You are on page 1of 9

The Fregean Background

(i) Frege’s Analysis of Language

At its simplest, our task is to analyse sentences such as ‘Fred is red’; ‘Gertrude is
loud’; ‘Everest is high’. How do we explain the fact that these sentences mean what
they do on the basis of what the component words mean?
Frege, a mathematician interested in the foundations of mathematics, comes at
this problem from a distinctly mathematical perspective. So let’s set aside for the
moment the question about language and take a simple mathematical function, say,
the function of adding 2. Here’s a table of the results of applying this function to a
series of numbers:

x x+2
1 3
2 4
3 5
5 7

OK this is all easy enough. We obviously understand the function and know what the
result of applying the function to any number is. Frege’s question concerns the nature
of the function and how we are to represent it. It is tempting to represent the function,
as we did in the table, by saying it is the function x+2. This isn’t entirely wrong but is
apt to be misunderstood. If ‘x’ simply designates an indeterminate/imprecise number
then so to does ‘x+2’; it doesn’t represent the function itself. In order to get at the
function we have to see what is common between the numbers in the right hand
column. That is we want to express the pattern we can see in:

x x+2
1 1+2
2 2+2
3 3+2
5 5+2

That is we want to see the numbers on the right as arising from adding two to the
numbers on the left. The function is something which takes each of the numbers on
the right to the appropriate number on the left. So it is something which gives the
numbers on the right when we replace the ‘x’ in ‘x+2’ by the numbers on the left. So
the expression for the function should be thought of as something essentially
incomplete since it yields expressions for certain numbers when other numbers are
placed in a specified position. So the expression for the function should be something
which makes this incompleteness explicit, viz.,

‘( )+2’

Importantly, Frege thinks that since we must represent the function to ourselves in this
way we must think of the function itself as being essentially incomplete.
So, to sum this up, a function is an incomplete entity which takes numbers (certain
objects) to other numbers (other objects). The expression for a function should make
its incompleteness explicit; we need, that is, to indicate the function’s argument place
or places.

Examples of functions:

( ).3+5; ( ).( ); ( )2-67; ( ); ( )+ { }

(Here the different styles of brackets indicate that these gaps may be completed by
using different numbers.)
In general whenever we have a (complex) expression for a number we can extract
some of the symbols for numbers and arrive at a function.

E.g., we may have ‘5.3+52’ from which we might extract the symbol ‘5’ to arrive at
the following expression for a function: ‘( ).3+( )2’. Or we might only extract one
symbol ‘5’ to arrive at: ‘5.3+( )2’. We might also extract the symbol ‘5’ and ‘3’ to
arrive at: ‘( ).{ }+ ( )2’.

As long as remember how to interpret the ‘x’, i.e., as indicating a gap, not as
indicating an indeterminate number, then we can replace the brackets by ‘x’s and ‘y’s.

Note that the function 2.(x+1)-2.x, whose value is always 2 no matter what we
substitute for x must still be sharply distinguished from the number 2 itself: the one is
a function the other a number, that is, a kind of object.

Consider this somewhat different case. Take the arithmetical expression ‘2+3=5’, if
we extract the symbol ‘3’ then what we should have left according to Frege is an
expression for a function. And obviously we can substitute names for other numbers
in the place occupied by ‘3’ to get:

x 2+x=5
1 2+1=5
2 2+2=5
3 2+3=5
5 2+5=5

But the expressions on the right are complete, so, according to Frege, they should be
names. But what do they name? Clearly they differ from expressions such as ‘2+3’ in
that they can be true or false (all but the third is false). But Frege treats them like
these other expressions: he simply says that they name Truth and Falsity. So the
function 2+x=5 is a function which takes numbers to truth-values. Any function
whose values are only the two truth-values is said by Frege to be a concept.

Natural Language
Let’s try to apply this framework to natural language, as Frege did. Take our sentence
‘Fred is red’ if we take the name ‘Fred’ from this sentence we’re left with ‘x is red’.
So this according to Frege is an expression for a function, specifically, a concept: it is
a concept which takes objects, including Fred to truth-values.

Are there functions which aren’t concepts in natural language? Indeed there are.
Consider this way of talking of South Africa: ‘The homeland of Nelson Mandela’.
Here we have an expression for a function, namely, ‘The homeland of x’ which takes
people (among other things) to countries. The example can obviously be multiplied.

Let’s think now about logical vocabulary. Take, for example, the conjunction ‘The
weather is foul and I am tired of getting soaked’. We can extract the two sentences to
get an expression for a concept: ‘x and y’. The role of this concept is to take truth-
values to truth-values. We might sum this up as follows:

x y x and y
True True True
True False False
False True False
False False False

Here we can ignore which particular sentences we replace x and y by since all those
sentences simply function as names for one or other of the two truth values. So we
can capture the meaning of ‘x and y’ completely by the above table which exhausts
the possible combinations of truth-values. Conjunction is thus an example of a truth
function and we arrive at what’s now commonly called the truth functional definition
of the sentential connectives (‘x and y’; ‘x or y’; ‘not x’; and ‘if x then y’).

Finally let’s consider the sentence ‘All men are mortal’. Our aim is to work out what
sort of expression ‘all’ is. Let’s start though with ‘men’ and ‘mortal’ These have the
same meaning here as in ‘Socrates is a man’ and ‘Socrates is mortal’ respectively. (If
they didn’t have the same meaning then the syllogism—All men are mortal. Socrates
is a man. Therefore, Socrates is mortal.—would commit the fallacy of equivocation,
which it clearly does not.) That is we have two functions ‘x is a man’ and ‘x is
mortal’. So the form of the sentence is better represented as ‘For all x, if x is a man
then x is mortal’. And here we apply the function ‘For all x’ to the complex function
‘if x is a man then x is mortal’; that is, we have a function of functions.

So, ‘For all x, Fx’ says of the concept F that it is satisfied by every object. ‘There is
an x such that Fx’ says of the concept F that it has at least one instance.

What we’ve been given is an account of how the semantic value of a complex
expression is determined by the semantic values of its parts. The semantic value of an
expression is an item in the world with which it is correlated, which it refers to or
denotes. So the semantic value of a name is an object, the semantic value of a
functional expression is a function and the semantic value of a sentence is a truth-
value.
We need to say a little more about functions. A function, as we’ve noted is something
which takes objects to other objects. So we can think of a function such as ‘x+3’ as
pairing each number with another. The list of pairs so determined is called the value
range or extension of the function. So for ‘x+3’ we’d have: (1,4); (2,5); (3,6) etc.
Frege treats functions extensionally, which is to say that he identifies functions which
have the same extension. This is because he believes that the semantic value of a
compound expression is completely determined by the semantic values of its
components. So, provided two expressions have the same extension, interchange of
one with the other won’t affect the semantic value of the whole.

And now we can see how we’d give an account of how the semantic-value of
complex expression is determined. That is, if we had specifications of what the names
in our language refer to and of the extensions of the functional expressions then we’d
be able to work out the truth-values of elementary sentences.

Example of a Primitive Semantic Theory

Axioms

(i) The denotation of ‘Freda’ is Freda


(ii) The denotation of ‘Mary’ is Mary
(iii) The denotation of ‘Lucy’ is Lucy

(I) The extension of ‘The mother of x’ is (Freda, Mary); (Mary, Lucy)


(II) The extension of ‘x is a lawyer’ is (Freda, False); (Mary, True); (Lucy, True)

(C) A name formed from a name and a function denotes the second object in the
ordered pair whose first member is the object denoted by the name, in the extension of
the function.

Examples:
(a) ‘The mother of Mary’
1. ‘The mother of Mary’ denotes the second object in the ordered pair whose first
member is the denotation of ‘Mary’, in the extension of ‘The mother of x’ (C)
2. ‘The mother of Mary’ denotes the second object in the ordered pair whose first
member is Mary, in the extension of ‘The mother of x’ (ii)
3. ‘The mother of Mary’ denotes Lucy (I)

We could use the same method to work out the denotation of, for example, ‘The
mother of the mother of Freda’

(b) ‘The mother of Freda is a lawyer’


1. ‘The mother of Freda is a lawyer’ denotes the second object in the ordered pair
whose first member is the object denoted by ‘The mother of Freda’ in the
extension of ‘x is a lawyer’ (C)
2. ‘The mother of Freda is a lawyer’ denotes the second object in the ordered pair
whose first member is (the second object in the ordered pair whose first object
is the denotation of ‘Freda’ in the extension of ‘The mother of Freda’) in the
extension of ‘x is a lawyer’ (C)
3. ‘The mother of Freda is a lawyer’ denotes the second object in the ordered pair
whose first member is (the second object in the ordered pair whose first object
is Freda in the extension of ‘The mother of Freda’) in the extension of ‘x is a
lawyer’ (i)
4. ‘The mother of Freda is a lawyer’ denotes the second object in the ordered pair
whose first member is Mary, in the extension of ‘x is a lawyer’ (I)
5. ‘The mother of Freda is a lawyer’ denotes the True. (II)

The Referential View of the Meanings of Names

‘Kilimanjaro’ means the highest mountain in Africa

is grammatical, and, just possibly, an ordinary sentence of English. In this case the
analysis would be something like the following:

‘Kilimanjaro’ means the highest mountain in Africa iff ‘Kilimanjaro’ refers to


the highest mountain in Africa.1

What we shall learn in this section is that this analysis faces severe problems. These
problems are sufficient to cast doubt on the original sentences as being an acceptable
starting point. These lessons are due to Frege. We shall then go on to look at Russell’s
alternative account which serves to show that the logico-syntactic category of singular
terms is highly moot. Although we shall be looking at these questions from the point
of view of giving an analysis of meaning, the discussion takes us to the heart of
philosophical attempts to understand names and definite descriptions. This is scarcely
surprising since the general point that we are pushing is that fixing on the starting
point of analysis already requires considerable philosophical insight into how
language works.

Frege’s Argument for the Notion of Sense

Frege’s argument2 focuses on identity statements. But this is primarily a heuristic


device: he could have made the same point by focusing on elementary sentences in
general. However it’s worth following through his reasoning about identity statements
simply because, in doing so, we restrict our attention to the way names function.
We often assert statements of identity. For instance: ‘The morning star is the
evening star’; ‘Everest is Gaurisanker’; ‘Clark Kent is Superman’; ‘Portia is
Balthazar’. The crucial thing to note is that these statements often carry useful
information. If you remember the plot of The Merchant of Venice you’ll know that
Bassanio would not have got himself in a pickle with Portia had he known that Portia
and Balthazar are identical. Similarly Lois Lane would have acted quite differently
had she known that Clark Kent is Superman: that piece of knowledge would have
been informative to her. So statements of identity can be informative. No account of
language and, in particular, of the functioning of names can be acceptable unless it
makes sense of this fact.

1
Note: we should not confuse these two claims. (i) The meaning of e is m iff e refers to m; and (ii) the
meaning of e is given by the fact that e refers to m. (i) is the claim in the text; (ii) is a claim of quite a
different sort that we shall only come to discuss in chapter six.
2
See inter alia Frege (1980: 56-78)
Frege asks himself the question: what does the relation of identity hold
between? It would seem that on the analysis of names with which we have so far been
provided we can give only two answers. The relation either holds between the objects
named by the names or it holds between the names themselves. Let’s consider each
suggestion in turn. Suppose that ‘a’ and ‘b’ are names and that ‘a=b’ is true. If the
relation holds between the objects denoted by the names then ‘a=b’ cannot differ from
‘a=a’. For, since ‘a=b’ is true, ‘a’ and ‘b’ name the same object. But clearly ‘a=b’
and ‘a=a’ differ because on many occasions (such as those above) statements of the
form ‘a=b’ are informative whereas those of the form ‘a=a’ are uninformative since
they’re known a priori to be true.
Should we then take the identity statement to assert a relation between names?
Frege says not since this can never be informative. Why so? Well what a sign means
is a matter of speakers conferring meaning on the sign. So, if identity asserts a relation
between names, then either ‘a=b’ is false since ‘a’ and ‘b’ are different names or it
registers a stipulation (which we’re entirely entitled to make) that the signs ‘a’ and ‘b’
are to be treated as the same name. On neither reading can ‘a=b’ be treated as an
informative statement.
The point seems to be this. The character of a particular language is arbitrary
in the sense that there’s no intrinsic connection between the signs we use and what
those signs refer to: whatever connections actually exist can be imagined to be quite
other without supposing any difference in extralinguistic reality. But we can’t simply
suppose that the link of a name to its reference is purely arbitrary: that there is no
saying how the use or the meaning of the name establishes the link. Why not?
Because then there couldn’t be informative identity statements, these would just be
the laying down of arbitrary conventions. What we need to be able to see is that there
is both an element of convention or arbitrariness and an element that is non-
conventional or non-arbitrary in the relation of a name to its referent. That is, we
need to see the link as mediated: the name is conventionally or arbitrarily related
to something and that thing is non-conventionally, non-arbitrarily related to the
referent. Frege calls this thing the sense of the name. How does this help with our
problem?
What we wanted to be able to explain is that we can learn something when we
are told ‘a=b’. Well ‘a’ is arbitrarily related to the sense of ‘a’ and the sense of ‘a’ is
non-arbitrarily related to the referent of ‘a’. Likewise, ‘b’ is arbitrarily related to the
sense of ‘b’ and the sense of ‘b’ is non-arbitrarily related to the referent of ‘b’. What
we learn when we are told that ‘a=b’ is that the sense of ‘a’ and the sense of ‘b’ are
related to the same object. And, since neither of these relations, is a matter of arbitrary
this is a genuine piece of knowledge: conceivably, things might have been different.
The point can be made in terms of what speakers understand when they grasp
the meaning of an expression. If speakers grasped the reference of an expression then,
for those regions of language which they understood, their understanding would
suffice for them to arrive at a verdict on the truth-value of an identity statement. For,
in such a case, they’d know the referent of each name as part of their understanding of
it. They would simply need to reflect on this understanding in order to appreciate the
truth or falsity of an identity statement. Clearly this misconstrues the nature of our
understanding of language. So, we might say that in grasping the meaning of an
expression speakers do not grasp as much as the reference of an expression.
Conversely, it is clear that speakers do not grasp as little as the reference of an
expression. For, if they did, then co-referring terms (terms which refer to the same
thing) would be synonymous—indeed since Frege treats sentences as names for truth-
values all true sentences would be synonymous with one another, as would all false
sentences. In this case we couldn’t distinguish the sentences ‘a=a’ and ‘a=b’ in terms
of meaning.
The conclusion is that the reference of a term is not an ingredient of what
speakers understand when they understand the term’s meaning. In short, reference is
not part of meaning. What do speakers grasp when they understand a term? Amongst
other things they grasp its sense.
Now if we are to explain the informativeness of ‘a=b’ by saying that it informs
us that the sense of ‘a’ and the sense of ‘b’ share a reference. Then it had better be the
case that ‘a’ and ‘b’ have different senses. So it is possible for the same referent to
have many senses.
We are able now to explain or, better, to allow for the possibility of
meaningful terms which don’t have a reference. If we insist on thinking of the
meaning of a term as what it refers to then terms which fail to refer will fail to have a
meaning. But, once we introduce the notion of sense, it may be possible for a term to
have a sense, and thus to be meaningful, yet fail to have a reference.
Let us sum up so far what we’ve discovered about sense. (i) The sense of a
term is part of what speakers understand when they understand the term’s meaning.
(ii) The sense is non-conventionally, non-arbitrarily related to the term’s reference.
(iii) In fact the sense of the term is that ingredient of a term’s meaning which
determines its reference. For Frege, sense determines reference. (iv) The relation of
sense to reference is many-one, that is, many senses may share the same reference. (v)
In addition, a sense may fail to have a reference.
We should note also that Frege thinks that these arguments can be extended to
cover all terms of the language: all expressions have a sense. The sense of a sentence
is a thought.
So, according to Frege’s argument our analysis of the meaning of a name is
wrong. The reasons for this relate to the information content conveyed by sentences
which include names. And this is brought out by consideration of routine uses of such
sentences. So the starting point of analysis (‘Kilimanjaro’ means the highest mountain
in Africa) itself seems threatened; such sentences, even if they are sometimes used,
cannot be thought to capture the essentials of the way names mean since they are
blind to the name’s possession of a sense. Just to be clear, we are not saying that
names have senses. Frege says this. But we want a starting point of analysis that
doesn’t beg the question against Frege, whether or not he’s is right.

Sense and Statements of Belief

Consider the following argument:

Premise 1: King George doubts that Scott is the author of Waverley.


Premise 2: Scott is the author of Waverley.
Conclusion: Therefore King George doubts that Scott is Scott.

Clearly something has gone very awry here. The first premise attributes an
understandable doubt to King George who may simply be not very well informed
about Scott’s literary productions. The second premise states a truth. But the
conclusion attributes to King George a doubt about the law of identity and clearly it is
unlikely that even he was so confused: such a conclusion is surely not warranted by
the premises3.
Now it is hard to explain what goes wrong here. What, in effect, premise 2
tells us is that ‘Scott’ and ‘the author of Waverley’ have the same reference. And in
that case we should be able to substitute one term for the other without changing the
truth-value of any sentence in which they occur. This is a consequence of Frege’s
compositional view of semantic value. The semantic value of a name is its referent;
that of a sentence is its truth-value. According to compositionality, the semantic value
of a complex expression is determined by the semantic values of its components.
What seems to be our problem is that applying this principle in the context of
statements about someone’s beliefs, doubts, knowledge etc., leads from plausible
truths to evident untruths and absurdities.
How should we resolve the situation? We could either say that in the context
of statements of belief etc. the principle of the compositionality of semantic values no
longer holds. Or we could say that the semantic values of expressions in the context of
statements of belief etc. change so that terms which shared a semantic value, that is,
which referred to the same thing, may no longer do so. But what could the terms now
refer to? Frege’s answer is that the terms shift their reference from what is their usual
reference to what is usually their sense. So in the context of the sentence, ‘King
George doubts that Scott is the author of Waverley’, the name ‘Scott’ does not refer to
Scott, it refers to the sense of ‘Scott’. Similarly ‘the author of Waverley’ no longer
refers to the author of Waverley (i.e., to Scott) it refers to the sense of ‘the author of
Waverley’. Since the sense of ‘Scott’ and the sense of ‘the author of Waverley’ are
different we can no longer be assured that when we substitute ‘Scott’ for ‘the author
of Waverley’ the truth-value of the containing sentence remains unchanged. Note that,
as a consequence of this account, senses are in the realm of reference.

The Objectivity of Sense

Different speakers can grasp the same sense. Indeed grasping the same sense is a
condition for communication. When we communicate one speaker uses a sentence to
express a certain thought; to do so she must grasp that thought. She successfully
communicates only if her audience takes the sentence which she utters to express just
that thought. In order to do so the audience must grasp that thought. Thus
communication requires that speakers share the senses of their terms.
The sense of an expression must thus be sharply distinguished from any set of
subjective impressions associated with it. Ideas, impressions feelings are all things
that are had by a certain subject: one can always ask whose idea, whose feeling etc.
Something which can be had in this way can’t be shared or, perhaps better, can’t be
known to be shared. The most that we can determine is qualitative differences
between different people’s feelings, sensations or ideas. We cannot determine that
they are qualitatively identical since for that, the sensations would have to be
compared by a single subject. And that’s clearly impossible since no subject can have
another’s sensations or impressions. So sensations, feelings and ideas cannot be
known to be shared. Thus, since communication requires that we share and can know
that we share senses, senses must be sharply distinguished from ideas. That is, senses
are not part of the subjective realm and, for Frege, this means that they are objective.
(Another option might be to say that they belong to the inter-subjective realm but
Frege doesn’t consider this as a possible challenge to his dichotomy of subjective and
3
See Russell (1956: 39-55).
objective.) Moreover for Frege to say that senses are objective is to reify them: senses
are objective items in the world. He thus distinguishes three realms of existents. There
is the realm of the actual, which includes ordinary middle-sized objects etc., then
there is the subjective realm consisting of ideas and sense impressions and the like
and finally there is a third realm consisting of senses and abstract objects such as
numbers and the truth-values.

Frege, G. ‘Function and Concept’ in his Collected Papers and in his Philosophical
Writings
For exposition see the chapter by Grayling and Weiss in Philosophy: Further through
the Subject (Anthony Grayling ed.) and also chapter 1 of Alexander Miller’s
Philosophy of Language ‘Frege’s Distinction between Sense and Reference’4

4
Terminological Note:

There is an unfortunate lack of agreement amongst commentators about how to


translate Frege’s German term ‘bedeutung’. It is variously translated as: ‘denotation’,
‘nominatum’, ‘meaning’ and ‘reference’. I’ll use ‘reference’ and ‘denotation’ pretty
much interchangeably as translations. ‘meaning’ won’t be used in this way at all. And
‘nominatum’ won’t be used at all.

You might also like