The Analytic-Synthetic Distinction

I am considering writing a book of philosophy, focussing on whether it is possible to make true statements about fictional entities. In today's post, I wish to discuss the analytic-synthetic distinction, a distinction first proposed by the philosopher Emmanuel Kant. Talking about this distinction is an entry-point into a conversation about fictional objects such as Sherlock Holmes and Spiderman, and I hope that getting my thoughts in order in this blog will help me when I get around to writing the book.

According to Kant, a true analytic proposition is true by virtue of the meaning of the words in it while a true synthetic proposition is true both by virtue of the meanings of the words in it and by virtue of one's experience of the world. Another way to put this is that a true analytic proposition is true in all possible worlds while a synthetic proposition may only be true in some possible worlds. The classic example of an analytic proposition is the statement, "All bachelors are unmarried". This proposition does not require knowledge of the external world for us to know that it is true– rather the word "bachelor" means an unmarried man and so no experience of the external world is required. The predicate is contained in the subject. If however we say, "All bachelors are lonely," this is a synthetic proposition, a proposition which we may tentatively suppose to be true until we discover a bachelor who isn't lonely, in which case we will be forced to say that this proposition is false. Its truth or falsity depends on observations of the real world. The proposition "All bachelors are unmarried", however, can never be shown to be false, can never be disproved, because it follows directly from the meanings of the words in it.

I have been thinking about this distinction since I was nineteen and was first introduced to this idea by the Otago philosophy professor Alan Musgrave. I sensed that there was a problem with it even then but it is only in the last couple of weeks that I have decided that it is possible to clearly state what is wrong with it. I will not be the first to criticise the notion of an analytic-synthetic distinction. The most famous critic was a chap called Willard Quine and his central argument is known as "The Indeterminacy of Translation". The precise meaning of Quine's critique has been a topic of considerable debate among philosophers ever since he proposed it in 1951. I will not be engaging with Quine's argument in this post but it is possible that I have hit on a similar critique and that I can present it more clearly today than he could back then. To reiterate – I am not deeply familiar with Quine's critique but it is possible that there might be overlap between my critique and his.

Before I advance to my own critique, I would like to discuss the world-view entailed by an uncritical acceptance of the analytic-synthetic distinction. It seems to me that people and philosophers who accept Kant's argument have a particular view of language. They regard language as universal, unchanging and independent of the real world and of how real people use words. This world-view is almost Platonic. According to Plato, the world of Ideas or Forms, of Universals, exists independently of the real world and in a sense precedes it. Forms are eternal, aspatial and atemporal; real phenomena are just shadows of these Platonic Forms. A form is the essence of a thing and exists in a separate realm. Plato, apparently, believed that we are aware of these Forms before we are born, before we are thrown into a world of appearances, of imperfect representations of these Ideas. The world of Forms, in this Platonic conception, is somehow more real than the material, physical world.

This world-view is almost mystical and, although many philosophers would bridle at the suggestion that their wold-view partakes of something obscurantist, I think it does. Nevertheless, some such world-view follows from the Kantian distinction. My own world-view, my own attitude towards language, is quite different. It seems to me that there are three unquestionable assumptions or axioms we can assert about language. These are:
1. Language is learnt.
2. Different people define words in different ways.
3. The generally accepted definition of a word is open to revision, can change.
If we accept these three claims, the analytic-synthetic distinction falls apart.

The most important of these three claims is the first but I shall tackle it later in this post. I'll start with the second assertion. Suppose I am talking to my friend Steven. Suppose in my own mind I define the word "bachelor" as "a man who has never married" but he defines the word "bachelor" in his mind as "a lonely man". We both know Chuck, an unmarried man. I tell Steven, "Chuck is a bachelor" (a synthetic proposition). Steven replies indignantly, "Chuck isn't a bachelor – he has heaps of friends!" (another synthetic proposition). Which one of us is right? It seems the proposition "Chuck is a bachelor" is true for me and false for Steven; conversely the proposition "Chuck isn't a bachelor" is false for me and true for Steven. This difference occurs even though we agree on matters of fact, that Chuck is unmarried but has many friends. Now, the reader may say that I am right and Steven is wrong, that Steven has an incorrect understanding of the word "bachelor". The reader may exclaim, "Everyone knows that the word 'bachelor' means a man who has never married!" But how can you be sure that this is true, that everyone shares this understanding of the word? It seems that in the dispute between Steven and me we must appeal to an external authority. I grab my Oxford English Dictionary and show Steven that the primary accepted definition of the word "bachelor" is "a man is who is not and has never been married" and he may or may not alter his mental definition of the word as a result.

We can go further. The idea that an analytic truth such as "All bachelors are unmarried" is true in all possible worlds is obviously wrong because we can imagine a world in which all the people and all the dictionaries define "bachelor" as "a lonely man", a world in which it is recognised that even married men can be lonely.

It might seem, then, that the only way to escape this impasse is to say that the meanings of words are codified by a kind of priesthood of lexicographers, linguists, scientists, and other experts, and that these people have always agreed on the meaning of words. Even if there is disagreement about semantics among ordinary people, among the laity, there is a generally accepted definition of every word established by some consensus of relevant professionals. We can always appeal to some external authority to ground our understanding of the meanings of words. However, this move also fails. Consider the proposition, "Pluto is a planet". Today, in 2019, we know this proposition to be false. In 2006, the word "planet" was redefined in such a way that it ruled out Pluto on the grounds that Pluto was too small and that some massive bodies in the Kepler belt were larger than it. The definition of a planet (according to Wikipedia) is "an astronomical body orbiting a star or stellar remnant that is massive enough to be rounded by its own gravity, is not massive enough to cause thermonuclear fusion, and has cleared its neighbouring region of planetesimals"and by this definition Pluto isn't a planet. But if we were to say, in 2000, "Pluto is a planet" we would be expressing a true proposition because the word "planet" was defined differently then. According to my Oxford Dictionary, published in 1998, the word "planet" is defined as "a celestial body moving in an elliptical orbit around a star" and, by this definition, Pluto actually is a planet. In fact the dictionary goes on to explicitly state that Pluto is a planet. It seems then that the proposition "Pluto is a planet" was analytically true in 2000 and is analytically false today. This causes all kinds of problems for Kant. We could say that the meaning of a word is established by a contemporary consensus of experts, by the most recent definition,  but we would also like to say that all propositions are either unequivocally true or false. What happens if two years from now the astrophysicists get together and agree on a new definition of "planet" which again includes Pluto? How can we know if the proposition "Pluto is a planet" is true or false?

I'll give another example of this problem, the problem that even the generally accepted definition of a word can change over time. In the nineteen-fifties, "homosexuality" was defined in the Diagnostic and Statistical Manual of Mental Disorders as "a sociopathic hatred of the opposite sex". To modern readers, this definition sounds absurd. We all know gay men who have female friends and lesbians who have male friends. In fact, the definition of "homosexuality" as presented by Wikipedia in recent times is constantly changing, is in a permanent state of flux. We only need a cursory understanding of Michel Foucault to realise that this is a serious problem, and that the health of our society depends on the best possible definition of homosexuality.

At this point in the post I will turn to the first claim, the assertion that language is learnt.

According to Kant, analytic truths are known a priori. This means that a proposition like "All bachelors are unmarried" is known independently of experience. Logically this would suggest that infants are born somehow already knowing that the word "bachelor" means "unmarried man". But of course this is absurd. At some point during early life, a child is either explicitly told that the word "bachelor" means "a man who has never married" or infers that this is what the word means from his or her observations of how others use this word. This seems to me quite obvious, yet it leads directly to the conclusion that analytic truths are known a posteori, that they depend on our experiences of having learnt a language.

We can approach this issue, the almost self-evident fact that we learn the meanings of words from others (from dictionaries, books, conversations), from another direction. Suppose I wish to discuss American politics with a friend and coin the word "glockle" to describe people who voted for Obama in 2012 and Trump in 2016. My friend obviously won't know what the word "glockle" means and so, before we can have a meaningful conversation about glockles, I first need to say to him, "I define the word 'glockle' as meaning 'anyone who voted for Obama in 2012 and Trump in 2016'. " This utterance of mine is neither true nor false – rather it is what John Seattle called a "speech act" and J.L Austin called a "performative". Like the statement "I now pronounce you husband and wife" when declared by a priest or the statement "I sentence you to two years imprisonment" when issued by a judge, my utterance has the effect of changing the world. Subsequently my friend and I can meaningfully discuss glockles – one of us could say for instance, "The Democrat presidential candidates need to win over the glockles in order to defeat Trump"– and such statements will either be true of false. Yet our ability to discuss glockles at all depends on my prior speech act, the utterance in which I defined the word, an utterance that is neither true nor false.

Of course it would be silly to suppose that "glockle" can be associated with some kind of Platonic Idea which I have discovered and named.

All this raises the fundamental and most difficult question of all. What is truth? The common sense answer is that truth is a relation between a proposition expressed in language and a fact about the world. This is the Correspondence Principle – "A" is true if and only if A. But the Correspondence Principle, when subjected to scrutiny, disintegrates. Suppose we say " 'Pluto is a planet' if and only if Pluto is a planet." Given our discussion of Pluto above, it is easy to see why the Correspondence Principle often fails. It doesn't provide a satisfactory account of the relationship between words and the world. I believe that truth is a kind of process, something dialectical, conversational, communal, a horizon that we continually approach but never arrive at. I believe I think in the idea of truth sketched out by Richard Rorty in Philosophy and the Mirror of Nature. I know that this is a vague claim but perhaps I will be able to elaborate on it in later posts. Or in the book I may or may not write.