Foundations of Semantics I:

Truth-conditions, entailment and logic

Rick Nouwen | [email protected]
February/March 2011, Gent

1 Meaning

There can be no simple answer to the question of what the meaning of a sentence is. The word meaning
has many different senses. This is why in linguistics we tend to be very explicit about the kinds of
meanings we are studying. Here are two examples:

(1) What is the meaning of a natural language expression?

a. that to which it refers
b. that which the speaker intends to communicate with it

The first answer to the question, (1-a), is what is normally associated with the study of natural language
semantics. Central to semantics is the relationship between a sentence and the world. So, the meaning
of John is the individual we refer to by the word ‘John’ and the meaning of John hates Bill is what needs
to be the case in order for this sentence to be true.

One could see the answer in (1-b) as typical to pragmatic inquiry. Beyond mere reference, pragmatic
meanings are about the use of a sentence in a particular context. For instance, consider the following
context. Someone has just asked Will John invite Bill to the party? If in this context, I answer John hates
Bill, it is likely that I intend this sentence to be understood as a negative answer to the question whether
John will invite Bill to the party.

There are interesting ways in which pragmatic meaning appears to depend on semantic meaning, and
there are many phenomena that straddle the semantics/pragmatics divide. It is therefore important to be
very precise about what kind of meaning one focuses on. Central to this course is the referential theory
of meaning that underlies what is often called formal, or truth-conditional, or model-theoretic semantics.

2 Truth-conditions

Apart from the referential nature of meaning, one crucial assumption in formal semantics concerns what
it means to know the (semantic) meaning of a sentence. Consider, (2).

(2) Rick has a 50 cent coin in his wallet.

To know (semantically) what (2) means is to be able to distinguish a situation in which (2) is true, from
one in which (2) is false. You clearly know how you would go about this: all you need to do is look in
my wallet and sift through the coins on the lookout for a 50c coin. Consequently, you know what (2)
means, even though you don’t know whether or not it is true.

To know the meaning of S is to know when S is true; that is, to know the conditions that make
it true: its truth-conditions.

1
Crucially, (2) tells us something very specific about the world we live in, namely that if you were to look
in my wallet you would find a 50c coin. This information is very specific in the sense that it leaves a lot
of other stuff open: whether there is more than one coin in my wallet; whether there are any other coins
in my wallet; what my wallet looks like; etc. In other words, there exists an infinity of situations that
make (2) true, but all these situations have one thing in common. This one thing is what the meaning of
(2) corresponds to.

If you haven’t looked in my wallet yet, and I assert (2) and you believe me, then you will have gained
information. Before accepting (2), you did not know whether the world you lived in was one in which I
have a 50c coin in my wallet. Afterwards, you did know (or at least believed so). This is what meanings
in the relevant sense do: they convey information about the world.1

Formal or truth-conditional semantics is sometimes called model-theoretic semantics. The idea is that a
sentence is true or false only with respect to a particular way things are, a particular model of what is
reality. In some state of affairs, the sentence is true, and in some others it will be false. Such alternative
state of affairs are often called a possible world. Imagine that apart from the world we live in, there are
many other possible worlds. Truth-values are relative to such possible worlds. For instance:

(3) World A: Obama is president of the US, Rick is a semanticist, Rome is the capital of Italy, Rick has
a 50 cent coin in his wallet, etc.

(4) World B: Obama is president of the US, Rick is a semanticist, Rome is the capital of Italy, Rick
doesn’t have a 50 cent coin in his wallet, etc.

(5) World C: Hilary Clinton is president of the US, Rick is a baker, Rome is the capital of Italy, Rick
has a 50 cent coin in his wallet.

(6) World D: etc.

The idea is that all the facts that are true describe a unique possible world, namely the actual one. At
the same time, there are many alternatives to these actual matters of fact. Semantic sentence meanings
can steer us towards finding out which possible world is actual. This is because semantic sentence
meanings, i.e. truth-conditions, are particular ways of distinguishing different possible worlds. One
knows the meaning of Rick has a 50 cent coin in his wallet if and only if one can distinguish worlds / models
in which it is true from worlds in which it is false.

3 Entailment

One of the prime sources of data for the study of semantics are entailments. You can use (intuitions about)
entailments to establish whether two (declarative) sentences are semantically independent, semantically
related or semantically identical.

Entailment — Sentence S entails sentence S’ if and only if whenever S is true, S’ is true too

In (7), you find an example of an entailment, indicated with ⇒.

(7) a. John owns a blue sweater.

b. ⇒ John owns a sweater.
1 Contrast this kind of informational view on meaning to the following use of the verb to mean, which is not the primary sense

of meaning that we’re after:

(i) Rick has a 50c coin in his wallet. This means that Rick can get a shopping trolley.

2
The denial of something true is false, and so, by the definition of entailment, we come to expect that a
sentence together with the denial of one of its entailments forms a contradiction. Contradictions cannot
be uttered felicitously.

(8) #John owns a blue sweater, but he does not own a sweater.

3.1. Entailment relations and truth-conditions

Given the notion of entailment, there are three kinds of meaning relations that may exist between two
sentences.

(9) either S⇒S’ or S’⇒S truth-conditionally related

neither S⇒S’ nor S’⇒S truth-conditionally unrelated
both S⇒S’ and S’⇒S truth-conditionally equivalent

The sentences in (7) are truth-conditionally related. That is, John owns a blue sweater entails John owns a
sweater, but not vice versa. Consequently, we cannot find a situation in which the former is true, but the
latter false, while we can find a situation in which it is true that John owns a sweater, but false that he
owns a blue one. (Just take a situation in which John’s sweater is red.)

Two truth-conditionally equivalent sentences, i.e. two sentences that entail one-another, have exactly the
same semantic meaning. This means that there are no situations in which their truth value differs: when
one of the two sentences is true, then both of them are true; when one of the two sentences is false, then
both of them are false.

For instance, (10-a) both entails and is entailed by (10-b). This suggests that the dative alternation in
English has no semantic import.

(10) a. John gave Mary an apple.

b. John gave an apple to Mary.

In summary, the relation between entailments and truth-conditions is a tight one. Above we said that S
entail S’ if whenever S is true, S’ is true too. An alternative way of saying the same thing is to make use
of the notion of possible world:

Entailment — Sentence S entails sentence S’ if and only if S’ is true in all possible worlds in
which S is true

We can depict truth-conditions by sketching the worlds in which sentences are true by means of Venn
diagrams.

S’

S
(11) S entails S0

S S’

(12) S and S0 are contradictory

3
(13) no entailment relation

(14) S’ equivalence

Exercise— How would you show that the two sentences in (15) are *not* truth-conditionally equivalent?
(15) a. All students didn’t do their homework.
b. Not all students did their homework.
Do the same for (16):
(16) a. The two girls lifted a piano.
b. Both girls lifted a piano.

Exercise: What are the entailment relations between the following (a.) and (b.) sentences?
(17) a. All famous boxers are rich.
b. All boxers are rich.
(18) a. John does not own a blue sweater.
b. John does not own a sweater.

Exercise — We can test a semantic theory, by testing the predicted entailments. Say, our theory gives the
following truth-conditions for the sentence S=“John did not see a unicorn”: S is true only in situations
in which there exists a unicorn that John did not see. Show that these truth-conditions make a wrong
prediction.

Exercise — Discuss the following sentences with respect to entailments.

(19) a. John was born in London or in Manchester.
b. John was born in London or John was born in Manchester.
c. Everybody in my class was born in London or in Manchester.
d. Everybody in my class was born in London or everybody in my class was born in Manchester.

Exercise — Discuss the meaning of cold, warm and hot, given the following intuition that (20-b) is a
contradiction, but (20-a) is not.
(20) a. The soup is warm. In fact, it is hot.
b. The soup is warm. #In fact, it is cold.

4
3.2. Entailment and embedding

Entailments are sensitive to their syntactic environment.

(21) a. John owns a blue sweater. ⇒ John owns a sweater.

b. Surprisingly, John owns a blue sweater. ⇒ John owns a sweater.
c. John doesn’t own a blue sweater. ; John owns a sweater.
d. Does John own a blue sweater? ; John owns a sweater.
e. If John owns a blue sweater, he will be considered cool. ; John owns a sweater.
f. Mary thinks that John owns a blue sweater ; John owns a sweater.

This is an important property, since it allows us to distinguish entailments from a second kind of
inference, a presupposition.

(22) a. John’s daughter is 12 years old.

b. ⇒ John has a daughter.

The relation between (22-a) and (22-b) is not one of entailment, because the inference remains in different
environments.

(23) a. Is John’s daughter 12 years old?

b. It’s not the case that John’s daughter is 12 years old.
c. If John’s daughter is 12 years old, then she could become friends with Sue.

4 Writing down truth-conditions

There is an immediate problem with using truth-conditions for the meaning of sentences. How will
we write them down? One option is to use paraphrases. For instance, the meaning of (24-a) is the
paraphrase of truth-conditions in (24-b).

(24) a. John owns a blue sweater.

b. There exists a blue sweater such that the individual named John owns this sweater.

Using paraphrases has an important drawback. By using paraphrases, the language we use to state our
truth-conditions in is the same as the language we are studying. We are thereby basically postponing
semantic analysis, for the meaning of the paraphrase is part of what our semantic theory needs to
explain. The biggest risk in this is in ambiguity. If a paraphrase is ambiguous, then it is unsuitable for
representing the truth-conditions of a sentence, for it would fail to determine a single set of conditions.
Consider, for instance, (25).

(25) The four boys ate three apples.

We could say that (25) is true if and only if the group of individuals that is the reference of the four boys
have the property that is the reference of ate three apples. This, unfortunately, is a semantic stale mate, for
(25) is very interesting from a semantic point of view, something which is completely obscured by using
the language in which the original sentence was stated as the language for presenting its truth-conditions.

What makes (25) interesting is that it is ambiguous. On one reading, it says that there were three apples
and that these were eaten by the boys. On another reading, however, the boys each have the property
of eating three apples. While in the first reading, three apples were eaten, in the second twelve apples

5
were eaten.

We could distinguish between the two readings by using more elaborate paraphrases, such as:

(26) a. There are three apples and the four boys ate these three apples
b. For each of the four boys there are three apples such that the boy in question ate these three
apples

What these complex paraphrases underline is the complexity of reaching an accurate statement of truth-
conditions. As stated before, the paraphrase needs to be completely unambiguous. The problem now
is what the relation is between the original sentence (25) and the two paraphrases in (26). Ideally,
we would want a systematic and completely predictive system that derives paraphrases for sentences.
This is another reason why we want to use a formal language as our metalanguage. Because our
theory of interpretation ought to account for the productive nature of meaning, in our mapping from
object to meta-language we must be able to systematically ensure that the meta-language expression is
unambiguous, we cannot do this on a case-by-case basis.

The common way to write down truth-conditions is therefore to choose a formal language as the
metalanguage, and to choose a formal language of which we know exactly the correspondence to truth-
conditions. Usually, we use a language called predicate logic. So, instead of the paraphrase in (27-b), we
will use the predicate logical sentence in (27-c) to represent the truth-conditions of (27-a).

(27) a. Rick has a 50 cent coin in his wallet.

b. there is a 50 cent coin which is contained in Rick’s wallet
c. ∃x[50-cent-coin(x) ∧ in(Rick’s-wallet,x)]

We will work towards a detailed understanding of predicate logic below. Before we do so, it is important
to understand what role the logic is going to play. A good theory of semantics will offer a system for
translating a natural language sentence into a predicate logical sentence. The logic is such that we have
full knowledge of what the meaning relations are between sentences in the logical language. In other
words, we know exactly which logical sentences entail which other logical sentences. It is the task of the
semantic theory to ensure that whenever there is an entailment relation between two natural language
sentences, there is also an entailment relation between the two corresponding logical sentences, and vice
versa.. Schematically,

(28) S ⇒ S’ (natural language)

↓ ↓ (semantic interpretation)
ϕ |= ψ (logical language)

Conversely, because we have full knowledge of the entailment relations within the logical language,
by mapping natural language sentences to logical ones, a semantic theory predicts what entailment
relations we should observe with respect to natural language sentences.