You are on page 1of 7

Phil 320

Problem Set 4: Solutions


Chapter 12
12.1

[2 marks: 12.1, 12.2, 12.10, 12.15, Extra #1, #2, #3= 14 marks total]

Part 1: First we show that for any even positive integer n, we can find a model M
of size n. If n = 2k, let the domain |M | be {1, 2, , 2k}. For the denotations:
PM(m) m is even
fM(m) = {m+1 if m is odd
{m-1 if m is even.
(So fM(1) = 2, fM(2) = 1; fM(3) = 4, fM(4) = 3; and so forth.)
The first two conjuncts of A assert that fM must be one-to-one and onto, and it is.
The third conjunct asserts that whenever fM(m) = n, PM(m) iff not PM(n), which is
also true because fM takes even numbers to odd numbers and odd numbers to even
numbers. Hence, M is a model of A.
So: every even positive integer is in the spectrum of A.
Part 2: We need to show that any finite model of A must have an even number of
elements. So suppose M is a finite model of A. Let
P+ = all m in |M| such that PM(m)
P- = all m in |M| such that not PM(m).
|M| is just the union of P+ and P-. So if P+ and P- have the same number of
elements, k, then |M| has 2k elements, and 2k is even.
But fM is a one-to-one correspondence between P+ and P-, because fM is one-toone, and it maps each member of P+ to a member of P-. Furthermore, the function
is onto because f is onto and if n is in P- and fM(m) = n, then by the third sentence
of A we must have PM(m) and so m is in P+. So P+ and P- have the same number
of elements. That proves that the size of the domain has to be an even number.
So: no odd positive integer is in the spectrum of A.

12.2

Imitate 12.1. Let the language include a one-place function symbol f, and oneplace predicates P1, P2 and P3. Let A be the sentence with the very same first two
conjuncts as in 12.1, and two additional conjuncts:
(1)
(2)

x[(P1x & ~P2x & ~P3x)(~P1x & P2x & ~P3x)(~P1x & ~P2x & P3x)]&
xy[f(x) = y [(P1x P2y) & (P2x P3y) & (P3x P1y)]]

The idea is this. (1) tells us that each object in the domain has exactly one of the
properties P1, P2 and P3. Conjunct (2) tells us that applying f moves around
among the three possibilities, just as the function f of 12.1 toggled between two
possibilities.

2
Part 1: If n = 3p, we have to find a model M with 3p objects. Let the domain of
M be {1, 2, , n}. For the denotations:
P1M(m) m is of the form 3k+1
P2M(m) m is of the form 3k+2
P3M(m) m is of the form 3k
fM(m) =

{m+1 if m is of form 3k+1 or 3k+2


{m-2 if m is of form 3k.

Then fM is 1-1, onto and satisfies the properties required in conjunct (2).
Conjunct (1) is satisfied as well. So A has a model with domain of size n,
for every n that is an exact multiple of 3.
Part 2: If M is a finite model, we have to show that the number of elements must
be an exact multiple of 3. If D is the domain of M, set
D1 = all m such that P1M(m)
D2 = all m such that P2M(m)
D3 = all m such that P3M(m).
By (1), these three sets are distinct and each element of D is in exactly one
of them. Next, fM is a 1-to-1 mapping from D1 onto D2, from D2 onto D3,
and from D3 onto D1. (For instance: it is clear from conjunct (2) that if m
is in D1, then fM(m) is in D2. And for any n in D2, n = fM(m) for some m in
D because f M is onto, and again conjunct (2) tells us that m must be in
D1.) So each of D1, D2 and D3 has the same number of elements, say k,
proving that the domain D has 3k elements.
12.10 (a) Imitating the sentences In in Example 12.1, try
Bn: x1 x2xn-1 xn[~(xn x1) & ~(xn x2) & & ~(xn xn-1)]
The argument of Example 12.1 may be adapted to show that there are at least n
equivalence classes for the equivalence relation E denoted by .
Suppose for a contradiction that there were only n-1 or fewer equivalence classes.
Pick objects m1,,mn-1 so that there is at least one from each of the equivalence
classes i.e., every object in the domain is in the same equivalence class as at
least one of m1,,mn-1. Then if Bn is to be true, there must be an mn in the domain
such that NOT mn E m1, and NOT mn E mn-1, which is a contradiction.
Note: Another good choice is
Bn: x1 x2 xn [~(x1 x2) & & ~(xn-1 xn)].
(b) Fn(x): x1 x2xn-1 xn [[(x1 x) & & (xn-1 x)]
[(xn x1) & & (xn xn-1) & (xn x)]]

3
Suppose a satisfies Fn(x), and suppose for a contradiction that the equivalence
class of a contains only n-1 or fewer elements. Let m1, , mn-1 be a complete list
of members in the equivalence class of a (with repetition if the class contains
fewer than n-1 elements). Since a satisfies Fn(x), putting mi for xi implies that
there must be an mn in the domain with mn m1, , mn mn-1 and yet mn E a, a
contradiction.
Note: Another good choice is
Fn(x): x1 x2 xn [[(x1 x) & & (xn x)] & [(x1 x2) & & (xn-1 xn)]]
Other sentences will work as well.
12.15 (a) is unsatisfiable iff some finite subset {C1,, Cm} of is unsatisfiable
iff every interpretation makes at least one of C1,,Cm false
iff ~C1 ~Cm is valid
(the first iff follows from Compactness).
(b) D is a consequence of
iff {~D} is unsatisfiable
iff {~D} is unsatisfiable for some finite subset of (by
Compactness)
iff D is a consequence of for some finite subset of .
(c) D is a consequence of
iff {~D} is unsatisfiable
iff for some C1,,Cm in , ~C1 ~Cm ~~D is valid by part (a)
iff for some C1,,Cm in , ~C1 ~Cm D is valid
Extra:
1. This question has two parts (since it is an if and only if claim).
Part 1
First suppose is complete. Assume that {~(A B)} is unsatisfiable. We have to
show that either {~A} or {~B} is unsatisfiable.
Suppose, for a contradiction, that both {~A} and {~B} are satisfiable.
Since {~A} is satisfiable, there is an interpretation M such that M and
M ~A. But M A B, for otherwise {~(A B)} would be satisfiable. It
follows that M B. But this means {B} is satisfiable. By completeness, we
must then have {~B} unsatisfiable, which is a contradiction.
Part 2
Now suppose that has the property that for every A and B, if {~(A B)} is
unsatisfiable, then either {~A} or {~B} is unsatisfiable. We must show
that for any A, either {A} or {~A} is unsatisfiable.

4
This follows from the fact that for any A, {~(A ~A)} is unsatisfiable: for
no interpretation can make true the sentence ~(A ~A). It must be, then, that
either {~A} or {~~A} is unsatisfiable, and the second of these is
equivalent to {A} is unsatisfiable.
2.

The sentence S says: if A is a reflexive, transitive and total relation (xAy or yAx
for all x and y), then there is a first y such that for all x, yAx.
Part 1
S is false for the interpretation with domain Z (the integers), where A means .
This relation is reflexive, transitive, and satisfies the third part of the antecedent
as well since m n or n m for any two integers m and n. But the consequent is
false, since there is no smallest integer m, i.e., no m such that for all integers n we
have m n.
Alternative solution: domain is N, and A means . Again, the relation satisfies
the antecedent, but not the consequent because there is no largest natural number.
Part 2
S is true in every finite domain. Suppose the interpretation M has finite domain
{d1, , dn}, and suppose the antecedent of S is true in M. This means:
(1) Reflexive. AM(d, d) for every d;
(2) Transitive. AM(d, d) whenever AM(d, d) and AM(d, d), for every d,
d, d;
(3) Total. Either AM(d, d) or AM(d, d) for any d, d.
Lets say d is A-smaller than d if AM(d, d). What we have to show is that there
is some d in the domain such that AM(d, di) for all i from 1 to n, i.e., that the
domain has an A-smallest element.
Define d in the following informal way: proceeding from left to right in the list
d1, d2, , dn, we keep track of the A-smallest member of the list up to step k.
At step 1, this will be d1; at step 2, it will be whichever of d1 and d2 is Asmaller, and so forth. By the time we get to the end, well have an A-smallest
member for the whole domain; if there are two or more tied for smallest, it
doesnt matter.
More formally, let a1 [the A-smallest at step 1] be d1, and let ak+1 [the A-smallest
at step k+1] be the A-smaller of ak and dk+1. That is, by (2), either AM(ak, dk+1)
[ak is A-smaller than dk+1] or not AM(ak, dk+1) [dk+1 is A-smaller than ak]. In
the former case, let ak+1 = ak; in the latter case, let ak+1 = dk+1. By (1) and (2), ak+1
is A-smaller than d1, , dk+1. By the time we come to the end of the list, we
have a member of the domain, an, which is A-smaller than everything on the
list.

3.

Since 1 2 is unsatisfiable, by Compactness it follows that 1 2 is

5
unsatisfiable for some finite subsets 1 of 1 and 2 of 2. Let 1 be {A1,,An} and 2
be {B1,,Bm}. Neither 1 nor 2 can be empty, because otherwise 1 2 would just
be either 1, which is satisfiable as a subset of 1 (which was assumed to be satisfiable),
or 2, which is satisfiable as a subset of 2 (which was assumed to be satisfiable).
Set A = A1 & & An. It is clear that every model of 1 makes A true, since each such
model makes each Ai true. Any model of 2, however, must make A false; for if M were
a model of 2 and M made A true, then M would make each of A1,, An true and (since
M is a model of 2) each of B1,, Bm true. So 1 2 would be satisfiable, which is a
contradiction. Therefore, M makes A false, as required.
Chapter 13
13.2

[2 marks each: 13.2, 13.8, 13.9 = 6 marks total]

Since is satisfiable, let M be a model of . Either M xF(x) or not


M xF(x). We show that in either case, {xF(x) F(c)} has a model,
provided c is a constant not in the language of .
In the second case, where not M xF(x), it is clear that M xF(x) F(c)
because M makes the antecedent of the conditional false. So in this case,
{xF(x) F(c)} is satisfiable.
In the first case, where M xF(x), it must be that there is an m in the domain
such that Mcm F(c), where c is a constant not appearing in F(x). It follows that
Mcm xF(x) F(c) because Mcm makes the consequent of the conditional true.
If we further make sure to choose c not appearing in , then by Extensionality M
and Mcm agree on the truth-values of all sentences in , so Mcm is a model of as
well. Since Mcm is a model of and of xF(x) F(c), we again have verified the
satisfiability of {xF(x) F(c)}.

13.8

Let Q be any model of (P). We show that there is an embedding of P in Q.


Step 1: Define the mapping j from |P| to |Q|.
Given an object p in |P|, there is a constant cp which denotes p in the
interpretation P*. Since Q is an interpretation of the language LP, Q
assigns a denotation to each such constant. Set:
j(p) = (cp)Q.
You might be worried about the possibility that some other constant c
denotes p as well: cP* = p. Then how do we decide whether to set j(p) =
(cp)Q or j(p) = cQ? Fortunately, there is no problem here because the
sentence c = cp is true in P* and therefore is in (P); hence, it is also true

6
in Q (which is a model of (P) by assumption), and therefore cQ = (cp)Q.
So to define j(p), you can pick any constant that denotes p.
Step 2: Show that j is 1-to-1.
If j(p1) = j(p2), that means (cp1)Q = (cp2)Q, which means the sentence
(1)

cp1 = cp2

is true in Q. But (1) is an atomic sentence and therefore if it is true in Q, it


must be in (P) and hence true in P*. But in P*, cp denotes p, so that the
sentence (1) is true in P* only if p1 = p2. This shows that j is 1-to-1.
Step 3: We have to show that the clauses (I1) and (I2) of section 12.1 are
satisfied (since there are no function symbols, we dont have to worry about (I3)).
For (I1): RP(p1,,pn) iff
iff
iff
iff

P* R(cp1,,cpn)
Q R(cp1,,cpn)
RQ((cp1)Q,,(cpn)Q)
RQ(j(p1),,j(pn))

(by definition of P*)


(since Q is a model of (P))
(by step 1)

For (I2): If c is any constant, suppose cP = p. Then if cp is the constant added to


the original language that denotes p, we have P* c = cp and hence Q c = cp,
which means
cQ = (cp)Q = j(p) = j(cP),
as required.
13.9

Suppose there is an embedding j of P in Q and A is an existential sentence of the


form x1xn F, and P A.
This means that there are m1, , mn in the domain of P such that if we add to our
language constants c1, , cn not appearing in F, and we set P* = P c1m1c2m2cnmn
[P* is just P extended to make the new constants denote m1, , mn], then
P* F(c1,,cn).
Now if we extend Q to Q* = Q c1j(m1)c2j(m2)cnj(mn), [Q* is just Q extended to make
c1 denote j(m1) and so forth], then j is an embedding of P* in Q* (because for each
new constant ci, j(ciP*) = j(mi) = (ci)Q*). All we need to do is show that
Q* F(c1,,cn),
since then Q A.
So the problem reduces to showing that if F is a formula in the language of P*
containing no quantifiers, and there is an embedding of P* in Q*, then if P* F it
must be the case that Q* F. In fact, we show that for all such F, P* F iff Q*
F. The proof of this is by induction on complexity, and can be lifted right from

7
the first part of the proof of the Isomorphism Lemma (12.5). (There are no
function symbols.)
Atomic case. If F is R(c1,,cn), then
P* R(c1,,cn)

iff RP*(c1P*,,cnP*)
iff RQ*(j(c1P*),,j(cnP*))
iff RQ*(c1Q*,,cnQ*)
iff Q* R(c1,,cn)

If F is c1 = c2, then
P* c1 = c2

iff c1P* = c2P*


iff j(c1P*) = j(c2P*)
iff c1Q* = c2Q*
iff Q* c1 = c2.

since j is 1-to-1

Nonatomic case. Suppose that the result holds for A and B: namely, P* A iff
Q* A, and the same for B.
If F is ~A, then
P* F

iff not P* A
iff not Q* A (since result holds for A)
iff Q* F.

If F is (A B), then
P* F

iff P* A or P* B
iff Q* A or Q* B (since result holds for A and B)
iff Q* F

Finally, if F is (A & B), then


P* F

iff P* A and P* B
iff Q* A and Q* B (since result holds for A and B)
iff Q* F.

You might also like