People are more likely to believe things that are easier to process; foreign-accented speech is relatively difficult to process & people believe information less when it is delivered in a foreign accent rather than a native accent

Exposing Individuals to Foreign Accent Increases their Trust in What Nonnative Speakers Say. Katarzyna Boduch-Grabka, Shiri Lev-Ari. Cognitive Science 45 (2021) e13064, Nov 2021. DOI: 10.1111/cogs.13064

Abstract: People are more likely to believe things that are easier to process. Foreign-accented speech is relatively difficult to process, and prior research shows that, correspondingly, people believe information less when it is delivered in a foreign accent rather than a native accent. Here we show that a short exposure to foreign accent can reduce this bias, and that the reduction in bias is due to improvement in the processing of the accent. These results demonstrate how cognitive aspects of language processing can influence attitudes. The results also suggest that ensuring exposure to foreign accent can reduce discrimination against nonnative speakers.

Keywords: Accent; Fluency; Credibility; Communication

1.2. Processing fluency and accent

As described above, there is ample evidence that stimuli that are easier to process are perceived more favorably. This finding could have grave implications for interactions between native and non-native speakers. Languages differ in their sound inventories. This can lead non-native speakers to find it difficult to produce certain sounds in their second language because these sounds do not exist in their first language, or because the sounds are not contrastive in their first language. For example, Japanese speakers often produce English ‘l’ in a nonstandard manner that can be confused with ‘r’ because Japanese does not distinguish between these sounds (Bradlow, Pisoni, Akahane-Yamada, & Tohkura, 1997). Nonnative speakers might also use nonstandard stress or prosodic patterns because these differ from those in their native language (e.g., Magen, 1998). These deviations from standard production render foreign-accented speech harder to process even when it is fully understood (e.g., Munro & Derwing, 1995). The greater difficulty of processing foreign-accented speech could lead individuals to treat it less favorably. Indeed, when Lev-Ari and Keysar (2010) presented participants with trivia statements recorded by native versus foreign-accented speakers of a variety of foreign accents (e.g., Korean, Italian, Turkish), participants judged the statements as more likely to be true when they were produced by native speakers. This effect was obtained even though it was highlighted to participants that speakers were merely reading aloud sentences provided by the experimenter, without knowing whether these were true or not. In other words, all speakers were merely messengers without any control over the content of their speech, thus reducing the relevance of any prejudice as it is not the speaker’s honesty and knowledge that are evaluated. The lower belief in foreign-accented statement was hypothesized to be due to the greater difficulty in processing them.

In a follow up study, Lev-Ari and Keysar (2010) tested whether raising participants’ awareness to the source of their difficulty reduces the tendency to believe nonnative speakers less. Their prediction rested on literature that shows that raising individuals’ awareness to the source of an affective experience can prevent individuals from misattributing the affective experience to other sources (Schwarz & Clore, 1983). To test whether raising awareness to the source of the difficulty would eliminate its misattribution to lower credibility, Lev-Ari and Keysar (2010, exp. 2) asked participants to rate the relative ease of understanding each of the speakers, in addition to rating the veracity of the statements. This manipulation succeeded in reducing the bias against mildly accented speakers, but not against heavily accented speakers. It seems, then, that participants corrected for the difficulty of processing, but not sufficiently so for the heavily accented speakers. The reduction of the bias against mildly accented speakers once awareness to the difficulty was raised also provides some support for the hypothesis that the lower belief in foreign-accented speech is at least partly due to misattribution of the lower processing fluency. Since the publication of those studies, several studies examined whether people believe information delivered by nonnative speakers less. Several of these found supportive evidence. One of these studies demonstrated that it is not only native speakers, but also nonnative speakers, who believe information more when it is provided in a native rather than a foreign accent (Hanzlíková & Skarnitzl, 2017). Other studies examined people’s ability to detect when native and nonnative speakers tell the truth and when they lie. All these studies found that people have a truth-bias when evaluating native speakers, that is, they are more likely to assume that the speakers are telling the truth than lying (Castillo, Tyson, & Mallard, 2014; Elliott & Leach, 2016; Evans & Michael, 2014; De Silva & Leach, 2013; Leach & De Silva, 2013). In contrast, participants never have a truth bias when evaluating nonnative speakers, and they sometimes have a lie bias, that is, they are more likely to assume that the speakers are lying rather than telling the truth (Castillo et al., 2014; De Silva et al., 2013; Evans & Michael, 2014). These latter studies, however, were not concerned with processing fluency, and their results might be driven by prejudice. In contrast, a couple of studies failed to find an effect of foreign accent on truth judgment (Souza & Markman, 2013; Stocker, 2017; Wetzel et al., 2021). While the conflicting results might be due to methodological differences (e.g., use of a highly familiar foreign accent without testing whether it is harder to process, use of a single speaker per condition), the inconsistent results of the follow up studies suggest it is important to try to replicate the results of Lev-Ari and Keysar (2010).

Furthermore, while past research suggests that people believe information less when it is delivered in a foreign accent, there is no direct evidence to indicate that this difference is due to lower processing fluency. Confirming the underlying mechanism is important for both theoretical and applied reasons. First, it will help us understand how the experience of language processing can influence decision making. Understanding the source of bias can also enable us to understand where else it might surface and how to counteract it. The main goal of this study is therefore to test the underlying mechanism that leads listeners to believe information delivered in a foreign accent less. The manner by which the study tests it, using an intervention paradigm, also proposes one approach that could be taken to reduce the bias against nonnative speakers.

One way to further test whether the tendency to believe information less when it is delivered in foreign-accented speech is at least partly due to processing fluency is to test whether facilitating the processing of foreign accent can reduce the bias. It is well established that the difficulty of processing foreign-accented speech is reduced with more exposure to the accent (Clarke & Garrett, 2004). Furthermore, the benefit of exposure can generalize to new speakers, especially if individuals are exposed to multiple speakers (Bradlow & Bent, 2008). The benefit is due to the fact that individuals speak with a foreign accent because the sounds of their native language differ from those of their second language. For example, the “th” sound in “think” does not exist in German. Therefore, many German speakers produce “s” instead (e.g., Hanulíková & Weber, 2012). Exposure to several German speakers who substitute “s” for “th” allows the listener to adapt to the accent and correctly interpret words with this nonnormative pronunciation of “th” also when encountering unfamiliar German speakers. Furthermore, German is not the only language to not have the sound “th.” Therefore, exposure to foreign-accented speakers who produce “th” in a nonnormative way could also help the listener to later understand speakers of other languages, such as Hebrew, who also does not have this sound and produce it in a similar nonnormative way. Indeed, it has been found that exposure to several foreign accents can alleviate processing of other foreign accents (Baese-Berk, Bradlow, & Wright, 2013). If at least one of the reasons that listeners distrust information delivered in a foreign accent is because it is harder to process it, and if exposure to foreign accent facilitates its processing, it might be possible to reduce individuals’ bias against foreign-accented speech by exposing them to foreign accent. This is the goal of the current study.

It should be noted that one prior study attempted to examine a highly related question, whether familiarity can moderate the effect of foreignness on credibility by facilitating processing (Wetzel et al., 2021). It did so by comparing familiar native accent, unfamiliar native accent, familiar foreign accent, and unfamiliar foreign accent. That study did not find an effect of either foreignness or familiarity. Unfortunately, because that study only utilized one speaker for each accent condition, it is hard to draw conclusions from it. There are many idiosyncratic features that influence how credible one sounds (e.g., pitch, speech rate). These need to be controlled, either by using the same speaker for all conditions (difficult when testing accents), or by using several speakers per accent to reduce the effect of idiosyncratic differences. As the study used one speaker per accent, it is impossible to know whether the speakers of the different accent did not differ on other features that cue trustworthiness. Second, group effects are not present in a comparison of every two individuals from the two groups. For example, men are taller than women, but when selecting at random only one man and one woman, the height difference might not be present or reversed. In other words, even if speakers of an unfamiliar foreign accent are harder on average to understand than speakers of a familiar foreign accent, not all speakers of an unfamiliar foreign accent would be harder to understand than all speakers of a familiar f.oreign accent, and similarly, not all of them would be rated as less truthful. Furthermore, as the authors themselves acknowledged, it was not even clear whether the unfamiliar foreign accent in the study (Finnish accent) was any harder to process than the familiar foreign accent (German accent), as the two accents are quite similar. Therefore, we do not know at the moment whether the tendency to believe foreign-accented speech less can be attenuated by exposure