Predicate Logic

Michael Franke

Formulas of predicate logic; predicate letters, variables & individ-

ual constants; domain of quantification; quantifier scope and binding;
atomic sentences; predicate-logical meaning of natural language sen-
tences; semantics of predicate logic; model, domain and interpretation
function; assignment and valuation functions; validity; predicate logic
with identity

1 Motivation

We can think of propositional logic as a system that formalizes the meaning

of important functional terms, namely the sentential connectives (negation,
conjunction, disjunction, implication). This carries a long way towards cap-
turing what sound logical inference is. But it also fails to capture crucial
patterns of inference. For example, if we know that

Alex is as tall as Bo

we also know that

Bo is as tall as Alex

in virtue of our knowledge of what the predicate “being as tall as” means.
Propositional logic cannot capture this. It can model the first sentences as p
(= “Alex is as tall as Bo”) and the second as q (= “Bo is as tall as Alex”), but
since we cannot look inside the structure of a simple proposition, there is no
way in which we can say that p, based on its internal form and relatedness to
the internal form of q, must necessarily entail q.1 1 We can, of course, make the additional

Predicate logic (PredLog) is an extension of PropLog which adds two premise by writing down that p → q, but
that clearly does not explain the general
things. Firstly, PredLog models the internal structure of atomic proposi- relationship.
tions in terms of predicates and individual constants. For example, we could
have a (two-place) predicate T meaning (“being as tall as”) and two sym-
bols, so-called individual constants, a and b which represent Alex and Bo
respectively. We can then translate the sentence “Alex is as tall as Bo” into
the formula T ab, which is a minimal truth-evaluable unit, but does have inter-
nal structure. Similarly, the sentence “Bo is as tall as Alex” would translate
into T ba.
Secondly, PredLog allows to express quantification, which is extremely
important in order to represent general rules, generalizations and key aspects
of our semantic and world knowledge.2 Imagine that you have a reasoning 2 Semantic knowledge is what we know

machine (computer, robot, friend . . . ) able to compute logical inferences about the meaning of words and expressions.
For example, semantic knowledge tells
in predicate logic. Even though we have not introduced any of the formal us that “being taller than” is a transitive
machinery (syntax, semantics, definition of validity, deduction system) nec- relation, and that “being as tall as” is an
equivalence relation. World knowledge is
essary to make such formal reasoning precise, imagine you give this machine what we know about the world. For exam-
ple, we know that Berlin is the capital of
Germany.
 predicate logic 2

the information T ab. You want it to represent “Alex is as tall as Bo,” but the
machine only has the string of symbols to work with. Would that machine be
able to conclude that T ba? No, it wouldn’t because it doesn’t know that you
want the symbol T to mean “being as tall as” and not “being taller than” or
anything else. But, using predicate logic, you can tell the system about the
fundamental structural properties of the relation “being as tall as,” such as
that it is symmetric. Concretely, you can express that “being as tall as” is a
symmetric predicate with the PredLog formula:

∀x ∀y (T xy → T yx)

which can be read as “for all objects x and y, if x stands in relation T to y,

then so does y to x.” This formula uses the quantifier ∀ to express a general-
ization: something that holds of any pair of objects. Generalizations of this
kind are essential for human reasoning and PredLog captures the most basic
aspects of quantification in a system of logical reasoning. To be clear, the
inference schema:

T ab, ∀x ∀y (T xy → T yx) / T ba

is logically valid in PredLog, but the schema:

T ab/T ba

is not.

2 The language of predicate logic

2.1 Basic ingredients of predicate-logical formulas

The formulas of PredLog consist:

• individual constants a, b, c, . . . , v • parentheses ( )

• sentential connectives ¬, ∧, ∨, →
• predicate letters A, B, C, D . . .
,↔

• variables w, x, y, z • quantifiers ∃, ∀

Individual constants are denoted by lower-case Roman letters (a, b, c, . . . , v)

up to v.3 Individual constants are like proper names: they refer to exactly one 3 If need be, we can also use indices like
a1 , a2 etc. This also holds for variables and
individual. For example, the individual constant a may be interpreted as re-
predicate letters.
ferring to Alex and b as referring to Bo. Individuals in the sense of predicate
logic need not be humans or animals. An individual is any kind of entity that
can have properties or stand in some kind of relation to any other property.
For example, constant m may denote a particular copy of Moby Dick.
Predicate letters are denoted with upper-case Roman letters (A, B, C, D . . . ).
Predicate letters will be used to denote properties or relations. Each predicate
 predicate logic 3

letter has a unique arity. The arity of a predicate letter is always an integer
bigger than zero. It gives the number of elements that the predicate letter
expects as an argument. A predicate letter with arity one is also called unary
predicate letter and will be interpreted to refer to a property. For example,
if B is a unary predicate letter denoting the property “x is a book”, then the
expression Bm can be interpreted as expressing that Moby Dick is a book.
A predicate letter with arity bigger than one will be interpreted to denote a
relation. For example, if the predicate letter L has arity two, it may stand for
a two-place relations such as “x loves y”. Consequently, we expect L to have
two arguments, so that Lab, Lam or Lmb would be well-formed expressions
(no matter whether true or meaningful), while Labm, Laaaa or Lb would not
be.
Variables are denoted by lower-case Roman letters (w, x, y, z), starting
from w. Variables are only interpretable in the scope of a quantifier, an impor-
tant technical concept we will introduce later. As a first intuitive guide, think
of variables as similar to pronouns4 which are used to refer to an unnamed 4 A proper logical treatment of pronouns

individual introduced by a quantifying expression like in these examples: in natural language requires more sophisti-
cated logical systems, like dynamic logics or
discourse representation theory.
For every boy it holds that he . . . [he = some boy]
There is a boy for which it holds that he . . . [he = some boy]

To build formulas, PredLog also uses parentheses and exactly the same
sentential connectives as PropLog does.
Quantifiers are special functional elements of the language of PredLog.
The quantifier ∃ is the existential quantifier. It is read as “there is” or “there
exists.” For example, the formula ∃x(Bx ∧ I x) would be read as “there is an
x such that x has property B and I.” It would mean that there is an individual
which has the property denoted by B (e.g., it is a book) and the property
denoted by I (e.g., it is interesting). In short, this formula would express
that there is at least one interesting book. The quantifier ∀ is the universal
quantifier. It is read as “for all,” “all” or “every.” For example, the formula
∀x(Bx → I x) would be read as “for all x it holds that if x has property B, then
it also has property I.” This would express that all books are interesting.

2.2 Formulas

The language L of PredLog is the set of all formulas which are recursively
defined as follows:

(i) If A is an n-ary predicate letter and if t1 , . . . , tn are individual constants

or variables, then At1 . . . tn is a formula.

(ii) If ϕ is a formula, then so is ¬ϕ.

(iii) If ϕ and ψ are formulas, so are:5 5 We allow ourselves to omit the outer-
most pair of parentheses as in PropLog.
 predicate logic 4

a. (ϕ ∧ ψ) b. (ϕ ∨ ψ) c. (ϕ → ψ) d. (ϕ ↔ ψ)
(iv) If ϕ is a formula and if x is a variable, then these are formulas:
a. ∀x ϕ [universal statement] b. ∃x ϕ [existential statement]
(v) Anything that cannot be constructed by (i)–(iv) is not a formula.
Here are examples of formulas of PredLog, together with intuitive para-
phrases based on the interpretation that a is Alex, b is Bo, m is the book
Moby Dick, Lxy means “x likes y,” Bx means “x is a book” and Oxy means
that “x owns y.”
Lam Alex likes Moby Dick.
Lab ∧ Lba Alex likes Bo and Bo likes Alex.
¬Oba Bo does not own Alex.
∃x(Bx ∧ Oax) Alex owns a book.
∀x ((Bx ∧ Obx) → Lax) Alex likes every book Bo owns.
¬∃x (Bx ∧ Oax) Alex does not own any books.
∀x (Bx → ¬Oax) Alex does not own any books.

2.3 Syntactic trees

The recursive definition for formulas of PredLog gives an internal structure

to each formula which we can represent using syntactic trees, just like for
PropLog. For PredLog the syntactic structure of a formula is particularly
important for the important concept of the scope of a quantifier and the
crucial notion of variable binding (see below). The syntactic structure, and
with it the scope of a quantifier, depends on proper use of parentheses.
To illustrate, assume that we want to explicate the logical structure of the
sentence:
Alex owns a book.
Here are two candidate formulas, of which the left one is correct, the right
one incorrect:
∃x (Bx ∧ Oax) ∃x Bx ∧ Oax
A paraphrase of the second (incorrect) formula is, roughly: “There is a book
and Alex owns them/ him/ her/ it/ you/ me/ us.” The point is that in the sec-
ond formula the x might but need not refer to the book, because it is not part
of the formula “dominated by ∃,” so to speak. This shows in the different
syntactic trees.
∃x (Bx ∧ Oax) ∃x Bx ∧ Oax

∃x (Bx ∧ Oax) ∃x Bx ∧ Oax

Bx ∧ Oax ∃x Bx
 predicate logic 5

2.4 Terminology

Just like in PropLog, formulas of PredLog can be named by their main

operator. For example, the following formulas can be called conjunctions:6 6 The second example here expresses the
idea that Alex owns a book, and that Alex
Oam ∧ Lam ∃x (Bx ∧ Oax) ∧ ∃x (Bx ∧ Lax) likes a book, where these could be the same
or different books.

A formula whose main operator is a universal quantifier can be called a uni-

versal formula or a universal statement; a formula whose main operator is
an existential quantifier can be called an existential formula or an existential
statement. For example, the formula ∃x (Bx ∧ Oax) from above is an existen-
tial statement, but the formula ∃x Bx ∧ Oax is a conjunction (as evidenced by
the syntactic trees given above).
If A is an n-ary predicate letter and if t1 , . . . , tn are individual constants,
then At1 . . . tn is an atomic sentence. Atomic sentences are minimal truth-
evaluable units of the language of PredLog, akin to the proposition letters of
PropLog.

3 Quantifier scope & binding

Not every well-formed formula of PredLog is interpretable. Consider the

formula Lax. We might paraphrase this as “Alex likes them/ him/ her/ it/ you/
me/ us.” Without knowing what x refers to, this formula —though a formula
of PredLog— is not interpretable. We therefore introduce terminology to
speak about which occurrences of variables are interpretable, which are not,
and how a variable that is interpretable is to be interpreted. The relevant tech-
nical terms are scope, as well as bound and free occurrences of a variable.
If ∀x ψ is a subformula of ϕ, then ψ is the scope of this occurrence of the
quantifier ∀x in ϕ. The same holds for ∃x. An occurrence of a variable x in
a formula ϕ (in a place where also an individual constant could appear, so
not the x in “∀x” and “∃x”), is free in ϕ if x is not in the scope of a quantifier
∀x or ∃x. If ∀x ψ (or ∃x ψ) is a subformula of ϕ and if an occurrence of x is
free in ψ, then this occurrence of x is bound by the quantifier ∀x (or ∃x). A
formula of predicate logic without any free occurrences of variables is called
a sentence.
Here are examples:7 7 The last formula is well-formed and
interpretable, but not very cooperative for an
Px x is free interpreter. In practice, we would rather like
to write ∃x( Px ∧ ∀y Qy)
Px ∧ ∀x Qx the first occurrence of x is free, the second bound
∃x( Px ∧ Qx) both occurrences of x are existentially bound
∃x Px ∧ ∀x Qx first occur. existentially bound, second universally bound
∃x( Px ∧ ∀ xQx) first occur. existentially bound, second universally bound
 predicate logic 6

4 Domain of quantification

Even if all variables in a formula are bound, in order to be able to interpret —

even if only intuitively— what a formula of PredLog could mean, we need
information about the domain of quantification D. Take the formula ∀x (Lxa)
with the interpretation of a and Lxy as before. We might take this to mean
that everybody likes Alex, or that everything on earth (including the book
Moby Dick) likes Alex. So, when we write down a formula with quantifiers
in PredLog, it will only be interpretable if we specify which individuals the
quantification should range over. We call this the domain of quantification D.
Remember that you must always specify the domain of quantification D in
translation exercises or other applications where your formulas are supposed
to be meaningfully interpretable.

5 Translations from natural langauge to PredLog

Just like PropLog, PredLog is useful for uncovering the logical structure of
sentences. Unlike PropLog, PredLog can lay bare the internal structure of
atomic propositions and aspects of quantification.
Suppose we want to translate this sentence to predicate logic:
Alex likes Bo but if Bo likes Alex, Bo likes everybody.
A formula that captures the logical structure of this sentence is:

Lab ∧ (Lba → ∀x Lbx)

Such a translation is only complete, strictly speaking, when we also explicitly

state the translation key, which defines what each individual constant and
predicate letter refers to, as well as the arity of each predicate letter. In the
example at hand, the translation key would be:8 8 Notice that the arity of the predicate
L is fixed by the notation Lxy and that it
(i) a: Alex is crucial for the translation key to specify
exactly what a predicate like L means, i.e., is
(ii) b: Bo first argument the slot for the person doing
or receiving the liking?
(iii) Lxy: x likes y
Since the sentence to translate includes the word “everybody,” the domain
of quantification should be the set of all human beings for the above formula
to be correct. If the domain of quantification also includes non-humans, we
would have to adapt the formula like so:

Lab ∧ (Lba → ∀x (Hx → Lbx))

and also include the predicate letter H in the translation key like so:
(iv) Hx: x is a human being
Let us consider a few examples. The domain of quantification D is the set
of all human beings.
 predicate logic 7

a: Alex Lxy: x likes y

b: Bo Px: x is a pilot

F x: x is friendly S xy: x and y are siblings

Here is a list with sentences and potential translations into PredLog.9 9 For some sentences, more than one
translation is given. These alternatives are
Alex is a pilot. Pa logically equivalent under the semantics of
PredLog which we will introduce later.
Bo is a friendly pilot. Fb ∧ Pb
No pilot is friendly. ∀x (Px → ¬F x)
¬∃x (Px ∧ F x)
Nobody likes pilots. ¬∃x ∃y (Py ∧ Lxy)
Bo has a friendly sibling. ∃x (F x ∧ S bx)
Every pilot has a friendly sibling. ∀x (Px → (∃y (Fy ∧ S xy)))
∀x ∃y (Px → (Fy ∧ S xy))
¬∃x (Px ∧ ¬∃y (Fy ∧ S xy))
¬∃x ¬∃y (Px → (Fy ∧ S xy))
 predicate logic 8

Exercise 1. For each of the following strings, determine whether they are
formulas of PredLog or not. Assume that P and Q are unary predicate let-
ters, and that R is a binary predicate letter.

(i) Px → ∃ x (v) Px ∨ ∃ xPx

(ii) ∀x(Px) (vi) ∀yPx ∨ ∃ xPx

(iii) ∀xPx (vii) ∀y(Rxy ∨ ∃ xPx)

(iv) (∀xPx) (viii) ∀y(Rxy ∨ ∃ xPx)

Exercise 2. For each of the following formulas of predicate logic, determine

whether each occurrence of a variable is a free or bound occurrence. If it is a
bound occurrence, determine which quantifier binds it.

(i) Px (iv) ∃x Px ∧ Lx j

(ii) ∃x Lx j (v) ∃x (Px ∧ Lx j)

(iii) ∃x Lxy (vi) ∃x (Px ∧ ∀x Lx j)

Exercise 3. Translate the following sentences into the language of predicate

logic. Preserve as much of the logical structure as possible and give the trans-
lation key and the domain of quantification (here: D : people).

(i) Everybody is friendly.

(ii) Everybody loves somebody.

(iii) Every pilot loves Bill.

(iv) If Mary is a pilot, someone loves her.

(v) Every pilot is unfriendly.

(vi) Some pilots are friendly.

(vii) No pilot is friendly.

(viii) Nobody loves anyone who is in love with a pilot.

 predicate logic 9

6 Semantics of predicate logic

The semantics of propositional logic was formulated in terms of valuation

functions, which mapped each formula to a truth value, true or false. We
thought of these as possible worlds, or possible states the world could be in.
The semantics of PredLog is slightly more complicated because in order
to say which formulas are true or false, we need to interpret not just atomic
propositions, but their internal structure, so to speak; we need to interpret all
terms (individual constants and variables) and the predicate letters. This is
done in a so-called model, which can similarly be thought of as a possible
world. Given a model, we can derive a valuation function. The model fixes
the domain (which individuals are there) and the meaning of the basic vo-
cabulary of our language, i.e., the meaning of all individual constants and all
predicate letters.

6.1 Model, domain & interpretation function

A model M = hD, Ii for language L consists of:

a domain D , ∅, and an interpretation function I.

The domain D is a set of entities.10 The interpretation function I maps each 10 We will always assume that the domain

element of C (the set of constants of language L) and P (the set of all predi- has at least one element.

cate letters of L) to a suitable relation on D:

if c is an individual constant, then I(c) ∈ D, and

if P is an n-ary predicate letter, then I(P) ⊆ Dn .11 11 Dn is the set of all n-tuples with
elements from D.
The interpretation I(c) of a constant c is the entity in the domain that c stands
for (e.g., its name or identifier). The interpretation I(P) of an n-ary predicate
letter, is the set of all n-tuples which stand in the relation P to each other. For
example, if R is the one-place predicate letter denoting the property “red,”
then I(R) is the set of all red things (in the possible world we are modeling
with the help of I). Or, if L is the two-place predicate letter denoting the rela-
tion “x likes y,” then I(L) is the set of all pairs of elements from the domain,
e.g., hd1 , d2 i, such that the first likes the second (in the possible world we are
modeling with the help of I).

6.2 Assignment functions

A model M = hD, Ii interprets individual constants and predicate letters.
But we also need a way to assign meaning to variables. This needs to take
quantifier binding into account. For this purpose we can use assignment
functions. Assignment functions and the interpretation function together
allow us to interpret all terms (variables and individual constants).
 predicate logic 10

An assignment function g for model M = hD, Ii and language L maps all

variables in L to elements in D. We write g[x/d] and read “g with x mapped
to d”, for the assignment function which is like g except that x is mapped
to d ∈ D. We will use this latter construct, g[x/d] to define the semantics of
existential and universal formulas (see below).12 12 If needed, this notation can also be

If t is a term of L, then [[t]] M,g is the term interpretation relative to M = extended to cover more than one variable,
like so: g[x/d1 ,y/d2 ,z/d3 ] .
hD, Ii and g:

[[t]] M,g = I(t) if t is a constant, and [[t]] M,g = g(t) otherwise.

6.3 Valuation functions

We now can use a model M = hD, Ii and an assignment function g to define
a valuation function V M,g , which uses M and g, to give a truth value for every
formula of predicate logic. Here is the full definition, which we will go
through systematically below.
D E
V M,g (At1 . . . tn ) = 1 iff [[t1 ]] M,g , . . . , [[t1 ]] M,g ∈ I(A)
V M,g (¬φ) = 1 iff V M,g (φ) = 0
V M,g (φ ∧ ψ) = 1 iff V M,g (φ) = 1 and V M,g (ψ) = 1
V M,g (φ ∨ ψ) = 1 iff V M,g (φ) = 1 or V M,g (ψ) = 1
V M,g (φ → ψ) = 0 iff V M,g (φ) = 1 and V M,g (ψ) = 0
V M,g (φ ↔ ψ) = 1 iff V M,g (φ) = V M,g (ψ)
V M,g (∀x φ) = 1 iff V M,g[x/d] (φ) = 1 for all d ∈ D
V M,g (∃x φ) = 1 iff V M,g[x/d] (φ) = 1 for at least one d ∈ D
For the most basic kind of formula, built from an n-ary predicate letter
A with n terms t1 , . . . , tn (where each ti is either a variable or an individual
constant), the formula At1 . . . tn is true, given M = hD, Ii and g, if the n-
tuple of entities picked out by the interpretation of all terms, using I and g, is
contained in the interpretation of A.
D E
V M,g (At1 . . . tn ) = 1 iff [[t1 ]] M,g , . . . , [[tn ]] M,g ∈ I(A)
For sentential connectives, valuation functions work exactly like in Pro-
pLog.
V M,g (¬φ) = 1 iff V M,g (φ) = 0
V M,g (φ ∧ ψ) = 1 iff V M,g (φ) = 1 and V M,g (ψ) = 1
V M,g (φ ∨ ψ) = 1 iff V M,g (φ) = 1 or V M,g (ψ) = 1
V M,g (φ → ψ) = 0 iff V M,g (φ) = 1 and V M,g (ψ) = 0
V M,g (φ ↔ ψ) = 1 iff V M,g (φ) = V M,g (ψ)
Finally, to assign a truth-value to existential or universal formulas, we
use the construction g[x/d] , which represents an assignment function which
is exactly like g but with a special, possibly different interpretation for the
variable x.
V M,g (∀x φ) = 1 iff V M,g[x/d] (φ) = 1 for all d ∈ D
V M,g (∃x φ) = 1 iff V M,g[x/d] (φ) = 1 for at least one d ∈ D
 predicate logic 11

7 Truth in a model

Assignment functions are just auxiliary constructs. At the end of the day, we
are interested in whether a formula is true in a possible world, i.e., a model
M = hD, Ii, no matter what assignment function is used.
If V M,g (ϕ) = 1 for all g, we write V M (ϕ) = 1 and say “ϕ is true in M.” If
V M,g (ϕ) = 0 for all g, we write V M (ϕ) = 0 and say “ϕ is false in M.”
Notice that any sentence of PredLog (i.e., any formula without free vari-
ables) is either true or false in any given model M. In other words, formulas
without free occurrences of variables are interpretable just with respect to a
model, and these formulas are —usually— what we care about.
Let’s consider a simple example of a model and the way it makes certain
sentences true or false. The language we are interested in has just one indi-
vidual constant a, two one-place predicate letters, P and Q, as well as one
two-place predicate letter R. An example model M = hD, Ii for this language,
fixing a domain and an interpretation of all basic vocabulary, can be written
out like this:

D = {1, 2, 3, 4, 5}
I(a) = 3
I(P) = {1, 2, 5}
I(Q) = {2, 5}
I(R) = {h1, 2i , h2, 2i , h1, 4i , h1, 5i}

For better visual grasp, small models can often be represented economi-
cally as a graph. Figure 1 is an example for the model at hand. The two-place
relation is represented in terms of arrows, small letters ornamenting the el-
ements of the domain show the interpretation of individual constants and
one-place predicate letters.

Figure 1: Example of a graphical

representation of a simple model
for predicate logic.

A graphical representation like in Figure 1 greatly facilitates checking

whether a model makes certain sentences true or false. For example, in order
to decide whether a formula like ∃x (Px ∧ Qx) is true or false in this model,
we need to check whether there is an element that has both property P and
property Q. Looking at Figure 1, we see that this is true. Elements 2 and
5 both have both properties. To write this down in a formally correct way
 predicate logic 12

with reference to the model’s mathematical structure and the definition of the
semantics of PredLog, we would say that V M (∃x (Px ∧ Qx)) = 1 because,
no matter which g we use, there exists an element in D, namely 2, such that
V M,g[x/2] (Px ∧ Qx) = 1. This is because 2 ∈ I(P) and 2 ∈ I(Q).
Take another example. What is the truth value that the model assigns to
formula Pa? Well, that formula is false in the given model. The individual
referred to by constant a is 3, but 3 does not have property P. In formal
terms, we would then say that V M (Pa) = 0, because, for any valuation
function g, since I(a) = 3 and 3 < I(P), we have V M,g (P(a)) = 0.

8 Validity

Validity is defined in the same way in PredLog as it was in PropLog.

If ϕ1 , . . . , ϕn and ψ are sentences of language L of predicate logic, ϕ1 , . . . , ϕn /ψ
is an argument schema. The argument schema ϕ1 , ϕ2 , . . . , ϕn /ψ is valid iff for
all models M of L such that V M (ϕ1 ) = V M (ϕ2 ) = · · · = V M (ϕn ) = 1 it also
holds that V M (ψ) = 1. If valid, we write ϕ1 , ϕ2 , . . . , ϕn |= ψ.
In order to demonstrate that an argument schema is not valid, it suffices to
give a single counterexample. In order to demonstrate that a given argument
schema is valid, we need to give an informal proof, since we must reason
over all models. The following gives examples for each.

8.1 Demonstrating invalidity with a counterexample

Claim 1. ∃x Px 6|= ∀x Px13 13 Intuitively, just because someone is a

pilot, doesn’t mean that everybody is a pilot.
Proof. A minimal counterexample for this case is a model M with domain
D = {1, 2} and interpretation function I(P) = {1}. It is clear that there is an
element d ∈ D, namely 1, such that V M,g[x/1] (Px) is true, so that the premise
∃x Px is true in M. But there is also an element d ∈ D, namely 2, such that
V M,g[x/2] Px is false, so that the conclusion ∀x Px is false in M. 

8.2 Demonstrating validity with an informal proof

Here are two examples for informal proofs demonstrating validity of an

argument schema.

Claim 2. ∀x Px |= ¬∃x ¬Px14 14 Intuitively, if everybody is a pilot, there

cannot be anyone who is not a pilot.
Proof. We need to show that for all models M, if V M (∀x Px) = 1, then
V M (¬∃x ¬Px) = 1. So suppose that we have an arbitrary model M = hD, Ii,
such that V M (∀x Px) = 1. This entails that for all d ∈ D, d ∈ I(P). Suppose
further, towards contradiction, using an indirect proof strategy, that for the
same M we have V M (¬∃x ¬Px) = 0. The latter means that V M (∃x ¬Px) = 1,
which is the case when there is an element d ∈ D such that d < I(P). But
that contradicts the inference we drew from our premise that for all d ∈ D,
 predicate logic 13

d ∈ I(P). Consequently, there cannot be a model M which makes the premise

true and the conclusion false. The argument schema must be valid. 

Claim 3. ¬∀x (Px → F x) |= ∃x Px15 15 Intuitively, if it is not the case that every
pilot is friendly, there has to be at least one
Proof. Assume that the premise ¬∀x (Px → F x) is true in a model M = pilot.

hD, Ii. From V M (¬∀x (Px → F x)) = 1 we know that V M (∀x (Px → F x)) = 0.
But if ∀x (Px → F x) is false in M, it follows that there is at least one element
d ∈ D such that V M,g[x/d] (Px → F x) = 0, for arbitrary g. We know that
an implication is only false if the antecedent is true and the consequent is
false (from propositional logic, definition of →). That means that it must
be the case that V M,g[x/d] (Px) = 1. This entails that V M,g (∃x Px) = 1, by
definition of truth of existential statements. But since g was completely
arbitrary and nothing hinges on any specifics of g, we have actually shown
that V M (∃x Px) = 1. 

9 Equivalences, tautologies and contradictions

The notions of logical equivalence, tautology and contradiction for PredLog

are all analogous to their counterparts in PropLog.
Two formulas ϕ and ψ of PredLog are logically equivalent iff for all
models M we have V M (ϕ) = V M (ψ). A formula ϕ of PredLog is a tautology
iff for all models M we have V M (ϕ) = 1. It is a contradiction iff for all models
M we have V M (ϕ) = 0.
In order to demonstrate that two formulas are logically equivalent, or that
a formula is a tautology or contradiction, we can give informal proofs. For
example, the previous section showed that ∀x Px |= ¬∃x ¬Px. In fact, ∀x Px
and ¬∃x ¬Px are logically equivalent. Here is an informal proof for this:

Claim 4. ∀x Px and ¬∃x ¬Px are logically equivalent.

Proof. Let M be a model in which ¬∃x ¬Px is true. This means that there is
not a single element d ∈ D, such that d < I(P). In other words, I(P) = D,
i.e., all elements in the domain have property P. So M also makes ∀x Px
true. The reverse direction has already been shown when proving ∀x Px |=
¬∃x ¬Px. 
 predicate logic 14

Exercise 4. Use the model from the example in Figure 1 to check the truth
value of the following formulas.

1. ¬Qa 4. ∃x∃y (Qx ∧ Py ∧ Rxy)

2. ∀x (Qx → Px) 5. ∀x∃y Rxy

3. ¬∃x Rxx 6. ∀x∀y (Rxy → (Px ∨ Qy))

Exercise 5. Consider a model M = hD, Ii for a language with two unary

predicate letters P and Q and a two-place predicate letter R, given as follows.

D = {1, 2, 3, 4, 5}
I(P) = {1, 2, 5}
I(Q) = {2, 5}
I(R) = {h1, 2i , h2, 2i , h1, 4i , h1, 5i}

Evaluate for each of the following formulas whether they are true or false in
the model. (Hint: It is highly recommended that you draw a graphical repre-
sentation for yourself similar to Figure 1, even if that is not required to obtain
points and/or a correct solution.)

1. ∃x (Px ∧ Qx) 4. ∃x∃y (Qx ∧ Py ∧ Rxy)

2. ∀x (Qx → Px) 5. ∀x∃y Rxy

3. ¬∃x Rxx 6. ∀x∀y (Rxy → (Px ∨ Qy))

Exercise 6. Give informal proofs or counterexamples for these claims:

(i) ∀x (Ax ∨ Bx ∨ Cx), ∀x (Ax → Bx), ¬∃Bx |= ∃x Cx

(ii) ∃x Ax, ∃x Bx |= ∃x (Ax ∧ Bx)

(iii) ∀x∀y∀z ((Rxy ∧ Ryz) → Rxz), ∃x∃y (x , y ∧ Rxy ∧ Ryx) / ∃x Rxx

(iv) ∀x (Ax ∨ Bx) |= ∀x Ax ∨ ∀x Bx

(v) ∀x (Ax ↔ Rxx), ∃x∀y (Ryy ↔ x = y) |= ∃x Ax

(vi) ∃x∀y Rxy, ∀x∀y (Rxy → Ryx) |= ∀x ∃y Rxy

(vii) ∀x (Ax → Bx), ∃x Bx |= ¬∃x Ax

 predicate logic 15

10 Predicate logic with identity

Suppose we want to translate the meaning of the following sentence into
predicate logic:

Only Alex is a pilot.

We can write Pa (with the obvious translation key), expressing that Alex is
a pilot. But we also want to express that nobody else is. We can express that
nobody is a pilot with a formula like ∀x ¬Px or, equivalently (see Claim 4),
¬∃ Px. But that’s not actually true because Alex is. So writing Pa ∧ ∀x ¬Px
is a contradiction that does not express the meaning of “Only Alex is a pilot.”
Indeed, PredLog as we defined it so far is not expressive enough to capture
the meaning of this sentence. We need to add something. We will add a
special predicate: identity.

10.1 Adding identity to PredLog

We change clause (i) of the definition of formulas of predicate logic, given in

Section 2.2, to also allow the use the special two-place predicate symbols “=”
and “,” for which we use infix notation.

(i) If A is an n-ary predicate letter and if t1 , . . . , tn are terms, then At1 . . . tn

is a formula. If t1 and t2 are terms, then t1 = t2 and t1 , t2 are
formulas.

These are special symbols also in the semantics, because they are inter-
preted in the same manner in every model.16 Concretely, we just add the 16 Identity is the same concept in all pos-

following clause to the definition of the semantics of predicate logic:17 sible worlds. It is a logical concept that has
a fixed meaning in all (logically) possible
V M,g (t1 = t1 ) = 1 iff [[t1 ]] = [[t2 ]] worlds.
17 Notice that there are three occurrences
V M,g (t1 , t1 ) = 1 iff [[t1 ]] , [[t2 ]] of the symbol “=” in the first line of this
definition. The first occurrence refers to the
two-place predicate letter of PredLog with
10.2 Translations with PredLog with identity identity. The latter two refer to our usual
understanding of “=” as identity. Basically,
Using identity in the logical language, we can express the meaning of this is a complicated way of saying that “=”
is supposed to mean the same same thing in
Only Alex is a pilot. PredLog as what it means outside of it.

in terms of the formula:

Pa ∧ ∀x (x , a → ¬Px)

which can be read as: “Alex is a pilot, and everybody who is not Alex is not a
pilot.”18 18 This is not the only paraphrase and

Here are further examples for translations, making use of identity:19 formula we can give to this example. See
the first example in the list that follows, for
instance.
(i) Alex is friendly but nobody else is. Fa ∧ ∀x (F x → x = a) 19 Translation key as before.

(ii) At least two people are friendly. ∃x∃y (F x ∧ Fy ∧ x , y)20 20 We omit parentheses here, even though
they would be strictly speaking required
by the definition of a formula. Instead of
writing ϕ ∧ (ψ ∧ χ) or (ϕ ∧ ψ) ∧ χ, given that
these two formulas are logically equivalent
(in PropLog and in PredLog), we simply
write ϕ ∧ ψ ∧ χ.
 predicate logic 16

(iii) Alex likes only pilots, except for Bo. ∀x (Lax → (Px ∨ x = b))

(iv) Exactly two people are pilots.

∃x∃y (x , y ∧ Px ∧ Py ∧ ∀z (Pz → (z = x ∨ z = y)))

(v) Bo likes at most two people.

∀x∀y (((Lbx ∧ Lby) ∧ x , y) → ∀z (Lbz → (z = x ∨ z = y)))

10.3 Translating modified numerals

Translations of quantified expressions like “at most two people,” as in the

last example, can be difficult. For example, ask yourself: Is the following
translation adequate for the last example of “Bo likes at most two people?”

(v) Bo likes at most two people.

∀x∀y ((Lbx ∧ Lby) → ∀z (Lbz → (z = x ∨ z = y)))

No it is not. This is because that last formula is false in a world in which

Bo loves two people. Suppose Bo loves Alex and Charley, and that Alex,
Bo and Charley are three different people. Let x and y refer to Alex. Then
the antecedent (Lbx ∧ Lby) becomes true. Let z refer to Charley. Bo loves
Charley, but Charley is not identical to Alex (by assumption).
Expressions like “at most two” or “exactly one” are called modified nu-
merals. Here are some templates for translating these kinds of expressions
into predicate logic with identity. The following are example translations for
modified numerals applied to a unary predicate P.

at least one ∃x Px
at least two ∃x ∃y (Px ∧ Py ∧ x , y)
exactly one ∃x (Px ∧ ∀y (Py → x = y))
exactly two ∃x ∃y (x , y ∧ Px ∧ Py ∧ ∀z (Pz → (z = x ∨ z = y)))
at most one ∀x ∀y ((Px ∧ Py) → x = y)
at most two ∀x ∀y ((Px ∧ Py ∧ x , y) → ∀z (Pz → (z = x ∨ z = y)))

Exercise 7. Translate the following sentences into predicate logic with iden-
tity.

(i) Only Alex and Bo are friendly.

(ii) If Alex is friendly, nobody else is.

(iii) Alex is the only friendly pilot.

(iv) There is a pilot but it is not Alex.

(v) There is at most one pilot.

(vi) Alex likes nobody but Bo and Alex.