VIII

I. M. BOCHBNSKI

ON ANALOGY·

1. Introductory

The present paper is an attempt to clear up some of the problems involved

in the traditional theory of analogy as presented by the Thomistic school.
The two main ideas behind the formal developments offered here are
(I) analogy is an important discovery, worthy of a thorough examination
and further development, (2) contemporary mathematical logic supplies
excellent tools for such work. This paper is, as far as the author knows,
the first of its kind 1 ; it deals with a difficult subject in a sketchy way; what
it contains is, therefore, not meant to be definitive truths, but rather
proposals for discussion.
The approach to the problems of analogy used here is the semantic one.
This is not the only method, but it would seem to be both the most
convenient and the most traditional. As a matter of fact, it is difficult to
see how equivocity, which is and must be treated as a relation of the same
type as analogy, can be considered except by the semantic method. Also,
St. Thomas Aquinas examined analogy in his question concerning divine
names and the title of Cajetan's classical work is De Nominum Analogia.
It will be taken for granted that the reader has a good knowledge of
classical texts of St. Thomas and Cajetan, and of the content of the
Principia Mathematica 2; no reference will be made to these works, except

• First published in The Thomist 11 (1948) 474-497.

1 The author is, however, indebted to the late Fr. Jan Salamucha and to J. Fr. Drew-
nowski who were the first to apply recent formal logic to Thomistic problems. The
present paper may be considered as an attempt to formalize some of the opinions
expressed by them. Cf. Mysl katolicka wobec Logiki wsp61czesnej (polish = The
Catholic thought and contemporary logic), Poznan 1937 (with French abstracts)
and J. Fr. Drewnowski, Zarys programu filozoficznego (Polish = A sketch of a philo-
sophic programme), Przeghld Filozoficzny, 37, 1943, 3-38, 150-181,262-292, especially
95-98. (There is a French account of this important work in Studia Philosophica
(Lw6w) I, 1935,451-454.
a A. N. Whitehead and B. Russell, Principia Mathematica, 2nd ed., Cambridge 1925-
1927.

97
A. Menne (ed.), Logico-Philosophical Studies
© D. Reidel Publishing Company Dordrecht-Holland 1962
 I. M. BOCHENSKI

for some laws used in the proofs. Other more recent topics of mathemat-
icallogic needed for the theory, as, e.g., plural relations!, semantics 2, etc.,
will be explained.
The main results of our inquiry are: (1) an exact definition of univocity,
equivocity, and analogy of attribution; (2) proof of the principles of
contradiction and of excluded middle for univocal and equivocal names;
(3) a metalogical examination and exact translation of the formula
'analogy itself is analogical'; (4) proof that a syllogism in Barbara with
analogical middle terms, if analogy is defined according to the alternative
theory, is a correct formula; (5) criticism of the alternative theory;
(6) definition of ana10gy of proportionality by isomorphy; (7) proof that
a syllogism in Barbara with analogical middle terms, if analogy is explain-
ed according to the isomorphic theory, is a correct formula; (8) a sugges-
tion that contemporary logic uses analogy.
Incidentally other results are reached, which may have a more general
relevance: (1) the foundations of a semantic system, useful for Thomistic
logic, are sketched; (2) a generalised table of relevant semantic relations
between two names is given; (3) the formal validity of a syllogism in
Barbara, as opposed to its verbal correctness, is defined; (4) a rudimentary
analysis of causality, as understood by Thomists, is supplied.

2. Meaning
The fundamental notion of our theory is that of meaning, described by
the following formula: 'the name a means in the language 1 the content!
of the thing x' (symbolically: 'S(a, 1,/, x)'.) The situation symbolized by
'S(a, 1,/, x)' will be called a 'semantic complex'. In spite of its simplicity
the semantic complex merits a detailed comment.
(1) By 'name' we understand here a written word or other written symbol.
It must be emphasized that a written symbol is just a black mark (a spot
of dry ink) on paper. As such (materialiter sumptum) it is a physical
object which occupies a given position in space and time. It may happen,
therefore, that two names, e.g., a and b have the same graphical form
(symbolically /(a, b), where 'J' suggests 'isomorphy') but we cannot

1 cr. R. Carnap, Abri.fJ der Logistik, Wien 1929, pp. 43-45; Bochenski-Menne,
Grundri.fJ der Logistik, 2nd ed., Paderborn 1962, § 23.
B cr. A. Tarski, Der Wahrheitsbegriffin denformalisierten Sprachen, Studia Philosoph-
ica (Lw6w), I, 1935,261-405.

98
 ON ANALOGY

speak correctly of 'the same' name which occurs twice, e.g. as middle
term in a syllogism. In that case we have always two different names of the
same graphical form.
(2) Every relation of meaning implies a reference to a language. This is
obvious, for the same name may mean one thing in one language and
something quite different in another. Moreover, it may have no meaning
at all in another language. If the mention of a language is omitted in
classical definitions, it is because the authors writing during the Middle
Ages and the Renaissance thought of the only one language used at that
time, Latin.
(3) What we call 'content' is what classical Thomists called 'ratio'. This
ratio is always conceived as something determining the thing whose
content it is; even in case of substantial contents (as 'substance' and
similars) we conceive them as such and St. Thomas explicitly teaches that
in this case we always have to do with a quality in a broader meaning
(including 'substantial quality')
(4) Finally, the 'thing' means the same as the 'res' of the Thomists,
namely the subject to which the content connoted by the name belongs.
This is, at least if the logical analysis is pushed sufficiently far, an individual.
The relation S gives rise to several partial relations and partial domains.
We are not going to investigate them here, as they are not relevant
to our theory. We shall note, however, that the relation S allows some
elegant definitions of some important semantic terms. Let Dn' R be the
class of all Xn such that there is at least one Xl, one X2 •.. Xn-l, one Xn+l,
one Xn+2 .•• Xm (m being the number of terms of R) such that
R(xl, X2, •.. , Xn, ... , xm). We shall call Dn'R 'the n-th domain of R'.
We put now:
2.1. nom = DI.Dl'S = DI.1i {(3 I,f, x) S (a, I,f, x)}
2.2. lin = DI.D2' S = DJ.l {(3 a,f, x) S (a, I,f, x)}
2.3. rat = DJ.Da'S = DI./ {(3 a,l, x) S (a, I,f, x)}
2.4. res = DI.D4' S = DJ.x {(3 a,I,f) S (a, I,f, x)} •
The above definitions define the classes of names (2.1), languages (2,2),
contents (2,3) and things (2.4).

3. Analogy is always a relation involving two names

We contend that analogy, as well as univocity and equivocity, is not an
absolute property of one name, but a relation involving two names at

99
 I. M. BOCHENSKI

least. If this seems contrary to tradition, it is because of the use the

classical authors made of the formula 'the same name': they meant two
names of the same form, but spoke, for the reason mentioned above
(§ 2), of a single name. If, however, our considerations about the names
are admitted, we are compelled to say that no single name is, strictly
speaking, univocal, equivocal, or analogical. A single name may have a
clear meaning or a confused meaning; but it has always one meaning only,
and it is not possible to speak about identity or diversity of its meanings,
which is required, if we have to define univocity, equivocity, or analogy.

4. The sixteen relations between two semantic complexes

Now if our relations involve two meaning names, they must be relations
between two semantic complexes; and as the nature of these relations
depends on the relations holding between the terms of both complexes,
they will be octadic relations, each complex being a tetradic relation. The
general form of such relations will be consequently the following:
R(a, b, I, m,f, g, x, y) ,
where a and b are names, I and m languages, f and g contents, x and y
things, while we have S(a, I,f, x) and S(b, m, g, y).
The question arises now, how many relevant relations are there of the
above type. This depends, evidently, on the number of dyadic relations
between the terms a-b, I-m, f-g and x-yo Such dyadic relations are very
numerous, indeed, infinite in number; but for each couple two relations
only are relevant, namely, l(a, b) and'"" l(a, b) for names; 1 = m and
I ::F m for languages; f = g and f ::F g for contents; x = y and x ::F y for
things. Thus there are sixteen and only sixteen relevant relations between
two semantic complexes. The following table enumerates them:
No. a, b I, m f, g x, y No. a, b I, m f, g x, Y
1. 1 - - - 9. '"" 1 - -
2. 1 - - ::F 10. '""1 - - ::F
3. 1 - ::F - 11. '""1 - ::F -
4. 1 - ::F ::F 12. '""1 - ::F ::F
5. 1 ::F - - 13. '""1 ::F - -
6. 1 ::F - ::F 14. '""1 ::F - ::F
7. 1 ::F ::F - 15. -I ::F ::F -
8. 1 ::F ::F ::F 16. '""1 ::F ::F ::F

100
 ON ANALOGY

This table should replace the traditional division of names into univocal,
equivocal, and synonymous ones. As weare, however,notinterestedin the es-
tablishment ofa full semantic theory, we shall not define all sixteen relations,
but only the first four which are directly relevant to the theory of analogy.

5. Definition of univocity and equivocity

These four (octadic) relations, which we shall name 'Rl', 'R2', 'Ra', and
'14', are defined as follows:
5.1. Rl(a, b, [, m,f, g, x, y).
= Df.S(a, [,f, x) . S(b, m, g, y) . I(a, b) . [ = m •f = g . x = Y
5.2. R2(a, b, [, m,f, g, x, y) .
= D/.S(a, [,f, x) . S(b, m, g, y) 'l(a, b) ·1 = m . f = g . x =P y
5.3. Ra(a, b, I, m,f, g, x, y) •
= D/.S(a, I,f, x) . S(b, m, g, y) . 1(0, b) . [ = m . f =p g . x = Y
5.4. R4(a, b, I, m,f, g, x, y) .
= Df.S(a, [,f, x) 'S(b, m, g, y) . I(a, b) ·1 = m . f =p g' x =p Y
5.1. is the definition of names which are semantically identical in spite of
being (physically) two names. We may call them 'isosemantic' names.
5.2. is the definition of univocal names: quorum (x and y) nomen est
commune [i.e. 1(a, b)], ratio autem significata (f and g) est simpliciter
eadem (f = g). 5.3. is again the definition of names which have the same
denotation, but a different connotation; we may term them 'heterologic'
from M)lo~ = ratio. Finally 5.4. defines the equivocal names: quorum
(x and y) nomen est commune [i.e. 1(a, b)], ratio autem significata simpli-
citer diversa (f = g). In all cases I = m, i.e. both languages are identical.
This being so, we may drop 'I = m' and put 'I' for om'. The above
definitions of univocity and equivocity will now run as follows:
5.5. Un(a, b, I,f, g, x, y) .
= D/.S(a, [,f, x) . S(b, [, g, y) . I(a, b) . x =p y . f = g
5.6. Ae(a, b, [,f, g, x, y) .
= Df.S(a, [,f, x) . S(b, [, g, y) . l(a, b) . x =p y . f =p g
We have used 'Un' to suggest 'univoca' and 'Ae' to suggest 'aequivoca';
we also changed, for technical reasons, the order of the two last factors.
The following laws, which are immediate consequences of 5.5, will be
needed in the latter parts of this paper:
5.7. Un(a, b, [,f, g, x, y) . ::> • S(a, [,f, x)
5.B. Un(a, b, [,f, g, x, y) . ::> • S(b, [,f, x) .

101
 I. M. BOCHENSKI

6. Partial domains and relata

Each of our relations Un and Ae being heptadic, contains a)

= 7 hexadic,
(D = 21 pentadic, (~) = 35 tetradic, (~) = 35 triadic and G)
= 21 dyadic
partial relations, together 119 (120 with the full relation). We may denote
them by' Un' respectively' Ae' followed by two figures: one above, indicating
the type of the partial relation (e.g. 'Un5' for a pentadic partial relation of
Un), another below, meaning the place which it occupies among partial
relations of the given type - the whole between parentheses. E.g. '( Un~)'
will mean the second among the pentadic partial relations of Un.
Moreover, each of these partial relations gives rise, exactly as the whole
relation does, to many partial domains and relata. The n-th domain of the
relation Rwill be symbolized, as above (§2), by 'Dn'R' and the n-th class of
relata of R by 'sgn'R'. There are 120 such domains and 120 such classes of
relata. We shall not define them all; the scope of the above remarks was
only to show how ambiguous the common language is when we use
it to speak about univocity or equivocity and, of course, about analogy.
We shall, however, use our notation in order to define the traditional
terms 'univoca' and 'aequivoca'. We need here first a definition of the
following partial dyadic relations:
6.1. (Un~l) = Df.XY{(3 a,b, l,j,g) Un(a,b,/,j,g,x,y)}
6.2. (Ae~l) = Df.XY {(3 a, b, I,f, g) Ae (a, b, I,f, g, x, y)} .
We can now define the classes called 'univoca' and 'aequivoca' which we
shall name 'uni' or 'aeq':
6.3. uni = Df.F'(Un~l)
6.4. aeq = Df.F' (Ae~l) .
If this would appear too generic, we may use triadic relations, including
the language as a term:
6.5. (Un: 1) = Df.ixy {(3 a, b,f, g) Un (a, b, I,J, g, x, y)}
6.6. (Ae: 1) = Df.ixy {(3 a, b, I, g) Ae (a, b, I,f, g, x, y)}
and consequently:
6.7. unil = Df.D'1(Un:1) v D'2(Un:1)
6.8. aeql = Df.D'1(Ae:1) v D'2(Ae:1).

7. The principles of non-contradiction and excluded middle

Other important laws of our theory are two formulae which will be
called, respectively, 'the law of contradiction' and 'the law of excluded

102
 ON ANALOGY

middle for univocal and equivocal names'. We mean by the first that no
two names can be univocal and equivocal in respect to the same language,
couples of contents and of things. By the second we mean that if such
names are not univocal, they must be equivocal, and conversely. It should
be clearly understood that this is true only in respect of some determined
contents meant by the names, moreover that these names must be of the
same form and the things they mean must be non-identical. For nothing
prevents two names from being univocal in respect off - g and, at the
same time, equivocal in respect of h - j, iff =1= h or g =1= j; also, if the
names do not mean the contents involved, they are neither univocal nor
equivocal in respect of them. The last two conditions follow from our
table in § 4.
Consequently, we state our principles in the following form:
7.1. (a, b, I,f, g, x, y): S(a, I,f, x) . S(b, I, g, y,) . J(a, b) . x =1= y' J
J' '" [Un(a, b, I,f, g, x, y). Ae(a, b, I,f, g, x, y)]
7.2. (a, b, I,f, g, x, y) : S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y' J
J . Un(a, b, I,f, g, x, y) v Ae(a, b, I,f, g, x, y) .
Proofs 1:
(1) P J '" (pq . p '" q) (axiom)
(2) p J . pq v p '" q (axiom)
(3) '" (f = g) . = Dt. . f =f g (definition)
(4) S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y: J
J : '" [S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y . f = g] :
: S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y' '" (f = g):
by (1) puttmg. S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y f = g
,--
p q
(5) = 7.1
by (4), (3),5.5 and 5.6 with the rule for adjunction of quantifiers.
(6) S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y: J
J : S(a, I,f, x) . S(b, 1, g, y) . J(a, b) . x =1= y . f = g . v
v . S(a, I,f, x) . S(b, I, g, y) . J(a, b) . x =1= y • ,..., (f = g)
by (2) with the same substitutions as for (4)
(7) = 7.2
by (6), (3), 5.5 and 5.6 with the rule for adjunction of quantifiers.

1 The method used is that of the Principia Mathematica; therefore what we call a
'proof' is rather a sketch of a proof. Rigorous proof could be, however, easily built
along the lines given here. (This applies to all proofs contained in the present paper.)

103
 I. M. BOCHENSKI

The law of excluded middle shows that the classical Thomists were right
when they named their analoga 'aequivoca a consilio', considering them
as a subclass of the class of aequivoca, and that some modern Thomists
are wrong when they put analogy as a third class coordinated to univocity
and equivocity. Incidentally it may be remarked that the authors of the
Principia Mathematica used an exact translation of the ,aequivocatio a
consilio' when they coined the expression 'systematic ambiguity'. As a
matter of fact, they were treating of analogy.

8. On the generic notions of analogy

Analogy will be, according to the above analyses, a heptadic relation
between two names, a language, two contents and two things (at least).
The names will be of the same form; the things must be different. How the
contents are related we must still investigate. If we suppose that the
answer to that question is expressed by 'F', the generic definition of
analogy will be the following:
8.1. An(a, b, I,J. g, x, y) . =
= Df• • S(a,I,J. x) . S(b, I, g, y) . I(a, b) . x # y . F.
Moreover, using 7.2 we may say that analogy is either a kind of univocity
or a kind of equivocity. According to tradition it is certainly not the
first. Thus it must be the second. We may put therefore:
8.2. An(a, b, I,J. g, x, y) . = Df, • Ae(a, b, I,J. g, x, y) . F.
The question arises now, if there is a factor G such that F would be
identical with the product of G with another factor, say H n , G being
identical in all kinds of analogy, Hn different for each; the definitions of
the successive kinds of analogy would be constructed by putting in 8.2
for 'F' first 'G . HI', then 'G . H2' and so on. Ifit be so, we could say that
the name 'analogy' is univocal; if not, i.e. if there could be no common
factor G, it would be equivocal.
As a matter of fact some well known Thomists asserted that the name
'analogy' is an analogical name, i.e. (according to 8.2) an equivocal one.
We are not going to discuss this assertion, but limit ourselves to a correct
formulation of it. This requires, however, some preliminary steps.

9. Expansion of the theory to higher levels

We must first note, that we are already dealing with a situation that is

104
 ON ANALOGY

far more complex than that which is met in classical formal logic. As a
matter of fact, all artificial symbols of any system of contemporary
formal logic belong to the same semantic level, namely to the object
language, i.e. each of them means some object, but none of them means a
symbol of an object. But in the theory developed above we are using
symbols belonging to a higher level, namely our symbols 'a' and 'b',
which are names of names, i.e. symbols of symbols.
In order to supply the last sentence with a more definite meaning, let us
introduce the following recursive definition: (1) the object language is the
first level; (2) a language such that at least one term of it is a symbol of a
symbol belonging to the n-th level, but none is a symbol of such term, is
the n + I level; (3) a relation holding between objects of which at least
one is of the n-th level, and none is of the n + I level, is ofthe n-th level.
It will appear that our a, b and also S, Un, Ae etc. are of the second level;
consequently the names of these will belong to the third level. Now when
we say that 'analogy' is an analogical name, the word 'analogy' is a name
of An; thus it belongs to the third level. We have to investigate if and how
we are allowed to extend our theory to that level, for everything we said
until now was clearly situated on the second level.
Let us note first that the laws of the third level would be, as far as structure
is concerned, exactly similar to these met on the second. For if we say
that 'analogy' is analogical, we mean that two names, say A and B mean
in our new language (which is, by the way, the third level), the relations
AnI and An2 of the objects (aI' bl , 11. /1, gl' Xl' yJ and (a2, b 2• 12. /2, g2, X 2,
yJ. The last two may be considered as classes; but there is nothing to
prevent us from considering them as objects, as the relations AnI and Ana
are true contents of them. Let us put 'X' for the first and 'Y' for the
second. We shall obtain the following exact formulation of the thesis
'analogy is analogical':
AN(A, B, L, An!, An2, X, Y) .

Here all symbols (except the parentheses and commas) are different from
those used in the former paragraphs; and yet the structure is not only
similar, but strictly identical with the structure of
An(a, b, [,f, g, x, y) .
It is also clear that the whole of our previous analyses might have been

105
 I. M. BOCHENSKI

repeated on the third level. We would reach a theory, whose terms and
meanings would be different from the theory we developed above, but
whose structure would be completely identical.
This suggests an important remark. Analyses of such kind involve the
use of the idea of structural identity, orisomorphism. Now, according the to
theory we shall propose, this means analogy of proportionality. It seems,
consequently, that we cannot treat adequately the problem of the generic
notion of analogy without a previous examination of analogy of proport-
ionality.
10. Analogy of one-one attribution

Among the several kinds of analogy there are only two that are really
relevant: analogy of attribution and analogy of proportionality. Two
names which are related by the first will be called 'attributively analogous';
similarly, two names related by the latter will be called 'proportionally
analogous' .
We are starting with the first kind. Here again there is one relation called
'analogia unius ad allerum' - in our terminology 'one-one analogy'
(symbolically' At') - and another called' analogia plurium ad unum', here
'many-one analogy' (symbolically 'Atm'). Let us begin with the first,
which is the more fundamental.
We have two things, xandyand two contents,Jandg; the names a andb
are equivocal in regard to them, but there is still another characteristic:
x is the cause of y or y the cause of x. Writing 'C(x, y)' for 'x is the
cause of y' we shall have:
10.1. At(a, b,l,f, g, x, y) .
= D/• • Ae(a, h,l,f, g, x, y) . C(x, y) v C(y, x) .
This is, however, rather unsatisfactory, for the connection off and g is not
shown, the relation of causality not being analyzed. We cannot, of course,
give a complete analysis of this highly complex notion here. We shall note
only that the relation of causality is a pentadic relation which holds
between two things, two contents and a peculiar dyadic relation between
the things; e.g. the food is the cause of the health of the animal, if and
only if there is a contentf(health) present in the food (x) such that, if a
peculiar relation R (here: of being eaten) is established between x and
the animal (Y), another content g (the health of the animal) appears in y.
Writing 'C(f, x, R, g, y)' for this relation we shall have:

106
 ON ANALOGY

10.2. At(a, b, I,f, g, x, y) .

= DI• • Ae(a, b, I,f, g, x, y) . (3R) . C(f, X, R, g, y) v
V C(g, y, R,f, x) .
The alternative is necessary, according to the traditional doctrine, as there
may be an analogy independent of the direction of causality.

11. Analogy of many-one attribution

The second kind of analogy of attribution is clearly derived from the

first. The many-one analogy holds, namely, between two names a and b,
if and only if there is a third name c, such that both a and b are attribut-
ively analogous (according to 10.2) with c:
11.1. Atm(a, b, I,f, g, x, y) .
= DI• • (3 c, h, z) • At(a, c, I,f, h, x, z) .
. At(b, c, I, g, h, y, z) .
Let x be food, y urine, z animal, f, g, h the contents called 'health' of,
respectively, x, y, z, and a, b, c the names of these contents. There will be a
many-one analogy of a in respect of b.
We may still distinguish four further subclasses of this class of analogical
names, for in 11.1 we may have either
(1) CU, x, R, h, z) . C(g, y, R, h, z) - or
(2) C(f, x, R, h, z) . C(h, z, R, g, y) - or
(3) C(h, z, R, f, x) . C(g, y, R, h, z) - or
(4) C(h, z, R, f, x) • C(h, z, R, g, y) .

12. Conditions of analogy ofproportionality

There are, according to tradition, two conditions for this kind of analogy:
the contents must be non-identical, i.e. we must have equivocity; still, the
syllogism having as middle terms a couple of proportionally analogous
names must be a correct formula. This is secured, according to classical
writers, by the fact that these middle terms mean something 'proportion-
ally common' in both cases, or that there is an analogatum commune
containing in confuso the contents meant by both names.
It seems at first, that these requirements are contradictory: for, if the
meanings of the two names are quite different, one can hardly see how a
syllogism with them as middle terms may be a correct formula. As a

107
 I. M. BOCHENSKI

matter of fact, not only is there a logical theory capable of fulfilling both
requirements without contradiction, but it seems even that there are two
such theories. It seems, namely, that one theory is suggested by the
'proportionaliter commune', the other by the 'confuse'. We shall call the
former 'isomorphic', the latter 'alternative theory'. As far as is known
to the writer, St. Thomas used the isomorphic theory, while the alternative
seems to be originated by Cajetan.

13. The alternative theory

The central idea of the alternative theory may be explained as follows:
we have to do with three names; one of them means the content/, the other
the content g, f and g being the analogata particularia; the third name
means the analogatum commune, namely, the alternative of f and g,
symbolically f v g. We shall give to that expression a sufficiently clear
meaning by putting
13.1. [f v g] X· = Df. ·fxv gx.
A rather complex situation arises here because of admission of three
names: this makes an expansion of our previous formulae to three
complexes necessary, and the basic formula for analogy of proportionality
becomes a relation often terms. Once a definition of this form is establish-
ed, the (heptadic) relations analogous to Un and Ae will appear as
partial relations of the general one, and the verbal formulae as elliptic.
We shall not, however, define this general relation in that way, as, for
several reasons, to be explained later (§ 16), the whole alternative theory
appears as inadequate. But we are going to investigate the validity of a
syllogism in Barbara with proportionally analogous middle terms. For
the use in that inquiry we define the analogy of proportionality (Anp)
according to the alternative theory as a heptadic relation in the following
way:
13.2. Anp (a, b, I,/, g, x, y) .
= Df• • Ae(a, b, I,f, g, x, y) . (3 h) . f = [g v h] .
This is a partial relation contained in the full relation of analogy described
above.
14. On formal validity of syllogism

If we wish to investigate the validity of a syllogism with analogical

middle terms we meet a serious difficulty unknown in current formal

108
 ON ANALOGY

logic. For in current formal logic it is always supposed that a formula

which is verbally valid is also formally valid; the reason of this supposition
is that all terms used in current formal logic are univocal symbols. Here,
however, the situation is different, as we have to deal with analogical
names. We need, consequently, a distinction between the verbal and
formal validity of a formula; moreover we need to know when a verbally
valid formula is also formally valid. This is by no means a universal
rule, as the case of the syllogism with equivocal and non-analogical
middle terms shows. We are not going to investigate the problem in its full
generality, but we will limit ourselves to a single case, the syllogism in
Barbara.
We shall first construct two languages:
(1) A first-level univocal language. This will be the language of the
theory of classes, interpreted as a logic of contents. In it the mode
Barbara will run as follows:
fCg·hCf·")·hCg.

(2) A second-level analogical language. This will contain all symbols used
until now (small Latin letters being sometimes substituted by small
Greek letters and indexes being added to them), with addition of the
following: (i) 'IT'; a formula composed of 'IT' followed by 'a', followed
by 'b' will be interpreted as meaning the formula 'a C b'; (ii) '+'; a
formula such as 'IT + a + b' will be read: 'a formula composed of II
followed by a, followed by b'; (iii) '8 T'; 'F 8 T' will be read: 'F is a true
theorem'.
The proofs will be developed in a second-level language, containing as
subclasses the above two. We shall proceed as follows. Given the (second-
level) premises A and B such that A 8 T . B 8 T, we wish to prove that the
(verbally correct) conclusion C (of the same level) is a true theorem, i.e.
that C 8 T. We translate A and B into the first-level language, apply to
the result the laws of classical formal logic and obtain a conclusion,
which we re-translate into the second level language; if we are able to
obtain C 8 T in that way, the formula 'if A 8 T· BeT, then C e T' is
clearly a valid formula and the formal validity of the mode, whose
premises are A and B, and the conclusion C, is proved.
We put as a law of translation the intuitively evident:
14.1. S(a, I,f, x) . S(b, I, g, y) : ") : II + a + b 8 T·=. f Cg .
109
 I. M. BOCHENSKI

With the help of 14.1 we can easily prove that a syllogism in Barbara
with univocal middle terms is a formally valid formula; but we cannot
prove it if the middle terms are either purely equivocal or attributively
analogical. Alongside of 14.1 we shall need still another law of translation
for cases where an existential quantifier is involved:
14.2. (3 h)· S(a, I, [[v h], x) . S(b, I, g, y): J
J: (3 h)· [[v h] C g.
This seems to be also intuitively evident.

15. The validity o[ the syllogism in Barbara with analogical middle terms
according to the alternative theory

In such a syllogism the middle term of the major premise is analogical

with regard to the middle term in the minor premise, the situation being
this, that the former means alternatively the content meant by the latter
and some other content. This syllogism, if in Barbara, is a valid formula.
The proof is rather cumbersome, because of the existential quantifier;
we shall however give here a developed sketch of it.
In the first place we need two theorems analogous to 5.7 and 5.8 .These
may be proved as follows.
(1) Anp(a, b, 1,[, g, x, y) .
- . Ae(a, b, I,f, g, x, y) . (3 h) . [= [g v h]
[by 13.2]
(2) _ . S(a, 1,[, x) . S(b, I, g, y) . /(a, b) . x =ft y.
·[ =ft g . (3 h) .[= [g v h]
[by (1) and 5.6]
(3) =. (3 h) . S(a, 1,[, x) . S(b, I, g, y) . /(a, b) .
· x =ft y' [=ft g' [= [g v h]
[by (2) and *10.24 Principia Mathematica.]
(4) =. (3 h) . S(a, 1, [g v h], x) . S(b, I, g, y) . /(a, b) .
·x=fty·[=ftg
[by (3) and *13.12 Principia Mathematica.]
(5) _ . (3 h) S(a, 1, [g v h], x) . (3 h) S(b, 1, g, y) .
· (3 h) . x =ft y . [=ft g
[by (4) and *10.5 Principia Mathematica.]
15.1. Anp(a, b, 1,[, g, x, y). J . (3 h) S(a, I, [g v h], x)
[by (5) and 'p = qr . J . p J q']

110
 ON ANALOGY

15.2. Anp(a, b, I,f, g, x, y) . ) . S(b, I, g, y)

[by (5) and 'p _ qrs . ) . p ) r', dropping the quantifier].
We enumerate now the five hypotheses of the syllogism in Barbara
with analogical middle terms, explained according to the alternative
theory:
+ +
HI. II ml al e T
H2. II + +
bl m2 e T
H3. Anp(ml, m2, I, /1,1 /12, X, y)
H4. Un(al, a2, I, (Xl, (X2, Z, t)
H5. Un(bl, b2, I, PI, P2, u, v) •
The proof of 'II + b2 + a2 e T' runs as follows:
(1) (3 h) S(ml, I, (f.l2 v h], x) by H3 and 15.1
(2) S(al, I, (Xl, z) by H4 and 5.7
(3) (3 h) . [/12 v h] C (Xl by (1), (2), HI and 14.2
(4) S(bl, I, PI, u) by H5 and 5.7
(5) S(m2, I, /12, y) by H3 and 15.2
(6) PI C /12 by (4), (5), H2 and 14.1
(7) PI C /12' (3 h) . (f.l2 v h] C (Xl by (6) and (3)
(8) (3 h) . PI C /12' (f.l2 v h] C 0:1 by (7) and *10.35 PM
(9) (3 h) . PI C (Xl by (8), '1C g . [g v h] Cj· ) ·fCj'
and *10.28 PM
(10) PI C (Xl by (9)
(11) S(b2, I, Pl, u) by H5 and 5.8
(12) S(a2, I, (Xl. z) by H4 and 5.8
(13) + +
II bl a2 e T . = .PI C (Xl by (11), (12) and 14.1
(14) II + b2 + a2 e T by (10) and (13)
Q.E.D.

16. Criticism of the alternative theory

It has been shown that a syllogism in Barbara with analogical middle

terms, defined according to the alternative theory, is a formally valid
formula. This is, however, the only advantage of this theory. Not even
all requirements of theology and metaphysics in regard to the syllogism
can be met by means of it. For a syllogism of these sciences has not only
analogical middle terms, but also analogical major terms; e.g. when we
write 'if every being is good, and God is a being, then God is good,'

111
 I. M. BOCHENSKI

not only 'being', but also 'good' must be analogical. But this means,
according to the alternative theory that H4 in § 15 should be replaced by

H so, instead of (3) we would obtain only

(3 h) • 1P2 v h] C [Q;2 v g]
which does not allow us to draw the conclusion (14). Neither can we try to
invert the order of '/' and 'g' in 15.1; in that case the syllogism would
become valid, but the major term in the conclusion would have an
alternative meaning, which can hardly be admitted.
Moreover, the theory has other inconveniences. First, the very definition
of analogy, as sketched in § 13, is highly unsatisfactory. By saying that
two names are analogical if and only if there is a third name meaning
alternatively the contents meant by both, we do not show any intrinsic
connection between the contents involved; and every couple of names
would be analogical, according to that definition, for we can always
introduce into our system a new name, meaning, according to definition,
precisely the said alternative. Secondly, there are serious gnoseological
difficulties. The situation with which we have to deal, is the following: two
names are given, and while we know the meaning of the first by direct
experience, we do not know in that way the meaning of the second. In
order to be able to use that second name correctly, we must supply it with
a meaning correlated in some way with the meaning of the first. Now the
alternative theory allows nothing of the sort: it only says how we can deal
with middle terms having alternative meanings, when both meanings are
already known.
These remarks do not lead to the complete rejection of the alternative
theory; but they seem to show that it is at least incomplete and should
be completed by another theory. The present author believes that this
was the position of Cajetan.

17. The isomorphic theory

This theory is based on the following considerations: the proportionaliter
eadem suggests that there is an identity, not between the contents meant
by both analogical terms, but between some relations holding between the

112
 ON ANALOGY

first (f) and its thing (x) on one side, the second (g) and its thing (y) on
the other. The texts of st. Thomas Aquinas are clear enough here. The
said relations are, however, not identical; this is also a traditional thesis,
strongly emphasized by all classical Thomists. We may therefore admit, as
a first approximation, that, while being non-identical, they are both
contained in the same relation. The definition of analogy of proportional-
ity would run, in that case, as follows:
17.1. Anp(a, h, I,f, g, x, y) . = Df• • Ae(a, h, I,f, g, x, y) .
(3 P, Q, R) . f Px . gQy . P #= Q . P c:: R . Q c:: R .
This is, however, not satisfactory. For if 17.1 would be the definition of
analogy of proportionality, there would be a material univocal element;
analogy would allow us to transfer to the other name some material
relations found in the meaning of the first. Now St. Thomas Aquinas
and tradition are quite clear as to the negation of such univocity. But
17.1 can be corrected by the affirmation that the common element in
both relations is formal, i.e. consists in the isomorphy of these relations.
The definition becomes:
17.2. An(a, h, I,f, g, x, y) . = Df• • Ae(a, h, I,f, g, x, y) .
. (3 P, Q) ·fPx· gQy. PsmorQ.
This is what we mean by 'isomorphic theory'. 1 It is strongly supported by
the fact that St. Thomas Aquinas uses for illustration of his doctrine
mathematical proportionality, the only mathematical function he possess-
ed and a function which makes one immediately think ofisomorphy.
One may think, perhaps, that if this be analogy of proportionality, the
meaning of our sentences about spirit, God etc., would be extremely
poor, indeed limited to some very few formal relations enumerated in the
Principia Mathematica. But this is not so. It is true that we cannot, as
yet, give exact formulations of many formal properties involved in
relations used by metaphysics and theology; the reason, however, is
not the lack of such formal properties, but the very undeveloped state of
biology and of other sciences, from which the metaphysician and the
theologian must draw his analogical names (and contents). An immense
progress in speculative sciences would arise out of a formalization of these
disciplines. And yet, even in the actual state of knowledge, where only

1 This definition is under the danger of trivialization; a better definition would be one
by which the quoted relations P and Q are contained in the meaning of a respectively
b in I (this proposal was made by P. Lorenzen).

113
 I. M. BOCHENSKI

mathematics, i.e. the poorest of all sciences, is formalized, we can show,

e.g., the difference between the principle and the father by purely formal
means - as, evidently, the first is transitive, the second intransitive.

18. The existential interpretation of the mode Barbara

If the isomorphic theory is admitted, a peculiar interpretation must be

given to the mode Barbara with analogical middle terms. Let us consider
the following substitution: 'if all being is good, and God is a being, then
God is good.' According to the isomorphic theory the only common
element meant by the two 'being' and the two 'good' is a product of
some formal relations, say P in the first case and Q in the second. But if
it is so, the major must be interpreted as follows: 'for all x: if there is an
f such that f Px, then there is a g such that gQy'; the minor will be inter-
preted in the same manner by the formula 'for all x: if there is an h such
that hRx, then there is anf such thatf Px'. From this we draw the conclu-
sion 'for all x: if there is an h such that hRx, then there is a g such that
gQx.' This would mean: 'if there is an x such that h is the Divinity of x,
then there is a g such that g is the Goodness of x.' The law used here is:
18.1. (x)· (3f)f Px::> (3 g) gQx: (x) . (3 h) hRx ::> (3f)fPx:
::> : (x) . (3 h) hRx ::> (3 g) gQx .
This is a correct formula of the logic of predicates.
The remarkable result of the existential interpretation is that the Thomistic
idea of analogy becomes sharply formulated in a very anti-univocal sense.
For, we do not know, as a result of our reasoning according to 18.1,
anything except that there is something (undetermined as to the content)
which has to God the set of quite formal relations Q. And yet, the talk
about God's goodness is clearly meaningful; moreover rigorous demon-
strations concerning it are possible.

19. The validity of the syllogism with analogical middle terms according
to the isomorphic theory

We are going to show now how, in such theory, a syllogism in Barbara

is a formally valid formula. We meet here, however, two formal difficul-
ties.
First we note that isomorphy, being a relation between two relations,

114
 ON ANALOGY

cannot be, as such, treated as a relation in which these relations are

contained; now this seems to be necessary if we wish to construct a
correct syllogism with analogical middle terms, interpreted according to
the isomorphic theory.
This difficulty may be, however, obviated in the following manner.
Isomorphy implies the identity of a series of formal properties of the
relations involved. These formal properties are different in each case of
couples of isomorphic relations; but for each of them in concreto a
product of such properties may be determined. E.g., in some cases both
relations will be included in diversity and will be transitive; in other cases
they will be intransitive and asymmetric etc. Now each of these properties
may be conceived as a relation in which the given isomorphic relations
are contained. This can be done by introducing into the system the name
of a new relation, which is treated as a primitive term, but whose meaning
is determined by an axiom. E.g. for symmetry we will put a relation Sand
determine the meaning of'S' by the axiom (x, y): xSy· =. xSy= xSy.
The product of such relations would constitute the relation in which
both isomorphic relations are contained. l
The other difficulty is strictly operational. It will appear that we shall
need an expansion of our 17.2 in order that the name of the common
relation R, in which the relations P and Q are contained, might be treated
as an argument of 'Anp'. If so, a new relation must be defined, namely an
octadic relation containing as terms, besides the seven stated in 17.2,
also R. We shall define it as follows:
19.1. Anp(a, b, I,/, g, x, y, R) . = DJ. • Ae(a, b, I,/, g, x, y) .
. (3 P, Q, R) . f Px . gQy . P # Q . P # R . Q # R .
. P C! R . Q C! R . ReForm.
By 'Form' we mean the class of all formal relations, as described in § 17.
There will be three laws of translation, analogous to 14.1:
19.2. Anp(ml, m2, 1, Jll, Jl2, x, y, P) . Anp(al, a2, 1, <Xl, <X2, Z, t, Q) : :>
+ +
:> : II ml al e T· =. (x) . (3f)f Px :> (3 g) gQx .
19.3. Anp(bl' b2, 1, Pl, P2, u, V, R) . Anp(ml' m2, 1, Jll, Jl2, x, y, P): :>
+ +
:> : II bl m2 e T· =. (x) . (3 h) hRx :> (3f)f Px .

1 The author is conscious that the proposed solution is highly un-orthodox; he would
be glad to find anything better. It must be remembered, however, that the whole
difficulty is purely operational; it seems intuitively evident that once there is a common
property, the syllogism is valid.

115
 I. M. BOCHENSKI

19.4. Anp(bl' b2,I, Pl, P2, U, V, R) . Anp(al' a2,I, <Xl. <X2, Z, t, Q) : ::>
::> : II + b2 + a2 e T· =.(x) . (3 h) hRx ::> (3 g) gQx .
Our hypotheses are
H1. II + ml + al e T
H2. II + +
bl m2 e T
H3. Anp(ml, m2,I, Ill. 1l2, x, y, P)
H4. Anp(al. a2, I, <Xl. <X2, Z, t, Q)
H5. Anp(bl' b2, I, Pl, P2, U, V, R) .
The proof of 'II + b2 + a2 e T' runs as follows:
(1) (x) . (31)/ Px ::> (3 g) gQx by H3, H4, HI and 19.2
(2) (x) . (3 h) hRx ::> (31)/ Px by H5, H3, H2 and 19.3
(3) (x) . (3 h) hRx ::> (3 g) gQx by (1), (2) and 1B.1
(4) + +
II b2 a2 e T· =.(x) (3 h) hRx ::> (3 g) gQx
by H5, H4 and 19.4
(5) by (4) and (3)

20. On analogy in recent logic

While the classical Thomists used analogy in ontology and theology,

but not in logic, recent writers seem to make a constant use of it in
formal logic. We noticed already that the authors of the Principia
Mathematica re-invented the very name used for analogy by the Thomists
(§ 7) and that analogy appears in the construction of semantics (§ 9).
The last phenomenon is connected with the theory of types. It is known
that, in order to avoid contradictions, we are bound to divide all objects
treated by logic (or all logical expressions) into classes called 'types'.
The formulae used in each type have quite a different meaning, but
exactly the same structure as the formulae used in another. This means
that the formal properties involved are identical i.e. that we have to do
with analogy, at least if the isomorphic theory is accepted.
The question arises as to why analogy has penetrated the domain of
formal logic. The answer seems to be given by the theory of Prof.
H. Scholz, who says that recent formal logic is nothing else than a part
of classic ontology.l As a matter of fact, recent formal logic generally
deals, not with rules, but with laws of the being in its whole generality;

1 H. Scholz. Metaphysik als strenge Wissenschaft, KOln 1941.

116
 ON ANALOGY

most of the laws contained in the Principia Mathematica, e.g., as opposed

to metalogical rules, are such laws. If this is so, it is not to be wondered
at that some consideration must have been given to analogy, for 'being'
is an analogical term and so are the names of all properties, relations, etc.,
belonging to being as such.
One curious feature of these developments is that the highly trained
mathematical logicians who had to speak about analogy, spoke about it
in a very loose and inexact way. What, for example, the Principia Mathe-
matica contains on the subject is far more rudimentary than the classic
Thomistic doctrine. Yet, recent formal logic, once applied to the lan-
guage itself, supplies superior tools for the elaboration of that notion.
The present paper is believed to contain only a very small sub-class of
the class of theorems on analogy, which may and should be elaborated
by means of recent formal logic.

117