Dishonesty is a pervasive human behaviour that has greatly increased social costs in the economic and financial fields (Ederer and Fehr, 2007). Common examples of these increased costs include Internet fraud, corruption and tax evasion. Dishonesty usually manifests as the disposition to lie, cheat, fraud or deceive (Bucciol and Montinari, 2019). Lying is regarded as a form of dishonesty with the aim of instilling false belief in the victim and doing so intentionally by asserting what one believes is false (Adler, 1997). Understanding how people make decisions about whether to be dishonest has important implications for individuals and society.

With increases in international cooperation, communication between people who speak different languages is becoming more prevalent. Thus, decision making and moral judgement in foreign language environments are becoming the new normal. The first study to investigate the process differences between decision making in one’s native language (NL) and that in a foreign language (FL) was conducted by Keysar et al. (2012). They found that thinking in a FL reduces decision bias, and they named this phenomenon the FL effect. Inspired by this finding, many subsequent studies found that the FL effect is robust and is not restricted to risky decision making (Costa et al., 2014a; Gao et al., 2015; Hadjichristidis et al., 2015; Winskel et al., 2016) but also extends to decision making in other areas, including moral judgements (Cipolletti et al., 2015; Geipel et al., 2016; Hadjichristidis et al., 2017). People are more inclined to make utilitarian choices under conditions that require FL use (Costa et al., 2014b; Cipolletti et al., 2015; Corey et al., 2017). Researchers refer to this phenomenon as the moral FL effect.

Although research on the perceived trustworthiness of people speaking in their native or non-native language is fairly common (Cheng and Broadhurst, 2005; Lev-Ari and Keysar, 2010; Evans and Michael, 2014; Li, 2017), only a small number of studies has focused on the FL effect on lying and whether people lie less in a non-native language than in a native language (Suchotzki and Gamer, 2018; McDonald et al., 2020). Most of these studies compared deception behaviour in a NL condition versus FL condition, but with mixed findings. Some researchers observed that deception was more successfully and accurately detected in a NL than a FL (Da Silva and Leach, 2013; Leach and Da Silva, 2013; Elliott and Leach, 2016; Akehurst et al., 2018; Bereby-Meyer et al., 2020), whereas some researchers found the opposite to be true (Evans et al., 2013, 2017). Other research revealed no difference in the two language conditions (Cheng and Broadhurst, 2005; Caldwell-Harris and Aycicegi-Dinn, 2009; Evans and Michael, 2014; Duñabeitia and Costa, 2015; Alempaki et al., 2020).

Two research theories, cognitive load and emotional distance, can explain the differences between lying/truth telling in a NL and that in a second language. However, the two theories lead to competing conclusions.

Lying is cognitively more demanding than truth telling (Zuckerman et al., 1981; Vrij et al., 2008). The cognitive load associated with communicating in a FL is especially burdening for lying, and increasing cognitive load seems to make lying more difficult (Blandón-Gitlin et al., 2014), thereby further increasing the difference between lying and truth telling (Vrij and Granhag, 2012). Studies have found that people respond slower to lying in a FL than to lying in their NL, and the latter results in a stronger skin conductance response (Suchotzki et al., 2015; Suchotzki and Gamer, 2018), greater pupil dilation and a longer duration of pronunciation (Duñabeitia and Costa, 2015). Nonetheless, although most research has supported the cognitive load hypothesis, Blandón-Gitlin et al. (2017) argued that lying is not always more cognitively difficult than telling the truth.

The emotional distance hypothesis predicts that the reduced emotionality associated with FL use may facilitate lying. Emotional stimuli elicit less pronounced autonomic nervous system activity when presented in a FL (Harris et al., 2003). Moreover, advertisements are judged to be less emotional when presented in a FL (Puntoni et al., 2009). People are less aroused by emotionally laden expressions, such as childhood reprimands or taboo words in a FL than in their native tongue (Colbeck and Bowers, 2012). Diminished emotional arousal can facilitate lying and thereby reduce behavioural and autonomic differences between lying and truth telling.

Although they support different conclusions, the theories of cognitive load and emotional distance are not independent of each other. Some researchers found smaller RT differences between lying and truth telling in a FL compared to the NL of individuals (Suchotzki and Gamer, 2018; Frank et al., 2019). Their result could be explained by the antagonistic effect of cognitive load and emotional distance on lying, namely, the stronger emotional distance ‘cancelling out’ the increased cognitive load while lying in the FL (Suchotzki and Gamer, 2018).

Based on two experimental behavioural studies, this paper explores whether different languages modulate lying behaviours. In the first study, we used a die-rolling task similar to that of Fischbacher and Föllmi-Heusi (2013) and Bereby-Meyer et al. (2020) to investigate the FL effect on lying behaviour. In the task, each participant was asked to report the outcome of a six-sided die roll that only the participant who rolled the die could see. Their payoff depends on the reported roll of the die. The participants have an incentive to lie and report higher numbers to get a higher payoff. This methodology could measure the propensity of people to be dishonest. Then, in the second study, participants were required to take part in a cheap-talk sender-receiver game proposed by Gneezy (2005), where two participants were randomly paired into groups where the sender could benefit more from telling lies to the receiver.

Two different tasks were adopted separately in the two studies, as they can complement each other in comparing lying behaviour in different languages. In the die-rolling task, since each number outcome should happen with the same probability (1/6), we can compare the reported die-roll outcome at the aggregate level with the 1/6 benchmark to infer the propensity to lie at the group level, thus considering the NL and FL groups separately. Unlike the die-rolling task measuring aggregate-level dishonesty, the cheap-talk sender-receiver game enables us to utilise individual-level data to analyse lying behaviour (Gao et al., 2018). In addition, social interaction is involved in the cheap-talk sender-receiver game. Therefore, guilt, nervousness and other emotions related to lying might be more easily aroused by the cheap-talk sender-receiver game than by the die-roll task.

Based on the cognitive load hypothesis, using a FL will increase cognitive load and make lying more difficult. We may observe less cheating in the FL condition than in the NL condition in our two studies. However, according to the emotional distance hypothesis, using a FL will reduce emotionality associated with lying, e.g. guilt. We can predict that participants will cheat more in the FL condition than in the NL condition, especially in Study 2.

We will test those two hypotheses with opposite predictions about the FL effect on dishonesty. Our studies may offer a contribution by providing more reliable evidence of this effect. Previous studies on the FL effect on dishonesty have generally yielded competing conclusions, possibly because the results are based on different tasks. In a recent study, Bereby-Meyer et al. (2020) asked participants to roll a die and self-report the results to receive compensation. Using a FL (Hebrew, Korean or Spanish) to report the total die roll resulted in significantly lower earnings than using one’s native language (English), which suggests that speaking in a FL reduces lying behaviour. By contrast, Alempaki et al. (2020) designed a performance-difference-reporting game in which participants could inflate their relative performance in a real effort task, and they did not find a FL effect in lying behaviour. In their experiment, the FL was English, and the NL was Chinese or German. In the present study, we investigate whether we can obtain consistent results in two different typical tasks under the same language pairs, namely, Chinese and English.

Study 1 revealed the difference in lying behaviour between the NL and FL conditions. However, with the design of Study 1, we can obtain the distribution of reported numbers revealed at the group level, but we cannot determine whether a subject lied or not. Besides, guilt is one of the emotions related to dishonesty and guilt a version will decrease lying behaviours. However, in the die-roll game, this kind of emotion might not be aroused, as participants may feel that they play the game alone and their choices have no effect on others. To better understand why speaking a FL reduces lying, we designed a cheap-talk sender-receiver game in which dishonesty could be observed at the individual level and there is an interaction between the participants and their counterparts, namely, the interaction between the senders and the receivers.

With the integration of the global economy, an increasing number of people need to work or live in an environment where a FL is spoken. It is of great significance to understand people’s decision-making behaviour in FL conditions. This paper explores the FL effect of lying behaviour, which could be linked to some activities in international trade or international finance. For example, in trading, sellers aim to over-report the quality of their products to earn more profit.

To answer the question of whether people behave differently under different language conditions, we investigated the behavioural characteristics and brain mechanisms of sophisticated lying among college students. We designed a simple die-roll task in Study 1 and a cheap-talk sender-receiver game in Study 2. Although the tasks differed in the two experiments, both sets of results revealed that fewer participants chose to lie when they used a FL than when they used their NL.

According to the emotional distance hypothesis, using a FL reduces emotionality associated with lying, e.g. guilt. Hence, people lie more when using a FL. However, our results provide some evidence to support cognitive load theory, which proposes that the cognitive load associated with thinking in a FL is a burden that makes lying more difficult. We observed less lying behaviour when the experiment was conducted in a FL. In Study 2, the results reveal that participants needed more time to decide to lie under the FL condition than under the NL condition. However, there was no significant difference in the number of honest decisions under these two conditions. These results might imply that participants had more difficulty lying in a FL than in their NL. This result is consistent with the findings of Bereby-Meyer et al. (2020). The discovery that people are more honest in a FL is predicted by the cognitive load hypothesis in that lying typically requires greater mental effort than telling the truth. Using a FL can increase cognitive load, thereby increasing the difficulty of lying.

Some limitations of our study should be noted. First, we adopted a between-subjects experimental design, and individual differences cannot be completely excluded. Second, due to the limitation of the participant group and experimental cost, the conclusion of this paper is applicable only to cases in which English is the FL.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics committee of Key Laboratory of Applied Brain and Cognitive Sciences, Shanghai International Studies University. The patients/participants provided their written informed consent to participate in this study.

XY, LL, and RL designed the experiment, wrote the manuscript, revised the manuscript, and finally approved the version to be published. LL performed the experiment and drew the figures. LL and RL analysed the data. All authors contributed to the article and approved the submitted version.

This work was supported by the National Natural Science Foundation of China (Project71873089) and National Social Science Fund of China (Project 15CGJ023).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.