Anda di halaman 1dari 107

DOXA LIBRARY

GYULA KLIMA

ARS ARTIUM

Essays in Philosophical Semantics, Mediaeval and Modern

Institute of Philosophy

Hungarian Academy of Sciences

Budapest, 1988

1
WHY "ARS ARTIUM"?;1

Dialectica est ars artium, ad omnium methodorum


principia viam habens, sine qua nulla scientia
perfecte scitur.2
As is evident also from this celebrated opening formula, mediaeval authors of logic tracts
offering a captatio for their subject matter pointed to the universality and utility of logic in
respect of the acquisition and organization of knowledge in several branches of science. But if
mediaeval scholars, with their almost monolithic hierarchy of sciences,3 felt so much the need of
such a universal art, an analytic tool that by its unifying methods and principles is able to supply
a cohesive force, as it were, to the whole body of sciences, then - per locum a minori - we,
scholars of the 20th century, should feel even more the need of such an art, with our well
developed, but utterly fragmented science, in which the unity of methods and principles in some
cases characterizes hardly more than single schools of special fields. For nowadays, I think, no
serious scholar could deny that the scientist of our age would deserve for at least equal reasons
Avicenna's reprimand addressed to Galen, who, says Avicenna, "knows many branches of
science, but does not know its roots".4
On the other hand, is it still possible, indeed, desirable at all, to have such a universal art and,
by its help, a unified science? And even if so, why should it be logic, rather than, say, physics, let
alone metaphysics or theology, that could supply the desired "unifying perspective"?
Well, as far as the question of desirability is concerned, we can say that if science is to be true
- and what else should science be, if science is to be a kind of knowledge, as opposed to mere
opinion? -, then it clearly cannot be inconsistent. So unity of science is required at least up to the
degree of consistency. But already this weak logical relation, consistency, implies a kind of unity
nowadays too scarcely met with, namely logical commensurability.
And this remark leads us immediately to the question of the very possibility of the unity of
contemporary sciences. For the real trouble with our "fragmented" science (as an integral part of
our fragmented culture) is not that it is inconsistent, but rather that for the most part we do not
even know whether it is inconsistent; indeed, the real trouble is that due to the immense diversity
of principles, methods and basic concepts involved, perhaps we cannot know this even in
principle. There simply seem to exist no intelligible lines of communication between such distant
parts of contemporary science as, e.g., molecular biology and political philosophy, or quantum
physics and systematic theology. But even further, why should it be only contemporary Western
science that could be considered here? What could we say about the logical relations between,
1
This essay is intended to serve as an informal description of the general project that "holds together" the otherwise self-contained, rather
technical essays of this volume. As such, it necessarily touches upon certain very broad topics in the philosophy of science, a subject matter of
volumes of philosophical work in its own right, which, as a matter of course, cannot receive sufficient treatment within the confines of such a
short essay. So I will adopt the tactic of drawing freely upon the vast literature of the topic, but without mentioning the representatives of several
points of view, thereby trying to avoid justified charges of simplification or misinterpretation.
2
For several versions of the famous dictum, ultimately deriving from St. Augustine's De Ordine, II.13, see L. M. de Rijk, Logica Modernorum,
Assen, 1967, II-1, pp. 32-33, 412, 418, 428, 431, 435, 436; II-2, pp. 357, 379, 417; Peter of Spain, Tractatus, ed. L.M. de Rijk, Assen, 1972,
p.1. As the reader will notice, I quote this phrase with the same kind of liberty as my mediaeval colleagues.
3
See e.g. the all-embracing system of sciences in L.M. de Rijk, op. cit., pp.459-462. Cf. also St. Thomas Aquinas's system in his commentary on
Boethius's De Trinitate, q.5., and his system of logic in his Prooemium to his Commentary on Aristotle's Posterior Analytics.
4
See M. A. Hewson, Giles of Rome and the Medieval Theory of Conception, London, 1975, p.47. See also p.56. n.19.

2
say, the principles of contemporary Western medicine and those of acupuncture? And what about
diachronical comparisons inevitably cropping up in historical sciences? The history of Western
philosophy and natural science shows so deep changes of conceptual structure that it seems
simply impossible to establish a unified framework within which the "relevant commensurations"
could be carried out.
Well, these are grave difficulties indeed. I think that it is precisely this immense conceptual
diversity, separating different parts of human experience accumulated up to the present, that
nowadays prevents any particular kind of science, be it physics, metaphysics or theology, from
providing the required "unifying perspective". For any particular kind of science presupposes a
particular kind of conceptual structure. So any other kind of science not sharing this particular
kind of conceptual structure will be beyond the reach of this purportedly universal science.
Whence, it cannot be universal. Indeed, a particular kind of science can be in such a position only
if the whole scientific community working within the same system of sciences shares a common
stock of basic concepts which secures undisturbed communication between workers of different
fields. This is why metaphysics and theology could play such a central role in the hierarchy of
medieval sciences, and this is why they gradually had to retreat from this position when the
conceptual unity of the schools was first only loosened, bit by bit eroded, and finally broken up
by the several spiritual and social movements from about the late fifteenth century up to the
present. So it seems that in our age no science can play the role of a universal science organizing
the results of other particular sciences into one intelligible whole.
What can we do then? Shall we produce apologetic ideologies of a pluralistic and relativistic
picture of science, treating sciences as convenient "myths" of local "tribes" with no justifiable
claim to truth, and convince ourselves that we, human beings, individuals of the same species
and inhabitants of the same planet, do in fact live in "different worlds" constituted by our tribal
superstitions with no hope of mutual understanding and intelligible communication? Or shall we,
instead, take an oath to promulgate one particular kind of science, which, of course, we regard as
universal, and, as good missionaries, try to convert the other, "savage" tribes to our only true
faith?
I think I am not alone with an emphatic "NO" to either of these alternatives. But is there any
possible third, an aurea medietas, between these two bad extremes?
Well, in the previous argument we have overlooked a - to my mind highly relevant -
distinction between different kinds of sciences drawn by the medievals. According to this
distinction, sciences like physics, metaphysics or theology, are real sciences intending to
describe extramental reality as it is. On the other hand, a rational science, like logic, treats of the
operations of reason as its proper subject matter, so it does not deal directly with extramental
reality, but only as it appears to the human mind, structured and organized by the operations of
reason.5
It is precisely in the spirit of this distinction that medieval scholars claiming a universal status
to logic distinguished between the universality of logic and a universal real science, metaphysics
or theology. For example, as an objection to the description of dialectic as the art of arts we find
the following: "the art of arts is theology; therefore it is not dialectic. By way of an answer we

5
Cf. texts referred to in nn. 2. and 3. above.

3
have to say that dialectic is the art of arts insofar as it presents a way to the other sciences without
which no other science can perfectly be known. Theology, on the other hand, is the art of arts in a
different manner, insofar as it treats of the First Cause, namely, God."6
Again, St. Thomas Aquinas, when discussing the relationships of logic to metaphysics and to
the other sciences, claims to logic the same scope of universality as to metaphysics, but adds that
logic has this scope only insofar as it deals with concepts used by the real sciences, and so also
with the concepts of the universal real science, metaphysics.7
Now I think it is with this conception of logic that we can take the above dilemma by its horns,
and can show a promising third between the two bad alternatives we were left with by the above
argument. For even if we concede that as a particular kind of science even logic presupposes a
particular sort of conceptual structure, still, as it deals with concepts as its proper subject matter,
in the case of logic this will not imply that different conceptual structures will necessarily be
beyond its reach, provided it is able to reconstruct these different conceptual structures by its own
conceptual tools. So with a logic so conceived we may be in a position to establish the required
conceptual bridges between different parts of contemporary and past science, whereby the
desired commensurability may be reached, at least in principle.
But we must take here very seriously that even logic, as a particular kind of science,
presupposes a particular kind of conceptual structure. Of course, as any other human science,
also logic has a history through which it has reached its present state. So if we want to use logic
as a universal conceptual tool, we must be keenly aware of the actual limitations imposed upon
the possible uses of this tool by its history. For if we consider the paradigm-constituting logical
theory of our days, namely classical quantification theory, as it is bequeathed to us in the works
of Frege, Russell and Tarski, in comparison to later developments as well as to earlier well
developed theories (in particular, to the sophisticated semantic theories of the medievals), then
we can see that although some very general features of the classical theory are firm constituents
of present day standards of what a logical theory should look like (minimally, it should possess a
recursive syntax and semantics in which the principle of compositionality is in force), it also has
some particular features which derive merely from the actual historical line along which it
developed. So in order to outline a way I think present day logic could be wrought into the shape
of an art of arts, a universal conceptual tool, let me single out some features of the nowadays
paradigmatic classical quantification theory which I take to be historically accidental and
imposing unnecessary limitations upon its possible uses as such a tool.
1. As is well-known, for Frege a general term of a natural language is a "Begriffswort", i.e., it
denotes a "Begriff", a concept. A concept is a special kind of function, namely a function from
individuals to truth values. Hence, the term denoting such a concept requires a referring
expression in its empty argument place to form a sentence. So a general term for Frege is
necessarily predicative in character, and, accordingly, in the canonical notation general terms are
represented uniformly by predicate letters. This, however, leads to several discrepancies between
the syntax of the canonical notation and that of natural languages. So, following this line, we
necessarily get involved in distinctions between grammatical, or "surface" structure and logical,

6
de Rijk, op. cit. II-1. p.435.
7
For detailed discussion and texts see R.W. Schmidt, The Domain of Logic according to Saint Thomas Aquinas, The Hague, 1966, pp.41-71.

4
or "deep" structure. This, of course, would not cause any serious difficulty in itself, but the
trouble is that on this approach there will be sentences of which we cannot construct the true,
"logical" form, for several sentences of natural languages are simply unrepresentable in this
theory. (Sentences with "most", or "more than half the", etc.)
Facing this difficulty, we can take linguistic structure at its face value (in the spirit of R.
Montague), and try to find a fitting semantics for the syntactic units of natural language
sentences. Indeed, if we treat noun phrases as denoting families of sets, we get the theory of
generalized quantifiers, which is able to overcome the difficulties met with by the classical
theory in representing natural languages. But this is done in this theory at the expense of setting
up a rather complicated and unintuitive semantics, in which, e.g., the noun phrase "Peter", as
opposed to the lexical item "Peter", denotes the set of sets containing Peter. But why should we
give up the simple, intuitive view that in "Peter walks" the noun phrase denotes Peter? Moreover,
why should we give up the simple and intuitive view that in the sentence "Every man walks", the
noun "man" stands for men, for the sake of treating the phrase "Every man" as denoting the set of
all supersets of the set of men?
If we follow the lead of the medieval logicians, who in their supposition theory attributed
referring function to general terms as well as to singular terms, then instead of the "type lifting"
tactic used by generalized quantification theory, we can use a "type lowering" tactic, according to
which general terms in their referring function belong in the same semantic type as proper nouns
and variables, that is, they pick up individuals, elements of the universe of discourse, as their
semantic values.8
Indeed, if we construe common noun phrases, i.e., general terms (whether simple or complex)
in referring position, and anaphoric pronouns with such antecedents as restricted variables, then
we get an extension of standard quantification theory that not only solves all problems of the
classical theory in representing natural languages, but which, at the same time, is true to
commonsense semantic intuitions as well as to the more sophisticated formulations of these in
medieval semantics.9
2. The Fregean view of concepts seems to provide us with very simple solutions to the tangled
problems involving the concept of existence. Existence is not a predicate of things, but a second-
level predicate of concepts: to say that a kind of thing exists is not to attribute a peculiar property
to a thing or to certain things, but to state that the corresponding concept is instantiated, i.e., that
for at least one individual thing the concept in question gives the True as its value. In this
analysis no paradoxes of referential tautologies or contradictions, the kind known to philosophers
since Plato's Sophist, can arise, since, despite elusive grammatical appearances, in existential
statements there are no referring expressions at all. Apparent counterinstances involving definite
descriptions or proper nouns are explained away by revealing the true, "logical" form: definite
descriptions are eliminated by Russell's method, while proper nouns are elevated to the rank of

8
For discussion of supposition theory see Essay II. of this volume. On semantic type changes in general see J. van Benthem, Essays In Logical
Semantics, Dordrecht, 1986, pp.68-71. For an excellent discussion of lowering on the level of syntax see P.A.M. Seuren, "Operator Lowering",
Linguistics, 22(1984) pp.573-627.
9
For more on this see Essay III.

5
verbs, so "Pegasus does not exist" will only mean that nothing pegasizes.10 Indeed, what else
could it mean?
Well, we can give an answer to this question if - again, lending an ear to the medievals - we
consider that referring, just like thinking or wanting, is an intentional act. So just as thinking or
wanting, also referring may have a nonexistent object. Or to put it more simply, just as we can
think of or want something that does not exist, so we can refer to something that does not exist.
As the medieval theory of ampliation teaches us, in several, as we nowadays would say,
"intensional", i.e., tensed, modal, or intentional, contexts, we do freely refer to and quantify over
objects of reference that do not exist. Indeed, in an appropriate formal semantics we can prove a
very general metatheorem showing the relationship between an existential statement, like "A
centaur exists", or "A thing that is destroyed exists", and the corresponding particular
quantification like "Something is a centaur", or "Something is destroyed". Note that of these
examples only the last one is true.11
3. Detachment of the above kind of the concept of existence from that of quantification, since
after this detachment we are not committed to talk only about existents, opens up the way of
returning to Aristotle's conception of truth in a much stricter sense than Tarski's semantics did.
For if we are entitled to talk also about nonexistents, then no sophistic paradoxes threaten the
definition of truth in terms of actual existence and that of falsity in terms of nonexistence.12
Moreover, in this way, if we reconsider the semantic function of predicates, we can give a
reconstruction of an Aristotelian theory of predication explaining truth in terms of the actual
inherence, or inexistence (inesse) of forms signified by predicates in particular things.
For according to this theory, the inherence theory of predication, the semantic function of
predicates is to signify particular, individualized forms, inhering in their subjects. It is the actual
existence of such a form that makes a predication true. For example, the sentence "Socrates is
white" is true at a certain time, if and only if Socrates's whiteness exists at that time. So on this
theory a predicate may signify numerically distinct forms in different individuals at the same
time, and at different times in the same individual. But the same predicate in the same individual
at the same time cannot signify different forms. So we can denote such a particular form as the
value of a function for these arguments, wherefrom we get the function itself by functional
abstraction, by abstracting from these arguments, the individualizing conditions of the form
signified by the predicate, namely time and subject. Let me call the function itself the
signification, while a value of this function a significate of a predicate. When this predicate
occurs in a sentence, then it is the subject term of the sentence that by its reference supplies the
first argument, and it is the time of the predication that supplies the second argument for this
function, and so the signification function determines which significate of the predicate has to be
actual so that the sentence be true.
Now this way of constructing the significate of a predication provides us with a model for
constructing the significates of complex propositions, indeed of any kind of complex expressions
in respect of certain things, wherefrom, by functional abstraction we can get the signification,
10
By the way, along these lines, if "surface", i.e., ordinary grammar is so negligible, then even the incriminated "The Nothing itself nothings"
("Das Nicht nichtet") should be perfectly acceptable.
11
For details see Essay IV.
12
For more on this topic see M. Matthen, "Greek Ontology and the `Is' of Truth", Phronesis, 28(1983), pp.113-135.) See also Essay VI.

6
that is, the meaning of these expressions. For example, suppose we identify the significate of an
atomic predication `Px' according to an assignment of variables at time t with the significate of
its predicate in the value assigned to x, at time t. On this basis, if we already know the
significates of atomic sentences we can construct the significates of negated sentences in the
following manner. Syntactically, sentential negation is a logical functor which requires in its
argument-place a sentence to form a sentence. Semantically, it changes the truth value of a
sentence to the opposite value. So the signification of sentential negation should be a function
which, when its argument-place is filled in with a sentence-significate, gives as its value a
sentence-significate that is actual if and only if the argument is not actual. Whence, the
significate of a negated sentence according to an assignment of values at a time t is the value of
the signification of negation in the significate of its argument-sentence according to this
assignment at time t.
Along these lines, by defining the signification of primitive expressions, which, in keeping
with the rule of compositionality, determines the significates of complex expressions built up
from them, we can construct the significates of any complex expression, in particular, of
sentences, relative to a referential frame containing both actual, and several kinds of non-actual
objects of reference. So we can accept Aristotle's definition of truth as concerning not only
atomic sentences, but also any kind of complex sentences, according to which a sentence is true
if and only if what it signifies exists.13
As can be seen, by this method, regarded as a general means of constructing the semantic
values of expressions in whatever grammatical category, we get sufficiently fine-grained
intensions to define an intuitive and adequate synonymy relation: two expressions are
synonymous if and only if they have the same signification.
This, however, need not mean that we have, despite the Duhem-Quine thesis,14 a once and for
all established synonymy relation and, correspondingly, a once and for all established set of
analytic truths in natural languages. As I have insisted above, even logic has a history, even logic
cannot be regarded as a depository of eternal truths. For example, for Buridan the statement "No
chimaera exists and some chimaera does not run" is analytically true, while for Russell it is
contradictory, although both were sane, acute logicians of their age. The reason is that in their
respective theories they judge differently the import of the negation in the second conjunct of this
conjunction.
The lesson of this, I think, is the following. Different people may have different thoughts
corresponding to the same syntactic structure. So what they mean by an expression is not so
much dependent on the objective meaning of the expression but on the subjective concepts they
attach to the expression in question.
Indeed, in an appropriate formal semantics we can model this situation by constructing the
subjective, mental significates of expressions relative to the subjective representational system of

13
For technical details see Essay V. For historical references see Essay VI. Cf. also the following text from Cajetan's Commentary on Aristotle's
Categories: "Et adverte hic diligenter quod illa maxima Aristotelis hic posita: `ab eo quod res est vel non est oratio dicitur vera vel falsa', non
intelligitur de re quae est subiectum aut praedicatum orationis, se de re significata per ipsam orationem, verbi gratia: cum dicitur homo est albus,
non ideo est vera ista quia homo vel album sit, sed quia hominem esse album est: hoc enim est significatum per illam orationem." Thomas de
Vio Caietanus, Commentaria in Praedicamenta Aristotelis, Romae, 1939, p.87.
14
Cf. S.G. Harding (ed.), Can Theories Be Refuted? - Essays on the Duhem-Quine Thesis, Dordrecht, 1976.

7
a human mind by a method exactly parallel to the one by which we have constructed the
extramental significates of the same expressions relative to a referential frame.
But what is it, then, that gives logic its objectivity? Well, of course, social interchange of
information, communication. Logic, although not a depository of eternal truths, is the depository
of rules which govern intelligible communication in a given linguistic community, which any
kind of subjective concepts of a member of the same community should comply with so that he
or she be understood.15 (After all, when Buridan and Russell made their respective claims, they
followed an established scientific usage of their respective scientific communities, and reinforced
it further by their respective theories. This is why neither of them was considered insane by his
contemporaries.)
As Buridan excellently puts it: "We have to say in brief that as there are propositions,
expressions and terms which are either mental, or vocal, or written, Aristotle in this book treats
only of the vocal ones, because in logic we have to use disputations; and because to determine
the nature of concepts pertains to psychology, or to metaphysics, whence for logic it remains that
it directs the use of words corresponding to concepts in valid arguments and congruous speech.
And so every noun of which we treat here is vocal. But you may ask: how is it that these vocal
nouns and verbs are applied at will (ad placitum), at my will, or at yours? I answer that there are
nouns and verbs that signify the same and in the same way for the same whole community: Latin
words for every Roman, and French words for every French. And it is not in my power to change
this common signification, but it was in the power of the one, or ones who created this language,
who, at their will assigned these significations to these words, as even now those who agree
among each other are able to fabricate at their will their own language, as it happens with those
who speak slang. Indeed, even I, when I dispute or teach something, give names to things at my
will when I say: let a be the major, b be the minor and c be the conclusion. For I could do
differently if I wanted."16
So in logic we deal with these common, "objective" significations, which gain their objectivity
from the rules governing linguistic usage in a community, co-ordinating, as it were, the
subjective concepts of the members of the community. 17 And so, when usage changes, as the
form of life of the community changes, then this changes also significations, and, consequently,
the set of analytical truths of the language used by the community.
So far, so good, one might say, but are we not too much committed, despite all our verbal
liberalism, by the very construction of this semantics, to one particular kind of metaphysics,
namely Aristotelian metaphysics, with its inherent forms, souls, as a special kind of forms, and
their forms, namely concepts? On the other hand, by relativizing even logic to the vicissitudes of

15
See again Essay V. and Essay VII.
16
Buridanus: Questiones Longe super Librum Perihermeneias, ed. by Ria van der Lecq, Utrecht, 1983, lb. I. q.3.; cf. J. Pinborg, "The
Summulae, Tractatus I De Introductionibus", in: J. Pinborg (ed.), The Logic of John Buridan, Copenhagen, 1976.
17
Whether "behind" this kind of "social objectivity" there is some more basic objectivity inherent in human nature (Kant's answer) or in the
nature of things (the answer of Plato and "realist" Aristotelianism) is a further, metaphysical question, which should not be determined by a
semantic theory. For this kind of interpretation of "nominalism", "conceptualism" and "realism" see E.A. Moody, "The Age of Analysis", in:
Studies in Medieval Philosophy, Science and Logic, Berkeley-Los Angeles-London, 1967, pp. 305-320. On the obligatory "aloofness" of logical
theories see D.P. Henry, That Most Subtle Question - (Quaestio Subtilissima): The Metaphysical Bearing of Medieval and Contemporary
Linguistic Disciplines, Manchester, 1984. (See entry "aloofness" in the index.)

8
history, are we not committing ourselves even more strongly than others to the "tribal-
superstition-view" of science?
Well, as for the metaphysical commitment of this system I think we should admit that even if
we are not committed to some curious sort of existence of non-actual things by merely referring
to and quantifying over them, still, we may seem to countenance too many existents. Surely those
philosophers "who have a taste for desert landscapes" would be reluctant to admit such things in
their universe as Socrates's whiteness, or Plato's love towards Socrates, or the fact that Socrates
drank hemlock, which, on the present theory, are actual significates of the corresponding
expressions.
Now to this we can say first of all that already traditional Aristotelianism had its own ways to
reduce the ontological commitments of its own logical theories.
On the one hand, there was the famous distinction, most clearly drawn by Saint Thomas
Aquinas, between two basic kinds of entities, namely between entia realia, real beings, and
entia rationis, rationate beings. Real beings are objects of reference that there are without any
operation or consideration of human reason. Rationate beings, however, are objects of reference
that there are as constituted by the operations of reason. Instead of going into the niceties of this
distinction we can illustrate it by saying that a rationate being is to a real being as the average
citizen with his two and a half children is to Mr Smith with his two children, or as the fact that
Socrates runs is to Socrates, who happens to run.18
On the other hand, there was the Ockhamist "tactic" of reducing the kinds of things admitted in
one's ontology by identifying the semantic values of different expressions which seem to pertain
to different ontological categories. By identifying the significates of terms with their supposita,
relations with their fundaments, and quantity with substance, Ockham reduced the really distinct
ontological categories to two, namely to that of substance, and that of quality; and by identifying
the significates of true propositions with supposita of corresponding complex terms, Buridan
eliminated the distinct category of complexe significabilia.19
Now these ontological reductions are properly reflectable in the formal semantics outlined
above. Of course, as the semantic theory of itself does not tell us which expressions have their
significates in this or that category, or which expressions have distinct significates and which
have the same, the different formal models constructible to the same language with the same
grammar will represent radically different metaphysics, with radically different conceptual
structures. Indeed, the abstract, mathematical character of this semantics, as any domain of a
referential frame may be empty, or overlapping with others, provides us with wide possibilities in
this respect.
For example, by placing significations into the domain of real beings we get a Platonic heaven
of really existing ideas, while placing them among rationate beings we get an Aristotelian
metaphysics of universals. But even further, if we place the significates of all expressions in the
domain of rationate beings, then we get a mind-constituted Kantian world. On the other hand, by
simply taking the non-actual domains of the referential frame empty, identifying the range of our

18
For more on this topic see Essay VI. and the texts referred to therein.
19
See end of Essay II. and relevant notes of Essay VI. and VII. Cf. also W. Ockham: Summa Logicae, St. Bonaventure N.Y., 1974, Pars I, cc. 6-
10, 15-17, 44-62, Pars II. c.27; Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, pp.32-34; Buridanus, Quaestiones in
Metaphysicam Aristotelis, Minerva, Frankfurt a.M., 1964, lb.4.q.X.

9
variables with the domain of existents, defining the extensions of our predicates by means of
their significations, and defining truth of an atomic predication in terms of elementhood in this
extension (for details see Appendix of Essay V.) we get back the ontology of classical
quantification theory.
Moreover, as the construction of this semantics is highly free from syntax, in the sense that the
same kind of semantics can be constructed to languages with widely different syntax, by the help
of this theory we may be able to model the ways the structure of language constitutes the
structure of reality.
But then, indeed, are we not committed by this seemingly unlimited freedom to a highly
relativistic view of science? Well, definitely not. For just as this semantics does not commit us to
one particular picture of reality, so it does not commit us to a plurality of realities. By its
multifarious possibilities of constructing several pictures of reality and conceptual structures, it
only provides us with means of comparing these structures, and formulating intelligible pros and
cons. Indeed, it is only in this way that it can serve as a universal art, which, as such, does not
dictate to metaphysics or to users of a language, but which is a prerequisite to intelligible, exact
discussions, and investigations into the nature of the field of real sciences.20
As the medievals said, "in acquisitione scientiarum dialectica debet esse prior", so
metaphysical investigations should follow only after the study of arts. But then they have to
follow. As another medieval saying goes: "non est canescendum in artibus" - one should not
grow old in the study of arts.

20
Cf. again Henry on aloofness. See n.17. above.

10
THE SQUARE OF OPPOSITION, COMMON PERSONAL SUPPOSITION AND
THE IDENTITY THEORY OF PREDICATION WITHIN QUANTIFICATION
THEORY

1. INTRODUCTION

In his "Ockham's Supposition Theory and Modern Logic", Gareth B. Matthews argued, I think
convincingly, against Philotheus Boehner's claim that "a particular, affirmative, categorical
proposition about a state of affairs (propositio affirmativa particularis de inesse) is interpreted by
the scholastics in exactly the same manner as by modern logicians".1 According to Matthews's
conclusion "Ockham and the moderns are not free to agree on the interpretation of any
categorical proposition - whether A, E, I or O".2
On the basis of Matthews's argument, the main points of disagreement can be gathered as
follows:
1. On the modern interpretation the propositions in question concern everything in the
universe, whereas on the medieval interpretation they concern only things falling under their
subject terms.3
2. As a consequence of this, on the modern interpretation the matrices following the
quantifiers in formulas representing A and I propositions differ in their logical form, while on the
medieval interpretation these propositions differ only in their quantity.4
3. On the modern interpretation an A proposition, having a conditional matrix, does not have
existential import, while on the medieval interpretation it does.
4. On the medieval interpretation, however, an O proposition does not have existential import,
whereas on the modern interpretation it does.
5. A and E propositions on the modern interpretation can, while on the medieval
interpretation, being contraries, cannot both be true at the same time.
Reacting to Matthews's paper, D.P. Henry has pointed out that, if we take "modern logic" as
including also Lesniewski's Ontology, these differences of interpretation can be eliminated.5
However, in view of the radical differences between the foundations of Lesniewski's system and
those of standard quantification theory, even if "modern logic" in this way can be said to have
taken a step toward the medieval interpretation, in the guise of the former, still, quantification
theory as it stands remained where it was.6
In what follows, therefore, I try to make quantification theory take this step. More precisely, I
try to show that with the aid of certain, rather trivial, supplementations and modifications, first

1
G.B. Matthews: "Ockham's Supposition Theory and Modern Logic", The Philosophical Review, 73(1964), No.1.; Ph. Boehner: Medieval
Logic, Manchester, 1952, p.29.
2
ibid. p.99.
3
Or, as Matthews puts it: "... Ockham quantifies over terms whereas modern logicians quantify over variables". p.99.
4
"... we cannot quantify over variables and still think of the corresponding A and I propositions as differing only in quantity." Matthews, op. cit.,
p.98.
5
See D.P. Henry: "Ockham, Suppositio and Modern Logic", Notre Dame Journal of Formal Logic, 5(1964), pp.290-292., later incorporated in
III.1. of his Medieval Logic and Metaphysics, London, 1972.
6
Since the whole issue concerns the interpretation of the categoricals, throughout this paper by "quantification theory" I mean a semantic theory.

11
order quantification theory with identity can be rendered able to represent fully the medievals'
analysis of categorical propositions with respect to their logical interrelations and to the
supposition of their terms. In fact, the modifications needed are very simple and, I think, also
intuitive: first, instead of the unrestricted variables of standard quantification theory, we have to
use restricted variables, i.e., variables, which take their values not from the whole universe of
discourse, but from the extension of an open sentence constituting the matrix of such a variable
(such a variable may then represent a common term of medieval logic, whether simple or
complex); second, as the extension of such a matrix can be empty (in the same way as that of a
common term), we have to introduce into our semantics an appropriately chosen zero-entity
representing the semantic value of empty terms.7 For the representation of the so-called
"suppositional descents" we have to introduce even further devices, which will, however, be
detailed later. Now let me call the language of first order quantification theory thus modified and
supplemented MPL (Medieval Predicate Logic).

2. MPL SYNTAX

The language MPL is defined as follows:


MPL:= <C,P,V,N,T,F>,
where C:={-,&,E,=,(,),.}, P is a denumerably infinite set of parameters, V is a denumerably
infinite set of variables (proper variables, as I shall call them)8, N is the set of arabic numerals of
positive integers, T is the set of terms and F is the set of formulae of MPL.
P detailed: P:=Pind U Ppred, where Pind is the set of individual parameters and Ppred is the
set of predicate parameters of MPL.
The set of terms is defined as follows. T:=Pind U Itm U Var, where Itm, the set of what I call
indexical terms, is defined in the following manner. If vV, iN and AF, then A(v/i)Itm,
where A(v/i) is the result of substituting i for v in all of its free occurrences in A.9 (Intuitively,
these indexical terms are designed to represent expressions like "this man", or "this lover of
Plato", or "this lover of every man", etc.)

7
Indeed, the idea of operating with restricted quantification in representing the medieval analysis of the categoricals (undoubtedly, for the reason
of being so plausible) is not a new one. See e.g. G. Priest -S. Read: "Merely Confused Supposition", Franciscan Studies, 40(1980), pp.265-297;
H. Weidemann: "William of Ockham on Particular Negative Propositions", Mind, 88(1979), pp.270-275; J. van Eijck: "Generalized Quantifiers
and Traditional Logic", in: J. van Benthem-A. ter Meulen: Generalized Quantifiers in Natural Language, Foris Publications, 1985; M. Brown:
"Generalized Quantifiers and the Square of Opposition", Notre Dame Journal of Formal Logic, 25(1984), pp.303-322. However, these attempts,
failing to introduce a zero entity as the semantic value of empty terms, have considerable difficulties in representing the medieval treatment of
propositions with empty subjects. For good arguments that we cannot afford to ignore the medievals' treatment of empty terms see A. Broadie:
Introduction to Medieval Logic, Oxford, 1987, p.120. Note that the system I am about to present validates all rules deployed here by Broadie
without excluding either empty or transcendental terms.
8
After the usage of M.M. Prullage in her "A Theory of Restricted Variables without Existential Assumptions", Notre Dame Journal of Formal
Logic, 17(1976), pp.589-612.
In general, if vVar, AF, in which some occurrence of vV is free and jNUItm, then A(v/j) is the string of signs that results from A by
9

substituting j for v in all of its free occurrences in A. An occurrence of a variable v in a string of signs S is bound, iff it appears in a well-formed
substring of S of the form: `v.A', or `(Ev)(A)'; otherwise this occurrence is free. We shall need this more general rule in the reconstruction of
Ockham's suppositional descents.

12
Var, the set of variables of MPL is defined as follows: Var:=V U Vres, where Vres is the set of
restricted variables of MPL. Vres is defined in the following manner: If vV and AF, then
`v.A'Vres, in which v is called the operator variable, and A the matrix of `v.A'.10
The set of formulae of MPL is defined by the following clauses:
/F1/ If t1,...,tnT and PnPpred, then Pn(t1)...(tn)F.
/F2/ If t1, t2T, then (t1=t2)F
/F3/ If A,BF and vVar, then ~(A), (A & B), (Ev)(A)F
For the sake of convenience the following abbreviation rules may be applied:
/Abbr1/ The matrix of a restricted variable may be omitted in all but one of its occurrences in a
formula, provided different restricted variables have different operator variables in the
same formula.
/Abbr2/ If A,BF, then (A v B), (A -> B), (A <-> B) abbreviates (as usual): ~(-A & -B), ~(A & -
B) and ~(A & -B) & ~(B & -A), respectively.
/Abbr3/ If (Ev)(A)F, then (v)(A) abbreviates ~(Ev)~(A).
/Abbr4/ If no confusion arises, parentheses may be omitted.

3. MPL SEMANTICS

The definition of a model for MPL runs as follows.


M:=<W,R,I,0>,
where W, the domain of M, is a nonempty denumerable11 set, R is a function assigning
semantic values to the parameters of MPL, I is a function from subsets of W to index functions,
(i.e., if B is a subset of W, then I(B)=IB, where IB will be defined later) and 0 is the zero-entity,
the semantic value of empty terms, which is not an element of W.12
R is defined by the following clauses:
/R1/ If aPind, then R(a)W.
/R2/ If PnPpred, then R(Pn) is a subset of Wn.
An assignment of values f (a function from T U F to W U {1,0} ) is defined as follows.
/f1/ If aPind, then f(a)=R(a).
/f2/ If vV, then f(v)W.
For defining f for indexical terms, first we define for every subset B of W an index function IB
in the following manner.13
If i is a positive integer, then
/i/ IB(i)B, if B is not empty, and its cardinality is greater than, or equal to i
10
As from their omission no confusion arises, for the sake of simplicity henceforth I will omit quasi-quotes.
11
The restriction of the possible domains of M to denumerable sets is needed for the correct representation of suppositional descents. In view of
the Löwenheim-Skolem theorem this stipulation means no severe restriction on the theory.
12
Merely for simplicity's sake, 0 will also serve as the semantic value of false formulae, but no philosophical depths are to be looked for in this
choice.
13
As will be seen from the definition, an index function assigns positive integers to individuals in subsets of W. This will serve to identify them
as referents of indexical terms. E.g., the i-th individual from the set of men will be the referent of "this man", when "this" is associated with the i-
th act of pointing at a man. If there are n men, then the first man will be associated with all numbers greater than n. So the index function will
give numbers to individuals as if it were counting: first, second, third,... pointing at different individuals,...n-th; and if there are no more
individuals, then further: (n+1)-th ... and so on, pointing always at the first one.

13
/ii/ IB(i)=IB(1), if B is not empty and its cardinality is less than i
/iii/ IB(i)=0 otherwise.
/iv/ For any i less than or equal to the cardinality of B, IB is a one-one function.
Now with this definition of IB at hand, the definition of f for indexical terms is the following:
/f3/ If A(v/i)Itm and i is the positive integer denoted by i, then f(A(v/i))=IE(v,M,f)(A)(i),
where E(v,M,f)(A), a subset of W, namely the extension of A in M according to f in respect
of v, is defined as follows:
E(v,M,f)(A):={uW: f[v:u](A)=1},
where f[v:u] is an assignment that is the same as f except that it assigns u to v (that is to say,
f[v:u](w)=f(w), if w is not v, otherwise f[v:u](w)=u, i.e., f[v:u](v)=u), and f(A)=1, which reads:
A is true according to (or, is satisfied by) f will be defined below.
The definition of f for restricted variables runs as follows:
/f4/ If v.AVres, then
/i/ f(v.A)=f(v), if f(A)=1
/ii/ f(v.A)=0 otherwise.
For formulas f is defined by the following clauses. (For the interpretation of the syntactic
symbols see clauses /F1/-/F3/ above)
/f5/ f(Pn(t1)...(tn))=1, iff <f(t1),...,f(tn)>R(Pn)
/f6/ f(t1 = t2)=1, iff f(t1)=f(t2)W
/f7/ f(-A)=1 iff f(A)=0
/f8/ f(A & B)=1 iff f(A)=f(B)=1
/f9/ f((Ev)(A))=1 iff for some uRg(v)f, f[v:u](A)=1, where Rg(v)f, the range of the
variable v with respect to f is defined as follows: Rg(v)f={uW: for some f' that differs from f
at most in the value assigned to v and its operator variable, f'(v)=u}, if this set is not empty,
otherwise Rg(v)f={0}14
/f10/ f(A)=1 iff f(A) is not 0
Truth in a model M is defined in the following manner:
A is true in M iff for some f in M, f(A)=1.
As usual, a formula of MPL is valid iff it is true in every model and an inference A  B is
valid iff the formula ~(A & -B) is valid.

4. THE SQUARE OF OPPOSITION

Within MPL a "tidy square of opposition"15 can be constructed in the following way:

14
Note here that Rg(v)f is never empty: it is either {0}, if the extension of its matrix is empty, otherwise it is identical with this extension. Proof:
Suppose v is of the form: `x.A'. If the extension of A in respect of x in M according to f is empty, then, by definition, there is no u W such that
f[x:u](A)=1, whence, for any f' differing from f at most in the value assigned to x, f'(A)=0, and so f'(x.A)=0 too. Hence, for any f' differing from f
at most in the value assigned to x and x.A, f'(x.A)=0, and so, since 0 is not an element of W, Rg(x.A)f={0}. On the other hand, if E(x,M,f)(A) is
not empty, then for any uW, and for any f' differing from f at most in the value of x, u E(x,M,f)(A) if and only if f'(A)=1 and f'(x)=u. But
then f'(x.A)=u iff uE(x,M,f)(A). So for any f' differing from f at most in the value of x and x.A, f'(x.A)=u iff u E(x,M,f)(A), whence
E(x,M,f)(A)=Rg(x.A)f.
15
The expression is from Matthews's "Suppositio and Quantification in Ockham", Nous, 7(1973) p.21.

14
A: (x.)(F(x.Gx)) E: (x.)~(F(x.Gx))

contraries

s s
u u
b b
a a
l l
t contradictories t
e e
r r
n n
s s

subcontraries

I: (Ex.)(F(x.Gx)) O: (Ex.)~(F(x.Gx))

In this formulation the points of disagreement listed above do not exist any more:
1. In this analysis the four categoricals concern only things falling under their subject terms in
that the corresponding restricted variables in these formulas take their values only from the
extension of the open sentence of which they are formed, not from the whole universe. Indeed,
the definition of the assignment function f for restricted variables follows to the letter Buridan's
definition of personal supposition of common terms (in his Sophismata, c.3.)16, according to
which a term stands for those things of which it is true, if there is any such thing. (For more on
this see sect. 10. below.)
2. The matrices following the quantifiers in A and I (and also in E and O) formulae have the
same form, so these formulae "differ only in quantity".
3. The A formula, having a "categorical matrix", does have existential import, i.e., the
following inference is valid in MPL: "(x.)(F(x.Gx))"  "(Ex)(Gx)". For if the conclusion is false
in a model M, then this means that R(G) is empty in M. But then, for any f, f(Gx)=0, and so,
again, for any f, f(x.Gx)=0, whence Rg(x.Gx)f={0}. So no element of Rg(x.Gx)f is an element of
R(F). But A would be true only if all elements of Rg(x.Gx)f were elements of R(F). So in such a

16
Cf.: "est autem suppositio prout hic accipitur acceptio termini in propositione pro aliquo vel pro aliquibus quo demonstrato vel quibus
demonstratis per ista pronomina `hoc' vel `haec' vel equipollentia illis, terminus vere affirmatur de isto pronomine mediante copula illius
propositionis" Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, p.50

15
model A cannot be true. Therefore, since the falsity of the conclusion implies the falsity of the
premiss, the above inference is valid. Q.e.d.
4. As the A formula has existential import, therefore, if its subject term is empty (i.e.
"(Ex)(Gx)" is false), then its contradictory formula, the O formula is true, and so this latter does
not have existential import.
5. Similar simple reasonings would show that among these four formulae the following
relations hold: A  I; A  -E; E  O, A  -O, E  -I, -I  O - the relations required by the
Square of Opposition.17

5. THE PROBLEM OF O PROPOSITIONS

As we have seen, according to the above analysis an O proposition does not have existential
import. Still, this situation does not seem to fit well our intuitions. For the proposition "Some
man is not white" seems to imply the existence of some man. I think it was precisely this point
which led Abaelard to distinguish between "Non omnis homo est albus" and "Quidam homo non
est albus".18 However, I think, his successors, distinguishing between the import of external
(propositional) negation and that of internal (term) negation (negatio negans vs. negatio
infinitans), had good reasons for assimilating the latter to the former.
Denial of the verb of an affirmative categorical proposition effects a negative proposition.
(E.g. "Homo currit" - "Homo non currit"; "Homo est currens" - "Homo non est currens", "Homo
est iustus" - "Homo non est iustus") Denial of the predicate term of an affirmative categorical
proposition, however, results in an affirmative proposition with an infinite noun as its predicate
term. ("Homo est non currens", "Homo est non iustus")19
Now, if a particular affirmative proposition is false, then the corresponding negative is true.
But if its subject term is empty, then an affirmative proposition is false.20 So, if there are no men,

17
In this way, as can be seen, a major defect of the standard quantificational analysis of the four categoricals has been eliminated: A and E
propositions in this analysis are indeed contraries, and so, in line with what the medievals say, we are not committed to accepting both "No
chimaera is licking my left ear" and "Every chimaera is licking my left ear" to which, however, one who defends the conditional interpretation is
committed. Still, due to the existential import attributed to A propositions in this analysis, this is done - unlike the line taken by G. Englebretsen
("The Square of Opposition", Notre Dame Journal of Formal Logic, 17(1976), pp.531-541) - without making an E proposition, a universal
negative proposition into an affirmative proposition with a negative predicate term, and without making an I proposition, a particular affirmative
proposition with a positive predicate term into a negative proposition with a negative (or infinite, as the medievals called them) predicate term.
18
See Abaelard: Dialectica, Assen, 1956, pp.175-177.
19
"... a proposition is rendered negative only by the negation of the composition, which is effected by the verb" St. Thomas Aquinas: "In Libros
Perihermeneias", in: Opera Omnia, Stuttgard-Bad Cannstadt, 1980, lb.II.lc.1.n.5.; "Again, some of the propositions are affirmative, some
negative. An affirmative proposition is that in which its formal part is left affirmative, while a negative proposition is that in which its formal
part is negated, and by the formal part of a categorical proposition I mean its verbal copula." Albert of Saxony: Perutilis Logica, Georg Olms
Verlag, Hildesheim-New York, 1974, fol.17.vc. (If not otherwise indicated, translations are mine.)
20
Cf. the following argument from the Sophismata of Albert of Saxony (Georg Olms Verlag, Hildesheim New York, 1975): "The twenty-fourth
sophisma is this: `Every chimaera which runs moves'. Proof. The contradictory of the sophisma is false, so the sophisma is true. The
consequence is valid. Proof of the antecedent. The contradictory of the sophisma is: `Some chimaera which does not run moves'. But this is false,
because it is an affirmative the subject of which supposits for nothing. ... In response to the first argument I deny that the contradictory of the
sophisma is false. And when it is said that the contradictory of the sophisma is `Some chimaera which does not run moves', I deny this, because
this proposition is affirmative as well as the first, for its principal copula is not negative. However, this is truly negative: `Some chimaera which
does not run does not move', or this:'Some chimaera which runs does not move'. For in these the principal copula is negated. I say, therefore, that
the contradictory of the sophisma is this: `Some chimaera which runs does not move' and not this `Some chimaera which does not run moves'.
And the first of these is true, because it is a negative proposition the subject of which supposits for nothing." In MPL, the sophisma can be
formulated thus: (Ex.)(M(x.Cx&Rx)) its proposed contradictory thus: (Ex.)(M(x.Cx&-Rx)), while its real contradictory thus: (Ex.)-
(M(x.Cx&Rx)). If in a model -(Ex)(Cx) is true, then the formula representing the sophisma and that representing its proposed contradictory are

16
then the affirmative "Aliquis homo est iustus" ("Some man is just") is false, while the negative
"Aliquis homo non est iustus" (Some man is not just") is true. Again, the affirmative "Aliquis
homo est non iustus" ("Some man is not-just") is false, while the negative "Aliquis homo non est
non iustus" ("Some man is not not-just") is true.21
Introducing internal negation, the same point can be brought out very clearly in MPL. The
introduction of internal negation is served by the following two clauses:
/F4/ If t1...tnT and PnPpred, then (-Pn)(t1)...(tn)F
/f11/ f((-Pn)(t1)...(tn))=1 iff <f(t1),...,f(tn)>Wn-R(Pn)22
Thus, if in a given model M R(H) is empty, then
(Ex.)(I(x.Hx)) /Aliquis homo est iustus/ is false,
(Ex.)~(I(x.Hx)) /Aliquis homo non est iustus/ is true,
(Ex.)((-I)(x.Hx)) /Aliquis homo est non iustus/ is false,
(Ex.)~((-I)(x.Hx)) /Aliquis homo non est non iustus/ is true.
We can also see that the following equivalences hold:
"(Ex.)~(I(x.Hx))"  "~(Ex)(Hx) v (Ex)(Hx & -Ix)" and
"(Ex.)((-I)(x.Hx))"  "(Ex)(Hx & -Ix)".23
That is, the proposition "Some man is not just", according to this analysis is equivalent to
"Either there are no men /i.e. nothing is a man/, or something is a man and is not just" (which,
pace Kneale, is by no means to say that these propositions are synonymous),24 while the
proposition "Some man is not-just" is equivalent to "Something is a man and is not just".25 But
everyday usage, at least in this case, does not seem to be particularly sensitive to the technical
distinction between "... is not just" and " ... is not-just", and so the former does duty for the latter,
which, in addition, in spoken English is hardly distinguishable from the former. As we have seen,
in the regimented Latin of the later medieval logicians the situation is a bit different: the negation
preposited to the copula negates the whole proposition together with its existential import, while
postposited to the copula it negates only the predicate term leaving unimpeached the import of
the affirmative copula. Still, before the time this became commonly accepted as a rule among

false, while that representing its real contradictory is true in that model - in full accord with Albert's solution. Cf. also the following two notes.
Concerning the unanimous approval of this type of solution of the problem of empty terms in late medieval logic see E.J. Ashworth: "Existential
Assumptions in Late Medieval Logic", American Philosophical Quarterly, 10(1973), pp.141-147. Concerning alternative, "value-gap"
treatments proposed by early medievals see S. Ebbesen: "The Present King of France Wears Hypothetical Shoes with Categorical Laces. Twelfth
Century Writers on Well-Formedness", Medioevo, 7(1981), pp.91-113. For a general survey of medieval treatments of the problem see S.
Ebbesen: "The Chimaera's Diary", in: S. Knuuttila-J. Hintikka: The Logic of Being, Helsinki, 1987.
21
Cf. Albert of Saxony: Sophismata, ed. cit., 13th sophisma of the second part: "In response to the second argument we concede that of any
suppositing term one of the contradictories is verified; but, because this term: `chimaera' supposits for nothing, neither this: `A chimaera is a
man', nor this: `A chimaera is a not-man' is true." Cf. further W. Ockham: Summa Logicae, St. Bonaventure N.Y., 1974, Pars II. c.12.; A.
Broadie, op. cit. (n.7. above), pp.28-31.
22
Note how well this formulation squares with Peter King's explanation of the matter in his Jean Buridan's Logic, Dordrecht, 1985, p.27. (see
the whole section)
23
Cf. the following text: "In negative propositions a term is either denoted not to refer to something or to refer to something of which the
predicate is truly denied, and so such a negative has two truth conditions like this: `A man is not white' has two conditions for its truth, for it is
true either because no man exists, and so is not white either, or because some man exists, and is not white." Ockham, op. cit., pp.218-219. Cf
also pp.284-285, 623-630; Paulus Venetus, Logica, Georg Olms, 1970, p.76. "Sexta regula..."; Moody, op. cit., pp.32-53.
24
Cf. W. and M. Kneale: The Development of Logic, Oxford, 1971, p.211.
25
Cf. Ockham, op. cit.: "Therefore this: `An ass is a not-man' is equivalent to `An ass is something and is not a man'." (p.283) "From this it is
evident that in virtue of its proper meaning /de virtute sermonis/ this is to be denied: `A chimaera is a not-man', for it has a false exponent sc.
this: `A chimaera is something'." (p.284)

17
medieval thinkers, Abaelard's intuition had suggested to him existential import in "Quidam homo
non est albus". (For the reason that the subject term in this proposition seems to be outside the
scope of the negation.) However, I think, when confusion is particularly to be avoided, a good
theory, contrasting "Quidam homo non est albus" with "Quidam homo est non albus" may correct
this intuition.26

6. COMMON PERSONAL SUPPOSITION

Now, it seems, we are already in a good position to show how the theory of common personal
supposition can be represented in MPL.
Common personal supposition was commonly divided by the medievals in determinate and
confused, and this latter in merely confused, and confused and distributive supposition. The
members of this division were characterized by Ockham by the following types of what are
called suppositional descents (descensus ad inferiora):27
1. DETERMINATE SUPPOSITION

1a. Some man is an animal, therefore, this man is an animal or that man is an animal or ...
(and so on for every man).
1b. Some man is an animal, therefore some man is this animal, or some man is that animal or ...
(and so on for every animal)
2. MERELY CONFUSED SUPPOSITION

2a. Every man is an animal, therefore, every man is this animal or that animal or ...
(and so on for every animal)
but not
2b. ... therefore, every man is this animal, or every man is that animal or ... (and so on for every
animal).
3. CONFUSED AND DISTRIBUTIVE SUPPOSITION

3a. Every man is an animal, therefore, this man is an animal, and that man is an animal and ...
(and so on for every man).
3b. Some man is not an animal, therefore, some man is not this animal and some man is not that
animal and ...
26
Especially, when the theory accords better with intuition in other cases. For example, in the case of propositions containing definite
descriptions or proper names, external and internal negation is readily distinguishable, if we use the expression: "It is not the case that ..." to
express external negation. Using this expression, everyone will accept that it is not the case that the present King of France is bald, and also that
it is not the case that the present King of France is not-bald /i.e. "haired"/. (Cf. Englebretsen, op.cit.; B. Russell: "Descriptions", in: R.R.
Ammermman (ed.): Classics of Analytic Philosophy, McGraw-Hill, New York, 1965; and A.N. Prior: "Negation", in: P. Edwards (ed.): The
Encyclopaedia of Philosophy, New York-London, 1967.) In the case of the four categoricals, however, the same device will not work. For in this
case the expression: "It is not the case that ..." would express the negation of the quantifiers, not the external negation of the predicate, as it did
in the former case.
27
Anyone unfamiliar with this doctrine of Ockham's, or with supposition theory in general, may consult (beyond the ones already mentioned in
the foregoing footnotes) the following works: E.A. Moody: Truth and Consequence in Medieval Logic, Amsterdam, 1957; P.V. Spade: "The
Semantics of Terms", in: N. Kretzmann-J. Pinborg-A. Kenny (eds.): The Cambridge History of Later Medieval Philosophy, Cambridge, 1982;
Ph. Boehner: "Ockham's Theory of Supposition and the Notion of Truth", Franciscan Studies, 6(1946), pp.261-292. S. Brown: "Walter
Burleigh's Treatise De Suppositionibus and its Influence on William Ockham", Franciscan Studies, 32(1972), pp.15-64; A.R.Perreiah:
"Approaches to Supposition-Theory", The New Scholasticism, 45(1971), pp.381-408; P.T.Sagal: "Refuting and Defending Supposition Theory",
The New Scholasticism, 47(1973), 84-87.

18
(and so on for every animal).
For reasons that will be evident later let us take also:
4. Some man is not an animal, therefore some man is not this animal and is not that animal and
... (and so on for every animal).28
Now, it is tempting to formulate these descents in MPL in the following manner:
1a. "(Ex.Hx)(Ax.)"  "A(H1) v A(H2) ..."
1b. "(Ex.Hx)(Ax.)"  "(Ex.Hx)(x. = A1) v (Ex.Hx)(x. = A2) v ..."
2a. "(x.Hx)(Ax.)"  "(x.Hx)(x. = A1 v x. = A2 ..."
but not
2b. "(x.Hx)(Ax.)"  "(x.Hx)(x. = A1) v (x.Hx)(x. = A2) v ..."
3a. "(x.Hx)(Ax.)"  "A(H1) & A(H2) & ..."
3b. "(Ex.Hx)~(Ax.)"  "(Ex.Hx)~(x. = A1) & (Ex.Hx)~(x. = A2) & ..."
4. "(Ex.Hx)~(Ax.)"  "(Ex.Hx)(~(x. = A1) & ~(x. = A2) & ..."
There are, however, at least two problems with this formulation. First, according to the given
formation rules of MPL, the consequents of these implications are ill-formed, and so these
implications are stricto sensu meaningless.29 Second, except for 1a. and 3a., in the consequents
we were compelled to introduce the identity-sign, which did not occur in their antecedents, and
so these consequents cannot be regarded as mere expansions (into disjunctions and conjunctions)
of their antecedents, which, however forms part of the idea of a suppositional descent.

7. THE IDENTITY THEORY OF PREDICATION

To solve the first of these problems, first we have to cope with the second one. The source of
this problem is that the formulae thus far introduced to represent the four categoricals do not
fully reflect Ockham's analysis of these propositions. For, on Ockham's analysis the subject and
predicate terms of these propositions are semantically on a par: they have the same semantic
function, namely the function of denoting individual entities and, correspondingly, the copula has
the function of an identity sign (whence the theory is frequently referred to as the "two-name
theory of predication" or "the identity theory of the copula").30 So, a more faithful representation
of Ockham's analysis in MPL would be the following:
A: (x.Gx)(Ey.)(x. = y.Fy)
I: (Ex.Gx)(Ey.)(x. = y.Fy)
E: (x.Gx)(y.)~(x. = y.Fy)
O: (Ex.Gx)(y.)~(x. = y.Fy)

28
Cf. what G.Priest and S.Read say (op. cit. n.7. above) on suppositio copulatim.
29
To be sure, the other works mentioned above treating the theory of suppositional descents suffer from the same defect too: the formal or
semiformal representations of these descents presented therein are stricto sensu meaningless in the same way in that in the theoretical framework
in which they are presented they are equally ill-formed, hence mathematically unacceptable (with the exception of the paper by G.Priest and
S.Read), even if they are acceptable as intuitive, but formally ungrounded extensions of this framework. Let this serve as an excuse for the rather
complicated formulations to follow, by which an attempt is made at giving a formally well-grounded reconstruction of Ockham's descents,
aiming at the mathematical rigour of contemporary formal theories.
30
Cf. e.g. L.M. de Rijk's "Introduction" to his edition of Abaelard's Dialectica, Assen, 1956, pp.37-48; P.T. Geach: "Nominalism", in his Logic
Matters, Berkeley-Los Angeles, 1972; D.P. Henry: Medieval Logic and Metaphysics, London, 1972, pp.55-56. For a clear formulation of this
doctrine as being one which needs no further confirmation see Buridan: Sophismata, (ed. T.K. Scott), Stuttgard-Bad Cannstatt, 1977, pp.42-43.

19
These formulae, as one can easily test it, are equivalent to the corresponding formulae
formerly used to represent the four categoricals, so they also would constitute a Square of
Opposition.31
The remaining disturbing factor in this formulation is the explicit appearance of the second
quantifier, which is only implicit (at least in Ockham's and his followers' analysis)32 in the
corresponding propositions. Let me therefore eliminate this explicit appearance (in accordance
with the rule: negatio distribuit terminum)33 by the following abbreviation rule:
/Abbr5/ If t1, t2T, then the string: (Et2)(t1 = t2) may be replaced by the string: .(t1 = t2)
and the string: (t2)~(t1 = t2) may be replaced by the string: .~(t1 = t2)
So, instead of the above formulae we can write:
A: (x.Gx).(x. = y.Fy)
I: (Ex.Gx).(x. = y.Fy)
E: (x.Gx).~(x. = y.Fy)
O: (Ex.Gx).~(x. = y.Fy)

8. EXPANSION FORMULAE, EXPANSION SCHEMATA

With this formulation at hand we can set about solving our first problem, by introducing the
notion of an expansion formula and that of an expansion schema.
Let us say that if wVar, xV, AF in which some occurrence of x is free, (Ex.A)(B)F in
which x.A occurs bound by an existential quantifier, B is the same string of signs without that
quantifier, (x.A)(B)F in which x.A occurs bound by a universal quantifier and B is the same
string of signs without that quantifier, then
1. the formula B(x.A/A(x/1)) v * * * v B(x.A/A(x/n) is an expansion formula of (Ex.A)(B) in a
model M, if the cardinality of E(x,M,f)(A) is n and the same holds for any n, if this set is empty
in M, or its cardinality is denumerably infinite in M,
2. the formula (w)(B(x.A/A(x/1)) v * * * v B(x.A/A(x/n))) is an expansion formula of
(w)((Ex.A)(B)) in a model M, if the cardinality of E(x,M,f)(A) is n, and the same for any n for
the other two cases,
3. the formula B(x.A/A(x/1)) & * * * & B(x.A/A(x/n)) is an expansion formula of (x.A)(B) in
M if the cardinality of E(x,M,f)(A) is n, and the same for any n for the other two cases,
4. the formula (Ew)(B(x.A/A(x/1)) & * * * & B(x.A/A(x/n))) is an expansion formula of
(Ew)((x.A)(B)) in M, if the cardinality of E(x,M,f)(A) is n, and the same for any n for the other
two cases;
31
The distinction between external and internal negation, say, in "Aliquis homo non est iustus" and "Aliquis homo est non iustus" in this
analysis can be expressed by the following formulae: (Ex.Hx)(y.)-(x. = y.Iy), on the one hand, and (Ex.Hx)(Ey.)(x. = y.-Ix) on the other. I think
this formulation renders even more perspicuous why internal negation can rightly be called also term negation, as opposed to propositional
negation.
32
Cf. e.g.: "Omnis terminus communis quem nullum signum praecedit nisi signum particulare supponit determinate, ut huius propositionis
homo est animal tam subiectum tam praedicatum supponit determinate; et si alicui istorum adderetur signum particulare non propter istud
impediretur suppositio ista." That is: "Every common term not preceded by a particular sign has determinate supposition, as in this proposition:
homo est animal, both the subject and the predicate terms supposit determinately; and if to either of these a particular sign would be added, this
supposition would not thereby be impeded." Paulus Venetus, Logica, Georg Olms, Hildesheim-New York, 1970.pp.40-41.
33
Cf. e.g. Buridanus, Tractatus de Suppositionibus, ed. Maria Elena Reina, Rivista Critica di Storia della Filosofia, 1957, pp.175-208, and
pp.323-352, p.328.

20
where n is the numeral denoting n, and the asterisks indicate intermediate conjuncts or
disjuncts (differing from each other only in the substituted numeral in them ).34
Let us say further that (our metaexpressions meaning the same as above)
1.' the string B(x.A/A(x/1)) v B(x.A/A(x/2)) v ... is an expansion schema of (Ex.A)(B),
2.' the string (w)(B(x.A/A(x/1)) v B(x.A/A(x/2)) v ... is an expansion schema of
(w)((Ex.A)(B)),
3.' the string B(x.A/A(x/1)) & B(x.A/A(x/2)) & ... is an expansion schema of (x.A)(B),
4.' the string (Ew)(B(x.A/A(x/1)) & B(x.A/A(x/2) & ... is an expansion schema of
(Ew)((x.A)(B)).
The strings defined by 1.' and 2.' are called disjunctive, while those defined by 3.' and 4.' are
called conjunctive expansion schemata. The expansion formulae defined by 1-4. above are
called equiform to the expansion schemata defined by 1'-4.', respectively.
Now, we can define a valid inference between an expansion schema and a formula of MPL in
the following manner.
If A is a formula of MPL, S is an expansion schema of a formula B and EXF(S,M,B) is an
expansion formula of B equiform to S in a model M, then
A  S is valid, if and only if there is no model M such that A is true in M and no EXF(S,M,B)
is true in M, provided S is disjunctive, or A is true in M and some EXF(S,M,B) is false in M,
provided S is conjunctive; and
S  A is valid, if and only if there is no model M such that some EXF(S,M,B) is true in M
and A is false in M, provided S is disjunctive, or every EXF(S,M,B) is true in M and A is false in
M, provided S is conjunctive.
With these definitions at hand the following is a true and meaningful representation of
Ockham's descents:
1a. (Ex.A).(x. = y.B)  .(A1 = y.B) v .(A2 = y.B) ...
1b. (Ex.A).(x. = y.B)  (Ex.A)(x. = B1) v (Ex.A)(x. = B2) ...
2a. (x.A).(x. = y.B)  (x.A)(x. = B1 v x. = B2 ...
but not
2b. (x.A).(x. = y.B)  (x.A)(x. = B1) v (x.A)(x. = B2)...
3a. (x.A).(x. = y.B)  .(A1 = y.B) & .(A2 = y.B) & ...
3b. (Ex.A).~(x. = y.B)  (Ex.A)~(x. = B1) & (Ex.A)~(x. = B2) ...
4. (Ex.A).~(x. = y.B)  (Ex.A)(~(x. = B1) & ~(x. = B2) ...
(Notice that A and B can be here of any complexity.)

9. SUPPOSITIONAL DESCENTS AS A "RUDIMENTARY QUANTIFICATION


THEORY"

"Is this a quantification theory?" - asks Matthews in his "Suppositio and Quantification in
Ockham".35 His answer is in the negative. But, I think, the reasons for his answer, however

34
For the use of "( / )" see again n.9.
35
p.17.

21
reasonable they were in their original conceptual framework, within this reconstruction are
eliminated.
Matthews' reasons for his answers can be summarised in the following points:
1. the expanded propositions are not equivalent to the propositions from which they are derived,
unless to the expanded propositions we add an expression of the sort: "and every man is this man
or that man or ...", which guarantees that no man is overlooked in the enumeration; but in this
case in the consequents of the suppositional descents there appear again the quantifiers which
were supposed to be eliminated by these descents;
2. in the descent numbered 3b. above, even if the inference goes down, it does not go up, i.e., the
implication does not hold in the opposite direction;
3. in the case of propositions with empty subject terms, "there being no subject-term supposita ...,
no subject-term descent is possible".
As a consequence of these points, the implications above cannot possibly be turned into
equivalences, and so these "descent schemata" cannot be regarded as giving "a contextual
definition of the quantifiers".
Now, as for point 1., it should be clear that the clauses securing the completeness of the
enumeration do not themselves belong to the expanded propositions, and so, in the formalization
of Ockham's descents, by the introduction of the notion of an expansion formula, which per
definitionem secures the completeness of enumeration, we could dispense with them altogether.
As for point 3., it should be recalled that, according to Ockham, what determines the form of
the expanded propositions is not the contingent circumstance whether the subject term of the
unexpanded proposition does in fact supposit for something, but rather the way in which it is
"denoted" to supposit,36 and so in the present reconstruction what determines the form of an
expansion formula (and that of an expansion schema) of an unexpanded formula is the form of
the latter, and not the circumstance whether its "subject term" is, in a given model, empty. (What
is determined by this circumstance is that the number of the conjuncts or disjuncts in this case is
irrelevant: a formula, which otherwise meets the conditions stated in the definition of an
expansion formula, may it have any number of conjuncts or disjuncts, will be an expansion
formula of the unexpanded formula in that model.)
Last but not least, as for point 2., it should be noted that even if Matthews is right in saying
that in 3b. no "ascent" is possible, still, the same is not true for 4. Had it been Ockham's purpose
to give expanded propositions (or, rather, proposition schemata) which are equivalent to the
unexpanded propositions, he might have chosen descent 4. His purpose with these descents,
however, as is clear from the context in which he treats them, was to distinguish different kinds
of common personal supposition, and so, since descent is possible in the manner of 3b., which
determines that the predicate term of the unexpanded proposition supposits confuse et
distributive, he simply did not care the other possible descent.37
All in all, in my opinion, Ockham might have used his theory of suppositional descents as a
"rudimentary quantification theory" in the sense Matthews uses this expression, but he simply did
not. Neither did he use these descents for a sort of "nominalist reduction" (as at a point Matthews

36
Cf. Ockham, op. cit., pp.218-219.
37
For an alternative (but, as far as I can see, with mine also compatible) intuitive response to the problem see P.King, op. cit., p.51. Cf. n.28.

22
seems to imply)38, for he used them, it seems, solely for the purpose of distinguishing the several
kinds of common personal supposition.

10. OCKHAM'S NOMINALISM AND THE IDENTITY THEORY OF PREDICATION

What Ockham did use for a "nominalist reduction" seems to be rather his theory of
predication. Because it was this theory, namely the identity (or two-name) theory, which, unlike
the then concurrent inherence theory, provided him with the possibility of reducing the categories
of entities admitted by him in his ontology to two, namely to that of substance and that of quality.
For, according to the inherence theory the function of the predicate is to signify particular forms,
inhering in their subjects, and what makes a predication true is the actual inherence (inesse) of
the form signified by the predicate in the thing supposited for by the subject. So, e.g., the
proposition "Socrates is white" is true, if and only if the whiteness of Socrates exists, or, the
same expressed in the "material mode of speech", Socrates is white by his whiteness.
But if so, then further: "the column is to the right by to-the-rightness, God is creating by
creation, is good by goodness, just by justice, mighty by might, an accident inheres by inherence,
a subject is subjected by subjection, the apt is apt by aptitude, a chimaera is nothing by
nothingness, a blind is blind by blindness, a body is mobile by mobility, and so on for other,
innumerable cases".39 And this is nothing, but "to multiply beings according to the multiplicity of
terms ..., which, however, is abusive and leading far away from truth"40 - says Ockham.
So, to get rid of this "slum of entities", Ockham abandoned the theory of predication which
seemed to involve it, in favour of the (then also well-established) identity theory, which,
assigning the same referring function to the predicate as well as to the subject term, did not
commit him to these entities.
But, as an objection of Peter of Ailly shows, this parsimony had its price to be paid elsewhere
in Ockham's theory. The objection runs as follows:
"Second Thesis: whether it is affirmative or negative, a proposition is not therefore true or
false because its subject and predicate supposit for the same thing or because they do not
supposit for the same thing.
Proof: Those who define `true proposition' or `false proposition' in terms of suppositing or not
suppositing for the same thing ... also give a definition of supposition in which it is said that
supposition `is the taking of a term in a proposition for its significatum or significata of which
this term is verified by means of the copula of the proposition in which it is posited'. Thus, they
define supposition in terms of verification of a proposition. Therefore, they should not,
conversely, define the verification of a proposition in terms of supposition or of suppositing or
not suppositing for the same thing. For one ought not to give circular definitions or define the
same by the same..."41

38
"Suppositio and Quantification in Ockham", pp.14-15.
39
Ockham, op. cit., p.169.
40
ibid. p.171.
41
Peter of Ailly: Insolubilia, Paris, 1492, quoted and translated by Marilyn McCord Adams in her "What Ockham Means by Supposition?",
Notre Dame Journal of Formal Logic, 17(1976), p.379.

23
Now let us replace in the definition of the assignment f in a model M clause /f5/ by the
following:
/f5'/ f(Pn(t1)...(tn))=1 iff f((Ex.)(ti = x.S(ti/x)Pn(t1) ... (tn)|))=1,
where S(ti/x)Pn(t1)...(tn)| stands for the formula that results from Pn(t1)...(tn) by substituting x for
ti in it, and x does not occur in the original formula. As a result of this replacement we have only
/f5'/ and /f6/ for supplying the truth conditions of atomic sentences in MPL, thereby getting rid of
a "slum of entities", the former denotata of predicate parameters, namely the subsets of W 1,
W2,...,Wn. (Of course, this is true only insofar as the analysis of predication is concerned. The
semantic apparatus of MPL as a whole by the axioms of set theory remains still committed to
these sets.) But then, trying to determine the truth value of an atomic sentence, say "Hs"
("Socrates est homo") in a model M, we get caught in a vicious circle. For, according to the
definition of truth in a model M, "Hs" is true, if and only if for some f in M, f(Hs)=1. But,
according to /f5'/, this is so if and only if f((Ex.)(s=x.Hx))=1, which, according to /f9/ holds iff
for some uRg(x.Hx)f, f[x.Hx:u]((s=x.Hx))=1, and this, according to /f6/ and /f1/, is so, iff
R(s)=u. So in order to know whether this formula is true in M, we have to know whether there is
a uRg(x.Hx)f such that it is identical with R(s). (That is, whether R(s)Rg(x.Hx)f.) So we have
to know which u's belong to Rg(x.Hx)f. But, according to the definition of Rg(x.Hx)f we can
know this only if we know the possible values of x.Hx according to any assignment f' that differs
from f at most in the value assigned to x and x.Hx, that is, if we know for any uW and for any
such f' whether f'(x.Hx)=u or not, which, according to /f4/, we can know only if we know for any
such f' whether f'(Hx)=1. But this leads us back to /f5'/ where we have begun, and so on in
infinitum.42
I think, it is in this way that Peter of Ailly's objection can be modelled and, what is more,
shown to be sound by means of the semantic apparatus of MPL.
It should be noted, however, that even if the introduction of /f5'/ instead of /f5/ leads to this
collapse of the theory, still, if we retain /f5/ the equivalence:
Pn(t1)...(tn)  (Ex.)(ti = x.S(ti/x)Pn(t1)...(tn)|)
holds without any collapse. So, e.g., "Socrates is man", "Hs", is equivalent to "(Ex.)(s =
x.Hx)", "Socrates is a man"; and "Lps", "Plato loves Socrates", is equivalent to "(Ex.)(p =
x.Lxs)", "Plato is a lover of Socrates'", and so on, in virtue of this equivalence, transformations of
this sort are always possible.43 The general lesson of which, in my opinion, is that the identity
theory, even if it may be a sound logical theory in that every atomic sentence can be transformed
into an identity statement, still, it cannot be exploited for the purpose of a thoroughgoing
nominalist reduction, unless it is supplemented with a further theory (a theory of signification),

42
I think it would be interesting to compare this regress to the Anselmian regresses discussed by D.P. Henry in several of his papers (later also
incorporated in his op. cit.). I cannot resist the feeling that St. Anselm's regresses are the same as this, only put in an object-language form. Just
consider the series: Hs; (Ex.)(s=x.Hx); (Ex.)(s=x.(Ey.)(x=y.Hy); ... (Socrates est homo; Socrates est aliquid quod est homo; Socrates est aliquid
quod est aliquid quod est homo; ...) This point, however, cannot be pursued here.
43
For evidence that e.g. Buridan was perfectly aware of the possibility of these transformations, see his Tractatus de Consequentiis, (ed. H.
Hubien), Louvain-Paris, 1976, pp.44-52. It should also be noted here that although Ockham uses the identity theory as a rival to the inherence
theory, these two theories in the works of other scholastics peacefully coexisted. See e.g. St. Thomas Aquinas, 3SN.d.5. q.3.a.3. What is more,
even Ockham conceded the inherence-analysis in cases involving quality-terms, connoting real, inherent qualities, such as the term "white". See
e.g. his op. cit., II.c.11.p.253.

24
which supplies independent criteria for determining the semantic values of terms and sentences.
But this is already a different story.
All I wanted to show in this paper was that despite appearances there exists no unbridgeable
gap between medieval logic and modern quantification theory. With the aid of suitable
modifications and supplementations modern quantification theory can be rendered able to
incorporate certain insights of the medievals, thereby resulting in a theory which does justice to
the medieval tradition, while meets modern requirements as well.

25
GENERAL TERMS IN THEIR REFERRING FUNCTION
"... thanks to Russell and Frege, most of the logical insights that were lost by
Aristotle's Fall have been recovered; but not, to my mind, quite all of them.
As I said, Aristotle's De Interpretatione recognized as belonging to the
category of names not only proper nouns like `Socrates' but also certain
common names like `man'. This simple and natural view was rejected by
Frege and Russell, for reasons that I do not find convincing; and most
modern logicians have followed Frege and Russell in this matter ... What we
still have not got is a formal theory that recognizes the status of some
general terms as names without blurring the distinction between names and
predicables. Success in stating such a theory would be Paradise Regained."
1
(P. T. Geach)

1. INTRODUCTION
In standard first order quantification theory (SQT), the general terms of natural languages are
represented by predicate letters. So SQT attributes predicative function to all general terms in
whatever grammatical category, be they nouns, verbs, adjectives, or participles. In the
representation of categorical sentences, therefore, even the subject terms of these sentences
become predicates, namely, predicates of bound variables. As a result, in this representation the
subject-predicate, or NP-VP (noun-phrase/verb-phrase), structure of the original categorical
sentences simply disappears. This is what qualifies this structure as a "surface" phenomenon, as
opposed to the "deep", "logical" structure of these sentences.
The variables, which take over the referring function of common noun phrases (CNP) in the
formulas representing categorical sentences, take their values from the whole universe of
discourse in a model. A consequence of this is that we cannot find a truth connective with which
completing the following schemes:

For every x }
For some x }
For the x } Fx ? Gx
For most x }
For five x }

we would get correct representations of the following sentence-schemes:

(1) Every }
(2) Some }
(3) The } F('s) is/are G('s)
(4) Most }
(5) Five }

(1) and (2) are representable in SQT only with formulae which have different matrices.2 The
matrix of the SQT formula representing (1) is conditional. So in this representation (1) does not
1
P. T. Geach, "History of the Corruptions of Logic", in: Logic Matters, Berkeley and Los Angeles, 1972, p.61.
2
G. B. Matthews: "Ockham's Supposition Theory and Modern Logic", The Philosophical Review, 73(1964), No.1., pp.91-99, p.98.

26
have existential import. And this destroys the traditional Square of Opposition.3 But even worse
than this is the fact that (4) in itself is not representable in SQT.4 A further consequence of this is
that the following sentence is not representable in SQT either:
(6) Most miners who own some donkeys insure most of them.
So even though variables of SQT may be characterized as acting like pronouns of natural
languages,5 despite appearances to the contrary, pronouns of natural languages may not be
characterized as acting like bound variables of SQT, not even in cases of simple nominal
anaphora.6
All in all, we can summarize these observations in the following points:
1. Formulae of SQT do not preserve the grammatical structure of natural language sentences
represented by them.
2. SQT destroys the Traditional Square of Opposition (and, consequently, several traditional
syllogistic forms) thereby breaking radically with its own tradition, namely traditional logic.
3. SQT is incapable of representing a number of quantified natural language sentences even
within the sphere of purely extensional contexts.
4. Despite appearances to the contrary, the machinery of bound variables of SQT does not
properly reflect the workings of pronouns even in cases of simple nominal anaphora
To many people it seemed that some or all of what are described in these points are
irremediable defects of SQT. So the problems touched by these points have given rise to several attempted
solutions which break more or less radically with the usual syntactic and semantic construction of SQT.7 In what
follows I present the syntax and semantics of a formal theory which is capable of providing a satisfactory solution to
all of the problems involved in these points without straying too far away from the beaten path of SQT.

2. GTL (GENERAL TERM LOGIC)


The root of the above-mentioned problems, as I hope it is clear from the previous presentation,
is that SQT attributes predicative function to all general terms of natural languages, and,

3
Cf. e.g. G. Englebretsen, "The Square of Opposition", Notre Dame Journal of Formal Logic, 17(1976), pp.531-541). I think it should be noted
here that there are at least two different uses of the expression "existential import" in the literature. According to one usage the expression "an A
proposition has existential import" means: "SaP implies that something is an S". According to another usage, however, the same expression
means that SaP presupposes that something is an S, i.e., "SaP or SoP implies that something is an S". See e.g. J. van Eijck: "Generalized
Quantifiers and Traditional Logic", in: J. van Benthem-A. ter Meulen: Generalized Quantifiers in Natural Language, Foris Publications, 1985,
and J. van Benthem, Essays in Logical Semantics, Reidel, Dordrecht, 1986, p.111. I follow the first usage.
4
For proof, if "most" is understood as "more than half the", see J. Barwise and R. Cooper, "Generalized Quantifiers and Natural Language",
Linguistics and Philosophy, 4(1981), pp.159-219, pp.214-215. (C13)
5
P.T. Geach, "Quine's Syntactical Insights" in: Logic Matters, Berkeley and Los Angeles, 1972, pp.115-127; V.W.O. Quine, Mathematical
Logic, Harper Torchbooks, 1951, pp.65-71.
6
For a suggestive description of the contrary view see G. Boolos, "To Be Is to Be the Value of a Variable (or to Be Some Values of Some
Variables)", The Journal of Philosophy, 8(1984), pp.430-449, 430-431. The problematic character of this view even in respect of
"firstorderizable" sentences was brought out most clearly by G. Evans, "Pronouns, Quantifiers and Relative Clauses" and "Appendix", Canadian
Journal of Philosophy, 7(1977), pp.467-536, pp.777-796.
7
ad 1. cf. F. Sommers, The Logic of Natural Languages, Oxford, 1982; J. Barwise and R. Cooper, "Generalized Quantifiers and Natural
Language", Linguistics and Philosophy, 4(1981), pp.159-219. ad 2. cf. G. Englebretsen, "The Square of Opposition", Notre Dame Journal of
Formal Logic, (1976), pp.531-541; D.P. Henry, Medieval Logic and Metaphysics, London, 1972; D.P. Henry, That Most Subtle Question -
(Quaestio Subtilissima): The Metaphysical Bearing of Medieval and Contemporary Linguistic Disciplines, Manchester, 1984; ad 3. cf. the
above-mentioned paper by Barwise & Cooper, Keenan,E.L.-Stavi,J. "A Semantic Characterization of Natural Language Determiners",
Linguistics and Philosophy, 9(1986), 253-326,and M. Brown: "Generalized Quantifiers and the Square of Opposition", Notre Dame Journal of
Formal Logic, 25(1984), pp.303-322; ad 4. cf. the above-mentioned papers by Evans, and H. Kamp, "A Theory of Truth and Semantic
Representation", in: J. Groenendijk et al. eds., Formal Methods in the Study of Language, Amsterdam, 1981, B. Richards, "On Interpreting
Pronouns", Linguistics and Philosophy, 1984, pp.287-324.

27
consequently, in the SQT representation of natural language sentences the referring function of
CNP's (common noun phrases) and of pronouns with such antecedents is taken over by variables
uniformly ranging over the whole universe of discourse of a model. But intuitively the following
sentences
(7) Every man is mortal
(8) Some man is mortal
(9) Most men are mortal
are not about all the things in the universe, but only about men, referred to by the common
subject of these sentences. So it seems plausible to represent these sentences by the introduction of a variable
which takes its values not from the whole universe, but only from the set of men, the extension of the open sentence
"x is a man". Let us say, therefore, that if "x" is a variable, then "x.x is a man" is a variable too, namely a restricted
variable representing the subject term of the above sentences. So in this representation, in perfect accordance with
the analyses of medieval terminist logicians (cf. sect. 2.3.) the subject term of these sentences possesses referring
function, indeed, as a variable it refers to men, i.e., things of which, when pointed at, we can truly say: "this is a
man".8 But what about the case, when this is true of nothing? Of course, then this term refers to nothing. So in this
case let us assign to it an appropriate zero-entity as its value, which lies outside the whole universe of discourse, i.e.,
which is not any of the things that can be referred to: that a term has the zero-entity as its value means no more nor
less, than that the term refers to nothing. As will be seen, this means that with a term that refers to nothing in this
system no affirmation will be true. The significance of this circumstance in respect of GTL's ties with medieval logic
will be evident later. But first let us see the formal development.

2.1 GTL SYNTAX


The language of GTL is defined as follows.
L:=<C,P,V,T,F>,
where C:={,&,=,.,(,)}D, where D is a set of determiners, P:=PindPpred, where Pind is the
set of individual parameters and Ppred is the set of predicate parameters of GTL; V is the set of
proper variables, as opposed to the set of restricted variables, Vres, which is a part of T, the set
of terms, and F is the set of formulae of GTL. C, P and V contain the primitive symbols of GTL.
The set of expressions of GTL (Exp:=TF) is defined by the following clauses.
(i) aPind  aT
(ii) xV  xVar, where Var:=VVres
(iii) vVar, AF  v.AVres
(Note that this rule is applicable recursively, so it permits even nested occurrences of
restricted variables, i.e., occurrences, in which a restricted variable appears as the operator
variable of another restricted variable.)
(iv) t1,...,tnT, PnPpred  Pn(t1)...(tn)F, (Pn)(t1)...(tn)F, (ti = tj)F
(v) QD, A,BF, vVar,  (A)F, (Qv)(A)F, (A & B)F9

8
Cf.: "est autem suppositio prout hic accipitur acceptio termini in propositione pro aliquo vel pro aliquibus quo demonstrato vel quibus
demonstratis per ista pronomina `hoc' vel `haec' vel equipollentia illis, terminus vere affirmatur de isto pronomine mediante copula illius
propositionis" - "Supposition, as taken here, is the taking of a term in a proposition for something, or some things, which being pointed at by
these pronouns `this' or `these' or their equivalents, the term is truly affirmed of these pronouns by the mediation of the copula of the
proposition", Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, p.50
9
The same abbreviations may be applied as in the previous Essay. As the reader may see, D and P are given without any further specification.
This gives us a certain liberty in the application of GTL to natural languages: for the construction of a GTL formula we may use either schematic
letters or expressions of natural languages from the appropriate syntactic categories, as the occasion requires. Cf. the somewhat similar policy of

28
2.2 GTL SEMANTICS
The definition of a model for GTL runs as follows:
M:=<W,R,0>,
where W is a nonempty set, the universe of discourse, 0 is the zero entity, which is not an
element of W, and R is a semantic function assigning values to the parameters of GTL defined by
the following clauses:
(R1) aPind  R(a)W
(R2) PnPpred  R(Pn) is a subset of Wn
Given a model, an assignment f (a function from Exp to W{1,0})10 is defined as follows
(for the interpretation of the syntactic symbols see clauses (i)~(v) above):
(f1) tT, AF  f(t)W U {0}, f(A){1,0}
(f2) f(a)=R(a)
(f3) f(x)W
(f4) f(v.A)=f(v), if f(A)=1, f(v.A)=0 otherwise
(f5) f(Pn(t1)...(tn))=1  <f(t1),...,f(tn)>R(Pn)
(f6) f((Pn)(t1)...(tn))=1  <f(t1),...,f(tn)>Wn-R(Pn)
(f7) f(t1 = t2)=1  f(t1)=f(t2)W
(f8) f(A)=1  f(A)=0
(f9) f(A & B)=1  f(A)=f(B)=1
(f10) f((Qv)(A))=1  for Q'uRg(v)f, f[v:u](A)=1,
where Rg(v)f, the range of the variable v with respect to f is defined as follows:
Rg(v)f={uW: for some f' that differs from f at most in the value assigned to v and its /nested/
operator variable/s/, f'(v)=u}, if this set is not empty, otherwise Rg(v)f={0}; Q' is a natural
language determiner represented by Q and f[v:u] is the same as f except that it assigns u to v.11
Truth in a model is defined as follows:
(T) |A|=T  for some f, f(A)=1
Validity, as usual, is truth in every model, and validity of an inference is validity of the
conditional formed from the conjunction of the premises and from the conclusion.

3. GTL AT WORK
In this section, through several applications and comparisons with alternative approaches, I
wish to show how GTL copes with the difficulties listed in the Introduction.

3.1. PRESERVATION OF LINGUISTIC STRUCTURE


As we can see, if Q is a determiner, then the sentence schemes (1)~(5) above can be
represented by the following simple formula scheme within GTL:
(1-5) (Qx.Fx)(Gx.)

E.L. Keenan - L.M. Faltz, "A New Approach to Quantification in Natural Language", in: C. Rohrer (ed.), Time, Tense and Quantifiers,
Tübingen, 1980.
10
0 functions both as the semantic value of empty terms and of false sentences in the same way as in the previous Essay.
11
Again, in the same way as in the previous Essay, the range of a restricted variable is never empty: it is either {0}, if the extension of its matrix
is empty, or is identical with this extension, provided this extension is not empty. Cf. n.14. of the previous Essay.

29
From this scheme we get representations of (1)~(5), if we replace Q in it by elements of D of
GTL representing "Every", "Some", "The", "Most" and "Five", respectively. However, what is
more interesting in this formulation from our present point of view is that (1-5) reflects the
common linguistic structure that is found in all of (1)~(5).
The close correspondence between the syntax of GTL and that of English can be seen even
more clearly with the following analysis of, say, (1):

(1') Every thing which is-F is-G


 x . F( ) G( )

The bars indicate the way of constructing the GTL formula representing (1').
But the same close correspondence can be seen even in much more complicated cases. Let us
take e.g. the following sentence:
(10) Every honest man loves his wife
Appropriate linguistic analysis easily shows how to construct the corresponding GTL formula:
(10') (x.Mx.Hx.)(Iy.)(L(x.)(y.Wyx.))
Clearly, there is some linguistic work to be done to show how to produce the "appropriate
analysis" of a sentence and to state explicitly the rules according to which the corresponding GTL
formula is to be constructed. These examples only serve to indicate the ways in which I think this
can be done.12 More to our present concern, however, is the comparison of the representative
force of GTL with that of SQT. So let us compare (10') with the SQT representation of (10):
(10'') (x)(Mx & Hx  L(x)(Iy.Wyx))
In this formalization there appear truth connectives that have no equivalents in the original
sentence (let alone those that would appear on the Russellian elimination of the description!).
The conditional is to be introduced in order to reflect somehow the restriction imposed by the NP
("honest man") upon the range of "every" in (10). The conjunction is invoked to reflect the
restriction imposed by the adjective ("honest") upon the range of the noun ("man"). In fact,
however, these devices cannot even properly be called restrictions. For these do not, properly
speaking, restrict the range of the variable in (10''). But the proper restrictions may be found
immediately in the structure of (10`). This is why, if we changed the determiners in (10), say
"Every" to "Most", GTL still would give faithful representations, where SQT would fail to give
any representation at all.
A further advantage of this representation is that since adjectival constructions need not be
analyzed in GTL as conjunctions, constructions like "dead man", "potential being", "fictitious centaur" need not
be analyzed as "something that is a man and is dead", "something that is being and is potential" and "something that
is a centaur and is fictitious", respectively. Indeed, in an appropriate intensional version of GTL it can be shown that
representations of these phrases have different semantic values in a model.13

12
In this paper I do not wish to pretend that I have a preestablished set of phrase structure rules and of corresponding rules of translation for
GTL and for a fragment of a natural language (the most natural in this respect being English, of course). Instead, I try to show the several uses
GTL may be put to in representing the workings of natural language expressions, from which the appropriate rules should finally quite clearly
emerge. As a matter of fact, I would rather leave the task of formulating these rules to an interested linguist.
13
See Essay IV.

30
3.2 THE SQUARE OF OPPOSITION
From what has been said thus far it should be clear that the four traditional types of
categoricals can be given a representation in GTL that corresponds very well to the traditional
analysis of these propositions:
A: (x.)(F(x.Gx)) E: (x.)(F(x.Gx))
I: (x.)(F(x.Gx)) O: (x.)(F(x.Gx))
Syntactically, A and I, and E and O differ only in quantity, while A and E, and I and O differ
only in quality. Semantically, as A  `(Ex)(Fx)' is valid in GTL, therefore the following relations
also hold: A  I, E  O, A  O, E  I, A  O, I  O —the relations required by the
traditional Square of Opposition.14
A bothering feature of this Square, however, especially for those trained in the Fregean
tradition, may be that according to this analysis e.g. "Some man is not white" is true if there are
no men.
But let us consider the following. If we want (a) "Every man is white" to have existential
import and also want (b) "Some man is not white" to be the contradictory of "Every man is
white", then we cannot but accept that "Some man is not white" is true when there are no men.
Now, if we want to avoid this conclusion we have to deny either (a) or (b). Denial of (a)
characterizes the Fregean analysis, while denial of (b) is the solution of Abaelard.15 But it seems
that we have good reasons to accept both.

14
I think I should call attention here to the faithfulness (cf. sect.3.3.2.) of this representation to the 13-14th (and even later, see E.J. Ashworth:
"Existential Assumptions in Late Medieval Logic", American Philosophical Quarterly, 10(1973), pp.141-147) analysis of the categoricals. In
this representation all, allegedly ineliminable, differences between the medieval and modern analyses listed by G.B. Matthews (see Essay II.)
simply disappear. In recent treatments of the Square it is quite generally assumed that the laws of the Square (as well as the laws of syllogistic)
were meant to apply only to categoricals with nonempty subject terms. (Cf. n.3. and E.A. Moody: Truth and Consequence in Medieval Logic,
Amsterdam, 1957, p.53.fn.) For a classical statement of this assumption see W. and M. Kneale: The Development of Logic, Oxford, 1971, pp.54-
61. Well, as for Aristotle, this assumption seems to be at least in conflict with, if not squarely contradicted by, the following passages: Cat.
13a38-13b35; Periherm. 19b5-20b13; Anal. I., 51b6-52b25. (Cf. also Boethius's comments on these texts.) It should be conceded, however, that
some 12th century authors made a similar assumption. See S. Ebbesen: "The Present King of France Wears Hypothetical Shoes with Categorical
Laces". Twelfth Century Writers on Well-Formedness", Medioevo, 7(1981), pp.91-113. (I owe thanks to Prof. Ebbesen for having called my
attention to this in correspondence.) But that their approach did not have lasting influence is amply attested by later medieval logical writings.
(See again, e.g., Ashworth, op. cit., Moody, op.cit., pp.32-53, D.P. Henry, That Most Subtle Question - (Quaestio Subtilissima): The
Metaphysical Bearing of Medieval and Contemporary Linguistic Disciplines, Manchester, 1984, pp.264-274, A. Broadie: Introduction to
Medieval Logic, Oxford, 1987, p.120.) In any case, the modern writers who made this assumption did not have in mind these 12th century
authors.
15
To be sure, Frege is not the first author in the history of logic to give an hypothetical analysis of universal affirmatives. In the 13th century
William of Sherwood wrote as follows: "When I say `Every man is an animal', here an habitual `is' is predicated. And insofar as it is necessary,
this proposition is equivalent to the following conditional `If it is a man, then it is an animal'." William of Sherwood, Introductiones in Logicam,
ed. M. Grabmann, Sitzungsberichte der Bayerischen Akademie der Wissenschaften, 1937, 10, p.83. (For a text to the same effect from Garland
the Computist /11th century/ see Henry, That Most Subtle Question, ed. cit., pp.85-86. Cf. also L.M. de Rijk (ed.), Logica Modernorum, Assen,
1967, II-2, p.730.) I think it is instructive to quote here at some length from Ockham's criticism of this approach: "... supposing that there are no
donkeys they reject this syllogism: `Every animal is a man; every donkey is an animal, so every donkey is a man', saying that `is' equivocates
here, for in the major premise `is' is taken as the operation of being, in the minor premise, however, `is' is taken as the `is' of habitude, or
consequence, as in: `if it is white, then it is colored'. And this is entirely absurd, for this leads to the destruction of all syllogistic forms. For
whenever I liked, I would say that `is' equivocates in the propositions and reject at caprice any syllogism on account of this equivocation.
Similarly, just as a syllogism is valid with any terms, so it is valid however the things may change... So even if ... all donkeys were destroyed,
this would be a valid syllogism. ... And so such distinctions as `is' is either the operation of being or it is the `is' of consequence are frivolous and
are set up by those who are unable to distinguish between a categorical and a hypothetical proposition. So these propositions are to be
distinguished: `A donkey is an animal' and `If it is a donkey, then it is an animal', because the one is categorical and the other is conditional or
hypothetical; and they are not equivalent for the one may be true the other being false. Just as this is now false: `God not-creating is God', but
this conditional is true: `If this is God not-creating, then this is God'." W. Ockham: Summa Logicae, St. Bonaventure N.Y., 1974, pp.263-264.
(See the whole chapter.) Let us not forget here, however, that in GTL the conditional form is also available. So whether or not Ockham's
arguments (and mine below) concerning the proper analysis of universal affirmatives are cogent, GTL, in keeping with Professor Henry's

31
1. Certainly, "Not every man is white", i.e., "It is not the case that every man is white" is true if
there are no men. But since "Not every man is white" is the contradictory of "Every man is
white", the latter has existential import.
2. If A propositions are analyzed in a Fregean manner, then we have to accept that every
chimera is white (and blue and yellow etc.) precisely because there are no chimeras. On the other
hand, while accepting this much, we cannot accept that some of them are white (and blue and
yellow etc.).
3. On the Fregean analysis we also have to accept both that every chimera is white and that
every chimera is not white.
4. If no miner owns a donkey, then on the Fregean analysis we have to accept that all miners
who own some donkeys insure them, but some of them do not.
5. On the Fregean analysis Mary has to accept that every boy at the party kissed her but only
because there were no boys at the party. (Poor girl!)16
6. "Not every ... is ..." simply means the same as "Some ... is not ...". This is why we can
define the one with the other. So filling in the blanks we get equivalent propositions.
So, it seems we had better accept both (a) and (b). But then we have to accept that "Some man
is not white" is true when there are no men.
To appease our remaining qualms we may add even the following. What seems to require from
this sentence that it have existential import is that it seems to attribute to a man the property of
being not-white. But it may be said that it is precisely this reading of this sentence on which it is
not equivalent to "Not every man is white". According to this reading the subject term falls
outside the scope of the negation, and this is why according to this reading "Some man is not
white", i.e., "Some man is not-white" is equivalent to "Something is a man and is not white".
In GTL this equivalence can be represented in the following manner:
(x.Mx)((W)x.)  (x)(Mx & Wx)
But, it may be argued further, this sentence has also another reading, according to which the
subject term falls within the scope of the negation, and it is this reading on which this sentence is
equivalent to "Not every man is white". Just as "The present King of France is not bald" can be
taken to express either "The present King of France has the property of being not-bald" (internal
negation) or "It is not the case that the present King of France is bald" (external negation),17 so
"Some man is not white" may be taken to express either "Some man has the property of being
not-white" or "It is not the case that every man has the property of being white". So "Some man
is white", if negation is taken to have wider scope than "man" in it, is equivalent to "Not every
man is white", and this further to "Either nothing is a man or something is a man and is not
white".18 In GTL:

"Aloofness Principle" (in brief: "Logic should not dictate to language users", see item: `aloofness' in the index of his That Most Subtle Question
cited above), makes room for both the "Fregean" and the 14th century nominalist analysis. As for Abaelard's solution see Abaelard, Dialectica,
ed. L.M. de Rijk, Assen, 1956, pp.175-177. Cf. W. and M. Kneale: The Development of Logic, Oxford, 1971, p.210-211
16
Cf. J. Barwise and R. Cooper, "Generalized Quantifiers and Natural Language", Linguistics and Philosophy, 4(1981), pp.159-219.
17
Cf. A.N. Prior: "Negation", in: P. Edwards (ed.): The Encyclopaedia of Philosophy, New York-London, 1967. Of course this distinction is
Russell's. But see also n.18.
18
So we have here a clear distinction between external (propositional) and internal (term) negation, which, in Keenan and Faltz's words, "seems
about the best we can do in a strictly two valued system". op. cit., p.248. For the expression "term negation" as opposed to "propositional
negation" (in medieval terminology, negatio infinitans vs. negatio negans, cf. Buridanus, Tractatus de Suppositionibus, ed. Maria Elena Reina,

32
(x.Mx)(Wx.)  (x.Mx)(Wx.)  (x)(Mx)  (x)(Mx & Wx)
And this is exactly what 14th century logicians said about the matter.19

3.3. DETERMINERS
As we have seen, GTL quite naturally incorporates a theory of determiners. In view of the
recently enormously increased and still growing interest in the field it seems appropriate to
consider here the relation of the approach incorporated in GTL to what is becoming the standard
approach to the matter.20 Let me refer to this approach as GQ, the "generalized quantifier"
approach.
In what follows I compare GTL and GQ with respect to certain of their heuristic features.
3.3.1. CAPACITY
The capacity of a logical theory can be measured by the magnitude of the portion of a natural
language which it is able adequately to represent. To be a bit more precise, let us say that a
semantic system S (i.e. <L,M>, where L is a formal language and M is a model-type) is adequate
to a portion of a natural language, PNL (where PNL is to be thought of as a subset of the set of
sentences of the given natural language), iff formulae of S are true just in those models which
represent states of affairs in which sentences of PNL represented by these formulae are true. In
short, S is adequate to a PNL iff it gives the correct truth conditions to sentences of PNL.
Now we can say that the capacity of S is greater than that of S' iff S is adequate to a greater
PNL than S', i.e., iff the PNL which S' is adequate to is a proper subset of the PNL which S is
adequate to. Such is the relation e.g. between GTL and SQT.
Now my claim in this section is that the capacity of GTL (or of some trivial extension of it) is
at least as great as that of any of the current versions of GQ. To substantiate this claim I should
have to prove that all sentences of, say, English adequately representable by some version of GQ
are also adequately representable by (some extension of) GTL, and that, perhaps, there are
English sentences adequately representable by (some trivial extension of) GTL that have no
adequate representations in any version of GQ. I conjecture that this could be done. Still, to show
how GTL works, instead of doing this, I only wish to render my claim probable by examining the
most suspicious cases. I take my examples from the astonishingly rich list of determiners
supplied by Keenan and Stavi.21
What should be clear at the outset is that every sentence of English expressible in the form:
Det NP VP is adequately representable in GTL. Note that this form comprises even "horrors" like
(11) All but at most 2O % of the students are females22
The GTL representation of (11) is the following:

Rivista Critica di Storia della Filosofia, 1957, pp.175-208, and pp.323-352, pp.328-329.), consider the following valid equivalence of GTL:
`(Ex.Fx)((-G)x.)' <=> `(Ex.Fx).(x. = y.-Gy)'. For the abbreviations applied here see the previous Essay.
19
Cf. the following text: "In negative propositions a term is either denoted not to refer to something or to refer to something of which the
predicate is truly denied, and so such a negative has two truth conditions like this: `A man is not white' has two conditions for its truth, for it is
true either because no man exists, and so is not white either, or because some man exists, and is not white." Ockham, op. cit., pp.218-219. Cf
also pp.284-285, 623-630; Paulus Venetus, Logica, Georg Olms, 1970, p.76. "Sexta regula..."; Moody, op. cit., pp.32-53. See also the previous
Essay.
20
See e.g. Barwise-Cooper, van Benthem, van Benthem-ter Meulen, Brown, Evans, Keenan-Faltz, Keenan-Stavi, op. cit.
21
Keenan-Stavi, op. cit., p.281.
22
In the subsequent formalizations I shall use expressions of English as belonging in the appropriate syntactic category of GTL. See n.9.

33
(11') (All but at most 2O% of the x.student(x))(female(x.))
The interesting truth (satisfaction) condition for (11') is the following:
f((11'))=1 iff it holds for all but at most 2O% of the uRg(x.student(x))f that
f[x.:u](female(x.))=1.
(Where the meaning of the determiner might be specified further.)23 What this condition says
is that in order to verify (11') in an assignment f in a model M you have to take the individuals
which can be the values of the variable `x.student(x)'and see if all but at most 2O% of these
satisfy the predicate `(are) female' in M. And this is how it should be. It should also be noted here
that the above form comprises even cases of Keenan and Stavi's Possessives and Det+Adjective
Phrase. For example,
(12) John's two cycles were stolen
in GTL becomes
(12') (Two x.cycle(x).John's(x.))(were stolen(x.))
while
(13) All liberal and most conservative delegates voted for that bill
through
(13') All liberal delegates voted for that bill and most conservative delegates voted for that bill
becomes
(13'') (All x.delegate(x).liberal(x.))(voted for that bill (x.)) & (most
y.delegate(y).conservative(y.))
(voted for that bill (y.))
Note that the nominal conjunction in (13) is represented by a sentential conjunction in (13''). It
is a somewhat similar device (breaking up the apparent unity of the NP) that makes available
certain cases which are at first sight not available for GTL, such as sentences containing many-
place determiners. For example,
(14) More male than female students passed the examination
seems to constitute a problem. But if we paraphrase (14) as
(15) More male students passed the examination than female students passed the examination,
then its GTL representation should be straightforward as
(15') (More x.student(x).male(x.))(passed(x.)) (than y.student(y).female(y.))(passed(y.))
the relevant truth condition being
f((15'))=1 iff it holds for more uRg(x.)f that
f[x.:u](passed(x.))=1 than for wRg(y.)f that f[y.:w](passed(y.))=1,
which may be specified further as
f((15'))=1 iff |{uRg(x.)f: f[x.:u](passed(x.))=1}| >
|{uRg(y.)f: f[y.:u](passed(y.))=1}|
And what this condition says is that in order to verify (15') according to an f in an M you have
to take the individuals which can be the values of `x.student(x).male(x.)' and see if more of these
satisfy the predicate `passed( )' than of those that can be the values of `y.student(y).female(y.)'.

23
Of course, by leaving the complex determiner of this sentence unanalyzed I do not intend to imply that it cannot be analyzed further. Simply,
from our present point of view the internal structure of the determiner is irrelevant. In what follows I will give examples of how determiner
denotations may be specified in GTL anyway.

34
Again, this is how it should be. The analysis of (15) in terms of (15') seems to be a rather trivial
extension, which is certainly susceptible of a canonical incorporation into the machinery of
GTL.24
Now this kind of treatment, provided we introduce also intensional models for GTL seems to
extend even to cases of Intensional Dets of Keenan and Stavi. For example,
(16) A surprisingly large number of doctors attended the meeting
may be analyzed as
(16') (A surprisingly large number of x.doctor(x)) (attended the meeting (x.))
the relevant truth condition being
f((16'))=1 iff it holds for a surprisingly large number of uRg(x.)f that f[x.:u](attended the
meeting(x.))=1
What this condition says is that in order to verify (16') in an f in an M you have to count the
number of the individuals that can be the values of `x.doctor(x)' satisfying `attended the meeting(x.)' and see if this
number is surprisingly large. And whether this number is surprisingly large should depend on further features of the
model (e.g. the number of doctors at meetings in past situations of the model). Clearly, if we want to make this and
similar conditions more precise, then there is some work to be done here. Nevertheless, I think, that much should be
conceded that in view of its (possible) capacity GTL is at least as promising as existing versions of GQ. (For further
developments cf. sections 3.3.4. and 3.4., and the next Essay.)

3.3.2 FAITHFULNESS

Adequacy is only a necessary, but not sufficient condition of faithfulness of semantic


representation. To see this, consider the following. Equivalent formulae of a semantic theory may
be said to be equally adequate representations of the same natural language sentence in the sense
that these formulae are true just in those models which represent states of affairs in which the
sentence in question is true.
For example, we may say that
(17a) (x)(x = Peter and walks(x)) and
(17b) walks (Peter)
are in this sense equally adequate representations of
(17) Peter walks.
Still, we would say that (17b) is a more faithful representation of (17) than (17a). We could
also say that (17a) gives a more faithful representation rather of the equivalent
(17c) Something is identical with Peter and walks.
And we would say these on the basis that in (17) and (17b) (and also in (17c) and (17a)) we
would be able to identify strictly corresponding syntactic units with the same semantic function,
whereas with (17) and (17a) this is not the case.25 So faithfulness, beyond adequacy, seems to

24
The "distribution" of the VP between the arguments of the Det in the analysis of (15) seems to be supported by the following consideration:
the arguments of a many-place Det may take also different VP's, so it is only on account of "economy" that the same VP does not occur twice in
(15). For example, we may correctly say: "More female students passed than male students failed", where we feel no temptation to treat "More
male students than female students ..." as a syntactic and semantic unit. (A similar argument would apply to the analysis of (13) as well.) Cf.
Keenan-Stavi, op. cit., pp.282-284.
25
Or consider the following: even if (x)(CxWx & -Wx) is equivalent to -(Ex)(Cx), whence both would be an adequate representation of, say,
"Nothing is a chimera", still, we would be inclined to say that it is rather the latter and not the former that is the "proper" formalization of this
sentence, motivated by considerations of faithfulness to the syntactic and semantic structure of the sentence. For further examples of the
discrepancy between adequacy and faithfulness consider Russell's elimination of descriptions, Quine's elimination of proper names or, in
general, talk about "surface" or "grammatical" structure as opposed to "deep", "logical" structure.

35
require even a close syntactic and semantic correspondence (or "match", as opposed to the
mismatch observed by Barwise and Cooper (p.164.)) between what represents and what is
represented. In particular, it seems to require that there be strictly corresponding syntactic units of
the two with the same (type of) semantic function. Now my claim in this section is that in this
sense GTL is more faithful to the same portion of English than GQ.
The first thing to be noted about this point is that GTL and GQ (at its best)26 exhibit the same
syntactic correspondence to the portion of English they represent.
Clearly, there is a strict correspondence between the (from our point of view) principal
syntactic units of GQ and English, on the one hand, and between those of English and GTL, on
the other, as is shown in the following diagram:

GQ: Quantifier
┌────────┴─────────┐
Determiner Set expression

English: Determined NP (DNP)


┌────────┴─────────┐
Det Common noun phrase (CNP)

GTL: Quantifier
┌────────┴─────────┐
Det Restricted variable

What I wish to show, therefore, is that it is its closer semantic correspondence that renders
GTL more faithful to English than GQ. To put it briefly, DNP's of English (including proper
nouns), unlike quantifiers of GQ, do not denote families of sets, but, instead, behave like the
quantifiers (and individual names) of GTL do.
This is clearest in the case of proper nouns. The denotation of the NP "Harry" say, in, "Harry
walks"27 is not the set of "properties", i.e., subsets of the universe of which Harry is an element.
For such a simple statement is about the denotation of its subject. But this statement is about
Harry, the man, who is certainly not a set of sets containing him. Moreover, we can go on: "But a
moment later he suddenly stops", where "he" is meant to be coreferential with "Harry". But the
denotation of "he" cannot be fixed as the set of sets (properties) containing Harry, for this set,
unlike Harry, cannot lose a property (walking) without losing identity.
But GQ is not better off with other DNP's either. For in GQ, say, "Every man is a man" is
parsed in the same way as "Harry is a man". The DNP in both sentences has the same semantic
function, namely that of denoting a family of sets. Now this being so, just as the sentence "Harry
is a man" is about the denotation of "Harry", so is the sentence "Every man is a man" about the
denotation of "Every man". But about which one can truly say that it is a man that is a man. So
the denotation of "Every man", call him "Everyman", if you like, is a man. And so we have a
fellow human here, who is a set, has cardinality, but lacks, e.g., any specific shape, color, etc.,
26
When I say "at its best", I have in mind the treatment of Keenan and Stavi and Keenan and Faltz of adjectival constructions (as not being
conjunctions).
27
I am speaking here deliberately about the NP, not the lexical item: see Barwise-Cooper, op. cit., pp.166, 174.

36
and who, in general, shows very strong resemblance to the Platonic idea of Man. But unless we
are unable to find a more fitting analysis for its DNP we should not accept that the truth of
"Every man is a man" commits us to such a weird entity.
Moreover, if we take a closer look at what we are supposed to do, and what we in fact do
when checking the truth of a quantified statement, we can see that GQ gives an adequate, but
entirely unfaithful analysis of quantified statements. With GQ, we are told,28 in order to check
the truth of, say, "Most senators are males", we have to do the following. First, we have to take
all those sets that contain a majority of senators. Second, we have to check whether any of these
is a subset of the set of males. If we find such a set, then the sentence in question is true.
Now, contrary to this description, what we in fact do is rather this: we take the senators, one by
one, check if they are males, and see if those who are males constitute a majority of all of the
senators. What this latter description shows is that the DNP ("Most senators") in this sentence
does not serve to denote any set, indeed, anything at all. The DNP is not a semantic unit.
Semantically, it splits up into Det and CNP, which have radically different semantic functions.
The CNP marks off the range of individuals to be considered whether they satisfy the VP, while
the function of the determiner is to determine how many (what proportion) of these should satisfy
the VP that the whole sentence be true, and, occasionally, to impose further restrictions upon the
range of individuals to be considered, as in the case of "The (n)", "Both", etc.29 But this is exactly
what GTL tells us about the function of Dets and CNP's. And this is why I take it to be more
faithful to English.
In fact, GTL offers us the formal explication of our most natural intuitions about quantified
statements. According to these intuitions, as has been stated above, the function of a CNP is to
mark off the range of individuals to be considered, and the basic function of the determiner is to
determine how many (what proportion of) these should satisfy the VP that the whole sentence be
true. So on this approach the "denotation" (semantic value) of the determiner is a function from
the range of values of the CNP and from the "denotation" of the VP (a function from individuals
to truth values)30 to truth values. Formally this can be spelled out as follows:
(Den(Q)) Den(Q)(N,V){1,0},
where QD (the set of determiners of GTL), N is either a nonempty subset of W or is {0} (i.e.,
N holds the place of the range of the CNP, the restricted variable), and V is a function from N to

28
Barwise-Cooper, op. cit., 191-193.
29
Cf. G. Evans, "Pronouns, Quantifiers and Relative Clauses" and "Appendix", Canadian Journal of Philosophy, 7(1977), pp.467-536, pp.777-
796, p.778. I think I should call attention here to the ease with which Dag Westerstahl's context sets (see: D. Westerstahl, "Determiners and
Context Sets", in: J. van Benthem-A. ter Meulen: Generalized Quantifiers in Natural Language, Foris Publications, 1985, pp.45-71. can be
introduced into the apparatus of GTL:
Syntax: V:=VindUVpred; XVpred, tT ==> X(t)F
Semantics: M:=<W,R,0,C>, where C, the context assignment, is defined as follows: XVpred ==> C(X) is a part of W.
Now with these clauses added to GTL, wherever Westerstahl writes
DX A B we can write (Dx.X.A)(B). So e.g. instead of his
(D1 of the D2 A)B <=> D1X A B, if D2X A M, false otherwise (if we make the "Russellian move" of taking "false" instead of "undefined")
we can write:
(D1 of the D2 A)(B) <=> (D1 x.X.A)(B) & (D2 x.Xx.A)(Ex.), where R(E)=W. Now it is along these lines that GTL would be able to
cannibalize Westerstahl's (to my mind ingenious) treatment of the definite article and, in general, of definites. For the treatment of the "plural
condition" see sect. 3.3.4. below.
30
For a different, and to my mind more faithful, account of the semantics of VP's, where the semantic value (signification) of a VP is a function
from individuals to individualized properties, see the last three Essays of this volume.

37
{1,0}. On this account a quantified statement (Qv)(A) is true in an assignment f iff the denotation
of Q yields truth for the value range of v and for the denotation of the VP of A.31 So, if we denote
the VP of A as v(A), then we can write:
(f11) f((Qv)(A))=1  Den(Q)(Rg(v)f,Den(v(A),f))=1,
where Den(v(A),f), the denotation of the VP of A according to an assignment f, is defined as
follows:
(Den(VP)) Den(v(A),f) is a function from Rg(v)f to {1,0} such that Den(v(A),f)(u)=1 iff
f[v:u](A)=1,
where, of course, uRg(v)f. That is, the denotation of the VP of A with respect to the assignment
f is a function from individuals to truth values such that it yields truth for a given u iff A is true
of u.
Now if to these we add the following stipulation:
(Den(Q)') Den(Q)(N,V)=1 iff for Q'uN, V(u)=1,
then, since by this and (f11) we get:
f((Qv)(A))=1 iff for Q'uRg(v)f, Den(v(A),f)(u)=1,
by (Den(VP)) we get back:
(f10) f((Qv)(A))=1  for Q'uRg(v)f, f[v:u](A)=1,
which shows the adequacy of the above clauses.
3.3.3. TREATMENT OF LINGUISTIC UNIVERSALS

The above clauses, determining explicitly the logical type of determiner denotations of GTL,
make it possible to prove very general metatheorems concerning GTL involving arbitrary
determiners. These metatheorems are the GTL counterparts of the so-called linguistic universals,
which provide the basic motivation for the GQ approach.32 As it happens, the 14th century
terminist attitude toward existential import and negation incorporated in the semantics of GTL
replaces some of these by new ones.
For example, conservativity, which, as a proposed metatheorem for GTL may be expressed as
(CONS) (Qv.A)(B)  (Qv.A)(A & B),
is evidently invalid for GTL.
The simplest way in which this can be shown is to take an instance of it. According to (CONS)
the following equivalence should be valid:
'(Some x.chimaera(x))(run(x.)'  `(Some x.chimaera(x.))(chimera(x.) & run(x.))'
(That is, "Some chimera does not run" should be equivalent to "Some chimera is a chimera
and does not run".) But, as the considerations of sect.3.2. above show, since negation is taken to
have wide scope here, according to GTL the left hand side of this equivalence, there being no
chimeras, is actually true, while the right hand side is false. So the equivalence does not hold.33
What replaces (CONS) for GTL is the following:
31
By the VP of A here I mean what remains from A if we leave off from it the occurrences of v bound by Q. So what I call here VP's are rather
complex predicates or predicables.
32
See van Benthem, op. cit., Introduction.
33
Another, perhaps more convincing, counterinstance to (CONS) is provided by the following sentence: "If the Present King of France runs,
then he moves", i.e., (Ix.Kx)(R(x.) ->.M(x.)), which is certainly true. However, according to (CONS), this should be equivalent to: (Ix.Kx)(K(x.)
& R(x.) -> M(x.)), i.e., "The present King of France is present King of France and if he runs, then he moves". But there being no present King of
France this latter is certainly false. So the equivalence fails to hold.

38
(CONS*) (v)(A)  `(Qv.A)(B)  (Qv.A)(A & B)'
By means of the clauses introduced in the preceding section (CONS*) can be proved valid for
GTL.
More to our present concern, however, is that in consequence of the "adequacy result" of the
preceding section we need no separate inductive verification of the empirical validity of
(CONS*) in the manner of Keenan and Stavi.34 Determiner denotations of GTL are to comply
with (Den(Q)') above, and so, as we have seen, (f11) is equivalent to (f10), the "natural criterion"
of truth of quantified statements. So whatever way we specify the denotation of a determiner
constant, insofar as this specification respects the clauses of the preceding section, any formula
formed with this determiner will be consonant with both (CONS*) and (f10), thereby being able
to supply empirically valid instances of (CONS*).
For example, let us take `most'. We may specify the denotation of `most' in the following
manner:
Den('most')(N,V)=1  For most uN, V(u)=1,
which, for the sake of "exactitude" may be specified further as follows:
For most uN, V(u)=1  |N|-|C|<|C|, where C:={uN: V(u)=1}. (Intuitively: more N's satisfy
V than not.)
Now, by these and (f11):
f('(most v)(A)')=1  |Rg(v)f|-|C|<|C|, where
C:={uRg(v)f: f[v:u](A)=1}. (See (Den(VP)) above.)
But this, as can easily be seen, given the above stipulations, is equivalent to the following
instance of (f10):
f('(most v)(A)')=1  for most uRg(v)f, f[v:u](A)=1.
But then whatever holds of `most' specified by Den('most') may not prove to be empirically
invalid, as `most' so specified complies with (f10), the "natural" or "empirical" criterion of truth of quantified
statements.

3.3.4. FURTHER INSIGHTS: PLURAL DISTRIBUTIVE REFERENCE


Another heuristic advantage of GTL over GQ seems to be that beyond covering all the ground
covered by GQ it also breaks new ground of investigation by providing further insights into the
referring function of general terms. In particular, GTL seems to give us a clue to plural
distributive reference. (As opposed to the collective reference in: "The apostles are twelve" or
"The wolves surrounded the deer". In these cases we cannot "descend to singulars" by:
"Therefore: Peter is twelve", or "Therefore: this wolf surrounded the deer".)
This use of general terms may be represented in GTL by introducing the following clauses:
Syntax:
If v.AVres, then so is v:A. Intuitively, v:A is to represent the plural form: "the A's".
Semantics:
f(v:A)=f(v.A) if |Ext(v,f,M)(A)|>1, otherwise f(v:A)=0, where Ext(v,f,M)(A):={uRg(v)f:
f[v:u](A)=1}.

34
See op. cit. sect.2.6.

39
This clause is intended to capture the natural idea that a distributively used plural term, say,
"boys" in "All the boys present laughed", refers to boys, i.e., at least two boys. In any case this
intuition seems to account for the fact that this sentence is false if there is only one or no boy
present. Similarly, by the introduction of the above clauses it becomes possible in GTL to
distinguish between the false "All moons of the Earth are round" and the true "Every moon of the
Earth is round". Moreover, with this extension of GTL we can distinguish even between the
following two sentences:
(18a) Some man present laughed
(18b) Some men present laughed
As can easily be tested, from among the corresponding GTL formulas
(19a) (Some x.man present(x))(laughed(x.))
(19b) (Some x:man present(x:))(laughed(x:))
in a model in which there is only one man present (19a) is true, while (19b) is false. However,
this much is not enough for distinguishing (18a) and (18b) when there are many men present but
only one of them laughed. For in this case (18a) is true, while (18b) is false. So a further
requirement for the truth of (18b) is that there should be at least two individuals from the range
of its CNP that satisfy its VP. But this is surely not required by the determiner but by the plural
form. So we have to state even the following condition:
if v is a plural variable (i.e., if it is of the form `w:A') the range of which in a model M
according to an assignment f is such that Rg(v)f is not identical with {0}, and A is a formula in
which v occurs free, then (in M):
f((Qv)(A))=1  |{uRg(v)f: f[v:u](A)=1}|>1.
With this clause added, GTL, through (19a) and (19b) gives the correct truth conditions for
(18a) and (18b).
But we must stop here. Certainly this is not the proper place to deal with all the puzzles
surrounding plural reference. What is interesting here is only that GTL provides promising suggestions also in this
direction.

3.3.5. ONTOLOGICAL SIMPLICITY

Beyond this technical potential there is another feature of GTL that should be appealing to
those who are prone to interpret a formal semantics realistically, i.e. not as a mere technical device with no
ontological significance at all. I have in mind the essentially first order character of GTL. GTL, contrary to current
approaches, in its treatment of distributive plurals and quantification involves no quantification over sets (or other
collections)35, and contrary to GQ it involves no essential reference to families of subsets of the universe.

3.3.6. EXPLANATORY POWER

By now it should be clear that GTL possesses an explanatory power that GQ does not. For
GTL, in consequence of its faithfulness, gives us a clue to understanding why we do just what we
do when checking the truth of a quantified statement. The basic clue GTL offers us for this is the
clear cut distinction between a common term's occurring in predicative or in referring position.

35
Cf. G.M Wilson, "Pronouns and Pronominal Descriptions", Philosophical Studies, 1984, pp.1-30. Cf. this with G. Boolos, "To Be Is to Be the
Value of a Variable (or to Be Some Values of Some Variables)", The Journal of Philosophy, 1984, pp.430-449. An interesting treatment of
plurals within the GQ approach is found in S. de Mey, On the Semantics of Plurals, Groningen, 1986.

40
This distinction comes out very clearly if we compare the following two sentences together with
their GTL representations:

(20a) Every Hungarian is Hungarian


 x.Hx H( )
└──────────┘

(x.Hx)(H(x.))
the analysis tree of which should look like this:
S
┌─────┴─────┐
NP VP
┌───┴───┐ ┌───┴───┐
Det N Cop Adj

on the other hand, consider

(20b) Every Hungarian is a Hungarian


 x.Hx =  y.Hy
└───────────────┘
(x.Hx)(y.)(x. = y.Hy)

the tree of which looks like this:

S
┌─────┼─────┐
NP │ NP
┌────┴──┐ │ ┌───┴───┐
Det N Cop Det N

In (20a) only the first occurrence of "Hungarian" is in referring position, so only that can be
bound by a determiner. In (20b), however, both occurrences of the same term are in referring position, so both are
to be bound to form a complete sentence. Now from this we can see why determiners "go together" with NP's in a
way they do not with VP's. (For example, why we cannot say "Some man some runs", whereas we can say "Some
man is some runner".)36 In the approach provided by GQ, most clearly in the "binary, relational approach" of e.g.
Evans or van Benthem, where both CNP's and VP's have the same kind of semantic values (subsets of the universe),
this fact is merely taken for granted instead of being explained. Indeed, in general, the semantic faithfulness of GTL,
beyond mere adequacy also attained by GQ, makes it possible to use GTL for providing representations of
explanatory value even in fields where GQ does not have much to tell us. As van Benthem puts it: "Summarizing,
there is every reason to expect that the present approach will generalize to all types of determiners. Still, this does
not mean that the generalized quantifier perspective is a unique best approach to noun phrases and determiners. In
fact, there are important phenomena such as iteration of determiners or anaphoric relations, about which it has little,
if anything to say."37 In the next section we turn to this latter topic.

36
And this is why medieval "two-name theorists", like Ockham, are ready to accept quantification of predicate terms, while "inherentists", like
St. Thomas Aquinas, are reluctant to do so. (Cf. St. Thomas's Commentary on Aristotle's Perihermeneias, lb.1. lc.10.)
37
op. cit. pp.23-24.

41
3.4. PRONOUNS, ANAPHORA AND RELATED ASININITIES
By now we can regard it as an established fact that certain anaphoric pronouns, E-type
pronouns, as Evans calls them, do not behave like bound variables of SQT. The problem, then,
lies in finding a satisfactory positive account of their behavior.
The following "donkey sentences" supply cases in point:
(21) Every miner who owns a donkey beats it
(22) If a miner owns a donkey he beats it
(23) Some miner who owns a donkey beats it
(24) * Pedro owns no donkey and Sancho beats it
(25) Most miners who own a donkey beat it
A currently widely recognized approach to the problems posed by these sentences is the one
provided by Kamp's discourse-representation semantics, which offers intuitively correct
interpretations for the pronouns occurring in (21)~(24).38 However, as in his illuminating paper
Barry Richards has argued, this approach seems "both to succeed and founder for the same
reason", namely because of its being tailored specifically to examples like these, in consequence
of which it is hard to see how it could be accommodated, if at all, to cases like (25).39
The other approach which Richards examines (the approach of Evans) involves interpretation
of pronouns in terms of certain appropriately chosen descriptions.
For example, for (21) the corresponding interpreting paraphrase would be:
(26) Every miner who owns a donkey beats the donkey he owns
The analysis tree of this sentence would show how (21) could be derived from (26) by an
appropriate pronominalization rule for the description.40
However, Richards rejects this approach on the basis that, since in (26) the description implies
uniqueness (for each choice of a miner), therefore (26) does not supply intuitively correct truth
conditions for (21).
In his own positive account Richards therefore proposes reconstruction for pronouns in terms
of certain appropriately constructed artificial complex terms. For example, the reconstruction for
(21) would be:
(27) Every miner who owns a donkey beats y[donkey(y) & he owns(y)]
According to Richards' syntactic and semantic definitions the occurrences of `y' in this
sentence are not free, and the semantic import of a complex term like this is invariably universal,
so that (27) boils down to this:
(28) Every miner who owns a donkey beats every donkey he owns,
which is intuitively correct.
However, the invariably universal import of Richards' terms leads to counterintuitive results
in respect of other examples. For example, Richards' reconstruction for (23) would be
(29) Some miner who owns a donkey beats y[donkey(y) & he owns(y)],
which has the import of
(30) Some miner who owns a donkey beats every donkey he owns.
38
H. Kamp, "A Theory of Truth and Semantic Representation", in: J. Groenendijk et al. eds., Formal Methods in the Study of Language,
Amsterdam, 1981
39
B. Richards, "On Interpreting Pronouns", Linguistics and Philosophy, 7(1984), pp.287-324. pp.289-291.
40
ibid. pp.291-294.

42
As Richards readily admits, many would find (30) as a too, almost intolerably strong
interpretation of (23).41 Nevertheless, Richards argues that intuition is not a secure guide here,
and the explanatory value of his approach (with its explanation of the type of anomaly involved
in (24)) might supply sufficient grounds for dismissing contrary intuitions.
But I think that GTL offers us a better way out. A partial GTL reconstruction of (21), parallel
to (27) is the following:
(27a) Every miner who owns a donkey beats(y.donkey(y) & he owns(y))
The basic difference between (27) and (27a) is that in (27a) the restricted variable is free, so it
can be bound by a determiner. The same holds of the parallel of (29):
(29a) Some miner who owns a donkey beats(y.donkey(y) & he owns (y))
So with GTL we are free to determine the import of the term obtained by reconstructing
the anaphoric pronoun.
Now, as George M. Wilson argues, E-type pronouns are to be interpreted in a way as
individual parameters in natural deduction systems.42 What this boils down to is that its extra-, as
well as intrasentential context determines for an anaphoric pronoun "both a range of reference
and a mode of generality".43
So, e.g., the intrasentential context of "it" in (21) determines (though not necessarily)44 that 1.
for each choice of a miner who owns a donkey we should consider only donkeys he owns
(range), and that 2. we should consider every donkey he owns (mode) if it is beaten by him. Now
this is exactly what the following reconstruction tells us.

41
Another, more serious, objection to Richards' approach is that it is apparently incapable of dealing with examples like this: "Every miner who
owns some donkeys beats some of them". His terms for reconstructing pronouns being unquantifiable, I wonder what Richards' reconstruction
for "them" in this sentence would be like.
42
G.M Wilson, "Pronouns and Pronominal Descriptions", Philosophical Studies, 1984, pp.1-30.
43
ibid. p.l7. The mode of generality is marked in the reconstruction by appropriate indices on the occurrence of the anaphoric pronoun. These
modes may be e.g.: U, universal, E, particular, I, definite or R, referential, as the context requires. In the present paper I do not deal with the
problem of how context determines these modes.
44
Of course, intuitions may vary here, especially concerning the mode of generality, which is entirely to be divined from context.

43
┌─────────────┐
(21a) Every miner who owns a donkey beats itU
 x.Mx . O( )( )  y.Dy B( )( ) │
└─────┼─┼──┘ └───┼───┘ │ └──┘
│ └─────────┘ │
└───────────────────────┘

┌──────────────────────┐
(Vx.Mx.(Ey.Dy)(O(x.)(y.)))(B(x.)(itU))

itU
┌───┴───┐
U it = donkey which he owns
z.Dz . x. O( )( )
│ └────┘ │
└───────────────────┘

(x.Mx.(y.)(O(x.)(y.Dy)))(z.)(B(x.)(z.Dz.O(x.)(z.)))

(21b) Every miner who owns a donkey beats every donkey he owns

Similarly with (23); only here the context determines for "it" to be "itE", so the GTL
reconstruction of (23) will look like this:

┌─────────────┐
(23a) Some miner who owns a donkey beats itE
 x.Mx . O( )( )  y.Dy B( )( ) │
└─────┼┼───┘ └───┼───┘ │ └──┘
│└──────────┘ │
└───────────────────────┘

┌─────────────────────┐
(x.Mx.(y.Dy)(O(x.)(y.)))(B(x.)(itE))

itE
┌───┴───┐
 it = donkey which he owns
z.Dz . x. O( )( )
│ └────┘ │
└───────────────────┘

(x.Mx.(y.)(O(x.)(y.Dy)))(z.)(B(x.)(z.Dz.O(x.)(z.)))

(23b) Some miner who owns a donkey beats a donkey he owns

which is provably equivalent to

44
(23c) (z.Dz)(x.Mx.O(x.)(z.))(B(x.)(z.))

A donkey is such that a miner who owns it beats it,

which is equivalent further to the standard formula:

(23d) (z)(x)(Dz & Mx & Oxz & Bxz)

However, that (23c) and, a fortiori, (23d) are merely adequate, but not faithful representations of
(23) is shown by the fact that if we negate the verb of the relative clause of (23) we get an
anomalous sentence, while with the sentence corresponding to (23c) this is not the case:
(23e) * Some miner who does not own a donkey beats it
(23f) Some donkey is such that some miner who does not own it beats it
To be sure, (23e) may have a reading on which it is all right, namely when it is read as equivalent
to (23f). However it also has an anomalous reading (analogous to 24) which (23f) does not have. Now GTL
provides a perfectly satisfactory explanation for this phenomenon:
FIRST (NON ANOMALOUS) READING (BOUND PRONOUN)
┌─────────────┐
(23e) Some miner who does-not own a donkey beats it
E x.Mx . ~ O( )( ) E y.Dy B( )( ) y.
│ └─────┼───────────┘ └──┼──┘ │ └───┘
│ └─────────────────┼──────────┘
└───────-─────────────────────┘

(y.Dy)(x.Mx.~(O(x.)(y.)))(B(x.)(y.))

Some donkey is such that a miner who does not own it beats it

which is equivalent to (23f), and this to the standard:

(y)(x)(Mx & Dy & ~Oxy & Bxy).

Note that in this reconstruction the determiner binding "donkey" has the largest possible scope.

45
SECOND (ANOMALOUS) READING (E-TYPE PRONOUN)
┌─────────────┐
(23e) Some miner who does-not own a donkey beats itE
 x.Mx . ~ O( )( )  y.Dy B( )( ) y.
└─────┼────────┼───┘ └──┼──┘ │ └───┘
└────────┼─────────┼──────────┘
└─────────┘

┌────────────┐
(x.Mx.~(Ey.)(O(x.)(y.Dy)))(B(x.)(itE))

itE
┌───┴───┐
 it = donkey which he owns
z.Dz . x. O( )( )
│ └────┘ │
└───────────────────┘

(x.Mx.~(y.)(O(x.)(y.Dy)))(z.)(O(x.)(z.Dz.O(x.)(z.)))

Some miner who owns no donkey beats a donkey he owns

Note that in this reconstruction the determiner binding "donkey" has the narrowest possible
scope; in particular, it has narrower scope than the negation. The result is that the sentence cannot be true (in the
same way as (24)). Indeed, the corresponding GTL formula is not satisfiable. As these reconstructions show, the
analyses provided by GTL reveal also a further possible reading, namely one which gives the determiner binding
"donkey" a "medium scope", as is shown in the following reconstruction:

46
THIRD (NON-ANOMALOUS) READING (E-TYPE PRONOUN)
┌─────────────┐
(23e) Some miner who does-not own a donkey beats itE
 x.Mx . ~ O( )( )  y.Dy B( )( ) y.
└─────┼───┼───────┘ └──┼──┘ │ └───┘
└───┼─────────────┼──────────┘
└─────────────┘

┌────────────┐
(x.Mx.(y.)~(O(x.)(y.Dy)))(B(x.)(itE))

itE
┌───┴───┐
 it = donkey which he does not own
z.Dz . x. ~ O( )( )
│ └─────────────┘ │
└────────────────────────────┘

(x.Mx.(y.)~(O(x.)(y.Dy)))(z.)(O(x.)(z.Dz.~O(x.)(z.)))

Some miner who does not own a donkey beats a donkey he does not
own,

which, again, is equivalent to (23f). I think this is why (23e) feels only "anomalous", and not
strictly contradictory as does e.g. (24) or the second reading of (23e).
Oddly enough, according to Richards' treatment, the anomaly in these sentences is not
contradiction (in a possible reading), but vacuous truth.45 Nevertheless, it seems to be quite clear
that the sentences: ""Pedro owns no donkey and Sancho beats it" and "Some miner who owns no
donkey beats it" are so far from being (even "vacuously") true that they are almost sheer
contradictions. (For how on Earth could one beat anyone's donkey if he or she does not own any?
How can one beat a nonexistent donkey?) GTL here, again, provides representations that fully
accord with intuition and, for that matter, with medieval solutions too. Consider the following
passage from Buridan: "It is a universal rule that an anaphoric pronoun (relativum) need not
stand for all those things for which its antecedent stands, indeed, it refers only to those supposita
of its antecedent for which the categorical proposition in which the antecedent occurs is verified;
so this is false: `An animal is a man and it is a donkey'. From this rule it appears immediately that
an anaphoric pronoun (...) may well be more restricted than its antecedent, e.g. when I say: `A
man runs and he is white', the antecedent `man' refers indifferently to every man, though
indefinitely, but this term: `he' does not so refer to all, but only to those of whom this proposition
`A man runs' is true; and if this were true of nothing, then the pronoun would refer to nothing, so

45
See his op. cit. pp.318-319.

47
both members of the conclusion would be false if no men were running, even if all men were
white."46
Now let us take the following reconstruction:

┌───────────┐
A man runs and heE is white
x.Mx R( ) & │ W( )
└─────┘ └──────┘

heE
┌───┴───┐
 he = man who runs
y.My . R( )
└────────┘

(x.Mx)(R(x.)) & (y.My.Ry.)(W(y.))

A man runs and a man who runs is white

Consider the case when no man runs: this means that R(M)R(R)=0 (the intersection of the
extensions of "man" and "run" is empty in this model). Then, clearly, since for any f, f(y.My.Ry.)=0, whence
Rg(y.My.Ry.)f={0}, the second conjunct of the conjunction is false as well as the first, in perfect agreement with
Buridan's solution.47

4. CONCLUSION
As we could see in the Introduction, the problems of application of SQT to natural languages
rooted in the Fregean dogma that the Bedeutung of a general term in whatever grammatical
category is essentially predicative in nature, and so a general term, even a common noun,
functions as a predicate even in subject position.48 It is at this point that Frege broke most
radically with the analyses of traditional (Aristotelian, medieval scholastic) logic, and, indeed,
with the real structure of natural languages. The most noteworthy thing about the recent history
of the topic, as I see it, is that the discovery of "anomalies" in the application of SQT to natural
languages (problems with "pleonotetic logic",49 "donkey sentences", etc.) has left the "paradigm
46
Buridanus, Tractatus de Suppositionibus, ed. Maria Elena Reina, Rivista Critica di Storia della Filosofia, 1957, pp.175-208, and pp.323-352,
p.337.
47
Cf. also Paulus Venetus, Logica, Georg Olms, 1970, p.52. "Secunda regula ..."
48
The basic texts are G. Frege, "Funktion und Begriff", and "Über Begriff und Gegenstand", in: Funktion, Begriff, Bedeutung. Fünf Logische
Studien, Göttingen, 1966.
49
The expression is from P.T. Geach, "Quine's Syntactical Insights", in: Logic Matters, Berkeley and Los Angeles, 1972, p.125. By the way, in
this paper Geach has left "as an exercise to the reader" the following sentence: "Almost every man who borrows a book from a friend eventually
returns it to him". Geach has claimed to have found "no obvious paraphrase, or account of the syntax" of this sentence. Now, since the whole
reconstruction diagram would be too complicated, let me simply give here the main points of it together with the "obvious paraphrase".
Vocabulary: G: almost every; B( )( )( ): ( )borrows( )from( ); Bk( ): book( ); R( )( )( ): ( )eventually returns( )to( ): I: the; F( ): friend( ).
Reconstructions: itU = (every) book that he borrows from a friend; himI = (the) friend he borrows it from. Hence,
(Gx.Mx.(y.)(z.)(B(x.)(y.Bk(y))(z.Fz)))(y.)(Iz.)(R(x.)(y.Bk(y).(z.)(B(x.)(y.)(z.Fz)))(z.Fz.B(x.)(y.)(z.))) That is: "Almost every man who
borrows a book from a friend eventually returns every book he borrows from a friend to the friend he borrows it from". (Supposing that no book
is borrowed from more than one friend.) Note that in this paraphrase all pronouns are bound, indeed, they are strict equivalents of the
corresponding bound restricted variables.

48
constituting" Fregean dogma essentially intact: general names are symbolized uniformly with
predicate letters, which are interpreted as denoting Fregean concepts, or, equivalently, subsets of
the universe. The result is "adding epicycles", construing NP's as generalized quantifiers,
departing in many ways from the usual construction of SQT,50 better fitting the data, but at the
expense of getting even farther away from the actual semantics of natural languages.
In this paper I have argued that if we take the courage to discard the dogma and follow the lead
of scholastic logic concerning the semantics of general terms, then we can arrive at a formal
semantics that, while incorporating the whole apparatus of SQT, is wholly consonant with
scholastic logic, resolves the above-mentioned "anomalies", and gives us new insights into the
workings of language, thereby connecting traditional logic, SQT and modern logico-linguistic
research.

50
Cf. n.7. A curious thing, too, is that the "Fregean dogma" survives even in systems constructed in conscious opposition to Frege: in McIntosh's
semantics to Sommers' term-logic, general terms denote subsets of the universe. So, again, contrary to scholastic theories, the general terms of
this theory do not refer to individuals.

49
EXISTENCE, QUANTIFICATION AND THE MEDIEVAL THEORY OF
AMPLIATION

1. INTRODUCTION

(1) Bucephalus is dead


(2) What is dead does not exist
Therefore,
(3) Bucephalus does not exist
Therefore,
(4) something does not exist
In my opinion this is a conclusive argument for the thesis that something does not exist. As is
well-known, however, many philosophers regard this thesis as paradoxical in a way, and,
consequently, they would make several objections to the simple reasoning that led to it above.
In what follows I firstly deploy some typical objections to this reasoning. After this I give
replies to these objections. The informal discussion will be followed by the description of the
syntax and semantics of a formal language, AMPL, which, in my opinion, can serve as a suitable
frame of reference for handling the problems that occur in the course of the informal discussion.
As the language AMPL is a certain extension of the language MPL I used in an earlier paper for
reconstructing medieval logical theories concerning extensional contexts,1 I shall relate here the
language AMPL to a mediaeval theory concerning intensional contexts, namely to the theory of
ampliation.2 This will be followed by a formal reconsideration of the introductory objections
and replies. The paper will end with some historical and philosophical concluding remarks.

2. OBJECTIONS

1. In the argument above, if conclusion (3) is true, then it is false, therefore it is false. (By the
"consequentia mirabilis": ('(p -> -p) -> -p') For if a statement is true, then that which it is about
must exist. (About something non-existent one cannot make a true statement.) So, if (3) is true,
then Bucephalus exists. But then (3) is false. Therefore (3) is false, of necessity.3
1
"The Square of Opposition, ... etc.", Essay II. of this volume.
2
For medieval statements of this theory see e.g. the following texts: L.M. de Rijk (ed.), Logica Modernorum, Assen, 1967, II-2, pp. 264ff, 273ff,
344ff, 728; Peter of Spain, Tractatus, (ed. L.M. de Rijk), Assen, 1972, 194-197; William of Sherwood, Introductiones in Logicam, ed. M.
Grabmann, Sitzungsberichte der Bayerischen Akademie der Wissenschaften, 1937, 10, pp.82-85; S.F. Brown, "Walter Burleigh's Treatise De
Suppositionibus and Its Influence on William Ockham", Franciscan Studies, 32(1972), pp. 15-64, pp.60-61; Walter Burleigh, De Puritate Artis
Logicae Tractatus Longior with a Revised edition of the Tractatus Brevior, ed. Ph. Boehner, ST. Bonaventure N.Y., Louvain, Paderborn, 1955,
pp.47-53; Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, c.5, pp. 90-103; Buridanus, Tractatus de Consequentiis, ed. H.
Hubien, Louvain-Paris, 1976, pp.26-30; Buridanus, Tractatus de Suppositionibus, ed. Maria Elena Reina, Rivista Critica di Storia della
Filosofia, 1957, pp.175-208, and pp.323-352, pp.343-351; Albert of Saxony: Perutilis Logica, Georg Olms Verlag, Hildesheim-New York,
1974, Tr.II.10, f.15. For contemporary presentations see: A. Broadie: Introduction to Medieval Logic, Oxford, 1987, pp.31-37, 127; Peter King,
Jean Buridan's Logic, Dordrecht, 1985, pp.22-24, 43-45, 51-57, E.J. Ashworth: "Existential Assumptions in Late Medieval Logic", American
Philosophical Quarterly, 10(1973), pp.141-147; S. Ebbesen: "The Chimaera's Diary", in: S. Knuuttila-J. Hintikka: The Logic of Being, Helsinki,
1987.
3
Cf. Plato, "The Sophist", in: The Works of Plato, Cambridge-London, 1961, 236E-239B, W.V.O. Quine, "On What There Is", in: From a
Logical Point of View, Cambridge, Mass., 1961, pp.1-2; B. Russell, "On Denoting", in: H. Feigl-W. Sellars (eds.): Philosophical Analysis, New
York, 1949; L. Linsky, Referring, London-New York, 1967, pp.1-2.

50
2. In the whole argument existence is treated as a predicate. But existence is not a predicate.
For if it were, then, since a statement implies the existence of what it is about, all positive
existential statements would be necessary, and all negative ones would be contradictory. But this
is surely not the case: "There are extra-terrestrial intelligent beings" is contingent.4
3. "Exists" in the argument is used as if it were a first-level predicate. But it is a second-level
predicate. So the argument is senseless. That existence is a second-level predicate may be seen
from the following: "'pink iguanas exist' means nothing more nor less than `Something is a pink
iguana'; since the latter is formed by wrapping the second-level predicate `something' around the
first-level predicable `is a pink iguana' it seems all but certain that the `exist' in the former
sentence must also be deemed a second-level predicate."5
4. The inference from (3) is invalid. For "to be is to be the value of a bound variable",
therefore, if (3) is true, then Bucephalus cannot be a value of a bound variable. So the bound
variable in (4), if (4) is to be thought of as "For some x, x does not exist", cannot take up
Bucephalus as its value , and so, even if (3) is true, (4) may be false.6
5. The claim: "Something does not exist" is simply inconsistent. For "Something does not
exist" is equivalent to "There is something that does not exist" and this further to "There exists
something that does not exist" and this is clearly inconsistent.7

3. REPLIES

1. To the first objection we may give a twofold answer. First, we can point out that the rule "if
a statement is true, then that which it is about must exist" holds only for positive statements. For
a negative statement is true, iff the corresponding positive statement is not true, i.e., if tertium
non datur, is false. But the non-existence of its subject is just one of the possible reasons on
account of which a positive statement may be false. (It is not true that the present King of France
is bald, precisely because there is no such a person.) So, if its subject does not exist, i.e. if its
subject term fails to refer to anything, then a negative statement is true. As the schoolmen put it:
Negativa cuius subjectum pro nullo supponit est vera.8 Second, we may point out that one
must distinguish between referring to what does not exist (nonexistence of the reference) and
failing to refer altogether (failure of reference).9 As Wittgenstein says: "That is to confound the
reference (Bedeutung) of the name with the bearer of the name. When Mr. N.N. dies, we say that

4
Cf. A.J. Ayer, Language Truth and Logic, Pelican ed., 1980, pp.57-58; M. Kiteley, "Is Existence a Predicate?", Mind, 73(1964), pp.364-373,
pp.365-366.
5
See T.P. Flint, "Review of C.J.F. Williams's What Is Existence?", The Philosophical Review, 93(1984), pp.131-134, pp.131-132.
6
Cf. W.V.O. Quine, "Reference and Modality", in op. cit., p.145; R. Routley, "Existence and Identity in Quantified Modal Logics", Notre Dame
Journal of Formal Logic, 10(1969), pp.113-149, p.133.; L. Linsky, op. cit., pp.110-111.
7
See T. Parsons, "Are There Nonexistent Objects?", American Philosophical Quarterly, 19(1982), pp.365-371, p.365.
8
See e.g. Ockham, Summa Logicae, (ed. Ph. Boehner), St. Bonaventure, N.Y., 1974, II.12., pp.284-285, Buridanus, Sophismata, ed. T.K. Scott,
Stuttgard-Bad Cannstatt, 1977, c.2.; Albert of Saxony: Perutilis Logica, Georg Olms Verlag, Hildesheim-New York, 1974, tr.4.c.24.; E.J.
Ashworth: "Existential Assumptions in Late Medieval Logic", American Philosophical Quarterly, 10(1973), pp.141-147, p145.
9
Cf. L. Linsky, Referring, London-New York, 1967, p.18.; T. Parsons, "Are There Nonexistent Objects?", American Philosophical Quarterly,
19(1982), pp.365-371, p.366.

51
the bearer of the name dies, not that the reference dies. And it would be nonsensical to say that,
for if the name ceased to have reference, it would make nonsense to say `Mr. N.N. is dead'."10
But we can quite sensibly say that Mr. N.N., or Bucephalus is dead. So in (1) the name
"Bucephalus" refers to Bucephalus, even if Bucephalus, since perished, i.e. ceased to exist, does
not exist. Therefore it is true that the reference of "Bucephalus", i.e. Bucephalus, does not exist.
2. "Bucephalus existed, then died, i.e. ceased to exist, and now does not exist." This is a
perfect example of the use of existence as a contingent predicate of things. So any argument
trying to demonstrate the slogan "Existence is not a predicate" is simply futile. The objection, as
it proceeds from the false assumption that "a statement implies the existence of what it is about",
proves nothing.
3. Again, since the example above beyond being an example of contingent existence-
predication makes also perfect sense, therefore, any argument trying to prove in genere the
senselessness of existence-predications is doomed to futility. The causa apparentiae in the
objection lies in our liability to take "Something is a pink iguana" to mean the same as "Pink
iguanas exist". Now, it is true that in most cases a statement of the form "Something is an F" is
equivalent to "F's exist", or "an F exists". But this by no means implies that these two forms of
statement are synonymous. What is more, in some cases "Something is an F" and "an F exists"
are not even equivalent to each other. For example, "Something is perished" is true, but "A
perished thing exists" is false.11
4. Since to be a value of a bound variable is to be an element of the universe of discourse, and
what can be referred to in a language is an element of the universe of discourse, therefore, the
slogan "to be (or to exist) is to be the value of a bound variable" reduces to the claim that
everything that can be referred to in a language exists. But this cannot be true already for an
artificial modal language (as Nicholas Rescher pointed out, (x)(Ex) /Everything exists/ leads
necessarily to (x)(M(Ex) -> Ex) /Everything that can exist exists/, i.e. to the unpalatable: a posse
ad esse valet consequentia),12 and a fortiori it cannot be true for a natural language, as it was
shown in the preceding replies. But then, as Bucephalus, even though does not exist, can be the
value of a bound variable, therefore, if (3) is true, i.e. "x does not exist" is true for Bucephalus,
then "For some x, x does not exist", i.e. "Something does not exist" must also be true.
5. In reply to the fifth objection first we should ask: on what grounds does one say that "There
exists something that does not exist" is inconsistent? For if you analyse this sentence as "For
some x, x does not exist" (i.e. "Something does not exist"), then, provided that "exist" is not a
necessarily universal predicate, as it was argued for above, it is surely not inconsistent.
It would, however, be inconsistent if you analysed it on the analogy of, say, "There runs
something that does not run", i.e., if you treated "exist" as a predicate in both of its occurrences
in "There exists something that does not exist". For then this sentence would be equivalent to
"Something exists and does not exist". But in this case it cannot be thought to be equivalent to
"Something does not exist". So, if "There exists something that does not exist" is analysed as

10
Wittgenstein, Philosophical Investigations, Oxford, 1963, I.40., as quoted by P.T. Geach, "Form and Existence", in: God and the Soul,
London, 1969,p.58.
11
See again E.J. Ashworth: "Existential Assumptions in Late Medieval Logic", American Philosophical Quarterly, 10(1973), pp.141-147,
p.143. and p.145.
12
N. Rescher, "On the Logic of Existence and Denotation", The Philosophical Review, 68(1959), pp.157-180, p.161.

52
equivalent to a contradiction, then it is not equivalent to "Something does not exist", and, on the
other hand, if it is analysed as equivalent to "Something does not exist", then "There exists
something that does not exist" is not equivalent to a contradiction.
Now, whether or not these objections or replies are convincing, in my view, depends on certain
rather vague and conflicting intuitions. In the following two sections, therefore, I shall describe
the syntax and semantics of a formal language, AMPL, which, I hope, will be of some help in
clarifying these intuitions.

4. AMPL SYNTAX

The language AMPL is defined as follows:


AMPL:=<C,P,V,T,F>,
where C:={-,&,=,E,E,I,.,,(,)}, P is a set of parameters, V is a set of proper variables,13 as
opposed to the set of restricted variables, Vres, a subset of T, the set of terms, and F is the set of
formulae of AMPL. P detailed: P:=Pind U Ppred, where Pind is the set individual parameters,
while Ppred is the set of predicate parameters of AMPL. Note: Ppred  C:={E,=}. (That is, E,
the existence-predicate of AMPL, is a distinguished predicate in the same way as identity.) I is
the descriptor, and  is the ampliator of AMPL.
The set of terms and formulas of AMPL are defined by the following simultaneous recursive
definition:
(1) If aPind, then aT
(2) If xV, then xVar, where Var:=V U Vres
(3) If vVar and AF, then v.AVres
(Note that this rule is applicable recursively, so it permits even nested occurrences of
restricted variables, i.e., occurrences, in which a restricted variable appears as the operator
variable of another restricted variable. For the significance of this option see Appendix.)
(4) If t1,...,tnT and PnPpred, then Pn(t1)...(tn)F, (-Pn)(t1)...(tn)F, E(ti)F and (ti = tj)F
(5) If A,BF and vVar, then ~(A)F, (A)F, (Ev)(A)F (Iv)(A)F and (A & B)F
For convenience's sake we may apply the following abbreviations:
(Abbr1) The matrix of a restricted variable may be omitted in all of its occurrences
following its first occurrence in a formula, provided different restricted variables have different
operator variables.
(Abbr2) Further connectives are to be regarded as abbreviations of the usual definients.
(Abbr3) When from their omission no confusion arises parentheses may be omitted.

5. AMPL SEMANTICS

The definition of a model for AMPL is the following:


M:=<W,S,as,D,R,0>,

13
Cf. M.M. Prullage, "A Theory of Restricted Variables without Existential Assumptions", Notre Dame Journal of Formal Logic, 17(1976),
pp.589-612. See also Essay II.

53
where W and S are nonempty sets, as is a distinguished element of S, D is a function from S to
the set of all subsets of W, i.e., if sS, then D(s) is a part of W, R is a function assigning
semantic values to the parameters of AMPL and 0 is the zero-entity, the semantic value of empty
terms, which falls outside W, i.e., it is not the case that 0W.
Intuitively, W is the universe of discourse of M, S is a set of situations, or states of affairs, as
is the actual situation and D is the domain assignment of situations, so that D(s) is the
domain of the situation s.
R is defined by the following clauses:
(R1) If aPind, then R(a)W
(R2) If PnPpred, then R(Pn)(s) is a part of Wn
(R3) R(E)(s)=W
Let us define further the extension of a predicate Pn in the situation s, Exts(Pn), in the
following manner: Exts(Pn):=R(Pn)(s)  D(s)n. (Whence, Exts(E)=D(s).)
An assignment in the situation s, fs (a function from T U F to W U {1,0})14, is defined in the
following manner:
(fs1) If xV, then fs(x)W
(fs2) If aPind, then fs(a)=R(a)
(fs3) If v.AVres, then
/i/ fs(v.A)=fs(v), if fs(A)=1
/ii/ fs(v.A)=0 otherwise
(fs4) fs(P (t1)..(tn))=1 iff <fs(t1),..,fs(tn)>Exts(Pn)
n

(fs5) fs((-Pn)(t1)...(tn))=1 iff <fs(t1),..,fs(tn)>D(s)-Exts(Pn)


(fs6) fs((t1=t2))=1 iff fs(t1)=fs(t2)W
(fs7) fs(~(A))=1 iff fs(A)=0
(fs8) fs((A & B))=1 iff fs(A)=fs(B)=1
(fs9) fs((Ev)(A))=1 iff for some uRgfs(v), fs[v:u](A)=1 where Rgfs(v), the range of v in
respect of fs, is {uW: for some fs' in the same s differing from fs at most in the value assigned to
v (and its /nested/ operator variable/s/, /if it has any/), fs'(v)=u}, if this set is not empty, otherwise
Rgfs(v)={0}; and fs[v:u] is the same as fs except that it assigns u to v (i.e. fs[v:u](w)=fs(w), if v is
not w, otherwise fs[v:u](w)=u).
(fs10) fs((Iv)(A))=1 iff there is exactly one uRgfs(v), such that fs[v:u](A)=1
(fs11) fs((A))=1 iff for some s'S, fs'(A)=1.
(fs12) fs(A)=1 iff not: fs(A)=0.
Truth in a model M is defined as follows:
(T) |A|M=T iff for some fas, fas(A)=1.
(That is, A is true in M, iff for some assignment in the actual situation, A is true according to, or
is satisfied by, that assignment.)

14
Again, in the same way as in Essay II., 0 serves as the semantic value of empty terms as well as the semantic value of false formulas. This is a
mere technical device with no philosophical significance attached to it.

54
As usual, a formula is satisfiable if there is a model in which it is true, valid, if its negation is
not satisfiable, and an inference is valid, if the conjunction of the premises with the negation of
the conclusion is not satisfiable.

6. AMPL AND THE THEORY OF AMPLIATION

, the only intensional operator of AMPL, is called - barbarously enough - the ampliator of
AMPL. As to its intuitive meaning, it has no single natural language equivalent, instead, in
several applications it may serve for several intensional operators.
For example, in modal contexts it may serve for the familiar possibility operator (M), while in
tensed contexts it may serve for the past or future tense operators (P or F). Of course,
accordingly, the set of situations S is to be interpreted as the set of possible, past or future
situations.
As an indication of the actual intended interpretation I shall use indices to . (Even if these
stricto sensu do not belong to the vocabulary of AMPL.) For example, for the sentence "A white
thing can be not white" I shall write:
(Ex.Wx)M~((Wx.)),
while for the sentence "A white thing will be not white" I shall write:
(Ex.Wx)P~((Wx.))
Of course, in "mixed contexts", where e.g. both tense and modal operators are involved (as in
an argument concerning future contingents) the single  would not suffice. In such a case further
operators should be introduced, and, accordingly, the set of situations should be further specified.
(For example, in the case of tensed modal contexts the elements of S should be ordered pairs of
time points and possible worlds in the vein of Montague.) But for our present purposes the single
 with its multifarious intuitive interpretation will suffice. For with the aid of  in AMPL we can
represent that common feature of intensional sentences that an adequate statement of their truth
conditions involves reference to situations beyond the actual one, and that, accordingly, it
contains (at least implicit) reference to individuals beyond the actual ones.
Now, according to the medieval analysis of sentences of this kind, the common feature of
these sentences is that the range of reference of their subject terms is extended, ampliated
beyond the sphere of actual entities to several, past, future or possible non-actual ones. For
example, while in the sentence "A man runs" (Homo currit) the subject term "man" refers to
only what is a man, in the sentence "A man ran" (Homo cucurrit) it refers to what is or was a
man, so that according to this analysis the sentence "A man ran" could be true even if actually
there were no men.15 Now this analysis can be represented in AMPL as follows:
(Ex.Mx v PMx)P(Rx.),
that is, a thing that is a man or was a man ran (Quod est homo vel fuit homo cucurrit). This
may be called the common analysis of a sentence containing an ampliative verb (verbum
habens vim ampliandi).16 But, as is well known, the medievals also made a distinction between
15
Cf. e.g. Buridanus, Tractatus de Suppositionibus, ed. Maria Elena Reina, Rivista Critica di Storia della Filosofia, 1957, pp.175-208, and
pp.323-352, c.6., p.349.
16
For a "catalogue" of these see Buridan, ibid.

55
de dicto and de re sentences, or de dicto and de re readings of sentences containing such
ampliative verbs.
For example, let us take the following sentences.
(1) A white thing can be black (Album potest esse nigrum)
(2) A thing that is white can be black (Quod est album potest esse nigrum)
(3) It is possible that a white thing is black (Possibile est album esse nigrum)17
(2) is a de re modal sentence, and (3) is a de dicto modal sentence. (2) can also be regarded as
the de re, while (3) as the de dicto reading of (1); or, according to another terminology, (2) can
be regarded as expressing the content of (1) in sensu diviso, while (3) can be regarded as doing
the same in sensu composito. Of course, if we take (2) and (3) as expressing different readings
or senses of (1), then we must hold that (1) is ambiguous, and, consequently, we cannot hold that
there is a single common analysis of (1). If, on the other hand, we accept that (1) has a single
common analysis, then we cannot take (2) and (3) as expressing different readings of (1),
instead, we must regard them as being different sentences in their own right.18
Which way we choose, I think, is a matter of convention. What concerns us more in the
present context, however, is that the differences of (1), (2) and (3) can be brought out very clearly
in AMPL. For the sake of simplicity, interpreting "black" as "not-white", we can write:
(1') (Ex.Wx v M(Wx))(M((-W)x.))
(2') (Ex.Wx)(M((-W)x.))
(3') M((Ex.Wx)((-W)x.))
Now, as can easily be seen, while (2') is satisfiable, (3') is not. For (2') is true in a model M, in
which there is an assignment in the actual situation fas, such that fas((1'))=1 in M. But this is so, if
there is a uRgfas(x.Wx), i.e. Extas(W),19 such that for some s, uD(s)-Exts(W) too. Now such a
model is in which there is a uW and sS, such that uR(W)(as)D(as)(D(s)-R(W)(s)D(s)).
But since there is such a model, (2') is satisfiable.
On the other hand, (3') is true in a model M if there is a situation s such that fs((3'))=1 in M.
But this is so if and only if there is a uRgfs(x.Wx), i.e. Exts(W), such that uD(s)-Exts(W) too.
But this means that (3') can be true only in a model in which there is a uW and sS, such that
uR(W)(s)D(s)(D(s)-R(W)(s)D(s)), which (the intersection of a set with its complement
being empty) is impossible. So there is no such a model, and therefore (3') is not satisfiable.
Now, since (2') is satisfiable, and (3') is not, therefore "(2') & ~(3')" is satisfiable, i.e., an
argument from (2') to (3') is fallacious. In particular, such an argument from a proposition in

17
Cf. e.g. Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, pp.68-69; Albert of Saxony: Perutilis Logica, Georg Olms
Verlag, Hildesheim-New York, 1974, Tr.2.c.10.ff.15-16, and Tr.5.c.4.f40.
18
As from the texts referred to above is quite clear, this was the attitude taken by Buridan and Albert of Saxony. In his Tractatus de Fallaciis
Buridan explicitly states: "Ista enim propositio `Laborans sanabatur' non est distinguenda: habet enim unicum sensum qui explanandus est per
propositionem unam de disiuncto subiecto sc. istam: `Qui est vel fuit laborans sanabatur'." S. Ebbesen, "The Summulae, Tractatus VII De
Fallaciis", in: The Logic of John Buridan, ed. J. Pinborg, Copenhagen: Museum Tusculanum, 1976, p.155. On the other hand, Ockham writes
thus: "Et ideo quaelibet talis propositio est distinguenda eo quod talis terminus potest supponere pro his quae sunt vel pro his quae fuerunt." W.
Ockham: Summa Logicae, St. Bonaventure N.Y., 1974, P.I.c.72, p.216. Again, for a "catalogue" of the technical terms for such and closely
related distinctions see St. Thomas Aquinas, ST1.q.14.a.13.ad3-um and ScG lb.1.c.67.n.10. in: Opera Omnia, Stuttgard-Bad Cannstadt, 1980
Lemma: Rgfs(x.Wx)=Exts(W), provided Exts(W) is not empty, otherwise Rgfs(x.Wx)={0}. Proof: for any uW, uRgfs(x.Wx), iff for some
19

fs', fs'(x.Wx)=u, that is, iff fs'(x)=u and fs'(Wx)=1. But this is so iff uExts(W). On the other hand, if Exts(W) is empty, then there is no fs' such
that fs'(Wx)=1, so for any fs', fs'(x.Wx)=0, whence Rgfs(x.Wx)={0}.

56
sensu diviso to a proposition in sensu composito would be a case of a fallacy termed by the
medievals as fallacia compositionis et divisionis.20
A notable example of this fallacy would be an argument
from
(4) It is possible that the number of the planets is less than seven
to
(5) That number which is the number of the planets is possibly less than seven
That is, from
(4') M((Ix.Nx)(L(x.)(s)))
to
(5') (Ix.Nx)M((L(x.)(s)))
(Vocabulary: `N': "is the number of the planets"; `L': "is less than"; `s': seven)
Clearly, (4') does not imply (5'), and so they cannot be regarded equivalent. So even if the
argument from (5) with
(6) The number of the planets is nine
to
(7) Nine is possibly less than seven
is valid, still, as (5) is false, (7) may also be false. On the other hand, the argument from the true
(4) with (6) to (7) is not valid. So the falsity of (7) is in no conflict with the truth of (4).
Therefore, if we carefully distinguish between the true de dicto (or opaque, (4)) and the false de
re (or transparent (5)) readings of "The number of the planets is possibly less than seven", then,
contrary to Quine's claim,21 no inconvenience follows from the "essentialism" involved in (5), or
in its existential generalization, the false
(8) A number which is the number of the planets is possibly less than seven
(8') (Ex.Nx)M((L(x.)(s)))
Of course, (8') follows from (5'), but does not follow from (4'), neither is it implied by
(9') M((Ex.Nx)(L(x.)(s)))
that is,
(9) It is possible that a number which is the number of the planets is less than seven.
According to Quine, another inconvenience which follows from "Aristotelian essentialism" is
that, at least on Quine's interpretation of this doctrine, the same attribute of an object, depending
on which description of the object is considered, must be regarded as both essential and not
essential to the object.22
For example, Socrates's wife is necessarily Socrates's wife. So, being Socrates's wife is an
essential attribute of Socrates's wife. But Xanthippe is not necessarily Socrates's wife (she might
have married another man), so, being Socrates's wife is not an essential attribute of Xanthippe.

20
See e.g. Aristotle, On Sophistical Refutations, Cambridge, Mass., London, 1955, c.4. & 20, 166a23-28, 177a33-177b35; Peter of Spain,
Tractatus, (ed. L.M. de Rijk), Assen, 1972, pp.122-124; W. Ockham: Summa Logicae, St. Bonaventure N.Y., 1974, pp.786-790.
21
W.V.O. Quine, "Reference and Modality", in: From a Logical Point of View, Cambridge, Mass., 1961.
22
V.W.O. Quine, Word and Object, Cambridge, Mass., 1960, p.199. For an informal criticism of Quine's Argument see D.B. Rasmussen,
"Quine and Aristotelian Essentialism", The New Scholasticism, 58(1984), pp.316-335.

57
But Xanthippe is Socrates's wife. So, being Socrates's wife is both essential and not essential to
the same person, i.e., Socrates's wife, namely Xanthippe.23
Now this line of reasoning involves at once two fallacies. One is the already familiar fallacia
compositionis et divisionis, and the other is an ignoratio elenchi. The ignoratio consists in
ignoring the proper definition of what it is to be an essential attribute of something. The fallacia
compositionis et divisionis (hand in hand with the abovementioned ignoratio) is involved in the
transition from "Socrates's wife is necessarily Socrates's wife to "Being Socrates's wife is an
essential attribute of Socrates's wife".
Now let us introduce a necessity operator into AMPL in the following manner.
fs(N(A))=fs(-M(~(A)))
As can be seen, the following formula is not valid.
(10') N((Ix.Wx)(Wx.))
(10) It is necessary that Socrates's wife is Socrates's wife
(Vocabulary: `W': "is Socrates's wife")
For in a model M in which there is a situation s in which Exts(W) is empty (i.e. in which
Socrates has no wife) (10') is false. The following, de dicto, formula, however, is valid.
(11') N((Ix.Wx)(Ex.) -> (Ix.Wx)(Wx.))
(11) It is necessary that if Socrates's wife exists (i.e., if Socrates has a wife),
then Socrates's wife is Socrates's wife
But the corresponding de re formula is not valid.
(12') (Ix.Wx)N((Ex. -> Wx.))
(12) Of that person who is Socrates's wife it is necessarily true that if she exists, then she is
Socrates's wife.
For in a model M in which there is exactly one dExtas(W), and there is an s such that sD(s)
but d is not an element of Exts(W), i.e., in which Xanthippe exists, but is not Socrates's wife,
(12') is false.
Now a predicate of an object is essential to this object, if and only if it is necessary that if it
exists, then the predicate is true of the object.24
This definition can be introduced into AMPL in the following way:
fs(ESS(Pa))=fs(N(Ea -> Pa))
As can easily be seen, the argument from (12') to
(13') (Ix.Wx)(ESS(Wx.))
(13) Of the person who is Socrates's wife it is essentially true that she is Socrates's wife
is valid; still, as (12) is false, (13) may be, and actually is, false too.
But the argument from (11') to (13') is not valid. So the transition from the true (11) to the
false (13) is clearly fallacious. Therefore the argument proves nothing against Aristotelian
essentialism.

23
For the sake of formal simplicity I have changed Quine's original example, the cycling mathematician. See note above.
24
It is interesting to see that Rasmussen, while criticizing Porphyry for being unfaithful to Aristotle's original doctrine, embraces the same
definition (p.323.), which, however, is a quite straightforward consequence of Porphyry's definition of accident. As it was quoted by the
medievals (in Boethius's translation): "accidens est quod adest et abest praeter subiecti corruptionem". For a stronger definition, of which this
Porphyryan definition is a corollary, see Essay VI.

58
Now in the foregoing formulations we used both the existence predicate, E, and the existential
quantifier, E, of AMPL. Consider the following metatheorem:
(EQ) If vV and AF, then for every model M,
|(Ev)(A)|M=|E(v.A)|M iff EXT(v,M,as)(A) is empty or EXT(v,M,as)(A)D(as)is not empty,
where EXT(v,M,as)(A), the extension of A in respect of v in M in the actual situation, is
defined as follows:
EXT(v,M,as)(A):={uW: for some fas, fas[v:u](A)=1}.
Proof
To simplify the proof first I prove two lemmas.
Lemma 1.
|(Ev)(A)|M=T iff EXT(v,M,as)(A) is not empty.
Proof
Suppose EXT(v,M,as)(A) is not empty. Then for some fas, and for some uW, i.e., for some
uRgfas(v), - since for any vV and for any fs, Rgfs(v)=W -, fas[v:u](A)=1, and so, |(Ev)(A)|M=T;
and conversely. Q.e.d.
Lemma 2.
|E(v.A)|M=T iff EXT(v,M,as)(A)D(as) is not empty.
Proof
Suppose EXT(v,M,as)(A)D(as) is not empty. Then for some fas, and for some
uRgfas(v)D(as), fas[v:u](A)=1, and so, fas(v.A)=u; whence for some fas,
fas(v.A)D(as)=R(E)(as), that is, |E(v.A)|M=T; and conversely. Q.e.d.
Now, suppose that (1) EXT(v,M,as)(A) is not empty and (2) EXT(v,M,as)(A)D(as) is empty.
Then, by (1) and Lemma 1 |(Ev)(A)|M=T, while from (2) and Lemma 2 it follows that it is not the
case that |E(v.A)|M=T.
On the other hand, suppose that (1) EXT(v,M,as)(A) is empty, or (2) EXT(v,M,as)(A)D(as)
is not empty.
From (1) by Lemma 1 it follows that it is not the case that |(Ev)(A)|M=T. From (1) it also
follows that EXT(v,M,as)(A)D(as) is empty, whence, by Lemma 2, it follows that |E(v.A)|M=T
is not the case either. But from (2) and Lemma 2 it follows that |E(v.A)|M=T, while from (2) it
also follows that EXT(v,M,as)(A) is not empty, whence also |(Ev)(A)|M=T. And this completes
the proof.
The significance of this metatheorem is that it shows the close connection, but without
blurring the distinction between an existential statement, and an existential, or, perhaps better,
particular quantification.
What metatheorem (EQ) says is that if the actual extension of the open sentence,
EXT(v,M,as)(A), involved in the quantification and in the restricted variable of the existential
statement is not ampliated to non-actual individuals, then these two forms of statement are
equivalent, but if it is ampliated to non-actual individuals and does not contain actual ones, then
they are not equivalent. For example, in virtue of metatheorem (EQ) the following formulae are
equivalent:
(14') (Ex)(Cx)
(15') E(x.Cx)

59
And this is how it should be. Clearly, the sentence
(14) Something is a centaur
is equivalent to
(15) A centaur exists
This is why we are entitled to use, in our logical and mathematical practice, the types of
statements represented by (14') and (15') interchangeably.
But let us take the following two sentences:
(16) Something is destroyed
(17) A thing that is destroyed exists25
These are clearly not equivalent. For if we define "is destroyed" as "existed and does not exist"
then we get the true
(18) Something existed and does not exist
(18') (Ex)(P(Ex) & ~(Ex))
and the false, indeed, inconsistent
(19) A thing that existed and does not exist exists
(19') E(x.P(Ex) & ~(Ex))
And this is why we cannot use (16) and (17) interchangeably.

7. OBJECTIONS AND REPLIES RECONSIDERED

After this rather prolonged overview of how AMPL is supposed to work, I think it is time to
see how it resolves the informal objections listed at the beginning of this paper. The main
argument can be formalized as follows:
(1') Db
(2') (Vx.Dx)(~(Ex.))
(3') ~(Eb)
(4') (Ex)(~(Ex))
The argument is valid, and, if `Dx' (x is dead) is regarded as a mere abbreviation for `P(Ex) &
~(Ex)' (x existed and does not exist), then (2') is valid, while in a model in which R(b) is not an
element of D(as), but is an element of some ("past") D(s), (1') is true. So the argument in such a
model proves the conclusion.
The first objection, as the reply rightly states, is a blatant non sequitur. From `~(Eb)' (b does
not exist) it does not follow that `Eb' (b exists). In general, `Ea' does not follow from `~(Pa)'. For
the latter is true, if and only if R(a) falls outside Extas(P), but thereby it need not be within D(as),
i.e., within Extas(E).
The second objection, again, assuming the false principle: `~(Pa)'  `Ea', proves nothing. An
analogous true (valid) principle would be the following:
'(-P)a'  `Ea', or even `(-P)a v Pa'  `Ea',
which is the AMPL-expression of the existence-presupposition of atomic statements, with
special regard to the distinction between external and internal negation.26 But this principle says

25
In this sentence the word "thing" is not to be thought of as a common term, but as a natural language equivalent of a proper variable. See
Appendix.

60
nothing about the external negation involved in genuine individual existence-denials, such as
"Bucephalus, since perished, does not exist".
To the third objection the reply should be evident from the results of the preceding section.
The equivalence of "Something is a pink iguana" with "Pink iguanas exist" is not sufficient to
show that existence is a second-level predicate.
In reply to the fourth objection I can simply point to the fact that the inference from (3') to (4')
is valid in AMPL. Of course, the construction of AMPL in itself is such that it involves the denial
of Quine's slogan. For this denial, however, I can find no better justification than the simple fact
that we do quantify over individuals that do not exist. Just consider "Some of the things that
existed before the Second World War do still exist, but some do not".
Lastly, the fifth objection, as the reply states, rests on a confusion between `(Ex)~((Ex))'
(Something does not exist) and `(Ex)(E(x.~(Ex)))' (Something that does not exist exists, or, if
you like, there exists something that does not exist), the former of which is satisfiable, while the
latter is not, and so the latter is equivalent to `(Ex)(Ex & ~(Ex))' (Something exists and does not
exist). Notice that `E' occurs always as a quantifier, or determiner, while `E' always as a
predicate. This is why the syntax of AMPL does not permit so easily the shift from "Something is
..." to "There exists ...", as the syntax of English, especially as conceived by analytic philosophers
nourished on Frege's, Russell's and Quine's ideas, does.

8. CONCLUDING REMARKS

In the preceding section I have investigated the interconnection between the notions of
existence and quantification with the help of the interpreted language AMPL. The basic result of
these investigations, formulated in metatheorem (EQ), is that although in purely extensional
contexts, where no reference is made to non-actual individuals, an existential statement is always
interchangeable salva veritate with a particular quantification, the same does not necessarily
hold for non-extensional contexts. I think, it is for this reason that in mathematical logic, where
there is no need of, indeed, no place for, a distinction between actual and non-actual elements of
the universe of discourse, the notion of existence could successfully be analyzed in terms of
particular, or existential quantification. (Such and such a number, set, function etc. exists, if and
only if some number, set, function etc. is such and such.) But this analysis, backed up by the
Kantian tradition ("existence is not a real predicate") on the one hand,27 and by the amazing
successes of mathematical logic on the other, led to an overall identification of the two notions.
The situation, however, unavoidably led to certain "anomalies" in the logical analysis of non-
extensional contexts.28
But as soon as we realize the proper connection between the two distinct notions, most of our
most recalcitrant puzzles get solved. Not all, however.

26
Cf. L. Linsky, Referring, London-New York, 1967, p.90. And Essay I. and II. on the topic of internal negation.
27
For interesting discussion of Kant's influence on Frege's ideas on existence see L. Haaparanta, Frege's Doctrine of Being, Acta Philosophica
Fennica, vol.39, Helsinki, 1985, pp.128-144; and L. Haaparanta, "On Frege's Concept of Being", in: S. Knuuttila-J. Hintikka (eds.), The Logic of
Being, Dordrecht, Holland, 1986.
28
For an imposing list of these "anomalies" and an abundance of "epicycles", so to speak, "to save the phenomena", see C.J.F. Williams, What Is
Existence?, New York, 1981.

61
In the preceding pages I have taken my examples from modal and tensed contexts. I have also
indicated how AMPL could be extended to cover also mixed contexts.
There are, however, several further intensional contexts, e.g. those created by intentional
verbs.29 One of the main characteristic feature of these verbs is that they may be true of objects
that do not exist.
Now let us introduce into the language AMPL a distinguished intentional predicate, P, in the
following manner.
(i) PPpred  C
(ii) For any model M, for any s, Exts(P)=R(P)(s)
(Of course, accordingly, the definition of Exts(Pn) above is to be understood with the exception
of P.)
An ampliated assignment in the situation s, fs, is defined in the following way:
(1) fs(a)=fs(a)
(2) fs(v)=fs(v)
(3) fs(v.A)=fs(v.(A))
Let (fs4) concern only Ppred-{P}, and for P let us formulate the following clause:
fs(P(t))=1 iff fs(t)Exts(P)
Now with these clauses at hand we can see that from
(20) I seek a unicorn
(20') P(x.Ux)
it does not follow that
(21) Something that is a unicorn is such that I seek it
(21') (Ex.Ux)(Px.)
Neither does it follow that
(22) Something is a unicorn
(22') (Ex)(Ux)
nor that
(23) A unicorn exists
(23') E(x.Ux)
These do not follow, because (20) may be true in a model M in which Extas(U) is empty, i.e., in
which unicorns do not exist, but in which there is a situation s, such that Exts(U) is not empty and
there is a dExts(U), such that dExtas(P)=R(P)(as). And this is possible because Extas(P)
may extend beyond D(as), i.e., because such a predicate can be true of anything that can be
referred to, with no regard to whether it exists or not. And this is so because such predicates
signify certain acts of the mind, which, by means of its abstract concepts, can think of any kind of
things indifferently whether they exist or not.
As Buridan's pupil, Albert of Saxony put it:

29
There are even further (and broader) intensional contexts, like the context of a myth, legend, or any fiction. (Or, for that matter, the context:
"in the legend".) These all may be represented, through some proper modification of AMPL, by a situation s and a corresponding special 
which "takes over" the discourse from as into that s. In general, here I take sides with L. Linsky, who writes: "I would not, in my ontology, divide
objects into ideal objects which subsist and real ones which exist, but into objects which are, e.g., characters in fiction, legendary figures, comic
strip characters, make-believe figures, as well as abstractions, mathematical objects, concepts, etc." Linsky, op. cit., p.20.

62
"All verbs, even in the present tense, which of their very nature can concern future, past and
possible things as well as present ones (habent naturam transeundi rem ita futuram vel
praeteritam vel possibilem sicut et praesentem) ampliate their terms to all times, future, past and
present, like these: think, know, mean and the like. And what accounts for this is that a thing can
be thought of without any difference of time (sine differentia aliqua temporis), sc. abstracted
from any place and time. And so, when a thing is thought of in this way, then a thing which was,
or will be, or can be may be thought of as well as a thing which is. Therefore, if I have the
common concept from which we take (a quo sumitur) this name `man', then I can think
indifferently of all men, past, present and future. And this is why these verbs can concern past or
future things as well as present ones."30
There remain, however, problems in connection with intentional verbs which cannot be
handled even with this extended apparatus of AMPL. Such is e.g. the problem of non-
interchangeability of identicals in the context of such verbs.
For example, the following argument is not valid.
(i) You do not know the man with his face covered
(ii) The man with his face covered is your father
Therefore
(iii) you do not know your father.31
Now for a solution of this sophism, we need some more sophisticated devices, namely the
devices of the medieval theory of appellation,32
But for a reconstruction of this theory, I think, first we need a reconstruction of a more basic
medieval theory, namely the inherence theory of predication.33

30
Albert of Saxony: Perutilis Logica, Georg Olms Verlag, Hildesheim-New York, 1974, Tr.2. c.10. 8a regula.
31
Cf. Aristotle, On Sophistical Refutations, Cambridge, Mass., London, 1955.; ch.24. 179a27-179b33. For this version of the sophism see
Lucian, "Philosophies for Sale", ch.22-23., in: The Works of Lucian, London, 1953.
32
Cf. Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, pp.59-83; esp. pp.72-74.
33
The reconstruction of the inherence theory is one of the starting points of the remaining essays in this volume. It is with the help of this theory
that the problem of the uninterchangeability of identicals in intentional contexts will be dealt with (in Essay V.), however, no special mention
will be made of Buridan's appellation theory. Nevertheless, I think that although Buridan was an "identitist" (as opposed to an "inherentist"), as
his theory of concepts and of appellation can be understood better from an inherentist starting point (see n.29. of Essay VII.), the reconstruction
of the inherence theory seems to be a prerequisite of the reconstruction of Buridan's treatment of intentional verbs. (Cf. also the references of n.2.
above.) This, however, already lies outside the scope of the present volume.

63
APPENDIX
In note 25. above I asked the reader to understand the word "thing" in "A thing that is
destroyed exists" not as a common term, but as the natural language equivalent of a proper
variable (as we indeed use it sometimes, when we recognize "things" that do not exist). Let me
explain why.
If we understood "thing" as a common term, i.e., a restricted variable in this sentence, then,
one might think, the AMPL equivalent of (17) should be something like this:
(i) E(x.Tx & Dx)
But as a matter of fact this formulation is mistaken. For just like a fictitious centaur is not
something that is a centaur and is fictitious, so a thing that is destroyed is not something that is a
thing ("ens et res convertuntur") and is destroyed. Formally, the problem is this.
If `T'Ppred, then Exts(T) is a part of D(s), for any s. So, since given the definition of `D'
('P(Ex) & ~(Ex)'), EXTas(P(Ex) & ~(Ex)) is empty, EXTas(P(Ex) & ~(Ex))  Exts(T) is empty
too. But this means that for any fs, fs(x.Tx & P(Ex) & ~(Ex))=fs(x.Tx & Dx)=0. And so, even
(ii) (Ex.Tx & Dx)(P(Ex.)),
i.e., on the intended reading
(iii) A thing that is destroyed existed
could not possibly be true.
So we cannot take (ii) as being a correct formalization of (iii), or, conversely, we cannot take
(iii) as being a correct reading of (ii). Indeed, a correct reading of (i) would be rather this:
(iv) Something that is a thing (a being) and
is destroyed existed
And this, since nothing can be both an actual being and destroyed, surely cannot be true. But,
then, what would be the correct formalization of (iii)?
For answering this question we should recall that the syntax of AMPL provides room also for
"nested" occurrences of restricted variables, i.e., occurrences in which a restricted variable
appears as the operator variable of another restricted variable. With this possibility at hand we
can offer as an answer to the above question the following formula:
(v) (Ex.Tx.Dx.)P(Ex.),
whence, substituting the definiens of `D', we get
(vi) (Ex.Tx.P(Ex.) & ~(Ex.)(P(Ex.)),
that is,
(vii) A thing that existed and does not exist existed
And this surely is true. (The formal trick, as one can see, is that, due to the  occurring in the
matrix of this variable, the variable is able to pick up its values from the extension of `T' in an s
outside as, the actual situation, where the extension of the matrix as a whole, due to the conjunct
`~(Ex.)', is empty.)
Pari ratione, the sentence:
(viii) A fictitious centaur is fictitious,
that is,
(ix) P(x.Cx.Px.)
(Vocabulary: `P': "fictitious"; `C': "centaur")

64
is true.
But the following sentence:
(x) A fictitious centaur is a centaur,
that is,
(xi) C(x.Cx.Px.),
provided Extas(C) is empty, i.e., that there are no centaurs, is false. So even if a fictitious
centaur is not a centaur, a destroyed thing is not a thing (in the strict sense), a dead man is not a
man and a merely potential being is not a being, still, these are fictitious, destroyed, dead, and
potential, respectively.34

34
Cf. Peter of Spain, Tractatus, (ed. L.M. de Rijk), Assen, 1972, pp.158-159, and E.L. Keenan and L.M. Faltz, "A New Approach to
Quantification in Natural Language", in: C. Rohrer (ed.), Time, Tense and Quantifiers, Tübingen, 1980.

65
UNDERSTANDING MATTERS FROM A LOGICAL ANGLE1

- Logical Aspects of Understanding -

As the recently almost unlimited multiplication of several kinds of logical systems tends to
turn the term "logic" somewhat ambiguous, perhaps the safest way to speak about the logical
aspects of anything is by presenting a logical system which one considers the most suitable in
treating the subject. At any rate, this is what I am going to do in this essay. But first, some
preliminary remarks.

1. UNUNDERSTANDING

Understanding, it seems, may best be understood ex opposito, namely from cases of


ununderstanding. Now, even if we restrict our attention to linguistic understanding (setting aside
e.g. metacommunication, on the one hand, or understanding a thing's nature, or one's feelings or
motives, on the other) what these cases reveal above all is that understanding is highly precarious
business, depending on many factors for its success.
The most trivial case of linguistic ununderstanding is when the listener simply does not know
the language of the speaker. From this case of absolute ununderstanding there is a more or less
continuous transition, through several degrees of understanding achieved between occasional
users of the same language (depending on the linguistic competence of both parties), up to the
degree of understanding achieved between native speakers of the same language.
But, as we all know well, even mastery of the same idiom by both parties is all too often
insufficient to avoid misunderstandings. Languages abound with ambiguities, which leave open
the possibility of different interpretations of the same expression by different users of the
language. These differences of interpretation may be the source of either occasional, easily
corrigible, or systematic, sometimes incorrigible misunderstandings.
Now what seems to be the common source of all these kinds of ununderstanding is either the
lack (in the case of absolute ununderstanding), or the difference (in cases of misunderstanding) of
interpretation of the same linguistic sign. This difference of interpretation may, again, stem from
different sources.
First, it may be due to some objective ambiguity, inherent in common language usage: the
same expression may have different (socially) objective meanings.
Second, it may be due to some sort of ignorance, or (semantic) incompetence of the interpreter
(who may be the speaker as well as the listener), who, in consequence, assigns a subjective
interpretation to the expression in question which is different from its (socially) objective
meaning.
Third, it also may be the case that the objective meaning of an expression is vague,
underdetermined, and so (perhaps, depending on context), users of the language are quite free to
assign to it different subjective interpretations without being ignorant or incompetent.

1
I hope Professor Quine and Professor Geach will pardon me for the trifling plagiarisms committed in the main title of this paper - if not for the
paper itself.

66
Now what all this boils down to seems to be that a logic which is to be able to give an account
of understanding (and, by the same token, of ununderstanding) should incorporate some notion of
objective, socially established meaning, on the one hand, and of subjective, individual
interpretations of intersubjective linguistic signs, on the other.
The task of the subsequent discussion is to draw the main outlines of such a logic.

2. FROM INTENSIONS TO SIGNIFICATIONS

In our days the most powerful tool for the exact study of meaning is model theoretic
semantics, in particular, its branch called intensional logic.
Nevertheless, by now it seems to be quite clear that the classical notion of intension does not
capture the intuitive notion of meaning. Intensions, conceived as functions from possible worlds,
or indices (ordered pairs of possible worlds and time-points) to extensions or factual values (cf.
Montague, 1973), are too "coarse-grained" in that they fail to distinguish between non-
synonymous, merely necessarily coextensive expressions. (cf. Lewis, 1976)
Recently several attempts have been made to produce semantics that would provide more
"fine-grained" intensions. However, as these attempts are either still under formation (see e.g.
Perry, 1986), or, at least, under discussion (see e.g. Anderson, 1987), let me bluntly proceed on
my own way and sketch my alternative approach to offer it for similar discussions.
The approach I am about to present has the following distinguishing features on account of
which, I think, it deserves to be called "alternative":
1. It is based on the so-called "inherence theory" of predication, (see e.g. Moody, 1957)
2. Accordingly, it incorporates the Aristotelian notion of truth, defined in terms of actual
existence. (cf. Matthen, 1983; Hintikka, 1986; Jacobi, 1986; Weidemann, 1986)
3. It provides an intensional logic without any reference to possible worlds.
4. It provides suitably "fine-grained" intensions and an adequate synonymy relation with no
reference to "unique, non-circular definitions". (cf. Bealer, 1982; Anderson, 1987)
5. It provides room for objective meanings as well as for subjective concepts - thereby being
able to solve the intentional paradoxes and to give an account of the abovementioned types of
ununderstanding. (cf. Kamp, 1981; Landman, 1986)
For the sake of simplicity, let me start with (the very primitive) language of first-order uniform
quantification theory, containing only the usual logical connectives, the two quantifiers,
variables, one-place predicates and formulas built up from these according to the usual
construction rules.
According to the usual semantics for this language, the semantic value of a predicate P is a
subset of the universe of discourse W of a model M, and an atomic formula of the form `Px' is
true according to (or is satisfied by) an assignment of values f, if and only if the element of W
assigned by f to `x' is an element of this subset of W. So far, so good.
Nevertheless, this is not the only possible way to specify the truth conditions of such a simple
predication. Intuitively, what the above formal description says is that the semantic function of
predicates is to denote sets, and the truth of a predication depends on membership in the set
denoted by the predicate. For example, on the above account the function of the predicate `is red`

67
is to denote the set of red things, and the truth of the predication `x is red` depends on whether
the thing assigned to `x`, say u, is a member of this set.
However, we may also say, and perhaps even more intuitively, that the function of this
predicate is to signify the property of redness, and the truth of this predication depends on
whether u has redness, i.e., whether u's redness exists. (Disregarding for the moment Quinean
and Geachean qualms about properties.)
So what the predicate of this predication, according to this description, signifies is u's
particular redness at time t of the predication (an "individualized form" as Prof. Geach would call
it. Cf. Geach, 1969) Let me call this particular redness the significate of the predicate "is red" in
u at time t, or, in symbols: Sgt(R)(u)(t). It is the actual existence of this significate that accounts
for the truth of this predication. But what about the case when this significate does not exist?
Well, then it either can exist or not. So, if this significate is not actual, then it is either
potential or impossible, and, accordingly, the predication is either possible or impossible at the
time of the predication. (Concerning reference to and quantification over non-actual individuals
see Essay III.)
But it may also be the case that the given predicate is simply not applicable to a given thing;
say, the predicate `is red` to the number 2. In this case one would rather be inclined to say that the
predicate has no significate in this thing at all, neither actual, nor non-actual, for the predicate is
simply undefined for this thing. We may represent this case by introducing a zero-entity, 0, which
falls outside the whole universe of discourse W: that a predicate's significate in a thing at time t is
0 means that the concept of the predicate is undefined for that thing.
So, to put this in formulas:
Sgt(R)(u)(t)  A(t) U P(t) U I(t) U {0},
where A(t), P(t), and I(t) are disjunct subsets of the universe of discourse W (representing the
sets of signifiable things which are actual, potential or impossible at time t, respectively) and 0 is
not an element of W.
Now from such a significate of R, a particular property of a thing at a certain time, we get what
I call the signification of R, the universal property of R-ness (for a parallel distinction between
"concrete" and "abstract" properties see Küng, 1967), by abstracting from its individualizing
conditions, namely time and subject, i.e., in formal terms, this significate being a value of a
function, by iterated applications of functional abstraction. (For the close parallelism between
functional abstraction and the traditional, Aristotelian conception of abstraction see Geach, 1969
and Klima, 1984.) Functional abstraction in general is effected by the application of Alonzo
Church's lambda-operation. (Curch, 1956) (To be sure, since significates are denoted here as
values of functions - which, again, are values of functions, and so on -, as the pairs of
parentheses, the signs of function-application show, we could denote these functions directly by
simply omitting the pairs of parentheses following the functional signs. However, with this
notation I wish to make even more explicit the derivation of significations as "unsaturated"
functions, by abstracting from their arguments. For the exact description of the working of this
operation see Appendix.) So we can write:
Sg(R)(u) = t(Sgt(R)(u)(t)),
whence, further,
Sg(R) = u(Sg(R)(u)) = u(t(Sgt(R)(u)(t))).

68
Now predicate-signification so-conceived is sufficiently fine-grained to distinguish between
non-synonymous, merely coextensive predicates. (cf. Sober, 1982)
For example, even if necessarily, for every x,
x is trilateral iff x is triangular,
that is, for every u and t,
Sgt(trilateral)(u)(t)A(t) iff Sgt(triangular)(u)(t)A(t)
since for any u and t, it is not the case that
Sgt(trilateral)(u)(t) = Sgt(triangular)(u)(t),
therefore, it is not the case that
Sg(trilateral) = Sg(triangular).
But what about sentence-meanings? Well, having predicate-significations at our disposal, we
can easily construct the significate of a simple predication at a time t according to a given
assignment as follows:
Sgt(Px)(t)(f) = Sgt(P)(f(x))(t).
But how can we proceed to more complex formulae? For in order to have a complete truth-
definition (let alone a complete synonymy relation) that covers all possible formulae we must be
able to construct somehow the significates even of complex formulae. (Of course, in accordance
with the principle of compositionality, as functions of the semantic values of their constituents.)
To this end we must take into consideration even the significates of logical constants in respect
of the significates of their argument-expressions.
So first, the significate of negation:
Sgt(-)(s)A(t) iff not: Sgt(-)(s)A(t),
where s is a sentence-significate, i.e., the significate of some sentence (formula) p at time t,
according to an assignment f.
Whence, the significate of a negated sentence, -p, is constructed as follows:
Sgt(-p)(t)(f) = Sgt(-)(Sgt(p)(t)(f))
Similarly with conjunction:
Sgt(&)(s1)(s2) is an element of
/i/ A(t), if s1 and s2 are elements of A(t)
/ii/ I(t), if s1 or s2 is an element of I(t) and the other is not 0, or if s1 is an element of A(t)
then s2 is not an element of A(t)
/iii/ {0}, if s1 or s2 is 0,
/iv/ P(t) otherwise2
whence
Sgt(p & q)(t)(f) = Sgt(&)(Sgt(p)(t)(f))(Sgt(q)(t)(f))
Now with the quantifiers the situation is somewhat more complicated. But let us just consider:
what does a formula of the form (Ex)(Fx) say? Well, it says that for some value of x, the function
associated to F is satisfied, i.e., in our present framework, that for some value of x the value of
2
As can be seen, the rationale of this rule is the following: a conjunction is /i/ true (its significate is actual) if both of its conjuncts are true, /ii/
impossible, if 1. one of its conjuncts is impossible , or 2. if the truth of the one conjunct implies the falsity of the other, i.e., if they are
incompossible, /iii/ undefined, if one of its conjuncts is undefined, /iv/ otherwise it is possible. We could have given a similarly complicated rule
also for negation, specifying e.g. that the negation is undefined if the negated sentence is undefined, but I wanted to give only the minimal
specification of what may count as negation. Surely further complications would be needed e.g. to distinguish external and internal negation, but
this is not of our present concern.

69
the function associated to F (namely, the signification of F) is actual. But what about more
complex formulae that can stand behind the quantifier? Well, if we look at the signification of F
as being derivable from that of Fx by -abstraction, then the question is not so difficult to
answer.3 As we know the significates of Fx for any value of x, we may construct the signification
of x(Fx) from these, again, by -abstraction. Let
(1) Sgt(x(Fx))(t)(f)(u)
be the significate of x(Fx) at time t according to f in uW, and let this be identical with
(2) Sgt(Fx)(t)(f[x:u]),
where f[x:u] is the same as f except that it assigns u to x.4 Now as we know (2), we also know
(1). But from this we can get the signification of x(Fx) by abstracting from u:
Sg(x(Fx))(t)(f) = u(Sgt(x(Fx))(t)(f)(u))
But then, in a similar vein, we can construct the signification of any -abstract of any
complexity, thereby providing suitable associated functions for quantified formulae.
So, to sum up,
Sg(v(A))(t)(f) = u(Sgt(v(A))(t)(f)(u)),
where
Sgt(v(A))(t)(f)(u) = Sgt(A)(t)(f[v:u]).
And so the signification of the existential quantifier can be defined as follows:
Sgt(E)(V) is an element of
/i/ A(t), if for some uW, V(u)A(t),
/ii/ I(t), if for every uW, V(u)I(t),
/iii/ {0}, if for every uW, V(u)=0
/iv/ P(t), otherwise,
where V(u)A(t) U P(t) U I(t) U {0} (intuitively, V is the place-holder of the signification
function of the "verb-phrase" of the formula); whence the significate of an existentially
quantified formula:
Sgt(Ev(A))(t)(f) = Sgt(E)(Sg(v(A))(t)(f)).
I do not want to bore the reader with the rest of the logical connectives, I think it should be
quite clear by now that by the same recursive method the significates of any complex formula can
be determined, wherefrom, by -abstraction, its signification can be constructed.
In general, if Sgt(exp)(e1)...(en) is the significate of any expression in respect of any entities
whatever, then its signification - Sg(exp) - is got from this by "chopping off" e1...en by iterated
applications of -abstraction.
Now this general definition of signification provides us with a very natural general notion of
synonymy: two expressions are synonymous if and only if their significations are identical, i.e., if
they have the same meaning.

In what follows I will use  also as a syntactic operator: a metasign prefixed by  and a metavariable stands for the object-language string
3

that results from the object-language expression denoted by the metasign by omitting the free occurrences of the variable denoted by the
metavariable. So e.g., if the metavariable v stands for the object-language variable y, and the metasign A stands for the object-language formula
`F(y) & (Ey)(Gy))`, then v(A) stands for the string: `F( ) & (Ey)(Gy))`. Henceforth I will call this the "verb-phrase" of A. In the example in the
main text I use x, F and Fx autonymously in the metalanguage.
4
That is, the significate of a "verb-phrase", i.e., what results from a sentence by leaving off a denoting expression, in respect of a thing is the
same as the significate of the sentence when the denoting phrase in question is taken to refer to the thing in question.

70
Well, so far, so good, as far as the formal theory is concerned - one might say - but how do we,
as users of a language, come to know, if ever, this all-embracing relation? How is it possible that
despite all our everyday semantic competence we have so much trouble in identifying meanings,
if things are so simple as they seem from the above description?
Well, things are not so simple as that at least for two reasons.
1. In actual language usage, when we identify the meanings of expressions, we cannot appeal
to well-defined functions in a well-defined model: all we have are other expressions, and we try
to specify the meaning of one expression by trying to construct other expressions which we judge
to have the same meaning as our explanandum. So we can never give the meaning of an
expression absolutely, by constructing it as an abstract object, as it were; all we can do is to
specify it relatively, by trying to supply synonymous expressions. The situation is somewhat
similar to that when we have to judge the length of things without an étalon. In this case all we
can do is to say which are and which are not of the same length, relying on our fallible sight.
Nevertheless, this does not mean that they do not have a length of their own. The setting up of a
model of the abovementioned type is like establishing an étalon. If we "calibrate" the
significations of the primitive expressions relative to a model, then, on the basis of these
"absolute meanings" we can make judgement about the synonymy relations holding between
them and between expressions built up from them, without having to rely on our vague
intuitions. So in the formal theory what comes first is absolute meaning, which defines
synonymy; in actual usage, however, what comes first is some fragment of the synonymy
relation, which specifies meanings. But things are not even so simple as that.
2. For when we acquire a language we do not get even these splinters, as it were, of the
synonymy relation. As Peter Bosch most aptly wrote: "... in the acquisition of our native language
we get started with partial explanations of the use of words; in the extreme case, these partial
explanations are ostensive definitions. We are shown positive and negative instances and we are
told: `This is a rose, and also that and that; but this here is not a rose'. But in the majority of cases
we are not given any explicit definition at all, not even an ostensive one. Rather we have to pick
up the use of an expression from the various concrete applications we happen to come across.
Now, these are always applications to a limited number of things, always in different contexts. ...
Bit by bit, we might say, the child who is exposed to those partial explanations witnesses the
growth of something more appropriately called a jungle of definitions: intersecting each other,
complementing each other, pushing each other aside, and still leaving whole areas of untouched
desert land in between and wide unexplored oceans around the buzzing jungle. Still, as long as
we move in our jungle, somehow we seem to be doing alright, communication flows uninhibited.
Only when we find ourselves unexpectedly in one of the deserts, or are driven into the ocean, we
stand speechless: for those regions our predicates are not defined". (Bosch, 1983:191)
So the objective, but "intangible" synonymy relation is not only elusive in that it can be
captured by us only in its instances, but, owing to the finitude of our mind and experiences, we
can know it only partially and in a rather unsystematic manner, despite all our everyday linguistic
competence.
So our logic, if it is to give an account of our troubles in identifying meanings, which lies at
the heart of our misunderstandings, then it is bound to provide us also with subjective concepts,
which are somehow partial relative to objective meanings.

71
3. FROM SIGNIFICATIONS TO CONCEPTS

In the present theory, the significate of a predicate P in a thing u at time t is a particular,


individualized property of the thing. We get from this the universal property of P-ness by
abstracting from these two individualizing factors, namely from time and subject. On the other
hand, by the converse operation, concretion, as it may be called, we can arrive again at the
particular individualized property of an individual signified by the predicate P at a certain time.
Now, what if we fill in the empty argument-place of this universal (the signification of P) with
a thing that is apt not only to bear such an individualized property, but which is apt also to
represent such properties, namely a human mind?
Well, then what we get is the significate of the predicate P in this mind, say m, at time t:
Sgt(P)(m)(t), i.e., the mental significate of P at t: Sgtm(P)(t). This is a particular, individual
concept, belonging precisely to this mind. However, this does not mean that this concept is not
universal in another respect, namely in regard to its objects. For even if this concept is particular
in that it belongs to this mind, it is universal in that it is a common representation of many
individuals. (cf. Aquinas, 1933:29.)
But how is this one-to-many relation established? Well, by abstraction, of course. The mind
generates this universal concept by abstraction from particular representations of individuals,
from phantasms, to use Aristotle's happy term. Now for the sake of brevity somewhat
simplifying matters let me give a sketch of how I think from these mental representations
(concepts and phantasms) we can construct a mental language, which may prove useful in our
attempts at understanding understanding.
As I have said, the significate of a predicate P in a mind m at time t is a subjective concept, m's
concept of P at time t:
Sgt(P)(m)(t) = Sgtm(P)(t) = Con(m,t)(P)
But this concept is universal in that it is got by abstraction from phantasms:
Con(m,t)(P) = (Con(m,t)(P)())
Now since phantasms are representations of individuals, we can speak of (here comes the
simplification, but see the next section) the phantasm of an individual u in a mind m: Phant m(u).
But then, we can suppose that the variables of our language, just as they range over individuals
ad extra, so they range over corresponding phantasms representing these individuals apud
mentem. That is, we may suppose that for every assignment of variables f, which assigns to a
variable a thing ad extra, there is another assignment apud mentem, fm, assigning to the variable
in question the phantasm representing the thing:
fm(x) = Phantm(fx)
But then we can construct the significate of an atomic sentence Fx in a mind m at a given time
t according to an assignment apud mentem fm (a "mental sentence", a thought) as follows:
Sgtm(Fx)(t)(fm) = Con(m,t)(F)(fm(x))
Now, to round out the picture, again, just as we supposed a frame of reference ad extra
consisting of sets of actual and non-actual individuals and their properties, so we have to suppose
in the mind a representational frame containing phantasms with their properties assorted in
classes according as the mind takes them to be actual or non-actual.
So, accordingly,

72
Con(m,t)(P)()Am(t) U Pm(t) U Im(t) U {0},
where Am(t), Pm(t) and Im(t) are disjunct subsets of the "discourse-universe" of the mental
representational frame, W(m,t), (note that W(m,t) may change in the course of time) which is a
finite proper subset of W and  is some phantasm: Phantm(u)W(m,t) U {0}. (The case
Phantm(u)=0 represents the situation when m has no phantasm of u.)
Now, on this basis, by a method exactly parallel to the one applied above in connection with
sentence-significates ad extra, namely by defining the mental significates of the logical
connectives, we can determine the mental significates of any complex formula.
For example,
Sgtm(-)(sm)Amt iff not: smAm(t),
and so on.
So in this way for any given formula A we can construct two corresponding significates (at a
given time, according to a given assignment): one ad extra, the actuality of which means that A
is true (at the given time, according to the given assignment), and another apud mentem, which
being in the actual domain of the representational frame of a subject means that the subject
assents to A (at t according to the corresponding assignment apud mentem).So whereas actuality
ad extra founds the truth of, actuality apud mentem founds belief in a proposition.
So we can see that the same syntactic structure allows us to construct to it two different, but
strictly corresponding significates. (cf. Bealer, 1982) And thus, just as the network of significates
ad extra defines for us a synonymy relation between significations, i.e., (socially) objective
meanings, so the network of significates apud mentem defines a synonymy relation apud
mentem, between subjective concepts.
But we have this pretty rich model for a pretty poor language. Indeed, this language is so poor
that despite the richness of its semantics it could only very poorly reflect the problems involved
in understanding or misunderstanding each other in a human language. As a matter of fact, it
could only serve to illustrate the semantic method by which we can handle even much richer
formal languages approximating better the complexity of natural languages.
In the next section, therefore, I sketch some extensions of our formal language at hand, in
order to show the ways in which I think we can get closer to the actual syntax and semantics of
natural languages.

4. APPROACHING NATURAL LANGUAGE

The first thing to do by way of extension is to introduce many-place predicates. Now, as


already one-place predicates are treated as if they had an extra-argument for time, this sort of
extension comes very naturally as follows:
Sgt(Pn)(u1)...(un)(t)A(t) U P(t) U I(t) U {0},
(of course, if u1 or ... un=0, then it is not the case that Sgt(Pn)(u1)...(un)(t)A(t)), whence the
significate of an atomic sentence with an n-place predicate is constructed as follows:
Sgt(Pn(tm1)...(tmn))(t)(Sp)=Sgt(F)(Sp(tm1)(t))...(Sp(tmn)(t))
where tm1...tmn are terms and Sp(tm1)(t)...Sp(tmn)(t) are their supposita at time t.

73
Supposition, as it was conceived by medieval logicians, is the referring function of general, as
well as of singular terms. As I have argued elsewhere (See Essay II. of this volume), supposition
therefore should be regarded as an extended assignment function, extended in that it assigns
referents not only to simple variables, but also to general terms, which are thus to be regarded as
restricted variables.
The introduction of restricted variables into the apparatus of standard quantification theory
greatly increases the expressive power and faithfulness of quantification theory in representing
the workings of quantificational and cross-referential devices (determiners, pronouns) of natural
languages. (See Essay III. of this volume.)
The introduction of restricted variables into the present framework is served by the following
clauses:
/i/ If v is a variable and A is a formula, then v.A is a restricted variable
(Notice that this clause is applicable recursively, so that it also allows of "nested" occurrences
of restricted variables. (Cf. Prullage, 1976))
/ii/ If x is a simple variable, then Sp(x)(t)W
/iii/ If v.B is a restricted variable, then Sp(v.B)(t)=Sp(v)(t), if Sgt(B)(t)(Sp)A(t),
otherwise Sp(v.B)(t)=0
Of course, the introduction of restricted variables has to affect also the clauses determining the
significates of quantifiers. On the other hand, since with restricted variables we have to regard all
quantifiers as two-place functors (at least, see again Essay II. of this volume), here we are able to
formulate these clauses so generally that they cover not only the standard quantifiers ("every" and
"some"), but also the astounding multitude of natural language determiners such as "most", "the",
"less than half of the", "three", etc. (See Keenan & Stavi, 1986)
The relevant clauses are the following:
If Q is a determiner, then Sgt(Q)(N*)(V) is an element of
/i/ A(t), if for Q'uN*, V(u)A(t),
/ii/ I(t), if it is not the case that for Q'uN*, V(u) is not an element of I(t)
/iii/ 0, if it is not the case that for Q'uN*, V(u) is not identical with 0,
/iv/ P(t) otherwise,5
where Q' is the natural language determiner represented by Q, N* is a subset of W U {0}, and
V is a function, such that for every uW U {0}, V(u)W U {0}. (Intuitively, N* is the place-
holder of the range of the noun-phrase determined by Q, while V is the place-holder of the
signification function of the verb phrase of the sentence in which Q occurs.) Hence,
Sgt(Qv.B)(C)(t)(Sp)=Sgt(Q)(Rg(v.B)(t)(Sp))(Sg(v.B(C))(t)(Sp))
where Rg(v.B)(t)(Sp), the range of v.B, is defined as follows:
Rg(v.B)(t)(Sp)={uW: Sgt(B)(t)(Sp[v:u])A(t)}, if this set is not empty, otherwise
Rg(v.B)(t)(Sp)={0}. (Recall also the definition of the signification of -abstracts above,
substituting Sp for f in it.)
Now, our new terms, the restricted variables, have not only supposition, but also signification:
a restricted variable, say, x.Fx, has not only the function to refer to (to denote) a thing which is F,
but also to signify (to connote) the F-ness of this thing. But this connoted property need not
5
Compare the parallel definition of the significate of the existential quantifier above.

74
always be identical with the significate of the matrix of such a variable: for example,
Sgt(x.Fxy)(t)(Sp) need not be identical with Sgt(Fxy)(t)(Sp), for in that case it should also be
identical with Sgt(y.Fxy)(t)(Sp); but, say, Plato's love towards Socrates need not be identical with
Socrates's being loved by Plato.
So we can suppose that the dot, the term-forming operator of restricted variables (the natural
language equivalents of which are relative pronouns), does double duty: it not only restricts the
range of the operator variable, for it does so only by making the matrix connote a property in the
supposita of the operator variable; indeed, the restricted variable stands only for those supposita
of the operator variable which actually have this property.
So the clause determining the signification of the dot (of relative pronouns) runs as follows:
Sgt(.)(u)(s)A(t), I(t), {0}, or P(t) according as sA(t), I(t), {0}, or P(t), respectively, whence
the significate of a restricted variable at time t according to a given supposition is constructed in
the following manner:
Sgt(v.A)(t)(Sp)=Sgt(.)(Sp(v)(t))(Sgt(A)(t)(Sp)).
Now recognizing this connotative function of general terms should make us rethink what I
have said about phantasms, the natural candidates for being the mental supposita of general terms
in their referring function. Phantasms, as I have said, are mental representations of individuals.
But individuals as a rule do not appear to us as nude individuals, bereft of all properties.
Phantasms always represent individuals as being endowed with general properties, signifiable by
general terms. As St. Thomas Aquinas puts it in his Commentary on the Posterior Analytics:
"... the senses perceive somehow even the universal. For they perceive Callias not only as being
Callias, but also as being this man, and similarly Socrates, as being that man. Hence, such a
perception of the senses being at hand, the intellect is able to see man in both ..." (Aquinas, 1882,
lb.2.lc.20.) So, instead of speaking about the phantasm of a thing absolutely (this was the
simplification noted above), we should rather speak of the phantasm of a thing u (in a mind m, at
time t), as bearing a certain property p: Phant(m,t)(u,p), where Phant(m,t)(u,p)W(m,t) U {0},
whence the mental suppositum of a restricted variable (at time t) is:
Spm(v.A)(t)=Phant(m,t)(Sp(v.A)(t),Sgt(v.A)(t)(Sp)).
Now, as we can see, in this way the same thing may be represented by different phantasms,
and different things may be represented by the same phantasm. But then even terms which are
coreferential ad extra may have different supposita apud mentem.
But since, as I have said, what founds belief, is the actuality of the mental significate of a
sentence in one's mental representational frame, no wonder if one who does not know about the
coreferentiality of two terms ad extra may assent to a sentence with the one term, while dissent
from the like sentence with the other. And so failure of substitutivity of coreferential terms gets a
rather smooth explanation in this framework.
For an adaptation of cases involving proper names we should suppose that even a proper name
can have a significate distinct from its suppositum. Well, why not? After all, getting to be named,
say, John, confers upon one the property of being named John, i.e., of being the suppositum of
the name "John".
So, accordingly, Sgt(j)(u)(t)A(t), if Sp(j)=u, otherwise it is 0, where Sp(j)W, whence
Spm(j)(t)=Phant(m,t)(Sp(j),Sgt(j)(Sp(j)(t))).

75
This much having been said, I think I should also add that in the present theory only simple
variables have no such kind of connotation: Sgt(x)(t)=Sp(x)(t)W.
Now what we have here is already a quite rich intensional theory, but with an extensional
language, in the sense that we still do not have in our language intensional expressions such as
modalities, tense operators or attitude verbs.
However, with this rich theory at hand we do not even need these various kinds of special
expressions, the so-called intensional operators. In line with George Bealer's arguments, we may
abandon the "multiple-operator approach" in favour of an "intensional abstraction approach" by
introducing terms referring to "intensional entities" - to significates of expressions. The
introduction of these new terms (sentential terms as I will call them, representing sentence-
nominalizations) is served by the following clauses:
/i/ If A is a formula, then [A] is a (sentential) term
/ii/ Sp([A])(t)=Sgt(A)(t)(Sp); Spm([A])(t)=Sgtm(A)(t)(Spm)
/iii/The significates of the above terms are identical with their supposita.
Now with these terms at hand we can offer satisfactory formulations and solutions of the
intentional paradoxes. (I think, the representation of modal discourse is trivial in this setting: a
possibility predicate, `it is possible', states of a sentential term, `that p', that its suppositum, i.e.,
the significate of p at t, is possible, i.e., is an element of P(t).) Let me sketch just one example:
(1) Plato is identical with Aristocles
(2) John believes that Plato is wise
(3) John believes that Plato is Aristocles
---------------------------------------------
John believes that Aristocles is wise
(1) and (2) do not imply the conclusion, however, (2) and (3) do.
In a straightforward formalization the argument looks as follows:
(1) p=a
(2) jB([Wp])
(3) jB([p=a])
-------------------
jB([Wa])
Here the natural stipulation on the interpretation of B is the following:
Sgt(jB(trm))(t)(Sp)=Sgt(B)(Sp(j))(Spjm(trm)(t))(t)A(t) if and only if Spjm(trm)(t)Ajm(t), where
trm is a sentential term, and `jm' denotes John's mind.
"Plato" and "Aristocles" being different names, though they stand for the same individual ad
extra, they may have different phantasms in John's mind as their mental supposita. But then the
mental significates of Wp and Wa, i.e., the mental supposita of [Wp] and [Wa], may also be
different. And so also the significates of (2) and the conclusion may be different, despite the truth
of (1). So (1) and (2) may both be true, while the conclusion false. On the other hand, if (3) is
true, then the names p and a must stand for the same phantasm in John's mind. But then also
[Wp] and [Wa] must stand for the same thought (of John's mind), and so the significate of (2)
and that of the conclusion are the same. Hence, if (2) and (3) are true, then so is the conclusion.
Now I think it is easily imaginable how this kind of treatment can be extended to cover also
other intentional paradoxes, such as e.g. Kripke's puzzle about belief. (Cf. Laurier, 1986:47) But

76
then with a theory at hand that is able to offer a satisfactory treatment of the intentional
paradoxes we may be more confident about what it brings out concerning the problems of
understanding.

5. TOWARDS UNDERSTANDING

I hope that even from the above rather sketchy presentation a quite rich picture of a semantics
emerged: we have a quite complex language with a twofold interpretation, one part of which is
ad extra and the other is apud mentem. Both define a network of semantic values, significations
ad extra and concepts apud mentem, characterizable by their respective synonymy relations.
Concepts are partial relative to significations in that they are defined only for finite real subsets
of the domain of significations. (At least, in their relation to one's subjective representational
system. For more on this see Appendix and Essay VII.) Still, there is some similarity between the
network of significations and that of concepts. Indeed, a suitably defined homeomorphism (i.e.
partial isomorphism) of their respective synonymy relations may express their conformity. And
what this conformity expresses should be nothing but the degree of the subject's semantic
competence who possesses this network of concepts. For apart from John's case above, who
simply does not know that Plato was originally called Aristocles, it may also be the case that, due
to our finitude of mind, we do not know the whole of the synonymy relation of our language.
This is why we have so many troubles with providing correct definitions or unexceptionable
taxonomies. And this is why Theaetetus, in his quest for knowledge, might come to know
without any kind of external experience what he, despite his everyday semantic competence,
formerly had not known: by Socrates's questions he was led to adjust the synonymy relation of
his subjective representational system (remember, it changes in time!) to the objective synonymy
relation existing in his language. (Cf. Bealer's treatment of the paradox of analysis.)
On the other hand, as I have already indicated, the objectivity of this synonymy relation is
social objectivity. This is the kind of objectivity possessed by e.g. prices. This kind of objectivity
derives from the behaviour of groups of people. (Whether or not "behind" this kind of objectivity
there is some more basic, unchanging objectivity, limiting the possible variety of human concepts
and languages, inherent in human nature or the nature of things is a further question. Notice that
the theory presented is, as a good logical theory should be (cf. Henry on "logical aloofness",
Henry, l984), able to give room for a great variety of ontologies ranging from extreme realism to
utmost nominalism and conventionalism. For more on this see Appendix.) But then, since people
are free agents, if their behaviour changes, this changes also the structure of the kind of
objectivity their behaviour constituted. And this is why meanings can change through the activity
of users of the language. If a rule is strong (many people keep to it), then acting against it is just
to make a mistake, if, on the other hand, the rule is weak, or there is no rule at all, then users of
the language are quite free to create new rules thereby changing the whole network of
significations. (Just consider the change caused by the change of the concept of motion. When
"moves" (movetur) ceased to mean "is being moved /by something/" (movetur /ab aliquo/), then
arguments for a prime mover became pointless.)
Now suppose there is a population of people who speak the same language. In our present
framework this means that they use the same expressions in a way which constitutes an objective

77
synonymy relation between these expressions. We may say that these people generally
understand each other if and only if in keeping with the rules according to which the significates
of expressions are to be constructed, possessing homeomorphic synonymy relations, they are able
to construct in their mental representational systems those mental significates of these
expressions which correspond to the same significate of these expressions ad extra.
But this kind of understanding may be hampered by several kinds of impediments:
1. SYNTACTIC AMBIGUITIES

The construction of the significates of a complex expression, as we have seen, depends on its
syntactic structure. So if a complex expression has different syntactic analyses (e.g. "Every boy
loves some girl)", "A white thing can be black"), then surely different significates may pertain to
it as it is analysed in one way or another.
2. SEMANTIC AMBIGUITIES

Languages abound with equivocal terms. (pot1=earthenware vessel; pot2=marijuana) Equivocal


terms are those which have different significations. But then the same word construed in one way
or another may contribute to the construction of non-corresponding significates of the same
expression in the minds of different language users.
3. PRAGMATIC AMBIGUITIES

Some expressions are unspecific on purpose. They contain some hidden variable element the
value of which is specified by context. For example, as Dag Westerstahl has convincingly
argued, the definite article is to be regarded as a context set indicator. (Westerstahl, l985) It
requires uniqueness of reference of the noun phrase following it only in respect of a limited set of
things: when I say "I was in the kitchen", I do not want to imply that I was in the only kitchen in
the world. (For the introduction of context sets into the theory of restricted variables see Essay
III.) Now failure to identify context sets properly is a very frequent source of misunderstandings.
4. IGNORANCE

In the case of John above, his lack of acquaintance with Diogenes Laertius's piece of
information might throw him into serious trouble at a history of philosophy examination. He
might /mis/understand the malicious examiner's questions concerning Aristocles as relating to a
person whom he should know but (for all his knowledge about Plato) does not.
5. INCOMPETENCE

Incompetence (I mean, semantic incompetence) is some partial anisomorphism between one's


subjective synonymy relation and the objective synonymy relation. For example, one who does
not know that pot2=marijuana may badly misunderstand a pot2 dealer. Now since this kind of
impediment of understanding depends on partial anisomorphisms, it may well be the case that
two persons who can quite fluently chat about certain things are shocked at hearing the
"nonsenses" the other says when it comes to others. I think that here is a quite smooth transition
from cases of incompetence, when one clearly breaks a definite rule, to cases of
6. CONCEPTUAL DIFFERENCES

where there is no definite rule to break, so users of the same language are quite free to create
their own concepts. (This is why it is so difficult sometimes to tell the mad from the genius.)

78
These conceptual differences sometimes may prove to be so grave that in some cases two
speakers, even if they use the same words, can hardly be said to speak the same language.
(Compare e.g. the language of Heidegger and Carnap.)
7. DIFFERENT CONCEPTUAL SCHEMES

Now if we finally take into consideration the fact that when we learn our native tongue we do
not learn words by definitions, but we are trained to recognize things as being such and such,
as falling under this or that concept (remember how, according to the present theory, phantasms
represent things as having some property), then it is just to be expected that different linguistic
communities may develop different conceptual schemes. (In the sense of Walton, l973)
***

Having a theory at our disposal which is rich and flexible enough to give an exact
reconstruction of the impediments of understanding we may have hopes to give at least correct
diagnoses (even if not remedies) of the several kinds of ununderstanding that plague human
communication. However, one must not be stupidly optimistic. At the end of this paper I am not
quite sure that I have succeeded in making myself understood.

79
REFERENCES

Anderson, C.A. 1987 "Bealer's Quality and Concept", Journal of Philosophical Logic 2(1987)
115-165.
Aquinas, T. 1933 De Ente et Essentia, ed.: C. Boyer, Rome
Aquinas, T. 1882 In Libros Posteriorum Analiticorum Expositio, in: Opera Omnia, Editio
Leonina, Tom. I., Rome
Bealer, G. 1982 Quality and Concept, Oxford
Bosch,P. l983 "Vagueness is Context-Dependence. A Solution to the Sorites Paradox", in:
T.T. Balmer-M. Pinkal: Approaching Vagueness, North Holland
Church, A. 1956 Introduction to Mathematical Logic, Princeton
Geach, P.T. 1969 "Form and Existence", in: God and the Soul, London
Hintikka, J. 1986 "The Varieties of Being in Aristotle", in: S. Knuuttila-J. Hintikka: The Logic
of Being, Dordrecht
Henry, D.P. 1984 That Most Subtle Question, Manchester
Jacobi, K. 1986 "Peter Abelard's Investigations into the Meaning and Functions of the Speech
Sign "Est"', in: Knuuttila-Hintikka: The Logic of Being, Dordrecht
Kamp, H. 1981 "A Theory of Truth and Semantic Representation", in: Groenendijk-Janssen-
Stokhof: Formal Methods in the Study of Language, Amsterdam
Keenan,E.L. and
Stavi,J. 1986 "A Semantic Characterization of Natural Language Determiners", Linguistics
and Philosophy, 253-326.
Klima, Gy. 1984 "Thomas Aquinas on the Meaning of Words", (in Hungarian, with English
abstract) Magyar Filozófiai Szemle 28(1984), 298-313.
Küng, G. 1967 Ontology and the Logistic Analysis of Language, Dordrecht
Landman, F. 1986 Towards a Theory of Information, Amsterdam
Laurier, D. 1986 "Names and Beliefs: A Puzzle Lost", The Philosophical Quarterly, 37-5O.
Lear, J. 1982 "Aristotle's Philosophy of Mathematics", The Philosophical Review, April,
pp. 161-192.
Lewis, D. 1976 "General Semantics", in: B. Partee: Montague Grammar, New York-San
Francisco-London
Matthen, M. 1983 "Greek Ontology and the "Is" of Truth", Phronesis, 113-135.
Montague, R. 1973 "PTQ", in: Hintikka-Moravcsik-Suppes: Approaches to Natural
Language, Dordrecht
Moody, E.A. 1957 Truth and Consequence in Medieval Logic, Amsterdam
Perry, J. 1986 "From Worlds to Situations", Journal of Philosophical Logic, 83-1O9.
Prullage, M.M. 1976 "A Theory of Restricted Variables without Existential Assumptions",
Notre Dame Journal of Formal Logic, 589-612.
Schmidt, R.W. 1966, The Domain of Logic according to Saint Thomas Aquinas, The Hague,
Martinus Nijhoff
Sober, E. 1982 "Why Logically Equivalent Predicates May Pick Out Different Properties?",
American Philosophical Quarterly, 183-191.

80
Walton, K.L. 1983 "Linguistic Relativity", in: Pearce-Maynard: Conceptual Change,
Dordrecht
Weidemann, H. 1986 "The Logic of Being in Thomas Aquinas", in: Knuuttila-Hintikka: The
Logic of Being, Dordrecht
Westerstahl, D. 1985 "Determiners and Context Sets", in: van Benthem - ter Meulen:
Generalized Quantifiers in Natural Language, Foris Publications

81
APPENDIX

In this Appendix I wish to give an exact description of the formal theory sketched in the body
of the paper lest any technically obscure point should remain.
The language of the theory, L, is defined as follows: L:=<C,Pr,V,Trm,F>, where C:={-
,&,Q,.,(,),[,],=}, Pr:=Pind U Ppred, where Pind is the set of individual parameters, or names,
Ppred is the set of n-place predicate parameters and V is the set of simple, or proper variables.
These sets contain the primitive symbols of L. The complex expressions of L are contained in
Trm, the set of terms, and F, the set of formulae of L. These sets are defined by the following
simultaneous recursive definition:
1. If aPind, then aTrm.
2. If xV, then xTrm.
3. If tmTrm and BF, then tm.BTrm
4. If tm1,...,tmnTrm and PnF, then Pn(tm1)...(tmn)F and (tmi=tmj)F
5. If B, CF, then ~(B)F, (B & C)F
6. If vTrm and BF, then (Qv)(B)F
7. If BF, then [B]Trm
A semantic system, SYST, for L is defined as follows:
SYST:=<RF,Sg>,
where RF, the referential frame of SYST is defined in the following way:
RF:=<W,A,P,I,T,Ra,Re,Abs,<,0>,
where W is a nonempty set, A(t), P(t), I(t) are disjunct subsets of W for any tT, T is a
nonempty subset of W ordered by <, Ra is a subset of W, Re is a subset of Ra, Abs is a subset of
W, and 0 falls outside W, i.e., 0 is not an element of W. Intuitively, W is the universe of
discourse, the set of everything that can be referred to or signified by any means in L. A(t), P(t)
and I(t) are the sets of signifiable things, which are actual, potential, or impossible at time t,
respectively. T is a set of time points (or intervals), ordered by <, the relation "earlier than". Ra is
the set of rationate beings and Re is the set of real beings. (For the intuitive distinction see
Schmidt, 1966) Abs is the set of abstract entities, such as sets and functions. 0 is the zero-entity,
the semantic value of empty terms, and of predicates in respect of things for which they are
undefined.
Sg, the signification function of SYST, is defined as follows:
Sg(exp)=e1(...en(Sgt(exp)(e1)...(en))...),
where exp is any expression of L (including primitive symbols, except for brackets and
parentheses), e1,...,en are any entities whatever (i.e., elements of W U {0}, elements of L and
functions defined thereon) and Sg(exp)Abs.
NOTE that I use the -operator according to the following equivalences:
If f and g are functions, then
f(x)(y)=g(y) iff f(x)=g iff y(f(x)(y))=g
NOTE also that the distinction between signification (Sg) and significate (Sgt) is intended to
reflect the difference between unsaturated and saturated entities, respectively. A significate of an
expression in respect of this and that is always a saturated, complete entity. (But note here the

82
special status of concepts. Cf. Essay VII.) The signification of an expression (perhaps, in respect
of this or that) is always an unsaturated, abstract entity.
Now the significates of expressions of L in an RF are determined by the following clauses:
(1) If PnPpred, u1,...,unW* and tT, then Sgt(Pn)(u1)...(un)(t)W(t)*, where W*:=W U {0},
W(t)*:=A(t) U P(t) U I(t) U {0}, and if ui=0, then not: Sgt(Pn)(u1)...(un)(t)A(t), whence
(2) Sgt(Pn(tm1)...(tmn))(t)(Sp)=Sgt(Pn)(Sp(tm1)(t)) ... (Sp(tmn)(t))(t), where Sp(tmi)(t), is a
suppositum
of the term tmi at time t. The supposition function of terms is defined by the following
clauses:
(i) If xV, tT, Then Sp(x)(t)W
(ii) If aPind, then Sp(a)W, and Sp(a)(t)=Sp(a)
(iii) If tm.BTrm, then Sp(tm.B)(t)=Sp(tm)(t), if Sgt(B)(t)(Sp)A(t), otherwise
Sp(tm.B)(t)=0.
(iv) If [B]Trm, then Sp([B])(t)=Sgt(B)(t)(Sp)
(3) If sW(t)*, then Sgt(-)(s)A(t) iff not: sA(t), whence
(4) Sgt(~(B))(t)(Sp)=Sgt(-)(Sgt(B)(t)(Sp)).
(5) If s1,s2W(t)*, then Sgt(&)(s1)(s2) is an element of
/i/ A(t), if s1A(t) and s2A(t)
/ii/ I(t), if s1I(t) or s2I(t) and the other is not 0 or if s1A(t) then not: s2A(t)
/iii/ {0}, if s1=0 or s2=0
/iv/ P(t) otherwise, whence
(6) Sgt((B&C))(t)(Sp)=Sgt(&)(Sgt(B)(t)(Sp))(Sgt(C)(t)(Sp))
(7) If u1, u2W(t)*, then Sgt(=)(u1)(u2)(t) is an element of
/i/ A(t), if u1=u2A(t)
/ii/ P(t), if u1=u2P(t)
/iii/ I(t) otherwise, whence
(8) Sgt(tm1=tm2)(t)(Sp)=Sgt(=)(Sp(tm1)(t))(Sp(tm2)(t))(t)
(9) If N* is a part of W*, and V(u)W(t)* - intuitively, the place-holders of the range of the
"noun phrase" and the signification of the "verb phrase" of the quantified statement, respectively
-, where uW*, then
Sgt(Q)(N*)(V) is an element of
/i/ A(t), if for Q'uN*, V(u)A(t),
/ii/ I(t), if it is not the case that for Q'uN*, not: V(u)I(t)
/iii/ {0}, if it is not the case that for Q'uN*,not: V(u)=0,
/iv/ P(t) otherwise, where Q' is a natural language determiner represented by Q. Hence,
(10) Sgt((Qv)(C))(t)(Sp)=Sgt(Q)(Rg(v)(t)(Sp))(Sg(v(C)(t)(Sp)), where
Rg(v)(Sp)(t)={uW: for some Sp' differing from Sp only in the value assigned to v,*
Sp'(v)(t)=u}, if
this set is not empty, otherwise Rg(v)(Sp)(t)={0}; Sg(v(C)(t)(Sp)=u(Sgt(C)(t)(Sp)(u)),
where for every uRg(v)(t), Sgt(C)(t)(Sp)(u)=Sgt(C)(t)(Sp[v:u]),

*
And, of course, its nested operator variables, if it has any.

83
where, Sp[v:u](w)(t)=u, if v=w, otherwise Sp[v:u](w)(t)=Sp(w)(t).
And this completes the definition of the signification function for formulas. What remains, then,
is to define it for terms. (For terms thus far we have defined only supposition.)
(11) If xV, then Sgt(x)(t)=Sp(x)(t)
(12) If aPind, then Sgt(a)(u)(t)A(t), if u=Sp(a)(t), otherwise Sgt(a)(u)(t)=0
(13) If tm.BTrm, then Sgt(tm.B)(t)(Sp)=Sgt(.)(Sp(tm)(t))(Sgt(B)(t)(Sp)),
where
(14) Sgt(.)(u)(s) is an element of A(t), I(t), P(t) or {0},
according as s is an element of A(t), I(t), P(t) or {0}, respectively.
(15) If [B]Trm, then Sgt([B])(t)(Sp)=Sp([B])(t)
Now clauses (1)~(15) /together with clauses (i)~(iv)/ determine the semantic values of L in an
RF ad extra. What remains to be done, therefore, is to define their semantic values apud
mentem. To this end first we have to define the subjective representational frame of a mind m
at time t, SRF(m,t), as follows:
SRF(m,t):=<W(m,t),Am,Pm,Im,Tm,Rem,Ram,Absm,<,0>,
where W(m,t) is a finite, proper subset of W, Am(t), Pm(t) and Im(t) are disjunct subsets of
W(m,t), Tm is a subset of T, Rem, Ram and Absm are subsets of W(m,t), < is an ordering on Tm and
0 is the same as above. Intuitively, the elements of SRF correspond to the elements of RF above.
Now a subjective representational system of a mind m at time t, SYST(m,t), consists of such
a representational frame, containing the mind's concepts and phantasms:
SYST(m,t):=<SRF(m,t),Con(m,t),Phant(m,t)>,
where, if 1,...,n are elements of W(m,t), then
Con(m,t)(Pn)(1)...(n)  Am(t) U Pm(t) U Im(t) U {0}
and
Phant(m,t)(u,p)  W(m,t) U {0}
where u and p are elements of W-W(m,t), i.e., u is an extramental thing, with p, its extramental
property.
The subjective interpretation of the language L, then, is determined by the following clauses:
(1m) Sgtm(Pn)(t)=Sgt(Pn)(m)(t)=Con(m,t)(Pn), whence
(2m) Sgtm(Pn(trm1)...(trmn))(t)(Spm)=Con(m,t)(Pn)(Spm(trm1)(t))...(Spm(trmn)(t))
where the mental suppositum of a term trmi is defined:
Spm(trmi)(t)=
(im) =Phant(m,t)(Sp(trmi)(t),Sgt(trmi)(t)), if trmi is a proper variable,
(iim) =Phant(m,t)(Sp(trmi),Sgt(trmi)(Sp(trmi))(t)), if trmi is an individual parameter,
(iiim) =Phant(m,t)(Sp(trmi)(t),Sgt(trmi)(t)(Sp)) if trmi is a restricted variable or a sentential term.
The rest of the clauses determining the mental significates of expressions of L run exactly
parallel with those determining the extramental significates of the same expressions. (Were this
not the case, this would mean that the subject in question has different concepts from those
established by common language usage even for the logical connectives. For the representation
of this case, however, rather further special syntactic signs would have to be introduced. For
example, a sign for Hegelian negation, etc.) Only by way of example:
(3m) Sgtm(-)(sm)(t)Am(t) iff not: smAm(t), whence

84
(4m) Sgtm(~(B))(t)(Spm)=Sgtm(-)(Sgtm(B)(t)(Spm))(t),
and so on. As no theoretical or practical difficulty is involved in the formulation of the
remaining clauses, let me omit their statement here.
However, before completing the formal description of this theory (by giving the definitions of
truth and validity) I have to clarify one more, theoretically important point. (I am indebted to
István M. Bodnár for having called my attention to this point.) It concerns a highly important
distinction of Aristotelian semantics between the immediate and ultimate significates of
linguistic expressions. For as we know from Aristotle's Peri Hermeneias (lb.I.c.1.), our words
signify immediately our concepts, and only by the mediation of these the (properties of) things.
The necessity of this distinction also in the present framework can be shown by the following
reasoning: in the body of the paper it has been said that it is the actuality of the significate of a
predicate in a thing that verifies the predicate of this thing. For example: the predicate "red" is
true of this apple if and only if the significate of this predicate in this apple is actual, i.e., iff the
redness of this apple exists. On the other hand, we have also said that the significate of a
predicate in a mind is a concept. But this would mean that the actual possession of a concept
signified by a predicate would verify the predicate of the mind which possesses this concept. For
example: on this account by possessing the concept of immateriality a mind would have to be
immaterial, and by possessing the concept of materiality the same mind would have to be also
material, which is absurd.
However, by drawing the abovementioned distinction between the immediate and ultimate
significate of a word, this absurdity can be avoided. We have said that the significate of a
predicate in a mind is a concept. Let us say now that this is the immediate significate of this
predicate in this mind. But this concept is a universal, i.e., a function, which, when its empty
argument-place is filled in with a phantasm, yields a thought within the subjective
representational system of the mind. On the other hand, we have said nothing thus far about the
application of this concept to the extramental objects themselves. But since according to Aristotle
a concept is the natural sign of a (property of) a thing, we can say that a concept, when applied to
an extramental thing, i.e., when its argument- place is filled in with a thing, yields an extramental
property of the thing as its value, namely the property which this concept is the natural sign of.
And this is the ultimate significate of the word signifying the concept immediately. And it is the
actuality of this significate that verifies the predicate of the thing. For example: the word
"immaterial" in a mind m signifies immediately the concept of immateriality inherent in this
mind. But the inherence of this concept does not verify the predicate of the mind. However, the
concept can be applied to a thing, indeed, to the mind itself, when it yields the individualized
property inherent in this mind, which is the ultimate significate of the word "immaterial". And it
is the actual inherence of this property that verifies the word of this mind. So even if the mind
possesses also the concept of materiality, this would not verify of it the word "material", since the
ultimate significate of this word is not actual, indeed, as the case may be, it is impossible in this
mind. (For more on this cf. Essay VII.)
Now if we denote the ultimate significate of an expression exp in respect of any kind of
entities e1,...,en, as Sgt'(exp)(e1)...(en), then we can give these ideas a formal expression by the
following clauses:
(1') Sgt'(Pn)(u1)...(un)(t)=Sgt(Pn)(u1)...(un)(t), provided u1...un is not a human mind.

85
(2') Sgt'(Pn)(u1)...(un)(t)=Con(m,t)(Pn)(u1)...(un), provided u1...un is not a phantasm. However,
(3') Sgt'(Pn)(m)(t)=Con(m,t)(Pn)(m)(t), but this is not identical with
Sgt(Pn)(m)(t)=Con(m,t)(Pn)
(4') Sgt'(Pn(tm1)...(tmn))(t)(Sp)=Sgt'(Pn)(Sp(tm1)(t)...(Sp(tmn)(t))(t)
(5') Sgt'(~(B))(t)(Sp)=Sgt(-)(Sgt'(B)(t)(Sp)).
(6`) Sgt'((B&C))(t)(Sp)=Sgt(&)(Sgt'(B)(t)(Sp))(Sgt'(C)(t)(Sp))
(7') Sgt'(tm1=tm2)(t)(Sp)=Sgt'(=)(Sp(tm1)(t))(Sp(tm2)(t))(t),
where identity is to be treated as a distinguished predicate obeying (1')~(3').
(8') Sgt'((Qv)(C))(t)(Sp)=Sgt(Q)(Rg(v)(t)(Sp))(Sg'(v(C)(t)(Sp)), where
Sg'(v(C)(t)(Sp)=u(Sgt'(C)(t)(Sp)(u)),
where for every uRg(v)(t) Sgt'(C)(t)(Sp)(u)=Sgt'(C)(t)(Sp[v:u]).
Now with these formulations at hand we can give as the general definition of truth for L in
respect of a SYST the following:
(T) If B is a formula, then B is true at time t, if and only if for some Sp, Sgt'(B)(t)(Sp)A(t).**
Validity, then, is truth at every time in every SYST, and formal validity of an inference is the
validity of the corresponding conditional, as usual.
Synonymy of expressions is identity of their significations. Synonymy of two expressions for
a mind is identity of their mental significates or significations (derivable from the mental
significates specified above by -abstraction). Synonymy of significations or concepts is their
identity. Synonymy, of course, may be partial, when it is nothing but the partial identity of
significations. (That is, identity of significations restricted to a subdomain of their domains.)
A homeomorphism of synonymy relations may be defined as follows:
If Syn1 and Syn2 are synonymy relations and x and y are elements of subsets of the domain and
the range of Syn1, respectively, then Syn1 is homeomorphic to Syn2 iff there is a 1-1 function f
such that <x,y>Syn1 iff <f(x),f(y)>Syn2, where f(x) and f(y) are elements of subsets of the
domain and range of Syn2, respectively.
Finally, I think, some remarks concerning the general character of this system are in order,
especially in view of its complexity. As I have noted in the main text, in keeping with
D.P.Henry's aloofness principle, I wanted to construct this system as general as possible, so that
it would not impose any artificial restriction either upon the intuitions of users of a language, or
upon the possible ontologies available (i.e. formalizable) in it. Now this is one reason for the
complicated nature of this system: it is only at the expense of reducing its generality and its
expressive power that we could reduce its complexity. For example, by restricting our language
only to uniform quantification, or to talk about extramental entities, excluding even human minds
as possible referents of our expressions we could omit several clauses of the above description.
But I think this would heavily reduce the philosophical import of this system. On the other hand,
with this construction, treated as a general framework, we can give formal expression to a great
variety of linguistic forms and related ontological viewpoints. Only by way of illustration let me
mention just some of the several possibilities.
(1) By introducing terms referring to significations, and by placing significations in the domain
of real beings we get the Platonic world of universal ideas.
**
Note how well this definition corresponds to Aristotle's conception of truth according to which a proposition is true iff what it signifies exists.

86
(2) By placing significations in the domain of rationate beings and imposing some restrictions
upon the discourse about them (cf. Lear, 1982) we get the Aristotelian theory of universals. (For
more on this see Essay VII.)
(3) By introducing into the language of the system a copula and an existence predicate, along
with appropriate semantic rules, we can offer exact reformulations of a number of traditional
metaphysical theses, such as Aquinas's thesis of the real distinction between essence and
existence in the creatures, or the thesis of the unity of substantial forms. (See Essay VI.)
(4) By identifying the supposita and significata of (absolute, non-connotative) general terms,
and, correspondingly, analyse the copula of the categoricals in terms of identity, we get the
metaphysics of 14th century nominalism. (See again Essays VI. and VII.)
Finally, lest anyone should think that this system is able to represent only traditional
philosophies let me show in somewhat more detail how I think the ontology of W.V.O. Quine
can be accommodated within this system. We have to introduce the following stipulations: W-
A!=Abs, where A! is the union of all A(t)'s, i.e., there are no non-actual entities (i.e. merely
possible, or impossible individuals, or objects of reference) save the abstract ones. Sgt(P)(u)(t)
{u,0}, i.e. predicates signify individuals, not their properties. Whence the extension of a
predicate: Ext(P):= {u: for some t, Sgt(P)(u)(t)=u}. And so a formula of the form Pa is true iff
Sp(a)Ext(P). The extension of this approach to many place-predicates is trivial as follows:
Sgt(Pn)(u1)...(un)(t){<u1,...,un>,0}, whence the extension of Pn: Ext(Pn):={<u1,...,un>: for some t,
Sgt(Pn)(u1)... (un)(t)=<u1,...,un>}. And so we get back the standard theory of predication.
As I have said, these are only some, rather ad hoc chosen examples of the possible
applications of this system. I have said nothing in particular concerning how modal, temporal and
intentional notions can be introduced, and how the exact reconstruction of these notions within
this system can contribute to the systematic study of concept formation, conceptual change, and
comparisons of different conceptual schemes in general. I brought up these illustrations only to
suggest that this system should not be regarded as the end, but rather as the beginning of an
ambitious project.

87
ON BEING AND ESSENCE IN SAINT THOMAS AQUINAS'S METAPHYSICS
AND PHILOSOPHY OF SCIENCE

In this paper I would like to present the outlines of a formal reconstruction of St.Thomas
Aquinas's concepts of being and essence, as they function in his metaphysics and philosophy of
science. This will necessitate the introduction of some formalism, nevertheless, I will try to keep
a certain balance between formal and informal presentation, so that we may steer our way safely
between the Scylla of empty concepts and the Charybdis of blind intuition.
Now, as is well known, "esse duobus modis dicitur. Uno modo secundum quod significat
veritatem propositionis, secundum quod est copula ... Alio modo dicitur esse, quod pertinet ad
naturam rei, secundum quod dividitur secundum decem genera". (3SN d.6.q.2.a.2.) 1 So the
copulative "est" signifies the truth of a proposition.
But what is it that makes a proposition true? Well, it is the actual existence of an
individualized form, or nature, signified by the predicate in the individual supposited for, i.e.,
referred to by the subject, at the time of the predication.2
For what individualizes a form in the first instance is the individual of which it is a form. Such
a form is what St.Thomas speaks of as "forma in supposito singulari existens per quod
individuatur".(ST1.13.9.)3 But the other individuating condition is time, for even if an individual
can have numerically the same form at different times, the form once emitted cannot recur
numerically the same, "quia quod omnino in nihilum decidit idem numero resumi non
potest".(4SN.22.1.1.)4
So predicates signify individualized forms, which are numerically different in different
individuals (except for the case of divinity, of course)5 and may be different in the same
individual at different times. But the same predicate in the same individual at the same time
cannot signify different forms. So we can speak of the significate of a predicate P in an
individual u at time t, which, therefore, can be denoted as the value of a function for these
arguments, like this:
Sgt(P)(u)(t)

1
Cf. 1SN 19.5.1.ad1., 33.1.1.ad1.; 2SN 34.1.1., 37.1.2.ad1. & ad3.; De Ente 1.; De Pot 7.2.ad1.; De Malo 1.1.ad19.; Quodl 9.2. 2.; in Meta 4.1.,
5.9., 6.2., 6.4., 9.11., 11.8.; ST1 3.4.ad2., 16.3.ad2.; 48.2.ad2.; ST1-2 36.1.; ScG 1.12., 1.58., 3.9. Cf. also: Thomas de Vio Cajetanus, "Super
Librum De Ente et Essentia Sancti Thomae Aquinatis", in: Opuscula Omnia, Bergomi, 1590., c.1. in princ.; C. Alamannus: Summa
Philosophiae, Paris, 1888, Tom.1. sect.II.5.1.; R.W. Schmidt S.J., The Domain of Logic according to Saint Thomas Aquinas, The Hague,
Martinus Nijhoff, 1966, Part II. ch.4. and Part III. ch.8.
2
Cf.: "Sicut cum dico, Socrates est homo, veritas huius enuntiationis causatur ex compositione formae humanae ad materiam individualem, per
quam Socrates est hic homo." in Meta 9.11. vide totum locum; cf. also Schmidt, op. cit., pp.212-214 and 224-226.
3
Cf. "Non enim oportet si hic est homo, et illud homo, quod eadem sit numero humanitas utriusque, sicut in duobus albis non est eadem albedo
numero..." 2SN 17.1.1.; cf. also e.g ST1 85.1. and 2.ad2.; Alamannus, op. cit., q.2.aa.1-3. Note that from the point of view of the present
reconstruction it makes no difference whether we speak of Socrates's humanity or of the humanity individualized by Socrates's matter, i.e., by the
materia signata that makes Socrates this individual. Indeed, these are one and the same form, the forma totius of Socrates. Cf. also the references
of the next note.
4
Cf. in Phys 5.6.; Quodl.4.3.2., 11.6.; Scg 4.80., 81.; 4SN 44.1.1.; Comp. Theol. 1.154.
5
Cf. e.g. ST1 39.2., 3.; 1SN 9.1.2.; De Pot 9.6. By the way, this approach offers a very good criterion of truth for relative identity statements, "a
is the same F as b", as opposed to absolute identity statements "a is identical with b". The former holds iff the significate of "F" in the
suppositum of "a" is identical with the significate of "F" in the suppositum of b. So "Filius est idem Deus cum Patre" is true, for "Deus" signifies
the same nature in the suppositum of "Filius" and of "Pater". But "Filius est idem cum Patre" is false, since the Son and the Father are distinct
supposita of this nature.

88
Now, it is the existence of such a form that accounts for the truth of a predication (namely, of
predicating P of u at t). "Ex hoc enim quod aliquid in rerum natura est sequitur veritas vel falsitas
in propositione, quam intellectus significat per hoc verbum est prout est verbalis copula."
(Met.5.9.)6
But then we can say that a sentence of the form `S est P' is true at time t according to a given
supposition, or acception of its subject term, iff the significate of P in the suppositum of S at time
t, at time t exists. It is the actual existence of this significate, an individualized form, that founds
the truth of this proposition.7
But from this it does not follow that this form is the significate of this proposition. For the
proposition involves also the copula, which signifies composition, an act of human reason, which
need not have a direct counterpart in reality. For the copula "est significans compositionem
cuiuslibet enuntiationis quam anima facit, unde hoc esse non est aliquid in rerum natura, sed
tantum in actu animae componentis et dividentis." (Quodl. 9.2.2.).8
So we can say that the significate of a proposition is the significate of the copula, an ens
rationis, which is in the second sense, if and only if the form signified by the predicate in the
suppositum of the subject is in the first sense.9 (In the case, of course, when the predicate is such
that it signifies some real form, not a privation, negation, or relation of reason. In these latter
cases also the significate of the predicate would be an ens rationis. (Cf. n.18. below.)
Now to give this idea a formal expression consider the following. First, let us suppose that
everything that can be signified by a predicate is either actual or not actual at a given time t.10 Let
us suppose further that everything which is actual is either a mere ens rationis or also an ens
reale.11 The significate of a predicate in an individual at time t , then, is an element of one of
these domains:
Sgt(P)(u)(t)W(t),
where W(t) is the set of all signifiable things that are either actual or non-actual at time t, so
that the set of actual things, A(t), is a subset of W(t), and the set of real beings at time t, Re(t), is
a subset of the set of rationate beings at time t, Ra(t), while Ra(t)=A(t).
Now this significate may be construed as the value of a function for the argument t. But then
the function itself can be derived from this significate by functional abstraction:12

6
Cf. 1SN 19.5.1.; 1SN 33.1.1.ad1.: "esse quod significat veritatem compositionis in propositionibus ... fundatur in esse rei"; Schmidt, op. cit.,
pp.232-237.
7
Cf. Schmidt, op. cit., pp.224-228; cf also H.Weidemann, "The Logic of Being in Thomas Aquinas", in: S. Knuuttila-J. Hintikka (eds.), The
Logic of Being, Dordrecht, Holland, 1986.
8
This point is brought out nicely by Weidemann, op. cit., sect.4.; cf. also Schmidt, op. cit., pp.238-239.
9
I would tentatively identify the significate of a proposition as the enuntiabile expressed by the proposition, expressly called by St. Thomas an
ens rationis in 1SN 41.1.5. I say "tentatively", because of St. Thomas's tendency to use the term enuntiabile as a synonym for enuntiatio
(although "emphasizing the objective meaning of enunciation" Schmidt, op. cit., p.223.n.84.). For St. Thomas's use of the term see 3SN 24.1.1b.;
1SN 38.1.3.; De Ver 2.13.ad7., 1.6., 14.8., 2.7., 1.5., 14.12.; Quodl 4.9.2.; ST1 14.14., 14.15.ad3., 16.7., ST3 1.2.ad2. For a clear expression of
the view that an enuntiabile is the significate of a proposition see e.g. L.M. de Rijk (ed.), Logica Modernorum, Assen, 1967, II-2, pp.208-213.
See also Peter of Spain, Tractatus, (ed. L.M. de Rijk), Assen, 1972, pp.205-207. Cf. also G. Nuchelmans, Theories of the Proposition - Ancient
and Medieval Conceptions of the Bearers of Truth and Falsity, Amsterdam-London, 1973, pp.165-194.
10
Cf. e.g. De Princ c.1. For a medieval-style resolution of the problems involved in referring to and quantifying over nonexistents see Essay III.
11
Cf. 2SN 34.1.1.
Concerning functional abstraction in general, see A. Church, Introduction to Mathematical Logic, Princeton, 1956, pp.15-23. Concerning -
12

abstraction in particular, see Essay V. of this volume.

89
t(Sgt(P)(u)(t)).
But this, again, can be regarded as a value of a function for the argument u. So again, we get
the function itself by applying -abstraction to u too; and let me call the result the signification
of P:13
Sg(P)=u(t(Sgt(P)(u)(t)))
Now let us suppose further that what the copula, the sign of composition, composes are this
function and its consecutive arguments, supplied by the suppositum of the subject term and by
the time of the predication.
So we can write:
Sgt(S est2 P)(t)(Sp)=Sgt(est2)(Sg(P))(Sp(S)(t))(t),
where Sp(S)(t){u: Sgt(S)(u)(t)A(t)}, if this set is not empty, otherwise Sp(S)(t)=0,14
and Sgt(est2)(V)(u)(t)W(t), where V(u)(t)W(t), and Sgt(est2)(V)(u)(t)A(t) iff V(u)(t)A(t).
(As can be seen, V here is the place-holder of the signification of the predicate, while u is the
place-holder of the suppositum of the subject.)
That is, what is signified by a proposition (at time t, according to a given supposition, or
acception of its subject term) is what is signified by the copula when it composes the nature
signified by the predicate term (according to its absolute consideration15) with a suppositum of its
subject (at time t) at time t. It is this composition of the intellect that answers the composition
found in the thing.16 And just as from the real composition in the thing there results an esse
reale, so from the composition of the intellect results an esse rationis answering the esse reale
which ultimately founds the truth of the proposition.17
And so we can say that a sentence is true if and only if what it signifies exists in the second
sense, i.e., it is an ens rationis, and that this is so if and only if the nature or form signified by the
predicate in the suppositum of the subject exists in the first sense, i.e., if it is an ens reale
(provided the predicate is such that it does not signify privation, negation or rationate relation) at
the time of the predication.18 In formulas:

13
Concerning the close parallelism between functional abstraction, on the one hand, and the traditional, Aristotelian conception of abstraction,
on the other, see P.T. Geach, "Form and Existence", in: God and the Soul, London, 1969, and my "Thomas Aquinas on the Meaning of Words",
Magyar Filozófiai Szemle, 1984/3-4, pp.298-313, (in Hungarian, with English abstract). As for the terminology used, of course, my use of the
terms "signification" and "significate" are not to be regarded as strictly corresponding to St. Thomas's use of "significatio" and "significatum".
What I call "signification" is most frequently referred to by St. Thomas as "forma significata (per praedicatum)", and what I call "significate" is
intended to be the same as St. Thomas's "forma vel natura individuata". Cf. e.g. ScG 4.49.; ST1 39.4.ad1.; ST1 16.2.; 1SN 25.1.4.; 3SN
7.1.1.ad5.; 1SN 4.1.2., etc.
14
As can be seen, 0 is the semantic value of empty terms. If we add the condition that for any predicate P non habens vim ampliandi it is not the
case that Sgt(P)(0)(t)A(t), and that an A proposition is true (at time t) iff its predicate is true of every suppositum of its subject (i.e., |Omne S
est2 P|t = T iff for every u{u: for some Sp, Sp(S)(t)=u}, Sgt(P)(u)(t)A(t), then all relations required by the Square of Opposition and all
syllogistical forms are saved. Cf. Appendix, and Essay II. of this volume.
15
Cf. "Patet ergo quod natura hominis absolute considerata abstrahit a quolibet esse, ita tamen, quod non fiat praecisio alicuius eorum. Et haec
natura sic considerata est quae praedicatur de omnibus individuis." De Ente c.4. cf. etiam commentum Cajetani ad hunc locum.
16
Cf. Schmidt, op.cit., pp.224-226.
17
"Tertio modo dicitur esse quod significat veritatem compositionis in propositionibus, secundum quod est dicitur copula: et secundum hoc est
in intellectu componente et dividente quantum ad sui complementum; sed fundatur is esse rei, quod est actus essentiae." 1SN 33.1.1.ad1. Cf.
Schmidt, op.cit., pp.215-222.
18
I think when the predicate signifies an ens rationis, then (and only then) we can identify the significate of the predicate with the significate of
the copula: if Sgt(P)(u)(t)Ra(t)-Re(t), then, and only then Sgt(est2)((Sg(P))(u)(t)=Sgt(P)(u)(t). In this way the esse of the significate of the
predicate will indeed consist in actu animae componentis et dividentis.

90
|S est2 P|t,Sp = T iff Sgt(S est2 P)(t)(Sp)Ra(t) iff Sgt(P)(Sp(S)(t))(t)Re(t)
/or Sgt(P)(Sp(S)(t))(t)Ra(t), provided the predicate signifies some rationate being./
But what about the case when the copulative est is used absolutely, without the addition of a
predicate term, when it answers the question an est?
Well, we may say that despite appearances this case is not so different from the former: for just
as in the former case the copula signified the existence of what the predicate signified in a
suppositum, so it signifies in this case the existence of the suppositum - the absence of the
predicate term means that it is not some determinate mode of existence that is attributed to the
suppositum, but existence simpliciter. So in our reconstruction we may suppose that in this case,
when there is no predicate term, what holds the place of the signification of the missing term is
an "empty", or "identical operation", i.e., a function which, somewhat loosely speaking, sends its
argument into itself:
Sgt(S est2)(t)(Sp)=Sgt(est2)(I)(Sp(S)(t))(t),
where I(u)(t)=u
Now from this we can derive that
|S est2|t,Sp = T iff Sp(S)(t)Ra(t),
that is, that the sentence "S est2" is true iff an S is a rationate being,19 i.e., iff what it signifies is
in the second sense.(Cf.the above equivalences.)
Now this reconstruction may perhaps gain further confirmation from the fact that if we take
into consideration St. Thomas's claim that a substantive name, as opposed to an adjective name,
can be taken for its suppositum even in predicate position,20 then from this reconstruction it
follows that deletion of a substantive predicate term does not affect the signification of a
proposition, thereby doing justice to the intuition behind the "ellipsis theories" of the copula.21
For if Sg(P)=I, i.e., if the predicate P is taken to signify not a form or nature in a thing really
distinct from the thing, but the thing itself, then from the above identities it follows that
Sgt(S est2 P)(t)(Sp)=Sgt(S est2)(t)(Sp).
So we can say that est in the second sense, whether it is used as a copula, i.e. as tertium
adiacens, or absolutely, as secundum adiacens, predicates existence in the second sense, i.e.

Of course, through |S est2|t,Sp = T iff Sgt(S est2)(t)(Sp)Ra(t). For, in general, for any proposition p: |p|t,Sp = T, iff Sgt(p)(t)(Sp)A(t). Cf.
19

M. Matthen, "Greek Ontology and the `Is' of Truth", Phronesis, 28(1983), pp.113-135. Whence, |p|t = T iff for some Sp: Sgt(p)(t)(Sp)A(t),
secundum regulam: indefinita equipollet particulari.
20
Cf. 3SN 5.3.3. I think I should briefly comment on St. Thomas's remark in this text that in this case the predication is a "praedicatio per
identitatem" as opposed to a "praedicatio per informationem sive denominationem" the latter being a "magis propria praedicatio" for "praedicata
tenentur formaliter". (Cf. e.g. in Meta 9.11.; ST3 16.7.ad4., 9.ad3.; ST1 13.12., 85.5.ad3.) Now this reconstruction as it st ands, of course,
favours the "inherence theory" as opposed to the "identity theory" of predication. (Cf. e.g. L.M. de Rijk's Introduction to his edition of
Abaelard's Dialectica, Assen, 1956, pp.37-38; D.P. Henry, Medieval Logic and Metaphysics, London, 1972, pp.55-56, P.T. Geach,
"Nominalism", in his: God and the Soul, London, 1969) in that it assigns to the predicate the semantic function of signifying inherent forms
through its abstract signification. However, this does not preclude the fact that even in this reconstruction a "proper predication" is always
equivalent to an identity statement: `S est2 P' <=> `S = P', if Sgt(=)(u1)(u2)(t)A(t) iff u1=u2A(t), whence Sgt(S =
P)(t)(Sp)=Sgt(=)(Sp(S)(t))(Sp(P)(t))(t) Now, since supposition is defined in terms of the actual inherence of the form signified by the term, this
reconstruction expressly shows that a predication is true, iff its terms supposit for the same thing, i.e., iff the forms signified by its terms inhere
in their common suppositum. Furthermore, if we take the predicate (in St. Thomas's, but not, e.g. Ockham's and his followers' view, improperly)
to stand immediately for its suppositum, i.e., if we identify its significates with its supposita, then we can identify also the significate of a
predication with that of the corresponding identity statement. So in this case taking the copula in the sense of identity will not affect even the
sense of the proposition, thereby doing full justice also to the identity theory of the copula.
21
Cf. e.g. R.M. Dancy, "Aristotle and Existence", in: S. Knuuttila-J.Hintikka: The Logic of Being, Dordrecht, 1986, and A. Kenny, The Five
Ways, London, 1969, pp.91-95.

91
existence proper to rationate beings.22 But this kind of existence is founded on existence in the
first sense, proper to real beings, which is signified by est in the first sense:
|S est1|t,Sp = T iff Sgt(est1)(Sp(S)(t))A(t),
where Sgt(est1)(u)(t)A(t) iff uRe(t).
But then
|S est1|t,Sp = T iff Sp(S)(t)Re(t),
while
|S est2|t,Sp = T iff Sp(S)(t)Ra(t).
So esse primarily signifies actual existence (as its "focal meaning", to use Professor Owen's
expression). But this signification is extended also to rationate beings, which also exist in a
secondary, derivative sense, owing this derivative existence to the real existence (or even
nonexistence, in the case of privations) of real beings, whether these are subsistent individuals or
real forms inhering therein.23
For this real existence is attributed to something in a twofold manner:
"Uno modo sicut ei quod proprie et vere habet esse, vel est...Omnia vero quae non per se
subsistunt, sed in alio, vel cum alio, sive sint accidentia, sive formae substantiales, aut quaelibet
partes, non habent esse ita ut ipsa vere sint, sed attribuitur eis esse alio modo, idest ut quo aliquid
est. Sicut albedo dicitur esse, non quia ipsa in se subsistat, sed quia ea aliquid habet esse album."
(Quodl.9.2.2.)
Now, since any individual substance has only one esse substantiale,24 therefore a form is
substantial to an individual iff the esse of this form (ut quo quid est) is identical with the esse of
the individual (quod est). But then, further, a predicate P is essential, or substantial to an
individual u, iff it signifies such a form in u:
(P)Subst(u) iff Sgt(est1)(u)(t)=Sgt(est1)(Sgt(P)(u)(t))(t)
And from this, besides, we can derive that if P is essential to u, then it is necessary that if u
exists, then it is P.25
Now, as the significate of a substantial predicate is a substantial form, St.Thomas's thesis of
the unity of substantial form can be expressed as follows:26
If G and S are substantial predicates of u, then
Sgt(S)(u)(t)=Sgt(G)(u)(t).
On the other hand, since a species term signifies the quiddity, or essence of a thing, therefore
the thesis of the real distinction of essence and existence in the creatures may be expressed as
follows:27
If S is a species term and u belongs to S, then it is not the case that
Sgt(S)(u)(t) = Sgt(est1)(u)(t).
22
This is why (pace Schmidt, p.235.) est means the same in "Caecitas est" and "Aliquid est caecum", or even in "Deus est" when this is an
answer to the question "An Deus est?". Cf. texts referred to in n.1.
23
Cf. in Meta 4.1.
24
"Impossibile est enim, quod unum aliquid habeat duo esse substantialia" 3SN 6.2.2.
25
For such a P can become false of u only if its significate in u ceases to exist, i.e., if the esse of this significate ceases to be actual. But as this
esse is identical with the esse of u, P can become false of u only if u ceases to exist. So u can be not P only if it does not exist. Whence, if u
exists, it must be P. Cf. Porphyry's definition of accident. Also Essay III. in this volume.
26
Cf. e.g. De Ente c.3.
27
Cf. e.g. De Ente c.5.

92
Now this is how esse and essentia are found in the individuals; and it is by abstracting from
these that we arrive at the cognition of universals, a class of rationate beings.28
But then, if the significate of the predicate P in an individual u at time t is a natura
individuata of P-ness, then what we get from this by abstraction is the nature of P as considered
absolutely, without any individuating conditions:
Nat(P) /=Sg(P)/ = u(t(Sgt(P)(u)(t))).29
Now consider the following passage from the De Ente et Essentia: "Haec autem natura habet
duplex esse, unum in singularibus, aliud in anima, et secundum utrumque consequuntur dictam
naturam accidentia..." namely, for example: "ratio speciei accidit naturae humanae secundum
illud esse quod habet in intellectu. ... Et quamvis haec natura intellecta habet rationem universalis
secundum quod comparatur ad res quae sunt extra animam, quia est una similitudo omnium,
tamen secundum quod habet esse in hoc intellectu vel in illo est species quaedam intellecta
particularis." (De Ente c.4.)
From this and related remarks we may form the following picture: by concretion, the inverse
operation of abstraction as presented above, we can go as it were, in two directions: either ad
extra, and then we arrive at the real individualized natures of individuals, or ad animam, and
then we arrive at the universal intentions of particular minds.30 But these intentions are universal
only insofar as they are got by abstraction carried out by these particular minds, from
representations of individuals, namely from phantasms.
Now from these mental representations (universal intentions and phantasms), in a similar
manner as we could construct the significates of propositions ad extra, we can construct their
significates apud mentem.31 Correctness of belief, then, consists in the adaequatio of these two
kinds of significates. But this correctness is based on the evident truth of first principles, which,
in turn, owe their evidence to induction, based on correct, essential, abstraction: for "sensus est
quodammodo et ipsius universalis. Cognoscit enim Calliam, non solum inquantum est Callias,
sed etiam inquantum est hic homo, et similiter Sortem, inquantum est hic homo. Et inde est, quod
tali acceptione in sensu praeexistente, anima intellectiva potest considerare hominem in utroque.
... Sic enim, scilicet per viam inductionis sensus facit universale intus in anima, inquantum
considerantur omnia singularia." (An.Post.2.20)32
Nevertheless, I have no space here to present the relevant reconstructions. All I hope to have
shown in this paper is that by the semantic approach presented here, St.Thomas's thought, from
what has become in the eyes of many philosophers of our days a museum piece, at best, can be
turned into something very substantial, and highly relevant even to our modern ways of doing
philosophy.

28
Some "basic texts" for St. Thomas's theory of abstraction are the following: in De Anima 3.8-12.; ST1 13.2.ad1., 85.1.; in De Trin 3.5.3.; cf.
also Alamannus, op.cit., tom.1.sect.1.q.2.aa.1-3.; Schmidt, op. cit. pp.177-202.
29
For more on this see Essay VII. Concerning the connection between abstraction and reduplicative constructions extensively used by St.
Thomas in this context see J. Lear, "Aristotle's Philosophy of Mathematics", The Philosophical Review, April, 1982, pp.161-169. Concerning
13th century treatment of reduplicative propositions in general, see my "Libellus pro Sapiente - a Criticism of Allan Back's Argument against St.
Thomas Aquinas's Theory of the Incarnation", The New Scholasticism, 58(1984), pp.207-219.
30
Cf. Schmidt, op. cit., pp.98-130, 212-215.
31
See Essay V.
32
Cf. in Meta 1.1.; in Anal Post 2.20.; Schmidt, op. cit., pp.270-302.

93
APPENDIX

In this Appendix I supply a brief, exact description of the formal theory outlined in the body of
the paper, lest any technically obscure point should remain.
The language of the theory, the language of categorical propositions, is defined as follows:
L=<C,Pr,F>,
where C={est1,est2,-,Q,=}, (where - is the sign for negation and Q is `Omne', `Quidam', or their
equivalents), Pr is a set of predicate parameters (S,P, etc.) and F is the set of formulae, or
sentences defined by the following clauses:
(i) If S,PPr, then `S est2 P', `Q S est2 P', `S est1', `S = P'F
(ii) If pF, then `-p'F
A model for this language is defined as follows:
M:=<W,T,<,A,Re,Ra,0,Sg>,
where W(t) is a nonempty set, T is a set of time-points ordered by <, A(t) is a part of W(t),
Ra(t)=A(t), Re(t) is a part of Ra(t), where tT. Intuitively, W(t) is the set of all signifiable
things which are either actual or not actual at time t, A(t) is the set of actual things at time t,
Ra(t) is the set of rationate beings, and Re(t) is the set of real beings at time t. 0, the zero-
entity, is the semantic value of empty terms, which falls outside the whole universe of discourse
W!, i.e. 0 is not an element of W!, where W! is the union of all W(t)'s: W!:= UtT W(t). Sg, the
signification function is defined for any expression (primitive as well as complex) as follows:
Sg(exp):= e1(...en(Sgt(exp)(e1)...(en))...),
where Sgt(exp)(e1)...(en) is the significate of any expression in respect of any entities, e1...en,
whatever (including elements of W!U{0}, T, L and functions defined on these).
NOTE that I use the -operator according to the following equivalences:
if f and g are functions, then
f(x)(y)=g(y) iff f(x)=g iff y(f(x)(y))=g
Now Sgt(exp)(e1)...(en) is defined by the following clauses (if not otherwise indicated it is
supposed throughout that uW and tT):
(1) Sgt(P)(u)(t)W(t)
(2) It is not the case that Sgt(P)(0)(t)A(t)
(3) Sgt(est2)(V)(u)(t)W(t), where (V)(u)(t)W(t) and Sgt(est2)(V)(u)(t)A(t) iff
(V)(u)(t)A(t)
(Intuitively, V is the place-holder of the signification of the predicate, while u is the place-
holder of the suppositum of the subject term. Cf. (8) below.)
(4) Sgt(est1)(u)(t)W(t) and Sgt(est1)(u)(t)A(t) iff uRe(t)
(5) Sgt(=)(u1)(u2)(t)W(t) and Sgt(=)(u1)(u2)(t)A(t) iff u1=u2A(t)
(6) Sgt(-)(u)W(t) and Sgt(-)(u)A(t) iff not: Sgt(-)(u)A(t)

94
(7) Sgt(Q)(N*)(E(V))(u)(t)W(t) and Sgt(Q)(N*)(E(V))(u)(t)A(t) iff for Q'uN*,
E(V)(u)(t)A(t),
where Q' is the English equivalent of Q, E(V)(u)(t)W(t), N(u)(t)W(t), and
N*:={u:N(u)(t)A(t)}, if {u: N(u)(t)A(t)} is not empty, otherwise N*:={0}.
(Intuitively, N and E(V) are the place-holders of the signification functions of the NP and the
VP of the quantified statement, E is the place-holder of the signification of the copula, while N*
is the place-holder of the range of values of the NP, i.e., of the subject of the quantified
statement. Cf. (4) above and (12) below.)
These were the clauses for the signification of the primitive expressions of L. Now here
follow the clauses for complex expressions:
(8) Sgt(S est2 P)(t)(Sp)=Sgt(est2)(Sg(P))(Sp(S)(t))(t),
where Sp(S)(t){u:Sgt(S)(u)(t)A(t)} if this set is not empty, otherwise Sp(S)(t)=0.
(9) Sgt(S est2)(t)(Sp)=Sgt(est2)(I)(Sp(S)(t))(t),
where I(u)(t)=u
(10) Sgt(S est1)(t)(Sp)=Sgt(est1)(Sp(S)(t))(t)
(11) Sgt(S = P)(t)(Sp)=Sgt(=)(Sp(S)(t))(Sp(P)(t))(t)
(12) Sgt(Q S est2 P)(t)(Sp)=Sgt(Q)(Sp(S)(t)*)(Sg(est2)(Sg(P)))(t),
where Sp(S)(t)*:={uW!U{0}: for some Sp, Sp(S)(t)=u}
and Sg(est2)(Sg(P)):=u(t(Sgt(est2)(Sg(P))(u)(t))).
(13) Sgt(-p)(t)(Sp)=Sgt(-)(Sgt(p)(t)(Sp))
Now, the definition of truth for any formula p, at time t, according to a given supposition, or
acception of its terms is the following:
|p|t,Sp = T iff Sgt(p)(t)(Sp)A(t),
whence the definition of truth at time t:
|p|t = T iff for some Sp, |p|t,Sp = T.
Now, if we define:
`Quoddam S non est2 P' df `- Omne S est2 P' and
`Nullum S est2 P' df `- Quoddam S est2 P', then all relations required by the Square of
Opposition among the four categoricals are provably valid in this system.

95
SOCRATES EST SPECIES

- Logic, Metaphysics and Psychology in Saint Thomas Aquinas's Treatment of a Paralogism -

1. INTRODUCTION: THE PROBLEM

The aim of this paper is to explore the logical, metaphysical and psychological theories
grounding Saint Thomas Aquinas's treatment of the following paralogism:
(I) Socrates est homo
Homo est species

Socrates est species
The problem in connection with this paralogism is caused by its apparent similarity to a
number of valid inferences. For example:
(II) Socrates est homo
Homo est animal

Socrates est animal
To be sure, this inference is not a formally valid syllogism either, as an inference of this form
does not hold in all terms. For example,
(III) Socrates est homo
Homo est albus

Socrates est albus
is an invalid inference. For, since Homo est albus is a per accidens predication, not every
man need be white.1
But (II) is a valid inference by reason of the meaning of its terms. For, since Homo est
substantia animata sensibilis is a per se predication, anything falling under the term homo has
also to fall under the predicate of which substantia animata sensibilis is the definition, namely
animal.
But then, pari ratione, we seem to have to accept that since Homo est praedicabile de
pluribus solo numero differentibus is a per se predication, for homo by its very meaning can
be predicated of Socrates and Plato, who differ only numerically,2 anything falling under the term

1
Cf. what St. Thomas says about the interpretation of indefinite propositions: in Peri 1.10; cf. also his De Fallaciis, cc.10-12, esp. c.12.
2
Cf.: "Et etiam probatur per definitionem speciei, quia homo praedicatur de pluribus scilicet Sorte et Platone numero differentibus." Buridanus,
Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, p.49. Cf. also: "Quod per se praedicatur de speciebus convenit ipsis speciebus: sed ly
Genus, per se praedicatur de speciebus, puta homine et bove. Ergo ly Genus, convenit ipsis speciebus: et sic homo erit genus et similiter bos.
Maior est ex terminis nota, implicat enim, quod aliquid per se praedicatur de aliquo in quocunque modo perseitatis et non conveniat illi. Minor
probatur sic. Illud in cuius diffinitione ponitur praedicari de speciebus, per se praedicatur de illis: sed ly Genus, est in cuius diffinitione ponitur
praedicari de speciebus: ergo genus per se praedicatur de speciebus. Maior syllogismi huius patet ex eo quod quaelibet pars diffinitionis in recto
primo modo perseitatis convenit diffinito. Maior autem in Porphyrio patet sic dicente, Genus est praedicabile de pluribus specie differentibus.
Similiter omnino potest argui de specie respectu individui, cum etiam in diffinitione speciei ponitur praedicari de individuis, et sic videtur quod
male dictum sit Socratem non esse speciem." Thomas de Vio Cajetanus, "Super Librum De Ente et Essentia Sancti Thomae Aquinatis", in:
Opuscula Omnia, Bergomi, 1590, p.332.

96
homo should also fall under the term of which praedicabile de pluribus solo numero
differentibus is the definition, namely under the term species. However, its conclusion being
false while its premises true, (I) cannot be a valid inference.
Now in the next section I will present an easy solution to this problem in the framework of
supposition theory, namely the solution of Buridan. In the third section I will attack this solution
by putting forward a general objection against the theoretical framework of Buridan's solution,
namely "nominalist" supposition theory coupled with the identity theory of predication. In the
fourth section I will sketch the alternative medieval predication theory, the inherence theory of
predication. In the fifth section I show how our problem reemerges in the framework of this
theory. I also present here a solution in this framework, which, however, will immediately be
attacked. These considerations will quite naturally lead us to St. Thomas's treatment, and to a
reconstruction of its logical, metaphysical and psychological background in his thought. The final
section of the paper will conclude with some general lessons to be drawn from the reconstruction
concerning the foundations of inferences.

2. BURIDAN'S SOLUTION

According to Buridan's solution of the above problem, inference (I) is invalid because the term
homo stands in different suppositions in its premises. In particular, in Socrates est homo, the
term homo stands in personal supposition, i.e., it refers to singular men3; on the other hand, in
Homo est species the same term, insofar as the proposition is true, stands in material
supposition, i.e., it refers to a concept or intention of the human mind.4 Now on Buridan's view,
what verifies an affirmative categorical proposition is the identity of the supposita of its terms.5
Correspondingly, what validates inference (II) is the transitivity of identity:6 for, since Socrates is
identical with a man, and a man, indeed, any man, and consequently even the man who is
Socrates, is identical with an animal, therefore Socrates is identical with an animal. Indeed, the
same rule validates the "expository form" of the otherwise invalid (III): for if a man is white, and
Socrates is that very same man, then Socrates must be white.7 On the other hand, since in (I) the
same middle term must refer to different things in the two premises (as no flesh and blood man
can be an intention of the mind), therefore transitivity of identity is inapplicable to it, wherefore
(I) must be invalid.

3
Cf. Buridanus, Tractatus de Suppositionibus, ed. Maria Elena Reina, Rivista Critica di Storia della Filosofia, 1957, pp.175-208, and pp.323-
352, p.323.
4
Cf. Buridanus, Tractatus de Suppositionibus, ed. cit., pp.201-205. Cf. also S. Ebbesen, "The Summulae, Tractatus VII De Fallaciis", in: The
Logic of John Buridan, Copenhagen: Museum Tusculanum, 1976.
5
Cf. Buridanus, Sophismata, ed. cit., pp.42-43. Decima conclusio
6
Cf.: "... syllogismi affirmativi tenent virtute istius principii `quaecumque uni et eidem sunt eadem inter se sunt eadem'. Unde ex eo quod
extremitates designantur in praemissis dici eadem uni medio concluduntur in conclusione dici eadem inter se." Buridanus, Tractatus de
Consequentiis, ed. H. Hubien, Louvain-Paris, 1976, p.84.
7
Cf.: "nullus syllogismus valet in quo medium in neutra praemissarum est distributum nisi in maiori propositione sumatur medium cum relativo
identitatis. Quia regulae per quas tenent syllogismi requirunt, si medium sit commune, quod extremitates coniungantur ei ratione eiusdem rei pro
qua iste terminus communis supponit, ut ante dicebatur. Et cum medium in neutra sit distributum possibile est quod coniunctio eius cum maiori
extremitate sit vera pro uno et coniunctio eius cum minori sit vera pro alio; ideo per hoc nulla potest inferri coniunctio extremitatum inter se nisi
per relativum identitatis medium cogatur teneri pro eodem in minori propositione pro quo erat eius verificatio in maiori. Sed tunc valet
syllogismus, et tenet per regulas supradictas, et est tanquam syllogismus expositorius; verbi gratia, `B est A et C est illud idem B; ergo C est A'.
Tenent ergo tales syllogismi in omnibus modis in quibus tenent syllogismi expositorii." Buridanus, ibid. pp.89-90.

97
3. ATTACKING BURIDAN'S FRAMEWORK

This solution as it stands may indeed be very appealing by its simplicity, and intuitive
character. However, the theoretical framework in which it is formulated is not immune to
criticism.
Despite possible expectations, nevertheless, I do not want to repeat here Professor Geach's
well-known arguments against the so-called "two-name" theory of predication committing the
sacrilege of treating general terms as referring expressions, horribile auditu, even in predicate
position, for those arguments are quite easily answerable (indeed, for the most part they have
been answered by G. Evans), and to my mind do not touch the heart of Buridan's framework.8 So,
instead, let me reformulate a general objection made by Pierre d'Ailly, which, I think, hits upon a
really weak point in nominalist supposition and predication theory.
The objection runs as follows: "Second Thesis: whether it is affirmative or negative, a
proposition is not therefore true or false because its subject and predicate supposit for the same
thing or they do not supposit for the same thing. Proof: Those who define `true proposition' or
`false proposition' in terms of suppositing or not suppositing for the same thing...also give a
definition of supposition in which it is said that supposition `is the taking of a term in a
proposition for its significatum or significata of which this term is verified by means of the
copula of the proposition in which it is posited'. Thus, they define supposition in terms of
verification of a proposition. Therefore, they should not, conversely, define the verification of a
proposition in terms of supposition or of suppositing or not suppositing for the same thing. For
one ought not to give circular definitions or define the same by the same..."9
Put in modern terms, the objection shows that on this theory the computation of the semantic
values of terms and propositions in a model would have to go on infinitely: for in order to know
whether an indefinite affirmative proposition according to a given supposition of its terms is true,
first we have to know whether the things supposited for by its terms in this supposition are the
same. However, in order to know this, first we have to know which are the individuals
supposited for by these terms, i.e., we have to know whether, say, this individual is supposited
for by this term or not. But, we are told, we can know this by checking if the term is true of this
individual; but, again, we can know this only if we know whether this individual is supposited
for by this term, which, again, can be known only by checking whether the term is true of it, and
so on infinitely.10 So unless we find independent criteria for determining the semantic values of
terms and propositions, we cannot be satisfied with the semantic framework of Buridan's solution
to our original problem. Hence, what we have to find is independent criteria for determining the
supposition of terms (whether given such and such conditions in rebus this term supposits for
8
See e.g. P.T. Geach, Reference and Generality, Ithaca, 1964, pp.115-122; Logic Matters, Berkeley and Los Angeles, 1972, pp.44-62, pp.95-
108, pp.289-302. G. Evans, "Pronouns, Quantifiers and Relative Clauses" and "Appendix", Canadian Journal of Philosophy, 1977, pp.467-536,
pp.777-796. See also Essay III. of this volume.
9
Peter of Ailly: Insolubilia, Paris, 1492, quoted and translated by Marilyn McCord Adams in her "What Ockham Means by Supposition?",
Notre Dame Journal of Formal Logic, 17(1976), p.379. Cf. also: "est autem suppositio prout hic accipitur acceptio termini in propositione pro
aliquo vel pro aliquibus quo demonstrato vel quibus demonstratis per ista pronomina `hoc' vel `haec' vel equipollentia illis, terminus vere
affirmatur de isto pronomine mediante copula illius propositionis" Buridanus, Sophismata, ed. T.K. Scott, Stuttgard-Bad Cannstatt, 1977, p.50,
and "ad veritatem cathegoricae affirmativae requiritur quod termini, scilicet subiectum et predicatum, supponant pro eodem vel pro eisdem",
ibid. p.42.
10
For a formal reconstruction of this objection see Essay II. of this volume.

98
this individual or not), on the one hand, and verification of propositions (whether given such and
such conditions in rebus this proposition is true or not), on the other.

4. THE INHERENCE THEORY OF PREDICATION

Now another medieval semantic framework, which seems to satisfy this requirement, is the
inherence theory of predication. For in this framework the verification of a proposition is not
based on the identity of the supposita of its terms, but on the actuality of the form signified by its
predicate in the supposita of its subject; these supposita, in turn, being those things in which the
form signified by the subject is actual.11 As can be seen, within this framework the above vicious
circle breaks down as follows: in order to know whether a given indefinite affirmative
proposition is true we have only to know whether the form signified by the predicate in an
individual supposited for by the subject is actual; and whether this or that individual is
supposited for by the subject can be known by checking whether the form signified by the subject
in this individual is actual, and this is the end of it.

5. THE REEMERGENCE OF THE PROBLEM

On the other hand, this framework gives rise to the original problem in a quite acute manner.
For "inherentist" logicians, like William of Sherwood or Peter of Spain held that since according
to their theory the predicate signifies only a form, not signifying the thing which bears this form,
the predicate term of a categorical stands in simple supposition.12 On the other hand, in Homo
est species, the term homo also stands in simple supposition, therefore, as its medium term now
refers indeed to the same thing in both of its premises, (I) seems to be a valid inference.13
Now to this difficulty the inherentist can immediately give an answer like the following: I
concede that the term homo refers indeed to the same thing, namely to the universal nature of
men in both of these premises, once as a predicate, once as a subject; still, this does not imply
that the conclusion follows from these premises. For even if identity is, as you call it, a transitive
relation, still, inherence need not be so. For what validates a valid inference of the following
form: `S est M', `M est P', therefore `S est P' is not the inherence of the universal nature signified
by P in the universal nature signified by M, which in turn inheres in the supposita of S; but, on
the contrary, it is the inherence of the nature signified by P in the personal supposita of M,
which, by virtue of the first premise, are also supposita of S.14 So inference (I) fails precisely
because in Homo est species the term homo stands in simple supposition, for the universal

11
Concerning the inherence theory in general, as opposed to the identity theory see L.M. de Rijk's Introduction to his edition of Abaelard's
Dialectica, Assen, 1956, pp.37-38; also D.P. Henry, Medieval Logic and Metaphysics, London, 1972, pp.55-56. Concerning St. Thomas's
inherence theory in particular see H.Weidemann, "The Logic of Being in Thomas Aquinas", in: S. Knuuttila-J. Hintikka (eds.), The Logic of
Being, Dordrecht, Holland, 1986; R.W. Schmidt S.J., The Domain of Logic according to Saint Thomas Aquinas, The Hague, Martinus Nijhoff,
1966; cf. also Essay VI. of this volume.
12
See William of Sherwood, Introductiones in Logicam, ed. M. Grabmann, Sitzungsberichte der Bayerischen Akademie der Wissenschaften,
1937, 10, pp.78. Quaestio; Peter of Spain, Tractatus, (ed. L.M. de Rijk), Assen, 1972, pp.83-88.
13
Cf. William of Sherwood, op. cit., pp.75-77.; Peter of Spain, op. cit., p.81.
14
Cf. Peter of Spain, op. cit., p.148.

99
nature of men, and not in personal supposition, which, however, would be required for the
validity of the inference.
Now this in fact seems to be a rather satisfactory solution, which, however, leaves open the
way to a further objection. The objection may be formulated as follows: the inherentist says that a
predicate term stands in simple supposition, which means that it stands for the universal nature
signified by it, that is, for a nature as such. However, what is true of a nature as such belongs to
the essence of the individuals which have this nature. Therefore, what is predicated of this nature
as such is true, indeed, essentially true of the individuals which have this nature. But it is said
that in Homo est species the subject term again stands in simple supposition, that is, it stands for
human nature as such. So what is truly predicated of it should also be true of what has this nature,
namely, by the first premise, of Socrates. So the previous solution notwithstanding, by this
reasoning, our problematic inference seems to be valid again.15

6. ST. THOMAS'S SOLUTION

Well, it is at this point that I think we can properly appreciate St. Thomas's well-known
threefold distinction between the different considerations of a nature, which allows us to dismiss
this difficulty.16
For according to St.Thomas a nature can be considered in three different ways: first, a nature
can be considered as it exists in the individuals, individuated and multiplied according to the
multiplicity of the individuals which have this nature; second, it can be considered absolutely, as
abstracted from all individuating conditions and from any kind of particular existence; third, it
can be considered as existing in the soul, again, individuated and multiplied according to the
multiplicity of human souls, however, bearing some striking resemblance to the absolutely
considered nature in that even if it is individuated with regard to its subject, the human soul, still,
it is abstract as regards its objects, the things which have this nature. It is precisely by this last-
mentioned feature of this nature that it is related to many individuals, being an essential, common
resemblance of these all.
Now in his De Ente et Essentia, when St. Thomas examines the relationship between essence
or nature and the logical intentions, he concludes that it is only nature existing in the soul that
can be a species. So according to him, in Homo est species, the term homo stands for human
nature as existing in the soul. So even if in Socrates est homo the same term stands for human
nature as considered absolutely, and whatever is true of a nature according to its absolute
consideration is also true of the individuals having this nature, still, as in Homo est species the
term homo does not stand for human nature according to its absolute consideration, therefore the
conclusion Socrates est species does not follow from these premises.17

15
Cf. St. Thomas Aquinas, De Ente, c.4. in fine; cf. also Cajetan's commentary, ed. cit., pp.331-332.
16
Cf. Quodl 8.1.1.; De Ente c.4.; De Pot 9.9.ad2., 9.1.; In de Trin 5.2. For St. Thomas's reliance on Avicenna in this doctrine see M.-D. Roland-
Gosselin, Le `De Ente et Essentia' de Saint Thomas d'Aquin, Kain, 1926, p.24., n.1.; cf. also C. Alamannus: Summa Philosophiae, , Paris, 1888,
Tom.1. sect.I. q.2.aa.1-3.
17
"Et quia naturae humanae secundum suam absolutam considerationem convenit quod praedicetur de Socrate, et ratio speciei non convenit sibi
secundum suam absolutam considerationem, sed est de accidentibus quae consequuntur eam secundum esse quod habet in intellectu, ideo
nomen speciei non praedicatur de Socrate, ut dicatur: Socrates est species; quod de necessitate accideret, si ratio speciei conveniret homini

100
7. RECONSTRUCTING ST. THOMAS'S FRAMEWORK

So far, so good. But do we really understand St. Thomas's solution? Do we really understand
what the members of St.Thomas's distinction are? What is the relationship between human nature
as such and human nature as existing in Socrates? Can there be anything like human nature as
such, if there is nothing which is not individuated?18 And if human nature as such is only an
abstraction of human mind, then what is it that distinguishes human nature as such from human
nature as existing in the mind? Indeed, if it is human nature as such which is predicated of the
individuals, as St. Thomas expressly states19, then why is it human nature as existing in the soul
rather than human nature as such which is a species, if a species is what is predicated of many,
only numerically different things? Finally, if it is indeed human nature as existing in the soul that
is a species, and it is the like intentions of the soul that logic studies20, then what is it, if anything,
that distinguishes logic from psychology, or, in other words, what is it that gives logic its
objectivity, what is it that renders logical laws cogent intersubjectively, indeed, more cogent than
natural laws?
In order to answer these questions first we should understand the questions themselves. So lest
we rely on vague intuitions, first we have to clarify the meaning of the basic terms used,
especially of the term `nature' as it functions in St. Thomas's logic, metaphysics and psychology.
As nowadays the best available tool for the exact reconstruction of meaning is model theoretic
semantics, let me therefore outline here a semantic reconstruction of St. Thomas's concept of
nature, or essence in the context of his logic, metaphysics and psychology.
To begin with, a common nature is what is signified by a common term. But a common nature
as such does not exist in reality, it is only numerically distinct individual natures of individual
things that there are in rebus. It is the actuality of such an individualized nature that verifies the
term of the thing which has this nature. So an individualized nature is what is signified by a term
in an individual at a certain time.21 It is by abstracting from these individuating conditions,
namely time and subject, that the intellect forms in itself the universal nature signified by the
term absolutely, without signifying these individuating conditions, though, at the same time, also
without the exclusion of these conditions.22
Now since a general term can signify numerically different individualized natures in different
individuals and at different times23, it can nonetheless signify only one individualized nature in a
certain thing at a certain time, therefore we can denote this nature as the value of a function for
these arguments as follows: Sgt(P)(u)(t) (read: the significate of the term P in the thing u at time

secundum esse quod habet in Socrate, vel secundum suam absolutam considerationem, scilicet inquantum est homo: quidquid enim convenit
homini inquantum est homo praedicatur de Socrate." De Ente c.4.; cf. in De Anima 2.12.; 1SN d.19.q.5.
18
For a good exposition of St. Thomas's doctrine on universals, along with ample references see Alamannus, op. cit., q.2.a.4. See also Schmidt,
op. cit., pp.175-194.
19
"Patet ergo quod natura hominis absolute considerata abstrahit a quolibet esse, ita tamen quod non fiat praecisio alicuius eorum. Et haec
natura sic considerata est quae praedicatur de omnibus individuis." De Ent c.4.; cf. commentum Cajetani ad hunc locum.
20
See Alamannus, op. cit., q.1.a.6.; cf. also Schmidt, op. cit.
21
Cf. 2SN 17.1.1.; st1 85.1., 2.ad2.; In Phys 5.6.; Quodl 4.3.2., 11.6.; ScG 4.80., 81.; 4SN 44.1.1.; Comp. Theol. 1.154.
22
The "basic texts" for St. Thomas's theory of abstraction are the following: in De Anima 3.8-12.; st1 13.12.ad1.; ST1 85.1.; in De Trin 3.5.3.;
cf also Alamannus, op. cit., 2.1-3.; Schmidt, pp.177-202.
23
The expression "individualized nature", of course, is to be understood broadly here, in sense of "individualized property". Cf. texts referred to
in n.21., esp. in phys 5.6. See also Peter of Spain, op. cit., pp.86-88.

101
t).24 It is the actuality of this significate at time t that verifies the term P of u. So Sgt(P)(u)(t) is to
be an element of the domain of all signifiable things which are either actual or nonactual at a
certain time t, say W(t). And so the set of things which are actual at time t, say A(t), is a subset of
W(t). Of course the set of all signifiable things which are actual or non-actual at any time t, W!,
is the union of all of these W(t)'s.25
Now if we want to denote the general nature signified by P absolutely, without its
individuating conditions, but, at the same time also without the exclusion of these, then we
should abstract from the arguments of the above function so that we get the function itself,
without its arguments, but also without the exclusion of these as follows:
u(t(Sgt(P)(u)(t)))
And let me call the result the signification of P: Sg(P).26 This is what is signified by the
predicate term of a categorical proposition. For the copula of an affirmative categorical
proposition signifies the composition of the nature signified by the predicate with the thing
supposited for by the subject. So in this reconstruction the significate of a categorical of the form
`S est P', according to a given supposition of its subject at a certain time t, can be constructed as
follows:
Sgt(S est P)(t)(Sp)=Sgt(est)(Sg(P))(Sp(S)(t))(t),
where Sgt(est)(V)(u)(t) is an element of A(t) iff V(u)(t) is an element of A(t), and Sp(S)(t) is
an element of: {uW!: Sgt(P)(u)(t)A(t)}, if this set is not empty, otherwise Sp(S)(t)=0, where
V(u)(t)W(t), 0 is not an element of W! and Sgt(P)(0)(t) is not an element of A(t). Intuitively, V
is the place-holder of the signification function of the predicate term; Sp(S)(t) is a suppositum of
S at time t and 0 is the zero-entity, the semantic value of empty terms.27
As we can see, in an affirmative proposition, by the converse operation of abstraction,
concretion, as it is called, we arrive again at the concrete, individualized natures of individuals
signified by the predicate in these individuals.28 And this is done by filling in the "gaps", the
empty argument-places in the absolute nature signified by the predicate, which give symbolic
expression to the non-exclusive character of this nature.
But then, if by filling a thing in the empty argument-place of this function we get the nature
signified by the predicate as existing in the thing (if it exists at all ),29 then what we get by filling

24
My use of the terms "significate" and "signification" is not intended to correspond strictly to the (by no means consistent) medieval use of the
terms "significatum" and "significatio", respectively. On the other hand, I do intend to express by this terminology the difference between (in the
Fregean sense) "saturated" and "unsaturated" things, i.e., concrete, individual properties and abstract properties.
25
For a medieval-style resolution of the problems involved in referring to and quantifying over nonexistent individuals see Essay IV. of this
volume. For a more detailed description of the formal model sketched here see Appendix of Essay VI. Concerning the general semantic
framework consult Essay V.
Concerning  -abstraction see Essays V. and VI.
26
27
For more on technical details see the papers referred to in n.25.
28
See Schmidt, op.cit., pp. 2ll-2l4: 311-313.
29
As I have said, the significate of a predicate P in a thingUat a time t may be either actual or non-actual at time t. When non-actual, this may be
either still possible or not. However, it may also be the case that a predicate signifies nothing in an individual, this predicate being uninterpreted
for this individual. This case may be symbolized by assigning the zero-entity as its value to the signification function of this predicate for this
individual. So in general, Sgt(P)(u)(t) is an element of W(t)U{0}. Now I think it is worth mentioning here that with this formulation at hand we
can show a way out for Buridan from the vicious circle of Pierre d'Ailly's objection. For according to Buridan the ultimate significates of non-
connotative general terms are extramental individuals. So we can write: Sgt(P)(u)(t){u,0}. But then we can define supposition with reference
to these significates, without having to rely on the notion of verification, by saying that the personal supposita of a term are simply its ultimate

102
this place by a particular human mind should be the nature signified by the predicate as existing
in this human mind, namely the concept, or intention signified by this predicate in this mind. 30
So, in formulas:
Sgt(P)(m)(t) = Con(m,t)(P).
But this concept, even if it is particular in that it exists in this human mind, still, it is abstract
as regards its objects, namely the things to which it can be applied. In his commentary on St.
Thomas's De Ente et Essentia Cajetan wrote about the relation between a concept and its objects
as follows: "... note that there are two sorts of concepts, sc. conceptus formalis and conceptus
objectalis. A conceptus formalis is a sort of idol formed by the receptive intellect in itself which
represents the thing thought of as its object, and which by the philosophers is called intentio or
conceptus, by the theologians, however, verbum. On the other hand, a conceptus objectalis is
the thing represented by the conceptus formalis, terminating in it the act of thinking. For
example, the conceptus formalis of lion is that image which the receptive intellect forms in
itself about the quiddity of lion when striving to think of it, but the conceptus objectalis of the
same is the nature of the lion itself which is thought of and is represented. And we should not
think, when we say that a name signifies a concept, that it signifies only one of these: for the
name `lion' signifies both, even if in different ways. For it is the sign of the conceptus formalis
as of a medium, as of by which something is signified, and it is the sign of the conceptus
objectalis as of that which is ultimately signified. So it makes no difference whether we speak
of the concept of being or the signification of being."31
So the conceptus objectalis of a thing is nothing but the nature ultimately signified by the
term signifying this nature. So if we denote the ultimate significate of the term P in the thing u at
time t as Sgt'(P)(u)(t), then we can write:
Con(m,t)(P)(u)(t) = Sgt'(P)(u)(t),
where Con(m,t)(P)(t) is an element of W(t). But as this ultimate significate of P in u is nothing
but the nature of u signified by P, therefore
Con(m,t)(P)(u)(t) = Sgt'(P)(u)(t) = Sgt(P)(u)(t),
that is, the individualized nature in u signified by P is nothing but the ultimate significate of P
in u, which is the same as the conceptus objectalis of P in the thing u, provided u is not a human
mind.32
As we can see, in this reconstruction we have two sorts of partially identical functions:
significations and concepts. These functions give exactly the same values for the individuals
which are not human minds. On the other hand, for a human mind the signification function of a
term P gives as its value the immediate significate of P, namely P's concept in this mind. But the
concept, when applied to a thing, or even to the mind itself, gives as its value the ultimate

significata: Sp(P)(t){u: Sgt(P)(u)(t) = u} if this set is not empty, otherwise Sp(P)(t) = 0, indeed, as Buridan himself says in his De
Suppositionibus, ed.cit., pp. 201-202. Note also that the above definition corresponds to Buridan's definition of natural supposition. From this
we can get accidental supposition by the following restriction: Sp(P)(t){u: Sgt(P)(u)(t)A(t)}, if this set is not empty, otherwise Sp(P)(t)=0,
which is the definition of supposition given in the text above. It can also be demonstrated that as a consequence of this definition the term will
supposit for those things of which it is true.
30
Cf.: "haec autem natura habet duplex esse: unum in singularibus, aliud in anima" De Ente c.4.
31
Cajetan, op.cit., p. 3O1.
32
U should not be a phantasm either. See the last section of the paper.

103
significate of this term, which is the individual property, or form, inherent in this thing or this
mind, which verifies the term P of it. This is how this reconstruction reflects the fact that it is not
possession of a concept but the possession of an inherent property signified by the concept that
verifies a predicate of the mind which possesses the concept. For example: it is not by a mind's
possessing the concept of immateriality that it is immaterial, but it is immaterial by possessing
the inherent property of immateriality signified by the concept of immateriality. And it is this
concept which is signified immediately, but it is the inherent property which is signified
ultimately by the term `immaterial'. To give this an explicit formulation we can write the
following:
Sgt(immaterial)(m)(t) = Con(m,t)(immaterial),
but this is not identical with:
Sgt'(immaterial)(m)(t) = Con(m,t)(immaterial)(m)(t).
Now I think this reconstruction can give us a good starting point to form a consistent picture of
St. Thomas's theory lying behind his treatment of our original problem. In fact, as in a formal
model like this we can easily judge the identity or non-identity of the elements of the model
(these elements being set-theoretical objects, namely, functions, their values and their
arguments), therefore, if our reconstruction is correct, then we can easily answer the above
questions concerning the relationships between the members of St. Thomas's threefold
distinction.
But first, let us see again our original problem in the framework of this reconstruction. As we
saw, human nature as such, being considered absolutely, in abstraction from its individuating
conditions, but also without the exclusion of these, is signified by a term in predicate position.
On the other hand, as St. Thomas's discussion shows, this nature can be made also subject in a
proposition, but owing to its abstract character, only by using so-called reduplicative
constructions.33 However, in this context nothing can be verified of a term which is not implied
in its definition. But being a species is by no means implied in the definition of homo. So this is
why it cannot be human nature according to its absolute consideration which is supposited for by
the term homo in Homo est species according as this proposition is true. On the other hand, as
the concept signified immediately by this term is also abstract, indeed, it is the same as the nature
in itself as regards its objects, but, at the same time, a concrete quality of the mind, therefore it
can be said to be predicable of many also in a non-reduplicative context. So if in Homo est
species the subject term stands for this concept, then (and only then) the proposition is true. And
so we can clearly see why our problematic inference is invalid.34

33
Concerning the connection between abstraction and reduplicative constructions see J. Lear, "Aristotle's Philosophy of Mathematics", The
Philosophical Review, April, 1982, pp. 161-192. Concerning l3th century treatment of reduplicative constructions in general see my "Libellus
pro Sapiente - a Criticism of Allan Back's Argument against St. Thomas Aquinas's Theory of the Incarnation", The New Scholasticism,
58(1984)pp. 2O7-219.
34
Cf. text quoted in n.17.

104
8. CONCLUSION:LOGIC, METAPHYSICS AND PSYCHOLOGY IN THE
RECONSTRUCTED FRAMEWORK OF ST. THOMAS'S SOLUTION

Now having thus reconstructed the background of St. Thomas's treatment of our original
problem, we may perhaps more confidently set about answering the more general questions
concerning the relationships between logic, metaphysics and psychology with respect to the
foundations of correct inferences. Indeed, by the help of this reconstruction, we may perhaps
more clearly delineate the respective domains of these sciences.
As we saw, in this reconstruction a thing, its individual nature, and the nature considered
absolutely are related to each other as argument, value and function, respectively. When this
function gets as its argument an extramental thing which is not a human mind, then it gives as its
value the individual nature of this individual signified by the term signifying this nature.35 On the
other hand, when this function gets as its argument a human mind, then what it gives as its value
is the particular concept signified immediately by the term in this human mind. However, this
concept is also abstract in that it can be applied to many things, being the natural sign of the
particular, individualized natures of the individuals which have this nature. Indeed, these
particular natures, which are the natural significates of this concept are the ultimate significates
of the term immediately signifying the concept.36
However, having said this much we still have left out of consideration a rather important
aspect of concepts. We have been speaking all the time about words, their meanings, concepts,
and their relations to individual things, but we have said nothing about their relations to
subjective representations of individual things, namely phantasms.
For our concepts, as we are taught by St. Thomas,37 are the result of an operation of our minds,
namely formatio, as it is called, which, in turn, uses as its principle of activity a species
intelligibilis abstracted by the active intellect, intellectus agens, from the sensual representations
of particulars, from phantasms. In the process of abstraction the active intellect separates the
universal nature intuited within the phantasms from its individuating conditions, though without
the exclusion of these, and creates in the receptive intellect, intellectus possibilis, a universal
similitude of the individuated natures of individual things, a socalled species intelligibilis. This
species intelligibilis serves as the principle of the operation called formatio, the term, or result
of which is the concept, intention, or mental verb signified immediately by the external, vocal
verb.38 Now this mental verb enters into composition with mental representations of individuals
to form a mental proposition.39 So this concept, as a function in our reconstruction, for an
extramental thing as its argument gives as its value the inherent property of the thing signified
ultimately by the term signifying the concept immediately. On the other hand, for a
representation of an extramental individual as its argument, the same concept gives as its value
the significate of the predicate of the mental proposition (the predicate of the mental proposition
being the concept itself) in the mental representation of the thing supposited for by the subject of
35
This kind of functional application is the symbolic counterpart of concretion achieved by composition. See n. 28.
36
This is how our semantic framework reproduces Aristotle's "semantic triangle" in his Perihermeneias. See St. Thomas's In Peri I.2.
37
Cf. Schmidt, op.cit., pp. 94-122.
38
See e.g. ST1.85.2.ad3.
39
Cf. Alamannus, op.cit., q.l9.a.l.

105
the corresponding vocal proposition. So the terms of a vocal proposition as well as the
proposition itself have a twofold interpretation: one ad extra, which concerns extramental things
and their individual properties, the actuality of which founds the truth of a proposition; the other
apud mentem, which concerns representations of individual things and their properties, the
actuality of which in the subjective representational system of a human mind founds the mind's
belief in this proposition.40
Now with this reconstructed picture at hand we can give as an answer to the questions
concerning the relationships between logic, metaphysics and psychology the following: logic
concerns concepts in their relation to extramental objects, in which, as we saw, they coincide
with objective meanings.41 And this is why logic is able to formulate objective laws concerning
subjective concepts, determining the rules of deductive inferences which any intersubjectively
communicable concept should comply with. On the other hand, psychology studies the same
concepts in their relation to the subjective representations of extramental things and their
properties, giving a description of how our subjective concepts and beliefs are generated on the
basis of external experience. And this how psychology can contribute by the study of abstraction
and concept formation to the study of inductive inferences.42 Finally, metaphysics, being par
excellence the science of truth, studies the correspondence or adequation between these
conceptual structures and the structure of reality, in which the concept of truth consists. And this
is how metaphysics contributes to the attainment of truth in general, and so to the attainment of
Prime Truth, Veritas Prima, in particular, in qua sit finis huius sermonis.43

40
For technical details see again Essay V.
41
For, as we could see, these give the same values for the same arguments which are neither human minds nor mental representations.
42
Cf. Schmidt, op. cit., pp.27O-295. and entry `psychology' in the Index.
43
Cf. In Meta 2.2. and De Ente in fine.

106
TABLE OF CONTENTS

Essay I: Why "Ars Artium"? 2

Essay II: The Square Of Opposition, Common Personal Supposition


And The Identity Theory Of Predication Within Quantification Theory 9

Essay III: General Terms In Their Referring Function 20

Essay IV: Existence, Quantification And The Medieval Theory Of Ampliation 40

Essay V: Understanding Matters From A Logical Angle:


Logical Aspects of Understanding 53

Essay VI: On Being And Essence In Saint Thomas Aquinas's Metaphysics


And Philosophy Of Science 70

Essay VII: `Socrates Est Species': Logic, Metaphysics and Psychology


in Saint Thomas Aquinas's Treatment of a Paralogism 76

107