Professional Documents
Culture Documents
(Guttorm Fløistad (Auth.), Guttorm Fløistad, G. PDF
(Guttorm Fløistad (Auth.), Guttorm Fløistad, G. PDF
Tome 1
Philosophie du langage
(Editeur G. Fl¢istad)
*
Logique philosophique
(Co-editeur G.H. von Wright)
edited by
GUTTORM FI.q)ISTAD
University of Oslo
Volume 1
Philosophy of language
(Editor G. Flq,istad)
*
Philosophical logic
(Co-editor C.H. von Wright)
Published under the auspices of the International Council of Philosophy and Humanistic Studies
and of the International Federation of Philosophical Societies, with the support of UNESCO.
Publie sous Ies auspices du Conseil International de la Philosophie et des Sciences Humaines et de
la Federation Internationale des Societes de Philopsophie, avec Ie concours de l'UNESCO.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted in any form or by any means, mechanical, photocopying, recording, or otherwise.
without the written permission of the publisher.
Martinus Nijhoff Publishers, P. O. Box 163, 3300 AD Dordrechr,
Contents/Table des matieres
G. FI¢istad, Introduction 3
J. Pelc, The place of the philosophy of language 11
C. Peacocke, The theory of meaning in analytical philosophy 35
J. Hintikka, Semantics: A revolt against Frege 57
J. Bouveresse, Wittgenstein et la philosophie du langage 83
N. Cocchiarella, Richard Montague and the logical analysis of
language 113
F. Kambartel and H.1. Schneider, ConstructiHg a pragmatic
foundation for semantics 155
P. Ricoeur, 'Logique hermeneutique'? 179
Philosophicallogic/Logique philosophique
and because they always occur within some higher order purpose
(context or situation) the intensions are always two-dimensional.
In both cases, however, our view of reality is in part determined
by language and its pragmatics.
The relations between syntax, semantics and pragmatics are
difficult to decide. The chronicles give conflicting views. Logicians
and linguists certainly have a point when they say that syntax is
the most fundamental discipline: No particular use of language
ever invents its syntax; it rather employs a given one. On the other
hand pragmatics may, historically speaking, undoubtedly be said
to have effected changes in the syntax of a given language. And
pragmatics also shows its power in syntactically incomplete
sentences. The context often gives them a perfect meaning.
Regarding the relation of a concept used in some utterance to a
hearer's linguistic expectation (a pragmatic aspect), we may at first
think that the concept acquires its meaning solely within the
utterance and its context and is then conveyed to, and eventually
reinterpreted through, the hearer's expectation. However, the
opposite probably holds true. The hearer's expectation is an
essential component of the speaker's context that helps to deter-
mine his interpretation and use of the concept. The context as a
whole thus has a significant influence on the reality presented in
the utterance and the way it is presented. In ideologically dis-
torted meanings the creative power of language is particularly
perspicuous.
Because ot difficulties such as these, it is probably true to say
that syntax, semantics and pragmatics (in logical as well as in
empirical semiotics) are mutually dependent on each other, and
that none of these disciplines has absolute priority over the
others with regard to the understanding and explanation of
natural language. The priority of one discipline is at most relative.
A cautious solution such as this is also recommended from a
different point of view. It is notoriously difficult to draw any
sharp boundaries between these disciplines and their objects,
especially between semantics and pragmatics. In their account of
logical constructivism Kambartel and Schneider hold initially
that pragmatics is the fundamental discipline, but later admit
that pragmatics comes close to (an extended) semantics. And
Hintikka thinks that text-semantics is just as viable as 'ordinary'
lOG. Fll/listad
II. SEMIOTICS
3. Semiotics as a method
means that our attention does not focus on that object alone, but
moves toward that of which the object - in our opinion - is a
sign, and often also toward the mental experiences of the actual or
alleged producer of the sign, experiences which we believe to be
expressed by that object. In a word, such an object becomes
subject to a semiotic interpretation, in particular, to understanding.
When this occurs, we reflect on its meaning or signification: for
instance, the meaning of a rite, a symphony, the outfit of a
housing space, or a rainbow. We also try to determine the object's
role is in the process of communication.
Application of the semiotic method to human and/or animal
products and actions that are not signs by their nature, and to
natural events, makes it possible to uniform our look at the
various spheres of cognition, and in particular enables us to
analyze nonhumanistic objects in a humanistic manner, so that
they are subject to such procedures as valuation and teleological
approach. This, on the one hand, opens opportunities for inte-
grating research and making it more comprehensive by the use of
the semiotic method, which converts the objects of its analysis
into texts and other sets of signs, but, on the other hand, involves
the danger of abusing the concepts of sign, meaning, and language.
Abusing the semiotic method may sometimes result in super-
stitions and groundless interpretations of allegedly supernatural
phenomena, and in many cases ends in formulating pseudo-
scientific theories.
Whenever language is the subject matter of investigation, the
semiotic method and that employed in the philosophy of language
coincide, or, in other words, the philosophy of language makes use
of the semiotic method. What is more, the philosophy of language
may be identified with that part of semiotics which deals with
language. The close relationship between the philosophy of lan-
guage and semiotics is perhaps best visible in the case of logical
semiotics.
4. Logical semiotics
BI BLiOG RAPHY
This survey will be restricted in scope not only in the respects indi-
cated in the title and by a limitation to fewer than thirty pages,
but also by the inevitable omission of discussion of the semantics
of particular expressions - proper names, demonstratives, natural
kind words, predicate and sentential operators, and so forth - the
upshot of which cannot fail to have some effect on our conception
of a semantic theory. But even within these restrictions, the field
is a vast one; and it will help to structure our discussion if we
impose a division corresponding to the following questions:
(I) What form should a theory that specifies the meaning of all
the sentences of a particular language take? I will follow the con-
vention of calling such a theory for a particular language a
'meaning theory' (MT) for that language, and will reserve the
phrase 'theory of meaning' for a theory (together with the argu-
ments for it) about the correct form for a MT.
(II) In virtue of what is one rather than another MT applicable
to a particular language in use in a community of speakers?
(III) What are the various constraints upon and infirmities of
the relations between our answers in questions (I) and (II)?
Though a theorist's motivation for answering anyone of these
questions in a particular way can hardly be independent of the
views he holds about the others, it is possible for different
theorists to agree on the answers to some (notably (I) and to
disagree on others.
I shall enumerate six answers of the last twelve years to this ques-
tion, before considering how they fare in respect of conditions of
adequacy derivable from not much more than the idea of a MT.
These six answers are not mutually exclusive: there are philos-
ophers and linguists who would want to combine some of them
together, each part of a total theory fulfilling a different role
demanded of a MT.
(i) Translational semantics. According to an idea inherited from
Fodor, Katz and Postal, a MT should consist of a mapping of sen-
tences of the language for which a MT is to be given into a dis-
ambiguated language, a language that perspicuously exhibits
various relations of entailment between predicates and possibly
other expressions (Katz [I], [2]). Though in generative semantics
the idea of a syntactically identified level of deep structure to
which a translational semantics is applied was abandoned, the con-
ception of interpretation as in part translation into something or
other was not: and it is only a variation on the fundamental theme
given by this idea if it is suggested that the semantic representation
use first-order logic. (For a survey, see Fodor [3].)
(ii) Model-theoretic semantics. A view developed notably by
Montague [4] and his students is that a MT for a language L must
consist of a definition of truth in an arbitrary model for each sen-
tence of L. Almost all model-theoretic semanticists agree either in
their metatheoretical commentary, or show by their actual prac-
tice, that they would in a MT for a language pick out a dis-
tinguished one of the indices relative to which models are defined,
the actual world, the present moment, or whatever; Montague [5]
wrote explicitly that this ought to be done. But beyond this point
of agreement, there are major distinctions between the kinds of
model-theory that have been advocated. (Some are perhaps not
happily called 'model theory' at all.) (a) There are those who hold
that a MT must specify the extensions of expressions not only
with respect to the actual world, but also with respect to ways
things might have been; thus David Lewis [6]. On this view, the
assignment of extensions to expressions with respect to nonactual
worlds is meant to capture the actual meaning of the expression in
the language being treated: in the case of English, with respect to
The theory o/meaning in analytical philosophy 37
be justified, but only that any justification is not in fact the route
by which the speaker arrived at it.) On this position there is not an
infinite amount of knowledge possessed by the understanding
speaker, but only a potentially infinite capacity: any requirement
that an infinite amount of knowledge in a single person have a
finite basis from which it is inferred would thus not be applicable.
This view, according to which nothing in a correct account of
understanding demands that the axioms of a MT be known by the
speakers of the language, ought not to be argued for on the ground
that all knowledge, or even only semantical knowledge, of a lan-
guage-using creature must be explicit. For there could be a
community of persons with a language that is understood by the
members of the community and who thus know on this view what
particular utterances say: yet nothing in this description of the
case requires that their language actually contain a word that
means 'says' (in the indirect sense) or 'true' or anything else that
they could use verbally to express this knowledge about particular
utterances. The view ought rather to be defended by arguments
from the very great sophistication of the concepts that may
need to be used in a theoretical description, both semantical and
syntactical, of their language. In particular, if we do attribute to a
person beliefs involving these sophisticated concepts as part of our
account of understanding a language, then we must explain why
these people do not behave on other occasions as one would
expect someone possessed of these concepts (such as sequences,
transformational cycles, and so forth) to behave. The semantic
analogue of Stich's view of syntax says that there is no such
knowledge in the ordinary competent speaker, and the expla-
nation of his not behaving in appropriate ways is simply that
he need not be possessed of the concept. Certainly it is not a
general truth that speakers know the principles that determine the
application of the concepts that are used in the expression of
truths that they do know: else we would have to say that anyone
with binocular vision was sufficiently sophisticated to solve simul-
taneous equations in visually judging distance.
My second topic in this section is indeterminacy. Quine's thesis
of the indeterminacy of translation has almost certainly received
more attention in the journals in our period than any other topic
in the theory of meaning. One way of stating the thesis is that
52 C. Peacocke
there can be two equally good translation manuals from one lan-
guage L into a second, M, which both preserve the verbal dis-
positions of all concerned with respect to the sentences in ques-
tion, and which yet map a given sentence of L onto sentences of M
that differ in truth value. (This formulation does not employ the
notion of meaning.) Quine's emphasis on his grounds for holding it
have altered in this period, as has too the sense in which he sub-
scribes to it. In 1970 he wrote that 'the real ground' for holding it
was the general undetermination of scientific theory by all
actually true (and "pegged") observation sentences [58). Yet he
also said later in our period that the only version of the under-
determination thesis that he would definitely affirm is one saying
that our theory of the world is bound to have empirically (i.e.
for Quine, observationally) equivalent alternatives such that, if we
were to discover them, we would see no way of reconciling them
by translating the predicates of one by open sentences of the other
[59). The sense, then, in which he now subscribes to indeter-
minacy insofar as it rested on the underdetermination thesis must
be correspondingly muted. This description of Quine's present
position meshes well with some remarks of Putnam [60].
Quine distinguishes an argument from above and an argument
from below for indeterminacy. The argument from above - that
is, the argument from underdetermination - is simply that since
theory can vary even though the truth value of all pegged obser-
vation sentences be fixed, the translation of a foreign scientist's
theoretical sentences can vary, preserving his dispositions to verbal
behaviour, even though the translation of his observational sen-
tences is fixed (Quine [58)). But it has been rightly objected by
Putnam [60] that Quine's notion of observationally equivalent
theories is very weak, turning as it does on agreement of the
theories of which indicative observational sentences are entailed by
the theories. A wider notion would require agreement on counter-
factuals stated in observational terms: and it is extremely unclear
that theory can vary while observationality in this sense is fixed.
(Clearly this issue depends in part on whether all counterfactuals
can be reduced to categoricals.)
Quite apart from Putnam's point, Quine's position seems to
involve an asymmetry. Though on Quine's view observation does
not determine theory, he is realistic about the physical theory he
The theory of meaning in analytical philosophy 53
BIBLIOGRAPHY
Evans, G. [7] Semantic Structure and Logical Form. In Evans and McDowell.
_. [46] The Causal Theory of Names. Proc. Arist. Soc. s.v. 47 (1973): 187-
208.
- [64] Identity and Predication. Journal of Philosophy. 72 (1975): 343-63.
- and McDowell, J. [67] Truth and Meaning. Essays in Semantics. Oxford:
Oxford University Press, 1976.
Field H. [25] Logic, Meaning and Conceptual Role. Journal ofPhilosophy 74
(1977): 379-409.
- [49] Tarski's Theory of Truth. Joumal of Philosophy 69 (1972): 347-75.
- [50] Conventionalism and Instrumentalism in Semantics. Noas 9 (1975):
375-406.
[65] Quine and the Correspondence Theory. Philosophical Review 83
(1974): 200-28.
- [66] Theory Change and the Indeterminacy of Reference. Journal of
Philosophy 70 (1973): 462-81.
Fodor,J.A. [51] The Language of Thought. New York: Crowell, 1975.
Fodor, J.D. [3] Semantics: Theories of Meaning in Generative Grammar.
New York: Crowell, 1977.
Friedman, M. [63] Physicalism and the Indeterminacy of Translation. Noas
9 (1975): 353-74.
Grandy, R. [33] Some Remarks about Logical Form. Noas 8 (1974): 157-
64.
- [47] Reference, Meaning and Belief. Journal of Philosophy 70 (1973):
439-52.
Graves, C. et al. [55] Tacit Knowledge. Journal of Philosophy 70 (1973):
318-331.
Grice, H. [36] Utterer's Meaning and Intentions. Philosophical Review
78 (1969): 147-77.
- [37] Utterer's Meaning, Sentence-Meaning, and Word-Meaning. Foun-
dations of Language 4 (I 968): 1-18.
Hacker, P. [18] IInsight and Illusion. Oxford: Oxford University Press, 1972.
Harman, G. [23] Thought. Princeton: Princeton University Press, 1973.
- [24] Meaning and Semantics. In M.K. Munitz and P.K. Unger, Eds.,
Semantics and Philosophy. New York: New York University Press,
1974.
- [39] Review of Schiffer [24] . Journal of Philosophy 71 (1974): 224-9.
Hintikka, J. [13a] Quantifiers in Logic and Quantifiers in Natural Language.
In S. Korner, Ed., Philosophy of Logic. Oxford: Blackwell, 1976.
- [I3b] Quantifiers in Natural Languages: Some Logical Problems II. Lin-
guistics and Philosophy I (1977): 153-72.
- [16] Reply to Peacocke. In E. Saarinen, Ed., Game-Theoretic Semantics.
Dordrecht: Reidel, 1978.
Katz, J., and Fodor, J. [IJ The Structure of Semantic Theory. Language
xxxix (1963): 170.
- and Postal, P. [2J An Integrated Theory of Linguistic Descriptions.
Cambridge, Mass.: M.I.T. Press, 1964.
56 C. Peacocke
Katz, J. [26] Semantic Theory. New York: Harper and Row, 1972.
Kreisel, G. [21] Foundations of Intuitionistic Logic. In E. Nagel et al., Eds.,
Logic, Methodology and Philosophy of Science. Stanford 1962.
Lewis, D. [6] General Semantics. In D. Davidson and G. Harman, Eds.,
Semantics of Natural Language. Dordrecht: Reidel, 1972.
- [40] Convention. Cambridge, Mass.: Harvard University Press, 1969.
Lycan, W. [19] Noninductive Evidence: Recent Work on Wittgenstein's
"Criteria". American Philosophical Quarterly 8 (1971) 109-25.
McDowell, J. [29] Truth Conditions, Bivalence and Verification. In Evans
and McDowell.
McGinn, C. [48] Charity, Interpretation and Belief. Journal of Philosophy
74(1977): 521-35.
Montague, R. [4] Formal Philosophy. Ed. R. Thomason. New Haven: Yale
University Press, 1974.
- [5] English as a Formal Language. In Montague [3].
Peacocke, C. [15] Game-Theoretic Semantics, Quantifiers and Truth. In
E. Saarinen, Ed., Game-Theoretic Semantics. Dordrecht: Reidel, 1978.
- [35] What is a Logical Constant'? Journal of Philosophy 73 (1976): 221-
40.
- [42] Truth Defmitions and Actual Languages. In Evans and McDowell, and
in trans.
Putnam, H. [60] The Refutation of Conventionalism. In M.K. Munitz and
P.K. Unger, Eds., Semantics and Philosophy. New York: New York
University Press, 1974.
Quine, W. [34] Philosophy of Logic. Englewood-Cliffs: Prentice-Hall, 1969.
- [53] The Roots of Reference. La Salle, Ill.: Open Court, 1974.
- [58] On the Reasons for Indeterminacy of Translation. Journal of Phil-
osophy. 67 (1970): 178-83.
- [59] On Empirically Equivalent Systems of the World. Erkenntnis 9,
(1975): 313-28.
- [61] Word and Object. Cambridge, Mass.: M.LT. Press, 1960.
- [62] Olltological Relativity. Journal of Philosophy 65 (1968): 185-212.
Ramsey, F. [54J General Propositions and Causality. In The Foundations
of Mathematics and Other Essays, Ed. R. Braithwaite. London:
Routledge, 1931.
Saarinen, E. [14] Game-Theoretical Semantics. Monist 60 (1977): 406-18.
Schiffer, S. [38] Meaning. Oxford: Oxford University Press, 1972.
Searle J. [45] Speech Acts. Cambridge: Cambridge University Press, 1969.
Strawson, P. [27J Scruton and Wright on Anti-Realism Etc. Proc. Arist. Soc.
51 (1976-7): 15-21.
Stich, S. [57] What Every Speaker Knows. Philosophical Review 80 (1971):
476-96.
Wallace, J. [30J On the Frame of Reference. In D. Davidson and G. Harman,
Eds., Semantics of Natural Language.
Wright, C. [22] Truth-Conditions and Criteria. Proc. Arist. Soc. S.v. 50
(1976): 217-245.
-- [28] Strawson on Anti-Realism. Synthese 40 (1979): 283-299.
Semantics: A revolt against Frege
J AAKKO HINTIKKA
Florida State University
1. FREGE'S HEGEMONY
One reason why Frege's semantical ideas (other than the sense-
reference theory) have largely been overlooked by philosophers is
that Frege did not think seman tical relations could, strictly
speaking, be explicitly formulated and discussed. Hence a system-
atic theoretical study of semantics is impossible in Frege's view.
He is accordingly representative of the type of view labelled by
van Heijenoort [3] 'logic as language,' which perhaps rather
ought to be called in more general terms the view of language as
the universal medium of all discourse. According to this kind of
view we cannot escape the basic semantical relationships that
connect our language with reality. Since they are presupposed in
anything we say, they cannot be meaningfully talked about. On
the view of language as the universal medium we cannot, as we
can on the view of language as calculus, speak of the representa-
tive relationships between language and reality.
In contrast, on the view labelled by van Heijenoort 'logic as
calculus,' we can meaningfully and nontrivially discuss the links
between our own language and reality. We can even think of those
representative relationships as being varied. In other words, we can
Semantics: A revolt against Frege 59
3. POSSIBLE-WORLDS SEMANTICS
4. ATOMIC SENTENCES
over the concept of logical form. For one of the most important
ideas underlying this notion is that the logical form of an unam-
biguous sentence is always unique. It was introduced largely to
capture the unique true logical or semantical form of our sentence
which this varying syntactical clothing tends to hide.
(0 Among the approaches to semantics which postulate a
unique mode of semantical representation there are those postu-
lating a 'language of thought' which is thought to represent the
way language is actually processed semantically by the human
mind. A representative example of approaches of this sort is Fodor
[ 100] . It is seen that current views of this type can be considered
as latter-day versions of earlier, mostly philosophical views of
thinking as being analogous to speaking - perhaps comparable
with mind's discourse with itself without spoken words. Views of
this kind were put forward among others by Plato and Peirce, and
in the contemporary discussion by Peter Geach [10 I ]. In spite of
such illustrious ancestry, 'language of thought' theories are subject
to serious doubts. Applied to ordinary natural languages, they are
essentially translational theories, and hence open to the same
objections as these. Moreover, the proponents of the 'internal
code' idea have failed to spell out the structure of this code in
several important respects. For instance, the problem of the
ambiguity or nonambiguity of 'is' ought to be easy to solve by
reference to the internal code. Yet no such solutions are forth-
coming from the 'language of thought' theorists.
8. TEXT SEMANTICS
BIBLIOGRAPHY
dead as death." "Riel! n'est aussi mort que la mort; rien n'est aussi
beau que la beaute elle-meme." L'image sous laquelle on se repre-
sente la realite est que la beaute, la mort, etc., sont les substances
pures (concentrees), alors qu'elles sont presentes dans un objet
beau comme ingredient. - Et ne reconnais-je pas ici mes propres
considerations sur "objet" et "complexe"?' (ibid., p. 242).
C'est l'origine linguistique des problemes philosophiques qui
explique, aux yeux de Wittgenstein, a la fois leur profondeur,
leur difficulte particuliere, leur· nature tout a fait differente de
celie des questions scientifiques et leur persistance. 'On en tend
toujours a nouveau, a-t-it ecrit, la remarque selon laquelle la
philosophie ne fait a proprement parler aucun progres, que les
memes probIemes philosophiques qui occupaient les Grecs nous
occupent encore. Mais ceux qui disent cela ne comprennent pas la
raison pour laquelle cela ne peut pas ne pas etre ainsi. Elle est que
notre langage est reste identique a lui-me me et no us entrafne
toujours a nouveau vers les memes questions. Tant qu'il y aura un
verbe "etre", qui donne l'impression de fonctionner comme
"manger" et "boire", tant qu'il y aura des adjectifs "identique",
"vrai", "faux", "possible", tant qu'il sera question d'un flux du
temps et d'une extension de l'espace, etc., etc., les hommes se
heurteront toujours anouveau aux memes difficultes enigmatiques
et auront les yeux fixes sur quelque chose qu'aucune explication
ne semble pouvoir eliminer' (Wittgenstein [4], p. 36).
En un certain sens, Wittgenstein a pris au serieux l'idee
traditionnelle que la philosophie devrait apporter la solution
complete et definitive des problemes qu'elle traite et en a tire la
consequence (a bien des egards triviale) que cette solution ne peut
pas etre donnee sous la forme d'explications ou de theories. Pour
ceux qui sont convaincus que les questions des philosophes ne sont
pas reellement plus 'profondes' que celles des scientifiques et que
les solutions de la philosophie sont aussi partieIles, provisoires et
progressives (en depit des apparences) que celles de la science, les
considerations mHaphilosophiques de Wittgenstein et les moti-
vations essentielles de son entreprise philosophique sont tout
simplement inintelligibles. Comme on I'a remarque, il y a au moins
un point commun important entre la philo sophie de Wittgenstein
et celle de Heidegger: la conviction que les methodes et les objec-
tifs de la philosophie sont fondamentalement differents de ceux
86 J. Bouveresse
II
que de lui est enonce de tous les objets que 1'0n examine' ([4],
p. 35). La consequence est que 'I'on croit suivre encore et toujours
la nature, alors que 1'0n ne fait que passer Ie long de la forme a
travers laqueUe nous la considerons' ([2], § 114).
Quand nous philosophons sur Ie langage, remarque Wittgenstein,
'nous sommes dans l'illusion que ce qu'il y a de particulier, de
profond, d 'essentiel pour nous dans notre recherche reside dans Ie
fait qu'elle essaie de comprendre I'essence incomparable du
langage. C'est-a-dire, l'ordre qui existe entre les concepts de la
proposition, du mot, de l'inference, de la verite, de l'experience,
etc. Cet ordre est un super-ordre entre - pour ainsi dire - des
super-concepts. Alors qu'en fait les mots "Iangage", "experience",
"monde", s'ils ont une utilisation, doivent en avoir une qui est
d'un niveau aussi bas que celle des mots "table", "lampe".
"porte'" (ibid., § 97). Notre attitude n'est pas celie qu'un
theoricien adopte habituellement a l'egard de l'objet de ses
recherches. Nous avons l'impression que, dans Ie cas precis, il s'agit
de decouvrir non pas un ordre possible ou hypothetique, mais bel
et bien l'ordre commun au langage, a la pensee et au monde --
completement a priori (parce qu'il est 'l'ordre des possibilites qui
doivent etre communes au monde et a la pensee '), parfaitement
simple et transparent (non contamine par les impuretes et les
ambiguHes de l'experience) et omnipresent (en ce sens qu'i!
penetre et gouverne de l'interieur tous les phenomenes).
Cet ordre, qui se manifeste sur Ie mode contradictoire de
l'idealite realisee, c'est-a-dire de la presence-absence, est a la fois
'superieur' et particulierement intime et dissimule. Or Ie pathos
du sublime et celui de la profondeur constituent precisement, aux
yeux de Wittgenstein, les deux symptomes essentiels de la 'super-
stition' philosophique. Ce qui est Ie plus difficile a admettre est
que la proposition est une chose tout a fait ordinaire, et non une
entite unique en son genre et dotee par elle-meme de pouvoirs
etranges et inexplicables, que nous pouvons decrire Ie fonction-
nement du langage sans avoir a utiliser necessairement des con-
cepts d'un ordre plus eleve que ceux qu'il nous fournit et que
taus les faits importants sont, d'une certaine maniere, sur Ie
meme plan et a la surface (ce sont meme des faits si evidents que
seule leur trop grande familiarite nous empeche de les remarquer
et de no us en etonner). A la difference de la science, la philosophie
Wittgenstein et fa philosophie du fangage 91
III
pas du tout evident, ecrit Hacker, que I'on ne puisse pas a la fois
rendre justice a I'indetermination du sens, au caractere ouvert des
concepts et a la flexibilite du langage, et egalement etre systemati-
que et englobant' (p. 140). De fait, 'bien des aspects de la
philosophie critique de Wittgenstein peuvent etre presentes de
fa~on systematique et revelent aisement Ie caractere acheve et
englobant de son oeuvre. Et, ce qui est plus important, il est loin
d'etre evident que ses dernieres conceptions sur la semantique ne
puissent pas etre representees de fa«on coherente et bel et bien
formalisees, de fayon a fournir ce que la plupart des philosophes
appelleraient une theorie englobante de la signification' (p. 141).
Mais, s'il est vrai que la philosophie n'est pas reellement une
activite theorique, it doit etre vrai egalement qu'un probleme
devenu proprement theorique n'est plus exactement un probleme
de philosophie. Chez Frege, !'idee d'une 'Iogique du vague' aurait
suscite des reactions comparables acelles qu'a pu provoquer autre-
fois I'idee d'une 'science de l'incertitude' (selon I'expression qui
a ete utilisee a propos du calcul des probabilites). Mais Iorsque
Ie flou et l'imprecision des concepts usuels ont perdu Ie caractere
philosophiquement preoccupanL voire franchement inacceptable,
qu'ils avaient pour lui (cf. Frege [10], II, § 56) et ont fini par
devenir un fait accepte et familier, en meme temps que I'objet
d 'une theorisation systematique, sous la forme d'une logique
du vague ou de quoi que ce soit de ce genre, les problemes qu'ils
posent ne ressemblent plus que superficiellement ace que Wittgen-
stein appelle un probleme philosophique. Pour lui, la philosophie
proprement dite intervient avant, au stade des resistances et des
prejuges instinctifs qui s'opposent a Ia perception des faits et
(secondairement) a la construction eventuelle d'une theorie, ou
apres coup, dans la mesure ou une theorie nouvelle et des con-
cepts nouveaux entrainent a peu pres inevitablement des
problemes de clarification nouveaux.
IV
realite n'existe pas veritablement, tant que l'usage n'est pas etabli;
et seuls certains cas particuliers de ce que nous appelons, de fa~on
generale, l'usage ou l'application des signes peuvent etre decrits de
fac;on plausible et eclairante en termes de projection et de con-
cordance, au sens d'une communaute de forme ou de structure.
Selon Ie Tractatus, Ie signe propositionnel doue de sens (pris
avec la methode de projection) est Ie signe propositionnel pense.
L'intervention de la pensee, qui est pourtant eUe-meme une
image logique des faits, au meme titre que Ia proposition,
supprime la distance apparemment infranchissable qui existe
entre toute image et ce qu'elle represente en dehors d'elle-meme.
Pourtant, aussi longtemps qu'on se represente Ia pensee,
l'intention, etc., sous la forme d'une image, meme parfaitement
ressemblante, Ie probleme reste entier. Lorsque quelqu'un pense,
'les objets sur lesquels il pense ne sont assurement pas du tout
dans sa tete (aussi peu que dans sa pensee)!' ([6], p. 143). La
question cruciale de savoir comment la pensee, Ie vouloir-dire,
l'imagination, I'attente, Ie desir, etc., peuvent reellement anticiper
sur un fait determine, qui y est deja, d'une maniere ou d'une
autre, present sous une forme affaiblie, mais parfaitement uni-
voque - celie d'une sorte de double ethere qui s'effacera, Ie
moment venu, devant Ia realite tangible - est simplement
rep osee a un au tre niveau.
On pourrait reformuler ainsi Ie probleme philosophique fonda-
mental qui a domine, d'un bout a l'autre, la problematique du
signe chez Wittgenstein: qu'est-ce qui fait vivre Ie signe materiel,
qui est apparemment une chose inerte, et lui permet de dire
queIque chose? La reponse classique est que 'l'action du langage
consiste en deux parties; une partie inorganique, la manipulation
de signes, et une partie organique, que nous pouvons appeler
comprendre ces signes, leur donner un sens, les interpreter, penser'
(The Blue Boo k [17], p. 3). Mais 'tou tes les fois que nous in ter-
pretons un symbole d 'une maniere ou d 'une autre, l'interpretation
est un nouveau symbole ajoute a l'ancien' (ibid., p. 33). Le sym-
bole mental n'est en lui-meme pas plus proche et pas plus eloigne
de l'objet qu'il vise que Ie signe materiel. Et, de ce point de vue,
nous ne perdons rien a remplacer systematiquement Ie premier
par Ie deuxieme. (Comme dit Wittgenstein, ce dont nous avions
besoin est precisement Ie caractere occulte de I'objet ou du
108 1. Bouveresse
ce qu'on fait avec elle, ou qu'on regarde fixement dans son arne
a soi pour voir si deux expressions ont la meme signification, et
d 'autres choses de ce genre' (Diamond [18] , p. 111).
Une des attaques les plus serieuses contre la philosophie du lan-
gage de Wittgenstein et sa conception de la philosophie tout court
est celle qui est provenue recemment de ce qu'on appelle quelque-
fois 'la nouvelle theorie de la reference', qui lui reproche d'etre
reste, en depit de ses reticences bien connues, foncic:hement
tributaire de la notion traditionnelle de signification. Si les repre-
sentants de cette theorie ont raison, il peut, d'une certaine maniere,
y avoir une interrogation sur ce qu'un mot signifie 'reellement' et
une question qui a trait a la signification exacte du mot 'or', par
exemple, n'est pas fondamentalement differente d'une question
scientifique sur la nature veritable de la substance en cause. Cela
oblige a reconsiderer Ie role essentiel que I'on a pris I'habitude
d'attribuer a l'analyse conceptuelle pour la solution des problemes
philosophiques et I'idee que l'essence doit etre cherchee dans des
conventions grammaticales, plutot que decouverte au fur et a
mesure que nous reconnaissons les choses pour ce qu'eUes sont
reellement. Comme I'a montre la polemique entre Putnam et
Malcolm sur la nature du reve, meme la pertinence et la superiorite
apparentes de l'analyse criteriologique dans Ie cas des termes
designant des experiences ou des processus mentaux est contestee
par la nouvelle theorie. Et ce que Putnam appelle 'la division du
travail linguistiquc' implique que la signification, si elle n'a effec-
tivement rien d'un element occulte plus ou moins prive, n'est
cependant pas non plus accessible par la simple observation de
l'usage public, dans la mesure OU les criteres d'identification
detenus par les 10cuteurs 'experts' se trouvent etre, a travers eux,
la propriete collective de la communaute linguistique, d'une
maniere telle que 'Ie fait Ie plus esoterique concernant l'eau peut
devenir une partie de la signification sociale du mot, bien qu'il
soit inconnu de presque tous les locuteurs qui acquierent Ie mot'
(,Meaning and Reference', in S.P. Schwartz (ed.), [19], p. 126).
Si cette delegation de propriete constitue reellement un aspect
essentiel du phenomene de la signification, il est difficile de
maintenir que la philosophie ne s'interesse pas a ce qui est cache
et propose quelque chose comme 'une vue "synoptique" de choses
que nous savons tous'.
110 J. Bouveresse
BIBLIOGRAPHIE
l. Ouvrages de Wittgenstein
- [16] Notebooks 1914-1916. Eds. G.H. von Wright and G.E.M. Anscombe.
Oxford: Blackwell, 1961.
- [17] The Blue and Brown Books. Oxford: Blackwell, 1958.
- [20] Worterbuch fur Vo lksschu len. Wien: Holder-Pichler-Tempsky, 1977.
- [21] Some Remarks on Logical Form. Proceedings of the Aristotelian
Society, Supp. Vol. 9 (1929); reproduit dans Essays on Wittgenstein's
Tractatus. Eds. I.M. Copi and R.W. Beard. London: Routledge and
Kegan Paul, 1966.
- [22J Philosophische Bemerkungen. Ed. R. Rhees. Oxford: Blackwell,
1964.
- 123] Bemerkungen iiber die Farben (Remarks on Colour). Ed. G.E.M.
Anscombe. Oxford: Blackwell, 1977.
Ambrose, A., and Lazerowitz, M., Eds. [24J Ludwig Wittgenstein: Philos-
ophy and Language. London: Allen and Unwin, 1972.
Anscombe, G.E.M. [25J An Introduction to Wittgenstein's Tractatus, 2nd
edn. London: Hutchinson, 1963.
Baker and Hacker [I3J Critical Notice on 'Philosophical Grammar'. Mind
85 (1976).
Black, M. [26] A Companion to Wittgenstein's Tractatus. Cambridge:
Cambridge University Press, 1964.
Bogen, J. [27J Wittgenstein's Philosophy of Language. London: Routledge
and Kegan Paul, 1972.
Diamond, C., Ed. [18] Wittgenstein's Lectures on the Foundations of Mathe-
matics. The Harvester Press, 1976.
Dietrich, R.A., [28] Sprache und Wirklichkeit in Wittgensteins Tractatus.
Tiibingen: Max Niemeyer Verlag, 1973.
Dummett, M. [29] Frege: Philosophy of Language. London: Duckworth,
1973.
Durfee, H.A., Ed. [30] Analytic Philosophy and Phenomenology. The Hague:
M. Nijhoff, 1976.
Fogelin, RJ. [31] Wittgenstein. London/Henley/Boston: Routledge and
Kegan Paul, 1976.
Frege, G. [5] Begriffsschrift. 1879.
- [10J GrundgesetzederArithmetik, I. 1893.
Griffin,1. [32] Wittgensfein's Logical Atomism. London: Oxford University
Press, 1964.
Hacker, P.M.S. [8] Insight and Illusion: Wittgenstein on Philosophy and the
Metaphysics of Experience. Oxford: The Clarendon Press, 1972.
Hallett, G. [33] Wittgenstein's Definition of Meaning as Use. New York:
Fordham University Press, 1967.
- [34] A Companion to Wittgenstein s Philosophical Investigations. Ithaca/
London: Cornell University Press, 1977.
1 12 J. Bouveresse
Hintikka, J., Ed. [35] Essays on Wittgenstein in Honour of GR. l'on Wright.
Amsterdam: North-Holland, 1976.
Hunter, J.F.M. [36] Essays ajter Wittgenstein. London: Allen and Unwin,
1973.
Kenny, A. [37] Wittgenstein. London: Allen Lane, 1973.
Klemke, E.D., Ed. [38] Essays on Wittgenstein. Urbana/Chicago/London:
University of lllinois Press, 1971.
Kreisel, G. [12] Wittgenstein's Theory and Practice of Philosophy. British
Journal for the Philosophy of Science II (I 960).
Lang, M. [39 J Wittgensteins philosophische Grammatik. Den Haag: M.
Nijhoff, 1971.
Leinfeller, E., Leinfeller, W., Berghel, H., Huebner, A., Eds. [39J Wittgen-
stein und sein Einfluss aUf die gegenwti'rtige Philosophie. Akten des
zwciten internationalen Wittgenstein Symposiums, Kirchberg aW. Wien:
Holder-Pichler-Tempsky, 1978.
Pitcher, G. [41] The Philosophy of Wittgenstein. Englewood Cliffs, N.J.:
Prentice-Hall, 1964.
-- Ed. [46] Wittgenstein: The 'Philosophical Investigations'. A Collection of
Critical Essays. New York: Anchor Books, 1966.
Rhees, R. [42 J Discussions of Wittgenstein. London: Routledge and Kegan
Paul, 1970.
Richardson, J.T.E. [1] The Grammar of Justification, An Interpretation of
Wittgenstein's Philosophy of Language. London: Sussex University
Press, 1976.
Schwartz, S.P., Ed. [19] Naming, Necessity and Natural Kinds. Ithaca, N.Y.:
Cornell University Press, 1977.
Setler, G.F. [43] Language and the World, A Methodological Synthesis with-
in the Writings of Martin Heidegger and Ludwig Wittgenstein. Atlantic
Highlands, N.J.: Humanities Press, 1974.
Stenius, E. [44] Wittgenstein's Tractatus, A Critical Exposition of the Main
Lines of Thought. Oxford: Blackwell, 1964.
Winch, P., Ed. [45] Studies in the Philosophy of Wittgenstein. London:
Routledge and Kegan Paul, 1969.
Zimmermann, J. Wittgensteins sprachphilosophische Hermeneutik. Frankfurt:
Vittorio Klostermann, 1975.
Richard Montague and the logical analysis of
language *
NINO COCCHIARELLA
Indiana University
*The author is indebted to the National Endowment for the Humanities for
support during the research and writing of this paper.
4. PRAGMATIC LANGUAGES
Moreover, the (modal) fact that Jones is not seeing a tree in some
other possible world but is having there some completely different
experience indicates that the property of having the experience
which Jones is having is not identical with the property of seeing
a tree, i.e., that
(6) Jones sees a unicorn having the same height as a table actually
before him.
o vuAv[P(v)-B-v=u]
As n-ary relation-in-extension, on the other hand, can be identi-
fied with an n-ary relation-in-intension that has the same extension,
viz., the relation-in-extension in question, at every index. Thus, we
Richard Montague and the logical analysis of language 131
Having made these reductions, one might well maintain that the
distinction between the intension and extension, or the sense and
denotation, of an expression is neither fundamental to intensional
logic - and therefore not a basic distinction for philosophical
analysis - nor necessary, as apparently Gottlob Frege thought, for
the analysis of ordinary language. One might well speak accor-
dingly, of the denotation simpliciter of individual and predicate
expressions, even though the denotations of the latter would in
this case be what others call their senses. Indeed, in [4] , Montague
explicitly endorses and carries through just such a view in his
semantics for a fragment of ordinary English. 'It is wrong to
maintain', Montague argues there, 'that an analysis of ordinary
English (or German) requires a notion of sense as well as one of
denotation' (p. 217).
In [5], on the other hand, Montague withdraws his emphasis
'on the possibility of doing without a distinction between sense and
denotation'; and he remarks that 'while such a distinction can be
avoided in special cases, it remains necessary for the general
theory, and probably provides the clearest approach even to the
special cases in question' (p. 222). And having changed his view on
this matter, Montague proceeds to fonnulate his second version of
intensional logic in which the distinction between sense and
denotation now is taken as fundamental.
The ontological framework of this new intensional logic is to be
based on a type-theoretical distinction regarding the various
entities that are to be explicitly acknowledged in the logic, i.e., the
entities to which quantificational reference is to be meaningful,
and a theory of types which is essentially an extension of Alonzo
Church's theory of simple types. To set-theoretically characterize
the types, let e, t, s be three objects, none of which is an ordered
pair; and let Type, the set of types, be inductively defined as the
smallest set such that (1) e, t E Type; (2) (a, b) E Type whenever
a, bE Type; and (s, a) E Type whenever a E Type.
132 N. Cocchiarella
will not do, since this fonnula represents the different English
sentence:
fully captures all that we want to say with (8) using only indirect
quantification over past individuals. And of course what this
analysis indicates is that we can contextually define direct quanti-
fication over past individ uals in general:
translates into (3', then Fi(O', (3) and Fj(O', (3) translate into ra'C{3'Y.
The mapping which is based upon these two fundamental
assumptions is defined as follows, for arbitrary categories A and B:
fCe) =e
[(t) =t
F(A/B) = f(A//B) = «s, f(B», f(A»
XPl\x(r'(x) ~ p {x})
XPvx(l\y(r'(y) # y = x) A P {x})
APVX(r'(x) A P{x})
respectively, where 'x' and 'y' are variables of type (s, e), i.e., of
the type of individual concepts. Here, of course, proper nouns,
definite descriptions and quantifier phrases are all translated into
expressions of intensional logic that denote sets of properties of
individual concepts.
Simple predication of a term and an IV -phase on this interpre-
tation, it should be noted, treats what the term denotes as the
function and the intension of what the IV -phrase denotes as the
argument. Thus, e.g., the English sentence 'John runs' is translated
by Montague into intensional logic as follows:
148 N. Cocchiarella
8. = df AU 8Cu)
o [8(x) ~ b. ex)]
for any 8 which translates an extensional IV or eN-phrase. (Note
that since 8 is of type {(s, e), 0, i.e., of the type of sets of indi-
vidual concepts, then 8. is of type (e, 0, i.e., of the type of sets of
individuals.) Thus, given that 'runs' is an extensional IV -phrase,
(II) reduces to:
1\ x [man/ex) ~ runs/ex)]
Richard Montague and the logical analysis of language 149
10ve'Cm, PP Cj})
10ve'Cm, Pi\x[man'(x) -+ P{x}])
love: (m, j)
I\.x[man'(x) -+ love: (m, -x)]
where 'x' and 'y' range over individual concepts. Here we should
especially note that while the identity in question in the first
premise is not of two individual concepts but only of their
extensions at the present moment in the real world, the truth of
the second premise, according to Montague, depends on the fact
that 'the temperature,' loosely speaking, 'denotes' an individual
concept and not an individual and that 'rise', unlike most verbs
'depends for its applicability on the full behavior of individual
concepts, not just on their extensions with respect to the actual
world and ... [presentJ moment of time' ([6J, p. 268). Thus,
'temperature' and 'rise' are particular examples of an intensional
CN and IV -phrase, respectively; and it is precisely in order to
explain the invalidity of the above type of argument that
Montague interprets IV and CN-phrases in general as denoting sets
of individual concepts rather than sets of individuals. And, it is for
this reason of course that English terms, including quantifier
phrases, are interpreted by Montague as denoting sets of properties
of individual concepts rather than sets of properties of individuals.
152 N Cocchiarella
NOTES
-
where 9' ranges over properties of properties of individual concepts and
'x', 'y' range over individual concepts.
BIBLIOGRAPHY
Works by others
S. SYMBOLIC ACTS
are given the same symbolic use. This illustrates the particular way
in which agreements and mere conventions are combined in sym-
bolic acts.
As is generally the case with mediating acts, we follow at least
two intentions in the performance of a symbolic act, the direct
one of the mediating act and the indirect one of the symbolic
act. We shall say that the direct intention 'carries' the symbolic
one. Phonetic acts mediating symbolic acts may be called speech
acts (in the narrow sense). In the same way we may understand
writing acts, the carrying intentions then consisting in the pro-
duction of figures (spatial forms). Speech acts and writing acts
then form the speech acts in the broad (usual) sense of the term.
- With this we have laid the foundation for a pragmatic recon-
struction of 'intensional meaning': we propose to define as
intensional the symbolic intention of speech acts in the broad
sense.
The defining agreement constitutes a symbolic act s. Naturally
we might in addition to performing s also follow rules when we
actualize s, which would then be called regulative rules.
A symbolic act s has informative meaning with reference to the
conditions which by agreement make the performance of s a
correct one, and it has performative meaning in so far as it has
constitutive pragmatic consequences. Obviously mixed cases are
possible.
But does not this approach lead us into serious methodical
problems? Are we not forced to go back to descriptions or articu-
lations 0 f the agreements constituting certain symbolic acts?
Since these descriptions already make use of language, we seem to
get into an infinite regress or into the same circle which we
criticized with respect to the semantic point of view: For it
appears that we have to use a descriptive metalanguage to learn the
acts of a pragmatically understood 'object language'.
However, a closer look at elementary speech situations shows
that in a pragmatic understanding of language there is no such
circle. That is to say, in these situations we can learn an ele-
mentary symbolic act s by exemplification and we can prag-
matically control our understanding of the agreement which is
intended. We just get acquainted with examples and counter-
examples of correct actualizations of s in an appropriate pragmatic
166 F. Kambartel and H]. Schneider
attributes and that for this reason we would have to expect (in
language) proper names and predicators and the functions of sub-
ject and predicate. Instead, we are proceeding from a pragmatic
understanding towards a reconstruction of these categories.
It hardly needs mentioning that the considerations just given do
not suffice for a full explication of the subject-predicate dis-
tinction. ll For one thing, more has to be said about the specific
kind of ambiguity that is exhibited by the example discussed
above, because one can think of other types of ambiguity where
an addition of a second expression will constitute another
semantic relation, e.g. attributive modification.
Secondly, the utterance-types described so far are intelligible
only in their pragmatic context, i.e. they are not independent of the
situation of their use. This independence will be reached when the
subject-expression is enriched in such a way that it becomes a
'description' in the usual sense: When more than one additional
predicator is used for disambiguation, the context may not be
needed any more.
After that, proper names can be introduced to replace these
descriptions. In this case, their difference to predicators can
clearly be understood as a functional one, which will not be the
case if they are introduced in the same way as predicators are, i.e.
by the use of examples and counter-examples. Context-independent
utterances are statements, and they are needed to proceed now to
logical connection.
7. ARGUMENTATIVE LOGIC
8. CONCLUSION
NOTES
1. This is the traditional approach, taken e.g. by Locke in his Essay [1],
Vol. II, p. 3, and, until quite recently, by 1.1. Katz [2], p. 176. Katz has
later given up this view; cf. his [3], p. 38.
Constructing a pragmatic foundation for semantics 175
BIBLIOGRAPHY
- (7a] The Rules of the Language Game. Times Literary Supplement, Sept.
10,1976.
Stegmuller, W. [40] Remarks on the Completeness of Logical Systems
relative to the Validity-Concept of P. Lorenzen and K. Lorenz.
Notre Dame Journal of Fomzal Logic 5 (I 964): 81-112.
Wittgenstein, L. [6] Philosophische Untersuchungen - Philosophical Investi-
gations. Eds. G. Anscombe and R. Rhees. Oxford 1953.
von Wright, G.H. [12] Explanation and Understanding. London 1971.
- [35] Nomz and Action. London 1963.
'Logique hermeneutique'?
PAUL RICOEUR
Universite de Paris
Mais eIle designe aussi les contenus transmis, dont certains expri-
ment des formes ideologiquement gelt~es de relations inter-
humaines. Si ces reifications du lien social ne peuvent etre trans-
formees que critiquement, c'est parce que Ie discours dans lequel
elles s'expriment marque la dependance entre, d'une part, Ie
langagc ct, d 'autre part, le couple forme par Ie travail et la domi-
nation. L'hermeneutique, ici, peut etre accusee d'idealisme
linguistique. dans la mesure ou elle ignore ce rapport de dependance
du langage a des forces sociales qui font du discours Ie siege de dis-
torsions systematiques irreductibles a des phenomenes de mecom-
prehension q u 'une comprehension plus penetrante su ffirait a
dissoudrc. L'ideologie, en effet, travaille derriere Ie dos des parten-
aires sociaux. C'est pourquoi sa dissolution requiert Ie detour de
procedures explicatives et non plus seulement comprehensives,
mettant en jeu un apparcil theorique, comme celui de la meta-
psychologic freudienne. qui ne peut deriver d 'une simple extension
dc I'art spon tane d 'in terp retation a I'oeuvre dans Ie discours
ordinaire et dans la con versa tion.
Si donc I'hermeneutique est epistemologiquement limitee -
limitee au groupe des sciences hermeneutiques - c'est parce
qu'elle est philosophiquement limitee par sa meconnaissance du
rapport entre langage et violence, suscite par Ie conflit des forces
sociaies. En ce sens Ie phenomenc ideologique, considere avec ses
variantes individuelles ou collectives, constitue unc experience
limite pour l'hermeneutique.
Le conflit entre hermeneutique et critique des ideologies
doit dre porte a un degre plus profond encore de radicalite. Pour
Habermas Ie defaut principal de l'hermeneutique de Gadamer est
d 'avoir on tologis~ l'hermeneutiq ue; il entend par lil son insistance
sur J'entente. sur l'accord, comme si Ie consensus qui nous precede
erait quelque chose de constitutif, de donne dans l'etre. Habermas
ne peut avoir que mefiance pour ce qui lui paraft etre I'hypostase
ontologique d 'une experience rare, a savoir 1'experience d 'etre
precede dans nos dialogues les plus heureux par l'entente. Mais on
ne peut canoniser cette experience et en faire Ie paradigme de
1',Jction communicative. Ce qui nous I'interdit, c'est precisement Ie
phenomene ideologique. Si J'ideologie etait seulement un obstacle
in terne a la comprehension, une mecompnShensiol1 que Ie seul
exercice de la question et de la reponse pouvait reintegrer, alors on
194 P. Ricoeur
4. LA REPLIQUE DE L'HERMENEUTIQUE
a
n!pondre directement l'objection seion Iaquelle l'hermeneutique
est limitee par un concept etroit de mecompn!hension qui
l'enferme dans un idealisme hnguistique. Pour y parvenir eIle
s'efforce de s'elargir elle-meme en integrant au processus de la
comprehension et de l'interpretation Ie segment explicatif ou
quasi-explicatif deploye par les sciences sociales critiques.
L'argument a lui-meme deux faces. une face polemique et une
face constructive.
D'un point de vue simplement polemique, l'hermeneutique peut
faire etat de la reconnaissance par ses critiques eux-memes du
caractere incomplet de I'explication objective des formes syste-
matiques de distorsion, telle qu'elle est menee par la psychanalyse
ou par Ia critique des ideologies.. De l'aveu meme de Habermas et
de Apel, l'explication doit s'achever dans une nouvelle compre-
hension hautement mediatisee qui releve d'une hermeneutique
elargie et approfondie. Les sciences sociales critiques peuvent donc
avoir l'ambition de porter Ia comprehension au niveau scientifique
en transformant la distance alienante en alienation contr6Iee (kon-
trollierte Verfremdung). Mais cet usage critique de l'objectivatlon
ne sera garde de retomber a l'objectivation dogmatique que si Ia
critique cesse de se concevoir comme simplement antithetique a
tout rapport de caractere traditionnel au passe. Admettre que la
retlexion n'opere que quand elle demasque des fausses pretentions
et detruit Ie dogmatisme de Ia praxis quotidienne, c'est revenir
a l'opposition naIve re<;ue de l'Aufkldnmg, entre raison et prejuge.
Mais cette surelevation de la reflexion, toujours en quete de
relations substantives a dissoudre, et deliee de toute traditionalite,
mhite bien plus l'accusation d'idealisme que l'affirmation de
I'universelle mediation de la comprehension langagiere.
A cet egard Ie cas de Ia psychanalyse est singulier. D'une part,
l'hermeneutique ne peut que ratifier l'interpretation que donnent
Apel et Habermas de la meta-psychologie freudienne, a savoir
qu'elle constitue seulement une quasi-explication de phenomenes.
que ceux-ci sont a decrire comme des cas de desymbolisation
systematique et que la resymbolisation constitue la finalite du
processus entier. Mais, me me interprete ainsi, Ie modele psych-
analytique merite d 'etre critiq ue pour deux raisons. D'une part on
risque de faire porter sur la reflexion critique tout Ie poids de
l'entreprise de dissolution des fausses comprehensions resultant
204 P. Ricoeur
NOTES ET BIBLIOGRAPHIE
Quine's Set Theory and Its Logic (1963). At about the time of its
publication axiomatic set-theory entered a new phase of lively
development stimulated by the sensational proof by Paul Cohen
of the independence of Cantor's Continuum Hypothesis.
Hilbert's grand idea was to exhibit the whole of classical mathe-
matics in the form of an edifice of ax iomati zed and formalized
calculi. His program called for the creation of a meta-mathematics
in which the consistency and other 'perfection properties' of the
mathematical calculi (including formalized logic) could be estab-
lished. Under the impact of Godel's discoveries in the early
1930s the philosophical program of the formalist school became
in part discredited - but from its modifications grew a major part
of what now counts as 'mathernaticallogic' proper: formal seman-
tics and model theory, theory of computability and recursive
functions, theory of automata (mathematical machines). In the
course of these developments, influences from the seHheoretic
trend which grew out of logicism and from the constructivist
trend inherent in intuitionism have to a large extent become
intertwined with the proof-theoretic aspirations of Hilbertian
formalism. - In the paper by Dag Prawitz in this volume the
reader will fmd an evaluation of the philosophic relevance of
contemporary proof-theory.
It is probably right to say that the ideas underlying intuitionism,
more than the foundation programs of Frege and Russell and of
Hilbert, continue to have a seminal influence on philosophical
logic. From the beginning these ideas represented a counter-trend
both to logicism and formalism. (Poincare attacked Russell and
Brouwer Hilbert.) Wittgenstein's criticism in his later writings of
every program for a 'foundation' of mathematics may be regarded
as a continuation of a tradition inaugurated by intuitionism and
semi-intuitionism early in this century. The constructivism of
intuitionist mathematics also has affinities with the thoughts of
Wittgenstein.
In recent years ideas stemming partly from Brouwer and partly
from Wittgenstein have been elaborated for purposes of theories of
truth and linguistic meaning. The most influential representative
of this new trend in philosophical logic and the philosophy of
language has been Michael Dummett.
4. I would regard the revival of modal logic as the philosophically
230 G.H. von Wright
NOTES
SELECTED BIBLIOGRAPHY
Thi~ list includes only a few selectt~d works which are either of value as
surveys or have been of seminal importance to the recent development of the
subject. Some landmarks are also listed from branches of philosophical logic
which are not covered by the survey essays in the present vvlume.
Cohen, P.J., Set Theory and the Continuum Hypothesis. New York 1966.
Dummett, M., Frege, Philosophy of Language. London 1973.
- Elements of Intuitionism. Oxford 1977.
Frege, G., Nachgelassene Schriften und wissenschaftlicher Briefwechsel, 1-2.
Hrsg. von Hans Hermes, Friedrich Kambartel und Friedrich Kaulbach.
Hamburg 1969-1976.
Hintikka, J., Time and Necessity. Oxford 1973.
Lewis, D., Counterfactuals. Oxford 1973.
Mostowski, A., Thirty Years of Foundational Studies. Oxford 1966. Acta
Philosophica Fennica, fasc. XVII.
Plantinga, A., The Nature of Necessity. Oxford 1974.
Prior, A.N., Past, Present and Future. Oxford 1967.
Quine, W. van Orman, Set Theory and its Logic. Cambridge, Mass., 1963.
Wang, Hao, From Mathematics to Philosophy. London 1974.
Wittgenstein, L., Bemerkungen iiber die Grundlagen der Mathematik. Frank-
furt 1974. (Remarks on the Foundations of Mathematics, revised and
enlarged edition, Oxford 1978.)
Modal and Many-Valued Logics. Acta Philosophica Fennica, fasc. XVI.
Helsinki 1963. Contains several papers which have been of seminal
importance to later developments - by Hintikka, Kripke, Montague, and
others.
Logic and Philosophy. Proceedings of the Colloquium of Institut Inter-
national de Philosophie in Dusseldorf 27 August - 1 September 1978.
The Hague 1980. Contributions by E. Agazzi, R. Barcan-Marcus, M.
Dummett, R. Hilpinen, R.M. Chisholm, D. Prawitz, W. Stegmiiller,
P. Suppes. Supplements the picture of recent work in philosophical
logic given by the survey articles in the present volume.
Philosophical aspects of proof theory
DAG PRAWITZ
University of Stockholm
in the end that makes an inference evident. This analysis forms the
basis of Gentzen's other results and has been a main stepping-stone
in present-day proof theory.
Gentzen's main idea is to break down the inferences into two
kinds of steps, called introductions and eliminations, where each
step involves only one logical constant. Each rule is here such that
one cannot imagine other more basic rules in terms of which the
rule would be derivable. The rule for the introduction of sentences
of a given form states a condition for inferring sentences of that
form. The rule for the elimination of sentences of a given form
states what direct conclusions can be drawn from premisses of
that form. Furthermore, the rule for elimination is the inverse of
the corresponding introduction rule in the sense that a proof of
the conclusion of an elimination is, roughly speaking, already
available if the premiss of the elimination is inferred by an intro-
duction. This fact has later been called the inversion principle.
As a typical example, an introduction of an implication A ~ B
has the form
B
A~B
A
A
B
A A-~ B B
B
proof without introducing any sentences in the proof that are not
used in building up the conclusion of the proof.
logic (or simple type theory), the first step of the extension is
straightforward (in analogy with first order), but the proof of the
normal fonn theorem (for a long time known as Takeuti's con-
jecture) turned out to be more difficult. It was solved by Tait
[24], Takahashi [25, 26], and Prawitz [27, 28] by utilizing a
certain correspondence between the c:llculi of sequents for classi-
cal logic and Tarskian truth definitions (see further section 4.2
below). Extensions to intuitionistic logic of second and higher
order were obtained by Prawitz [29] and Osswald [30], respec-
tively. However, all these results for impredicative systems showed
only the existence of a normal form for the proofs in question and
did not establish the termination of any particular series of reduc-
tions. This situation drew attention to the difference between such
a normal form theorem and a normalization theorem that estab-
lishes the termination of some reduction procedure; for the predi-
cative systems, this latter and stronger result had usually been
implicit in the proofs.
For first order Peano arithmetic, Gentzen [31] had in fact
(although he did not formulate it in these terms) obtained a
normalization theorem for proofs of numerical equations. An
extension to all proofs in the system was obtained by lervell
(32]. More generally, Martin-Lor [33] analyzed intuitionh;tic
proofs with ind uctive and iterated inductive definitions into
introduction and elimination inferences and proved a corre-
sponding normalization theorem. This result is of special interest
for two reasons: it extends the idea of introductions and elimi-
nations to apply not only to logical constants but also to atomic
predicates, and it allows a constructive development of a con-
siderable part of mathematics. The normalization theorem for this
system depended among other things on two essential discoveries,
namely (1) of the correspondence between proofs and certain
functionals, which was to be expected in view of certain
intuitionistic theones about proofs (see section 4.3 below), and
(2) of the resulting possibility to carryover to proofs a certain
notion of computability introduced earlier by Tait [34] for a
system of functionals under the name convertibility. This allowed
Martin-Lof to formulate a fairly general method of proving
normalizability via proofs of computability.
A considerable extension in a quite different direction but also
246 D. Prawitz
4.2. Although Gen tzen [17] arrived at his calculi of seq uents by
reflecting upon the systems of natural deduction and, in the case
of classical logic, generalizing the deducibility relation by allowing
several sentences taken disjunctively as conclusions (which to his
surprise, as Gentzen notes in his paper [31], removed certain
problems connected with classical negation), there is an equally
interesting aspect of the calculus of sequents for classical logic,
namely the close correspondence between its rules and the clauses
of the Tarskian truth definition, which was perhaps first explicitly
Philosophical aspects of proof theory 247
4.3. While the calculus of sequents for classical logic lends itself
to establishing a link between proof theory and classical semantics
(model theory), the system of natural deduction is especially
suitable for certain discussions of the semantics of intuitionistic
logic.
Intuitionistically, the logical constants are often interpreted in
248 D. Prawitz
4.4. Since the discussions in this area have not yet resulted in any
definite or generally accepted viewpoints, one cannot expect a
252 D. Prawi(z
survey like the present one to do full justice to all the participants,
i.e. I want to make the reservation that further developments will
soon show that the accents should have been put in different
places or that the whole picture should have been painted differ-
ently.
In conclusion, I also want to mention that other surveys of the
topics discussed here have been made in papers by Kreisel [62],
Prawitz [5], and Troelstra [43] in the last decade, that a mono-
graph by Cellucci [7] dealing with certain parts of the subject has
recently appeared, and that Gentzen's collected works have been
available in an English translation by Szabo [70] for some time
- for an extensive review of that volume, see Kreisel [71].
(1) For each proof pinT and for each real sentence A in T:
if p is a proof of A in T, then A is true.
The sentence (I) is of course not a real sentence ofT, but it is not
required that the demonstration of the truth of the provable real
sentences of a theory T is carried out in the real part of T itself.
By introducing a function g that assigns Gadel-numbers to the
objects of T and by defining a relation Pr and a property Tr to
hold for Gadel-numbers when the corresponding objects of T
satisfy the relation 'proof of and 'true', respectively, sentence (I)
can be written
(I') For each proof pinT and for each real sentence A in T:
Pr(g(p), g(A)) ~ Tr(g(A».
Hilbert's program, we should finally note that when the real sen-
tences are delimited in Hilbert's way, the consistency problem
happens to be equivalent to Hilbert's program. Trivially, (l)
implies consistency. Conversely, assume that T is consistent and
that a real sentence A has been proved in T. If A is a quantifier-
free sentence and 4ence decidable, then A must be true because
otherwise I A would be true and the completeness of T with
respect to decidable sentences would imply that also IA was
provable, contrary to the consistency of T. If A is of the form
VxB(x), then B(n) would also be provable for each numeral n,
and hence, by what has just been said, B(n) must be true for each
n. But this isjust what is required for the truth ofVxB(x).
This equivalence is in effect noted by Hilbert, e.g. in [74], but
was not enough focalized. With Kreisel, who has especially stressed
the equivalence, e.g. recently in [11], one must say that the
reason for this is insufficient discussion of the significance of
Hilbert's project, which was usually just stated as the task of
proving consistency. In particular, it is important to note that
Hilbert's way of drawing the demarcation line between the real
and ideal sentences is crucial for the eq uivalence. As noted above,
existential sentences 3xA(x), where A(x) is quantifier-free, can as
well as the universal ones be interpreted constructively and hence
counted as real sentences. But then the equivalence between the
consistency problem and Hilbert's program fails: nothing prevents
a sentence 3xA(x) to be provable in a consistent theory T
although A(n) is false for each numeral n. And of course this is
fatal regardless of whether 3xA(x) is read in a finitary or in a
transfinite way; in either case, consistency is no guarantee that
the theory does not contain provable sentences which are simply
false.
We may also note that the terminology 'finitary' and 'trans-
finite' is not very felicitous since the finitary sentences contain
sentences formed by universal quantification over all natural
numbers. Thus, the infinite enters into both the finitary and trans-
finite sentences but in the first case as potential infinity and in the
second case as actual infinity as explained above. What this differ-
ence amounts to is not too obvious; it is one of the main issues
that has to be clarified or perhaps rather reformulated in the
dispute between classical and intuitionistic logic. However, I shall
here follow this terminology of the Hilbert-school.
Philosophical aspects ofproof theory 259
3.2. The position of Gentzen is rather that the ideal sentences have
some kind of meaning, which agrees with the usual way of reason-
ing about them, but that this meaning is somewhat shaky; the
ideal sentences are not yet entirely understood, and therefore one
has to give them a finitary foundation, making sure that they do
not give rise to any inconsistencies. Gentzen [76] says e.g.:
'Indeed, it seems not entirely unreasonable to me to suppose that
contradictions might possibly be concealed in classical analysis."
Although something would certainly be gained by a consistency
proof when one maintains an attitude like Gentzen's, the obvious
weakness with this position is that consistency or, eq uivalentIy,
truth of provable real sentences, is far too little. Since also the
so-called ideal sentences are granted a meaning, even if somewhat
shaky, one must demand of a foundation that it also shows the
results about ideal sentences to be in agreement with this meaning.
And as we have already seen, Hilbert's program does not rule out
that a sentence3xA(x) is provable although A(n) is false for all n.
260 D. Prawitz
(i) For each numeral n, I-G(n); but neither I-VxG(x) nor 1--,VxG(x)
(TO from the premiss 'if A(~) for all ordinals ~ < (3, then A({3)'
infer the conclusion 'A( -y) for all ordinals -y' up to € 0;
3XYx(xEX ~A(x»
(a) Long, tedious proofs that use elementary means are often less
reliable (because of the great risk of error) than short proofs
that use more abstract means and make us understand why
the result holds; hence, the whole idea of increasing the
reliability of abstract principles by a combinatorically com-
plex consistency proof is misconceived.
Philosophical aspects of proof theory 267
Of these arguments, (a) and (b) seem to me rather weak. The prem-
iss of the argument (a) about the need of abstract notions is an
observation that most people in the Hilbert-school would certainly
accept, at least partly, as far as the risk of errors in long combina-
torial proofs is concerned (see e.g. Bernays [87]) - after all,
they wanted to save classical mathematics just because of the ease
with which we obtain short and elegant proofs when using abstract
notions. But of course, the premiss is not a carte blanche for the
use of any abstract principles, and if one really has doubts about
some of them, the use of trusted principles in one consistency
proof can be carefully checked once and for all, even if very long,
and this would then certainly increase our confidence.
As for the argument (b), Takeuti [88] explicitly asserts that his
confidence in natural number theory is increased by Gentzen's
consistency proof. But most people probably agree with Tarski's
[89J often quoted statement in a discussion of Gentzen's use of
Eo-induction: '1 cannot say, however, that the consistency of
arithmetic is now ... more evident than by epsilon .. .'. However,
the reasons why this proof has not increased our confidence in
the theory are clear: firstly, as already pointed out, almost nobody
takes moderate formalism literally, but at most certain classical
concepts are doubted, and then in particular the impredicative
ones of classical analysis; secondly, the doubts there can be about
the use of tertium non datur in first order natural number theory
are expelled by the double negation (or equivalent) interpretations
268 D. Prawitz
given up, and the so-called ideal sentences are admitted to have a
meaning, one tries to give a constructive interpretation of all the
classical notions but rejects what seems to be the more natural
alternative of developing mathematics constructively as far as
possible. Such a position is not very convincing, and it is hardly
surprising that it has not been possible to arrive at a foundation
of mathematics in this way.
This does not mean of course that work on the program cannot
be relevant for other aims. The eliminability of transfinite reason-
ing from proofs of finitary results - to the extent that this can be
demonstrated - is in itself of interest. In general, work along the
lines of Hilbert's program has given us insight in axiomatic ques-
tions about the relation between certain systems; for a survey of
works in proof theory from this aspect, see Kreisel [75].
Indirectly, such results may contribute to a better understanding
of certain foundational questions.
But it seems fair to say that the philosophically most interesting
problems of proof theory now concern certain general questions
about proofs about which there is no general agreement today.
The main philosophical significance of Hilbert's program may be
the discussions that it has given rise to and the resulting doubts
that a satisfying foundation of mathematics can be obtained in
that direction.
Acknowledgement
BIBLIOGRAPHY
'2. The second stage in the evolution of modal logic came with the
extension of MPL to include standard quantification theory (MQL)
and identity (MQL=) as in Barcan (Marcus) [7, 8] and Carnap
[19,20] and higher orders (MQ2 L=) as in Barcan (Marcus) [8].
Theses which held in those original extensions were
2.2 (x)(y)(x=y:JD(x=y»
3. In its third stage modal logic was provided with a very general
model-theoretic semantics within which validity for sentences of
ML, MQL, and MQL= could be defined. The central ideas were
elaborated hy Kanger [72, 73 J, Kripke [80-82J and Hintikka
[63--66] .
Kripke's formulation for the semantics of MPL (here somewhat
revised) begins with a model structure (G ,K,R) where K is a set
of worlds, whi~h has the actual world G as a member, R is a
relation between worlds. A model on a model structure is a
function if; which first assigns a truth value to each atomic sen-
tence at each world. True in a world can be inductively defined
by extending 1>. The clauses for truth functional connectives are
straightforward extensions of standard semantics relativised to
worlds. For the modal operator, 1>(DA ,II) = T just in case 1>(A ,H')
= T for every H' such that HRH'. A sentence is valid just in
case it is true at every world in every model on every model
structure.
The semantic clarifies questions about iterated modalities.
Validity of l.l(b), 1.2(a) and 1.2(b) requires that R be reflexive,
Modal logic , modal semantics and their applications 283
SELECTIVE BIBLIOGRAPHY··
Lewis, c.I., and Langford, C.H. [90] Symbolic Logic, 1st edition, New York,
1932, 2nd edition, Dover, 1959.
Lewis, D. [91] Counterpart Theory and Quantified Modal Logic. Journal of
Philosophy 65 (1968): 113~126.
~ [92] General Semantics.Synthese 22 (1970/71): 18-67.
-- [93] Counterfactuals. Blackwells, 1973.
- [94] Counterfactuals and Comparative Possibility. Journal of Philosophical
Logic 4 (1973): 418-446.
Linsky, L. [95] Reference, Essentialism and Modality. Journal of Philosophy
66 (1969): 687-700. Reprinted in Linsky [96].
_. [96] Reference and Modality*. Oxford University Press, 1971.
Makinson, D. [97] On Some Completeness Theorems in Modal Logic. Zeit-
schrift fiir mathematische Logik und Grundlagen der Mathematik 12
(1966): 379~84.
Marcus, R. Barcan (also see Barcan) [98] Modalities and Intensional Lan-
guages. Synthese 13 (1961): 303~322. Reprinted Wartofsky [149],
and Copi and Gould [25].
- (also see Barcan) [99] Essentialism in Modal Logic.Nous 1 (1967).
~ (also see R. Barcan) [100] Modal Logic. In Klibansky [78],1968.
- (also see R. Barcan) [101] Essential Attribution. Journal of Philosophy
68 (1971).
- [102] Dispensing with Possibilia. In Proceedings of the American Philo-
sophical Association of 1976, 1977. Newark, Delaware.
Meyer, R.K. [103] Coherence in Modal Logics. Logique et Analyse 14
(1971): 658-68.
Montague, R. [104] Syntactical Treatments of Modality. Acta Philosophica
Femzica 16 (1963): 153--168.
~ [105] Pragmatics. In Klibansky [78],1968.
~ [106] Pragmatics and Intensional Logic. Synthese 22 (1970/71): 68~94.
Reprinted in Davidson and Harman [30] .
~ [107 J Forrnal Phiiosophy*. Ed. R.H. Thomason [139] . Yale, 1974.
Munitz, M.R., Ed. [108] identity and lndil'iduation*. New York University
Press, 1971.
~ Ed. [109] LogicandOntoiogy. New York University Press, 1973.
Olson, R., and Paul, A. [110] Contemporary Philosophy in Scandinavia"'.
Baltimore/London) 1972.
Parsons, T. [Ill] Grades of Essentialism in Quantified Modal Logic. Nous 1
(1967): 181-19l.
~ [112] Essentialism in Quantified Modal Logic. Philosophical Review
78 (1969): 35-52. Reprinted in Linsky [96].
Peacocke, C. [113] Proper Names, Reference and Rigid Designation. In
S. Blackburn, Ed., Meaning Reference and Necessity_ Cambridge 1975.
Plantinga, A. [114] The Nature of Necessity. Clarendon) 1974.
Poilock, J.L. [115] Basic Modal Logic. Journal of Symbolic Logic 32 (I967):
355 --365.
Modal Logic, modal semantics and their applications 297
Since its inception in the late 1950s, the possible worlds semantics
of modal logic has been applied with considerable success to
several areas of philosophical logic, for example, to the logic of
knowledge and belief (Hintikka [1-4]); the logic of perception
(Hintikka [5], [4], ch.4,Thomason [6], Niiniluoto [7], Bacon [8],
Smith [9]); deontic logic and the logic of imperatives (Hintikka
[ 10, II], Kanger [12], Chellas [13, 14]); the logic of action
(Kanger [15], Aqvist [16], Porn [17, 18], Lehrer [19]); the
logic of conditionals (Stalnaker [20J, Stalnaker and Thomason
[21], Lewis [22-24], Aqvist [25], Chellas [26]); and the logic
of fiction (Lewis [27]). These applications have illuminated
various problems and puzzles of philosophical logic and analytical
philosophy, but they have also generated new conceptual problems.
Not surprisingly, questions concerning the ontology of possible
worlds have been among the main sources of controversy and
confusion in this area. David Lewis ([23], ch. 4), Alvin Plantinga
[28,29] and Robert Stalnaker [30,31] have tried to clarify the
nature of possible worlds, but their efforts have been criticized
by Brian Ellis [32], Susan Haack ([33] ,[34], ch. 1O),lohnMackie
([35], pp. 88-92) and Lawrence Powers [36] (among others),
who have expressed various misgivings about possible worlds
semantics as a tool of philosophical analysis. The present paper
surveys recent work in the theory of conditionals, where possible
worlds semantics has engendered both logical illumination and
philosophical controversy. Throughout the paper I shall concen-
• Submitted after the introduction was written (Ed. note).
II
Certain inference patterns which are valid for strict conditionals (as
well as material conditionals) fail for subjunctive conditionals.
Perhaps the most obvious example of such an inference is the rule
of strengthening the antecedent:
(3) D(A :J C)
D(A &B:J C)
and
(10) If Bertie were to see the cow-creamer, he would not steal it.
III
if two people are arguing 'If p will q?' and are both in doubt
as to p, they are adding p hypothetically to their stock of
knowledge and arguing on this basis about q.
IV
(15) 'A > B' is true at u if and only if B is true at s(A, u).
(I follow the customary practice of speaking of the truth of a
statement at a world, even though the expression 'true of a world'
would be logically more correct.) If A is an impossible statement,
that is, false at every world accessible from u, S(A, u) cannot be
a possible world. (Here the concepts of possibility and truth are
always thought of as being relative to some world u even when
this is not explicitly mentioned.) In this case Stalnaker takes
s(A, u) to be the absur-l world A at which all statements (including
contradictions) are tru\;. A is not possible with respect to any
other world, and no other world is possible with respect to A- The
sole purpose of A is to permit the evaluation of conditionals with
impossible antecedents: according to Stalnaker's theory, all such
conditionals are true.
Conditionals and possible worlds 305
(16) (i) A is true at seA, u) (for any statement A and world u).
(ii) s(A, u) = A if and only if A is true at no world accessible
from u.
(iii) If A is true at u, seA. u)::;: u.
(iv) If B is true at seA, u) and A is true at s(B, u), then
s(A, u) = s(B, u).
or
(26) If (i.e., even if) I had run, I would not have caught the bus,
and this is not just a matter of ignorance: neither (25) nor (26)
need be true in the situation since my catching the bus would
have been dependent on how fast I would have run, and this is
left undetermined by the antecedent. In such a situation it would
be true to say only:
VI
VII
VIII
(34) A,B
A D-?B
and
(35) (A V B) Q-l- C.
AQ-l-C
of 'If A were the case, B would be the case' would be odd and
unidiomatic, but it should not therefore be regarded as false:
Lewis notes that 'oddity is not falsity; not everything true is a
good thing to say' ([23], p. 28; cf. also Pollock [56], p. 39). In
Lewis's semantics the validity of (34) follows from the assumption
that no other world is as similar to u as u itself is. If the semantics
of conditionals is formulated in terms of selection functions, this
assumption is equivalent to (21.i). If (21.i) is weakened to
and
IX
not the other way round. This suggests that instead of (43) we
should regard its converse
(46) peA V B) ~ PA
entail
(49) Consider a king telling his vassal: 'You may pillage city X
or city Y. But first take counsel with my secretary.' The
second part of this statement makes !t clear that the vassal
should not infer from the first part that he may make his
own choice of city.
322 R. Hilpinen
This explanation suggests that the difference between (44) and the
permission sentence occurring in (49) can be made explicit by
expressing (44) in the form
(52) P'FA,
x
If the truth-conditions of conditional statements are expressed
by means of a selection function, it is easy to see how Kamp's
explanation of the ambiguity of disjunctive permissions can be
adapted to the antecedents of conditionals. If (36) is equivalent
to the conjunction of (37) and (38), its truth-conditions can
(according to definition (20)) be given in the form
butnotequ~a~ntto
and
(57) (If A were the case) or (if B were the case), then C would
be the case,
XI
(64) lA, then BI = the set of all worlds v such that IAlv is in-
cluded in IB I.
and
is a logical truth, but the elimination principle does not hold for
(66) (or for (56)): if the most plausible (Le., selected) B-worlds are
more plausible than the most plausible A-worlds, the former
worlds are also the most plausible worlds satisfying A VB; in this
case f(A V B, u) = f(B, u), and f(A, u) is not included inf(A VB, u).
McKay and van Inwagen's example (41) (with B = 'Spain fought
on the Axis side') illustrates such a case. The semantics outlined
a bove shows how the word 'or' can function as a wide-scope
conjunction in the antecedent of a conditional: if the connective
'or' has a wide scope in the antecedent, as in (57), the conditional
in question implies the conjunction of two simple conditionals.
In natural languages the word 'or' occurring in the antecedent
of a cond itional can nomlally be regarded as having wide scope
(even if the word 'if' has been deleted from the second disjunct);
but occasionally (as in McKay and van Inwagen's example) it has
narrow scope, and then the elimination principle fails. The narrow
scope interpretation can be ruled out by expressing the con-
ditional in form (57).
The preceding analysis of conditionals shows why the principle
of intensionality may appear problematic (ef. Nute [87,67,88]).
If the disjunction in the antecedent of
(69) If the window were open and the door were closed, or the
window were open and the door were open, we would get
some fresh air
(70) If the window were open and the door were closed, we
would get some fresh air,
(71) If the window were open, we would get some fresh air.
appears to state exactly the same thing as (5), and neither (5) nor
(75) is entailed by sentence (4).
These observations suggest that a conjunction in the antecedent
of a conditional should always be regarded as having narrow scope,
or should be interpreted in such a way that it is equivalent to a
narrow-scope conjunction. This feature of conjunctive antecedents
330 R. Hilpinen
(77) 0«* A V * B) J C)
and
Ac knowledgement
BIBLIOGRAPHY
- [93J Deontic Logic and the Semantics of Possible Worlds. In A.G. Conte,
R. Hilpinen and G.H. von Wright, Eds., Deontische Logik und Semantik,
pp. 82-88. Wiesbaden 1977.
Hintikka, 1. [IOJ Quantifiers in Deontic Logic. Societas Scientiarum Fennica,
Commentationes Humanarum Litterarum, 23 (4). Helsinki 1957.
- [1 J Knowledge and Belief' An Introduction to the Logic of the Two
Notions. Ithaca 1962.
- [2J Individuals, Possible Worlds and Epistemic Logic. Nous 1 (1967):
32-62.
- [5] On the Logic of Perception. In N.S. Care and R.H. Grimm, Eds.,
Perception and Personal Identity, pp. 140-173. Cleveland 1969. Re-
printed in 1. Hintikka, Models for Modalities, pp. 151-183. Dordrecht
1969.
- [3] Semantics for Propositional Attitudes. In J.W. Davis, D.l. Hockney
and W.K. Wilson, Eds., Philosophical Logic, pp. 21-45. Dordrecht 1969.
Reprinted in Models for Modalities, pp. 87-111. Dordrecht 1969.
- [11] Some Main Problems of Deontic Logic. In R. Hilpinen, Ed., Deontic
Logic: lntroductory and Systematic Readings, pp. 59-104. Dordrecht
1971.
- [4] The Intensions of Intentionality and Other New Models for Modalities.
Dordrecht 1975.
Humberstone, I.L. [91] Two Merits of the Circumstantial Operator Language
for Conditional Logics. Australasian Journal of Philosophy 56 (1978):
21--24.
lackson, F. [73] A Causal Theory of Counterfactuals. Australasian Journal 0/
Philosophy 55 (1977): 3-21.
Kamp, H. [59] Free Choice Permission. Aristotelian Society Proceedings N.s.
74 (1973-74): 57-74.
-- [97] Semantics versus Pragmatics. In F. Guenther and S.1. Schmidt, Eds.,
Formal Semantics and Pragmatics/or Natural Languages, pp. 255-287.
Dordrecht 1979.
Kanger, S. [12] New Foundations for Ethical Theory. In R. Hilpinen, Ed.,
Deontic Logic: Introductory and Systematic Readings, pp. 36-58.
Dordrecht 1971.
- [15] Law and Logic. Theoria 38 (1972): 105-132.
Lehrer, K. [19] 'Can' in Theory and Practice: A Possible Worlds Analysis. In
M. Brand, and D. Walton, Eds., Action Theory, pp. 241-270. Dordrecht
1976.
Lemmon, E.J., and Scott, D. [83] An Introduction to Moda! Logic. Ed.
K.Segerberg, American Philosophical Quarterly Monograph No. 11.
Oxford 1977.
Levi, I. [44] Gambling with Truth. New York 1967.
Lewis, D. [22] Completeness and Decidability of Three Logics of Counter-
factual Conditionals. Theoria 37 (1971): 74-85.
- (23] Counter/actuals. Oxford 1973.
- (241 Counterfactuals and Comparative Possibility. Journal of Philosophical
Logic 2 (1973): 418--446.
334 R. Hilpinen
1. INTRODUCTION
1.1 Tautological entailment. We turn now and for the rest of this
paper to entailment as a relation between sentences; furthermore,
we address ourselves to the special case when the necessity
·We thank R.K. Meyer for discussions on these topics across the years and
oceans, and S. Kripke for more recent conversations across tea tables.
There are some things one sometimes wishes had never been
invented or discovered: e.g., nuclear energy, irrational numbers,
Entailment and the disjunctive syllogism 341
T
Both VNone
F
Dunn [6]), but we will not go through all that here again. We do,
however, want to focus attention on the definition of "'. Pre-
sumably we do not have to argue for the definitions of - T and
-F; and given what we mean by the four values, we feel the
definitions of -B and -N are almost as obvious. Thus, if the com-
puter (or anyone else for that matter) has been told that A is both
true and false, then there is certainly a point to saying that it has
been told that -A is also both true and false (true because A is
said to be false, false because A is said to be true). And there is a
similar point to saying that if the computer has been told nothing
about the truth value of A, then it has been told nothing about the
truth value of -A either.
Following Meyer, let us call the operation '" de Morgan ne-
gation. Now the problem of this section is that there is another
operation on the four values with some claim to be considered as
negation, again following Meyer, Boolean negation" defined so
it behaves the same as the de Morgan negation""" on T and F, but
so -lBoth = None and ,None = Both. One can imagine motivating
that ,Both = None by saying something like this: if a sentence A
is marked as both true and false, then ,A cannot be marked as
true since in order for this to be. the case it would have to be that
A is not marked as true (but it is). And similarly ,A cannot be
marked as false since then A would have to be not marked as false
(but it is). So ,A must be marked as None. And the reader can go
through similar moves for himself in order to motivate that ,None
=Both.
In a nutshell, the difference between - and, would seem to be
that - is a kind of 'internal' negation, whereas, is a kind of
'external' negation; -A might be read as 'A is false,' whereas ,A
should be read as 'it is not the case that A is true.'
At least two questions force themselves upon us. (1) Which is
the real negation? (2) Even supposing that Boolean negation is not
the real negation, why should we not have it in our logic anyway?
(1) of course drags along with it a subsidiary question about just
what kind of question it is anyway. Is it a question of meta-
physics, linguistics, of logic? We shall not presume to put these
questions to rest here, but we do wish to address ourselves to some
ancillary questions they raise.
344 N.D. Belnap. Jr. and M. Dunn
2.3 Horn 1. That was the bare bones of the dilemma; we now
begin to flesh it out. Taking the first horn then, let us suppose that
the metalinguistic 'not' is de Morgan. Let us symbolize it by -,
letting context determine whether - is meta-linguistic or object-
linguistic. Surprisingly, we can show, given plausible semantical
assumptions,
clauses (t-,) and ([-,)? We can say that (t-,) instructs that -,A
should be marked as t if, and only if, A is not marked as t. But
what does this mean, practically speaking, from an input-output
point of view?
On the input side, does it mean that the reasoner receiving-,A
marked t as input, should 'unmark' A as t (erase any marking of
A as t)? Then what is the reasoner supposed to do when it receives
A &-,A marked t as input? Both mark and unmark A as t (or mark
A as t and then erase)? Trying to both mark and unmark A as t
seems to invite psychotic breakdown (and the parenthetical
alternative of marking and erasing is not without problems: it
would seem that A&-,A would differ from -,A&A, and there are
deeper troubles, to which we shall advert after we discuss the rule
of -,A as output).
Let us think about the conditions under which the reasoner can
produce -, A marked t as output. The reasoner would first have to
verify that it has not been told A, either explicitly or implicitly.
This last is most important, and we have not stressed it previously.
If the reasoner can ultimately deduce A (output A marked t), then
we would not want the reasoner to report out -,A (output -, A
marked t) purely on the basis of his not yet having gotten around
to the appropriate deduction of A. Indeed, -,A is a claim that such
a deduction does not exist. Thus -, would be an 'ineffective' con-
nective in the technical sense, since it is well-known (Church's
Theorem) that there is in general no mechanical procedure for
determining whether such deductions of A exist (at least if quanti-
fiers are present, which we suppose they are in any interesting
case).
Reflecting upon what has just been said reveals new problems
for -,A on the input side as well. Taking it as an instruction to
'erase A' is not really a viable alternative, at least in the absence of
some formal (mechanical?) model of how a reasoner 'takes things
back'. On being told -,A, we do not merely want the reasoner to
'keep quiet about A', even though all its information points to
A's being true. It is still then implicitly told A. What we would
want, we guess, is for the reasoner to correct its information so
that A is no longer deducible. But how is it to do this? We need a
theory of theory correction, and there seems to us to be no such
fully developed theory on the market. Without such a theory -,A
Entailment and the disjunctive syllogism 349
On the other hand, one would have definitely expected that some
additional apparatus beyond the usual two-valued approach was
necessary to do a semantics for relevance logic in a classical meta-
language.
But these considerations do not by themselves dispel the puzzle.
Although one need not give a four-valued semantics in a relevant
metalanguage, couldn't one? The answer seems to be no, given at
least the semantical assumptions we adopted in exploring the first
hom of our dilemma. But why is it no? It seems to us that the
answer is something like the following.
There is no way in a relevant metalanguage (without Boolean
negation) to express that a sentence A takes on, say, just the
value t (and not f as well). The most that one can say is that A is
at least t (t E IAI). The metalinguistic sentence
(tEIAI)&-(fE IAI
does not do the trick it might be thought to do, since (at least
upon our plausible semantical assumptions) it asserts that A is at
least true and it is not the case that A is at least false. But this last
conjunct does not mean that A is really not at least false; all it
means in the end, given our analysis of -, is that A is at least true,
the same as the first conjunct.
Similarly, there is no way to say that A takes on just the value
f - all that one can say is that A is at least f. The two values of the
relevant metalanguage are, as it were, 'at least t' and 'at least f,'
whereas the usual two values of a classical metalanguage are, as it
were, )ust l' and 'just f'
The surprising result (*) above now should look much less sur-
prising. The shock of it was that it appeared to say that A took on
precisely one of the values t, f - that A was at least false if, and
only if, A was not at least true. But now we see that since the 'not'
in question is de Morgan negation, the right-hand side does not
mean really that A is not at least true. All it means is that it is at
least false that A is at least true, which reduces (on our semantical
principles) to the left-hand side - A is at least false. Got it?
Entailment and the disjunctive syllogism 351
The chief topic of this section is to indicate, with all brevity, the
ways in which relevance logic has been enriched by quantifiers,
and, with almost equal brevity, how arithmetic has been enriched
by relevance logic. We offer these indications partly for their
intrinsic interest and partly as preparation for some further dis-
cussions below of relevance logic, relevantism, and the d. s. But
first a preliminary definition.
Kleene's own arguments for them are relevantly invalid. But the
analogy can't really be close in any straightforward sense, because
the extant proofs not only involve steps raising the question of
relevance; they also all involve second order considerations -
quantification over sets of sentences (theories). By so much the
arguments cannot be merely 'relevantized' to become available in
Rlit (which is first-order). We return to this point. The second
alternative is that DS# may be unprovable in R#. In the latter
case, DS# may even be unprovable in p# because, perhaps, of the
second-order considerations mentioned above; but that would not
be of as much interest as if it turned out that DS# were provable
in p# but not in R#.5 Of course, as we said, we don't know
whether or not p# is wholly included in R#; but if it is not, it
would somehow be striking if the arithmetic statement of the d. s.
itself were a witness to the fact. We must, however, record the
sober guess that DS# is not provable even in P#, since (as we said
above) the extant proofs of the admissibility of the d. s. involve
second order considerations. But who knows, the picture might
change again for second order R #: it could be that the extant
arguments can be carried out in second order R # , or it could even
be that DS# is provable in second order p# without being so in
R# .5
5.1 I'm alright, Jack. One might think as follows. The point of
relevantism is taking seriously the threat of contradiction. But
there is in this vicinity (that of fairly low level mathematics) no
real such threat. So here it's O.K. to use the d. s. and conclude
the theoremhood of B. That sounds O.K., but is it? After all, we
suppose that 'here there is no threat of contradiction' is to be con-
strued as an added premiss. But a little thought shows that no such
added premiss should permit the relevantist to use the d. s., for a
very simple reason: as we said, avoidance of the d. s. was bound up
with the l:hreat of contradiction, and one thing that is clear is
that adding premisses cannot possible reduce that threat.lf in fact
the body of information from which one is inferring is contra-
dictory, then it surely doesn't help to add as an extra premiss that
it is not. That way lies madness.
We take the opportunity to point out a disanalogy between the
relevantist situation and the intuitionist situation. The intuitionist
while overhearing with dismay the meanderings of some classicalist
can always say: 'Poor fellow! He actually thinks he is reasoning.
Entailment and the disjunctive syllogism 361
Still, there is some sense that can be made of his musings. What he
seems to be doing is assuming (without warrant) a bunch of
excluded middles. So I can charitably interpret him as con-
structing an enthymematic argument which can be made
(intuitionistically) correct by adding the appropriate excluded
middles as premisses.' The relevantist, as we have seen, cannot
make an analogous charitable interpretation.
5.3 The leap of faith. One tack that has been taken with respect
to induction is just to count it as involving some judgment (not
to be construed as an extra - useless - premiss) that in the par-
ticular case induction is appropriate, and then the making of a
leap to the conclusion, a leap known to be risky, a leap based
perhaps partly on faith as well asjudgment.
Perhaps the relevantist could or should take a precisely parallel
tack, given his proof of the extensional admissibility of the d. s.,
and given the E-theoremhood of -A and A VB. He cannot and
must not count the inference to the E-theoremhood of B as 'good
logic,' but perhaps he could judge that nevertheless it would be
appropriate to leap to the E-theoremhood of B anyhow, even
though (because of the threat of contradiction) he knows the leap
to be risky and hence based partly on faith.
361 N.D. Belnap, Jr. and M. Dunn
5.4 The toe in the water. There is another option, which like the
foregoing pictures the relevantist as at least tempted to jump to
the E-theoremhood of B from the E-theoremhood of --A and
AVB, but with a great deal of hesitancy. This option involves just
a little technicality. Observe first that although the inference
(6) (of § I), from """p and (PVq) to q, is relevantly abhorrent, there
is nothing wrong with the inference (3) (also of § 1), from ""P and
(PVq) to qV(p&""p) - i.e., to 'q, unless there is something awfully
wrong in our information about p.'
This suggests the technical maneuver of introducing a propo-
sitional constant f, to be interpreted (in this context - see AB75,
§ 27.1.2) as the generalized disjunction of all contradictions (given
propositional quantifiers, f could be defined as 3p(P&""p». Then
it is easy to see that
5.5 The true rele van tist. The final option which we describe is the
one to which we have become more and more attracted as this
paper has emerged: the true relevantist should not at all even be
tempted to use the d. s. After all, the temptation presumably
comes from continuing to take pVq as some kind of logical link
between p and q - perhaps a weak one, but still more than
nothing. Perhaps the true relevantist should just stop before he
starts, declaring pVq to be no link at all.
Apply this to the case at hand. The relevantist so described
364 N.D. Belnap, Jr. and M. Dunn
NOTES
BIBLIOGRAPHY
To set the scene for some of the subsequent action, let us consider
how (in especially ~lmple cases) truth functional compounds
of unreliable observation reports E j can be more reliable than
anyone is, separately. (This bears on Mary Hesse's [1] 'self-
correcting observation language', which she discusses further in
[2],pp.125-131.)
Suppose that we have n independent reports ±E 1 , ••• , ±En,
each of which is a 'yes' or 'no' answer to the same question,
'Is it the case that H?' (H might say that the accused struck the
plaintiff, or that there was a brown precipitate from a certain
solution, etc.) Suppose, too, that each report has the same prob-
ability a of being positive when H is true, and has the same,
smaller probability (3 of being positive when H is false. Example:
O! = 80% and (3 = 51 %. For any truth functional compound of
these reports, e.g., (E 1 • E 2 )v(E 3 • E 4 ), we have a probability
polynomial hex) which gives the probability that the compound
is true when x is the common probability of truth of the Ej, e.g.
for the present compound, hex) = 2X2 - X4, which maps the unit
interval onto itself with fixed points hex) = x at x = 0, x = .62, and
x = 1: see Figure 1 (adapted from Moore and Shannon [3], p. 194,
Figure 3). This mapping x 4 hex) moves points in the unit interval
away from the middle fixed point, toward the other two, in such
a way as to make the truth value of the compound be a more
reliable indicator of the truth value of H than is the truth value
of any of the Ej, for with H true the probability that the com-
pound is true is h(cx) = 87% instead of 0: = 80%, and with H false
the probability that the compound is false is l-h(~) = 55% instead
of 1-~ = 49%.
In general, reliability is increased when (as in Figure I) we have
h(o:) > 0: and hem < ~ so that the compound is more likely than
any of the E j to have the same truth value as H, be it truth or
falsity. Moore and Shannon [3] show that this sort of reliability
enhancement is obtainable to any degree short of the ideal step
function simply by using enough indicators E j in appropriate
ways, and they show that on the whole, the Utst compounds are
what they call 'quorum functions' of the Ej, i.e., truth functions
of n arguments where for some m between 0 and n, the compound
is true/false if more/fewer than m of its arguments are true.
(See their theorem 2, being watchful for absent-minded inter-
changes of greater and less in the proof on pp. 203-204.) Quorum
functions are optimal in another sense, too: they correspond to
likelihood ratio tests, viz., here, Bayes strategies (see e.g. Chernoff
and Moses [4], appendices Ell and E l2 .)
Hesse ([2], p. 125) accepts the view of c.1. Lewis ([5], Ch. X
and [6], p. 186) that we cannot even have probable knowledge
ani if H is true at s . .
0.1) Ps(Eil ....• Ej )={ ., (11 < ... < 1m)
m {3m if H IS false at s
0.2)
370 R. Jeffrey
Statements about the past and present are S-determinate, i.e., they
have definite truth values at each stage s of each possible world W,
regardless of how W may evolve after stage s. On the other hand,
future contingent statements need to be evaluated at stages of
worlds: at the present stage of the actual world it may be false
that there will be a female President of the United States in the
twentieth century, but if that statement is genuinely contingent,
there will he a different possible world, agreeing with the real one
up to the present stage, but differing thereafter so that in the
present stage of that world, there is a female twentieth century
President of the United States. Then future contingencies are {r}-
indeterminate: such a statement may have one truth value at
stage r of world W, but have the other truth value at the same
stage, r, of a different world, X. In a third category are the eternal
statements, i.e., statements whose truth values depend on worlds
but not stages: E is eternal iff the truth value of E at stage s of
Choice, chance, and credence 371
Proof We have already seen that (2.3) implies (2.2). Now observe
that by (2.2), crCE-[=PsE]) is the cr-mean ofpt(E-[=PsE]), which
(by (2.4) and the probability axioms for cr ) assumes the value
ptE for t in [=PsE] and assumes the value 0 for all other t. Then
cr(EI [=PsE]) = cr(E - [=PsE] )1 Cr [=PsE] = PsEc r [=PsE] ICr [=PsE] =
PsE, Q.E.D.
Fraasen [14], D. Lewis [15 Land Skyrms [16] treat the re-
lationship between chance and credence much as I have reported
here (but the present representation of propositions as sets of pairs
is drawn from Burgess [17 J with small modifications). The idea
(2.2) that credence ought to be the cr-mean of chance figured
basically in Camap's ([ 18], §41D) early formulation of his ideas
on induction, but Camap was then a frequentist about chance.
Good ([ 19J, p. 41) characterized chances as (subjective) prob-
abilities that 'are independent of any further experiments' and
freely applied (2.2): see also Good [20]. Braithwaite [21] inter-
prets de Finetti's ([ 22] , III) theorem on exchangeability via (2.2),
i.e. (from de Finetti's point of view) backwards. The 'mixed
Bayesianism' of Jeffrey [23] stems from these sources, but with
a different nonfrequentist 'objectivism' about chance. The possi-
bility of reducing chance to credence together with nonprobabil-
istic objective factors has been studied by Jeffrey ([ 24 ], § 12.7),
Mellor [9], Skyrms [25], and David Lewis [15 J. The view that
chances are propensities to produce appropriate relative frequencies
was Pierce's ([ 26], § 644) fallback position from frequentism -- a
position to which many have rallied since its adoption by Popper
[27], [28J. For a useful survey of the following 15 years of work
along such lines, see Kyburg [29].
374 R. Jeffrey
'*
cr([=PrE]/E) = PrEcr[=PrE1/crE and, since crE = 1, cr[=PrE] =
PrEcr[=PrE] or, since cr[=PrE] 0, PrE = 1. Q.E.D.
and the same equation will hold with 'Ps' replaced by 'cs' for one
who knows that the actual outcomes so far have been as in sand
that he is watching a P61ya process with double replacement.
. )
(3 . 12) Cs( 1 next = 21+n(s,i)
+ n(s) Cs (t
h'
IS stage
.
IS
f' 1)
not ma
for one who knows that the actual outcomes so far have been as
in s and that balls are being drawn with single replacement, and
for whom initially, credence about chance was uniform, so that
(3.11) holds.
In these examples we may take S to be the set of all 'words' on
the 'alphabet' A = {a,b}, i.e., S is the set of all finite sequences
of a's and b's, including the empty sequence, e, of length O. The
'product' st of two words is their concatenation, and e is the
identity element, i.e., se = es = s for all sinS. Define ~t as holding
iff su = t for some u in S (with sLt if u =1= e). Define W(s);:: {t: thS}
and set 11"= {- W(S):SES}.
Define the product of sets of words: BC = {st:sEB and t€C}.
378 R. Jeffrey
Here are the credences c(C l lAd = 0.1 and c(C l IA 2 ) = 0.03 are
uniquely determined by equations (4.1), given the fanciful d-values
of Figure 3. (For brevity, the subscript 'r's have been dropped.)
The prima facie difficulty posed here for my [24] account of
decision-making is that the c-values may be as above when Fisher
thinks that abstinence has no prophylactic or therapeutic value at
11 A 2 • C2
10 A 1 ·C 2
9 A I (abstain)
8 A2 (smoke)
7
IOc(C 2 )+1 A2
6
IOc(C 2 ) A,
5
all, in which case I do agree with Nozick and most others that
dominance should imply practical preferability, i.e., here, ration-
ality of smoking despite the lower d-value it gets in (4.1). As
Gibbard and Harper [38] say, the case in which c(C I /A I ) < c(C I /
A2 ) he cause (as Fisher thought) abstinence is a sign offreedom from
the unalterable genetic makeup in the presence of which the
chance of cancer is high, is a case in which Fisher's preference for
AI over A2 as a news item (Jeffrey [24] §S.7) or as an item in
his biography is compatible with his preference for A2 over Al
as a course of action.
I call this a 'prima facie' difficulty because in a plausible elabor-
ation of my conjecture about Fisher's beliefs it is not the smoking
itself but the desire to smoke that he took to shed direct light on
his genotype. Then if Fisher knew (D) that smoking dominated
abstinence in his preferences, and if he saw the further information
that AI (or A 2) is true as shedding no further light on his chance
of getting cancer, we have c(D) = I and so d(Do X) = d(X) and
c(Cj/AjOD) = c(Cj/D = c(Cj). Thus the equations
BIBLIOGRAPHY
Suppes, Patrick, Henkin, Leon, Joja, A., and Moisil, C.R., Ed. [51] Logic,
Methodology, and Philosophy of Science, Volume 4, Amsterdam:
North-Holland, 1973.
roda, Mananao [48] Causality, Conditional Probability and Control. In
Ahmet Ayka~ and Carlo Brumat, Ed.,New Developments in the Appli-
cations of Bayesian Methods. Amsterdam: North-Holland, 1977.
Abbreviations used by some contributors
MT = meaning theory
TM = theory of meaning
OL = object language
IV = the category of intransitive verb phrases
T = the category of terms
TV = the category of transitive verbs
IAV = the category of IV-modifying adverbs
eN = the category of common name phrases
MPL = modal propositional logic
PL = standard propositiona11ogic
MQL = quantified modal logic
I ndex of names