Chapter 8: Meaning

From: Harnad, Stevan (
Date: Mon Apr 28 1997 - 12:34:09 BST

        Chapter 8: Meaning and Conversation.

Sentence Meaning vs. Speaker Meaning.

There is an important difference between the meaning of a sentence and
the intended meaning of the speaker of the sentence:

For example, "Would you mind getting off my foot" is an inquiry about
whether the hearer would mind doing something, but what the speaker
means is "do that" ("get off my foot").


The use of language cannot be understood by considering only sentence
meaning. Sentences are spoken in a CONTEXT. Context means two things:
"con-text" literally means the text that comes before and after the
sentence in question:

"He dropped it" has only a vague sentence meaning: Some male dropped
something. But if preceded by "President Clinton came into the room
with a huge saxophone. He dropped it" we understand more, because the
context fixes who "he" is and what "it" is.

More context would tell more (when was this? where did it happen?)

The second meaning of "context" comes from where and when a sentence is
spoken. If you are looking for your precious alto saxophone, and you
find it on the floor broken into pieces, and I point my finger at
President Clinton and say "He dropped it" this also fixes the meaning
in a way the sentence "he dropped it" cannot do on its own.

Meaning and Reference.

What is the meaning of a word? We are tempted to say it is whatever it is
that the word stands for. So maybe "he" and "it" are problem words, which
require context in order to know what they mean, but "President Clinton"
just means President Clinton, that person we all have in mind now.

But if that is the meaning of "President Clinton", what about the
meaning of "the saxophone-playing president"? Both symbol-strings
REFER to the same person, but do they MEAN the same thing? Are they

No they are not, and a lot of contemporary philosophy of language has
focused on the difference between REFERENCE -- what a symbol string
refers to -- and SENSE (or "MEANING"), which seems to be something else.

What we were doing with the help of context was fixing the REFERENCE of
"he" and "it," but their MEANING is something else.

Some philosophers and psychologists hold that the meaning of a word,
phrase or sentence is what "picks out" its referent: "President
Clinton" like "the saxophone-playing occupant of the Whitehouse" are
two different ways of picking out one and the same person that they
both refer to.

But what is "meaning" then, and how does it pick out referents?


Your textbook suggests that the meaning of a word is the "concept"
associated with it. But what is a "concept" then?

This is getting into a controversial area, so neither these notes nor the
book should be taken to be the best and last word on the subject,
but here is one candidate for what the concept behind a word is:

The "Classical" Theory of Concepts.

A concept is based on a set of features. An "apple" is a round, red
fruit that grows on trees. All apples have to have those features.
Each of these features is a "necessary condition" for being an
apple. Having all these features is enough (a "sufficient condition") to
pick one and only one kind of thing, and that thing is the referent of
the word "apple."

The problem with this view of the concept of apple is that it is neither
necessary nor sufficient to pick out apples: An apple can be green
instead of red; and there are other things that are red, round and grow
on trees (plums of the right shade), and they are not apples.

So the features I mentioned -- and more I could mention --
are neither necessary nor sufficient to pick out an apple.

A few concepts really do have necessary and sufficient conditions:

a "bachelor" is an "unmarried man": Being an unmarried man is
both necessary and sufficient to be a bachelor, because that is how we
have DEFINED "bachelor." With "apple" you could go out and discover a
fruit that fit the description but was not an apple, and vice versa.
That is because the description does not really DEFINE what an apple is.
It is more like a prediction: look for such and such features and you
will probably find an apple.

Now the full discussion of reference, meaning and concepts would go FAR
beyond this introductory chapter. Just note that it is not yet clear
what a concept is, and that explaining concepts is a major goal of
explaining the mind.

Among the candidates for what a concept is, there are others besides
necessary/sufficient conditions. Let's call this first theory of concepts
the "classical" theory. (I happen to still like the classical theory the


But it looks as if the classical theory of concepts runs into trouble
the way it does with "apple": we can't seem to pin down a set of
necessary and sufficient conditions for BEING an apple and hence for
being NAMED by the word "apple," although in most cases we have no
problem deciding what is and is not an apple, and no trouble using the
word "apple" correctly.

A rival to the "classical" theory of meaning is the "prototype" (or
"template") theory, according to which the way that concepts pick out
their referents is through similarity: We have a picture in our minds
of an ideal apple, and then we compare any particular apple or
apple-like thing to that ideal "prototype" of an apple, and we call the
thing an apple if it is closer to that prototype than it is to, say,
the ideal "peach" or "plum."

The trouble with this prototype theory of concepts is that it implies that
everything is an apple to some degree, and the real apples are simply
the most "appley" things. But that's not true. A real apple is no kind of
plum, not even a little bit.

Mental Models.

Our textbook, inspired by the work of a very important British cognitive
psychologist, Phil Johnson-Laird, puts especial emphasis on the
"mental-model" theory of concepts and meaning. See:

So far, we have used "model" to refer to what theorists construct
(with symbol systems, neural nets, or analog transforms) in order to
explain human behavioural capacity. Johnson-Laird's notion of
"mental-model" is a more general kind of model, as we will see in a
moment, but if you are clever enough you will see that it covers the
cognitive theorist's sense of model too.

Imagine a field, with North ahead, West to the left and East to the
right. Imagine two apples lying on the ground beside one another, the
red one on the east side and the green one on the west side.
Now imagine someone on a cart coming toward the apples from the North.
On what side would the red apple be for him: the right or the left?

There are many ways this question could be answered, but most of us
build a little model in our heads with the locations of the apples
laid out; then we add someone coming from the north; then we
"read off" the position of the green apple that fits with all the
information provided. That is a mental model -- a spatial one.

There are logical mental models too: If I say President Clinton is
smarter than John Major but not as smart as Paddy Ashdown, whereas Tony
Blair is smarter than Paddy Ashdown, who is the 3rd smartest of them,
you can read it off a mental model you have built that orders them in
some way.

These mental models are not just used for puzzles like this: They are
used in conversations, so we understand one another. If we are to
understand one another, we must have the same model in mind, or at least
similar enough ones so that what we say to one another makes sense.


I have used the food dispenser example before to explain what
information and communication are about: If there are six numbers and I
can only choose one number each day, then my chance of getting lunch is
1/6, but if someone tells me that the number is odd today (and that is
true) then my probability of lunch has gone up to 1/3.

When human communication matters most (when your lunch or your life
depends on getting the right information), it's important not only that
you understand the SENTENCE meaning of one another's messages, but also
the SPEAKER meaning: what each of you had in mind.

The chapter describes an experiment in which there are two people
on opposite sides of an opaque window. Each have twelve shapes in front
of them, but one of them has the shapes in the correct order and the
other has them scrambled. Let's say that BOTH their lunches depend on
getting the scrambled ones into the right order, and, obviously, it all
has to be done with words.

The one with the right order has to describe what each shape looks like
and where it must be put. They manage by gradually arriving at a
description that picks out each of the 12 shapes uniquely (based
on similarities) and then using that to describe the correct order.

This situation captures many of the features that must have played a role
in the origins of language. Similar but much simpler experiments
have been done with pigeons (see Lubinski & Thompson:
and with Chimpanzees
(Savage-Rumbaugh, E. S. (1986). Ape language:
>From conditioned response to symbol. New York:
Columbia University Press.) See also:

What happens in the human experiments is that the players gradually
converge on a simple, shared set of descriptors. You may have
wondered why some things have proper names ("Bill") or
kind-names ("apples") whereas others only have descriptions
("things that are bigger than a breadbox"). Some of the cooperative
verbal games described in Chapter 8 should convey the flavour of what
verbal communication is about, and for.


The philosopher HP Grice has suggested some "maxims" that describe what
is necessary for speakers to understand so as to inform one another.
Sentence meaning is certainly not enough. Not even sentence meaning
in context (in both senses of the word: con-text and where the text is
spoken) is enough. Grice's "maxims" are QUANTITY (say enough, but
not more than necessary), QUALITY (tell the truth), RELATION (say only
what is relevant) and MANNER (say it clearly).

If one or more of these "maxims" is violated ("Bill Clinton is a
doughnut") then the hearer has to assume there was a reason for the
violation (in this case, the sentence is not meant to be taken literally,
it is just an unflattering likening of Bill to a doughnut).

Understanding the uses of non-literal language (as in metaphors, or
irony) is also a task of pragmatics, and whole courses could be based on

1st and 2nd Order Representations

Children and animals have been tested on "false belief" tasks:
A child watches a puppet put an object in a box and then
the puppet leaves and another puppet puts it into a different
box. The first puppet returns and the child is asked where the
puppet will think the object is. The correct answer would
be the old location, but at 3 1/2 children say it's the new location.
So they have not yet understood that there are things someone
could or could not know.

When they get old enough to correctly infer what the puppet would
know, this is called a "1st order representation" of what is going on in
someone else's mind. [Never mind that puppets are just for pretending
anyway.] A "2nd-order representation" of what is going on in someone
else's mind occurs when the child can not only infer what the puppet
would think, but what the puppet would think ANOTHER puppet would

According to some theories, we need at least 1st order representations of
what is going on in someone else's mind to understand metaphor, and 2nd
order representations to understand irony (as in "And these theories
are the last word on this subject!").


Sperber & Wilson have written a book on relevance. They have a simpler
theory than Grice's: All a hearer has to assume is that the speaker meant
to say the most relevant thing, and the most relevant thing is the one
that has the most implications for the least effort. Read the part
of the chapter that explains this in more detail and then let me know
whether you think that's the last word on relevance, and if not, why not?

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:52 GMT