1. Introduction
I came of political age in the early 1970s in the US, a period of intense activism around ending the Vietnam war, establishing the civil rights of Black people, and pressing for equal rights and opportunities for women, for gays and lesbians, and for persons with disabilities. At the time, we activists thought that a large part of our job – though certainly not all of it – was to correct many mistaken beliefs that were widely held by members of the American public: that the war in Vietnam was being fought to preserve democracy against Communist aggression; that Black people were inferior to white people and needed to be kept separate from them; that women were, by their natures, fit only for domestic and subservient roles in society; that heterosexuality was the only morally acceptable form of sexuality; and that disabled persons simply couldn't do certain things and needed to accept that. This is not to ignore the fact that powerful people had vested interests in maintaining US imperialism, racism, sexism, homophobia, and ableism, but insofar as we were interested in producing accurate and useful theories of the problems that beset our society, it was the propositions that were advanced in defense of these and other forms of injustice that needed to be articulated and refuted.
Those battles were, of course, never decisively won by progressives – all these social malignancies are, unfortunately, still with us. New generations of activists have joined us old-timers in battling these too-familiar cancers as well as more recently identified ills, such as transphobia. Among these new activists is a group of philosophers who are arguing that what is needed at this point is a revision not of propositions believed but of the concepts that compose these propositions. Here is philosopher Sally Haslanger:
On my view, to say that I am a white woman is to situate me in complicated and interconnected systems of privilege and subordination that are triggered by interpretations of my physical capacities and appearance. Justice requires that we undermine these systems, and in order to do so, we need conceptual categories that enable us to describe them and their effects […].
Much recent debate over race, in particular, seems to have become bogged down in the question whether this or that account of race can claim to be an analysis of our concept of race […]. I've maintained that my goal is not to capture the ordinary meanings of ‘race’ or ‘man’ or ‘woman’, nor is it to capture our ordinary race and gender concepts. I've cast my inquiry as an analytical—or what I here call an ameliorative—project that seeks to identify what legitimate purposes we might have (if any) in categorizing people on the basis of race or gender, and to develop concepts that would help us achieve these ends. (Haslanger, Reference Haslanger2005, p. 11)
Haslanger is saying several things here. Let's look at the first paragraph:
1. That gender is a hierarchical social system, and that being a woman involves standing in the lower rung of this system.
2. That gender justice requires dismantling this system.
3. That our current concept of WOMAN does not enable us to describe this system or its unjust effects.
Now I completely agree with the first two points. Haslanger and I are both materialist feminists – ‘materialist’ in the sense that we believe that oppression is not just a matter of what people believe, but also a matter of the objective physical and social factors that structure our options and choices. However, I disagree with Haslanger on the third point. I don't see why our concept WOMAN is any less adequate for activist purposes now than it was back in the twentieth century, when many of us were working with the same social analysis and pursuing the same political goals that Haslanger and I are both committed to today.
In fact, I actually believe that it makes no sense to call for the revision of our concept of WOMAN. That's one of the points I'll be arguing for below.
But to get everything out onto the table, let's look at the second paragraph. Here Haslanger is calling for a different kind of adjustment to our conceptual repertoire. She is claiming that:
1. Too much time is being spent (by activist philosophers) in trying to figure out what our ordinary concept of RACE, or WOMAN, is.
2. We should instead question whether it serves any good purpose to have such concepts.
3. If the answer to question 2 is ‘yes’ – that is, that there is an important purpose to our having such concepts – then we ought to try to either revise or produce concepts that better serve those purposes. This is the ameliorative project.
Here Haslanger is, first of all, criticizing a type of philosophical project, called ‘conceptual analysis’ that aims to elucidate the concepts that we (and I use the term ‘we’ advisedlyFootnote 1) use in ordinary life. The procedure was to propose a definition of a concept – e.g., KNOWLEDGE – and then test it against speakers’ intuitions, making revisions, if necessary, until the revised definition pretty much accorded with speakers’ judgements about what was and was not knowledge. Haslanger is suggesting that we give up on projects like that, and instead ask what purpose a concept is meant to serve, with the idea that we might either give up concepts that serve no good purpose (like, perhaps, RACE) or else revise existing concepts to enable them to serve their purposes better.
Although I agree with Haslanger's criticism of conceptual analysis as it has been pursued in philosophy, I must again disagree with Haslanger on a couple of counts, regarding two presumptions that lie behind her ameliorative project. First, I think, once we have a concept, we are stuck with it. That is, once we have the conceptual resources to represent a category, we have them, and cannot get rid of them without something like brain surgery. Second, I think that the only ‘purpose’ a concept has is simply to enable us to refer to, or pick out, objects and properties in the world. And referring to or picking out an object, set of objects, or property is neutral with respect to whatever it is we want to assert about the objects or attribute the property to. I think that Haslanger, and other proponents of ‘amelioration’ or ‘conceptual engineering’ are failing to distinguish the essentially referential function of concepts with the assertoric function of propositions.
I'll try to make these abstract points more concrete by working through the cases of the concepts RACE and WOMAN, but first, some general philosophical background.
2. Background
Let's start with the distinction between concepts and words.Footnote 2 Words, or terms, are the primitive elements of language. Put together in accordance with the syntactic rules of a language, words strung together form sentences, which can be used to assert things, to ask questions, to issue commands, and do all the other things we use language to do.Footnote 3 Although the general structure of naturally occurring human languages is governed by innate rules, the semantics of language – the determination of what particular words refer to – is up to us. That the sound ‘dog’ refers, in English, to the particular animals it does is a matter of convention. It's the fact that speakers of English all use the same word to refer to the same animals that enables us to communicate with each other. We know that this regularity – using ‘dog’ to refer to <dogs> is a convention, as opposed to a natural law, because different societies have different conventions to do the same job – the French use the word ‘chien’ and the Germans use the word ‘Hund’ – just as the English regularize automobile traffic by requiring drivers to keep to the left, while Americans do it by requiring drivers to keep to the right.
Conventions are thus, in an important sense, our creatures – we made them, we can change them. No one individual can do this unilaterally – conventions essentially involve social cooperation – but enough individuals can, with enough time, manage to do it. We have seen plenty of cases where a word changes its reference: ‘awful’ used to pick out a quality that (Americans, at least) now express with the word ‘awesome’; the word ‘hopefully’ used to be an adverb of manner, so that one could say ‘She walked hopefully toward the train’. Nowadays, the word functions as a sentence operator, meaning, roughly, ‘It is to be hoped that …’, rendering the previous sentence virtually uninterpretable for many speakers (e.g., students in my undergraduate classes). In science, too, there can be what's called ‘semantic drift’ – the word ‘atom’ originally (in the Greek, or so I'm told) referred to the ultimate constituents of reality, whatever they were. As modern science developed, however, it became clear that the things regarded as atoms were in fact composed of even smaller things. Did the discovery of atomic structure mean that ‘atom’ changed its reference? Or did it mean that we discovered that (cleaving to the original reference) the things we had called ‘atoms’ were not atoms? This is an unanswerable question – the relation between words and their references is not definite enough for there to be a clear yes or no.
One arena where there can be a clear yes or no, however, is the law. In order to prevent vagueness in enforcement, certain terms are explicitly defined: ‘citizen’, ‘contract’, ‘homicide’, etc. These legal contexts are also ones in which we can intentionally and abruptly change the reference of a word. In the US, this can be done through a constitutional amendment, such as the ones that made formerly enslaved Black men citizens, and made women of all races eligible voters. In Canada, it took legislation to redefine ‘marriage’ so as to make marriage legal between members of the same gender. Such changes can also be accomplished legislatively in the US, as happened with the gradual elimination, in state law, of the abhorrent ‘marital exclusion’ in rape law. It used to be, in every state in the US, that ‘rape’ was defined or understood in such a way as to make it legally impossible for a husband to rape his wife – the law took it that a wife, in marrying, gave her husband blanket consent to sexual intercourse, making rape a legal impossibility (see Ryan, Reference Ryan1995).
But all this is about words. What about concepts?
Just as words are the primitive components of sentences, concepts are primitive components of thoughts. Concepts are prior to words. We know this for several reasons: (1) concepts are required in order for language to be acquired;Footnote 4 (2) fascinating research on infants makes clear that from the earliest testable ages, human babies possess and employ a rich stock of concepts representing the general properties of physical objects and animals, as well as social and even moral properties of persons, all antecedent to the sort and quantity of experience that would have been necessary for them to pick them up from observation,Footnote 5 and (3) there's strong empirical evidence that non-linguistic creatures – quite a large number of them – are capable of thought.Footnote 6 According to the developmental psychologists and ethologists whose work I follow (or try to), many non-human animals are capable of quite sophisticated reasoning. All of our closest primate cousins – chimpanzees, gorillas, bonobos, orangutans and many species of monkey – engage in mind-reading, that is, they seem to possess, natively, an understanding that others of their species have mental states, and an appreciation of the structure of those mental states. Many other species – horses, dogs, cats, elephants – also appear to have such capacities. (Interestingly, dogs seem to be better interpreters of human minds than are our closest relatives, the chimpanzees.Footnote 7) Since these creatures do not have overt languages as human beings do, the comprehension and reasoning displayed by these non-human creatures requires an internal, language-like system of representation.
The philosopher Jerry Fodor made this theoretical postulate explicit and argued that humans and many other creatures possess a language of thought – ‘Mentalese’, if you will. On this view, concepts are essentially the words in a native system of mental representation.Footnote 8 I'll speak of our conceptual repertoire as the postulated set of Mentalese words that form the basis of the thoughts we can think.
Evidence that we have such an internal representational system, one that is independent of whatever public language we acquire, comes from many sources: everyday experience as well as carefully designed and controlled psychological experiments. Here are some samples.Footnote 9 First, from everyday life:
1) ‘tip-of-the-tongue’ phenomena
You probably have had the experience of knowing that there is a word that expresses the idea you have, but being unable to produce it. This suggests that you are tokening the concept, and just cannot locate in memory the word that expresses it. I might be able to induce the experience in you right now. See if you can come up with the name for the object that fits the following description:
an archway in a garden or park consisting of a framework covered with trained climbing or trailing plants.
Do you feel that you know what this is, and just cannot think of the thing's name? (The answer is ‘pergola’.)
2) detection of ambiguity
Consider the following sentence: ‘Visiting relatives can be tedious.’ Did you, right away, appreciate that there were two different things that this sentence can mean? Now, once you recognize that the sentence is ambiguous, you can probably formulate each of the two readings in natural language (‘It can be tedious to visit relatives’ vs. ‘Relatives who are visiting can be tedious’). But I'll bet that you recognized that there were two different readings before you actually formulated them in your head.
3) planning
When you set out to do something, you often make a plan, but you don't often explicitly formulate the plan in language, even to yourself. Think of cooking a familiar dish. You often go about collecting the ingredients, preparing them, cooking them in a certain sequence, without formulating the sequence explicitly. The plan is in place, though, as you can tell from what happens if there is a glitch. ‘Oh no,’ you think, ‘I forgot to get eggs!’ This is a thought you might articulate explicitly, but what about the intention to get eggs from the fridge?
The point that plans and intentions needn't be explicitly formulated in a natural language is also borne out by the fact that non-human animals can and do produce novel plans. Crows (actually, rooks) confronted with a novel situation – a tube of water with a tasty grub floating on the surface, and a pile of small stones beside it – were able to figure out to drop stones into the tube to raise the water level high enough for the crow to pluck the grub out of the water (see Jelbert et al., Reference Jelbert, Taylor, Cheke, Clayton and Gray2014). (There's also my dear departed dog, Freya, who spent one entire morning working out how to get hold of a donut someone had tossed onto the sidewalk in front of our house: ‘Gotta get outside. How do I get outside? Get Louise. Why do I want to get outside again? Oh yeah, the donut!’)
The fact that intentions must be formulated in thought, using concepts, is a point that I'll make much of in my criticism of the project of amelioration, to which I now turn.
3. Against Amelioration
Amelioration, if it is possible, must grapple with the following: what is the connection between Mentalese words – concepts – and words in human public (or what is confusingly called natural) languages? This is a vast and unsettled question, not just for advocates of amelioration, but for everyone. One view that has been popular in philosophy, and in some branches of linguistics, is that concepts are actually determined by words in natural language – that language determines the expressive potential of thought – this is a view championed by Benjamin Whorf (Reference Whorf1956), and defended in a limited way by psychologists such as Lera Boroditsky (Reference Boroditsky and Nadel2003). But the thesis that (public) language determines thought founders on the evidence for structured thought in pre-linguistic humans and non-linguistic animals that I alluded to above.
What about the other way around? Does our native conceptual repertoire determine what thoughts we can think, or can language extend the boundaries of thought? Here we need to separate two questions: 1) Can we acquire new concepts? 2) If we can, do we acquire them through language? I think the answer to the first question is ‘yes’, but that the answer to the second question is ‘no’. That is, I think that we certainly do acquire new concepts – that we can and do enrich our conceptual repertoire – but that we don't do it by means of language. When we coin new words – as we certainly do – we can only give sense to these words by recombining concepts (words in the language of thought) that we already possess. Usually new words are essentially abbreviations of complex expressions in the language of thought. Insofar as new words can be defined in terms of combinations of old words – that is, essentially, definitions – there is no reason to think that there's any expansion of our conceptual repertoire involved. This is important to keep in mind with respect to the project of amelioration – there are certainly new words that come into use as the result of expanding knowledge of the world or cultural changes. But the addition of new words does not necessarily involve the generation of new concepts.
Fodor made an important distinction between the acquisition of a new concept – which he thought was not only possible, but something that happened all the time – and the learning of a new concept, which he thought was impossible. ‘Learning’, as he used the term, requires there to be a rational connection between the external circumstances that give rise to a concept (or to a thought) and the concept that results. If we could learn concepts, the process would have to proceed this way: we would have to first generate a hypothesis about what the reference of the concept was, and then test to see if that hypothesis was correct. So if we were learning the concept WOMAN, we would have to first think, say, ‘I bet that to be a WOMAN is to be an adult human female’ and then try to figure out whether our hypothesis was correct – maybe by saying of various individuals we judged to be adult human females ‘So-and-so is a woman’ (using the word as a proxy for the concept) and seeing if the others in our linguistic community accepted such sayings, or corrected us. Acceptance or rejection of such sayings would be rational evidence about whether our hypothesis as to the reference of the new word was correct.
But the thing to notice is that, if we are to ever successfully carry out this process of forming and confirming hypotheses about what the ‘new’ concept referred to, we already have to possess concepts which, combined in the right way, refer to exactly the same things as the new concept we are learning. Therefore, if there is some process by which we can make genuinely new additions to our conceptual repertoire, it cannot be by first postulating what the new concept stands for, and then getting evidence whether our postulate is correct or not. The postulate has to already contain the conceptual resources for picking out what the new concept picks out. I'll later explain the pertinence of this point to the debate about amelioration.
Here's another essential point about concepts, also pertinent to the amelioration debate: concepts are simply used in thought to pick out the subjects of thought; tokening a concept is not to say anything about those subjects. To think to yourself ‘Women bear children’ is to, first of all, refer to women (whoever those individuals might be), and then to attribute to women (those individuals the concept picks out) the property of bearing children. Now this thought can be true or false – it is true if the subjects of the thought, women, possess the property that is attributed to them in the thought, namely the property of bearing children. Some of you might judge that this thought is not true, because of the fact that not all women do bear children. Some of you, though, might think it is true, so to speak, ‘in general’ the way it's true ‘in general’ that dogs have four legs or bark, or that birds fly. These claims all have exceptions, but are usually accepted as true by speakers of English. The case of the sentence ‘women bear children’ is one of those cases, by the way, where the linguistic representation of thought can be ambiguous – the same sentence in English can express any of these more specific thoughts: ‘All women bear children’ (false); ‘Bearing children is a distinctive property of women’ (true);Footnote 10 or ‘Most women bear children’ (true).
The point I am focusing on right now, however, is that the function of the concept WOMEN in thought is just to pick out the class of women. But now you probably want to know – how does this concept do this work? What about the concept WOMAN connects it to women? And also – who are these individuals, women, to whom this concept is connected?
I cannot answer the first question, although I and many other philosophers have tried to provide theories of how concepts might be connected to their references. (There are problems with all of the theories that I know of.Footnote 11) But we can say something about the question of what the references of our concepts are. That is, with respect to the question ‘who are these individuals that the concept WOMAN picks out?’ we have an idea how to provide an answer. The general program is, basically, to gather up the individuals who a large number of thinkers will judge to be women, and then see what these individuals have in common. That's the program, but it can be – and is in this case – difficult to carry out.
There is, first of all, the fact that thinkers sometimes make mistakes – they might see someone who is dressed in a way that is atypical for women in their culture, and mistakenly judge that individual to be a man (i.e., not a woman). This is a perceptual mistake. But there's another kind of mistake an individual might make. They might have a mistaken belief about what properties an individual has to have in order to be a woman. Such a person might believe, for example, that in order to be a woman, an individual must be capable of bearing children. This person would then judge an infertile female, or a post-menopausal female, or a transwoman to be not a woman.
Another inherent difficulty in trying to characterize what it is to be a woman – that is, what it is to fall under the concept WOMAN – is that women (whatever they are, exactly) will have many properties in common. In fact, most women have a great deal in common biologically – they have the same chromosomal properties (two X chromosomes), they have the same general endocrinology (hormonal properties), and the same morphology (body shape and structure).
But not all women share these properties. Many of you will be aware of the existence of intersex individuals – individuals who do not have the set of biological properties typical for either females or males – yet many intersex individuals are women. And apart from women who are biologically intersex, there is the fact that many girls and women lose one or more of these biological properties either permanently or temporarily through illness or accident. These considerations prompt many theorists – including myself – to argue that the commonalities that actually unite individuals into the category <women> are a matter of the social role women play in human societies.Footnote 12 On this view, women are united as a class by a system that assigns distinctive social roles, governed by distinctive norms, about dress, behavior, and occupations to individuals, usually at birth, on the basis of what is, or is perceived to be their biological properties. Feminists who believe this will often say that WOMAN is a social concept, not a biological concept, meaning that our concept WOMAN picks out or is connected to individuals who are assigned or who come to play a certain social role, one that is generally and for the most part attached to the biological property of being female.
Now this disagreement about whether our concept WOMAN picks out biological females or individuals who occupy a certain social role, is, in my view, a question of fact about what our concept actually picks out. It may be – it is – a very difficult question to answer, but there is a fact of the matter, and it is independent of what we might believe or want to be the case. What our concept does pick out is pre-determined; and what it picks out determines whether our thoughts involving the concept are true or false.
This may sound very bad to progressive ears. What I'm saying is that if our concept WOMAN picks out individuals who have certain biological features, then this will have the consequence that the thought ‘transwomen are women’ will be false. And that is – I absolutely agree – a difficult conclusion to accept. And this is where amelioration might look attractive. Suppose, you might think, our current concept has the reference <female human beings> – can't we just change it? Can't we revise – or re-engineer or ameliorate – that concept so that it has a reference that includes transwomen? The answer, I'm afraid, is ‘no’. In order to make such a change we would have to use the old concept in framing our intention to change its own reference – it would amount to thinking ‘WOMAN (old) should mean WOMAN (new)’ but such a thought would be incoherent. The first occurrence of the concept WOMAN in that thought would have to retain the old reference, which would put it into conflict with the intended reference of the new concept. We cannot alter the reference of a concept we already have by wishing into existence a different concept. We can add to our conceptual repertoire a new complex concept: WOMAN-OR-OCCUPANT-OF-THE-SOCIAL-ROLE-GENERALLY-OCCUPIED-BY-WOMEN, but that won't change the reference of the component concept WOMAN. But this all seems politically inadequate, offensive to transwomen (even if their own concept WOMAN has, unbeknownst to them, that content) and even morally repulsive.
I think this sort of consequence is one of the things that advocates of amelioration have in mind when they call for revision of our concepts. But if amelioration is impossible, what can we do? Quite a bit. In the first place, what the impossibility of amelioration shows is simply that if the reference of WOMAN is determined by biological properties, then it cannot be changed to be determined by social properties. But the if-part of that conditional might be false. Indeed, the very strong intuitions that many of us have that transwomen are women is in itself evidence that the reference of our concept WOMAN is not determined biologically after all. If WOMAN is in fact a social concept, and not a biological concept, that means that those who think the thought TRANSWOMEN ARE NOT WOMEN are wrong, as a matter of fact, and in virtue of the reference of their own concept WOMAN. Those of us who realize that this thought is false must then work to persuade those who think this that they are wrong. This would be analogous to persuading someone in the eighteenth century who thought WHALES ARE FISH that they were wrong, because whales are in fact mammals. The mistake such a person would be making is the same as the mistake made by someone who thinks TRANSWOMEN ARE NOT WOMEN – they'd be incorrect about the properties that actually make an individual an individual of the type in question. If this turns out to be the problem, it would be a very welcome result, politically speaking.
There's another issue about the properties that determine whether an individual falls within the reference of a concept or not. Haslanger, in the quote above, says that to say that she is a woman is ‘to situate her in complicated and interconnected systems of privilege and subordination’, and then goes on to recommend that we ‘ameliorate’ our concept of women to reflect this fact. Now I agree with Haslanger that the gender role, being a woman, is part of a hierarchical system of privilege, and that it has been a regularity in human societies that the women in those societies – that is the individuals who are assigned the woman-role – are subordinated, often severely so, relative to the men. It may even be that the description Haslanger gives expresses a property that in fact coincides perfectly with the set of individuals that our (current) concept WOMAN picks out. If it is, then there is no need to revise or ‘ameliorate’ our concept. If it is not, then it is impossible to revise our concept, if my arguments above are correct.
The crucial point to realize here is that whatever the properties are that determine what the concept WOMAN refers to, possessing the concept does NOT mean knowing what properties connect the concept to its reference. So even if Haslanger is right, that being subordinated is one of the properties in virtue of which WOMAN refers to women, no one has to know that in order to use the concept. Compare: WATER refers to substances with the chemical structure H2O – that chemical structure is necessary to being water. But for most of human history, and probably even today, most people who have the concept WATER don't know this.
But suppose we did introduce a new, abbreviatory concept WOMAN(sub) that is defined to mean INDIVIDUALS-WHO-ARE-IN-A-SUBORDINATED POSITION RELATIVE TO MEN. Then the thought WOMEN(sub) ARE SUBORDINATED would not be an interesting, substantive thought. It would be akin to thinking BACHELORS ARE UNMARRIED. It would express a conceptual, rather than an empirical, discoverable fact. And it would be a fact of no political significance. That is an unwanted consequence. So amelioration is not something that is politically desirable, even if it is possible.
There are two more things I want to say to mitigate what may appear to be morally or politically discouraging consequences of the view of concepts I've been defending.
First, concepts, as I pointed out earlier, are not propositions – they are elements of propositions. That means that having a concept with the reference <biological females> does not condemn a person to being transphobic. Nor does having an inclusive concept ensure that one is trans-friendly. One can construct the proposition TRANSWOMEN SHOULD BE ALLOWED TO USE ‘LADIES’ ROOMS’ with either concept, and one can deny it using either concept. The content of a proposition is not determined by the subject concept – it's determined by that concept, plus the other concepts, plus the structure in which they are put together.
Second – and here I want to return to what I regard as the most fundamental matter in this debate. The political work that urgently needs to be done neither requires nor would obviously be served by conceptual change. Where does that leave us? Well, the reference of concepts is not under our control, but the reference of words is. And that is hardly a trivial matter. We have the power, and politically the duty, to ensure that our laws, institutions, and ways of speaking respect each other's identities, and realize our ideals of justice. Redefining ‘marriage’ so that the laws recognized loving unions between members of the same gender as equivalent to those involving members of different genders; redefining ‘rape’ by removing clauses in the law that permitted husbands to sexually assault their wives with impunity; explicitly adding the category ‘non-binary’ to permissible options for legal identity documents – all these measures have had real, salutary and even transformative consequences for many people. Explicitly stipulating in law that the categories ‘woman’ and ‘man’ are to include transwomen and transmen would do the same. Language, as I've been saying, is our creature, and language has power.
Of course such changes will not, in themselves, change anyone's opinions about sex, gender, and justice. To the extent that a just social order depends upon what people believe, we must still do the taxing and uncertain work of trying to change what people believe. But to the extent that people are motivated to obey the law, legal reforms can change what people do. And that is absolutely not nothing.
