1

INTRODUCTION

1.1 THE CENTRAL QUESTION

This book is an introductory survey of five prominent answers to the
central philosophical question (henceforth, central question) about truth:

(Central Question) What feature do all and only the truths have in
common, which makes them all true?

More briefly: what is truth?

A common reaction to the central question is that it’s hopelessly
deep and complicated, far too complex to answer. This reaction may
stem from the following thought. If we could answer the central ques-
tion correctly, say by identifying Φ as the feature that makes a truth
true, then we would thereby be in a position to discover the truth
about every other subject matter as well. Want to know whether there
is extra-terrestrial life? Just test whether ‘There is extra-terrestrial life’
has Φ. Want to know whether nuclear war will break out this cen-
tury? Just test whether ‘Nuclear war will break out this century’ has
Φ. Clearly, philosophers cannot hope to discover the truth about
everything. As a result, philosophers cannot hope to discover the truth
about truth either.
The implicit assumption in the preceding argument is that the fea-
ture which makes a truth true must be much easier to recognize than

DOI: 10.4324/9781003190103-1
2 INTRODUCTION

truth itself. Although this would be nice, it need not be the case,
and in general, the answers philosophers have given to the central
question do not provide us with an algorithm for determining the
truth about every conceivable question. This may be disappointing
if you had turned to this book to help answer a question in quan-
tum physics, theology, or ethics. On the upside, it makes answering
the central question a tractable task, one which philosophers can
realistically hope to make progress on.
The first step to answering the central question is to clarify the
‘shape’ of the question and corresponding answers.

1.2 SHAPE OF THE CENTRAL QUESTION

AND THE SORT OF ANSWERS
The central question is one instance of the following general kind
of question: what feature do all and only the Fs have in common,
which makes them all F? More briefly: what is F? Other questions
of this general kind include:

• What feature do all and only the conscious beings have in

common, which makes them all conscious?
• What feature do all and only the morally right actions have in
common, which makes them all morally right?
• What feature do all and only the blue objects have in common,
which makes them all blue?
• What feature do all and only the diamonds have in common,
which makes them all diamonds?
• What feature do all and only the hot objects have in common,
which makes them all hot?
• For readers with some knowledge of mathematics: what feature
do all and only the continuous functions have in common, which
makes them all continuous?

The first question has mostly been investigated by philosophers and

neuroscientists. No consensus has emerged as to the correct answer,
put a popular view is that consciousness consists in certain kinds of
complicated information processing.
 INTRODUCTION 3

The second question has mostly been investigated by philosophers

and theologians. Again, no consensus has emerged around a single
answer, but popular views include the following: the moral rightness
of an action consists in it maximizing overall happiness, the moral
rightness of an action consists in it being in line with God’s com-
mands, and the moral rightness of an action consists in it being what
a wholly virtuous person would do.
The third question has mostly been investigated by physicists and
philosophers. In this case, something approaching a consensus has
emerged around the view that blueness consists in being disposed to
reflect light with wavelength between 450 and 495 nanometers under
normal conditions.
The fourth and fifth questions have been investigated almost exclu-
sively by physicists. Here, there is complete consensus: being a
diamond consists in being composed of carbon atoms arranged in
a certain kind of crystal structure, and being a hot object consists in
being composed of particles with high mean kinetic energy.
The sixth question has been investigated almost exclusively by
mathematicians. Again, here there is complete consensus: being a
continuous function consists in being a function for which the preim-
age of every open set is open. (Do not worry if you do not know
what this means.)
An answer to a question of the form ‘what is F?’ is called an analysis
of F. We write an analysis of F as follows:

x is F =df x is Φ

where Φ is the feature taken by the analysis to answer the ‘what is

F ?’ question. For example, the standard analysis of hotness can be
written as:

x is hot = df
x is composed of particles with high mean kinetic energy.

As can be seen from the above list, analyses are commonplace both
in philosophy and in science more generally. As can also be seen
from the above list, the methods used to give an analysis of F vary
substantially depending on the domain in which F is found.
4 INTRODUCTION

Although experimental observations were helpful in analyzing the

properties of being hot, being a diamond, and being blue, and likely
will be helpful in analyzing consciousness, experimental observations
were of no help in analyzing a mathematical function’s ‘continu-
ity’ and are unlikely to be of any help in analyzing moral rightness.
Although it is hard to draw the line between domains in which exper-
imental observations are relevant and domains in which they are not,
it is safe to say that truth, in this respect, is closer to mathematical
continuity than to being a diamond.
Despite the variation in methods used to give analyses, there
remain certain abstract features common to all correct analyses (i.e.,
correct answers to What is F? questions), which every analysis aspires
to satisfy. We briefly summarize these features below.

1.2.1 EXTENSIONAL ADEQUACY OF CANDIDATE ANSWERS

The most obvious requirement for an analysis of F to be correct is that

the feature identified by the analysis is in fact shared by all and only
the Fs. This condition is referred to as extensional adequacy. For exam-
ple, an analysis that took consciousness to consist in a certain kind
of complicated information processing would not be correct if there
were nonconscious beings (computers, perhaps) that perform exactly
the same kind of complicated information processing. In other words,
the would-be analysis claims that to be conscious is just to process infor-
mation in such-n-so way, and so the analysis would be incorrect in the
face of such nonconscious information processors.

1.2.2 INTENSIONAL ADEQUACY OF CANDIDATE ANSWERS

Extensional adequacy alone is not enough for an analysis to be cor-

rect. For example, for all we know, it might turn out that we on Earth
are the only conscious beings in the universe. If this were the case,
then the conscious beings would be all and only the animals from
Earth. Nevertheless, it would still be incorrect to analyze conscious-
ness as the property of being an animal from Earth. This is because
analyzing consciousness as being an animal from Earth entails that
 INTRODUCTION 5

it is impossible for a being to be conscious without also being from

Earth. Clearly, however, it is possible for extra-terrestrial conscious
beings to exist, even if, by chance, they do not.
In general, for Φ to be the correct analysis of F, it must be the case
that

(i) Φ is necessary for F, which means that it is not possible for

something to satisfy F without also satisfying Φ.
(ii) Φ is sufficient for F, which means that it is not possible for
something to satisfy Φ without also satisfying F.

An analysis that satisfies conditions (i) and (ii) is said to be inten-

sionally adequate. Note that intensional adequacy entails extensional
adequacy, since if, in actual fact, a feature Φ is not shared by all and
only the Fs, then Φ is not necessary and sufficient for F. (But, to
repeat, extensional adequacy does not entail intensional adequacy.)

1.2.3 EXPLANATORY POWER OF CANDIDATE ANSWERS

Even intensional adequacy is not enough for an analysis to be correct.

For example, an analysis of consciousness which took consciousness
to consist in being conscious would not be correct, despite being inten-
sionally adequate. This is because analyzing consciousness as being
conscious lacks explanatory power, in the sense that we cannot use the
analysis to derive any of the known phenomena involving conscious-
ness (e.g., the association between certain kinds of brain processes and
certain kinds of conscious states) from more fundamental principles.
Contrast the analysis of consciousness as being conscious with the
analysis of hotness as being composed of particles with high mean kinetic
energy. The latter analysis does have explanatory power because we can
use it to derive various known phenomena involving hotness from
more fundamental physical principles. For example, we can derive
the conditions under which certain kinds of hot objects will cool
down.
In general, for an analysis to be correct, it must have explanatory
power.
6 INTRODUCTION

1.2.4 UNIVOCAL AND NON-UNIVOCAL ANALYSES

We have established that for an analysis to be correct, it must be inten-

sionally (hence extensionally) adequate and have explanatory power.
There remains one further point to clarify. All of the proposed anal-
yses listed at the start of this section are univocal, which means that
they take a single feature to explain the presence of F in all Fs. It
could turn out, however, that different kinds of Fs are F for different
reasons.
An equivalent way of putting the same point is that the correct
analysis of F could have a so-called disjunctive, or either-or, form
such as
x is F =df (x is Φ1 ) or (x is Φ2 ) or … or (x is Φn )
where Φ1 , Φ2 , … , Φn are different features appropriate for different
kinds of Fs. The way we stated the central question seems to presup-
pose that the correct analysis of truth is univocal; however, univocity
is not required of correct analyses. Indeed, all of the analyses of truth
considered in this book are non-univocal to some extent, and some
are more radically non-univocal than others.

1.3 NECESSARY BACKGROUND ISSUES FOR THE

CENTRAL QUESTION
The central question revolves around a fair number of issues from
‘truth bearers’ to ‘context sensitivity’. We address those issues before
sketching the plan of the book.

1.3.1 FALSITY

Truth has a natural counterpart: falsity. Consequently, you might

expect an analysis of truth to be paired with an analysis of falsity. It
turns out, however, that any analysis of truth can straightforwardly be
generalized to an analysis of falsity. This is because falsity is plausibly
analyzed as follows:
x is false =df the negation of x is true
 INTRODUCTION 7

where the negation of the sentence ‘p’ is obtained by prefixing ‘p’

with the negation operator, namely, ‘it is not the case that’.1 For
example, the falsity of the sentence ‘It is raining in London’ consists
in the truth of the sentence ‘It is not the case that it is raining in
London’. Thus, if Φ is an analysis of truth, then we can substitute ‘is
Φ’ for ‘is true’ in the above analysis to obtain

x is false =df the negation of x is Φ.

We’ll assume such an analysis of falsity throughout.

1.3.2 TRUTH BEARERS

We now turn to issues that arise specifically for the analysis of truth.
The first issue: what kinds of objects are true? Equivalently: truth is
a property of what sort of thing?
In ordinary life, we most commonly predicate truth of two distinct
kinds of objects: sentences and beliefs. For example, we can say both
that the sentence ‘It is raining in London’ is true and that Ahmed’s
belief that it is raining in London is true. It turns out to be more
fruitful to focus on the truth of sentences, rather than the truth of
beliefs, for two reasons.
The first reason it’s more fruitful to focus on sentences than beliefs
is that, unlike sentences, which are just finite sequences of symbols,
we do not currently have a clear idea of how our beliefs are structured.
The most popular contemporary view about beliefs is that our beliefs
are something like brain processes. However, we are yet to locate
exactly which process in Ahmed’s brain is his particular belief that it
is raining in London.
The second reason it’s more fruitful to focus on sentences than
beliefs derives, in a sense, from the first. Because our beliefs are hid-
den from public view, we developed language in order to share our
beliefs. Consequently, every belief is capable of being expressed by
a sentence, which is true under exactly the same conditions as the
belief it expresses. For example, Ahmed’s belief that it is raining
in London is expressible by the sentence ‘It is raining in London’,
which, like Ahmed’s belief, is true if and only if it is in fact raining
8 INTRODUCTION

in London. Thus, by analyzing truth for sentences, we automatically

get an analysis of truth for beliefs: what makes a belief true is being
expressible by a true sentence.
Reflecting the foregoing, we assume that the correct analysis of
truth has the following non-univocal form:

x is true =df (x is a belief and x is expressible by a true sentence) or

(x is a sentence and x is Φ)

where different theories of truth fill out Φ in different ways.

Having clarified how the analysis of truth for beliefs should go, we
will talk about truth in the remainder of the book as if it applied only
to sentences.

1.3.3 FOREIGN LANGUAGES

Since people differ in which language they use to express their beliefs,
a complete account of truth should explain what the truth of the sen-
tences of each language consists in. However, since languages differ
greatly in their syntax, any account of truth which is simultaneously
applied to all languages would, by necessity, have a very abstract char-
acter. Consequently, we prefer to analyze truth for each language
separately.
Given that this book is written in English, we focus on how to
analyze truth for sentences of English. It is important to clarify, how-
ever, that we do not take the truth of sentences of English to be more
fundamental than the truth of sentences of, say, German, Hindi, or
Mandarin Chinese. In particular, we do not suppose that the correct
account of truth for, say, German is the property of being synony-
mous with a true sentence of English. Rather, we suppose that the
correct analysis of truth for sentences has the form:

x is true =df (x is a sentence of English and x is Φ1 ) or

(x is a sentence of German and x is Φ2 ) or
(x is a sentence of Hindi and x is Φ3 ) or
(x is a sentence of Mandarin Chinese and x is Φ4 ) or...
 INTRODUCTION 9

where Φ1 , Φ2 , Φ3 , Φ4 , … are distinct yet structurally similar proper-

ties. Since this book is written at an introductory level, we do not
attempt to generalize the theories of truth for English surveyed here
to other languages. Nevertheless, we hope the reader will not be left
in doubt that this could in principle be done.
For ease of expression, we will talk about truth in the remainder
of the book as if it applied only to sentences of English.

1.3.4 SCOPE AMBIGUITY

Is the following sentence true?

The Earth is flat and the Earth is circular or the Earth is spherical.

The correct answer is: it depends. In particular, the truth status of the
sentence depends on how the different components of the sentence
are grouped together. We can use brackets to make clear the possible
groupings:

1. (The Earth is flat and the Earth is circular) or the Earth is spherical.
2. The Earth is flat and (the Earth is circular or the Earth is spherical).

When grouped according to (1), the sentence asserts that one of the
following conditions obtains: (i) the Earth is flat and circular, (ii) the
Earth is spherical. Since condition (ii) does obtain, the sentence is
true according to grouping (1).
By contrast, when grouped according to (2), the sentence asserts
that both of the following conditions obtain: (i) the Earth is flat,
and (ii) the Earth is circular or spherical. Since condition (i) does not
obtain, the sentence is not true according to grouping (2).
Because sentences of English frequently admit different possible
groupings, and the way a sentence is grouped can affect its truth
status, a theory of truth should, strictly speaking, apply to sentences
that have been bracketed to force a unique grouping. However, since
this book is written at an introductory level, we will mostly gloss over
this issue, leaving the brackets implicit.
10 INTRODUCTION

1.3.5 LEXICAL AMBIGUITY

Scope ambiguity is not the only kind of ambiguity found in lan-

guage – there is also lexical ambiguity. To understand lexical ambi-
guity, consider whether the following sentence is true:

A bank is usually a good place to fish.

As before, the correct answer is: it depends. However, on this occasion,

whether the sentence is true does not depend on how it is grouped,
but rather on how the word ‘bank’ is interpreted.
In English, the word ‘bank’ has at least two possible interpreta-
tions: (i) a type of financial institution and (ii) a type of geographical
area – in particular, the type of geographical area which immediately
borders a river. According to interpretation (i), the sentence is false –
financial institutions, after all, do not usually contain fish. By con-
trast, according to interpretation (ii), the sentence is true, since rivers
do usually contain fish.
Just as we can resolve scope ambiguities by adding appropriate
brackets, we can resolve lexical ambiguities by adding appropriate
indices to ambiguous words and then providing a key that informs us
which interpretation the indices refer to. For example, in the case of
‘bank’, our key would contain the following two entries:

bank1 : The type of financial institution in which money is

deposited.

bank2 : The type of geographical area that borders a river.

We can then enforce a disambiguation of ‘bank’ by adding the

appropriate index:

A bank2 is usually a good place to fish.

Given that many words in English admit multiple possible interpreta-

tions, and how we choose to interpret those words can affect the truth
status of a sentence in which they appear, a theory of truth should,
 INTRODUCTION 11

strictly speaking, apply to fully disambiguated sentences – sentences

that have both been bracketed to resolve all scope ambiguities and,
in addition, had indices added to resolve all lexical ambiguities.
However, since this book is written at an introductory level, we will,
like the brackets, leave the indices implicit.

1.3.6 TWO KINDS OF EQUIVALENCE

Before introducing the next key feature of truth, we need to clarify

some terminology.
We say the condition that p is materially equivalent to the condition
that q when in actual fact, either both conditions obtain or both
conditions fail to obtain. For example, supposing we are alone in the
universe, the condition that there is a conscious being in the room
is materially equivalent to the condition that there is an animal from
Earth in the room. The sentence ‘p if and only if q’ asserts that the
condition that p is materially equivalent to the condition that q.
We say the condition that p is necessarily equivalent to the condition
that q when in every possible world, either both conditions obtain or
both conditions fail to obtain. Necessary equivalence is much more
demanding than material equivalence. For example, because there is
a possible world in which intelligent extra-terrestrial life exists, the
condition that there is a conscious being in the room is not necessarily
equivalent to the condition that there is an animal from Earth in the
room. The sentence ‘Necessarily, p if and only if q’ asserts that the
condition that p is necessarily equivalent to the condition that q.

1.3.7 TRUTH CONDITIONS

The most central feature of truth is that the truth of each sentence
‘p’ is materially equivalent to the condition that p. For example, the
truth of ‘It is raining in London’ is materially equivalent to the con-
dition that it is raining in London, the truth of ‘Unemployment
is falling in the US’ is materially equivalent to the condition that
unemployment is falling in the US, and the truth of ‘The aver-
age temperature on Earth is rising’ is materially equivalent to the
condition that the average temperature on Earth is rising.
12 INTRODUCTION

Consequently, truth satisfies the following principle, which is

referred to in philosophy as the T-schema:

‘p’ is true if and only if p.

The T-schema might initially strike you as a mere tautology, about

as informative as the principle that p if and only if p. But appear-
ances can be misleading, and in this case they are. If the T-schema
strikes you as tautologous, then this is only because you speak English.
Consider, for example, the following instance of the T-schema:

‘It is raining in London’ is true if and only if it is raining in London.

If you translate this sentence into, say, French, then it would strike
a French speaker as far from trivial. This can be seen by going the
other way around and translating the corresponding French instance
of the T-schema into English:

‘Il pleut à Londres’ is true if and only if it is raining in London.

It should now be clear that, far from being a tautology, the

T-schema asserts a non-trivial equivalence between a linguistic con-
dition on the one hand and a ‘worldly’ condition on the other hand.
We refer to the condition that p as the truth condition of the sentence
‘p’, since it is the worldly condition under which ‘p’, in actual fact,
is true.

1.3.8 THE LIAR PARADOX

Because the T-schema captures the most central feature of truth, it

is a basic requirement on the correctness of an analysis of truth as Φ
that Φ satisfies the analogous Φ-schema:

‘p’ is Φ if and only if p.

Thus, the instances of the Φ-schema should at least be consistent with

our background knowledge. It turns out, however, that the T-schema
runs into consistency problems all by itself, and the consistency prob-
lems for the T-schema automatically transfer to the Φ-schema for any
 INTRODUCTION 13

other choice of Φ. Consequently, these problems need to be resolved,

or at least appropriately contained, before we can attempt to analyze
truth.
The T-schema runs into consistency problems when we attempt
to predicate truth of sentences that themselves contain the predicate
‘is true’. Suppose, for example, that we give the name ‘The Liar’ to
the sentence ‘The Liar is not true’. By instantiating the T-schema,
we get
‘The Liar is not true’ is true if and only if The Liar is not true.
Consequently, since The Liar just is the sentence ‘The Liar is not
true’, we get
The Liar is true if and only if The Liar is not true
which, according to the orthodox account of logic, is impossible.
Since we should reject any theory that entails an impossibility, we
must either reject the orthodox account of logic or else restrict the
T-schema to some privileged subset of sentences. We postpone a full
discussion of the issues involved in rejecting the orthodox account of
logic to accommodate the unrestricted T-schema to Chapter 7. Until
then, we restrict our attention to only the correct account of truth
for sentences that do not themselves contain the predicate ‘is true’,
for which the liar paradox does not arise.
* * Parenthetical note: A further slightly technical restriction. Since replac-
ing ‘is true’ in the unrestricted T-schema with any other predicate
also runs afoul of the liar paradox, we need to further restrict our
attention to sentences that do not contain our proposed analysis of
truth. Therefore, we assume that the correct analysis of truth for
English has the form:
x is true =df (x contains no truth-related notions and x is Φ1 ) or
(x contains truth-related notions and x is Φ2 )
and, until Chapter 7, restrict our attention to filling out Φ1 . For
ease of expression, until Chapter 7, we will talk as if truth only
applied to sentences that do not themselves contain any truth-related
notions. End note. * *
14 INTRODUCTION

1.3.8.1 Tarski on Extensional Adequacy

Suppose Φ is an analysis of truth. As previously argued, at a minimum

all instances of the Φ-schema:

‘p’ is Φ if and only if p,

where ‘p’ contains no truth-related notions, should be consistent

with our background knowledge. But it would be better if not
only were the instances of the Φ-schema consistent with our back-
ground knowledge, but our background knowledge actively entailed
the instances of the Φ-schema. This is for two reasons.
First, if our background knowledge entails the instances of the Φ-
schema, then analyzing truth as Φ has explanatory power, because
it explains why the T-schema holds. This is analogous to the way in
which the analysis of being a diamond as being composed of carbon
atoms arranged in a certain kind of crystal structure has explanatory
power in virtue of entailing various observable properties of diamonds
(their color, hardness, ability to conduct heat, and so on).
The second reason why it is better for our background knowledge
to entail the instances of the Φ-schema is more subtle. The first per-
son to grasp this reason was the Polish logician Alfred Tarski, who has
done more than anyone to advance our understanding of truth. What
Tarski understood was that if our background knowledge entails all
instances of the Φ-schema, then we thereby have a theoretical guar-
antee that the analysis of truth as Φ is extensionally adequate, which,
recall, means that all and only the truths have Φ. This is because, for
any sentence ‘p’, it follows from:

‘p’ is Φ if and only if p

and
‘p’ is true if and only if p
that
‘p’ is true if and only if ‘p’ is Φ.
The fact that we have a theoretical method of guaranteeing the exten-
sional adequacy of an analysis of truth is a remarkable fact, which
distinguishes the analysis of truth from most other cases of analysis. In
 INTRODUCTION 15

particular, there is no theoretical method of guaranteeing the exten-

sional adequacy of an analysis of consciousness, or of moral rightness,
or even of blueness. In these cases, we argue for the extensional ade-
quacy of an analysis of F as Φ by, in effect, checking that each object
in a large sample of Fs has Φ, and each object in a large sample of
non-Fs lacks Φ. Clearly, Tarski’s method is a vast improvement over
such a piecemeal approach.

1.3.9 CONTEXT SENSITIVITY

Consider the following puzzle. John is a 6′ 2″ professional basketball

player. Mary points to John while he is playing basketball and asserts
‘That guy is not tall’. In contrast, Kylie points to John while he is
shopping at the mall and asserts ‘That guy is tall’. On the surface,
these two assertions are incompatible – John cannot be both tall and
not tall. Yet, we are also inclined to accept that both Mary and Kylie
speak truly. So what gives?
The solution to this puzzle is probably already apparent to you.
The predicate ‘is tall’ is context sensitive, in the sense that its meaning
depends on the context in which it is used. When Mary says ‘John
is not tall’, she means that John is not tall for a male basketball player.
Since John’s height is below the average height of a male basketball
player, what Mary says is correct.
By contrast, when Kylie says ‘John is tall’, she means that John is tall
for a male shopper at the mall. Since John’s height is significantly above
the average height of a male shopper at the mall, what Kylie says
is also correct. Thus, despite the superficial appearance of conflict,
there is in fact no disagreement between Mary’s statement and Kylie’s
statement.
* * Parenthetical note: The difference between lexical ambiguity and context
sensitivity. You might be wondering how context sensitivity differs
from lexical ambiguity. What is the difference between the context
sensitivity of ‘tall’ and the lexical ambiguity of ‘bank’?
The answer is that, unlike ‘bank’, whether ‘tall’ applies to a par-
ticular object on a given occasion of use depends in a systematic way
on the context in which ‘tall’ is used. In particular, the context of
use provides a reference class of objects – the objects the conversa-
tional participants are talking about – and ‘tall’ applies to an object
16 INTRODUCTION

in that context if and only if the height of the object in question is

sufficiently above the average height of the reference class.
By contrast, no systematic story can be given as to whether ‘bank’
applies to financial banks or river banks on a particular occasion of
use – it is largely just a matter of what the speaker has in mind. End
note. * *
Since the instances of the T-schema should hold in every context, it
follows that ‘is true’ inherits all of the context sensitivity present in the
language to which it is applied. For example, because Mary would be
correct to assert ‘John is not tall’, she would also be correct to assert
“The sentence ‘John is tall’ is not true”. Likewise, because Kylie
would be correct to assert ‘John is tall’, she would also be correct to
assert “The sentence ‘John is tall’ is true”. Consequently, to assert
“The sentence ‘p’ is true” in context C is to assert that ‘p’ is true
relative to C.
In particular, when Mary says “The sentence ‘John is tall’ is not
true”, she thereby asserts that ‘John is tall’ is not true in her present
context, in which the standard for tallness is determined by the set of
all male basketball players. Similarly, when Kylie says “The sentence
‘John is tall’ is true”, she thereby asserts that ‘John is tall’ is true in
her present context, in which the standard for tallness is determined
by the set of all male shoppers at the mall. Again, the appearance of
conflict between Mary’s statement and Kylie’s statement dissipates.
Given that “The sentence ‘p’ is true” means in context C that ‘p’
is true relative to C, you might expect that an analysis of truth would
have the form

x is true relative to context C =df x stands in relation R to C.

Although some philosophers favor analyzing truth in this manner,

this is not the approach we take in this book. Instead, we only survey
analyses of truth with the form:

x is true =df x is Φ

where the context used to interpret sentences has been implicitly

fixed in the background. There are two main reasons for favoring
our approach.
 INTRODUCTION 17

First, since we do not need to account for the parameter C, the

resulting theories of truth are considerably simpler. This suits the
introductory nature of this book.
Second, one of the leading theories of truth – the transparency
theory – can only be formulated when the context for interpret-
ing sentences is implicitly fixed in the background. Although some
philosophers take this to count against the transparency theory, it is
better to adopt a framework that does not, from the outset, rule out
the possibility that the transparency theory is the correct theory of
truth.

1.3.10 RELATIVISM

Roughly speaking, relativism about truth is the thesis that sentences

are not true absolutely, but rather are only true relative to some other
factor. One way to make this thesis precise is to take the other factor
to be the context in which the sentence is used. Under this interpre-
tation, relativism about truth is the thesis that whether a sentence is
true depends on the context in which it is used.
As we saw in the previous section, this version of relativism about
truth is unquestionably correct. To take an uncontroversial example,
suppose Dan and Zoe are wearing blue and red shirts, respectively. At
the very same moment of time, Dan and Zoe both utter the sentence
‘I am wearing a blue shirt’. In this case, despite uttering the very same
sentence, Dan says something true but Zoe says something untrue.
This discrepancy arises because the context in which a sentence is
used – more specifically, the identity of the speaker of the sentence –
determines who the indexical expression ‘I’ refers to. Relative to a
context in which the speaker is Dan, ‘I’ refers to Dan and so, since
Dan is in fact wearing a blue shirt, the sentence ‘I am wearing a blue
shirt’ is true. Likewise, relative to a context in which the speaker is
Zoe, ‘I’ refers to Zoe and so, since Zoe is not in fact wearing a blue
shirt, the sentence ‘I am wearing a blue shirt’ fails to be true.
Relativism about truth is generally taken to be a radical, counter-
intuitive doctrine. Thus, identifying relativism about truth with the
uncontroversial thesis that truth depends on context does not do jus-
tice to what proponents of relativism about truth intend their view to
be. A full discussion of how best to understand relativism about truth
18 INTRODUCTION

is beyond the scope of this book. However, we will briefly discuss

two alternative ways of making relativism about truth precise, which
strike us as better approximations of what relativists intend.
According to the first, more radical way of understanding rela-
tivism about truth, relativism about truth is the thesis that, even if we
were to remove all context sensitivity from the part of English which
does not contain any truth-related notions, ‘is true’ would remain
context sensitive. That is to say, even if we were to remove index-
ical expressions, such as ‘I’, ‘you’, and ‘that’, and non-truth-related
context sensitive predicates, such as ‘tall’, ‘thin’, and ‘old’, whether
‘is true’ applies to a sentence would still depend on the context in
which ‘is true’ is used. Therefore, on this understanding, relativists
about truth consider ‘is true’ to be an independent source of con-
text sensitivity, over and above the context sensitivity present in the
non-truth-related part of language.
For example, a relativist of this kind might maintain that, even
though the sentence ‘Mount Everest is taller than K2’ is not context
sensitive, the sentence
‘Mount Everest is taller than K2’ is true
is context sensitive, since its meaning relies on some contextual
information. This contextual information is traditionally taken to be
something like the way the user of the sentence conceptualizes the
world. On this view, even when we remove all the context sensitivity
in ‘p’, there is still no sense in which ‘p’ could be true absolutely – the
most we can say is that ‘p’ is true relative to one of the many possible
conceptualizations of reality.
Call this version of relativism strong relativism about truth.
Although strong relativism might seem plausible, it can easily be
refuted. To see how, imagine that Miho and Shirley are two speak-
ers of English who conceptualize the world in different ways. At the
very same moment of time, Miho asserts The sentence ‘Mount Everest is
taller than K2’ is true, while Shirley asserts The sentence ‘Mount Everest
is taller than K2’ is not true, where we assume that ‘Mount Everest is
taller than K2’ is not context sensitive.
According to the strong relativist, Miho and Shirley could both be
right. Yet, this cannot be. For if Miho is right then, by the T-schema,
 INTRODUCTION 19

she would also be right in asserting Mount Everest is taller than K2.
Likewise, if Shirley is right, then, by the T-schema, she would also
be right in asserting Mount Everest is not taller than K2. But then, given
our assumption that ‘Mount Everest is taller than K2’ is not context
sensitive, we can infer that at the time Miho and Shirley make their
assertions, Mount Everest both is and is not taller than K2, which is
impossible.2
So the relativist about truth cannot allow the context sensitivity
of ‘is true’ to float free of the context sensitivity of the underlying
language. Nevertheless, it remains an open possibility that the non-
truth-related predicates in our language are more context sensitive
than we ordinarily assume – so that, for example, ‘Mount Everest is
taller than K2’ means in a given context that Mount Everest is taller
than K2 relative to the speaker’s way of conceptualizing reality. This
entails the truth of ‘Mount Everest is taller than K2’ is also relative
to a conceptualization – but only because being taller than is already
relative to a conceptualization. On this view, truth is radically relative,
not in virtue of anything intrinsic to truth, but rather in virtue of the
language to which it is applied.
By adding the speaker’s way of conceptualizing the world to the
context for determining the meaning of a sentence, relativists can do
justice to the radical nature of relativism about truth, while maintain-
ing our ordinary understanding of the nature of truth. Consequently,
everything we say in this book is compatible with this version of rel-
ativism about truth (recall from the previous section we assume the
context for determining the meaning of a sentence is implicitly fixed
in the background).

1.4 PLAN OF THE BOOK

Having addressed the various background issues around the central
question of truth, we now turn to answers. In this book, we discuss
five major answers to the central question:

1. Correspondence. According to the correspondence answer,

truth consists in correspondence to the facts, where facts are
understood to be the existing portions of reality.
20 INTRODUCTION

2. Semantic. According to the semantic answer, the meanings of

the simple words in a sentence combine to directly associate a
sentence with its truth condition. A sentence is then ‘made true’
by its truth condition ‘obtaining’.
3. Verifiability. According to the verifiability answer, every sen-
tence is associated with a procedure and an outcome, such that
if following the procedure were to result in the outcome, the
community of speakers would agree to accept the sentence. The
truth of a sentence then consists in the condition that following
its associated procedure would result in its associated outcome,
regardless of whether the procedure is ever actually followed.
4. Transparency. According to the transparency answer, to assert
that ‘p’ is true just is to assert that p. Consequently, there is no
single feature common to all truths which explains why they are
true. Instead, any explanation of why ‘p’ is true is an explanation
of why p (so to speak). For example, an explanation of why ‘Sea
water is salty’ is true is an explanation of why sea water is salty, an
explanation of why ‘The Earth revolves around the Sun’ is true
is an explanation of why the Earth revolves around the Sun, and
so on.
5. Plurality. According to the plurality answer, different accounts
of truth are required for sentences with different subject matters.
For example, while the correspondence answer is suitable for sen-
tences about (say) ordinary macroscopic reality, the verifiability
answer is more suitable for sentences about (say) mathematical,
ethical, or microscopic reality.

Multiple different versions of each of these five answers can be found

in the philosophical literature, and philosophers who propose the
same general answer to the central question may nevertheless have
substantial theoretical disagreements. Since this is an introductory
book, we do not in any fashion attempt to cover all of the inter-
nal disagreements between proponents of the same general answer.
Instead, we offer very simplified, yet sufficiently precise, formula-
tions of what we take to be representative versions of each of the
five answers outlined above. There may well be (in fact, probably is)
no actual theorist who holds exactly the sketched representative view;
 INTRODUCTION 21

however, the representative views give a sufficient family resemblance

to important actual answers that are advanced by actual theorists.
To aid the reader in comparing the five answers, every chapter has
the same structure:

1. Answer to the central question. In this section, we outline

the analysis of truth on offer.
2. Motivation. In this section, we explain why philosophers have
been interested in the analysis.
3. Argument. In this section, we present an argument for the target
analysis.
4. Evaluation. In this section, we do two things. First, we consider
some objections to the argument offered in the previous section.
Second, we consider some objections to the analysis itself, which
apply regardless of the argument used in support of it.

In Chapter 7, after evaluating each of the five candidate answers to

the central question, we take up the challenge of answering the liar
paradox. This chapter is more technical in nature than the chapters
that precede it, and many of the issues raised in the final chapter are
to a large extent independent of the issues raised in the rest of the
book. Nevertheless, since any adequate theory of truth must ulti-
mately apply to the entirety of English, tackling the liar paradox is
a task that every truth theorist – regardless of whether they endorse
the correspondence, semantic, verifiability, transparency, or plurality
answer – must eventually undertake.
At the end of the book, you can find various technical appendices
that are referenced throughout the text. These offer a more advanced
discussion of some topics discussed in the book and are generally
suitable for readers with prior knowledge of philosophy and formal
logic.

1.5 CHAPTER SUMMARY

In this chapter, we introduced the central question of the book: what
is the feature shared by all and only the truths, which makes them
all true? We explained that for an answer to the central question to
22 INTRODUCTION

be correct, it must be intensionally adequate and have explanatory

power. We also clarified some necessary background issues regarding
truth. In particular, we clarified that, in this book, we take truth to
primarily be a property of sentences, rather than beliefs. In addition,
we pointed out that the most central feature of truth – consequently,
the feature of truth most in need of explanation – is the T-schema:

‘p’ is true if and only if p.

NOTES
1. It should be noted that analyzing falsity in this manner is philosophically controversial.
2. See Appendix A (page 155) for a more precise formulation of this argument.

FURTHER READING
• For further discussion of the nature of analysis, see Gideon Rosen,
‘Real Definition’, Analytic Philosophy, Vol. 56, No. 3, pp. 189–209,
2015.
• For further discussion of the T-schema, see Sections 1 (pp. 154 –
165) and 3 (pp.186 –209) of Alfred Tarski, ‘The Concept of Truth
in Formalized Languages’, in Logic, Semantics, Metamathematics
(Hackett Publishing Company: Oxford, 1956).
• For further discussion of relativism, see Herman Cappelen
and Torfinn Thomesen Huvenes, ‘Relative Truth’, in Michael
Glanzberg (ed.), The Oxford Handbook of Truth (Oxford University
Press: Oxford, 2018), pp. 517–542.