Jason Bridges

Fall 2002

Phil 33401, A02—Lecture 3 notes

Is Semantics Possible?




I. Putnam’s criticism of traditional semantics

I will begin with a discussion of Putnam’s view of the semantics of natural-kind terms, and then talk about how it bears on externalism, the latter not explicitly addressed in this paper.

Putnam says a lot about the semantics of natural-kind terms—I’ll only discuss the important points for our purposes.


First of all, what is semantics?  What is semantic theory in the sense relevant to Putnam’s discussion?

Well, semantics has to do with the meanings of linguistic expressions.  It’s to be contrasted with syntax, which is concerned not with the meanings of words but with grammar—that is with word order and more generally with the ways in which complex expressions can be built up out of simpler ones.

So a theory of semantics might equally be called a theory of meaning.  But here we have to draw a distinction.

Talk of a theory of meaning or semantics might mean one of two things in the philosophical context:

1.      It might refer to attempts to give an analysis of the notion of linguistic meaning.  That is, it might refer to an attempt to say what linguistic meaning as such is, to say what a word’s having a meaning consists in.  What we discussed last time amounted to a partial essay of such an account.  We distinguished between various concepts of linguistic meaning, showed that there isn’t just one, and that moreover the various types are interrelated.
This sort of work is sometimes (probably mistakenly) thought of as conceptual analysis.  It’s an attempt, as one might put it, to explain what is involved in concepts such as linguistic meaning that do not seem wholly clear or seem somehow problematic on their face.  The kind of pursuit philosophers have long took themselves to be primarily concerned with.

2.      But a theory of meaning or semantic theory might also be an attempt to actually state the meanings of specific words in a specific language.  A theory of meaning for English in this sense is an account of what all the expressions of English mean.  The difference between a ‘conceptual analysis’ of linguistic meaning and a theory of meaning for a particular language is precisely analogous to the difference between a ‘conceptual analysis’ of the notion of causation and a theory of the particular causal relations that hold among certain actual objects—say, among the members of Bush’s cabinet.


The second is the kind of semantic theory that is Putnam’s topic.  But one may wonder how anyone could think that theories of meaning of particular languages are the business of philosophers, as opposed to empirical researchers.  Surely it’s not the business of philosophers to investigate the particular causal relations among the members of Bush’s cabinet.  That dreary lot falls to the historian or political scientist.  Why then should a philosopher be concerned with what specific words like “gold” or “lemon” mean?


Part of the answer here is that Putnam is not actually interested in particular words as such, only insofar as they are representatives of more general types of words.  In particular he’s interested in natural kind terms—what they are we’ll soon get to.

The other part of the answer is that attempting to give an account of the meanings of these kinds of words, and in particular seeing how badly traditional accounts of this sort fail, will turn out to have important implications for our understanding of meaning as such.  Popular accounts of the meanings of particular terms, say those on the part of linguists contemporary to the time of this article, turn out not to flow from a sensitive, open-minded investigation of the phenomena, but from assumptions about what the meaning of a word must be like.  Such general assumptions are the province of philosophy.  What Putnam would ultimately come to think, although we shall only brush on this today, is that certain philosophical conceptions of the nature of the mind lay in the background of these bogus accounts of the meanings of particular words.


Second preliminary: a natural kind word is first of all a common noun.  You may remember this term from grammar.  A common noun is opposed to a proper name.  A proper name is a word for a particular thing or individual—a common noun is a word for a type of thing, individual or stuff—human being, pencil, water, explosion, etc.

Common nouns have extensions (discussed last time).  The extension of a term is composed of the things the term is true of.  So the extension of “pencil” is all the world’s pencils


Natural kind terms are more specifically words for naturally occurring types of things.  What that might mean we shall very shortly discuss in more depth.  But at the outset we can see an intuitive contrast between a naturally occurring kind or type, such as tiger, and what’s called an artifactual kind, like watch or pencil. 

Artifacts are things we make.  Members of natural kinds are things there in the world independently of our machination, and moreover kinds that seem naturally, as part of the structure of the natural world, differentiated from other kinds.


Now, third preliminary, the idea of analytic truth.

The standard idea of an analytic truth is a statement that is true just in virtue of the meanings of its words.  The standard example is the statement that all bachelors are unmarried men.  This is true, one wants to say, simply in virtue of the meaning of the word “bachelor”.

What exactly it means to say, though, that something is true in virtue of meaning is a very difficult question.  There is no need for us to get into this issue, for Putnam’s aim is to argue that there are far less analytic truths than people think, whatever an analytic truth is supposed to be.

Perhaps we can bypass these issue by just relying on the following apparent mark of the analytic:


A truth p is analytic iff to know that p is true, the only knowledge you need is knowledge about the meanings of the expressions in p (plus deductive logic).


“All bachelors are unmarried” does seem (although it may not in fact) to meet this criterion.


Analytic truths are necessary truths.  They’re truths that couldn’t possibly have been false.  But not all necessary truths are analytic truths.  It’s a necessary truth that the square root of 2809 is 53, but, at least in terms of our provisional criterion, it appears not to be analytic.  (Well, actually, it’s rather hard to say.  Shows the difficulties one gets into in taking seriously the idea of the analytic.)

But the connection does run the other way.  So necessity is a criterion for analyticity.

Now, a statement that serves to define a term, to give its meaning, one might think will be analytic.  For it will be precisely what someone who knows the meaning of the term thereby knows.


But what kind of analytic truth can be used to give a definition?  One natural proposal is this:

A definition as an analytic statement of necessary and sufficient conditions:

To give the meaning of a common noun x is to state an analytic truth of the form:

Something is an x iff it is _____.

Equivalently, all and only x’s are ____

Where what goes in the blank doesn’t use the word x itself (otherwise the truth would be trivial)


We need the “iff” because it seems natural to think that we don’t give the meaning of a common noun until we’ve stated both analytically necessary and analytically sufficient conditions for something’s falling under that word.

The point is really a simple one.  “Something is bachelor if it’s not married” doesn’t fully define “bachelor”, it only gives one criterion for being a bachelor.  Some women are unmarried—they’re not bachelors.  So it’s not, as we say, a sufficient condition on being a bachelor that you’re unmarried.  You also have to be male.  And an adult.  So perhaps we say that something is a bachelor iff it’s an adult unmarried male.  (We can ignore the fact that this isn’t true.)

Or take the other well-known example, Something is a vixen iff it’s a female fox.  What’s on the right seems necessary and sufficient for what’s on the left.

So the idea is that knowing the meaning of a common noun is knowing analytic necessary and sufficient conditions on belonging to its extension.


Now, Putnam discusses a semantic theory owing to the linguist Katz that involves lots of fancy terminology that we won’t get into.  What matters is what this view implies—namely, that for all common nouns we can state their meaning through analytic statements of necessary and sufficient conditions.  This claim Putnam thinks proves on reflection to be absurd.


Let’s try to do one for “lemon”:

Something is a lemon iff it is yellow, tart….


Does this work?  Forget about whether this is analytically true, or necessarily true.  Is it even true?

No. There are green lemons.


The same point, Putnam wants to claim, holds for all natural kind terms.  For no natural kinds is it the case that the observable properties we ordinarily associate with such kinds must be had by all members of the kind. 

Features of natural kinds:

1. Members of a natural kind can lack the observable features had by normal members of the kind.


These properties we associate with the term apply rather only to normal members of the kind.


This might lead to the following proposal:

Something is a lemon iff it belongs to a natural kind whose normal members are….


This implies that we can’t appeal to these properties in attempting to state necessary conditions for membership.


(What “normal” means here is problematic—a possible reading: “statistically most numerous”.)

Putnam finds this an improvement.  It acknowledges that being yellow and tart and so forth are not conditions that anything must satisfy if it is to be a lemon.  It is already in this respect to move beyond the kind of semantic theory found in Katz.

But he still rejects it.  Why?  Well, some of the points he makes here are confused in ways I won’t get into now.


But the main point is that he doubts that it’s a necessary truth that normal lemons are….  It’s possible that one day, given some change in the environment, normal lemons will be green.  In such a scenario, we wouldn’t stop regarding them as lemons.

Not only is the relationship between being a lemon and being yellow not necessary, the relationship between being a normal lemon and being yellow is not necessary.


II. Putnam’s positive proposal

How then do we state the meaning of “lemon”?


Putnam’s suggestion is this.  There is first of all associated with every natural kind term a stereotype:

The stereotype of a natural-kind term ‘x’ = the set of observable characteristics a competent speaker of the language associates with normal x’s.

So the stereotype of a lemon is a yellow, tart, citrus fruit.


Putnam has just argued that we can’t say that lemon is, as a matter of analytic definition, something that belongs to a kind whose normal members are….


So what’s the relationship between a term’s stereotype and its meaning?  It’s more subtle, and can’t be understood until we add a further component.  This component, says Putnam, is the term’s extension.

The meaning of a term is determined not only by stereotype, but by extension.


But what does this mean?  It’s a very mysterious remark.  For we’re naturally inclined to think that the meaning of a term determines its extension, not the other way around.  This thing is in the extension of “mug” because mug means what it does.  What could it mean to say that “mug” means what it does because things like this are in its extension?


Well, in the case of mug, it couldn’t mean anything, for a mug is not a natural kind.  But here’s an elaboration of the idea for natural kinds, which Putnam himself moves over very quickly.

Natural kind terms are used to pick out actual kinds of substance occurring in the natural world.  We think of the natural world as divided up into different kinds of thing naturally, independently of us.  We aim with our words to pick out such natural ‘joints; among things.  If we want to interact successfully with the natural world, we had best acknowledge the way it actually is organized.  To put the point in a silly way, if we had a word that was so defined that its extension was lemons and a kind of poisonous berry, we’d have a much harder time conveying to each other what’s safe to eat and what not.  We want our words to distinguish between lemons and poisonous berries because the world distinguishes between them,

One aspect of the idea of a natural as opposed to artificial difference between kinds of thing is that natural differences are explained by underlying structural facts about the items in question, facts that we may not have a hand on but that are there in the nature of the thing waiting to be discovered.

In the case of differences between species of living things, the underlying differences are (roughly) in DNA.  In the case of a natural stuff like water or air, the difference is in chemical composition.  In the case of an element like gold or iron, the difference is in atomic number.

Different kinds of natural kinds have different kinds of underlying explanatory principles.


How do we know these facts?  Only through scientific investigation.  They are not there observable on the surfaces of things—they are the underlying facts that explain the observable facts.  To discover those facts is the job of the natural sciences.  If we’re talking about stuffs like water, it’s the job of chemistry.  If we’re talking about living things, it’s the job of biology.

A natural kind, as the idea is sometimes put, has a scientifically discoverable essence.

2. All members of a natural kind share a “scientifically discoverable essence”.


But now the question arises—if we want the extensions of at least some of our words to correspond to where the world really is broken up at the joints, but if what breaks the world up at the joints is underlying facts that can be discovered only through scientific investigation, how can we do it?  There was a time when these sciences were not well-advanced, when no one knew the chemical structure of water, when no one knew what distinguished one species of animal from another.  How did they have words for natural kinds, for things with naturally discoverable essences?


Well, here’s where stereotypes come in.  These underlying essences correspond, not perfectly, not necessarily, but more or less, in the normal case, to observable features of things.  Ordinarily, things with the genetic structure of a lemon are yellow and tart.  Ordinarily, tigers have four legs, stripes, live in jungle.


So we notice that various objects in nature have various properties clustered together—some are yellow and tart, some are red and mealy, etc., and we say, let’s have words for the natural kinds to which things with these cluster of properties can be presumed to belong.

Since the underlying property that determines the natural kind and the observable features only correspond imperfectly, the stereotype does not determine the bounds of the extension.  What determines the bounds of the extension is the nature of the stuff itself.


The idea, then, is that a stereotype hones in on a natural kind term, but that what determines the extension of the term, once a given natural kind has been honed in on, is the world itself.

This is why stereotypes can’t be boot-strapped into analytic necessary and sufficient conditions.  That’s not their role.

We can get necessary and sufficient conditions only for a kind that doesn’t have a natural essence, such as the kind bachelor.


In a nutshell, Putnam’s proposal (not using his lingo):

The meaning of a natural kind term is determined by a stereotype and a natural essence.  The stereotype picks out standard members of a natural kind, and the natural essence determines the precise boundaries of the kind.


How then do we state the meaning of a natural kind?  Well you state the stereotype.  But a competent speaker will know, at least implicitly that these don’t give necessary and sufficient conditions for membership in the kind, even for normal membership.  Rather the stereotype more or less imperfectly picks out a kind, and what the word refers to is determined by the world itself.

As it’s sometimes put, the meaning of a natural kind term is world-involving.


The closest we could come to an analytic definition of “lemon” would be this (and it’s not in fact analytic:

Something is a lemon iff it shares the underlying natural structure (essence) of these things [said while pointing at stereotypical lemons].

More simply, iff it belongs to the natural kind typified by these things.


This is an ostensive definition (explain) and the role of the ostension, the determination embodies the respect in which the meaning of the word is determined by the worldly thing itself.

One can understand such a definition without oneself being a scientist who knows what the underlying structures are.  You just have to know that is has such a structure—or perhaps not event that, just that it is a natural thing.


So the meaning of “lemon” is not given by stating this structure.  “Lemon” doesn’t mean—thing with DNA composed of…..  “Lemon” means thing belonging to the same natural kind as this.

So even if there were no scientists, we could still use words in this way.  We just need the idea of a kind whose boundaries is fixed by some underlying nature that doesn’t precisely correspond to any immediately observable features.


So what about externalism?  Well, meaning something is a state of mind.  But on this view what one means is partly determined by the way the world is, something that is not in your head.

This particular implication is the focus of the next article we will read.