeatedly do not see their target, they will more probably fail to notice it once it does appear. Undoubtedly this prevalence effect also impacts on counterfeit detection as well, given that counterfeits in everyday life are extremely rare (Rich et al., 2008).
About time
The act of accepting a banknote is performed rapidly. An internal DNB cashier field study (Zondervan, Heinen, & Heuvel, 2019) shows that most cashiers make – implicitly or explicitly and without the use of authenticating devices – the decision of whether or not to accept a banknote within 3 s. According to Layne-Farrar (2011) it takes only 1–2 s for waiters to pick up money from a table for a tip and pocket it. The simple task of accepting a banknote and storing it in your wallet is probably within that range of time. Presumably, this is also the time that the banknote has been authenticated, at least implicitly. Some national central banks of the Eurosystem (e.g., Bank of Italy (2020) and Bank of Finland (2020)) state on their websites that it only takes a few seconds to (explicitly) authenticate a banknote. However, as far as is known there has been no empirical evidence regarding the speed with which banknote can be recognized as counterfeit or genuine. In the current study, the task was to decide whether a banknote was counterfeit or genuine and the exposure duration to the banknote was systematically varied.
As noted before, it is unknown whether counterfeit detection would benefit from a longer exposure duration. However, research on scene perception may be somewhat informative, as it has revealed that people are able to recognize complex, real-world scenes at a mere glance, regardless of the visual complexity of the scene (see, for instance, Fei-Fei, Lyer, Koch, & Perona, 2007; Oliva, 2005). On the other hand, it should be noted that this type of recognition concerned the gist of the scene (e.g., “it is an outdoor scene with mountains”) without concern for specific features or details. Possibly, more fine-grained perception is a prerequisite for successful counterfeit detection.
It is generally believed that presenting a display for only 200 ms should be enough for detecting basic features. Because the time it takes to make a saccadic movement is at least 200 ms, such a task is completed in a single glance (Healey & Enns, 2012). The recognition and discrimination of patterns appears to take longer. According to Fei-Fei et al. (2007) observers need a presentation time of 500 ms to be able to almost perfectly categorize outdoor and indoor scenes. Furthermore, a study by Greene, Botros, Beck, and Fei-Fei (2015) showed that participants can make an adequate description of typical real-world situations scenes after 506 ms, although it takes participants longer to understand and even perceive improbable visual images (e.g., a press conference being convened under water), indicating that our rapid scene-categorization abilities depend critically on our prior experience with real-world environments (Greene et al., 2015).
As noted earlier, one of the goals of this study was to determine the lower limit on how rapidly people can distinguish counterfeits from real banknotes. If counterfeits are distinct from genuine banknotes by virtue of features that stand out (e.g., Theeuwes, 1992) then one should be able to do this very rapidly. In the set of exposure times employed in our experiments, we therefore incorporated a 500-ms condition, which represents a time in which one or two eye-movements can be made. We also tested longer exposure durations of 1000 ms and (up to) 10 s to determine whether a longer exposure duration would improve performance. Indeed, if the detection of counterfeit banknotes requires the processing of specific details, we would expect that a longer exposure duration would greatly improve performance. As an upper limit we used an exposure duration of 10 s, as it was previously shown that the hit rate in detecting a counterfeit does not increase beyond an exposure duration of 10 s (Van der Horst, Eschelbach, Sieber, & Miedema, 2016).
As argued before, it remains to be seen to what extent these temporal constraints are modulated by certain factors, such as expertise and perceptual modality. Whereas the above findings pertain to vision, haptic perception is also likely to play a considerable role in counterfeit detectability. Below, we provide a review of the tenets of touch with respect to counterfeit detectability.
Touching on touch
When both visual and haptic perception are available, it is likely that they will play an interactive role (Wijntjes, 2009, cited by De De Heij, 2017; Kandula, Hofman, & Dijkerman, 2015). An example of such an interaction is the rubber-hand illusion: watching a rubber hand being stroked, while one’s own unseen hand is synchronously stroked, may cause the rubber hand to be attributed to one’s own body (Tsakiris & Haggard, 2005).
The conception that haptic perception is likely to play a considerable role is further fueled by the fact that it may simply contribute a good deal of new information. Consider, for instance, that when using a banknote, people might see only one side of it, but will always feel both sides.
Haptic perception typically involves active manual exploration. In general, when exploring objects haptically, people tend to rely on their experiences with the external world of surfaces and object properties such as roughness, shape, weight, material characteristics, contour, etc. According to Lederman and Klatzky (1987), there are basically six types of haptic exploration: (1) lateral motion, typically used to explore textures; (2) pressure, to determine hardness; (3) static contact, to assess temperature; (4) unsupported holding, for judging weight; (5) enclosure for estimating size; and (6) contour tracking, to determine the shape. With respect to assessing counterfeit banknotes haptically, we hypothesize that lateral motion (exploring texture) and pressure (hardness of the surface) are the most important types of haptic exploration. Note, however, that we do not manipulate or control for the type of haptic exploration used by participants in the present study. Hence, if we are to establish a considerable role for haptic perception, we would not be able to make claims about specific strategies; (and, analogously, we will not be able to make claims about specific strategies in the visual domain).
The haptic exploration of banknotes was studied by Wijntjes (2009). This study indicated that a cash handler receiving a banknote will examine it haptically before placing it in the cash register. Usually a banknote is held between two fingers, the index finger on the reverse and the thumb on the front. The (side of the) middle finger may assist the index finger, exerting counter-pressure to the thumb. Some specific interactions are illustrated in Fig. 2. The picture on the left shows the bending of the paper, the picture in the middle shows planar movement of the thumb and the picture on the right shows the multiple contact areas. The cash handler thus perceives various banknote properties such as its structure and raised ink.
Zondervan et al. (2019) carried out an in-house study for De Nederlandsche Bank in which two “mystery shoppers” purchased a product at 30 shops and paid with an artificially modified genuine banknote. The cashier’s behavior was assessed when they were confronted with these “suspicious” banknotes. One of the findings was that approximately half of the retailers authenticate banknotes with the tips of their fingers.
Prior haptic perception research has shown that humans are very good at recognizing common objects like paper within only a few seconds on the basis of touch alone (Lederman & Klatzky, 1993). Tactile information is processed even if people do not deliberately intend to do so. According to De Heij (2017) several studies have shown that people are triggered to perform an authenticity check on a banknote that they just received when “it felt different.” In line with this, the ECB recognizes that the feel of a banknote is an important feature for detecting counterfeits. The feel includes the paper itself (“feel the banknote, it is crispy and firm”) and the raised print (“feel the short, raised lines on the left and the right edges of the banknote. The main image and the large value numeral also feel thicker”) (European Central Bank, 2013). The ink layer of the banknote is in general up to about 60 μm high. However, this height decreases when banknotes are used intensively. According to De Heij (2017) deterioration of banknotes is caused by relaxation of the paper fibers, and also by all sorts of wear and tear. Wrinkles in a banknote will create a “tactile noise level.”
A study by Raymond (2017) was designed to allow for perception testing and discrimination based on “intaglio only” (raised print). Different from the present study, the banknotes were specifically manufactured for this study. Respondents had to learn about the fantasy notes and the counterfeits were made artificially (that is, they were not removed from circulation as in the current study). Raymond et al., used three soil levels and three variants of counterfeits, similar to what, according to her, is typically seen in actual counterfeits. The results showed that sensitivity to detect counterfeit was adequate across all soil levels, even when very high-quality counterfeits were presented. Raymond concluded that tactile information affords better counterfeit detection than visual information, regardless of soil level.
Next to intaglio, the substrate or matter of the banknote is useful for authentication purposes. A 2013 cash survey by the Bank of Spain (Pérez, Guinea, & Negueruela, 2014) indicated that this was the most frequently verified security feature by both the general public and retailers. A study by Summers, Irwin, and Brady (2008) was conducted on the discrimination of 10 different types of plain paper on the basis of only a few seconds’ contact. Summers concluded that two perceptual dimensions, namely roughness and stiffness, are used to discriminate the paper. However, as with raised ink, a drawback of these factors is that they change dramatically over the banknote’s lifetime.
To quantify the potential of tactile discrimination in counterfeit detection, in one of our current experiments we included a condition in which participants could only feel the banknote. Comparing this condition with a “see-only” condition allowed us to quantify how important tactile information really is. Finally, the experiment also comprised a condition combining vision and touch, which is more similar to real-life transactions. As far as we know, multi-sensory authentication of banknotes has only been previously investigated by Klein, Gadbois, and Christie (2004). In subtests of this study, the objective was to compare inspection of banknotes using sight alone, touch alone and sight and touch combined. In the sight condition, the notes were put in plastic sleeves so that the participants could not feel them. In the touch condition the participants were allowed to touch the notes, but sight of the notes was blocked by a screen. Participants performed better when they saw the notes while being unable to touch them (yielding an 87% detection rate) than vice versa (74%). When sight and touch were combined the detection rate was, on average, 92%.
In sum, we wanted to know how well experts and non-experts are able to authenticate banknotes using different senses and how this authentication is affected by exposure time. We studied this in two separate experiments. In Experiment 1, the task for participants was to distinguish images of genuine banknotes from counterfeits by visual inspection on a computer display. In Experiment 2, participants had to discriminate physical genuine banknotes and counterfeits by only touching them or by touching and seeing them.