The compositional fallacy in language
What do the following three sentences have in common?
- She was fast asleep.
- The PhD accused the doctoral candidate of playing fast and loose with his citations.
- She held fast against the criticisms.
Not only do all of these sentences contain the word fast, they use the word fast in a way that flies in the face of its definition. If you broke into a native English speaker’s house, woke them up in the middle of the night and asked them to give a synonym for fast, they would call the cops and perhaps also give the word quick. But quick doesn’t fit in the above sentences at all:
- *She was quick asleep.
- *The PhD accused the doctoral candidate of playing quick and loose with his citations.
- *He held quick against the criticisms.
(By the way, a leading * is used in linguistics to mean a sentence is ungrammatical in a particular language – in this case Standard English.)
As a native speaker of American English, I find it pretty obvious that the latter two usages of fast are just idioms, but the first one is more opaque. In fact, I didn’t even think about how fast asleep is an idiom until last year.
According to my favorite procrastination tool, etymonline.com, fast’s ancestor fæst meant firm in Old English. The original meaning survives in cognates like modern German’s fest – and in these expressions, which really do make more sense if we take fast as meaning firm.
Is fast deviating from its normal meaning? In a sense, fast doesn’t even have a “normal” or “default” meaning. It isn’t being used any more or less properly in “That’s a fast car” than in “She was fast asleep”. All that the word fast ever does is take its meaning from the context.
I mostly wrote about the fast/firm thing because I find it interesting. But it also helps illustrate what I consider to be a common false conception about language:
Every word in a given language has a fixed meaning and every sentence gains its meaning by operating on the meanings of its words.
I will call this the compositional fallacy. Let’s see this fallacy in action!
There’s not one right way to use not
One of my favorite concepts in linguistics is the idiolect, one particular speaker’s version of a language. Language exists first and foremost in our minds, and the version of American English that exists in my head isn’t exactly the same as the one in your head, even if we grew up in a very similar environment.
(In my idiolect, a couple means “two or more”. If I say I “drank a couple cups of coffee this morning”, I probably mean two or three, but maybe four. I love the hover text on this XKCD: “If things are too quiet, try asking a couple of friends whether ‘a couple’ should always mean ‘two’. As with the question of how many spaces should go after a period, it can turn acrimonious surprisingly fast unless all three of them agree.”)
Here’s one sentence that’s not part of my idiolect, but is in the idiolect of many close friends and family members:
I could care less about that.
I’d rather say “I couldn’t care less about that” or “I don’t care about that”, but “I could care less” is indisputably another way of expressing uncaringness. When somebody tells me “I could care less about that.”, I always know exactly what they mean and I don’t make a fuss.
An endless array of grammatical tinpot dictators insists that it makes no sense to say you “could care less” and everyone who says that is stupid. The argument goes:
- There is a scale of how much a person cares about something
- We can compare how much a person cares about something to how much they care about another thing
- There’s a point of “zero caring” at which somebody can’t care any less than they actually do
- If somebody could care less, there exists some quantity of caring \( x \) less than the quantity of caring they actually have, \( y \), i.e. \( x < y \)
- But if \( y = 0 \), then we cannot have \( x < y \) because we also know \( x \geq y \)
- Therefore, “I could care less” actually means \( y > 0 \), i.e. the speaker has an amount of caring greater than 0
The incredible thing about the argument is that it manages to get all of the logical details right, but it makes a gigantic unstated assumption:
Every word in the sentence “I could care less” has only one meaning and we can combine these meanings to find out what it means.
Negative concord is a linguistic team signifying when multiple negative elements in a language combine to form a negative statement. Standard Spanish and Russian have negative concord; you would need to say something glossed as “Nobody didn’t say nothing” for “Nobody said anything”. Standard English doesn’t have negative concord. Well, most of the time. There are quite a few situations where multiple negative elements form a negative. Consider the following sentence:
- I didn’t eat any chili.
It contains two negative elements: not (reduced to the clitic n't) and any. Yes, any is negative, as evidenced by the incorrectness of the following sentence:
- *I ate any chili.
(By the way, I owe the any example to Steven Pinker’s wonderful The Language Instinct.)
It’s entirely possible that negation is acting in an irregular way in the sentence “I could care less”. Maybe not is a null morpheme. Or maybe, as Steven Pinker has also suggested, the speaker is being sarcastic! Then the negation is marked by tone.
Is it possible that the formulation “I could care less” ought to be avoided? I could buy that a certain phrase should be avoided due to ambiguity, even if it’s correct, but the purists are self-evidently not confused by the phrase. They know exactly what the speaker meant by “I could care less”, because otherwise, they wouldn’t know to correct the could-care-lesser for their alleged semantic crime.
Linguistics is a science, and in science, we discuss how things actually are, not how they’re supposed to be. It’s legitimately interesting to think about how standard negation rules fail in sentences like “I could care less”. Prescriptivism isn’t just thinly-veiled classism. It’s also boring.
Morphology isn’t helpful for language learning
I’ve been learning German for three and a half years, and when I first started, I regularly fell prey to the composition fallacy. I thought that I could learn German effectively by remembering how compositional the language is. All I needed to do to learn a new word was consider all of its components and think about how they combined, as simple as \( 2 + 2 = 4 \)!
German is, in the popular imagination, a Lego language where there’s no limit to your ability to make a new word. Morphology (word formation, basically) is a lot more predictable in German than in English – just think about how the (most common) word for gloves is Handschuhe, literally “hand shoes”.
What people usually mean when they call German a Lego language, though, is that it has a lot of really long words. An article in The Guardian presents Geschwindigkeitsbeschränkung, which is formed from Geschwindigkeit (speed) and Beschränkung (limit). It means… speed limit.
But there’s no substantive difference between these two words in speech! You can’t put another word or phrase inside the German word, but you can’t do that with the English phrase either. Nobody would say “I was driving over the speed local limit”; only “I was driving over the local speed limit”. Whether a sequence of sounds accepts new elements in the middle is, in fact, one of the most common tests for “wordness” in linguistics – and “speed limit” passes the test.
One hiccup is that the German word has an extraneous s inserted to make it easier to pronounce. Geschwindigkeit doesn’t end with an s, nor does Beschränkung start with an s. The English phrase speed limit does undergo a shift in stress, though. I pronounce limit with the first syllable stressed, but only speed is stressed for me in speed limit – and I’m confident most native speakers do the same.
I’m eliding a very obvious difference – Geschwindigkeitsbeschränkung has no spaces in it and speed limit has a space. Long words without spaces certainly look funny, but space usage is purely a matter of orthography. In written German, the norm is to write compound nouns without a space, and in written English, the norm is to write them with a space. It’s important to adhere to these rules in formal writing, but they don’t really tell us anything about English or German in and of themselves.
But I can hardly ridicule anyone that fell for the compositional fallacy when it comes to German nouns – it wasn’t long ago that I regularly did the same. The most usual word for decision in German is Entscheidung. ent- is a prefix for verbs or verbal nouns that often denotes a reversal or inversion; scheid is a verb root meaning cut; -ung is a suffix for turning verbs into nouns. In other words, a decision in German is literally the removal of a cut. That makes a lot of sense – in making a decision, you eliminate all the divisions between possible choices as you settle on one course of action.
Sadly, the story above is almost certainly no more than a fun fact. Maybe that’s how the word Entscheidung was originally derived, but there’s no way native German speakers are thinking about the process of “removing divisions” every time they use the word Entscheidung. There is a hypothesis in linguistics called Sapir-Whorf claiming language influences cognition, but it’s probably wrong, and the influence is subtle even if there is some truth to it.
Take the word bedenklich. be- is a prefix that makes a verb directly transitive in German (like in English verbs like befriend and bewitch); denk is a verb root that is a direct cognate of the English think; -lich is a suffix for adjectives related to the english -ly. Then we could translate bedenklich into English as bethinkable, which sounds like something worthy of thought. It certainly seems like I could tell someone Ihr Vorschlag ist sehr bedenklich – Your proposal is very (bedenklich)! Except I couldn’t, because bedenklich actually means problematic or troublesome!
Thinking about the respective etymologies of Entscheidung and bedenklich, we do end up with a vague idea of what they mean, but the particulars are defined by usage. Just think about all the words in English that mean something else than their etymology would suggest. incredible “literally” means “unable to be believed”, and it is very occasionally used in that sense, but most frequently incredible just denotes a strongly possible value judgment. There are also loads of negated words with no positive counterpart; I might say “He looks unkempt”, but I’ve never heard anyone say something like “He is remarkably kempt”.
I keep linking to etymonline, so here’s a quote from their home page: “This is a map of the wheel-ruts of modern English.” I absolutely love the “wheel-ruts” metaphor. Words constantly change in definition – they’re on the move. As they move, they leave behind imprints, and the imprints are what lend the words meaning, not their exact positions.
So how should you learn vocabulary? I’d advocate emotional context learning. All it means to know a word is to know what situations to use it in, which is a sort of knowledge I lacked with bedenklich.
Coda
What is the “atomic unit” of a language? In other words, what is the smallest unit of language we have that continues to be “language”? If you agree with me that the compositional fallacy is a thing (I love saying that things are a thing), then the answer can’t be individual words. Individual words don’t have any meaning until they are in sentences.
Once I reach a level of proficiency in German I’m happy with (I still don’t know what level that is), I’d like to learn Spanish, which I took for at least seven years in middle and high school without ever having taken it seriously. I think “taking Spanish seriously” means I need to learn phrases instead of words – I need to learn the proper and idiomatic way of expressing something. To do anything else would mean becoming the victim of the compositional fallacy.