You are on page 1of 14

Frege's Theory of Incomplete Entities

Author(s): Michael David Resnik


Source: Philosophy of Science, Vol. 32, No. 3/4 (Jul. - Oct., 1965), pp. 329-341
Published by: The University of Chicago Press on behalf of the Philosophy of Science
Association
Stable URL: http://www.jstor.org/stable/186528
Accessed: 06-05-2015 15:43 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR to digitize, preserve
and extend access to Philosophy of Science.

http://www.jstor.org

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGE'S THEORY OF INCOMPLETE ENTITIES*

MICHAEL DAVID RESNIK**


University of Hawaii

This paper examines four arguments in support of Frege's theory of incomplete


entities, the heart of his semantics and ontology. Two of these arguments are based
upon Frege's contributions to the foundations of mathematics. These are shown to be
question-begging. Two are based upon Frege's solution to the problem of the relation
of language to thought and reality. They are metaphysical in nature and they force
Frege to maintain a theory of types. The latter puts his theory of incomplete entities
in the paradoxical position of maintaining that it is no theory at all. Moreover, his bneta-
physics rules out well-known suggestions for avoiding this difficulty.

1. Introduction. Several recent articles have been devoted to Frege's theory of


incomplete entities and the paradoxes(or supposed paradoxes)found in it.1 Yet, to
me, the issue concerning the validity of that theory remains unresolved. Therefore,
in this article, I will examine four arguments for Frege's theory which are explicit
or implicit in his writings. Two of these are question-begging; the other two rest on
a metaphysicaltheory of meaning and involve Frege in a self-referentialinconsistency
which admits no obvious repair.
Presupposing as much acquaintancewith Frege's doctrines as may be gathered
from a perusal of his writings and articles on them, I will first state some basic facts
about Frege's theory to which I will appeal in the course of this paper. The specialist
will see these as reminders. Ample references are provided the non-specialist from
which he may obtain further details. Having done this I will show that the following
claims beg the question in favor of the Fregean semantics:
1) Frege's functions are needed to interprethis functional logic;
2) They are requiredin order to make use of his definitionof number.
Next, I will state and document Frege's metaphysicalviews about the relationbetween
language and thought (the realm of sense) and reality (the realm of reference). This
metaphysicsdoes lend fairly persuasivesupport to Frege's theory, but it gives rise to
a theory of types. The latter lands Frege's theory in a self-referentialinconsistency.
The former precludes the obvious repairs.

* Received, April, 1964.


** I would like to thank Professor Burton Dreben for the suggestions which led to this article.
Professors Chung-ying Cheng, Dagfinn Folesdal, Richard P. Haynes, and Charles D. Parsons
have commented on various manuscripts of this article and I am grateful to them. Much of the
research reported here was done while I was employed by the U.S. Air Force Cambridge Re-
search Laboratories, and I would like to acknowledge their support. This article is adapted
from a dissertation submitted to the Department of Philosophy of Harvard University in partial
fulfillment of the requirements for the Ph. D.
1 See [24], [28], [36]. References to a passage in translation immediately follow references
to the passage in the original and are in bold face print. All references to the Grundlagen cite
pages in [18].
329

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
330 MICHAEL DAVID RESNIK

2. Some Basic Facts about Frege's Theory. It is not difficult to establish that
accordingto Frege: 1) There are both incomplete senses and incomplete references;
2) predicates and function names have both senses and references which in each
case are incomplete; 3) the reference of a predicate is not a class or a Wertverlauf
(because these are complete); rather, it is a concept, and 4) concepts themselves
are extensional in their behavior, [26], [27], [28].2 Frege admits that terms like
'incomplete' are only used metaphoricallyby him,3 yet they receive a fairly precise
explication in his semantics and ontology. Incomplete entities have differing kinds
of incompleteness;some may only be united with or completed by complete entities,
others may only be completed by other kinds of incomplete entities. The result is
a simple hierarchyof types for both senses and references.4To be more precise, we
define types inductively as follows: oa is a type; if t1, ..., tn are types, then so is the
ordered n-tuple <t1, ..., t,> (n > o). The complete entities make up type oz. An
incomplete entity belongs to the type <t1, ..., tn> if and only if it may be completed by
(and only by) entities of types tl, ..., tn taken in that order. Thus, a function of two
arguments of type <of,<K>, for example, can only take as its first argument a member of
type oz (an object) and as its second argument, a member of type <o> (a function).
This type theory is reflected in Frege's semantics. The formation rules of the
Grundgesetze assign a unique type to each well-formed expression. The class of well-
formed expressions with which a well-formed expression may be combined to form
new well-formed expressions is determined by its type. The hierarchy of types for
expressions, moreover, is isomorphic to the hierarchy of those for senses and
references. Frege calls expressions of type oz object names, while other well-formed
2 Strictly speaking "sense" and "reference" are relative terms; an entity can be the sense
of one term and (in indirect discourse, e.g.) the reference of another. Moreover, by Cantor's
theorem, there are more things to refer to, more thoughts (propositions) to be true or false,
than there are expressions to do referring or expressing. Frege recognized higher infinities
([181, pp. 96-99). Nonetheless, Frege sometimes uses "sense" and "reference" as absolute terms
(e.g. [61, p. 35; [17], p. 64). Those entities which cannot be senses are extensional in character.
" Uber Sinn und Bedeutung" clearly establishes this for objects, i.e. complete non-senses. Further,
in his review of Husserl's Philosophie der Arithmetik and in unpublished writings, Frege states
that the niames of concepts and functions which have the same value for all arguments may
be substituted for each other salva veritate ([8], p. 320; [171, p. 80; [22], pp. 58-59). If we drop
axiom (V) from Frege's Grundgesetze, omit the MB-notation, and replace axiom (IIb) by the
infinity of its instances, then the usual interchangeability theorems hold as metatheorems for
Frege's functional logic. The proofs proceed in the ordinary way. Thus, as Frege uses the terms
"function," "concepts," and "object," the entities in question are extensional. Accordingly,
we shall not extend those terms to cover complete senses.
3 ([6], pp. 35-36, [17], p. 65); ([5], p. 205, [17], p. 55); ([11], p. 665, [17], p. 115); ([14],
p. 157, [17], p. 134); ([10], p. 372); ([15], p. 36), ([20], p. 1).
4 ([7] Vol. I, pp. 40-41.) Frege does not extend his theory of types explicitly to include a

denumerable infinity of types. This is only implicit in his discussion. Further, it is not clear
from his writings whether senses and references are arranged in parallel but disjoint hielarchies
or whether certain incomplete senses or references can be completed by references and senses,
respectively. Moreover, the hierarchy of senses will also have orders owing to Frege's theory
of indirect discourse. In the first order we will have senses of expressions having their cus-
tomary references, in the next order we will have senses of expressions having their indirect
references, and so on. Lacking sufficient data in Frege's writings we can only give an approxi-
mation of the resulting hierarchy of types and orders. The details are not important here, but
Church's formulation in [25] seems to be a good approximation to what Frege hints at. (Functions
and objects will make up the Oth order.)

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGES THEORYOF INCOMPLETEENTITIES 331

expressions are called function names.5 This theory of types is complimented by


the following semantical principle: An expression of type t can only express
a sense of type t and can only denote a reference of type t.6
The incomplete nature of the members of all types except ot has an additional
consequence which is importantfor the present discussion. Frege's functional logic
can only be formalized as a many-sorted theory. The reduction of a many-sorted
functional logic to a one-sorted functional logic would require us to admit variables
ranging over both complete and incomplete entities. According to Frege, this is
incompatiblewith the incompletenatureof functions. For, except for their occurrences
in quantifiers,function variablesmust carry a set of bracketswith them. We cannot
write 'f = g' but at best, 'f( ) = g( )'. The blanks should properlybe filled by some
other expressions, namely, object names or object variables. Even the expressions
' is one-one' and 'f is incomplete' are illegitimate; we have isolated function letters.
In so doing, we have mistakenlytreatedfunctions as if they were objectsand complete.7
With the preceding in mind, let us now turn to the analysisof Frege's argumentsfor
his theory.

3. The Arguments From the Sem'antics of the Grundgesetze. The first


argument implicit in Frege's writings is that to do logic, at least as set out in the
Grundgesetze,you must recognize Fregean functions. For to interpret that system
you must admit them as values for function variables. Something like this seems to
hiavebeen implied by Geach when he wrote:
The best symbolismcannotinformatively state what a functionis; if you do not
alreadygraspthat,you will not see howthe symbolismworks([21],p. 147).
He could not have meant that one must understandwhat a function is in order to
manipulate the symbolism. This is clearly false. It is fair to take Frege's ideography
as one of the best symbolisms of the logic of functions. We can manipulateit without
knowing what a function is; Frege even intended this.8 Geach may be suggesting that

5 ([7], Vol. I, pp. 43-44.) The hierarchy of types for expressions is only in principle infinite.
Extra notation would be needed to make this extension. Although Frege never presented a
detailed classification of expressions in ordinary language, it is implicit in some of his discussions
([5], pp. 198-202, [17], pp. 47-51). Such a classification of terms in ordinary language seems
fraught with difficulties. From Frege's discussions it seems that we would have to be able to
at least specify a unique paraphrase in quantificational notation of any sentence in ordinary
language. According to Frege, this would be a stronger requirement than specifying a synony-
mous paraphrase; for he states that a sentence of the form r(A & A)' is synonymous with A,
([15], p. 39, [20], p. 5), yet the two sentences do not have the same quantificational structure,
or the same analysis in terms of function and object names.
6 ([4], p. 8, [17], p. 25); ([4], p. 18, [17], p. 32). Frege never formalized the logic of indirect dis-
course, but in a letter to Russell he indicated that new symbols would be needed to avoid equi-
vocacy. Expressions would have one and only one reference and no expression would have an
indirect reference ([22], p. 19). Thus, the names of the resultant symbolism would be divided
into both types and orders. Again we would have a system like Church's. My semantical prin-
ciple has been intentionally left vague; I make no references to orders. It is not necessary to be
precise about this in this paper.
7 ([4], pp. 29-30, [17], p. 40); ([7], Vol. I, p. 35, Vol. II, p. 148, Note 2, [17], p. 180), (Note 2);

([11], pp. 663-664, [17], pp. 113-114); ([9], p. 56); ([5], p. 200, [17] p. 50); ([22], p. 57).
8 ([7], Vol. II, pp. 99-100, [17], pp. 185-186.)

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
332 MICHAEL DAVID RESNIK

we will not understandwhy and how the symbolismof the Grundgesetzeis constructed
unless we know what a function is-that this symbolism is properly open only to
Frege's semantics. This is more plausible.
One does not need to admit Frege's ontology and semantics in order to interpret
his Grundgesetzeand use his logic. A very simple non-Fregean semantics may be
presented for Frege's system by interpretingthe term 'function' as referring to the
complete 'functions' of Church'slogic of sense and denotationand by formalizingthe
resulting theory in a one-sorted logic.9 It is not hard to verify that in such a theory
one can even carryout Frege's proof that every name in his system has one and only
one reference ([35], pp. 95-111). Even Frege's use of the Mg-notation in his axiom
(IIb) can be interpreted as a disguised use of functional abstraction. For example,
'M, (f- 0 + /3)'may be regardedas equivalent to 'M(Ax(x 0O + x)).'
From Frege's point of view there are two objectionsto the proposed interpretation
of his symbolism.
1) Our treatmentof functions is unsatisfactory;we are viewing them as complete.
We have formalizedour semanticsby means of a one-sorted logic, thus allowing the
same sort of variableto range over both our functions and objects. Formulas such as
'f = g' are allowed by us. This is an illegitimateisolation of function letters. In short,
Frege would view our functions as objects.
This objection may be answered very easily, however, by conceding that from
Frege's point of view our functions are complete. To show that our interpretationof
the Grundgesetzeis inadmissible, Frege would have to appeal to his own semantics;
he would have to assumeor prove that functions must be viewed as incomplete. Then,
he would not be able to supporthis semanticsby arguingthat hisfunctionsarenecessary
to interpretthe Grundgesetze.
2) Once we have admitted that our functions are complete and, therefore, objects,
Frege could raisea second objection.He could arguethat, strictlyspeaking,expressions
like 'f(a)' shouldbe analyzedinto 'f', 'a,' and the functionnameC5(4). For 'f' asan
object letter, does not carry an enclosure with it. Thus we must take the application
of a function to an argumentas a primitivefunction. We have only shifted things, and
the need for functionsarisesagain([5], pp. 204-205, [17], p. 54). Once we have admitted
one Fregean function, we might as well admit them all and adopt the Fregean
semantics.
There is a simple answerto this objection also. We may admit that we take 'C(0)'
as a connective,but we need not admitthat it designatessomething.'0Frege's semantics
requires that in a sentence having a truth-value all connectives refer to something.
We need not accept this requirement.To argue that we must, Frege has to appeal
to this semantics. But since a putative justification of this semantics is in question,
such an appeal would be illegitimate at this point.
4, The Argumelnt from the Definition of Ntumber. A second argumentimplicit
in Frege's writings is that incomplete entities, in particularconcepts, are required to
justify his analysis of number and numerical statements. This argument is also
question-begging but the proof of this leads us to consider Frege's definitions of
number. There are two; one of them, the first one, is given in the Grundlagen,the
other of them, the second, is given in the Grundgesetze.The first-the number of
9 Church treats functions as does Frege; i.e., as extensional ([25], p. 18).
10 We do not have to existentially-generalize on this connective. Thus, begging the question
in our favor, we do not have to view it as a designative.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGE S THEORY OF INCOMPLETE ENTITIES 333

F is the extension of the concept 'similar to the concept F'-seems to treat numbers
as classes of concepts.1'The second-the number of u is the extension of the concept
similar to u'-seems to treat numbers as classes of classes since 'u' is a variable
ranging over classes and other objects.12It turns out, however, that Frege interprets
his earlier definition so that the two definitions are essentially the same and so that
numbers are classes of classes.13
If numbers are classes of classes what are we to make of Frege's famous dictum:
"A statementof number contains an assertionabout a concept"([18], p. 59). Consider
the numericalstatement, 'Venus has 0 moons.' Frege transformsthis to 'The number
of moons of Venus 0= ' ([18], p. 69). Using the Grundgesetzedefinitionsof 'number
of' and '0' and the modern 'x' for 'the class of 'x's' we get:
1) A(x is similar to y(y is a moon of Venus)) = z(z is similar to x(x # x))
The sentence (1) does not appearto be an assertionabout a concept at all but rather
one of an identity between two classes. Yet, we must not overlook Frege's statement
that the structureof a thought (proposition)and of a sentence may be analyzedin more
than one way ([5], pp. 199-200, [17], p. 49). Accordingto Frege's syntactic principles
(1) may be seen as constructed from the first level concept name (of type <a>) 'e is
a moon of Venus' and the second level concept name (of type <a>) '(x is similar to
y'O(y))-(z is similar to A(x # x))'. According to Frege 'e is a moon of Venus'
denotes a concept. Thus, by the above analysis, (1) is an assertion of a concept.
Indeed, Frege appears to concede that this is the way his dictum is to be justified:
... in this numerical statement, something is being asserted about a concept. I shall not
dispute whether the assertion relates directly to the concept and indirectly to its exten-
sion, or indirectly to the concept and directly to the extension; for one goes with the
other ([8], p. 322, [17], p. 82).

Therefore, Frege's writings may be interpretedso that in one sense numbers attach
to concepts and in another sense they attach to objects.14
It is the doctrine that numbers attach to concepts, however, that Frege regardsas
central to his analysis of number, and which therefore, seems to form an implicit
argument for the existence of concepts. We have seen that the view that numbers
attach to concepts is borne out by analyzing sentences as we analyzed(1) above. To

11 ([18], pp. 79-80). I have translated "gleichzahlig" as "similar".


12 [7], Vol. I, p. 57. Frege's definition is in fact

c q) r- on
^4T o(aP)

Thus, I have paraphrased.


13 See ([31],p. 542 and [35],pp. 202-208). The key passage here is Frege's remark ([5],pp. 198-
199, [17], p. 48) (made with reference to [18], p. 80, Note) that "the concept F" may replace
"the extension of the conceptF." Frege himself did not make this matter clear in the Grundlagen.
He even remarked: "I attach no decisive importance to the bringing the extension of concepts
into the matter at all." ([18], p. 117.) Thus perhaps at that time, Frege viewed numbers as classes
of concepts or concepts of concepts. We are now considering his fully-developed theory, and
within that context numbers become classes of classes. (Note, by the way, that concepts are not
even described as incomplete in the Grundlagen.)
14 This does not imply that every object has a number. By expanding the Grundgesetze
definition, one may easily verify that when u is not a class, the number of u is the null class.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
334 MICHAEL DAVID RESNIK

justify the conclusionthat a numberattachesto a concept, or that numericalstatements


contain assertions about concepts, we must be assured of the correctness of o)
splitting (1) the way we have, and e) taking the expression 'e is a moon of Venus'
as designatinga concept. Neither of these acts is obviously correct in itself. It is easy
to lay down alternative formation rules for Frege's ideography ([35], pp. 62-72).
Only Frege's rules justify splitting sentences as we have in step (ac).Why should
they be taken as showing the way a sentence is really constructed? They are not an
especially simple set of rules; the ones given along standardlines are easier to state.
Clearly, since simplicity cannot be appealed to, this way of splitting up sentences
must be backedup by a semantics. Step (1B),moreover,requiresa Fregeansemantics.
Only the Fregeansemantics requiresthat 'e is a moon of Venus' be taken as denoting
a concept(in his sense), and this is just what is at issue. Thus, we need Frege'ssemantics
to demonstratethat, by Frege's definition of number, numbers attach to concepts.15

5. The Argument from the Nature of Functionality. The next two arguments
in support of Frege's theory to be considereddo not tacitly beg the question. Starting
from metaphysicalassumptions, problems are raised and Frege's theory of complete
and incomplete entities is offered as a solution to them. The explanatoryvalue of the
theory is supposed to compel us to accept it.
The first argumentI shall consideris that Frege's theory offers the only satisfactory
explanation of functions, relations, and concepts. By starting with the question of
functionality, Frege gives his theory a mathematicalair. We shall see, however, that
Frege's treatment is basically metaphysical. His metaphysical views concerning the
relation between names and their references prevent him from accepting several
explications (including current ones) of the notion of function and force him to his
own explanation.Let us see the details.
1. Unsatisfactoryexplanationsof functionality. Frege considers and rejects several
explanations of functionality. These shall now be listed, together with his grounds
for rejectingthem.
1) Functions yield their values by being correlatedto them. Reply: this explanation
is unsatisfactorybecause functionalsymbols are needed to express laws of correlation;
we need to understand what functions are to understand what correlationis ([11],
pp, 661-662, [17], pp. 111-112).
2) A function is a special kind of sign. Reply: this is insufficient,for it is bound up
with formalism ([11], pp. 662-663, [17], pp. 112-113); ([4], pp 2-4, [17], pp. 21-22).
3) Functions are the references of computational expressions such as '2.3 + 1.'
Reply: These expressions refer to numbers; functions are not just numbers but are
something new ([4], p. 5, [17], p. 23).
4) Functions are the referencesof expressionscontainingfree variables(as opposed
to place markers)such as '2 + x'. Reply: Such expressions do not refer, they only
indicate; one indicates the same numbers with 'x' and 'x + 1Fwhen he writes these
expresssionsin isolation ([4], p. 6, [17], p. 24).
5) Functions yield their values by applying to their arguments. Reply: Then we
are left with the problem of explaining the operation of application; we have only
shifted the difficulty; application would be like a new function; we would need to
know how it connected a function with its argument ([5], p. 204-205, [17], p. 54).
15 Since concepts satisfy the law of extensionality, one cannot properly argue: We must admit

concepts, since Frege's definition presupposes intensions such as properties or attributes.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGE'S THEORY OF INCOMPLETE ENTITIES 335

6) Functions are relations or sets of ordered pairs. (Frege does not consider this
explanation explicitly; we extrapolatein giving his reply.) Reply: Then we have to
explain how relations relate their terms, or else we have to explain the membership
relationin set theory; relationsthemselves can be treatedas special kinds of functions;
thus it is preferableto explain functions first; in general, all logics need at least one
connective, and we can always ask how the connective connects (ibid).
2. Frege's metaphysicalassumnptions. By now one might be impatient with Frege.
He keeps generatingfurther problems by assumingthat every term capableof joining
other terms to form a complete name or sentence designates something which is
supposedto connect non-linguisticentities. Becauseof this, he rejectsseveralproposed
explicationsfor the notion of function. On the other hand, because we do not have to
existentially generalize on 'E', we can view it as non-designative and avoid Frege's
problems by explaining functions in terms of sets. It seems that Frege is imposing
his semantics on us again.
This is not quite the case; Frege has deeper reasons for viewing all connectives
as designative. They are metaphysical and concerned with the relation of names to
references. The documentation for them in Frege's writings is, perhaps, thin.
Nonetheless, since Frege explicitly states analogousreasonsto defend the introduction
of incomplete senses, the evidence we are able to present adequately supports our
interpretation.Frege uses his theory of functions to answer the following questions:
Why can we name somethingwe have never namedbefore ?Why can we referto things
in new ways? His (implicit) answer: Because the world of references has a model in
the world of names. To the structure of the reference there corresponds the com-
pounding of words into names; and here the order is not indifferent. The documen-
tation of this answer now follows.
First, we note that Frege stated that a complex name has a referenceif and only if
its components have references (at least in the context of the complex name).16We
can also establish that Frege thought that a complex reference has parts which are
sensitive to arrangement:"The whole reference and one part of it do not suffice to
determine the remainder"([6], p. 36, [17], p. 65). Frege also stated that objects alone
cannotbe compoundedinto logicalcomplexes;partof the complexmust be incomplete;
it is sensitive to order and has the power to unite objects:
An object-e.g., the number2-cannot logicallyadhereto anotherobject-e.g.,
JuliusCaesar-withouta bindingagent.But thatmaynot be an object,but mustbe
somethingunsaturated.A logicaltie canonlybecomea wholeby havinganunsaturated
partsaturatedor supplementedby one or moreparts([10],p. 372).
By examining their grammaticalrole, Frege decides that functional expressions and
predicatescorrespondto the incompletepartsof the reference.17Thus, Frege concludes
that the essence of function, the referenceof the function name, is its incompleteness
([4], p. 6-7, [17], p. 24). its incompletenessexplainshow it adheres to its argumentto
yield its value; it explains how concepts attach to their instances and how relations
relate their terms. This shows how new object names manage to refer; they contain

16 ([6], pp. 32-33, [17], p. 62.) In [9], pp. 55-56, Frege states that in ordinary language some

words may only obtain reference in context. He views this a logical defect of ordinary language.
17 ([4], pp. 6-7, [17], p. 24); ([5], pp. 193-195, [17], pp. 43-45); ([5], pp. 200-201, [17], pp.
49-50); ([5], p. 205, [17], pp. 54-55); ([11], pp. 665-665, [17], pp. 114-115); ([7], Vol. I, pp. 5-6,
[17], pp. 152-153); ([10], pp. 371-372.)

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
336 MICHAEL DAVID RESNIK

function names as parts; and just as they manage to unite object names to form new
names, so do their referencescombine objects to form logical wholes.
Because Frege speaksabout syntax to such an extent when explainingthe essence of
function, it might be supposed that he has naively taken a linguistic distinction for an
ontological one. Frege was aware of this charge; for he remarked that the close
connection between the linguistic cleavage and the real one allowed him to use the
formerfor expositorypurposes ([5], p. 195, [17], p. 45). Indeed, when he distinguished
between complete and incomplete senses, he argued in the other direction:"the want
of fulfillmentreally emerges in the realm of sense, and is transferredfrom there to the
symbol" ([15], p. 39, [20], p. 4). Of course, Frege probablyarrivedat his metaphysics
by reflectingon the relationshipbetween logic and language.

6. The Argument from the Essence of Language. The same sort of metaphysics
lies behind Frege's introductionof complete and incomplete senses. As is well known,
Frege explained the intersubjectivityof language by positing objective thoughts, the
objective senses of sentences ([6], pp. 29-32, [17], p. 59-62). Thoughts, however, are
not sense-perceptible-to communicate them we need something else. It is via
languagethat we think ([2], pp. 48-49). Only by means of a sentence can we make a
thought known to someone else; language bridges the gap between the sensible and
non-sensible.'8But how does this come about?
The world of thoughts has a model in the world of sentences, expressions, words, signs.
To the structure of the thought there corresponds the compounding of words into
a sentence; and here the order is not indifferent ([14], p. 148, [17], p. 123).

Indeed, this model is based upon and is used to explain the essence of language,
namely that we can understandsentenceswe never heardbefore. The details are given
in the following passage:
It is astonishing what language can do. With a few syllables it can express an incal-
culable number of thoughts, so that even a thought grasped by a human being for
the first time can be put into a form of words which will be understood by someone to
whom the thought is entirely new. This would be impossible, were we not able to distin-
guish parts in the thought corresponding to the parts of a sentence, so that the structure
of the sentence serves as an image of the structure of the thought ([15], p. 36, [20],p. 1).

The problem of joining thought parts arises; we have to explain the unity of the
thought. The words of a sentence do not just lie on the page; their order counts, and
they producea unifiedwhole ([15],pp. 36-37, [20],p. 1-2). The same thing happenswith
the thought:
The whole owes its unity to the fact that the thought fulfills the unfulfilled part or,
as we can also say, completes the part needing completion. And it is natural to suppose
that, for logic in general, combination into a whole always comes about through the
fulfillment of something unfulfilled ([15], p. 37, [20], p. 2).

We have the same metaphysicsonce again:there are logical complexes, and they only
hang together by having an incomplete part with adhesive powers.19
1. The developmentof Frege'smetaphysics.We can tracethe developmentof this meta-
physics in Frege's works. He tells us in the Begriffsschriftthat he was at first bound by
81 ([13], p. 61, [19], p. 292); ([13], p. 66, Note, [19], p. 298 Note); ([22], p. 86).
19 ([15] loc. cit.); ([5], p. 205, [17], pp. 54-55); ([14], pp. 155-157, [17], pp. 131-134).

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGE'S OFINCOMPLETE
THEORY ENTITIES 337

the subject-predicatedistinctionbut then replacedit by the function-argumentone([l],


p. VIII, p. 4, [17],p. 4). This, of course,was centralto his developmentof quantification
theory. At this time he did not make clear whether the distinction between function
and argumentis a syntacticor an ontologicaldistinction([1], pp. 15-17,[17], pp. 12-14).
In the Begriffsschriftfunctions were considered to be (ambiguously) either parts of
sentences of else part of judgeable contents; we find no mention of mathematical
functions.Perhapsthis is why in the Grundlagenthe term 'function' is replacedby the
term 'concept.' Concepts are clearly part of Frege's ontology then, ([18], pp. 60-61);
although calling the concept the predicate of a judgeable content was unfortunate
([18], p. 11, note 2). The concept-object distinction is taken as fundamental, ([18],
p. x), but in the Grundlagenwe never see the terms 'incomplete' or 'unsaturated.'
On the other hand, we find the principle that a word only really has a meaning
(Bedeutung)in the context of a statement ([18], p. x). This never occurs in his other
writings. A plausible explanation for this is that Frege had wondered whether the
meaning of the sentence emanates at all from the independent meanings of its
parts. He gave a negative answer because he could not account for the unity of the
thought. By the time of the Grundgesetzehe had solved the problem of the unity of
the thought by positing incomplete entities, and was able to extend this to other
domains.It seems likely that Frege's metaphysicscame from his concernwith language
and was then applied in mathematics. For functions and concepts are not seen by
Frege as mathematicaluntil the appearanceof "'Funktion und Begrzff."There Frege
writes as if he is passing from mathematicalfunctions to statementalfunctions ([4],
pp. 13-18, [17], pp. 28-32). This could have been for pedagogic reasons-his lecture
was delivered to the JenaischeGesellschaftfur Medizin und Naturwissenschaft([17],
p. 31).

7. The Status of these Arguments. Given Frege's metaphysicalviews concerning


(roughly) language, thought, and reality, his theory of complete and incomplete
entities becomes rather persuasive. Its explanatoryvalue should not be dismissed
lightly. However, if the theory of complete and incomplete entities fails to make sense,
then we have considerationswhich militate against Frege's metaphysics. This is not
a direct refutation of course, since Frege does not (and probably cannot) produce a
logical deduction from his metaphysical principles to his ontological theories. Yet,
since Frege's theory is fairly persuasive when based on these principles, its failing
to make sense would be a good reasonto wonder if perhapsthe metaphysicsitself does
not make sense and, instead, arises from asking the wrong sort of questions.To try
to refute it directly looks like a thanklessundertaking;concentratingon the theory of
complete and incomplete entities will prove more fruitful.
1. A major difficulty.It is not possible to state Frege's theory of incomplete entities
strictly and at the same time adhere to his own principles. This is something he,
himself, recognized:
In the sentence "something is an object" ... whatever we may put in place of "some-
thing", we always obtain a true sentence, since a function name can never assume the
place of "something."20
20 Letter to Bertrand
Russell, dated June 29, 1902. I would like to express my gratitude to
Professor Hans Hermes for permission to quote from the Frege Nachlass. It is being edited by
Professor Hermes and his colleague, Professor Kaulbach, and will be published by Verlag
Felix Miner, Hamburg.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
338 MICHAEL DAVID RESNIK

The details are as follows: 1) By Frege's principles, the predicates 'is a function',
'is saturated',and 'is an object' of ordinarylanguageare first level predicates.2) Thus,
if we interpretthese predicatesas Frege does the sentences 'A is a function' and 'A is
unsaturated'are always false, and 'A is an object' is always true. For A can only be
an object name and, according to Frege, no object is a function and no object is
unsaturated.3) But we want to be able to assertthat some things arefunctions.4) Thus,
strictly speaking,these predicatesdo not work correctly;we cannot use them to state
Frege's doctrine precisely.
For those who feel that it does not make sense to speak of levels with respect to
ordinarylanguage,we turn to Frege's ideography.Frege's theory cannot be formalized
in it. The first level predicate'Funct (s)'is such that: 1) (x),- Funct (x). The second
level predicateTFunct,+(g))' is such that 2) (f) Functt (f(O)). This follows by noting
the levels and the intended interpretationsof these predicates.We also want to be
able to express the tenet that some things are functions and some things are not.
But "(Ex)Funct (x)' and '(Ef), Funct ((/))' contradict (1) and (2), respectively.
This holds generallyfor any predicatesFrege's ideographicprinciples permit. For we
need predicates which cut across types-which have argument places open to both
function and object names or variables-and this is just what he does not allow. They
would conflict with the incomplete nature of functions. Frege realizedthis, too:
As in language, one cannot properly say of a function that it is not an object, one
cannot with this signification say of aln object, e.g., 4, that it is not a function.21

Noting phenomenasimilar to those just presented, Max Black concludes:


A doctrine which reduces philosophy to silence might conceivably be correct, but
nobody can reasonably be expected to accept a theory which on its own showing is
trying to say what cannot be said, ([23], p. 251).

But neither Geach nor Frege seems to find the difficulty so fundamental. They fall
back upon the doctrine of elucidation; namely, that the primitive terms of a system
are not to be defined but must be explainedinformallybefore they are used to define
other terms ([12], pp. 209-302, pp. 305-306). The terms 'function' and 'object' and
'incomplete' are primitive terms in Frege's theory,22and thus, when explainingthem,
he rightly permits himself to rely on the use of metaphors and hints and on the
cooperation of his readers. Accordingly, his reaction to the difficulty in his theory
takes the following form:
By a kind of necessity of language, my expressions, taken literally, sometimes miss
my thought ... I fully realize that in such cases I was relying upon a reader who would
be ready to meet me half-way-who does not begrudge a pinch of salt ... The obstacle is
essential ... there is nothing for it but to realize this and always take it into account
([5], pp. 204-205, [17], pp. 54-55).
All apparent contradictions that we come upon here arise because we wish to treat
the concept as an object, contrary to its unsaturated nature; to be sure, the nature of our
language forces us to this. But this is only something linguistic ([10], p. 372).

Note carefully that Frege talks about relying on the reader to meet him half-way.
He is quick to add that he is not defining anything that all he wants or is able to do is
21 Russell correspondence, Loc. cit. The words, "This signification" refer to a second level

function name which Frege considered in the letter to represent the notion of function.
"2 ([22], p. 30); ([4], p. 18, [17], p. 32); ([5], p. 193, [17], pp. 42-43); ([11], p. 665, [17], p. 115).

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGE'S THEORY OF INCOMPLETE ENTITIES 339

to give hinlts ([5], pp. 204-205, [17],pp. 54-55, 10, p. 372). Frege does not think that the
linguistic obstacle constitutesa problem since we only elucidatethe notion of function.
Those who fail to understandhim have failed to meet him half-way. If we object to
his theory on the grounds that it fails to make sense, he might well reply that his
elucidation of it had not yet been completely successful. We get exactly this sort of
position with Geach:
But the concepts we are dealing with are too simple for a proper analysis; and circum-
locutions may help direct the reader's attention the right way, given a little cooperation
on his side. (It is like showing a point on a map by drawing a ring around it.) ([30],
p. 413.)
2. The theory'sdownfall.Neither Geach nor Frege can correctlyfall back on the doc-
trine of elucidation. Black's failure to find Frege's theory satisfactoryis based upon
a clear understandingof Frege's use of the term 'function'. Moreover, the problems
are not confined to the uninitiated; neither Frege nor Geach can state their theories
precisely to themselves. Clearly,the doctrine of elucidation cannot be used to defend
Frege's attempt to express what, on his own showing, is inexpressible.
Geach may have some other theory of language in mind which would admit the
inexpressible.However, Frege's appealto a necessity of languageis fatal to his theory.
On his view, a theory can only be a body of thoughts; for its doctrines are supposed
to be true or false, and only thoughts are true or false ([13], pp. 60-61, [19], p. 292).
But any body of thoughts, any theory, must have an image in the world of sentences.
On its own saying, Frege's theory is supposed to be one without an image in the
world of sentences. The structure of some of its thoughts is not mirrored in the
structure of sentences. Therefore, on its own saying, Frege's theory of incomplete
entities is not a theory at all; it is self-referentiallyinconsistent.
3. Some obviousrepairs rejected.The situation with respect to Frege's theories is
similarto that in which the theory of types found itself. It was objected that the latter
theory was meaningless accordingto its own principles [29]. At least two replies are
open to those who want to continueto use the theory of types. One may claim that the
theory is really only a set of formation rules for a given class of symbolisms and
that it says nothing about reality or languagegenerally. A more satisfactorysolution,
worked out for simple type theory by Quine, is to do away with type restrictions,and
lay down axioms so that atomic sentences violating the original restrictions are no
longer counted as meaningless but as false. Modificationsare made in the compre-
hension axioms so that classes violating the theory of types do not exist [34]. This
solution makes more sense than the first one. For, taken seriously, the theory of types
has content-it is not merely an empty formalism. In addition, set theories are not
supposed to be concernedwith what is meaningful or meaningless.
One might suppose that similar replies could be made to the objections given to
Frege's theory. Clearly, a reply like the first will not do. Frege's theory included
directions for building formal systems; but it went far beyond that. It was intended
as a theory about non-linguistic entities, too. Moreover, the metaphysics behind it
precludes the second solution. Dispensing with types would eliminate the distinction
between complete and incomplete entities (at least according to Frege's account of
it); with that would go the explanatoryvalue the theory has in connection with the
metaphysicswhich backs it up.
Thus, it would seem that the metaphysicalassumptionswhich drove Frege to his
theory also bring on its downfall and rule out attempts to rehabilitateit.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
340 MICHAEL DAVID RESNIK

REFERENCES

[1] FREGE, Gottlob, Begriffsschrift, eine der arithmetischen nachgebildete Formelsprachen des
reinen Denkens, Halle: Louis Nebert, 1879. (Reprinted Hildesheim I. Ensch, 1962.)
[2] FREGE, Gottlob, "Uber die wissenschaftliche Berechtigung einer Begriffsschrift", Zeit-
schrift fur Philosophie und Philosophische Kritik 81, 1882, pp. 48-56.
[3] FREGE, Gottlob, Die Grundlagen der Arithmetik, Breslau, Wilhelm Koebner, 1884. (Re-
printed Breslau 1934 and Darmstadt and Hildesheim 1961.)
[4] FREGE, Gottlob, Funktion und Begriff, Jena: Hermann Pohle 1891.
[5] FREGE, Gottlob, "Uber Begriff und Gegenstand" Vierteljahrsheft fiur wissenschaftliche
Philosophie 16, 1892, pp. 192-205.
[6] FREGE, Gottlob, "Iber Sinn und Bedeutung," Zeitschrift fiir Philosophie und Philosophische
Kritik 100, 1892, pp. 25-50.
[7] FREGE, Gottlob, Grundgesetze der Arithmetik I, Jena: Hermann Pohle 1893. II, Jena:
Hermann Pohle 1903. (Reprinted Hildescheim: Georg Olms Verlagsbuchhandlung, 1962.)
[8] FREGE, Gottlob, Review of Husserl's Philosophie der Arithmetik, Zeitschrift fur Philosophie
und philosophische Kritik 103, 1894, pp. 313-332.
[9] FREGE, Gottlob, "Lettera del sig. G. Frege all' Editore," (Dated Sept. 29, 1896), Revue de
Mathematique 7 No. 2, 1898, pp. 53-59.
[10] FREGE, Gottlob, "Jber die Grundlagen der Geometrie." (First series.) Jahresbericht der
deutschen Mathematikvereinigung 12, 1903, pp. 319-324, pp. 368-375.
[ 1] FREGE, Gottlob, "Was ist ein Funktion ?" Festschrift fur Ludwig Boltsmann gewidmet zum
sechzigsten Geburtstage, 20 Februar, 1904, Leipzig: Ambrosius Barth, 1904, pp. 656-666.
[12] FREGE, Gottlob, "Uber de Grundlagen der Geometrie," (Second series). Jahresbericht
der deutschen Mathematikervereinigung 15, 1906, pp. 293-309, pp. 377-403, pp. 423-430.
[13] FREGE, Gottlob, "Der Gedanke. Eine logische Untersuchung," Beitrdge zur Philosophie
des deutschen Idealismus 1, 1918, pp. 58-77.
[14] FREGE, Gottlob, "Die Vemeinung," Beitrdge zur Philosophie des deutschen Idealismus 1,
1918, pp. 143-157.
[15] FREGE, Gottlob, "Gedankengefuige," Beitrdge zur Philosophie des deutschen Idealismus 3,
1923, pp. 36-51.
[16] FREGE,Gottlob, Correspondence with Bertrand Russell (unpublished).
[17] FREGE, Gottlob, Translations of the Philosophical Writings of Gottlob Frege. P. T. Geach
and Max Black (eds.). Oxford, 1952. Second edition, Oxford: Basil Blackwell, 1960. (Con-
tains translations of [1], [4], [5], [6], [7], [8], [11], [14].)
[18] FREGE, Gottlob, The Foundations of Arithmetic, trans. J. L. Austin, Oxford, 1950. Second
edition, Oxford: Basil Blackwell, 1953. (English-German edition of [3].)
[19] FREGE, Gottlob, "The Thought: A Logical Inquiry," trans. A. N. and Marcell Quinton.
Mind 65, 1956, pp. 289-311.
[20] FREGE, Gottlob, "Compound Thoughts," trans. R. H. Stoothoff, Mind 72, 1963, pp. 1-17.
[21] ANSCOMBE, G. E. M. and GEACH, P. T., Three Philosophers. Ithaca: Comell University
Press, 1961.
[22] BARTLETT, J. M., Funktion und Gegenstand. Dissertation, Munich, 1961. (Contains passages
from the Frege Nachlass.)
[23] BLACK, M., "Frege on Functions," Problems qf Analysis by Max Black. Ithaca: Cornell
University Press, 1954, pp. 229-254, pp. 297-298.
[24] CATON, C., "An apparent difficulty in Frege's ontology," The Philosophical Review 71,
1962, pp. 462-475.
[25] CHURCH, A., "A Formulation of the Logic of Sense and Denotation," Structure, Method,
and Meanings: Essays in Honor of H. M. Sheffer. Kallen, P. Henle, and S. K. Langer (eds.)
New York: Liberal Arts Press, 1951.
[26] DUMMETT, M., "Frege on Functions: A Reply," The Philosophical Review 64 ,1955, pp. 96-
107.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions
FREGE S THEORY OF INCOMPLETE ENTITIES 341
[27] DUMMETT, M.,"'Note: Frege on Functions," The Philosophical Review 65,1956, pp. 229-230.
[28] FISK, M., "A paradox in Frege's Semantics," Philosophical Studies 14, 1963, pp. 56-62.
[29[ FITCH, F. B., "Self-reference in Philosophy," Symbolic Logic. New York: Ronald Press,
1952.
[30] GEACH, P. T., "Subject and Predicate," Mind 59, 1950, pp. 461-482.
[31] GEACH, P. T., "Frege's Grundlagen," The Philosophical Review 62, 1953, pp. 535-544.
[32] JACKSON, H., "Frege on sense-functions," Analysis 23, 1963, pp. 84-87.
[33] PATZIG, G., (ed.) Funktion, Begriff, Bedeutung. Gottingen: Vandenhoech and Ruprecht,
1962. (Contains reprints of [2], [4], [5], [6], [11].)
[34] QUINE, W. V., "Unification of Universes in Set Theory," Journal of Symbolic Logic 21,
1956, pp. 267-279.
[35] RESNIK, M. D., Frege's Methodology, Dissertation, Harvard, 1963.
[36] WELLS, R., "Is Frege's Concept of Function Valid ?" The Yournal of Philosophy 60, No. 23,
1963.

This content downloaded from 146.50.96.170 on Wed, 06 May 2015 15:43:56 UTC
All use subject to JSTOR Terms and Conditions

You might also like