Overview
Learning a second language is tough enough. It’s even harder when that language is the one used in your workplace. The stakes are high. Your colleagues’ perceptions of you can make or break your chances of getting a raise, bonus, or promotion. Could a handful of grammar mistakes affect those important perceptions?
Across multiple research studies we’ve found that grammar mistakes negatively affect your image. But it’s an open question how that applies to non-native English speakers. Might you get a pass from your colleagues if English is not your native language? Or are you judged more harshly?
Research on stereotyping shows that judgments about others tend to occur along two dimensions, namely warmth and competence (i.e., how nice and smart a person seems). Thus, we wondered how grammatical errors differentially affect the perceived warmth and competence of native vs. non-native English speakers.
The Experiment
To test this question, we designed an email experiment with 1,200 people from the research platform Prolific. Participants viewed a hypothetical email from a coworker who was either a native or non-native English speaker (randomly assigned); unbeknownst to participants, we also randomly assigned whether the email used correct grammar or contained three obvious grammar mistakes that a non-native English speaker might make.
At the beginning of the study, participants were instructed to read the email and answer the questions that followed. Participants were then told, “The following email is from a new co-worker at your job. English [is / is not] his native language.” We randomized whether each participant saw “is” or “is not.” Each participant then viewed one of the two emails below, randomly assigned.
From: [email protected]
Sent: Tuesday, Sep 19, 2023, 1:38 PM
To: [email protected]
Subject: Thanks
Hello,
Thank you for the introduction to John. It was a very enjoyable and interesting conversation.
I'm looking forward to working with you both.
Best,
Carlos
Sent: Tuesday, Sep 19, 2023, 1:38 PM
To: [email protected]
Subject: Thanks
Hello,
Thank you for the introduction to John. It was a very enjoyable and interesting conversation.
I'm looking forward to working with you both.
Best,
Carlos
From: [email protected]
Sent: Tuesday, Sep 19, 2023, 1:38 PM
To: [email protected]
Subject: Thanks
Hello,
Thank you for introduction to John. It was much enjoyable and interesting conversation.
I look forward to work with you both.
Best,
Carlos
Sent: Tuesday, Sep 19, 2023, 1:38 PM
To: [email protected]
Subject: Thanks
Hello,
Thank you for introduction to John. It was much enjoyable and interesting conversation.
I look forward to work with you both.
Best,
Carlos
Finally, after reading the email, participants answered a series of survey questions to measure how warm and competent (i.e., how nice and smart) the email sender seemed. To measure both outcomes we used the psychology scales and survey questions of Abele and colleagues (2016).
Participants were asked, “To what extent do you think each of the following words describe Carlos? (1 = Not at all, 7 = Very much)” with answer options on a 1-7 survey scale. The words for competence included “Competent,” “Capable,” “Clever,” “Intelligent,” and “Efficient.” The words for warmth included “Caring,” “Warm,” “Empathetic,” “Affectionate,” and “Friendly.”
Participants were asked, “To what extent do you think each of the following words describe Carlos? (1 = Not at all, 7 = Very much)” with answer options on a 1-7 survey scale. The words for competence included “Competent,” “Capable,” “Clever,” “Intelligent,” and “Efficient.” The words for warmth included “Caring,” “Warm,” “Empathetic,” “Affectionate,” and “Friendly.”
Results
The data revealed several interesting findings. First, as anticipated, we found a significant overall difference in perceived competence between the correct and incorrect grammar conditions (p < 0.0001). Whereas correct grammar netted our email sender an average rating of 5.06 (1-7 scale), incorrect grammar reduced this rating to 4.25, a 16.0% drop. In statistical terms, this equates to 0.62 standard deviations, a medium-to-large effect.
A much smaller effect emerged for warmth. Whereas our email sender who used correct grammar was rated 5.08 on our perceived warmth scale, incorrect grammar only dropped this rating to 4.91 (p = 0.012), a 3.4% drop.
There were also overall positive effects for just being a non-native English speaker, likely from being able to speak at least two languages. Our non-native speaker was rated 0.70 points more competent (16.3.%) and 0.37 points warmer (7.6%) than native speakers, who were rated 4.31 on competence and 4.82 on warmth (p’s < 0.0001). These effects held even when considering only the good grammar conditions; non-native speakers were rated 0.32 points (6.6%) more competent and 0.30 (6.0%) more warm than non-native speakers (p = 0.001; p = 0.002).
There were also overall positive effects for just being a non-native English speaker, likely from being able to speak at least two languages. Our non-native speaker was rated 0.70 points more competent (16.3.%) and 0.37 points warmer (7.6%) than native speakers, who were rated 4.31 on competence and 4.82 on warmth (p’s < 0.0001). These effects held even when considering only the good grammar conditions; non-native speakers were rated 0.32 points (6.6%) more competent and 0.30 (6.0%) more warm than non-native speakers (p = 0.001; p = 0.002).
Even more interesting, there was a significant interaction between native vs. non-native English speaking and good vs. bad grammar. Although non-native English speakers do incur a penalty for using incorrect grammar, this penalty was significantly less than the penalty imposed on native speakers. As the graph below illustrates, whereas native speakers were perceived to be 1.11 points (22.7%) less competent when using bad grammar, non-native English speakers were only perceived to be 0.42 points (8.1%) less competent when using bad grammar (p < 0.0001). This 0.69-point difference equates to a 62% discount benefiting non-native English speakers. There was no such interaction effect for perceived warmth.
Finally, perhaps most interesting of all, we found that this “beneficial discount” differs significantly with gender, specifically the gender of the email reader (i.e., the colleague judging you). In summary, the vast majority of the beneficial discount conferred to non-native speakers is from female colleagues. Whereas this discount provided by women is 79.9% (1.06 points), the discount provided by men is only 28.8% (0.25 points) from men. In addition, women seem to penalize native English speakers a bit more for grammar mistakes, an effect we've found in other studies.
Conclusion
Contrary to what you might think, non-native English speakers do not seem to be judged more harshly for using incorrect grammar, at least relative to native speakers. In fact, while this experiment’s native speaker was perceived to be significantly less competent when using incorrect grammar (22.7% less), the non-native speaker was penalized less than half as much (8.1% less).
This effect seems to be driven by women. Whereas men viewed our non-native English speaker to be less competent by about the same amount as our native speaker when using incorrect grammar, women judged our native speaker much more harshly (25-30% less competent) comparted to our non-native speaker (5% less competent).
So, if English is not your primary language, take heart in knowing that you may not be held to the same standards as native speakers, especially around your female colleagues. Although there is still a slight image penalty for grammar mistakes, there also seems to be acknowledgement of the laudable feat that is learning a second language.
References
Abele, A. E., Hauke, N., Peters, K., Louvet, E., Szymkow, A., & Duan, Y. (2016). Facets of the fundamental content dimensions: Agency with competence and assertiveness—Communion with warmth and morality. Frontiers in psychology, 7, 1810.
Methods Note
We used Ordinary Least Squares (OLS) regression analyses to test for significant differences in competence between our experimental conditions. OLS regression analyses with interaction terms were used to test for differences across the main results. Our statistical significance threshold was a p-value below 0.05. The data and survey materials used for this study are available upon request.