3

SEMANTIC

3.1 ANSWER TO THE CENTRAL QUESTION

According to the semantic theory of truth, the meanings of the names
and predicates in a sentence combine in a systematic way to deter-
mine the truth condition of the sentence. A sentence is then made
true by its truth condition obtaining.
So far, the semantic theory is indistinguishable from the correspon-
dence theory. The innovation of the semantic theory is that it does
not reduce truth to a property involving facts. Instead, the semantic
theory reduces truth to a property we call semantic correctness, which
is definable without reference to any philosophically controversial
entities:

(Semantic Theory) x is true =df x is semantically correct.

The remainder of this section is dedicated to the definition of

semantic correctness.

3.1.1 INDUCTIVE DEFINITIONS

The definition of semantic correctness is a type of definition known

as an inductive definition. As an example of an inductive definition,

DOI: 10.4324/9781003190103-3
50 SEMANTIC

consider the property of being a descendant of Genghis Khan. This

property is defined as follows:

(Descendant of GK) x is a descendant of Genghis Khan =df x is a

child of Genghis Khan or x is the child of a
descendant of Genghis Khan.

Note that the property this definition is attempting to define – being a

descendant of Genghis Khan – occurs in the definition. The definition
is therefore circular. However, we should distinguish two ways in
which a definition can be circular.
The first kind of circularity, known as vicious circularity, occurs
when the definition fails to determine whether the defined prop-
erty applies in particular cases. As an example of a viciously circular
definition, consider

(Viciously Circular) x is blue =df x is blue.

This definition does not allow us to determine whether a given object

is blue, since to do so we would already need to know whether the
object is blue. Viciously circular definitions are therefore defective.
By contrast, the second kind of circularity, known as virtuous cir-
cularity, occurs when the definition does determine whether the
defined property applies in particular cases. The definition of being
a descendant of Genghis Kahn is an example of a virtuously circular
definition. To see why, consider a given person x. To determine
whether x is a descendant of Genghis Khan, it suffices to check
whether there exists a sequence of people y1 , y2 , y3 , … , yn such that
(i) y1 is a child of Genghis Khan, (ii) yn is x and (iii) for every i < n,
yi is a parent of yi+1 . If so, we know x is a descendant of Genghis
Khan. If not, we know x is not.
In general, an inductive definition defines a property P by first
defining, in an entirely non-circular manner, an initial set of objects
with property P, and then specifies rules by which P can be trans-
ferred from one object to another. For example, in the case of being
a descendant of Genghis Khan, the definition first specifies that the
children of Genghis Khan are descendants, and then specifies that a
descendant transfers the property of being a descendant to their chil-
dren. Inductive definitions are virtuously circular because, to decide
 SEMANTIC 51

whether a given object x has an inductively defined property P, it

suffices to check whether x can be obtained by repeatedly applying
the transfer rules to some objects in the initial set.6

3.1.2 SEMANTIC CORRECTNESS

We now turn to the definition of semantic correctness. For simplicity,

we define semantic correctness for the fragment of English obtained
by repeatedly applying the connectives ‘and’, ‘or’ and ‘not’ to atomic
sentences. Note, however, that a fully general definition of semantic
correctness would need to apply to all complex sentences.

3.1.2.1 Base Cases

In accordance with the structure of an inductive definition, we begin

by defining, in an entirely non-circular manner, which atomic sen-
tences and negations of atomic sentences are semantically correct.
These definitions are known as the base cases of the definition of
semantic correctness.

(Positive Base Case) ‘t1 ,…,tn are R-related’ is semantically correct

=df the objects referred to by t1 ,…,tn (in that
order) stand in the relation referred to by R.
(Negative Base Case) ‘t1 ,…,tn are not R-related’ is semantically cor-
rect =df the objects referred to by t1 ,…,tn (in
that order) do not stand in the relation referred
to by R.

For example, ‘The Amazon is shrinking’ is semantically correct

because the object referred to by ‘The Amazon’ has the property
referred to by ‘is shrinking’. Similarly, ‘Monaco is not colder than
Scotland’ is semantically correct because the objects referred to by
‘Monaco’ and ‘Scotland’, in that order, do not stand in the relation
referred to by ‘is colder than’.
Our background knowledge entails that an atomic sentence ‘p’ is
semantically correct if and only if p. Consequently, analyzing truth as
semantic correctness explains the T-schema for atomic sentences. To
take an arbitrary example, consider the sentence ‘The Pacific Ocean
is frozen over’. This sentence is formed by applying ‘is frozen over’
52 SEMANTIC

to ‘The Pacific Ocean’. Thus, ‘The Pacific Ocean is frozen over’

is semantically correct if and only if the object referred to by ‘The
Pacific Ocean’ has the property referred to by ‘is frozen over’. We
know that ‘The Pacific Ocean’ refers to the Pacific Ocean and ‘is
frozen over’ refers to the property of being frozen over. Consequently,
‘The Pacific Ocean is frozen over’ is semantically correct if and only
if the Pacific Ocean is frozen over.
Likewise, our background knowledge entails that the negation ‘not
p’ of an atomic sentence ‘p’ is semantically correct if and only if it is
not the case that p. Consequently, analyzing truth as semantic cor-
rectness also explains the T-schema for negations of atomic sentences.
To take an arbitrary example, consider the sentence ‘Oxford is not
foggy’. This sentence is formed by applying ‘is foggy’ to ‘Oxford’.
Thus, ‘Oxford is not foggy’ is semantically correct if and only if the
object referred to by ‘Oxford’ lacks the property referred to by ‘is
foggy’. We know that ‘Oxford’ refers to Oxford and ‘is foggy’ refers
to the property of being foggy. Consequently, ‘Oxford is not foggy’
is semantically correct if and only if Oxford is not foggy.

3.1.2.2 Inductive Clauses

Next, we specify the rules by which semantic correctness is passed

from one sentence to another. More specifically, we reduce the
question of whether an un-negated complex sentence (a complex
sentence without ‘not’ at the front) is semantically correct to the
question of whether its constituents are semantically correct. Simi-
larly, we reduce the question of whether a negated complex sentence
(a complex sentence with ‘not’ at the front) is semantically correct to
the question of whether the negations of its constituents are seman-
tically correct. These rule are known as the inductive clauses of the
definition of semantic correctness.

1. ‘p and q’ is semantically correct =df both of ‘p’, ‘q’ are semantically

correct.
2. ‘p or q’ is semantically correct =df at least one of ‘p’, ‘q’ is
semantically correct.
 SEMANTIC 53

3. ‘not not p’ is semantically correct =df ‘p’ is semantically correct.

4. ‘not (p and q)’ is semantically correct =df at least one of ‘not p’,
‘not q’ is semantically correct.
5. ‘not (p or q)’ is semantically correct =df both of ‘not p’, ‘not q’
are semantically correct.

Consider, for example, sentence

(S) Neptune is blue, and it is not the case that either Mars is blue or
Jupiter is blue.

It follows from the base cases and inductive clauses that (S) is seman-
tically correct. First, by the negative base case, ‘Mars is not blue’
and ‘Jupiter is not blue’ are semantically correct. Thus, by inductive
clause 5, ‘it is not the case that either Mars is blue or Jupiter is blue’
is semantically correct. Also, by the positive base case, ‘Neptune is
blue’ is semantically correct. Consequently, by inductive clause 1, (S)
is semantically correct.
We now show that if ‘p’ and ‘q’ are atomic sentences then

(i) ‘p and q’ is semantically correct if and only if p and q

(ii) ‘p or q’ is semantically correct if and only if p or q.

Thus, like the correspondence theory, the semantic theory explains

the T-schema for conjunctions and disjunctions of atomic sentences,
in addition to atomic sentences and their negations.
3.1.2.2.1 Conjunctions of Atomic Sentences To take an arbitrary
example, consider the sentence ‘The Earth is round and Venice is
sinking’. ‘The Earth is round and Venice is sinking’ is semantically
correct if and only if ‘The Earth is round’ is semantically correct and
‘Venice is sinking’ is semantically correct. By the positive base case,
‘The Earth is round’ is semantically correct if and only if the Earth
is round, and ‘Venice is sinking’ is semantically correct if and only if
Venice is sinking. Consequently ‘The Earth is round and Venice is
sinking’ is semantically correct if and only if the Earth is round and
Venice is sinking.
54 SEMANTIC

3.1.2.2.2 Disjunctions of Atomic Sentences To take an arbitrary

example, consider the sentence ‘Chicago is windy or London is
windy’. ‘Chicago is windy or London is windy’ is semantically cor-
rect if and only if ‘Chicago is windy’ is semantically correct or
‘London is windy’ is semantically correct. By the positive base case,
‘Chicago is windy’ is semantically correct if and only if Chicago is
windy, and ‘London is windy’ is semantically correct if and only
if London is windy. Consequently, ‘Chicago is windy or London
is windy’ is semantically correct if and only if Chicago is windy or
London is windy.

3.2 MOTIVATION
In the previous chapter, we observed that the correspondence theory
has three nice features:

1. The correspondence theory explains the T-schema.

2. The correspondence theory entails that truth is meaning sensi-
tive (i.e., the correspondence theory entails that if a sentence had
meant something different then it would have had a different truth
condition).
3. The correspondence theory entails that truth is responsive
(i.e., the correspondence theory entails that truth is deter-
mined by how reality is configured, and not the other way
around).

These three features are nice because the T-schema, meaning sen-
sitivity and responsiveness collectively constitute our common sense
conception of truth.
However, we observed that the benefits of the correspondence the-
ory come at the cost of accepting the existence of facts, which form
an unusual category of theoretical entity. For example, according to
the theory of facts, for the dog to be in his bed there must exist three
entities: the dog, the dog’s bed and, in addition (this is the contro-
versial part), the fact of the dog being in his bed, which is a structured
entity composed of the dog, the dog’s bed and the abstract relation
of containment.
 SEMANTIC 55

Many philosophers are skeptical that facts exist. Consequently, it is

desirable to find an analysis of truth which has the three nice features
outlined above, but which is not committed to the existence of facts.
The semantic theory is one example of such an analysis. We have
already seen that the semantic theory explains the T-schema. In the
remainder of this section, we explain why the semantic theory entails
meaning sensitivity and responsiveness.

3.2.1 MEANING SENSITIVITY

According to the semantic theory, the truth condition of an atomic

sentence is determined by which entities the names and predicates
in the sentence refer to. For example, the truth condition of ‘Lon-
don is larger than Paris’ is the condition that London is larger than
Paris because ‘London’ refers to London, ‘Paris’ refers to Paris and
‘is larger than’ refers to the relation of being larger than. Conse-
quently, if ‘larger’ had instead meant smaller then, according to the
semantic theory, the truth condition of ‘London is larger than Paris’
would have instead been the condition that London is smaller than
Paris. Therefore, the semantic theory entails that truth for atomic
sentences is meaning sensitive – and a similar argument shows the
same for negations of atomic sentences.
By the definition of semantic correctness, the observation that
truth for atomic sentences and their negations is meaning sensitive
entails that truth for any complex sentence formed by repeatedly
applying ‘and’, ‘or’ and ‘not’ to atomic sentences is meaning sensi-
tive. Consider, for example, the complex sentence ‘London is larger
than Paris and Paris is larger than Florence’. According to the seman-
tic theory, the truth condition of this sentence is the conjunction of
the truth conditions of the atomic sentences ‘London is larger than
Paris’ and ‘Paris is larger than Florence’. Thus, since truth for atomic
sentences is meaning sensitive, the truth of ‘London is larger than
Paris and Paris is larger than Florence’ is also meaning sensitive.

3.2.2 RESPONSIVENESS

Consider the atomic sentence ‘Neptune is blue’. According to the

semantic theory, ‘Neptune is blue’ is true because
56 SEMANTIC

(i) ‘Neptune is blue’ is formed by applying ‘is blue’ to ‘Neptune’

(ii) ‘Neptune’ refers to Neptune and ‘is blue’ refers to the property
of being blue
(iii) Neptune has the property of being blue.
Since Neptune has the property of being blue because Neptune is blue,
‘Neptune is blue’ is true because Neptune is blue.
On the other hand, consider the atomic sentence ‘Mars is blue’.
According to the semantic theory, ‘Mars is blue’ is not true because
(i) ‘Mars is blue’ is formed by applying ‘is blue’ to ‘Mars’
(ii) ‘Mars’ refers to Mars and ‘is blue’ refers to the property of being
blue
(iii) Mars lacks the property of being blue.
Since Mars lacks the property of being blue because Mars is not blue,
‘Mars is blue’ fails to be true because Mars is not blue.
These observations generalize to all other atomic sentences. In
general, given the actual meaning of ‘t1 ,…, tn are R-related’, the
semantic theory entails that the truth status of ‘t1 ,…, tn are R-related’
is determined by whether t1 ,…,tn are R-related. In particular, ‘Scot-
land is colder than Monaco’ is true because Scotland is colder than
Monaco, ‘Venice is sinking’ is true because Venice is sinking and
‘Raleigh is closer to New York than to Washington DC’ is not true
because Raleigh is not closer to New York than to Washington DC.
Therefore, according to the semantic theory, truth for atomic sen-
tences is responsive – and a similar argument shows that truth for
negations of atomic sentences is also responsive.
As in the case of meaning sensitivity, the definition of semantic
correctness ensures the responsiveness of atomic sentences and their
negations is transmitted to any complex sentence formed by repeat-
edly applying ‘and’, ‘or’ and ‘not’ to atomic sentences. Consider,
for example, the complex truth ‘The Earth is round and Venice is
sinking’. By the semantic theory, ‘The Earth is round and Venice
is sinking’ is true because ‘The Earth is round’ is true and ‘Venice
is sinking’ is true. Since atomic truth is responsive, ‘The Earth is
round’ is true because the Earth is round and ‘Venice is sinking’ is true
because Venice is sinking. Therefore ‘The Earth is round and Venice
is sinking’ is true because the Earth is round and Venice is sinking.
 SEMANTIC 57

3.3 ARGUMENT FOR THE SEMANTIC ANSWER

We now turn to the question of how to argue for the semantic answer
to the central question. Our motivation for investigating the semantic
theory was the desire to find an analysis of truth which entails truth
is meaning sensitive and responsive, but which is not committed to
the existence of facts. This motivation provides what we consider to
be the central argument for the semantic answer:

(1) Truth is meaning sensitive.

(2) Truth is responsive.
(3) Facts do not exist.
(4) If truth is meaning sensitive, truth is responsive and facts do not
exist then truth is semantic correctness.

(5) Truth is semantic correctness. [From (1), (2), (3), (4)]

3.3.1 REASONS FOR BELIEVING THE PREMISES

3.3.1.1 Premises (1) and (2)

As discussed in Sections 2.2.1 and 2.2.2, meaning sensitivity and

responsiveness form part of our common sense conception of truth.
This entails premises (1) and (2), by default, have a high degree of
plausibility, which can only be undermined by compelling reasons
for disbelieving them.

3.3.1.2 Premise (3)

As discussed in Section 2.4.2.1, facts are weird. Whereas we are used to

the idea that ordinary material objects, like biological cells, wooden
planks and bricks, can compose larger entities, the idea that material
objects, like Mount Everest and K2, and abstract relations, like being
taller than, can compose larger entities is peculiar. Consequently, the
existence of facts is, on its face, implausible.

3.3.1.3 Premise (4)

We can argue for premise (4) as an inference to the best explanation.

To our knowledge, the semantic theory is the only known analysis of
58 SEMANTIC

truth which is consistent with the non-existence of facts, provides an

explanation of the T-schema and entails that truth is both meaning
sensitive and responsive. Thus, the semantic theory currently pro-
vides the best explanation of the T-schema, meaning sensitivity and
responsiveness under the assumption that facts do not exist.

3.4 EVALUATION
There are two items to evaluate: the central argument for the
semantic answer and the semantic answer itself.

3.4.1 EVALUATION OF THE ARGUMENT

Although premise (4) is highly plausible given our present state of

knowledge, premises (1), (2) and (3) are all open to question.

3.4.1.1 Premise (1)

In Section 2.4.1.2 we explained one reason to be skeptical of meaning

sensitivity. In brief: meaning sensitivity is doubtful because some of
the ways we would like to utilize the truth predicate require the truth
of ‘p’ to be necessarily equivalent to the condition that p.

3.4.1.2 Premise (2)

In Section 2.4.1.3 we explained one reason to be skeptical of respon-

siveness. In brief: responsiveness is doubtful because there are some
conditions, such as the condition that the Grand Canyon is beautiful,
which seem to obtain because of how we respond to reality, rather
than because of how reality is in itself. In these cases, truth is plau-
sibly involved in constructing the relevant part of reality, rather than
merely responding to reality.

3.4.1.3 Premise (3)

In Section 2.3.1.1 we saw the existence of facts can be motivated by

the principle that truth supervenes on being (TSB), according to which
there can be no difference in the state of reality without a differ-
ence in which entities exist. Facts are needed to support TSB due
 SEMANTIC 59

to the existence of possible situations which differ over how reality

is but contain all the same ordinary objects. Consider, for example,
a possible situation in which the mug is on the desk and a possible
situation in which the mug is under the desk. Both possibilities agree
on which ordinary objects exist (the mug, the desk and their various
parts). Therefore, the proponent of TSB needs to posit the existence
of unordinary entities — namely, the fact of the mug being on the
desk and the fact of the mug being under the desk — to make a
difference between these possibilities.

3.4.2 EVALUATION OF THE ANSWER

In this section, we consider three objections to the semantic answer,

which apply regardless of the argument used in support of it.

3.4.2.1 Generalizing Semantic Correctness

We showed how to define semantic correctness only for a restricted

fragment of English – the part obtained by repeatedly applying the
connectives ‘and’, ‘or’ and ‘not’ to atomic sentences. Thus, for the
semantic theory to offer a complete account of truth for English,
the definition of semantic correctness would need to be general-
ized to cover sentence formed using other kinds of connectives,
such as quantifiers (‘every’, ‘at least one’), modal operators (‘neces-
sarily’, ‘possibly’), propositional attitudes (‘Jane believes that’, ‘Kevin
hopes that’) and explanatory connectives (‘because’, ‘in virtue of ’). Some
progress has been made on this problem. In particular, Tarski showed
how to define truth in a semantic style for quantifiers and Kripke
showed how to define truth in a semantic style for modal operators.7
Despite this progress, it remains controversial how to define truth in
a semantic style for sentences containing propositional attitudes and
explanatory connectives.
3.4.2.1.1 A Possible Reply A proponent of the semantic theory may
respond that the theory is presently a work in progress, and point to the
successes of Tarski and Kripke as evidence the theory will eventually
be extended to cover all of English. Furthermore, there have been
proposals for how to semantically define truth for sentences con-
taining propositional attitudes and explanatory connectives, which
60 SEMANTIC

some linguists and philosophers find compelling, though these pro-

posals have not yet achieved the same level of consensus as Tarski and
Kripke.8

3.4.2.2 Wittgenstein’s Paradox

The semantic theory requires that, in a given context of use, every

name refers to a unique object and every n-place predicate refers to
a unique n-place relation.9 For example, according to the semantic
theory, ‘Paris is in France’ is true because (i) ‘Paris’ uniquely refers
to Paris, (ii) ‘France’ uniquely refers to France, (iii) ‘is in’ uniquely
refers to the 2-place relation of being located in and (iv) Paris and
France, in that order, stand in the relation of being located in.
These referential links between language and the world are
assumed by the semantic theory to be part of the meaning of English.
Consequently, whether a particular name refers to a particular object
or a particular n-place predicate refers to a particular n-place relation
is largely determined by the manner in which speakers of English
happen to use language. For example, given this understanding of
reference, ‘Paris’ refers to Paris not because of anything intrinsic
to the sequence of symbols P-a-r-i-s, but rather because of a long-
established linguistic convention, one that could have been different.
And if it had been different, ‘Paris’ might have instead referred to
Berlin, thereby giving ‘Paris is in France’ the truth condition that
Berlin is in France.
Wittgenstein’s paradox, originally due to Ludwig Wittengstein and
later elaborated by Saul Kripke, calls into question the possibility of
giving a plausible account of how the linguistic behavior of speakers
determines, in a given context of use, a unique assignment of names
and predicates to entities in the world.10 If no such account can be
given then it is doubtful whether the kind of language-world links
required by the semantic theory really exist. In this section, we briefly
outline Wittgenstein’s paradox and sketch one possible solution.
3.4.2.2.1 The Paradox Consider the monadic predicate ‘is a
human’. According to common sense, this predicate has, in the
recent past, always referred to the property of being a human. Sup-
pose a skeptic challenges this common sense view by suggesting that,
although ‘is a human’ now refers to the property of being a human,
 SEMANTIC 61

prior to the present moment ‘is a human’ instead referred to the prop-
erty of being one of the first 200 billion humans to ever live (call
this alternative property the property of being an early human). Thus,
according to the skeptic, yesterday everyone using the word ‘human’
actually meant early human. How can we refute the skeptic’s bizarre
hypothesis?
To refute the skeptic, we need to identify the aspect of our past
use of the predicate ‘is a human’ which ensured it referred to the
property of being a human, rather than to the property of being an
early human. Wittgenstein’s paradox is essentially the observation that
refuting the skeptic is a far more difficult task than we might naively
suppose.
The significance of this is as follows. If the skeptic cannot be
refuted, then it is false that ‘is a human’ previously singled out the
property of being a human. But since there are no relevant differences
between ‘is a human’ and other predicates, it would follow that none
of our predicates previously singled out any properties or relations
in particular. Consequently, since there are no relevant differences
between our past use of language and our present use of language,
none of our predicates right now single out any properties or relations
in particular, and so the referential relations posited by the semantic
theory simply do not exist.
We now discuss some possible ways of responding to the skeptic.
As a first attempt, we might say that, prior to the present moment, ‘is
a human’ referred to the property of being a human because speakers
of English only ever applied ‘is a human’ to humans and only ever dis-
applied ‘is a human’ to non-humans (a speaker disapplies a monadic
predicate to an object when the speaker describes the object as fail-
ing to satisfy the predicate). There are two main problems with this
account.
First, it is not actually the case that speakers have only ever applied
‘is a human’ to humans and disapplied ‘is a human’ to non-humans.
For example, there have surely been occasions where a speaker acci-
dentally applied ‘is a human’ to a manakin dressed as a human, or
disapplied ‘is a human’ to a human in costume.
Second, more importantly, since fewer than 200 billion humans
have ever lived, whenever a speaker in the past applied ‘is a human’
to a human, they also applied ‘is a human’ to an early human. In
62 SEMANTIC

addition, since every early human is also a human, whenever a speaker

disapplied ‘is a human’ to a non-human, they also disapplied ‘is a
human’ to a non-early-human. Consequently, our first response to
the skeptic is equally compatible with the skeptic’s hypothesis that,
prior to the present moment, ‘is a human’ referred to the property
of being an early human.
It is likely you did not find our first attempt at refuting the skeptic
very compelling. This is because, ordinarily, we would say that a
child has learned the meaning of ‘is a human’ not when they have
memorized a finite list of entities to which ‘is a human’ has previously
been applied or disapplied, but rather when they have internalized a
rule for using ‘is a human’. It is this rule – not a finite list of examples –
which enables the child to competently apply ‘is a human’ to new
cases.
Consequently, a more plausible response to the skeptic is that, prior
to the present moment, ‘is a human’ referred to the property of being
a human because correctly following the established rule for using ‘is
a human’ could only ever result in ‘is a human’ being applied to
humans. In particular, ‘is a human’ did not refer to the property of
being an early human because the rule for using ‘is a human’ allows
it to be applied to humans born after the early humans.
The problem with this account – and here lies the real bite of
Wittgenstein’s paradox – is that the rules for using language are them-
selves expressed in language. Consequently, just as the skeptic posits
a bizarre interpretation of ‘is a human’, they could posit an equally
bizarre interpretation of the rule for using ‘is a human’.
Suppose, for example, that the following rule for using ‘is a human’
is suggested:

(Rule For Using ‘is a human’) Apply ‘is a human’ to an entity if and
only if the entity belongs to the same
species as us.

The skeptic might respond that, although ‘is a species’ now refers to
the property of being a species, prior to the present moment ‘is a
species’ instead referred to the property of being the initial segment
of a species consisting of the first 200 billion members. Thus, the
rule for using ‘is a human’ is compatible with the skeptic’s original
 SEMANTIC 63

hypothesis that, prior to the present moment, ‘is a human’ referred

to the property of being an early human.
The proponent of the rule-based solution might respond by giving
a rule for using ‘is a species’ which excludes the skeptic’s deviant
interpretation. For instance:

(Rule For Using ‘is a species’) Apply ‘is a species’ to an entity if

and only if the entity is a natural
population of interbreeding organ-
isms which is reproductively isolated
from other such populations.11

But now the skeptic can simply assert that, although ‘is a natural pop-
ulation’ now refers to the property of being a natural population, prior
to the present moment ‘is a natural population’ instead referred to the
property of being the initial segment of a natural population consist-
ing of the first 200 billion members. So, once again, the rule-based
solution is made consistent with the skeptic’s deviant interpretation
of ‘is a human’.
The proponent of the rule-based solution could go on to give a
further rule for using ‘is a natural population’, but the skeptic will
also be able to give a deviant interpretation of that rule which accords
with their desired interpretation of ‘is a human’. Clearly, this process
can continue indefinitely. But nobody can internalize infinitely many
rules. So eventually we will end up at a rule so basic that the meanings
of the words in the rule cannot be fixed by any further rules. Once
the skeptic gives a deviant interpretation of this basic rule (which they
surely will be able to do), the proponent of the rule-based solution
will have no recourse to additional rules to respond to the skeptic.
Consequently, simply positing a collection of rules for using language
is not enough to determine a unique meaning for ‘is a human’ –
and so is not enough to determine a unique meaning for any of our
predicates.

3.4.2.2.2 A Possible Reply Although a full discussion of proposed

solutions to Wittgenstein’s paradox would take us beyond the scope of
this book, we will briefly sketch one fairly popular solution, known
as reference magnetism.12
64 SEMANTIC

Reconsider the initial proposal that, prior to the present moment,

‘is a human’ referred to the property of being a human because speak-
ers of English only ever applied ‘is a human’ to humans and only ever
disapplied ‘is a human’ to non-humans. The first problem with this
account is that speakers sometimes misapply ‘is a human’. Let us tem-
porarily set this problem aside and suppose that speakers really have
only ever applied ‘is a human’ to humans and disapplied ‘is a human’
to non-humans. There remains the problem that, since every previ-
ously existing human was also an early human, speakers have only
ever applied ‘is a human’ to early humans and disapplied ‘is a human’
to non-early-humans. Thus, according to this account, the property
of being an early human is an equally good candidate for the meaning
of ‘is a human’.
Here is a more precise characterization of the problem. Let HA
denote the set of entities to which speakers have previously applied
‘is a human’ and let HD denote the set of entities to which speakers
have previously disapplied ‘is a human’. By our simplifying assump-
tion, HA contains only humans and HD contains only non-humans.
Despite this, there are still multiple properties which are shared by all
members of HA but not exemplified by any member of HD – e.g., the
property of being one of the first 200 billion humans, the property
of being one of the first 201 billion humans, the property of being
one of the first 202 billion humans, and so on. Thus, an account of
reference which takes the referent of ‘is a human’ to be any prop-
erty which is shared by all members of HA but not exemplified by
any member of HD will fail to single out a unique referent for ‘is a
human’.
To fix this problem, an additional criterion is needed to filter out
the unwanted properties. The theory of reference magnetism pro-
vides us with such a criterion. This theory builds on the observation
that properties can be ordered according to their relative naturalness.
For example, the property being electrically charged is more natural than
the property being electrically charged prior to 10.31am on 03/17/1994.
The reason why is that 10.31am on 03/17/1994 is an arbitrary cut-off
point, which does not constitute a real joint in nature. Consequently,
whereas the predicate ‘is electrically charged’ is highly likely to occur
in a scientific textbook, the predicate ‘was electrically charged prior
 SEMANTIC 65

to 10.31am on 03/17/1994’ is highly unlikely to occur in a scientific

textbook.
According to reference magnetism, the referent of ‘is a human’
is the most natural property which is shared by all members of HA
but not exemplified by any member of HD . More generally, when
we allow that ‘is a human’ may have been misused (so that HA con-
tains some non-humans and HD contains some humans), reference
magnetism takes the referent of ‘is a human’ to be the most natural
property which best fits the way speakers have historically applied and
disapplied ‘is a human’.
It is plausible that the property of being a human is in fact the
most natural property which best fits the way speakers have histori-
cally applied and disapplied ‘is a human’. In particular, although the
property of being an early human is equally fitting, it is much less
natural than the property of being a human, since, to the best of
our knowledge, nothing special distinguishes the early humans from
later humans – and so the number 200 billion is an arbitrary cut-off
point. Thus, reference magnetism plausibly entails that the property
of being a human is the unique referent of ‘is a human’, thereby
refuting the skeptic.

3.4.2.3 Fictional Characters

Consider the sentence ‘Harry Potter is a wizard’. At first glance, this

sentence is true – Harry Potter is indeed a wizard in the eponymous
book series. Consequently, according to the semantic theory, ‘Harry
Potter is a wizard’ is true because ‘Harry Potter’ refers to some inde-
pendently existing object x, and x has the property of being a wizard.
Clearly, however, wizards do not actually exist. Therefore, there is
no object for ‘Harry Potter’ to refer to, and so the semantic theory
cannot be right. This, in short, is the fictional character objection to the
semantic theory.
3.4.2.3.1 A Possible Reply The fictional character objection, as it
stands, has a straightforward response. Although the sentence ‘In the
Harry Potter book series, Harry Potter is a wizard’ is unquestionably
true, it does not follow from this that the sentence ‘Harry Potter is
a wizard’ is true. Compare: although ‘In The Man in the High Castle,
66 SEMANTIC

the Allies lost WW2’ is true, ‘The Allies lost WW2’ is untrue. So
what happens inside a novel can differ from what happens outside the
novel. Consequently, proponents of the semantic theory can simply
reject the assumption that ‘Harry Potter is a wizard’ is true. This
sentence only seems true because it is true according to the novels.
3.4.2.3.2 A Stronger Objection Although the initial version of the
fictional character objection poses little threat to the semantic the-
ory, there is a modified version of the fictional character objection
which is much trickier to deal with. For even granting sentences
about Harry Potter inside the novels are untrue, there remain seem-
ingly true statements about Harry Potter which reside outside the
novels. Most notably, the sentence ‘Harry Potter is a fictional char-
acter’ seems to be true. But then, according to the semantic theory,
there independently exists an object x such that x has the property
of being a fictional character and ‘Harry Potter’ refers to x. Unfortu-
nately for the semantic theory, it is hard to see what kind of entity x
could possibly be. After all, we do not ordinarily bump into fictional
characters on the street.
Note that the response we gave to the initial fictional character
objection does not carry over to the modified fictional charac-
ter objection. This is because it is simply false that in the novels,
Harry Potter is a fictional character. Rather, according to the nov-
els, Harry Potter is a real flesh-and-blood wizard. Consequently,
we cannot explain away the appearance that ‘Harry Potter is a fic-
tional character’ is true on the basis that it is true according to the
novels.
Therefore it appears the semantic theory really is committed to the
independent existence of fictional characters. Whether this is palat-
able depends on whether a reasonable theory can be given as to the
nature of these mysterious entities.13

3.5 CHAPTER SUMMARY

In this chapter, we introduced the semantic theory of truth, accord-
ing to which truth consists in semantic correctness. We clarified the
nature of semantic correctness and of inductively defined proper-
ties more generally. We then outlined an argument for the semantic
 SEMANTIC 67

theory on the basis that, like the correspondence theory, it entails

that truth is meaning sensitive, responsive and satisfies the T-schema;
but, unlike the correspondence theory, is not committed to the exis-
tence of facts. Finally, we discussed some objections to this argument
and to the semantic theory more generally – in particular, the objec-
tion from Wittgenstein’s paradox and the objection from fictional
characters.

NOTES
6. An elementary mathematical technique can be used to formulate many inductive def-
initions in a non-circular manner. See Appendix C (page 159) for an application of this
technique to the definition of semantic correctness.
7. The mechanics of Tarski’s definition of truth – which applies to sentences containing
quantifiers (‘every’, ‘at least one’) – can be found in any introductory logic textbook, such
as Jc Beall and Shay Logan, Logic: The Basics (Routledge: New York, 2017). The mechanics
of Kripke’s definition of truth for sentences containing modal operators can be found in
any introductory modal logic textbook, such as Jc Beall and Bas van Fraassen, Possibilities
and Paradox (Oxford University Press: Oxford, 2003).
8. For a textbook-length introduction to the project of semantically defining truth
for English, see Richard Larson and Gabriel Segal, Knowledge and Meaning (MIT Press:
Cambridge, MA, 1995).
9. Recall from the discussion of context sensitivity in Section 1.3.9 that we assume a
context of use has been implicitly fixed in the background.
10. Our discussion of Wittgenstein’s paradox closely tracks Kripke’s account of the para-
dox in his book Wittgenstein on Rules and Private Language (Wiley: Hoboken, NJ: 1984).
It is controversial to what extent Kripke’s interpretation of Wittgenstein is historically
accurate. Wittgenstein’s original discussion of the paradox can be found in his classic text
Philosophical Investigations (Wiley-Blackwell: Hoboken, NJ, 2009). However, since Philo-
sophical Investigations is notoriously difficult to understand, we recommend starting with
Kripke’s book.
11. This definition is from Ernst Mayr, Systematics and the Origin of Species from the Viewpoint
of a Zoologist (Harvard University Press: Cambridge, MA, 1996).
12. The theory of reference magnetism is originally due to David Lewis, ‘New Work for
a Theory of Universals’, Australasian Journal of Philosophy, Vol. 61, No. 4, pp. 343–377,
1983.
13. For more discussion of the fictional character objection and related issues, see Saul
Kripke, Reference and Existence (Oxford University Press: Oxford, 2018).

FURTHER READING
• Historical background: Alfred Tarski, ‘The Semantic Conception
of Truth’, Philosophy and Phenomenological Research, Vol. 4, No, 3,
pp. 341–376, 1944.
68 SEMANTIC

• Classic work: Hartry Field, ‘Tarski’s Theory of Truth’, The Journal

of Philosophy, Vol. 69, No. 13, pp. 347–375, 1972.
• Contemporary discussion: Douglas Patterson (ed.), New Essays on
Tarski and Philosophy (Oxford University Press: Oxford, 2008).