Jason Bridges

University of Chicago

Phil340, W02-lecture 13-14 notes

Davidson on conventionalism and shared languages



I. Use and convention

“Nice Derangement of Epitaphs” is extremely dense; he covers a lot of ground, and makes a lot of claims.  I’m going to restrict my discussion of it to two issues: his criticism of the view that sentence meanings are conventional, and his own view of the ontological status of natural language.  As we’ll see that view is that the concept is totally unhelpful for understanding what actually goes on when people communicate with each other.


Davidson’s claim that language use does not involve conventions has astounded many commentators.  Partly this is because they are not quite clear on what it is Davidson is denying.


We have to distinguish three theses about word and sentence meaning:


1. The arbitrariness thesis: For most words in a natural language, that a word means what it does is arbitrary.


This is just the idea that the fact that a given noise or mark has a particular meaning, rather than some other noise or mark, is a matter of happenstance. This seems obviously true: it’s evidenced by the fact that different natural languages have totally different noises and marks assigned to those meanings.
There was apparently a view in ancient
Greece, reported by Plato in the Cratylus, that words were somehow especially cut out for their meanings, at least Greek words anyway.  Thinking of English that way would amount to the claim that something about the noise “dog” makes it especially cut out for referring to dogs.
No one buys this nowadays, certainly not Davidson.  It’s ridiculous.  Instead, everyone accepts that it’s totally contingent, a matter of the vagaries of history, that certain marks and noises ended up being assigned the meanings that they were.  Of course, some words might have a more than arbitrary connection to their meaning.  Onomatopoetic words, “meow”, “buzz” are partly determined by their resemblance to the sounds they refer to.  But in the vast majority of cases it’s arbitrary.

The only reasons I even mention this is that in the old Greek debate, the view that word-meaning relationships are arbitrary came to be known as the view that such relationships are conventional.  If this is all we mean by talk of convention, then everyone is a conventionalist.


Now to distinguish the next two theses, we need to be clear on the difference between a convention, understood in Schiffer’s sense, and a regularity.

Doing A is a regularity in group G iff members of G generally tend to do A.

Doing A is a convention in group G iff members of G do A because they know* that [fill in Schiffer’s analysis]

Whenever there’s a convention to do A, doing A is a regularity.  If there’s a convention to drive on the right in a given group than people in that group regularly do that.

But not vice versa.  English speakers as a rule eat every day.  But there’s no convention to eat every day.  There’d only be a Schifferian convention if, roughly, people ate everyday because they know everyone knows everyone’s expected to.  But that’s not why people eat.

Or consider: there’s a regularity among birds to build nests.  But there’s not a convention.  Birds have no conventions at all.  They don’t do anything because of why they know about the practice and expectations of other birds.   But all manner of behavioral regularities obtain among birds.

The notion of a regularity leaves out all the intellectual stuff, all the mutual knowledge and expectations and so forth involved in the Schifferian notion of a convention.


2. The regularity-in-use thesis: What a word in a natural language means is determined by a regularity in its use by speakers of the language.


3. The conventionalist thesis: What a word in a natural language means is determined by a convention holding among speakers of the language.


Now, the first thesis is nearly impossible to deny.  It’s just the claim that the fact that “dog” means dog is a matter of how speakers of English use it.

One way to see this is to think about what happens when words change their meaning.  What causes such changes?  Obviously, nothing could but a change in how speakers in general use it.  If more and more English speakers stopped using “dog” as a word for dogs, eventually it would stop having that meaning.  If English speakers keep on using “dog” as a word for dogs, then it can’t possibly stop meaning dog.

It’s completely up to us, taken collectively, what the words of our language mean.  This is not to claim that we can exert conscious control over the meanings of words in the language.  Unless we’re very influential people, we’re not going to be able to do that.  The point is just that nothing could possibly determine the meanings of words in English except the shared usage of speakers of the language.

This is not to say that anything goes.  It’s quite possible for the majority of people to use a word wrongly, at least for a period of time.  For example, many people use “comprise” as a synonym for “compose”.  It isn’t.  If people keep using it that way, and the usage spreads, however, it will probably change its meaning.  The dependence of meaning on regularities in usage is complex and hard to encapsulate in principles.  But it is there.


(Sometimes when people speak of regularities in use, they mean regularities that can be described in non-semantic terms.  Compare:

That “dog” means dog in English is determined by the fact that English speakers regularly use “dog” as a word for dogs.

That “dog” means dog in English is determined by the fact that English speakers regularly utter the word “dog” in the presence of dogs.

People say things like the second claim when they want to give a reductive analysis of meaning in terms of regularities.  This requires them to state the relevant regularities non-circularly.  We’ll talk about this kind of view when we get to Kripke’s book.

When I say that it’s undeniable that meaning is determined by regularities in use I don’t mean that the regularities can be specified non-circularly.)


Davidson doesn’t deny 2.  However, he denies that recognition of regularities of usage general plays an essential role in the understanding of utterances.  This leads him to deny that the notion of a natural language is not explanatory for understanding communication.  We’ll see this shortly.

Davidson does deny 3.  We’ll talk about his reasons now.


II. Davidson’s criticism of the conventionalist thesis

Davidson’s criticism of the idea of language as conventional is that the idea of a convention is of no use in explaining how communication is possible, and so it’s unhelpful to think of language in those terms.


Davidson’s question: What knowledge or ability would suffice for understanding a speaker?  Can we describe something that, were an interpreter to have it, would guarantee her understanding of the speaker?


Here again we see Davidson’s emphasis on the interpretive situation.


Suppose we understand a natural language as the conventionalist wants us to.  Would knowing that the relevant conventions obtain in the speaker’s linguistic community suffice for understanding her?  So if I know you are a speaker of English, and I know the conventions that obtain among such speakers, would that suffice for understanding you?

The answer, says Davidson, is no.


He develops the case in terms of malapropisms, misuses of words.  If someone were to say to you, “I’ve reached the very pineapple of success,” would it be sufficient for understanding what she is saying to understand the alleged conventions determining the meanings of the words she utters?  Probably not.  For most likely her intended meaning is that she’s reached the pineapple of success; she’s just misspeaking.  Trying to gloss her utterance in terms of the conventional meanings of the words will only mislead.

Therefore, knowing the relevant system of conventions of a speaker’s language does not suffice for understanding her.

A simple objection.


But now we should note two important points.


First, the objection does not need to be made in terms of malapropisms.  Even when someone uses their words in accord with the conventions of the group, assuming there are such conventions, knowing those conventions does not suffice for understanding the speaker.  For there is still a further thing one must know, namely that the speaker is adhering to the conventions.  In the malaprop case, we needed to know that the speaker wasn’t adhering to the conventions, something knowledge of the conventions cannot tell us.  In the ordinary case, we need to know the speaker is adhering to the conventions, which is also something knowledge of the conventions cannot tell us.



1. A convention to mean pinnacle by “pinnacle” obtains in group G.

2. Sally belongs to group G.

It does not follow that:

3. When Sally now utters “pinnacle”, she means pinnacle.


This is no more a valid argument than the following:

1. A convention to drive on the left obtains in the U.K.

2. Bob’s in the U.K.

3. Therefore, Bob will drive on the left.

Even if we add that Bob has always previously adhered to the convention, it still doesn’t follow that he will now.  Knowing these general facts doesn’t entail knowing what will happen now.  Even if the inference is warranted, it is not a matter of implication.


From how things generally tend to happen nothing logically follows about how they will happen in this instance.  There’s no logical guarantee

But what we need to know, in interpreting someone, is what is the case now—that is, what the person means now in uttering the words that she does.  That knowledge is not constituted by knowledge of conventions


But if this is what funds Davidson’s worry about the significance of conventionalism, then it turns not out to a worry about conventionalism in particular, but a worry about the significance very idea of a natural language, whatever one’s philosophical account of natural languages may be.


For consider the following:

1. In English, “pinnacle” means pinnacle.

2. Bob speaks English.

3. Therefore, Bob now means pinnacle by “pinnacle”.

This is an invalid argument.  The premises may both be true but the conclusion false.  Maybe Bob is purposely using the word in a funny way; maybe his tongue has slipped; maybe he’s trying to extend ordinary usage of the word in a new direction.

So precisely analogously to Davidson’s point about conventions, we can also say that knowing how to speak a speaker’s language does not logically suffice for understanding the speaker.  There’s always a ‘leap’ here, from how things generally tend to happen to how things will happen here.


What Davidson calls a “passing theory” is unavoidable.  The passing theory, as you know from the article, is supposed to what takes you from prior expectations about how the person would use word—based, say, on your knowledge of the natural language she speaks—to an understanding of what she’s actually saying now.

It’s peculiar to call this a theory, really.  But Davidson’s not wrong that something is needed here.  In the ordinary, non-malaprop case, the something is just the assumption that the person is using her words in the ordinary way.  But this is still a further assumption.  It is not something given to you by knowledge of the language.


Really, the point flows directly out of the regularity thesis.  What a language means is determined by a regularity, by how things generally tend to happen.  But how things generally tend to happen does not guarantee how they will now happen.  So knowing the former does not guarantee knowing the latter.


Does this mean that Davidson rejects the regularity thesis about natural languages?  No.  Davidson can accept that natural languages rest on general tendencies.  But for precisely that reason he will adopt what we can call:

The dismissive view of natural language: The idea of a natural language, however it is understood (be it in terms of conventions or otherwise) is of little or no help in explaining how communication is possible in principle


He adopts this view because he points out that a correct answer to his question (the question I called “Davidson’s question”) is not merely: knowledge of how to speak a natural language.


But one may well ask: granting this point about how to answer his question, does it really follow the idea does not play no essential role in explaining the possibility of communication?

Maybe it plays some other role than providing a complete answer to Davidson’s question.


Davidson does consider one such alternative.  One might grant that knowing a person’s language does not suffice for understanding but still hold that the idea of a natural language plays a crucial role in understanding the possibility of communication, by claiming that sharing a language, if not sufficient for communication, is still necessary.


But Davidson thinks that reflecting on the reasons why sharing a language is not sufficient for understanding will also lead one to the conclusion that it is not necessary. (though this transition is somewhat submerged in the essay we’ve all read).


Davidson is clearly right about this.  We can imagine the last two people on earth speaking different languages: say, English and French.  They could come to understand each other while each persists only in speaking their original language.

A traveler is sometimes the only one where she travels speaking her language and yet mutual understanding can transpire.


But could there be another alternative.  Could we hold that natural language are a crucial part of the explanation of how communication is possible, while allowing that understanding of natural languages is neither necessary or sufficient for communication?

Wiggins, in effect, argues that we can.  We’ll discuss this next time.