You are on page 1of 8

Compare and contrast Tarski's and Bolzano's account, and explain Quine's

argument that, in first-order cases, they are equivalent. Do you agree with the
argument? If not, why not? If so, what is its significance for our understanding
of logical consequence?

Advanced Topics in Philosophical Logic

200-353-705
Introduction
In this essay I will briefly exposit the competing substitutional and modal theoretical
accounts of logical consequence. Much of the language in the early sections will refer to
logical truth but I will set out how it can be used to directly refer to logical consequence.
After setting out Bolzano's and Tarski's projects, I will examine the similarities and
differences between their accounts. I will briefly explain Quine's motivation behind wanting a
purely substitutional account, and explain his argument for first order equivalence. I will
conclude that though he may not be correct about their equivalence, he forces justification of
a set-theoretic ontology for logical consequence. His opponents must explain why logical
consequence needs sets and sequences and not solely rely on the combination of grammar
and truth.

Tarski and Bolzano's account of logical consequence are very similar. I will firstly explain
Bolzano's account and then describe how Tarski modifies and extends it. Then I will show
the problems that the former account faces, and how Tarski's account solves it.

Bolzano

Bolzano offers a substitutional account of logical truth. Bolzano considered logical truth to be
a relation between propositions and component ideas, though to facilitate comparison with
Tarski I'll refer to sentences and atomic expressions (Etchemendy, 1990, p. 28). Sentences are
not logically true simpliciter; they are logically true with respect to a set of fixed terms –.
This is clearly only the beginning of a depiction of logical truth, to finish the project one
needs to discover the terms that must remain fixed.

A substitutional account captures our intuition that logical truths remain true through every
permutation of the non-logical terms in the sentence. Logical truths remain true relative to 
when we substitute a subset of expressions (the ones that we do not hold fixed) for those of
the same grammatical type. When variable terms are substituted they must be substituted
uniformly. For example, the sentence 'Joe is mortal or Joe is not mortal' (ignoring possible
failure of reference) is logically true if we keep, at least, the connectives 'or' and 'not' fixed
and uniformly substitute all other expressions for those of the same grammatical type. For
example, 'Joe' may be substituted to 'James'' and 'is mortal' can be substituted to 'is mordant'
Rendering

James is mordant or James is not mordant.


All grammatically permissible permutations of the sentence remain true iff the set of fixed
terms contain 'or' and 'not.' If any set  does not contain at least one of these terms then the
sentence can easily be changed to render a falsity. For example

James is mordant and James is not mordant.

Is clearly false. Any sentence with at least one false substitution instance is not a logical truth.
Bolzano's account can be put thusly:

A sentence is logically true with respect to a set of fixed terms  if all substitutional
instances are true. In terms of logical consequence, P is a logical consequence with respect to
 of  iff every uniform substitution of non-logical terms that makes  true also makes P true
(Turner, 2011).

Tarski

Tarski's analysis extends Bolzano's account by appealing to the notion of satisfaction of a


sentential function. Sentential functions are sentences in which at least one term is replaced
by a variable. For example if we replace 'Joe' in the sentence above with a variable, it
becomes:

x is mortal or x is not mortal

Sentential functions are satisfied by sequences. A sequence is a function that assigns an


object to each of the variables for the purposes of satisfying logical truth (Etchemendy, 1990,
p. 35). A sentential function is not true simpliciter, it is true relative to a sequence that
satisfies it. For a sentential function, and thereby any sentence that is a limiting case of it, to
be logically true, the sentential function must be satisfied by all sequences. More formally, P
is a logical truth with respect to the set of fixed terms  iff every re-interpretation of the non-
logical terms in P makes it true. For logical consequence, P is a logical consequence with
respect to  of  if every reinterpretation of the non-logical terms that makes  true also
makes P true.

Comparing and Contrasting

As I've written earlier, the Bolzano and Tarski account are very similar in terms of their
structure and motivating intuitions. Both are trying to capture the sense that logical truths
should remain true regardless of what is 'plugged in' to the non-logical variable terms in the
sentence. Additionally, they both treat logical truth and consequence as a relation that holds
between sentences in a language rather than a property of sentences, or sets of sentences. The
Bolzano view coincides with Tarski's in allowing many different 'logics' which depend on 
holding different terms fixed. They allow doxastic or deontic logics and can develop
conceptions of 'logical' truth for both.

Where they differ, is their ability to deal with our intuition that logical truth must persist
through expansions or contractions of the language. In radically impoverished languages,
with little substitutional opportunities, substitutionalism overgenerates for logical truth. For
example, in a language with truth functional connectives held fixed and the following non-
logical terms:

F = is President

a = Barack Obama

b = George Bush

The conclusion becomes a logical truth.

Fa

Fb

The premise and conclusion are both true and no substitution instance renders the premise
true and the conclusion false (Turner, 2011). A simple expansion of the language to include a
third term – c, where c = Joe provides a false substitution instance, making the sentence no
longer logically true. This clashes with our intuition that logical truths should be preserved
through an expansion of the language. Similarly, mutatis mutandis, on contracting the
language by removing c, sentences that were not logical truths become so. Bolzano's
formulation of logical truth and consequence clearly overgenerates in impoverished
languages. This shows that while substitution may be a necessary component of logical truth
it is not sufficient.

Tarski's account of satisfaction provides a limiting factor for substitution and claims to be
both a necessary and sufficient account. Tarski rules out the logical consequence of Fb above
because there is some perhaps unnamed object that fails to satisfy F. Thus no expansion of L
which merely includes the name of this object can affect the logical status of the original
argument (Etchemendy, 1990, p. 35). Tarski's account allows us to go beyond the actually
available substitutional instances of any sentential function by having sentential functions
satisfied by sequences. Sequences put at our disposal all possible names that might be
incorporated into a language (Etchemendy, 1990, p. 37) thereby preventing changes in the
actually existent language resulting in overgeneration. However, the requirement for
satisfaction by sequences instead of a substitution of terms by those of the same grammatical
types adds an extra layer of complexity to the analysis of logical truths. Firstly, by invoking
the notion of sequences that map "possible expressions" onto actually existing terms, Tarski
must introduce sets into his analysis – later Quine will argue that this is ontologically
profligate.

Bolzano's account fails because it does not depend wholly on a set of logical terms to depict
logical truth, but it is also affected by the size of the domain of substitution. Tarski chose
satisfaction domains and defines a satisfaction relation, though this adds complexity. If
substitutional domains expand beyond names it results in ungrammaticality if one does not
explain how atomic expressions contribute to meaning. If one does not attempt a rudimentary
semantic theory, one cannot explain the satisfaction relation, and one risks admitting
predicates like 'nix.' Bolzano's substitutional account of logical truth needs only the division
of sentences into grammatical categories, whereby the terms of the same grammatical type
can be freely substituted by each other. Substitutionalism allows us to keep our semantic
theory to a minimum; we don't need to know how each expression contributes to the truth of
a sentence as we are only interested in the end result.

Quine

Motivation

The motivating force behind Quine's objections is his pragmatism. He only seeks simplicity
in characterising logical truth, if competing analyses are coextensive in what they allow to be
logical truth, we should choose the one with the least ontological commitment. It is this
desire for 'desert landscapes' that leads him to a substitutional account. Additionally, it seems
that his belief in the unclarifiability of synonymy would lead him away from a belief in a
'correct' analysis of logical truth.
Claims

At the core of Quine's argument are the results of the Loewenheim Hilbert-Bernays theorem.
The Loewenheim theorem being - 'every schema satisfied by a model at all is satisfied by
some model where U is limited by the positive integers. The Hilbert-Bernays
theorem extends this by stipulating that each of the sets can be stipulated by a
sentence of elementary number theorem (Quine, 1970, p. 54).. The Loewenheim theorem
establishes that any set of sentences that have a true interpretation also have a true
interpretation in arithmetic. This does not yet establish the adequacy of arithmetic for
establishing logical truth. The Hilbert-Bernays theorem does this, it shows that for any
consistent set of sentences, its true interpretation in the universe of natural numbers can be
expressed in the language of arithmetic (Sagueillo, 2001). Combining these we get Quine's
argument:

(A) If a schema is satisfied by a model at all, it becomes true under some substitution of
sentences of elementary number theory for its simple schemata (Quine, 1970, p. 54)

One can derive by contraposition from (A) that if a schema is false under all substitutions of
sentences of elementary number theory, it cannot be satisfied by any model (and thus the
schema is a logical falsity). If we negate this schema, then we get the depiction of logical
truth

(1) If a schema comes out under all substitutions of elementary number theory, it is
satisfied by every model. (Quine, 1970, p. 53)

So, for any supposed logical truth, if a false substitution instance exists, we have the
resources to find it in elementary arithmetic – if it is substitutionally true in elementary
arithmetic, then is satisfactionally true on every model1.

The theorems intend to show that in a sufficiently enriched language we can talk directly of
substitution and not of interpretation. For Quine, if a sentential function is satisfied (or
falsified) by some interpretation involving classes, or sequences, it is also satisfied (or
falsified by some other interpretation that can be written in the notation of arithmetic (Quine,
1982). Quine argues that we can cut to the chase in all this – considering we do have a

1
I have left some steps of the argument out here to save space
language which is sufficiently rich, we don't need the added ontology that sequences bring –
and we can cut to the chase by testing for logical truth in arithmetic.

If Quine is correct in his analysis, the landscape of logical consequence is very much
changed. These are not so much a consequence of his exposition but a presupposition to it.
His pragmatic account only considers the universe of the given interpretation of the language,
his conception is local (Sagueillo, 2001). No other expansion or contraction of the language is
contemplated (Sagueillo, 2001). This is a pre-supposition to his analysis, as to generate the
equivalence between model-theoretic and substitutional accounts he needs the resources of
Peano arithmetic. This is troubling, as it seems that it seems logical truths are then universe
dependent, and logical truths in this world are not true in all worlds. However, maybe this is
what Quine seeks. To conclude his purported definition in Chapter 3, he states 'Logic is, in
the jargon of mechanics, the resultant of two components: grammar and truth.' (Quine, 1970,
p. 60). Earlier in the book, Quine argues that grammar is an immanent notion (Quine, 1970,
p. 22). If grammar is immanent, it is relativised to a language and thus not universalisable2. If
one of the key components of logical truth is relativised to a language, it is not necessarily
universalisable either and can be altered as a result of linguistic or scientific evidence. Quine
agrees, it is only the 'maxim of minimum mutilation ... that suffices to explain the air of
necessity [around logic]' (Quine, 1970, p. 100).

I disagree with Quine for two reasons. Firstly, there is always a trade-off between ontological
parsimony and the explanatory power of an analysis. I feel Quine has erred too far on the side
of caution in his parsimony. Quine admits sets into his ontology so by deploying them in
analysing logical consequence he is not multiplying his ontological categories – it hardly
seems like too much a theoretical sin. My second reason is weaker - it is not clear that his
ontologically limited substitutional account captures the intuitions that we have about logic.
We want it to be universalisable, fixed, unchanging and not dependent on the expressive
capabilities of our language. But if it is as Quine says, at least in this early work - that we are
projecting an extra-linguistic account of an intra-linguistic phenomenon maybe that's not
what he should give us. If what he has provided is the simplest explanation that fits the
phenomena, maybe these bare landscapes are the best thing for us. Regardless, it seems the
ball is back in the court of those that favour 'overpopulated' universes. We need to
2
As a side note, this might be a way of disputing Boolos’ proof. If Quine’s view is that logical truth is not
necessarily universally specifiable then it makes no sense showing how truths are not necessarily substitutable
or co-extensive across languages as it may be acceptable that their respective grammars contain different
concepts of logical truth and not they are not necessarily intertranslateable.
demonstrate that we need an expanded ontology or that the extra-linguisticity of logic is
necessary if we want to capture all our intuitions.

Bibliography
Etchemendy, J. (1990). The Concept of Logical Consequence. Cambridge, Massachusetts: Harvard
University Press.

Quine, W. (1982). Methods of Logic. Cambridge: Harvard University Press.

Quine, W. (1970). Philosophy of Logic. Cambridge, Massachusetts: Harvard University Press.

Sagueillo, J. M. (2001). Quine on Logical Truth and Consequence. Agora - Papeles de Filosofia , 139-
156.

Turner, J. (2011). Handout - the Substitional Account of Logical Consequence.

You might also like