Key Points
- Linguistic bias occurs when perceptions about a person's identity, informed by the language they use or how they speak, influence how others judge their skills and qualifications.
- Language may be a tool for communication, but communication is a human experience, and, as such, it is subjected to both conscious (explicit) and unconscious (implicit) biases.
- Scientists can proactively challenge linguistic assumptions to create spaces for multilingual opportunity and global engagement.
English has become the global language of science, but this dominance has a cost. Researchers who are not native English speakers face invisible hurdles: efforts to learn and use a second language, obstacles to research dissemination and diminished professional visibility. These barriers do more than prevent access to opportunities. They also cement unfair assumptions about scientific quality and preferentially amplify native voices, shaping scientific discourse in narrow and exclusive ways. Yet, simple solutions can address the issue and maximize opportunities for scientific exchange.
Language is more than a tool for communication. It influences how people are perceived and judged by others. Accent, dialect and word choice are some linguistic features that can trigger subjective perceptions of a person's identity and cloud how people judge the skills and qualifications of one another. This phenomenon, known as linguistic bias, acts as an invisible barrier to scientific progress. It quietly upholds old hierarchies and stereotypes about scientists and effectively sidelines multilingual talent. At its root, linguistic bias is less about language itself and more about perceptions of proficiency and compliance with subjective standards. This bias finds a stronghold in academic publishing. By critically reflecting on the sources and impacts of this linguistic gatekeeping, we can begin to break down barriers preventing the full representation of talent that would otherwise be lost in translation.
Historical Path Toward Linguistic Homogenization
Languages use structured formats of symbols (e.g., oral, written and gestured) for information exchange. The estimated 7,000 distinct languages spoken today are survivors of a linguistic history that once included more than 31,000 systems. But most of the planet's population speaks only 10 of these languages, English being the most spoken of all. Such linguistic homogenization is rooted historically in centuries of British colonialism and the more recent growth of the United States as an economic power. In a modern world of fast connectivity, English users can readily access and exchange information and make global connections, influencing politics, business, pop culture, science and technology. English speakers also benefit from better-paid jobs, greater professional visibility and more expansive networks of reach and opportunity. Such incentives help explain why most English users are not native speakers. 
Science has also experienced linguistic homogenization. As the primary language for the dissemination of scientific knowledge, there is pressure on non-native scholars to communicate their scientific work in English. This immediately imposes the burden on multilingual scientists to present their work in the non-dominant language and to conform to perceived linguistic standards of scientific discourse. As a result, a hierarchy of privilege based on English proficiency, or the perception of such, also exists that disadvantages multilingual scientists and hinders scientific progress by amplifying the most fluent voices.
Linguistic Bias in Academic Publishing
Language may be a tool for communication, but communication is a human experience and, as such, it is subjected to both conscious (explicit) and unconscious (implicit) biases.
Many aspects of linguistic bias are innate. Humans are conditioned to the unique speech patterns and phonetic tonality of their native language while in the womb and grow up as infants showing a preference for speakers of their native language. Linguistic bias is, therefore, deeply ingrained in the human brain and a strong conditioner of social perceptions and interactions. Language-based assumptions are also complex and plural-they may manifest alone, or in combination with non-linguistic biases, such as those based on economic and geographical stereotypes.
Linguistic bias is well documented in science. Studies suggest that authors perceived to be native or to have native-like proficiency can receive more favorable peer ratings and editorial decisions in scientific journals, but perceptions of the economic standing of the author's country of origin may also confound peer ratings. However, double-blinding the peer review process to hide the authors' identities equalized these ratings. In other studies, even subtle shifts in English language patterns led reviewers to perceive the authors as non-native speakers, resulting in lower evaluations of their scientific contributions.
Linguistic bias may manifest in academic publishing in other ways too, such as positive associations between reviewers and authors who share the same country of affiliation or those sharing the same primary language. Multitasking and deadlines, both common in academic publishing, exacerbate unconscious bias effects. This is because work stressors favor information processing through our unconscious system (a process described as the brain's cognitive economy). Reviewer fatigue is real, making peer reviews more prone to unconscious bias under the weight of heightened workload. This makes multilingual authors more vulnerable to language-based subjectivity during the peer review process.
Academic Publishing Beyond Words
Simple solutions, such as anonymizing author identities, can help address, if only partially, the pervasive influence of linguistic bias. Open access tools assisted by generative artificial intelligence (AI), including chatbots like ChatGPT, can help multilingual authors overcome some of the language barriers they face when disseminating their work. However, the use of AI tools has limitations. Users must know and abide by journal's legal and ethical policies (see ASM's Generative AI policy for guidelines on how to maximize the benefits of AI usage without compromising the integrity of the scientific record). GPT detectors may also unfairly flag multilingual authors due to the statistical measures of linguistic diversity and word choice they use to identify AI-generated content. Thus, more is needed to achieve linguistic equity in science publishing.
Self-awareness of linguistic bias, like other forms of unconscious bias, remains the most effective tool to mitigate its adverse effects. It starts with recognizing its pervasive influence (one cannot know what one doesn't know). It then invites self-reflection to identify and challenge potentially subjective responses to linguistic aspects of the work, such as grammar, style and word choice. Applying pre-established criteria and rubrics during evaluation and not changing them mid-course remains a staple of best practice toward peer review objectivity. For peer reviewers and editors, this means careful examination of the work's technical merit and evidence-based assessment of scientific rigor and novelty. Editors also benefit from in-group discussions, interactions with peer reviewers and coaching to collectively grow as a professional community focused on merit-based science and tune out the voices behind the work.
Multilingual authors are familiar with the various ways in which peers communicate perceived linguistic deficiencies in their work. Below is one such example from ASM's archived reviews in Applied and Environmental Microbiology®.
- "The authors should recruit the help of a native speaker to revise the English."
- A more appropriate expression of concern could have been phrased: The authors may consider revising the manuscript for clarity and to better highlight the quality and significance of the research."
Attempts to set and regulate the standards of academic English are all too common in social groups (a behavior termed "verbal hygiene"). Scientists, like any other social group, also tend to adopt a common linguistic style (what they perceive to be native or native-like) and regulate its use. It is in this context of perceived language standards and self-regulatory roles where linguistic bias manifests.
What can scientists do to ensure impartiality under the often-opaque blanket of linguistic perceptions? Editor and reviewer training is essential to manage perceived linguistic barriers fairly, as is disclosure of whether such barriers could bias the technical evaluation of the scientific contribution. Reviewers should ask themselves, "Do language deficiencies prevent objective evaluation of the science?" and, if yes, recuse themselves from the review process. Reviewers may also be guided to focus their evaluation on parts of the manuscript that they can comfortably understand. Communication with the editor is essential to understand peer concerns and to maintain the integrity of the peer review process.
A Path Toward Linguistic Equity
Impartiality, fairness and respect are foundational values of a strong peer review and editorial process. Though intrinsic to human identity and shaped through lived experiences, scientists can proactively challenge linguistic assumptions to create spaces for multilingual opportunity and global engagement. Journals play a key role in supporting the voices of science. Double-blinding the peer review process is a necessary first step, as it is facilitating training programs to increase awareness of linguistic and associated biases. Advancing these initiatives empowers the academic publishing community to strengthen its commitment to integrity and rigor while unlocking access to a world of talent that speaks the common language of science.
