Jason Bridges

University of Chicago

Phil340, W02-lecture 14-15 notes

Wiggins on conventionalism and natural languages



I. Wiggins’ criticism of the conventionalist thesis

Note that Wiggins, in “Languages as Social Objects”, couches his criticism in terms of Lewis’ account of convention, which differs in some details from Schiffer’s.  But the accounts share the basic thought that conventions are regularities sustained by a shared belief that everyone conforms to the regularity in turn giving rise to conformity.  We can translate Wiggins’ criticisms in terms of Schiffer’s account of convention, and I will do so.

Remember also that we only got as far as Schiffer’s conventionalist account of non-composite whole-utterance-type meaning.  But he’s going to argue in the end that meanings of words are determined by conventions as well.  The details of how he gets from one to the other are unimportant for understanding Davidson’s and Wiggins’ criticisms.


So again:


The conventionalist thesis about word and sentence meaning: What a word or sentence means in a natural language is determined by a convention (or set of conventions) among speakers of that language.


Recall Schiffer’s account of convention:

There exists a convention in group G to do A iff members of G mutually know* that:

1.      There is a precedent or explicit agreement to do A,

2.      Because of 1, members of G expect members of G to do A

3.      Because of 2, members of G do A


We contrasted this idea of convention with that of a mere regularity, a way that things generally tend to happen.  If you have a convention among a group, you have a regularity.  But not necessarily vice versa.  What’s distinctive of conventions is that the regularity exists because people know about it and on that basis expect others and themselves to sustain it.  And there can be a regularity that is not held in place for that reason—e.g., the regularity that people who want to go across the room will walk instead of crawl there.  Or regularities in the movements of the planets.  Etc.

In the case of meaning, the relevant conventions, recall, will ultimately concern the speaker meaning of utterances of the linguistic expressions in question.


What does this account suggest about the ontological status of a natural language?  That is, what kind of object does it suggest that English is or that French is?


The conventionalist view of natural languages: A natural language (English, French) is a system of conventions.


On this view English is the same kind of thing as say, Victorian etiquette, or Freemason ritual.  Victorian etiquette is a system of conventions governing behavior that held among a certain group of people in England in the late 19th-century.  Everyone in that group knew there were precedents for doing certain things, thus knew there was an expectation, etc.  Similarly, English is a system of conventions, now not governing how one acts at a party but rather what one uses words to mean, holding among a huge, far-flung group of people over many centuries.  It differs from Victorian etiquette and Freemason ritual in the breadth of the group in question and in the type of behavior in question.  But that is all.

Note that there is no such thing as a convention given Schiffer’s analysis that does not actually hold among an actual group of people.  We could imagine conventions that no one follows, but these are imaginary conventions.  This comports with the fact that a natural language is something that is or was at one time actually spoken.  I could make up a language that no one speaks, but for just that reason it couldn’t be a natural language.


Wiggins has two main objections to the conventionalist theses:


            Wiggins’ objections to conventionalism

            1.  The relevant group cannot be defined non-circularly.


Among what group of people are the relevant conventions supposed to hold?  We can identify the group among whom Freemason rituals hold: Freemasons.  And these can be independently identified in terms of their membership in the club.  And we can identify the people among whom Victorian etiquette obtained: late 19th century upper-class Brits.


Important point.  The identity of the group enters in in two ways: both determining the scope of the convention, and determining what the members of the group themselves aim to conform to.  So the members themselves must have a notion of the relevant group if there is to be a convention.


It is sometimes a tricky thing to determine, though.  Among what group does the convention to drive on the left side of the street currently hold?  Not British citizens.  Rather, whoever is driving in the U.K.  So the group is constantly changing as people fly in and out of the country.


What about English?  Can we define the group geographically?  No.  They’re all over the place.  And someone could speak English when surrounded entirely by no other English speakers.  By culture of religion?  Obviously not.

There seems no answer possible other than: the group of those who speak English.  But this is problematic.  For English simply is the set of conventions at issue, on this view.  There is nothing more to being an English speaker than to adhering to the conventions in question.  But then we can’t identify the conventions.  Given Schiffer’s account of convention, there has to already be some fixed group in view if it is to make sense to speak of a convention.  (Wiggins actually doesn’t take the objection this far.  He doesn’t note that there’s no possibility even of specifying what people aim to conform to without a way of identifying the group in question.


2. The thesis misrepresents the intentions of speakers.


It may well be fair to say that the reason people adhered to Victorian etiquette was because they knew there was an expectation to do so because there was a precedent for doing so.  But is that a plausible thing to say about why people mean what they do by their words?

Wiggins thinks the answer is obviously no.  English speakers do not use words as they do because they believe other English speakers use words that way.  Even if they do believe that, it is not their reason.  Their reason, for Wiggins, is simply their belief that that’s what the word means in English.

But to say this is to deny that it is a Schifferian (or Lewisian) convention that determines what a word means.


This objection reflects Wiggins’ view that languages are social objects.  Of course we could understand the term, ‘social object’ in a thin enough way so that a system of conventions would count as one.  But when Wiggins claims that languages are social objects, he means it in a more substantive, specific way.


One way to understand what is at stake is by contrasting a pair of views from the philosophy of social science about social phenomena, i.e., phenomena whose existence depends on a group of people.  (Note that the following terms have nothing to do with individualism and holism as they're discussed in other contexts.  The ‘individualism’ that is Burge’s target is something unrelated.):


Individualism: Any social phenomenon can be given a reductive analysis in terms of independent properties of, and relationships among, individual people.  (I.e., facts about social phenomena are nothing ‘over and above’ facts about individuals.)


Holism: Individualism is false.


(‘Independent’ here means: not stated in terms presupposing the concept of the social phenomenon in question.)


What does this mean?  Consider some social phenomenon P, say, the phenomenon of girls having coming-out parties in the south, the bellbottom fashion trend, the United States.  Individualism says in effect that we can explain what it is for there to be debutantism, the bellbottom trend, the U.S., in terms of some independent circumstance or condition that holds among the relevant group of people.


Take a simple case of a social phenomenon: everyone in this room is angry.  Call this room-257 anger.  Suppose this happens today and it’s so startling it makes the news and so forth.  People wonder at the causes of room-257 anger.

Now whatever the causes, it’s clear that we can give an individualistic analysis of this phenomenon.  What it is for this phenomenon to obtain is just for the following fact to obtain: everyone in room 257 is angry.

Can we do the same for a fashion trend?  Can we say what it is for this trend to obtain among a group only in terms of properties of people in the group —say, that most people in the group wear bell-bottoms, or more complexly, that influential people wear it and others imitate them or some such?  A tough question whether this works out in the end.

What about the existence of the United States?  Can we explain that in individualistic terms?  Seems still harder.


What about a natural language?  If the conventionalist thesis were true, if such things were systems of conventions, then we could give a reductive analysis of a natural language in terms of independent properties of individuals in a relevant group.  For the existence of a Schifferian convention consists wholly in their being various beliefs and intentions on the part of individuals in the relevant group.   The presence of a convention in a group is explained completely in terms of beliefs members of the group have about what other people have done and will do and thus what they’re prepared to do themselves.

It is a major part of the substance of Wiggins view of a language as a social object that he denies the existence of a natural language can be explain in individualistic terms.  Talk about what is true of a natural language is autonomous, Wiggins says, by which he means that it is irreducible to talk about what is going on with individuals.  It is something over and above these individualistic facts.

The social-object view of natural language: Facts about a natural language cannot be reduced to independent facts about individuals.


Does this mean that Wiggins denies the regularity thesis, that what words mean in a natural language depend on regularities in their use?  He does not; he acknowledges, as anyone has to, that it is ultimately how people actually use words that determines what they mean in the language.  He certainly does not want to hold, a la those ancient Greeks we discussed last time, that words have their meaning in virtue of intrinsic natural features.

But Wiggins wants to claim that the sense in which language supervenes on, depends on use, is irredeemably complex and in principle incapable of being cashed out in reductive terms.

Consider this analogy.  We may well accept that the character of this table is determined wholly by the character of the molecules composing it.  That’s all that this table is made out of, and you can change the table in any way you like by messing with its molecules.  Moreover, any change in the table will mean a change in its molecules.  But it doesn’t follow that we can fully explain and understand everything about what it is for something to be a table in terms of facts about its molecules.  We may think that, if we restricted ourselves just to microphysical talk, we could never capture the concept of a table, even though all tables are wholly composed of microphysical stuff.

Or, to move closer to our topic, consider the fact that we’re in a recession.  We may accept that in some sense that this fact is wholly determined by the microeconomic behavior of individuals around the country and world.  We could eliminate the recession simply by messing enough with that behavior.  But one might accept this without holding that we can understand what it is for there to be a recession in such individualistic terms.  We might think that the existence of a recession involves all kinds of complicated group phenomena that are autonomous in Wiggins' sense, that cannot be cashed in independent terms.

That is precisely what Wiggins wants to say about a natural language.


For Wiggins, the relationship of an individual to her language is not that of an individual to a convention to which she conforms.  Rather it is the relationship of a person to a resource upon which she draws.  The language does not consist in her and others’ conformity to a way of going on.  Rather, the language is there, and she draws upon it in going on.

This is obviously somewhat evocative.  Let’s try to put some meat on it.


II. Davidson’s dismissive view recapitulated

Recall the dismissive view of natural language we ascribed to Davidson:


Dismissive view of natural language: The idea of a natural language, however it is understood (be it in terms of conventions or as a social object) is of little or no help in explaining how communication is possible.


This isn’t to deny that there’s such a thing as a natural language; it’s to hold that talking about natural languages is unhelpful for understanding what makes communication possible.


Davidson’s reasons, as we discussed, are two-fold:

1.      Knowing a speaker’s language does not by itself suffice for understanding an utterance of that speaker.
We spent some time discussing the reason for this.  To know what the words and sentences of a natural language mean is to know how those words are generally used.  And from a fact about something generally tends to happen, it simply does not follow that that is how things will now happen.  Of course, one ordinarily assumes so, but in doing so one is making a substantive assumption.  That one is making a substantive assumption is graphically shown by cases of malapropism, where the assumption will lead one astray.  But even in cases where the speaker use words as they are generally used, where there is no confusion or slip of the tongue, to understand the speaker still requires the assumption that that is what the speaker is in fact doing.  And that is something additional to the knowledge of the language.  That a speaker is now using her words in the way they are generally used is not something that the account of general use itself tells us.
Davidson himself puts this point in terms of the following distinction:

Prior theory: One’s view, prior to a communicative interaction, of what a speaker’s words and sentences mean (generally a matter of one’s knowledge of the natural language of the speaker).

Passing theory: That which must be added to the prior theory if one is to understand a speaker’s utterances.

It’s a bit peculiar of Davidson to call the passing theory a ‘theory’, for often it consists only of the belief that the speaker is using words in accordance with their meaning in the language.  Even in cases where that is not enough—such as cases of slips of the tongue—it’s not really a theory that’s added to the mix, but perhaps only the belief that on this particular occasion, and this occasion alone, the speaker means pinnacle by “pineapple”.
Setting this aside, Davidson’s point is that a passing theory is always necessary.

2.      Two speakers could understand each other without sharing a language.
Now, one might grant that knowing a person’s language does not suffice for understanding.  But one might 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 necessary.
As we discussed, Davidson argues against this as well.  And we saw reason to agree with him on this score as well.


Suppose we accept these two points.  Thus sharing of a natural language (or any other language) is neither necessary nor sufficient for understanding a speaker.  Does the dismissive view follow?

Only if we can’t find any other way in which shared languages aren’t a crucial part of the explanation for the possibility of understanding and communication.  Davidson in effect assumes that there’s no other way in which the idea of a shared language could play this role except if sharing a language were necessary or sufficient for communication.  But we can see Wiggins as denying precisely this, with respect to natural languages in particular.


III. Reasons for resisting the dismissive view

Wiggins never discusses Davidson.  His focus is rather on Chomsky.  Chomsky also holds the dismissive view, but for reasons different than Davidson.  He thinks the idea of a natural language is hopelessly unscientific, not tractable to the resources of natural science, and hence to be dismissed on that ground.


That is a very different argument from Davidson’s, and not one I will discuss.  At the end of the article Wiggins argues against the steps of the argument, and I won’t discuss his objections to it.


But the Chomskian argument and the Davidsonian argument end up at the same place, the dismissive conclusion.  Much of what Wiggins says is in the manner of convincing us of the importance of the idea of a natural language for understanding communication.  And so this part of his thought bears as much on Davidson as on Chomsky.

Wiggins, we saw, also criticizes the conventionalist view of natural language.  We talked about those criticisms last time.  I won’t recapitulate them now.  Wiggins is concerned to defend the social-object view both against the conventionalist alternative and against the dismissive perspective, and I’m going to talk about the latter bit now.


My discussion will be brief, and speculative.  I’m not wholly sure myself how far Wiggins sorts of considerations can be pressed, and I just want to get them on the table in an initial way.  These considerations, unlike what we’ve discussed in the course thus far, stray rather far from mainstream analytic philosophy.


1. Natural languages are the ‘reality to which theories of meaning are answerable’.

Recall our old question of why a theory of meaning has to be compositional.

One answer we discussed is that we constructors of the theory are finite creatures, and we simply can’t list T-sentences for all the sentences available to an ordinary speaker.  For the sentences are without end.

Moreover, if we want to interpret someone, we’ll find we just can’t make any headway unless we view their utterances as built up out of repeating parts.


I mentioned (and it’s on the first handout) that some find both of these answers unsatisfying (though the second less so than the first).  To them it makes the compositionality of the theory seem nothing more than a practical expedient.  We need the compositionality because of limitations in us, in the one case simply our finitude, in the other case because we don’t know how to proceed with interpretation otherwise.

But it seems very appealing to suppose that the compositionality requirement does not flow merely out of a feature of us, the interpreters or theory constructors, but out of the subject matter of our theories—that the compositionality of the theory mirrors the compositionality of the thing theory is about.  It seems appealing to think a non-compositional theory would miss something crucial about the nature of language itself.  So even if we were omniscient beings that could list an endless number of T-sentences, we ought to construct a compositional theory of meaning.


But the question then arises: what is it about the subject matter that motivates the compositionality requirement?  We briefly discussed one possible answer: that the compositionality of a theory of meaning mirrors the structure of the speaker’s own knowledge of her language.  As a matter of human psychological fact, our knowledge of language exists within us in the form of a compositional theory.  (I called this the ‘psychological answer’.)

One difficulty with this answer is to say in what sense we ‘know’ a compositional theory for our own language.  Certainly this knowledge is not explicit; otherwise constructing a compositional theory of meaning would be a snap, rather than enormously difficult, as it in fact is.


Neither Wiggins nor Davidson have any truck with this answer.  But then Wiggins’ question to Davidson is: what is the reality to which a theory of meaning is answerable, such that it must be compositional?  And his suggestion is that we can say nothing other than simply: the language the speaker speaks.  (Quote section 6.)


Now Davidson sometimes throws around the idea of an idiolect: an individual’s unique version of a natural language.  This will differ from the public language for it will contain only some of the words of the language, and will have various idiosyncrasies, for example persistent malapropisms.  For Davidson, one might think, the idiolect is ultimately far more important for understanding how communication is possible than the natural language, for it is precisely what the interpreter, the listener, must divine to understand the speaker.

And so one might hold that it is to idiolects that theories of meaning are answerable.  So there’d be one for each person.


But Wiggins finds the idea of an idiolect fishy.  Do I speak the same idiolect as I did 3 years ago, when I lacked all kinds of words and used slang I don’t now?  Presumably not, especially if what matters is understanding me here and now.  How could how I used to speak 3 years ago matter any more for interpreting what I say now than how others speak?

But then idiolects must be indexed not only to individuals but to moments in time.  We all speak an endless procession of changing idiolects.

And Wiggins wants to say that we have no clear notion of such an entity as a momentary idiolect.  He doesn’t push this far explicitly, but one might say: in virtue of what is it now true that I speak the idiolect that I do?  Presumably it includes all the words in my repertoire now even if I’m not using them.  But what if right now I’m tired and have forgotten some words that tomorrow I’ll remember?  Are they in there or not?  And so forth.

For Wiggins, the only grip we can have on the idea of an idiolect is that of a particular way of speaking a common language.  That’s how I originally defined it, and for Wiggins there’s no way to detach it from that connection.  It is one person’s way of drawing on the common store.  It is no more its own entity then is your unique way of playing baseball.  How do we describe your way of playing baseball?  By giving an account or theory of how to play baseball, and then explaining what you do as a way of doing so.  There is an unavoidable dependence here: can’t give any content to idea of explaining how you play baseball in and of itself, detached from any theory of how to play baseball. 

Wiggins concludes that there’s nothing for a theory of meaning to be about, when our concern is with actual speakers of natural languages, than these natural languages.  It’s true that a theory of meaning for English won’t capture any particular English speaker’s way of speaking precisely: we all veer more or less from the norm in some place or other.  But that doesn’t threaten the reality of the natural language as a subject matter for semantics.  So the answer to the question about compositionality is just that natural languages are compositional.  That’s how things stand with the languages themselves.


2. A speaker intends, not only to get across such-and-such, but to “go on record as saying” such-and-such, and in the basic case this requires a pre-existing language.

This goes way back to our discussions of Grice.  We discussed how the arena of speaker’s meaning can be divided up into two broad areas: what is explicitly said, and what is implied.  Our old actor example.

Wiggins point is that a speaker, someone producing speech, doesn’t just intend to get something across in the sense of have his Gricean intentions be recognized.  He intends to say something, for some part of his speaker meaning to fall into the category of what is explicitly said.

The importance of this is especially clear in the legal context, where what a person said, vs. merely implied can take on enormous importance.  It is what is said, often enough, for example, that matters for contracts and so forth.  One can always say about an implication that it was not really there and one was misinterpreted: “Perhaps what I said could be taken to imply that a job was guaranteed.  But I didn’t say that.  I just said, ‘We’ll see what we can do.’”  But one can’t take back one’s words: “Well I did say that.  But I didn’t mean it.”

Or consider raunchy sitcoms, where all kinds of things the censors allow to be implied but don’t allow to be said.


Wiggins’ thought is that these are merely vivid instances of something that is present in all usages of speech.  Indeed, it is precisely this that distinguishes speech, linguistic use, from other actions that get things across.  With a pointed stare, I can mean a lot.  But I can’t say anything with it.  We do not understand linguistic communication unless we acknowledge the difference between saying, going on record, that p, and on the other hand just getting that p across some way or the other.

It seems quite easy to secure a place for this distinction: we just point out that the sentence uttered by the speaker already has a meaning, independent of this instance of communication.  Griceans like Schiffer are right about at least this.

Of course, we can sometimes make up codes and then use them.  But in the normal case, the sentence is part of a natural language.  It’s what the sentence means in the natural language, which is independent of this particular communicative situation, that allows the speaker to go on record.


It’s much harder to see how the phenomenon of going on record could happen if we refuse to make explanatory use of anything but idiolects (and not languages they are idiolects of).

Davidson would say: we can make sense of the phenomenon in terms of the speaker’s intention to be understood as saying such-and-such.  But it’s quite doubtful this surrogate could fulfill the various roles I am suggesting going on record plays.


3. Natural languages “influence normatively by their presence in the social world the communicative efforts of speakers”.

By talk of “normative influence”, Wiggins means that natural languages impose standards, norms, of correct usage, ways that we must speak if we are to speak correctly.

And of course there are such norms.  And of course we care about them.  We care about our choice of words; we want to use them in accord with what they mean in the language.  We don’t want to use “comprise” as a synonym for “compose” once we find out that they are not. 


Davidson can of course acknowledge this point.  He will say, and has, that this kind of normativity is real but philosophically uninteresting.  It’s just a matter of the pressure toward conformity.  We are subject to comparable norms about all kinds of things, how we dress, how we walk, etc.


But there’s more than a hint in Wiggins of the suggestion that to equate norms governing correct usage of a natural language as having no more significance than this is to miss something crucial. (Section 20.)

The thought I find half-buried here is that we need the normativity imposed by a shared language to make precision in thought possible.  Precision in thought is possible because there is at everyone’s disposal a pre-existing store of words with well-honed meanings, shaped by generations and generations of use, each gaining special import from their similarities and differences from others.

The suggestion here is that we care about the correct use of words not merely because we want to fit in or get approval from authority figures, but because we recognize on some level that a natural language is a instrument for the expression of thought, and that exhibiting a sensitivity toward the workings of that instrument is essential for its effectively serving that function.  Thus one might quail when people start, say, using the word “ironic” to apply to absolutely everything that seems kind of odd or coincidental, because one realizes that a tool for expressing something richer than that is thereby being lost.

Correct usage here, by the way, cannot be understood to mean speaking as the elites do, as upper-class British people do.  Slang, for example, has its special expressive powers, no less than any other area of the lexicon.  Teenagers are enormously sensitive to the correct usage of slang terms, one might think not wholly because they want to identify who to make fun of, but because those words have a life for them that can easily be lost by too much thoughtless employment.


Nor is it implied that usage can’t change, even in a moment.  Innovation is possible; one can start to use a word in a new way that catches on.  But Wiggins would want to say that this is not a matter of starting from scratch, but from building on usages already in place.  Just to take an example I was thinking about the other day (not a very example, though): people once said, “lock and load” only in the context of firing off rifles. Some other people found the expression resonant, and started using it in situations where some endeavor was about to be started, where it came to mean that one should get primed to begin and further connoted the whole enterprise had a whiff of ridiculousness about it.  And so on.

For innovations like this to occur we need a natural language already on the scene.  Innovations in use build on (in the case of this example, even comment on) previous usage.


The following is not a suggestion Wiggins actually makes but is in the spirit of his recommendations.  It is more explicitly found in some writings of McDowell, who in turn finds it in the German philosopher Gadamer.


4. Natural languages are “repositories of tradition”.

A phrase of McDowell’s.  What’s it mean?


We may understand the thought as follows.  As a child grows up it hears sentences that it is not yet in a position to understand.  It repeats them, attempting to do so in the right situations, but not yet sure what such situations are or what their significance could be.  In time the child gets a better and better grasp on what it itself is saying when it repeats those sentences or hazards variants of them.

What has happened here is that simultaneously with learning how to speak, a child is learning what to say in given situations.  That is, the child is learning facts about the world at the same time it is learning what words and sentences mean.  The child is learning what the sentence, “Hitting is wrong,” say, means, but simultaneously it’s learning that hitting is wrong.  It’s learning this because it accepts, is not in a position not to accept, that the sentence is true because the parent asserts it.  So it is learning at once what is true (or at least what its elders believe) and what words mean.

But notice that for this to work the child must at first imitate how its parents speak.  It must learn to speak as others speak.  It’s crucial that it simply does what its parent does, at least at first.  The child learns by imitating the parents’ vocalizations.  And unless we assume that the child’s vocalization will over time come to have the meaning that the parent’s does, we can’t make sense of the child’s coming to learn the facts, acquire the beliefs, that the parent wishes it to have and that it must have to get on in the world.

Upon rote imitation is built both understanding of language and acquisition of world-view.


Davidson wants to claim that possession of a shared language simply plays no role in making communication possible.  But one may accept that adults can get by without sharing languages without thinking that one could acquire a first language that was wholly idiosyncratic for one.

We have here a peculiar sort of analogy to the Davidsonian interpretive situation.  We must solve for meaning on the part of the child—assume that its words are moving toward meaning what its parents’ words mean, if we are to take it to come to acquire the beliefs essential for growth into a competent, thinking human being.