2.

7 Frege’s Conception of Logic

Before receiving the famous letter from Bertrand Russell informing him of the
inconsistency in his system, Frege thought that he had shown that arithmetic is
reducible to the truths of logic. It is recognized today, however, that at best
Frege showed that arithmetic is reducible to second-order logic extended only by
Hume’s Principle. Some philosophers think Hume’s Principle is analytically true
(i.e., true in virtue of the very meanings of its words), while others resist the
claim, and there is an interesting debate on this issue in the literature (see,
e.g., Boolos 1997, Wright 1999).

However, for the purposes of this introduction to Frege’s work, there are prior
questions on which it is more important to focus. Whereas Frege thought that the
truths of arithmetic are derivable from analytic truths of logic, Kant thought
arithmetic principles are synthetic, in which case they wouldn’t be derivable from
analytic truths. Their different conceptions of logic helps to explain why these
two philosophers came to such different conclusions. In this section, we therefore
turn to the following questions:

How did Frege’s conception of logic differ from Kant’s? In particular:

What resources (or laws) did Kant and Frege both consider to be logical?
Did Kant and Frege agree about the content and subject matter of logic?
How did Frege’s conception of logic differ from that of later logicians?
An answer to the first question sets the stage for answering the second.

2.7.1 How Frege’s Conception of Logic Differed from Kant’s

One of the most important differences between Kant and Frege concerns the resources
available to logic. Kant’s logic is limited to (a) Aristotelian term logic with a
simple theory of disjunctive and hypothetical propositions, and (b) representing
inclusion relations among concepts (MacFarlane 2002, 26). By contrast, Frege’s
logic includes (a) a term-forming operator ε’, which allows one to form the
singular term ε’ƒ(ε) from the function expression ƒ(ξ), and (b) a Rule of
Substitution, which allows one to substitute complex open formulas for free second-
order variables in theorems of logic but also allows one to define, and assert the
existence of, complex concepts, including concepts defined in terms of quantifiers
over concepts. We’ll discuss both of these resources below, but first, the
discussion needs some context.

The differences concerning the resources available to logic revolve around a key
issue, namely, whether the additional resources Frege assigns to logic require an
appeal to non-logical constructions, specifically to a faculty of ‘intuition’, that
is, an extralogical source which presents our minds with phenomena about which
judgments can be formed. (Recall the discussion above about Frege’s early interest
in appeals to intuition.) The debate over which resources do and do not require an
appeal to intuition is an important one. Frege continued a trend started by Bolzano
(1817), who eliminated the appeal to intuition in the proof of the Intermediate
Value Theorem in the calculus (which in its simplest form asserts that a continuous
function having both positive and negative values must cross the origin). Bolzano
proved this theorem from the definition of continuity, which had recently been
given in terms similar to the definition of a limit (see Coffa 1991, 27). A Kantian
might simply draw a graph of a continuous function which takes values above and
below the origin, and thereby ‘demonstrate’ that such a function must cross the
origin. But appeal to a graph involves an appeal to intuition, and both Bolzano and
Frege saw such appeals to intuition as potentially introducing logical gaps into a
proof. There are reasons to be suspicious about such appeals: (1) there are
functions which we can’t graph or otherwise construct for presentation to our
intuitive faculty, e.g., the function
f
which maps rational numbers to
0
 and irrational numbers to
1
, or the functions noted by Weierstrass, which are everywhere continuous but
nowhere differentiable; (2) once we take certain intuitive notions and formalize
them in terms of explicit definitions, the formal definition might imply
counterintuitive results; and (3) the rules of inference from statements to
constructions and back are not always clear.

Frege dedicated himself to the idea of eliminating appeals to intuition in the

proofs of the basic propositions of arithmetic. He explicitly remarked upon this
fact in a number of works throughout his career (1879, Preface/5, Part III/§23;
1884, §§62, 87; 1893, §0; and 1903, Appendix). Thus, he would deny Kant’s dictum
(A51 [B75]), ‘Without sensibility, no object would be given to us’, and claim
instead that
0
and
1
are objects but that they ‘can’t be given to us in sensation’ (1884, 101). Frege’s
view is that we can understand or grasp them as objects if we (a) define them as
extensions of concepts, and (b) show that singular terms of the form ε’ƒ(ε)) can be
axiomatized by an analytic proposition. (The latter was to be accomplished by Basic
Law V, and so the collapse of this law in light of Russell’s paradox undermined
this part of his plan for avoiding appeals to intuition.)

Moreover, philosophers have questioned whether Frege’s Rule of Substitution

(Grundgesetze I, 1893, §48, item 9) also required an appeal to intuition. The Rule
of Substitution allows one to substitute complex formulas for free second-order
variables in logical theorems to produce new logical theorems. Boolos argued (1985,
336–338) that since Frege’s Rule of Substitution is equivalent to a Comprehension
Principle for Concepts, it is extralogical in character. The Comprehension
Principle for Concepts asserts
∃
F
∀
x
(
F
x
≡
ϕ
)
, provided
ϕ
doesn’t have a free variable
F
; the proviso blocks the instance
∃
F
∀
x
(
F
x
≡
¬
F
x
)
, from which one can quickly derive a contradiction. Thus, the Comprehension
Principle for Concepts asserts the existence of a concept corresponding to every
expressible condition on objects. From Kant’s point of view, such existence claims
were thought to be synthetic and in need of justification by the faculty of
intuition. So, although it was one of Frege’s goals to avoid appeals to the faculty
of intuition, there is a question as to whether his system of second-order logic
(minus Basic Law V), which implies a principle asserting the existence of a wide
range of concepts, really is limited in its scope to purely logical laws of an
analytic nature.

If we now put aside their differences about logic’s resources and the appeal to
intuition, there are other ways in which the Kantian and Fregean conceptions of
logic differ. Both MacFarlane (2002) and Linnebo (2003) point out that one of
Kant’s central views about logic is that its axioms and theorems are purely formal
in nature (i.e., abstracted from all semantic content and concerned only with the
forms of judgments) and are applicable across all the physical and mathematical
sciences (1781 [1787], A55 [B79], A56 [B80], A70 [B95]; and 1800, 15). And Kant
takes the laws of logic to be normative and prescriptive (something one can get
wrong), and not just descriptive (1800, 16); they provide constitutive norms of
thought (MacFarlane 2002, 35; Tolley 2008). Indeed, Linnebo takes two theses, that
logic is formal and provides laws that are constitutive norms of thought, to be
distinctive of Kant’s conception (Linnebo 2003, 240).

By contrast, Frege rejects the idea that logic is a purely formal enterprise
(MacFarlane 2002, 29; Linnebo 2003, 243). He took logic to have its own unique
subject matter, which included not only facts about concepts (concerning negation,
subsumption, etc.) and identity, but also facts about relations (e.g., their
properties and ancestrals). Frege (1906, 428 [1984, 338]) says:

Just as the concept point belongs to geometry, so logic, too, has its own concepts
and relations; and it is only in virtue of this that it can have a content. Toward
what is thus proper to it, its relation is not at all formal. No science is
completely formal; but even gravitational mechanics is formal to a certain degree,
in so far as optical and chemical properties are all the same to it. … To logic,
for example, there belong the following: negation, identity, subsumption,
subordination of concepts.
And, of course, as we’ve seen, Frege supposed that there is a domain of special
logical objects (courses of values), among which he defined, and – until confronted
by Russell’s paradox – took himself to have proved facts about, extensions and
natural numbers (1884, 1893/1903). Logic, then, is not purely formal, from Frege’s
point of view, but rather can provide substantive knowledge of concepts and
objects.

There is some question, however, as to the extent to which Frege took logic to
provide constitutive norms of thought. Linnebo suggests that Frege eventually
rejected this idea. Though he offers various arguments for thinking that Frege
moved away from the constitutivity thesis, his [Linnebo’s] main argument concerns
the fact that Frege wanted to position Basic Law V as a logical claim, but that
Basic Law V doesn’t seem to be a constitutive norm of thought (Linnebo 2003, 247).
This is a persuasive reason, though it does make one wonder what Frege could have
meant, in the second volume of Grundgesetze (§147), when he said, concerning Basic
Law V:

If there are logical objects at all … then there must also be a means of
apprehending, or recognizing them. This … is performed … by the fundamental law of
logic that permits the transformation of an equality holding generally [
∀
x
(
f
(
x
)
=
g
(
x
)
)
] into an equation [ε’ƒ(ε) = α’g(α)]. [Authors note: the equations in brackets were
added for the sake of clarity.]
Given Frege’s commitment to logical objects as part of the content of logic, the
above passage suggests that he might have regarded the law which transformed an
equality holding generally into an equation as a constitutive norm of thought.

But many Frege scholars are convinced that Frege took the laws of logic to provide
constitutive norms of thought (MacFarlane 2002, Taschek 2008, Steinberger 2017).
MacFarlane, in particular, argues that Kant and Frege may have agreed that one of
the most important characteristics of logic is its generality, and that this
generality consists in the fact that it provides normative rules and prescriptions.
He notes that “[t]he generality of logic, for Frege as for Kant, is a normative
generality: logic is general in the sense that it provides constitutive norms for
thought as such, regardless of its subject matter” (2002, 35). So, though they may
differ as to which principles are logical, there may be at least one point of
reconciliation concerning how Kant and Frege conceived of logic.

2.7.2 How Frege’s Conception Differed from Later Logicians

Given the constraints of the present entry, we shall not attempt to discuss this
question in any detail; instead, we present only the barest of outlines. After all,
modern logicians and philosophers of logic have not yet come to agreement about the
proper conception of logic. Many have a conception of logic that is yet different
from both Kant’s and Frege’s, one that was, to some extent, anticipated by Bolzano,
namely, that logical concepts and laws remain invariant under reinterpretation of
the non-logical constants or under permutations of the domain of quantification.
But since this modern conception is still a matter of debate, it may be that
elements of Frege’s conception will yet play a role in our understanding of what
logic is.

It is important to recognize just how much Frege took himself to be focusing on the
content, as opposed to the form, of thoughts. His concern to more precisely
represent the content of thoughts is stated explicitly in an 1882 lecture before
Jena’s Society for Medicine and Natural Science, where he distinguished his 1879
system from Boole’s logic by saying:

I was not trying to present an abstract logic in formulas; I was trying to express
contents in an exacter and more perspicuous manner than is possible in words, by
using written symbols. I was trying, in fact, to create a “lingua characteristica”
in the Leibnizian sense, not a mere “calculus ratiocinator”—not that I do not
recognize such a deductive calculus as a necessary constituent of a
Begriffsschrift. [Frege 1882, V.H. Dudman (trans.) 1968]
So Frege was not just trying to develop an abstract reasoning system for the
precise derivations of theorems from axioms (see van Heijenoort 1967 for
discussion). Frege was at least as interested in formalizing the content of
reasoning as he was in formulating the rules for deriving a given thought from some
group of thoughts. Frege would not have regarded the logical axioms of his formal
systems as axiom schemata, i.e., as a metalinguistic sentence patterns whose
instances (i.e., the sentences of the object language that match the pattern) are
axioms (see Goldfarb 2001 for discussion). Nor would he have agreed that the
logical axioms of his system were uninterpreted sentences. His unease with the
modern conception of an uninterpreted formal system was expressed in his reaction
to Hilbert’s Foundations of Geometry (Frege 1906, 384 [1984, 315]):

The word ‘interpretation’ is objectionable, for when properly expressed, a thought

leaves no room for different interpretations. We have seen that ambiguity simply
has to be rejected …
Instead, Frege thought that his logical axioms are (a) fundamental truths governing
negation, conditionalization, quantification, identity, and description, and (b)
principles from which other such fundamental truths could be derived. Indeed, even
though Frege sometimes introduces methods for abbreviating these truths, he takes
great pains to insist that these abbreviations are to be understood in terms of the
full content being expressed. For example, he summarizes the law of universal
instantiation (Basic Law IIa) in 1893/§47 using a formula that we would write
nowadays as:

∀
x
F
x
→
F
y
But closer inspection of §20, where this principle is first discussed, and of §17,
where he introduces some notational conventions used in §20, makes it clear that
the above is shorthand for:

∀
F
∀
y
(
∀
x
F
x
→
F
y
)
This latter is a sentence; it is not a schema, nor an open formula with a free
variable, nor an uninterpreted sentence. Rather it asserts something that Frege
takes to be fundamental law, namely: for any property
F
and for any object
y
, if everything falls under
F
, then
y
falls under
F
.

Frege’s understanding of, and attitude towards, the formulas of his formal language
goes a long way towards explaining why his position, in the now famous debate with
Hilbert about the status of the axioms in a formal system, is not an unreasonable
one. At first glance, it looks as if Frege has mistakenly challenged Hilbert’s
method of relative interpretability, whereby one can prove the consistency and
independence of axiom systems by re-intepreting, and thereby reducing, them to
systems assumed to be consistent. Frege’s objections about what exactly has been
established by these relative consistency proofs may seem misguided to a modern
ear. But since Frege and Hilbert understood the notions of consistency and
independence differently, they didn’t always directly engage with the other’s
ideas. Blanchette nicely shows, both in the entry on the Frege-Hilbert controversy
and in her book (2012, Ch. 5), that if the notions of consistency and independence
are understood Hilbert’s way, then Hilbert’s methods do establish what he says they
do, but that if these notions are understood Frege’s way, they don’t. The reader
should pursue these works for a more detailed explanation and nuanced discussion of
the disagreement.

One can appreciate how Frege and Hilbert might have failed to engage with one
another by considering a simple analogy. Consider the inference from “
x
had a nightmare” to “
x
had a dream” and ask the question, is the latter a logical consequence of the
former? If one examines the inference purely formally, as Hilbert might, then the
sentences have the form ‘
F
x
’ and ‘
G
x
’ and the question becomes, does
F
x
logically imply
G
x
? The answer for Hilbert would be ‘No’, because one can interpret ‘
F
’ and ‘
G
’ in such a way that the inference fails, e.g., just assign the standard meaning of
‘dream’ to ‘
F
’ and assign the standard meaning of ‘nightmare’ to ‘
G
’, i.e., interpret ‘
F
’ as the property having a dream and interpret ‘
G
’ as the property having a nightmare. This shows that, from this purely formal
point of view, “
x
had a nightmare” doesn’t logically imply “
x
had a dream”. (From this purely formal viewpoint, one additionally needs the
premise
∀
y
(
F
y
→
G
y
)
 to infer
G
x
from
F
x
.)

Although Frege had a formal system in which the open sentence

F
x
doesn’t logically imply
G
x
, he takes logical consequence to be a relation among thoughts. He wouldn’t answer
the question, of whether one thought is a logical consequence of another, solely by
looking at the form of the sentences that express them, but rather by looking at
the content of those sentences. Given the meaning (content) that ‘nightmare’ and
‘dream’ in fact have, “
x
had a dream” is a logical consequence of “
x
had a nightmare”, for having a nightmare, i.e., having a bad dream, logically
implies having a dream. So, for Frege, this would clearly be a case of logical
consequence.

This analogy might help one to see how Frege and Hilbert might differ in their
approach to questions of consistency and interpretation. On Frege’s view, the
consistency of a group of axioms depends on content, and if the form of these
axioms, under logical analysis, sufficiently captures their content, this
consistency will be inherited by their formal representations as well. Of course,
in proving consistency, Hilbert was concerned primarily to determine whether an
axiom system entailed a contradiction having the form
ϕ
&
¬
ϕ
. So, given this formal goal, Hilbert’s methods are useful and immune to criticism.

But this brings us to one final issue that is crucial to Frege’s conception of
logic, namely, the extent to which his formal representations capture the content
of the claims being analyzed. This issue is relevant because Frege’s primary tool
for analyzing the content of a mathematical or philosophical claim is by way of
representing the content in a system that axiomatizes the fundamental concepts that
are needed for the analysis. This issue is the subject of the first half of
Blanchette 2012. To see what is at stake, we vary the example from the one used in
Blanchette 2012 (24). Frege would represent the arithmetical law:

No (natural) number precedes zero

in the first instance, as:

¬
∃
x
(
N
u
m
b
e
r
(
x
)
&
P
r
e
c
e
d
e
s
(
x
,
0
)
)
Then if we substitute Frege’s definitions of
N
u
m
b
e
r
(
x
)
,
P
r
e
c
e
d
e
s
(
x
,
y
)
, and
0
, as described in Sections 2.5 and 2.6 above, his representation of the
arithmetical law becomes:

¬
∃
x
(
(
x
=
0
∨
P
r
e
c
e
d
e
s
∗
(
0
,
x
)
)
&

∃
F
∃
z
(
F
z
&
0
=
#
F
&
x
=
#
[
λ
u
F
u
&
u
≠
z
]
)
)
Though the formal representation could be taken further, if we expand the
definitions of
P
r
e
c
e
d
e
s
∗
,
#
F
, and
#
[
λ
u
F
u
&
u
≠
z
]
, enough has been said to pose the question: why think that by deriving the formal
representation from more fundamental principles, Frege has derived the arithmetic
law that no natural number precedes Zero? This question is tackled in some detail
in in the early part of Blanchette 2012, which investigates Frege’s understanding
of conceptual analysis. Her answer (Chapter 4) is that the formal representation of
the arithmetic law has to be (self-evidently) logically equivalent to a good
analysis of the original. If it is, then notwithstanding Frege’s failed reduction
of numbers to extensions, a derivation of the formal representation from more
general logical laws of the kind represented in his system would have in fact
achieved the goal of reducing arithmetic to logic. The reader is directed to her
work for discussion of this important point.