What follows is a short essay I wrote in 2020 about Neil Gorsuch’s reasoning on behalf of the majority in the Supreme Court case Bostock v. Clayton County. The case addressed the fraught and complex issue of what exactly constitutes “discrimination,” and how we can legally test for discrimination, particularly when it involves same-sex attracted or transgender individuals. With my background in philosophy of language, certain issues jumped out at me about this case that I thought were lost both in popular commentary on the case, and in Gorsuch’s reasoning itself.
The case was in my view wrongly decided, and the controversy now long past. However, I think the essay nevertheless remains instructive as a case study in how subtle issues of philosophy relating to language, causation, counterfactuals, and other issues are often implicated in law, and a better understanding of such issues would make for better legal conversations, and better law.
The Fallacy in Bostock
In 2019, the Supreme Court decided in Bostock v. Clayton County that Title VII of the 1964 Civil Rights Act protects LGBTQ individuals from discrimination. Justice Gorsuch, writing for the majority, argued that Title VII’s ban on discrimination ‘on the basis of sex’ must, as a matter of textual logic, also apply to trans and same-sex-attracted individuals. Discrimination on the basis of sexual orientation or gender identity (hereafter ‘SOGI’), Gorsuch contends, necessarily implicates discrimination on the basis of sex, for the following reason: When an employer discriminates on the basis of SOGI, they necessarily discriminate on the basis of a property they would tolerate in someone of the opposite sex.
According to Gorsuch’s reasoning, if an employer fires a male for being attracted to males, for example, they fire that employee for a property, ‘being attracted to males’, that they would tolerate in a female employee. Similarly, if an employer fires a biological male for presenting as female, they necessarily discriminate on the basis of a property, ‘presenting as female’, they would tolerate in someone of the opposite sex. In this sense, according to the majority, sex is a ‘but-for’ cause in cases of SOGI discrimination: But for the fact that the individual was a biological male or female, they would not have been discriminated against due to either presenting as, or being attracted to, the opposite sex.
Justice Alito and Justice Kavanaugh, dissenting along with Justice Thomas, however, disagree. They contend that Gorsuch is being too literalistic and austere in his understanding of what discrimination ‘on the basis of sex’ means. That phrase, the dissenters argue, clearly does not, in its ordinary commonsense usage, mean or imply discrimination ‘on the basis of gender identity or same-sex attraction.’ As they point out, no one who drafted Title VII, nor the public at large, would have understood its provision in that way. As a matter of public, original understanding, then, ‘on the basis of sex’ simply cannot be made to mean or imply what the majority insists it does.
In my view, the dissent is correct in their assessment that Title VII neither meant, nor implies as a matter of textual meaning, that sex discrimination is implicated in SOGI discrimination. However, the root of the disagreement between Gorsuch and the dissenters, I believe, has remained obscure.
The crux of the disagreement is over what discrimination ‘on the basis of sex’ does or does not imply as a matter of meaning. The dissent rightly notes that discrimination on the basis of sex would not have been understood to include SOGI discrimination at the time of its adoption. However, Gorsuch rightly notes that not everything a phrase implies as a matter of its meaning, even as a matter of its public and original meaning, is fully appreciated by those who grasp its meaning. So, for example, Gorsuch correctly observes that sexual harassment of males by other males would not have been appreciated as entailed by the concept of discrimination on the basis of sex by many at the time of Title VII’s adoption; nor would many have appreciated that employment decisions made on the basis of sexual stereotyping also are implied by the concept of discrimination on the basis of sex. Yet, upon deeper reflection on the full implications of the statue’s meaning, we have come to see that these forms of discrimination are indeed forms of discrimination on the basis of sex.
This debate, then, cannot be adjudicated simply by appeal to disagreements over whether a literalistic textualism, or a public understanding originalism, is the correct interpretive method. Instead, it can only be adjudicated by considering more carefully just what the ordinary meaning of ‘discrimination on the basis of sex’ does, or does not, essentially imply. As we will see, however, this quickly takes us into deeper considerations regarding the nature of the relationship between causation, counterfactuals, and attributions of intent more generally. As such, Bostock presents us with a unique object lesson in the import of philosophy and logic for debates of immense public import.
The Court both in Bostock and elsewhere has connected the idea of discrimination ‘on the basis of’, ‘because of’, or ‘on the grounds of’, with the ‘but-for test’ for causation. According to this test, a person discriminates on the basis of a feature X when, but for that person having feature X, the employer would not have acted as they did. In other words, a person discriminates on the basis of X just in the case the following counterfactual holds: Had the employee (or potential employee) not had feature X, the employer would not have made the decision that they did. Because attributions of causation by discrimination have been taken to hinge on whether counterfactuals of this form hold, I first want to address just what the general connection is between causation and counterfactuals. After, I will address how this connects up the causal claims involving intention in particular. Finally, I will show how these considerations speak in favor of the dissent’s opinion in Bostock, and against the majority.
I. Causation and Counterfactuals
Philosophers have long recognized that there is indeed a connection between counterfactuals and causation. In general (though as we will see shortly, not always), whenever a statement of the form ‘A causes B’ is true, a statement of the form ‘Had A not occurred, B would not have occurred’ is also true. For example, if it’s true that the ball thrown at the window caused the window to shatter, then it should also be true that, had the ball not been thrown at the window, the window would not have shattered. Thus, as a general rule, causal claims entail corresponding counterfactual claims.
However, it has equally been recognized by philosophers that this entailment does not hold in the opposite direction: Counterfactuals of the form ‘Had A not occurred, B would not have occurred’ do not in general entail causal claims of the form ‘A caused B.’ That this is so can easily be shown by everyday examples. Consider a shadow cast by a chair sitting in front of a lamp at a certain position. While it’s true to say that had the shadow not been on the floor, the chair and light would not have been in the positions they are in, it’s clearly false to say that this implies the shadow caused the chair and lamp to be in the positions they are. Indeed, the opposite is true. The counterfactual claim thus does not entail the causal claim. Or consider a thermostat accurately reading ‘75 degrees Fahrenheit.’ The counterfactual holds that had the thermostat not read 75 degrees, it would not have been 75 degrees outside (because, by hypothesis, the thermostat is giving an accurate reading). However, the corresponding causal claim that the thermostat caused the temperature outside to be 75 degrees is, of course, false. Again, the opposite is true, and we have another case of counterfactual implication and causation coming apart.
Even the entailment from ‘A caused B’ to ‘Had A not occurred, B would not have occurred’, however, does not always hold. In particular, it does not hold whenever there is causal overdetermination. That is, cases in which an effect has two or more causes, each of which was individually sufficient for the effect. Return, for example, to the ball shattering the window case. Imagine that, instead of one ball being thrown at the window, two were, each of which was sufficient to shatter the window. Now consider only one of the balls—call it ‘Ball 1.’ Ball 1 individually was a cause of the window shattering. However, because it was only one of two balls that equally accounts for the window shattering, it’s false to say that ‘Had Ball 1 not been thrown, the window would not have shattered.’ The window would still have shattered, because Ball 2 could have done it all on its own. Thus, although it’s true to say that ‘Ball 1 caused the window to shatter’, it’s false to say that ‘Had Ball 1 not been thrown, the window would not have shattered.’
What these cases show is that counterfactual connections are at best only a sometimes-useful test for detecting causal connections. Sometimes causal connections imply counterfactual connections, sometimes they do not; sometimes counterfactual connections imply causal connections, sometimes they do not.
Apart from these more general ways in which counterfactuals and causation can come apart, however, there is one way in particular that will become relevant for assessing discriminatory causes. This is the case in which two or more properties covary, but in which only one of the covariant properties is the cause of a particular effect. Consider, for example, the properties ‘being on fire’ and ‘producing smoke.’ These properties covary, in the sense that whenever something is on fire it also produces smoke. Although fire is always attended by smoke, however, it is the fire that causes a thing to burn, not the smoke. Or consider the properties ‘being a gunshot’ and ‘emitting a loud noise.’ These properties also covary, in the sense that whenever a gun is shot, there is also a loud noise. However, in a killing, it is only the gunshot that is causally relevant, not the loud noise.
Note, however—and this is the key point—that in cases like this, because the properties covary, even the causally irrelevant property will appear in a true counterfactual about the causal event. So in the fire/smoke case, it will be true for example that ‘Had the house not been producing smoke, it would not have burned to the ground.’ But this, while true, tells us nothing about the real cause of the house burning down. The real cause was its being on fire. The counterfactual about the smoke is only true because ‘producing smoke’ covaries with the genuine cause. Similarly for the gunshot/loud noise case. It’s true that ‘Had the loud noise not emitted from the gun, the person would not have been killed.’ But this isn’t because the loud noise was the cause. It’s because the loud noise was connected to, covaried with, the genuine cause: the gunshot.
The first lesson I want to draw, then, is this: The ‘but-for’ test is meant to be a counterfactual test for causation in a variety of legal contexts—not only discrimination, but also in tort and criminal law. However, as we saw, the connection between counterfactuals and causation is not as tight as one might think. As a consequence, we ought to be very cautious in using such a test in too general or rigid a way. Whether the but-for test actually tracks causal truths will be highly context-dependent, and will often, in more complex cases, fail.
But did it fail in Bostock? I believe the answer to be a definitive ‘yes.’ To see why, we must first explore how counterfactuals relate to causation by intention. The sort of causation, that is, relevant to discrimination.
II. Counterfactuals and Causation by Intention
Intentional actions are those actions brought about by our beliefs and desires. If I trip and fall over a rock, this was ‘unintentional’: the action is not explained by any of my beliefs or desires. On the other hand, if I pick up the rock and throw it, this was ‘intentional’: it is explained by my belief that I can throw this rock, and my desire to throw it. Collectively, we can call the beliefs and desires upon which we act as our ‘reasons’ for acting. Intentional actions, then, are those based on reasons (rather than, say, reflex, force, coercion).
When we intentionally act, we often have some very specific property or properties in mind that figure in our reasons for acting. So, for example, I may buy a car because it’s red, and I love the color red. However, just as in the cases previously mentioned, the properties that figure in my reasons may covary with others that in no way figure in my reasons. I may have bought the car because it’s red. But it may also be that the dealership from which I bought the car only buys red cars from Toyota, and so at the dealership ‘being a red car’ and ‘being a Toyota’ covary. Let us stipulate, however, that I don’t care at all about whether my car is a Toyota or not. I only cared that it was red. In that sense, nothing about the car being a Toyota is causally relevant to me purchasing the car. It was only the redness I was concerned with. However, because the properties ‘being a red car’ and ‘being a Toyota’ covary at this dealership, the following counterfactual holds true: Had the car not been a Toyota, I would not have bought it. Like in the previous cases, this would be true, but would obscure the fact that it was not the property of being a Toyota that was causally relevant, but rather the property of being red.
In this sense, when we act intentionally we only select certain properties as important for our reasons. There may be others connected with those properties, but insofar as they are irrelevant to my reasons, they are irrelevant to explaining my action. And insofar as they are irrelevant to explaining my action, they are irrelevant for explaining anything that my action brings about. To take another simple example, I may pick out a t-shirt to wear because I find it comfortable. My wife may point out to me that the shirt is also an ugly color. In defense, I might rightly say ‘But that’s not why I picked out this shirt to wear. I could care less whether it’s an ugly color, I only care about how comfortable it is. I picked it out because it’s comfortable, and not because it has this color.’ This would be perfectly sensible and coherent, and again illustrates the underlying principle that applies both inside and outside intentional contexts: Just because two or more properties covary with another does not mean that they are all causally relevant in any context in which any of those properties produces an effect.
This, at last, brings us to the specific case of discriminatory causes.
III. Discriminatory Causes and Counterfactuals
When a particular effect—a firing, a refusal of promotion, etc.—is caused by discrimination, it is caused by an intentional act. In particular, it is caused by an intentional act in which some property associated with a protected class like being a woman, of a certain religion, or of a certain race, figures in the reasons for the action in a negative way. That is, discrimination occurs when one of these properties counts, for the discriminator, as a reason to injure the affected party in some way.
As we’ve seen, just because a pair of properties covary, does not mean that both properties are causally relevant in an intentional action that includes one of those properties in its reasons. This also applies, a fortiori, to the particular case of discriminatory acts. Imagine, for example, that I live in a town in which all mailmen are also members of a violent gang. I own a bank, and refuse to offer loans to people who are mailmen because I do not want to support such criminal activity. In such a case, is it true to say that I have discriminated ‘on the basis of’ someone’s being a mailman? Surely not. I may even quite like the profession of mailmen, and wish we had better ones in our town. In such a case, no reasonable person would say that the property of being a mailman figures negatively in any of my reasons for acting, including my refusal to give any particular mailman a loan. Rather, the only property I care about is being a violent criminal. It is this property that is the true cause for my refusal to give a loan in particular cases. However, because these properties (alas) covary in my town, the counterfactual is true: Had that customer not been a mailman, I would have given him a loan. In other words, but for the fact that this customer was a mailman, I would have served him. But once again, this is not because I discriminate against mailmen. It’s because I discriminate against the covarying property of being a violent criminal.
In discriminatory cases as much as in any other case, then, we must be careful in employing a strict but-for, counterfactual test for causation. The test may be useful in some cases, but it will go badly wrong in many others. In particular, it will go wrong in any case in which two properties covary, but the basis of someone’s discriminatory action was only one of those properties, and not the other.
IV. The Fundamental Fallacy of Bostock
This, in my view, is the fundamental fallacy of Bostock. The majority argues that discrimination on the basis of SOGI necessarily implicates discrimination on the basis of sex. To see why this is wrong, let us again draw our attention to Gorsuch’s key example. An employer fires a man who is attracted to other men. If the man had been a woman, he would not have been fired. Therefore, Gorsuch concludes, the property ‘being a man’ in this case was a but-for cause of the firing. However, as I have shown, simply pointing to the truth of the counterfactual ‘Had he not been a man, he would not have been fired’ is not sufficient to establish that the man’s sex was causally relevant to the firing. It must also be shown that the man’s sex figured in the employer’s reasons for injuring (firing) the employee. This, however, has not been shown. As Alito and Kavanaugh point out, the fact that they would have treated a lesbian employee in the same way demonstrates this.
What, then, kept the majority from appreciating this intuitive point? I believe the reason may have something to do with the fact that they have mistaken a concern for a relation with a concern for the things related. To see this, consider first that a prejudicial attitude towards same-sex-attracted individuals is, in itself, only a prejudice against the following relation obtaining: x is of sex Z, and y is of sex Z, and x is attracted to y. Note that this relation does not, in itself, implicate either sex. Of course, you must be one of the two biological sexes to satisfy this relation. But no one of those sexes is uniquely implicated in that relation. Now, the nature of this relation is that it can only obtain where the sex of x and the sex of y are identical. Thus, the relation will fail to obtain wherever the sex of x and the sex of y are not identical. And this is the key to the confusion in the majority’s reasoning.
Gorsuch contends that discrimination on the basis of sex is revealed in his hypothetical case because when you switch the sex of the employee from male to female, they would not have been fired. The reason it is discrimination on the basis of sex, Gorsuch infers, is because the maleness is the thing that determined the firing. But he is wrong. For it was not the change from male to female that, in itself, made the difference. Instead, it was the fact that in this case changing the male to female produced a distinct relation. A relation of heterosexual, instead of homosexual, attraction. The employer can coherently be characterized as caring only for the instantiation of the relation, and not the intrinsic natures of the things related—‘being male’ and ‘being attracted to males’, respectively. The reason, that is, the employer would have behaved differently had the employee been female is not because of maleness itself, nor because of being-attracted-to-men itself, but because putting those properties together produces a certain relation, and it is the relation that the employer’s real concern.
We should conclude, then, that the dissent is correct about what the ordinary meaning of ‘on the basis of sex’ does and does not entail, and the majority mistaken. There is a reason both the adopters of Title VII, as well as subsequent judicial and legislative history, has with near-uniformity not seen Title VII as protecting against SOGI discrimination. Gorsuch imputes to all such interpretive history a failure to grasp the logical implications of the statute’s ordinary meaning. I hope to have shown that this history instead reflects an intuitive grasp of the logic of intentional action, causation, and counterfactual implication as it relates to the ordinary sense of ‘on the basis of sex.’ Protecting minorities from invidious discrimination is a worthy political goal, but the Court has once again twisted the meaning, and logic, of the law in order to bypass the democratic process.
Well reasoned and I agree with your conclusion.