Refining "Microaggression": a linguistic perspective

For a few years I’ve been very interested in a phenomenon that is a topic of a lot of research in the social psychology literature, and which is increasingly discussed in social media and in the news media, but which has not yet gained much attention in linguistics. The phenomenon, microaggression, is somewhat controversial. I think this is, in part, because it is a social phenomenon that often has language and language use central to it, but the discussion around microaggression is often predicated on a number of (sometimes easily challenged) assumptions, and the definitions of microaggression that are common, while intuitive, are not terribly formally rigorous — so interpretation of what is and is not a microaggression is not satisfyingly ‘objective’. 

Here I’m going to discuss some of my ongoing research project, and specifically I want to propose a tighter definition of microaggression in a linguistic domain and then argue that the unsatisfying ‘subjectivity’ of microaggression is precisely why it’s an interesting topic of study and a valid — and necessary — concept. 


There are a number of varying, slightly different definitions, but microaggression is usually defined as “brief, everyday exchanges that send denigrating messages to certain individuals because of their group membership” (from this book). 

Earlier formulations were a bit more explicit, so people may be familiar with microaggression as the “…brief and commonplace daily verbal, behavioral, or environmental indignities, whether intentional or unintentional, that communicate hostile, derogatory, or negative racial slights and insults toward people of color.” Pretty quickly, people challenged this formulation, and I think that it's generally accepted that microaggression is about power structures and not inherently a bad thing white people do to people of color (as was more-or-less the original formulation by Sue et al). So now, people will accept that microaggression is something that say, white women can experience. I think most would accept that white men can (and do) experience microaggression in certain contexts (for instance, when I'm a non-native Chinese speaker in an exclusively Chinese space). The point is that whenever there is a power dynamic at play, microaggression is possible

There are a number of  different ways of constraining the idea, but as a linguist, I'm naturally more inclined to look at linguistic behavior in conversation, in part because that's what I feel most capable of analyzing. So here, I won't be discussing facial expressions or institutional decisions like naming a dormitory for a slave owner and insisting black students call their residence advisors "master."  Rather, I will discuss verbal behavior only.

So that's a lot of words and still it's not that clear what microaggression actually is. Most discussion of it fall back on (1) giving a bunch of examples and (2) interpreting those examples. The problem is, most take the interpretation to be obvious, uncontroversial, and the only way to interpret what was said. This is a mistake, not just because it opens up the theory to criticism like Racial Microaggression? How Do You Know?but because it misses what makes microaggression a valuable concept in the first place.

But first, some classic examples. One of the most common examples of microaggression is when an Asian American is asked "where are you from," and then pressed "where are you really from." The traditional explanation is that it's clear, especially from the second sentence, that the meaning is you can't be from here, and more broadly you're not American and even more broadly you don't belong. I'll take another example from my own experience: I went to a bookstore with a black friend, who was asked 3 times in 5 minutes to perform employee tasks at the store ("can you change the music please?" "where do I find...", "Can you just reshelf this for me, I'm not going to buy it."). The staff at that bookstore wear uniforms, and he was dressed nothing like their uniform. The traditional microaggression interpretation would be that these people mistook him for an employee despite him clearly shopping in street clothes because he is black, and they acted out of a number of assumptions based in stereotypes about black people (e.g., that they don't read, and possibly even that they shirk responsibility -- so the fact that he was obviously not working didn't necessarily suggest he wasn't an employee). Derald Sue and his colleagues have a typology of microaggressions and their interpretation in the article I linked above.


Many people, especially white people and especially men, react to these kinds of stories with what I (as a white man...) think is a completely reasonable question: but couldn't it have just been a misunderstanding? This is often brushed aside or ridiculed, which I think is a mistake. 

The key strength of microaggresion, and what gives the concept its 'bite' is that yes, it always could have just been a misunderstanding.

That is, my argument is that microaggression is a valid and useful tool for discussion precisely because it is not cut-and-dry. Otherwise, it's just overt aggression. Nothing 'micro' about it. If I say "your outfit looks dumb" that's just an insult. If I say "oh, your outfit is so...ethnic," that's much trickier to categorize. Is "ethnic" a codeword? Am I insinuating it's inappropriate? That I don't like it? That I think it's unprofessional or looks bad? That I think it doesn't belong? Well...maybe.

So the first part of my argument is:

Microaggression deals with a class of utterances that, given the context of their production, are ambiguous: they are potentially insulting or invalidating, but the insult is plausibly deniable.

What this means is that microaggression is strategic

More importantly, it also means that listeners, faced with potential verbal microaggression are faced with a classification problem: is this an insult, or a mistake?

Elsewhere I've written about the concept in the field of Pragmatics of Gricean implicature, and specifically compared it to the Semantic notions of presupposition and entailment. Implicature is when you say something that -- when listeners assume you're trying to maximize quality, relevance, etc. -- insinuates something beyond the literal content of the utterance. Presupposition and entailment, however, things you can read directly off the utterance.

For example, take the sentence:

  • The first democratically elected leader of the Congo was assassinated.


The presuppositions are things like "there is a democratically elected leader of the Congo," (namely, Patrice Lumumba). Presuppositions can cause problems, like when you say things like "the king of France is bald." It presupposes a king of France, which doesn't actually exist in the real world. The entailments are things like the first democratically elected leader of the Congo is dead and that they did not commit suicide.

Now think about:

  • Congrats, you've won five dollars!

The implicature is that you've won exactly five dollars, but this is cancellable. That means I can say: "Congrats, you've won five dollars! In fact, you've won ten!" (note that it's a "scalar" implicature, meaning I can cancel the implicature that you won exactly five dollars, but I can't say "in fact, you've won only four!" without canceling the entire utterance).

I review all this so I can now claim:

Microaggressions are unambiguously identifiable when the offensive material is encoded structurally via presupposition or entailment.

This has a more interesting inverse:

Microaggressions are NOT unambiguously identifiable when the offensive material is merely implicated.


I am a huge fan of using the tools of Game Theory to investigate (some) linguistic phenomena. Game Theory grew out of Decision Theory, and in a nutshell is a formal (mathematical) approach to figuring out how rational, thinking 'agents' should behave when the value of their decisions depends on both (1) their own choice, and more importantly (2) the choices of other agents

Applied to pragmatics, this means people -- if we were totally rational -- would think about both what we want to say and how it will be interpreted. I think we generally do do this, but tons of research backs up that we're only moderately good at thinking about what other people are thinking. That means conversation is strategy, but we kind of suck at strategizing. 

I've got a whole complicated flow chart for speaker decisions that models the process, but it basically boils down to: you have a choice of being nice or not being nice, and if you're not nice you have a choice of either saying it outright ("your outfit looks dumb"), encoding your meanness directly in the utterance through presupposition or entailment ("Only idiots wear herringbone"), or encoding it through cancellable implicature ("your outfit is so...interesting.")

The benefit of the latter is that it's deniable and you can save face if you're challenged: "what do you mean 'interesting'?" "I mean I love how intricate it is!" 

But here's a problem: as I mentioned above, we kind of suck at figuring out what other people are thinking. We're also kind of lazy. So it's entirely possible you choose "be nice" at the first branch of your decision-making process, and then basically misfire, and say "you look interesting." You meant something like "you look good," but it just kind of came out wrong because you're socially awkward. 

The other possibility is that you misfire, but it's because you have implicit bias that you're unaware of, but which affects your worldview and which other people can recover from your speech. That is, you choose to be nice, and you say something like:

  • You're a credit to your race.

Right there you've got presuppositions that could be challenged (e.g., "race is biologically real"), and entailments that are perhaps unsavory (e.g., "people of your race are usually not good in some way."). And here you thought you were being nice.

[EDIT: Stephan Hurtubise points out in the comments below that the latter is actually implicature! The entailment is that you are a credit, the implicature we take away from that is that your race needs 'a credit,' but this is, in fact, cancelable. Another post, for another day, is going to be the conventionalization of so-called 'dog whistle.' I would argue that using terms that are conventionally used to insult (e.g., "welfare queen") affects how we interpret the utterance -- even when, as was the case a few years back with a friend of mine, the speaker does not know the conventional meaning.] 

Unfortunately, here we're moving away from Game Theory, since GT is all about agents who are completely rational, so having unconscious bias doesn't fit well -- at best, you can model it as some sort of error (these are called "shaky hand" models, usually).


So let's say someone says something that doesn't stand out as negative or hostile due to any presuppositions or entailments you can point to that are inherent to the utterance, but you still feel like it's a negative remark because of some implicature. You are then faced with a classification problem: was this comment aggression toward me? 

This is where most previous analyses of microaggression really fall apart, and why people get into shouting matches when the subject of microaggression comes up. Most explanations of microaggression boil down to that was obviously an aggressive statement, are you blind or are you playing dumb?! The other side, of course, being that was obviously NOT an aggressive statement, why are you playing the victim?!

My argument is that classifying an ambiguous utterance as either "microaggression" or "not microaggression" is a Bayesian classification problem.

Without getting into Bayes' Rule in depth, the basic idea is that you classify the statement based on situating it in the context, and by evaluating prior beliefs about the speaker and the situation. These priors inform your final classification. 

As an example, there's a video going around this week, of a black mother eviscerating a cashier at Target for insulting her children's clothing. In describing the situation to my colleagues, one black female, one white male (note: this is a gross oversimplification of their complex, intersectional identities), they interpreted the situation differently, because they had different prior beliefs about the context

So here's what happened: one child in question was wearing traditional African clothing, and was evidently shopping without their mother present. The cashier asked the child:

  • "Are you trick or treating?" 

Some potentially relevant context: This took place in early September. Target has Halloween decorations for sale already. 

In order to determine how that statement was meant, without further context, we have to start evaluating a handful of possibilities: was she using the narrative present for the future (i.e., did she mean "are you going to go trick or treating?") Did she mean something like "are you trick or treating this year?" so she could helpfully point out the halloween section? Or was she implicating "your clothes look like a costume" and therefore "you look ridiculous" ?

From the utterance alone, it is impossible to tell ... but people have very strong intuitions about what was meant and what is plausible. These are formed, in part, by past experience, so my black female colleague immediately reacted with the interpretation that it was probably intended as an insult (this is my intuition, too), while my white male colleague thought it was an odd thing to say, but potentially wholly unrelated to the child's current outfit. The intuitions are also formed by context. I would expect that on October 31st the statement is minimally offensive, and on April 30th it's approaching maximally insulting. But September 12th in a store that has put up Halloween decorations absurdly early -- that's harder to argue objectively. I have my intuitions, but other people may disagree.

If it was an insult, what makes it an effective microaggression, though, is the fact that the cashier can say "I don't know what you're talking about -- I was just going to point out our halloween section!" Or even slyly double down on microaggression with "I just thought you were in the spirit!" If it wasn't intended as an insult, the cashier made an unfortunate mistake, and while completely well-meaning, accidentally said something that looks like an insult. And there's an old saying about ducks that basically sums up Bayes' Rule in this case (if it looks like a duck, and quacks like a duck...).


In summary:

  1. I propose we limit verbal 'microaggression' to things that cannot be clearly classified as overt aggression.
  2. I suggest that ambiguity is not a bug, but is in fact a key feature
  3. I argue that people think strategically, albeit imperfectly about how we communicate with others.
  4. I argue that when confronted with an ambiguous and potentially 'microaggressive' utterance, the listener is confronted with a classification problem.
  5. I argue we use Bayesian thinking -- where our prior beliefs about the speaker and the context inform our interpretation -- to figure out what was meant
  6. Ultimately, I argue the above is probabilistic and, infuriatingly for those who may have been insulted, can never be 100% proven when the microaggression is implicature and not a 'structural' microaggression (that is, one where presupposition or entailment can be clearly demonstrated). 

The key here is that it's important to think about context, and that whenever we get lazy it's entirely possible to assume someone will understand "how you meant it" and be wrong about that assumption. It's also important to note that context is constantly evolving, and that interactions add to people's prior beliefs for the next interaction. This is why when Donald Trump says something ambiguous people assume it probably is microaggression, but when anti-racist activist Tim Wise says something ambiguous, he may get a pass.

Moreover, I think it's hugely important that we don't throw the baby out with the bathwater, so to speak, when challenging less rigorous discussions of microaggression. People are experiencing something that is a real phenomenon. Coded racist speech exists. Dogwhistle exists. But its strength is precisely that it is so maddeningly hard to prove.

In general, my advice would be to listen when people claim they've experience microaggression, and to think about what informs their interpretation. And if you want to avoid accusations of microaggression, that means you're going to have to think about context, history, and the other person's perspective (yes, this is difficult. Sorry.), every time you go to speak.



©Taylor Jones 2016

Have a question or comment? Share your thoughts below!