Our native tongue is not English. It is Greek, German and Hebrew. You can guess who speaks what language. We are very proficient in English but we did not grow up using it in any meaningful way. Though English is a second language for us, most of our professional and social life takes place in English, not our native tongue. We are comfortable using English, and except for occasionally being tripped up by the wrong preposition we are just fine. But English never felt the same as our native tongue. It is not that it feels more difficult to express things, it just feels different. And sometimes it feels like we behave differently, react differently and make different choices because we are using English. This experience, shared by millions of people who routinely use multiple languages, reflects what researchers have studied for decades. It suggests that language affects the way we think. We start this paper with a non-systematic review of the traditional literature about language and thought, and then we describe more recent research about this topic that is specific to the difference between thinking while using a native tongue compared to a second or foreign language.
Language and thought
Research on language and thought has traditionally focused on the extent to which formal aspects of a language influence people's thought (e.g. Lucy, 1992; Whorf, 1956). For example, studies have shown that in languages that use grammatical gender, the gender of a noun affects how people perceive the target entity (e.g. Bassetti and Nicoladis, 2016). For instance, the generic name for cat in Spanish, Italian, and French is masculine (el gato, il gato, le chat) whereas in Greek and German it is feminine (η γατα, die Katze). Speakers of languages for which an entity is denoted by the feminine gender tend to perceive the entity as having more female characteristics than speakers of languages for which the same entity is denoted by a male gender (e.g. Sera et al. 1994; see also Boroditsky et al. 2003).
Language structure even affects behavior. Chen (2013) suggests that the language we speak can influence our financial and health decisions. Some languages mark the future grammatically. In English we say “Today I read a book” and “Tomorrow I will read a book”. However, other languages use the same grammatical form for the future and the present tense. In German one typically says “Heute lese ich ein Buch” (“Today I read a book”) and “Morgen lese ich ein Buch” (“Tomorrow I read a book”). Chen (2013) analyzed large databases and found that the native speakers of futureless languages like German save more, smoke less, practice safer sex, and are less obese than the native speakers of languages that mark the future. Importantly, Cheng showed that these effects hold even for individuals that live in the same country and are matched in several potentially relevant dimensions (see also Sutter et al. 2015; Thoma and Tytus, 2018). These results suggest that the language we speak influences future wellbeing with futureless languages fostering more future-oriented behavior.
Notice that traditional studies on language and thought have examined how native speakers of diverse languages perceive and think differently. A problem with such studies, which their authors readily admit, is that they are susceptible to confounds because they are typically correlational. For example, German and English speakers differ in culture and in other ways in addition to their language. Any one of these differences could cause the difference in future-oriented behavior.
In this paper, we also focus on language and thought, but from a novel perspective. Instead of investigating how grammatical differences between languages influence thought, we are interested in how the nativeness of a language influences thought. We categorize languages into native or nonnative, and examine their impact on judgment and decision making (JDM). In the experiments we will describe, late-bilinguals (e.g. native Italian speakers who have learned English in a classroom context) are randomly assigned either to a native language condition (e.g. Italian) or to a non-native condition (e.g. English), they receive all instructions and materials in the language in which they have been assigned to, and their responses are recorded and compared. The experiments typically include comprehension checks to ensure that eventual language differences are not simply due to misunderstanding. Experimentally, this method is “cleaner” than traditional language-influence-thought studies in that extraneous factors that may influence the results such as culture, level of education, and socio-economic status, do not vary systematically between groups.
Although the effects that we will describe have been dubbed “foreign language effects” stressing the impact of nonnative language, we believe that they are important mainly because they shed light on what the native language does. We will argue that our emotions and socio-moral norms are entrenched in our native language (see Hadjichristidis et al. 2017; Pavlenko, 2012), and so the use of the native language involves them in judgments and decisions. In contrast, a nonnative language attenuates emotions and disengages individuals from norms, leading to judgments and choices that are less impacted by emotions and socio-moral considerations.
The rest of the paper is organized as follows. We first review evidence suggesting that the nativeness of a language influences emotions. Then, we unpack the connection between intuition, and in particular emotions, and judgment and decision making. We discuss the work of Kahneman and colleagues on heuristics and biases, according to which many of our judgments and decisions are based on emotions (e.g. Kahneman and Frederick, 2002). Next, we present evidence showing that the nativeness of language impacts JDM when emotions are at play, but less so when they are not. We primarily focus on studies on moral, risk and financial JDM, as these have been studied more extensively.
Language nativeness and emotion
Dewaele (2010) and Pavlenko (2005) in a large questionnaire-based study—The Bilingualism and Emotions Questionnaire—asked multilinguals to self-rate the perceived emotionality of their native and nonnative languages, and to make language choices to express a variety of emotions. The results showed systematic language differences. For example, people reported that the sentence “I love you” had a greater emotional weight when expressed in the native language, and the same was true for certain words with negative valence such as taboo words and swearwords. In a similar vein, Puntoni and colleagues presented trilinguals with emotional advertisements and asked them to rate how emotional they perceived them to be (Puntoni et al. 2009). They used slogans including one for a hotel (“When you are here you are family. We will leave the light on for you”). They found that the slogans elicited higher emotional ratings when they were written in the native versus nonnative language.
Studies have also reported an increased emotional intensity of the native language using physiological measures. For example, Caldwell-Harris and colleagues asked Turkish-English bilinguals residing in the US to rate the emotionality of neutral, positive, negative, and taboo words, as well as childhood reprimands (e.g. “Shame on you!”), and measured their skin conductance responses (Harris et al. 2003; see also Harris, 2004). They found higher emotional reactions in the native language, especially in the auditory modality, and particularly for taboo words and childhood reprimands (but see also, Jankowiak and Korpal, 2018). These findings were later replicated by Caldwell-Harris and Ayçiçegi-Dinn (2009) with Turkish-English bilinguals residing in Turkey. However, several studies have failed to detect a link between language nativeness and emotional intensity. For comprehensive reviews of this literature and accounts for the divergent findings refer to Caldwell-Harris (2015) and Pavlenko (2012).
A higher emotional resonance of the native language has also been shown in studies that examined behavior. For example, Bond and Lai (1986) asked pairs of participants to discuss embarrassing and non-embarrassing topics in either their native or nonnative language. They chose embarrassing and non-embarrassing topics that were matched in terms of previous knowledge, amount of interest, and perceived importance of the topic to everyday life, and compared the relative ratio of time spent on discussing the embarrassing versus non-embarrassing topic across the language conditions. They found that participants in the native language condition talked less about embarrassing topics than did participants in the nonnative language condition.
In sum, research on language nativeness and emotionality supports an emotionality advantage of the native language: certain emotional words and sentences prompt more intense emotional reactions when presented in a native rather than nonnative language (see also Ivaz et al. 2016). Note that in many of these studies emotions are linked to transgressions of cultural and social norms such as cursing or discussing embarrassing topics. We will return to this point later. But why should differences in emotional reactions prompt differences in judgment and decision making?
Intuitive and deliberative thinking
To answer this question, we turn to Kahneman's dual-process theory of judgment and decision making (e.g. Kahneman, 2003; Kahneman and Frederick, 2002). According to this theory, people have two fundamentally distinct ways of thinking: a fast way (intuition) and a slow way (reason). When we compute we use fast thinking—the answer just pops to our minds. We also use intuition when we abstain from eating a chocolate that looks like feces, or from drinking orange-juice that has been in contact with a sterilized cockroach (see the work of Rozin and colleagues on magical thinking; e.g. Rozin and Nemeroff, 2002). However, finding the solution to requires deliberate, slow thinking. We have to apply the rules of multiplication, we have to remember the numbers from the initial calculations and so forth. Thus, slow thinking is effortful and it requires attention and mental resources. In contrast, intuitive thinking just happens automatically and effortlessly, and does not tax cognitive resources.
A central element in the heuristics and biases approach of Tversky and Kahneman (e.g. Tversky and Kahneman, 1974), is the concept of heuristics. According to the latest conceptualization of heuristics within dual-process theories, heuristics refer to an attribute-substitution process (e.g. Kahneman, 2003, Kahneman and Frederick, 2002). Individuals substitute the attribute that they are asked about (target attribute) with another attribute (heuristic attribute) that is on the top of their mind at the time of judgment. Importantly for current purposes, a commonly used heuristic attribute is affective valence which is how positive or negative one feels about a target entity. Consider the question: “How risky are nuclear power plants for the American society as a whole?” The idea is that people may substitute the target attribute—in this case risk—with an attribute that is more accessible at the time of judgment—e.g. affective valence. Because “nuclear power plants” typically elicits images and thoughts that are imbued with negative affect, individuals may judge them as risky and not beneficial (see affect heuristic: Finucane et al. 2000; Kahneman and Frederick, 2002; see also the feelings-as-information theory, Schwarz, 2012).
The use of an affect heuristic has been demonstrated in diverse domains including punitive damage awards (e.g. Kahneman et al. 1998) and valuation of public goods (e.g. Kahneman et al. 1999). For example, studies on punitive damages show that the more outrage the misdeeds of a company elicit, the higher the amount of punishment and punitive damages jury-eligible individuals demand. Although the evidence is correlational, and so no cause-effect relationship can be inferred, it is consistent with the use of an affect heuristic. Indeed, according to Kahneman and Frederick (2002) affective valence is a “natural assessment,” that is, it is an attribute that is highly accessible at all times, and therefore a good candidate for a heuristic attribute. This idea has been highlighted by Zajonc (1980): “We do not just see “a house”: We see a handsome house, an ugly house, or pretentious house.” Other natural assessments include physical attributes such as size and certain abstract attributes such as similarity.
Besides punitive damage awards and public good valuations, emotions also influence moral judgments. According to the social intuitionist model of moral judgment, our intuitions determine moral judgments directly—with reasons entering the judgment process later, if ever, in the form of post-hoc rationalizations (e.g. Haidt, 2001). Consider reading a story about a brother and a sister who decided to make love. They used contraceptives, they enjoyed the act, as a result they feel closer to one another, but decided not to do it again. What do you think about that? Was it OK for them to make love? Many people judge the action as morally wrong without being able to properly justify the basis of their judgment, an effect known as “moral dumbfounding” (but see Royzman et al. 2015). Notice the similarity between this hypothesis and the affect heuristic. Individuals are asked one question—Is this [morally] Ok?—but instead answer another question—How do I feel about it?—and translate their response back to the target attribute scale.
In sum, a great deal of our judgments and decisions are based on intuitive thinking, and a central element of intuitive thinking is affective valence or emotions. In this section, we reviewed a limited number of examples that demonstrate the link between emotions and JDM. In the next section, we will unpack more examples while we discuss research showing effects of language nativeness on judgments and choices.
Language nativeness and JDM
Having discussed the link between language nativeness and emotion, and emotion and JDM, we now turn to the evidence regarding language nativeness and JDM. we divide the research into studies that examined the influence of language on JDM phenomena that are believed to be based on affect or emotion and studies that have examined its influence on JDM phenomena that are not based on affect. To foreshadow the results, the research suggests that language nativeness impacts JDM in the former case but not in the latter. The main exception to this rule is that language nativeness affects judgments that are relatively non-emotional but concern violations of established socio-moral norms.
Emotion-based JDM
Moral judgments. The first attempts to examine whether the nativeness of a language impacts moral judgment examined dilemmas that involve a conflict between respecting the rights of an individual, a deontological concern, and doing what's best for the majority of people, a utilitarian concern. Consider the famous footbridge dilemma, which involves a runaway trolley (Foot, 1978; Thomson, 1985). Participants are informed that if they do nothing, five people that happen to be on the rail tracks will die. But if they push a large person off a footbridge and into the path of the trolley, the trolley would stop, the large person would die, and the five people would be saved. Is it permissible to push the large person? The results of numerous studies have shown that the majority of participants select the deontological-type response—they respond that it is not permissible to push the large person (e.g. Hauser et al. 2007). Further studies have shown that this response is based on emotions, as participants with damage in the emotional areas of the brain show more “utilitarian” responses (Koenigs et al. 2007; see also interpretation of these findings in Greene, 2007). These studies typically use the native language.
Yet recent studies show that nonnative language use increases the rate of utilitarian responses to the footbridge dilemma (Cipolletti et al. 2015; Costa et al. 2014a; Geipel et al. 2015a; Hayakawa and Keysar, 2018; Shin and Kim, 2017). Importantly, using a nonnative language had no effect on a variant of the footbridge dilemma, the so called “switch dilemma” that is formally the same as the footbridge dilemma but does not evoke a strong emotional reaction. This dilemma also involves sacrificing one person to save five, but instead of pushing the large person, the action involves pushing a lever that would move the trolley into alternative tracks where an innocent bystander is standing. The fact that the language manipulation made a difference for the emotional but not for the non-emotional variant suggests an important role for emotions in the effect of language. While this is suggestive, these studies did not directly examine the mechanism behind the moral foreign language effect. Specifically, it could be that people are more likely to sacrifice one to save five in a foreign language because the nonnative language reduces emotions and intuitive thinking, increases systematic thinking, or both.
To answer to this question, two main approaches were developed. One involved gathering ratings of emotionality in response to the scenario (Geipel et al. 2015a). Geipel et al. (2015a) found an intermediate effect of language on moral judgment and a small effect of language on emotion. However, they did not find that emotion mediates the language impact on moral judgment. That is, the effect of language on moral judgment was independent of the effect of language on emotions. Similar results were obtained in subsequent studies (e.g. Chan et al. 2016).
A second way in which this issue was addressed, was by using a technique that purportedly dissociates between deontological and utilitarian responses (Conway and Gawronski, 2013). Specifically, Hayakawa et al. (2017) presented participants with trolley-like scenarios in which deontological and utilitarian concerns pull judgments in opposite directions such as killing one person to save five, but also to scenarios where they point in the same direction such as killing one person to prevent injuring another. Under certain assumptions, the responses to such problems can help differentiate between deontological and utilitarian judgments, while controlling for ‘weird' responses (e.g. killing a person for no reason). An analysis of the results showed that the nonnative language decreased deontological responding (small effect) but did not increase utilitarian responding (see also Muda et al. 2018).
Besides trolley-type dilemmas, studies have also examined how people respond to behaviors, such as incest, that people judge as morally wrong due to the disgust these behaviors elicit. Geipel et al. (2015b) showed that the use of a nonnative language decreases moral condemnation for such actions. To the extent that moral condemnation is based on affect, an automatic revulsion triggered by the action (see Haidt, 2001), these findings are consistent with the hypothesis that nonnative language acts through reducing emotional intensity. This paper included two experiments that assessed this hypothesis directly by gathering ratings of emotionality associated with reading each scenario. The results were mixed: One experiment found the predicted mediation, while the other did not.
Geipel et al. (2015b) also presented participants with a number of violations of social norms and conventions such as failing to keep minor promises, cutting in line when in a hurry, and keeping excess change at a store. Analyses of the data showed that the use of a nonnative language reduced condemnation also for these actions. These results, together with the lack of the mediation, suggest that the crucial element driving the language effects may not be emotion per se, but rather whether an action violates an established social, cultural, or moral norm.
Moral behavior and temptation. Besides moral judgment, studies have investigated the role of language in dishonest behavior (Bereby-Meyer et al. 2018; see also Alempaki et al. 2017). For example, Bereby-Meyer and colleagues asked native speakers of Hebrew, Korean, Spanish, and English in a series of lab and online studies to privately roll a dice three times, and then report the number that came up on the first roll. The higher the number they reported, the higher their earnings. This provided an incentive to cheat with no fear of consequences. On average participants cheated more when they used their native language as compared to the nonnative language (intermediate effect). The reduction in dishonest behavior in the non-native language might be driven by reduced temptation (Hadjichristidis et al. 2017; see Greene and Paxton, 2009).
The idea that temptation is reduced in a foreign language is in line with the results of a study on eating behavior. Klesse and colleagues (Klesse et al. 2015) presented native German speakers who were dining in a restaurant with an option between a tempting desert (“Chocolate Mousse”) and a less tempting desert (i.e. “Fruit Salad”). The deserts were either labeled in the diner's native language or a nonnative language. They found that nonnative language labels increased the selection of the healthy fruit salad (small to intermediate effect). One way to interpret these findings is that the use of a nonnative language reduced the temptation of the hedonic food, as it reduced the temptation to cheat.
Framing effects. Framing effects refer to people's tendency to respond differently to logically equivalent descriptions of the same situation, depending on whether the description highlights positive or negative aspects. Consider the classic Asian disease problem, in which participants are informed that an Asian disease is about to strike their home country and anticipated to kill 600 people. They are asked to choose between two plans to combat the disease, one provides a certain outcome and the other a probabilistic outcome. These two options are also framed either as potential gains or as potential losses. Kahneman and Tversky (1979) found that the majority of people in the gain frame prefers the sure option, whereas in the loss frame the majority prefers the probabilistic option. So even though the two frames present choices that are formally identical, people become risk averse with the gain frame and risk seeking with the loss frame.
The framing effect has been associated with increased activation of the amygdala (De Martino et al. 2006) suggesting that it results from a strong emotional attraction to sure gains and a strong aversion to sure losses (Kahneman and Frederick, 2006). To the extent that the use of a nonnative language attenuates emotions, one should find a reduced susceptibility to framing effects when using a nonnative language. This is exactly what Keysar et al. (2012) found. In particular, they found a strong framing effect in the native language, which was reduced in the nonnative language. These results were replicated in a number of subsequent studies (e.g. Costa et al. 2014b; Winskel et al. 2016; but see Oganian et al. 2016).
Financial decisions. Keysar et al. (2012) also gave participants the chance to play gambles whose monetary expected value was positive—as in an equal chance of winning $6 or losing $4. Although these are favorable gambles, psychological research suggests that losses loom larger than gains, a phenomenon known as “loss aversion” (Kahneman and Tversky, 1979). To the extent that language attenuates emotions, the authors expected that it would reduce loss aversion and, through that, increase the acceptance of such gambles. This is exactly what they found (small to intermediate effect). These results were replicated by Costa et al. (2014b; but see Hayakawa et al. 2019).
Interestingly, a different study replicated the results with the framing effects with scenarios but not with gambles that involved just numbers (Winskel et al. 2016). Winskel et al. (2016) argued that there is a major experimental difference between their financial-decision making paradigm and the framing-effect paradigm. The latter one being language-laden whereas the former one language poor as it involved mainly the processing of visual cues such as a diagram with red and green colors (De Martino et al. 2006). Taken together, these findings provide evidence for the hypothesis that the effect concerns the role of language. When a task is language-laden the language effect was obtained, when it was not there was no effect.
Risk and benefit perception. Studies have also addressed whether language nativeness affects how people perceive risk and benefit associated with various technologies and activities. According to the affect heuristic, people frequently judge risks and benefits based on the overall affect that the presentation of a stimulus elicits (e.g. Slovic et al. 2002). If the overall affect is positive, people will perceive the target stimulus to be low in risk and high in benefit; if it is negative, they will perceive it to be high in risk and low in benefit (see Alhakami and Slovic, 1994; Finucane et al. 2000).
Hadjichristidis et al. (2015) reasoned that if a nonnative language attenuates emotions, and in particular negative emotions (e.g. Sheikh and Titone, 2016; Wu and Thierry, 2012), then the same stimulus would evoke overall more positive affect when described in the nonnative rather than the native language. Indeed, the use of a nonnative language evoked higher benefit and lower risk judgments (intermediate effect). A mediation analysis also revealed that the effect of language on risk/benefit judgments was mediated by affect.
Magical thinking. Research suggests that magical thinking is rooted in intuition. For example, studies have shown that people's score on cognitive style scales, such as the Cognitive Reflection Test (Frederick, 2005), is related to their magical thinking and belief in the paranormal (e.g. Pennycook et al. 2012). As a result of years of cultural conditioning, bad luck superstitions come to trigger negative affect whereas good luck superstitions prompt positive affect. Hadjichristidis et al. (2019) presented participants with scenarios involving typical bad luck and good luck superstitions. For example, in one scenario, participants read: “Imagine that you have to submit an important job application. On the day of the submission, your mirror in the bathroom breaks. How would you feel about submitting your job application under this condition?” The results showed that nonnative language use reduced the negative feelings associated with bad luck scenarios, reduced the positive feelings associated with good-luck scenarios (small effect), while it had no impact on scenarios that had emotionally neutral content. These results suggest that the use of a nonnative language can reduce emotional reactions that are linked to superstitions elements.
Consumer decisions. Studies have also shown that the use of a nonnative language can increase people's willingness to consume sustainable products that are disgusting (Geipel et al. 2018). Specifically, Geipel and colleagues tested participants' willingness to consume recycled water, artificial meat and cookies made of mealworm flour. After reading a brief description of each product in which the link to a disgusting entity was emphasized, participants were asked whether they would consume it. For all these products, the nonnative language descriptions increased participants' willingness to consume them (small effect). In the study on mealworm cookies, following the willingness to consume question, participants were asked to rate how disgusted they felt towards mealworm cookies. We found that the use of a nonnative language decreased ratings of disgust, and that feelings of disgust mediated the association between language and willingness to consume. A further study on recycled water provided evidence that the nativeness of a language can also influence real consumption, suggesting that language interventions can be put to the service of sustainability.
Nonemotion-based JDM
Although we have interpreted many of the results as supporting the idea that nonnative language acts through attenuating emotional reactions, the possibility remains that it acts through triggering systematic thinking (see, Keysar et al. 2012). Selecting a utilitarian response in the footbridge dilemma or accepting favorable gambles can also reflect rational thinking. There are many ways of distinguishing between these accounts. For example, one could examine the effect of language nativeness on tasks that are purely logical such as performing mathematical calculations. If using a nonnative language increases deliberation then we should find an improved performance in such tasks; if it does not, then we should observe no language differences. A related question is whether nonnative language reduces intuitive thinking in general, or solely intuitive thinking that is associated with emotions. Recent research has addressed both questions.
One of the first attempts to investigate whether the nativeness of a language influences performance on purely logical problems was a study by Takano and Noda (1993). These authors asked participants to perform two tasks concurrently–a main task that involved performing calculations and a secondary task that involved answering questions that were orally presented to them either in their native or nonnative language. Examples include open-ended questions such as: “What is the name of the president of the United States?” The authors found that participants performed worse in the main task, when the secondary task was presented in the nonnative rather than in the native language. They explained this finding in terms of a higher cognitive load associated with processing information in a nonnative language. However, note that this evidence is indirect—the language manipulation pertained to the secondary task. Below we focus on research that applied the language manipulation to the main task.
Costa et al. (2014a; see also Turula, 2016) examined participants' responses to the CRT task by manipulating the language in which the task was presented. This task was developed by Shane Frederick (Frederick, 2005) and contains three tricky problems. One of them is: “A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost?____cents.” Each of these problems has a strong intuitive response; the intuitive response for the problem above is 10 cents. However for all of these problems the intuitive response is wrong. If the price of the ball were 10 cents, then the bat would cost $1.10, and thus bat and ball together $1.20. The explanation is that people frequently go with their first intuitive response without monitoring it—monitoring is assumed to be a function of the slow, deliberative system (Kahneman and Frederick, 2002). Contrary to the hypothesis that using a nonnative language triggers systematic thinking, both Costa et al. (2014b) and Turula (2016) found no effect of language condition on the CRT.
This issue was also addressed with the “Moses illusion” task (see, Erickson and Mattson, 1981). Geipel and colleagues (Geipel et al. 2015b) informed participants that they will be presented with a number of trivia questions some of which might be ill-formed such as “Garfield is the dog of which cartoon?” This question is ill-formed because Garfield is a cat. Participants' task was to provide an answer, or say “can't say” if the question is ill-formed. After responding to the well-formed question, “Which country is famous for cuckoo clocks, chocolate, banks, and pocket knives?” they were asked the ill-formed question: “How many animals of each kind did Moses take on the Ark?” While the intuitive answer is “2,” the correct answer is “can't say” because the biblical character was Noah, not Moses. The results showed that the use of a nonnative language marginally reduced the number of correct responses (from 35% to 16%).
But perhaps the most systematic attempt to date to examine whether a nonnative language improves people's performance in tasks that involve intuitions but are not emotion-based, is a recent study by Vives and colleagues (Vives et al. 2018). These authors tested whether using a nonnative language reduces the outcome bias which is people's tendency to evaluate an action based on the outcomes of that action, the conjunction fallacy which refers to people's tendency to assign higher probability to a conjunction of entities rather than to one of its constituents, and the base-rate fallacy which is people's tendency to ignore base rates when making probability judgments. The cases of conjunction fallacy and base-rate neglect they studied are believed to be driven by the representativeness heuristic, which refers to the tendency of judging the likelihood of an event based on similarity. Despite the large sample sizes, they found no effect of language nativeness in any of these tasks. If anything, the results hint towards a tendency in the opposite direction: using a nonnative language slightly reduced performance, a result that the authors tentatively attributed to cognitive load.
General discussion
Summary of findings
Taken together, the findings are consistent with the idea that language nativeness influences JDM mainly in tasks that involve emotions or socio-moral norm transgressions. Compared to the native language, a nonnative language promotes more utilitarian, cool-headed responses towards certain moral dilemmas, less condemnation of moral and social violations, decreased susceptibility to framing effects, higher propensity to gamble on favorable bets, less magical thinking, and higher intended and actual consumption of repulsive sustainable products. Importantly, the nativeness of a language does not affect JDM in cases where performance depends on reflective thinking or on biases that are not emotion based. It did not affect performance on emotion poor problems such as the CRT, the Moses illusion, base rate neglect, and the conjunction fallacy. If anything, for these tasks, the results hint to an effect in the opposite direction (but see Díaz-Lago and Matute, 2019).
Explanations
But why does a nonnative language dampen emotional reactions? One explanation is that memory is language dependent. When we experience things, we often experience them in a linguistic context, in a context where a specific language is used. The idea is that memories keep traces of the linguistic context. If the linguistic context is the same at retrieval as at encoding, then memories would be more likely to be retrieved along with their associated emotional content. Evidence for language dependent memory comes from studies on bilinguals (e.g. Marian and Neisser, 2000; Schrauf and Rubin, 2000). For example, Marian and Neisser (2000) asked Russian native speakers who later immigrated to the United States to recall specific life events. Participants retrieved more experiences from a period of their lives (Russian-speaking vs. English-speaking), when the language of the interview matched the language of that period.
Another related explanation is that nonnative language use is typically acquired and experienced in emotionally neutral contexts. Indeed, in most of the studies we have discussed, nonnative language was defined as a language that an individual has acquired in a classroom context. The idea is that language gains emotionality by virtue of being used in emotional contexts (see Harris et al. 2006; see also Dewaele, 2004). For example, when we hear swearwords and at the same time experience fights and broken noses then these words gain emotionality. Thus, when these words are described in the native language some of this emotionality would be activated and would impact judgments and decisions. Words in a foreign language do not possess such emotional grounding.
The language-dependent memory account can explain language differences involving socio-moral violations and superstitions directly, bypassing an appeal to emotions. Socio-moral rules and superstitions are usually acquired at a young age in a context where the native language is spoken from primary caregivers and peers (e.g. for moral rules see Nichols et al. 2016; for superstitions see Vyse, 2014). Because such norms or rules are learned in a native language context, the language effect in these cases could be explained in terms of increased norm activation when the native language is used. Of course, emotions and norms often go hand-and-hand as breaking a norm is usually associated with negative affect, so the two accounts are not always easy to separate.
Several alternative accounts have been proposed to explain foreign language effects. We briefly discuss and comment on them. Most of these accounts in isolation provide inadequate explanations for the effects of language nativeness. Yet, we suggest that the cause behind effects of language nativeness is multidimensional. That is, using a nonnative language has a number of psychological consequences: It disengages emotions and socio-moral norms, it introduces cognitive load, and so on. The effect of language nativeness, then, may involve multiple pathways.
Disfluency
It is possible that the use of a nonnative language triggers a metacognitive perception of disfluency which, in turn, can explain the increase in the apparent “rational” responding. Although some studies support that disfluency can influence moral judgment in a way similar to how a nonnative language influences moral judgment (e.g. see Spears et al. 2018), other studies support a less direct connection between fluency and moral judgment. Specifically, Laham et al. (2009) found that disfluency can reduce moral condemnation of disgusting actions, like the ones used by Geipel et al. (2015b), but only when moving from a disfluent to a fluent font or vice versa. That is, these authors did not find a main effect of fluency on moral judgment but rather an effect of discrepancy in terms of fluency (for a parallel effect of language nativeness on framing effects, see Oganian et al. 2016).
Psychological distance
Psychological distance may also account for language nativeness effects. However, psychological distance is a metaphor—thus, on its own, it cannot explain anything. To test this claim one needs to articulate what this term means. For example, psychological distance is frequently associated with level of construal: psychological closeness is associated with a concrete mindset, whereas psychological distance with an abstract mindset (Trope and Liberman, 2010). Research provides mixed results as to how level of construal influences moral judgment. For example, some studies have found that a more abstract level of construal is associated with higher moral condemnation of morally questionable actions (Eyal et al. 2008), whereas others found the opposite (Gong and Medin, 2012). Moreover, the pattern of the results obtained in studies that used similar items while manipulating language nativeness (Geipel, 2015) does not converge with the results of either of these studies.
Furthermore, unpublished data finds no evidence that language nativeness influences level of construal (Geipel, 2015; Hayakawa, 2014). In these studies, participants were presented with the Behavioral Identification Form (Vallacher and Wegner, 1989), which contains a series of target actions such as reading an article, followed by two restatements: a concrete restatement that emphasizes how the action is performed (e.g. following lines of text), and an abstract restatement that emphasizes why it is performed (e.g. gaining knowledge). For each target action, participants have to choose the restatement that best captures it. If a nonnative language promotes a more abstract mindset, then it should increase the selection of abstract restatements. No such difference was found between the use of a native and a nonnative language.
Cognitive load
Another explanation concerns cognitive load (see, e.g. Volk et al. 2014). The claim that processing a nonnative language increases cognitive load is uncontroversial (see Hasegawa et al. 2002). Furthermore, sometimes using a nonnative language has similar consequences in JDM tasks as increasing cognitive load (e.g. Costa et al. 2019; Geipel et al. 2016). However, most of the effects we described cannot be explained in terms of an increased cognitive load. For instance, an increase in cognitive load has been shown to decrease utilitarian judgments (Greene et al. 2008) and to increase risk aversion (Benjamin et al. 2013)—this is the exact opposite of what using a nonnative language does (Keysar et al. 2012).
Mental imagery. Another account that might help explain certain effects of language nativeness holds that the use of a nonnative language reduces mental imagery. For example, Hayakawa and Keysar (2018) presented participants with triplets of words such as: carrot, pen, mushroom. In some trials the participants had to identify the odd item in terms of shape (here, mushroom), while in others in terms of category (here, pen). The authors predicted worse performance in the nonnative language condition but particularly for the task concerning shape, as this task involves mental imagery. This is exactly what they found. The authors replicated the nonnative language effect with the footbridge dilemma, and showed that this effect is partly associated to differences in the ability to visualize the large man standing on the footbridge.
Future directions
Although the main account is that language nativeness impacts judgment and decision making through reduced emotional reactivity, the evidence to date is limited. A reason for this is that several studies assessed emotions through self-ratings, using emotional scales in the language in which a participant was assigned to. A problem with inferring emotionality from self-ratings is that people rate the emotionality of the same experience differently depending on whether the anchors of the emotional scale are written in the native or nonnative language (de Langhe et al. 2011). In order to circumvent this problem, other studies have used physiological measures such as pupil dilation, which is positively associated with emotional reactivity (García-Palacios et al. 2018). However, pupil dilation is also positively correlated with processing difficulty, which is increased when processing information in a nonnative language. Future studies may benefit from exploring different methods of assessing emotionality such as facial expressions and brain imaging.
Besides honing on the mechanism, another suggestion is to examine whether these effects generalize to experts. Many researchers, including us, motivate the importance of studying language nativeness effects by pointing to the potentially huge real life implications such as in diplomacy and policy. However, to date, no such evidence exists. Would diplomats think differently about issues depending on whether the issues are presented in their native or foreign language? For example, would diplomats become more utilitarian when deciding on issues that involve short-term sacrifices for long-term future prosperity?
Also, the research so far has concentrated on intrapersonal decisions. Future studies could examine the role of language nativeness in an interpersonal context. For example, although reading a person’s CV in a nonnative language may activate less stereotype thinking (see Danziger and Ward, 2010), it could lead to more bias when one actually interviews the candidate. Indeed, research in international business shows that people use a person’s proficiency to infer competency. The use of a nonnative language in an interpersonal context can also have other adverse consequences. For example, in another study employees stated that they would be less likely to speak up during board meeting for fear of sounding silly (Piekkari et al., 2015). However, most of these studies involve structured interviews and so do not allow inferences about causality. It would be important to supplement these findings with experiments.
Concluding remarks
This paper reviewed evidence supporting the idea that language nativeness can sway judgments and decisions that involve emotions or socio-moral violations. The findings suggest that the use of a nonnative language attenuates emotions and disengages people from culturally learned norms. We think that a more fruitful way of looking at these findings is in terms of what they tell us about the native language: the native language is not a simple carrier of meaning but allows our emotions and socio-moral norms to have an impact on decisions. Sometimes the end result of this is positive, if we are more respectful of others, but at other times it is negative, if we are more prone to magical thinking and implicit stereotyping.
For us personally, the conclusion is that the choice of language is always consequential. When we make decisions in our emotion-laden native tongue, Greek, German or Hebrew, it might protect us from taking high risks we might eventually regret, but it might also cause us to forgo pursuing risky opportunities that could benefit us in the long run.