​In the paper, we propose a relatively broad characterization of linguistic discrimination as a range of practices, actions and experiences, which share a common core of an unfair (i.e. inequitable) treatment of a person on the basis of their language, language use or linguistic identity. One prominent example is that of intentionally denying a linguistic minority the right to use and transmit their language. What we have in mind here are forced linguistic assimilation policies resulting in deliberate linguistic deprivation and erosion, through coercive state policies in critical domains, such as education, healthcare and law. Another set of cases concerns situations in which certain linguistic properties (e.g. relating to a person’s accent, or whether their language is spoken or signed) function merely as a trigger or a proxy for other, prima facie language-unrelated negative attitudes toward members of different groups, based e.g. on race, ethnicity, class, gender and ability. So understood, linguistic discrimination can take various forms. The crucial element that these various forms share (including the two cases above) is a common experience of discrimination in communication, and in virtue of being a user of language. Moreover, experiencing such discrimination may affect how a person experiences and engages in communication with others, as well as their sense of linguistic identity.

​In the paper we highlight the challenge of understanding the complexities of linguistic discrimination, which arise from the complex and “messy” nature of both the phenomenon itself and the various concepts, definitions and theories used for studying it within and across academic disciplines (e.g. philosophy, psychology, linguistics, political science, anthropology). Thus, one of our key aims in this paper is to explain certain variability in how linguistic discrimination is defined and investigated across disciplines.

In our description of implicit attitudes that may lead to linguistic discrimination, we take implicit attitudes to be those that are typically outside one’s conscious awareness (and control) (see e.g. Kelly 2013; Brownstein & Saul, 2016). In cases that we are interested in, such attitudes are negative attitudes towards certain languages, linguistic properties (e.g., accent) or forms of communication that underlie an unfair treatment of interlocutors (i.e. those who participate in communication on a given occasion). As an example, consider cases where members of one community, A, often the more powerful one, routinely discriminate against members of a less-powerful community, B, when members of B do not perform according to the implicit linguistic expectations of community A. This may happen, for example, when members of B use A’s language with a marked B-accent, or when using only some competencies in A’s language but not others (e.g. reading and writing in it but not speaking or listening). Members of A may discriminate without being consciously aware of their implicit attitudes toward the particular linguistic properties used by B, and/or any other negative attitudes members of A may have toward them. They will also typically not have direct control over how their negative attitudes impact on their behaviour towards members of B.
​
This above is, of course, but one way to understand what implicit attitudes are, which can be applied to cases that we are interested in. There is a substantive debate concerning the nature of implicit attitudes (e.g., are they unconscious, dissonant or unendorsed cognitions, or something else? See e.g. Holroyd et al., 2017), as well as ongoing empirical research concerning implicit biases.  Moreover, we do not want to preclude the possibility that more than one kind of implicit mental state or process may underlie various specific forms of linguistic discrimination. 

There are at least two things we can point to. First, we note in the paper that some strands of research concerning bias and language in social psychology and sociolinguistics face some problems related to their reproducibility (i.e. whether and to what degree a study provides the same findings when independently conducted by a different researcher). In a nutshell, more work may be needed to experimentally investigate the role of implicit bias in communication. For example, Roessel and colleagues (2020) suggest that implicit measures may be needed to uncover the nature and scope of some forms of foreign accent bias, given that participants in studies investigating prejudice against speakers with foreign accents may be able to suppress their prejudice when explicitly asked for their opinion about a speaker (e.g., Wang et al., 2013). We are hopeful that experimental research on bias, including language-related implicit bias, will not only provide robust results, but also give us some knowledge regarding possible strategies for countering linguistic discrimination (see e.g. Hansen et al. 2014).
​
Second, we are excited about the prospects of expanding our toolkit for investigating various forms of linguistic discrimination, particularly those resulting from implicit attitudes. One way to do so, we suggest, would be to expand the evidential pools and look for direct/ personal/ first-person evidence from individuals with an experience of linguistic discrimination. While this may not seem to be a controversial proposal as such, the proof is ultimately in the pudding: we need a considered theoretical grounding for why we want to appeal to such evidence and corresponding methodological approach(es) that can be used to gather and process this type of evidence.

The view that testimonies from harmed individuals constitute an acceptable, valid and legitimate form of evidence is more common in cases of explicit negative attitudes than it is in their implicit counterparts. To our mind, however, this very type of evidence offers a valuable source of knowledge on the nature, dynamics and mechanisms of linguistic discrimination resulting from implicit attitudes, which may not be so easily detected in current research designs. We want to emphasise that our aim is not to rank different evidentiary sources, that is, to pronounce some forms of evidence as intrinsically preferable to others. Rather, we make a case for a greater inclusion of this particular type of evidence, and of existing (albeit still not too common) pertinent research that incorporates first-person-perspective, into current studies on linguistic discrimination in its various forms.
​
While we hope that such a proposal would not appear too controversial, we openly acknowledge the fact that its systematic theoretical justification, and the methodological challenges that it raises, are nonetheless far from a settled matter. We hope that the emerging scholarship that directly focuses on key theoretical and methodological aspects of including first-person perspectives in psychological research could be of some help here, given that it tries to systematise that endeavour by means of identifying different approaches (e.g. systematic introspection, autoethnography), and their respective strengths and weaknesses (e.g. Lumma and Weger 2023). Consider autoethnography as one example. Autoethnography can be defined as a qualitative research method and a “genre of academic writing that draws on and analyzes or interprets the lived experience of the author and connects researcher insights to self-identity, cultural rules and resources, communication practices, traditions (...)” and more (Poulos 2021: 4). Such accounts of individual personal histories can provide a valuable insight into their linguistic experiences, including experiences of more or less subtle forms of linguistic discrimination that may have otherwise gone unnoted (e.g., Snoddon 2022; Dovchin et al. 2024). Arguably, such accounts can aid investigating the experiential basis of linguistic discrimination in its various forms, including those resulting from implicit attitudes. 

Authors' work:

• Anna Drożdżowicz and Yael Peled. 2024. The complexities of linguistic discrimination. Philosophical Psychology, 37, 1459–1482, https://doi.org/10.1080/09515089.2024.2307993

Other cited work:

• Brownstein, M., and J. Saul (eds.). 2016. Implicit bias and philosophy, volume 1: Metaphysics and epistemology. Oxford University Press.
• Dovchin, S., Gong, Q., Dobinson, T., and M. McAlinden. 2023. Linguistic Diversity and Discrimination: Autoethnographies from Women in Academia. Routledge.
• Hansen, K., Rakić, T., and M.C. Steffens. 2014. When actions speak louder than words: Preventing discrimination of nonstandard speakers. Journal of Language and Social Psychology, 33, 68–77.  https://doi.org/10.1177/0261927X13499761
• Holroyd, J., Scaife, R., and T. Stafford. 2017. What is implicit bias? Philosophy Compass, 12, e12437. https://doi.org/10.1111/phc3.12437
• Kelly, D. 2013. Implicit bias and social cognition. In B. Kaldis (ed.), The encyclopedia of philosophy and social science (Vol. 9) (pp. 460–462). SAGE Publications.
• Lumma, A. L., and U. Weger. 2023. Looking from within: Comparing first-person approaches to studying experience. Current Psychology, 42, 10437–10453. https://doi.org/10.1007/s12144-021-02277-3
• Poulos, C. N. 2021. Essentials of autoethnography. American Psychological Association.
• Roessel, J., Schoel, C., and D Stahlberg. 2020. Modern notions of accent-ism: Findings, conceptualizations, and implications for interventions and research on nonnative accents. Journal of Language and Social Psychology, 39, 87–111. https://doi.org/10.1177/0261927X19884619
• Snoddon, K. 2022. Writing as being: On the existential primacy of writing for a deaf scholar. Qualitative Inquiry, 28, 722–731. https://doi.org/10.1177/10778004211073071
• ​Wang, Z., Arndt, A. D., Singh, S. N., Biernat, M., and F. Liu. 2013. “You lost me at hello”: How and when accent-based biases are expressed and suppressed. International Journal of Research in Marketing, 30, 185–196. https://doi.org/10.1016/j.ijresmar.2012.09.004

Perhaps the best way to get a grip on metalinguistic negotiation, and how it stands out relative to our ordinary thinking about disagreement, is by comparing two types of conversation. Consider first the following exchange:

It’s pretty clear what this disagreement is about. It concerns Bongbong Marcos’ numerical position in the list of Filipino Presidents. But while it’s important to note what the speakers are disagreeing about, it is just as important to note what they agree on. A and B agree on all of the relevant concepts at play within conversation. They both mean the same thing by the term ‘Philippines,’ they do not have obscure or strange ideas about what constitutes a ‘President,’ and they accept that 16 comes before 17. We might call this a first-order disagreement. It is ‘first-order’ to the extent that the disagreement strictly concerns facts about the world and the disputing parties agree about which concepts to use in representing these facts.
 
You might wonder: how else could a disagreement take shape? Consider another exchange: 

Again, it’s obvious what this disagreement is about: C and D disagree as to whether kissing someone outside of their relationship counts as ‘cheating.’ But there’s something distinctive about this exchange compared to first-order disagreements. Where A and B agree on the relevant concepts but disagree about the facts, C and D agree on the facts but disagree about the relevant concepts. To put it differently, C and D have different concepts of cheating, and they are arguing over which is the best concept to apply to the act of kissing outside of their relationship.
 
This latter kind of disagreement is known as metalinguistic negotiation. It contrasts with our usual understanding of disagreement insofar as it involves different attitudes at the level of concepts, rather than first-order facts. Despite its departure from ordinary thinking, metalinguistic negotiation is a part of our everyday lives. ​

​Think about a formal debate. Two sides speak in turn, for or against a contention. The context is structured such that each side gets a certain amount of time to put forward their position, and where interjections are prohibited or mostly unallowed. There aren’t really limitations on what speakers can say, perhaps apart from the use of curse words, and all contributions are given equal consideration.
 
These are extremely ideal conditions for disagreement. The rules allow speakers to make uninterrupted arguments for their point of view, which is heard and taken seriously by opponents (and judges). For the most part, I imagine, formal debates don’t tend to be about which concepts speakers ought to use. Nevertheless, metalinguistic negotiation can be ‘ideal’ in a similar way. I have argued that in ideal metalinguistic negotiations, speakers are conversational peers insofar as they can contribute equally to the processes involved in determining which concept will be used in a particular context. Take the disagreement between C and D above. For this disagreement to be ideal, both C and D must be able to defend, with equal consideration, their preferred concept of cheating as being apt or appropriate for the situation – they must be able to make equal contributions to settling the question of whether kissing outside of their relationship counts as cheating. That is, both must be able to make a similar impact on conversation by taking time to listen to one another and weighing up the reasons that would likely lead to one concept being jointly accepted.
 
This is how I have interpreted much of the philosophical literature on metalinguistic negotiation. There seems to be a sense in which philosophers have projected idealised versions of themselves onto the public. The representation of metalinguistic negotiation in this literature appears to assume that speakers are able to contribute equally to conceptual disagreement, or that the relative contributions and impact of speakers are roughly the same.
 
Formal debates do a particularly good job, though imperfectly, of abstracting themselves from real-world power structures. As a matter of fact, however, disagreements in the wild tend not to take this shape. As typical socially situated subjects, we are not abstracted from power-relations. We exist within them. And they affect the dynamics of disagreement, even at the level of concepts. In other words, metalinguistic negotiation in the wild is rarely ideal. Most metalinguistic negotiations are non-ideal. Speakers are not usually able to contribute equally to conceptual disagreement, where one speaker tends to have more control over which concept is chosen as apt or appropriate for the situation.
 
When we attend to this fact, we begin to see just how worrying metalinguistic negotiation can be: it is often infected by the unjust power that is responsible for the oppression of particular social groups.

​​Many cases of metalinguistic negotiation are reasonably unproblematic. In their oft-cited example, David Plunkett and Tim Sundell (2013) discuss a situation in which speakers disagree over which concept should be expressed by the word ‘spicy’ in a particular context. Provided that the situation is innocuous, where speakers are simply exchanging their respective thoughts about the spiciness of a dish, then it doesn’t really matter whether one person has greater control to determine which concept of spicy should be used. In fact, sometimes unequal control is warranted. We might imagine that one of the speakers is a notable chef that has better insight into what counts as spicy in, say, the Australian context. It seems right that given the chef’s expertise, they ought to have more of a say as to which concept of spicy is appropriate – especially when deciding whether to add more spice.
 
But other non-ideal metalinguistic negotiations are seriously troubling. These are situations in which settling on a particular concept is practically important, but where the process of determining which concept to use is skewed in favour of someone with greater power owing to their dominant social identity. To see this, consider the following exchange where a woman makes claims of sexual harassment against her male boss:

​​​​What makes this case especially worrying is that our society is patriarchal, misogynistic, and sexist, with a history of systematically under-ascribing credibility to women, especially when it comes to sexual harassment allegations, on the grounds of being ‘hysterical’, ‘over-sensitive’, and ‘paranoid.’ Moreover, our society tends to over-ascribe credibility to men, with operative stereotypes of men as objective, rational, and level-headed.
 
In the conversation between the woman and her male boss, these existing norms and stereotypes become apparent in the context given the nature of the topic. This infects its dynamics with unjust power relations. Because men are considered objective and women are seen as hysterical, the boss acquires greater control over the processes that determine which concept of sexual harassment will be used. There is unequal power in metalinguistic negotiation that tracks social identity. In this case it’s gender.
 
The practical upshots of this are significant. If the man gets his way, and his preferred concept becomes operative in the context, then the woman’s claim will be dismissed. The act won’t be considered sexual harassment. This frustrates the woman’s ability to pick out a particular wrong-doing, and to have it recognised by her wrong-doer. And it prevents the woman from being able to seek redress, either personally or institutionally.

I have argued that problematic forms of non-ideal metalinguistic negotiation involve distinctive types of injustice. There are different ways of understanding what (in)justice is. What I am interested in is epistemic and linguistic injustice. Broadly, epistemic injustice occurs when one is wronged as an epistemic subject – as someone who knows, or believes, or understands, or inquires. And linguistic injustice, a less familiar form, occurs when one is wronged as a linguistic subject – as someone who communicates, or describes, or expresses. Given that both epistemic and linguistic capacities are distinctive human capacities, then when one is undermined as an epistemic or linguistic subject, they are undermined as human beings. And when one is not treated as fully human, then one is objectified.
 
Epistemic and linguistic injustice occurs within non-ideal metalinguistic negotiations. To see this, we can simply reflect on how the woman’s epistemic and linguistic capacities become seriously limited in the conversation with her boss. When the boss denies the allegation and puts forward an alternative concept of sexual harassment that would get him off the hook, the power dynamics that kick-in make the situation one in which the woman can no longer contribute equally to conceptual disagreement owing to being a woman. That is, in virtue of her gender identity, the woman loses important epistemic and linguistic capacities that constitute her humanness. She cannot make a significant contribution to epistemic conditions, such as settling on a concept that will be used to make sense of what happened to her; and she cannot make a significant contribution to linguistic conditions, such as determining what ‘sexual harassment’ will mean in the context. The woman is constrained in her ability to enact aspects of her humanity owing to prevailing prejudicial stereotypes relating to her identity. I have called this metalinguistic injustice. 

Other cited works:

• Plunkett, D. and Sundell, T. 2013. Disagreement and the Semantics of Normative and Evaluative
  Terms. Philosophers’ Imprint, 13, 1–37.