Professional Documents
Culture Documents
, 1977), pp. 61-79 Published by: Philosophy Education Society Inc. Stable URL: http://www.jstor.org/stable/20127017 . Accessed: 01/03/2014 09:19
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Philosophy Education Society Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Review of Metaphysics.
http://www.jstor.org
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
(non-significant statement,
category
as a statement be characterized form, may are mismatched. The and predicate phe arises both as follows. an extension Any and be assigned predicate a sortal
incorrectness may
language
range. The extension of a predicate is the collection of all those entities of which the predicate is true, while the sortal range con sists of all those entities to which the predicate is significantly ap of plicable. When the entity designated by the subject-expression an atomic statement (a predication) falls outside of the predicate's sortal range, the statement is sortally incorrect. The following
are typical examples of sortally incorrect statements.
is blue (Drange);
and "Socrates is a prime number" (Pap); "Quadruplicity drinks pro crastination" (Russell); "This stone is thinking about Vienna"
(Carnap). (1), for instance, is sortally incorrect because the pred
icate "is blue" is not significantly applicable to abstract entities. The theory of relativity is not the sort of thing that could be blue. A wealth of literature has been devoted to the problem of de
veloping ments.1 correct semantic of sortally incorrect state analysis adequate0 is clear, motivation albeit in controversial: sortally an unusual statements to exhibit when behavior appear The other two statements senses in of negation incorrect logically complex are distinguishable statements. statements. when In this the an
is applied
to sortally
paper,
of the 1976 Dissertation essay prize-winning Essay Education sponsored Competition Inc., by The Philosophy Society, of The Review publisher of Metaphysics. 1 recent are: J. N. Three with extensive articles, bibliographies, "A for Semantics Martin, Many-Valued Mistakes," Category Synthese, 31 (1975), pp. 63-83; R. Routley, "The Need for Nonsense," Australasian Journal 47 (1969), pp. 367-384; "A and R. H. Thomason, of Philosophy, Semantic Theory of Sortal Incorrectness," Journal of Philosophical Logic, 1 (1972), pp. 209-258.
This
is the
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
MERRIE BERGMANN
reopen incorrect question statements. semantic analysis of
I are in order. under the rubric few preliminary remarks First, a I intend to exclude related incorrect statements" closely "sortally statements which be of called may "type-errors." Type variety as being errors are statements if treated true or false, may which, A lead to semantic paradoxes. Examples include: "What I am now
saying is false"; "The term 'heterological' is heterological" (Grelling). Unlike type-errors, the statements which I call "sortally incorrect"
do ments choose from not may to type Second, lead to semantic state incorrect paradoxes. Sortally I be odd or deviant, but not paradoxical. For this reason as a phenomenon examine sortal distinct incorrectness incorrectness. I am treating sortally incorrect statements as seman
R. H. Thomason has shown that treating these tically deviant. statements as syntactically deviant leads to the following prob
lem.2 Let us suppose that sortally of incorrect statements are syn
tactically
have say been that
deviant?that
coupled statement (1)
within
the statements
in violation
("the theory of relativity") with a "concrete" predicate ("is blue"), in violation of the implicit grammatical rule that "concrete" predi
cates can expressions. be meaningfully Now consider combined the only with "concrete" subject following statement:
(2)What
I am thinking of is blue.
If I am thinking of the sky, then the statement is sortally correct and true. But consider the occasion on which I utter (2) while On this occasion, (2) would thinking of the theory of relativity.
appear to be sortally incorrect. But then sortal incorrectness cannot
be a purely syntactical matter; the reference of the expression "what I am thinking of" is the crucial factor in evaluating the sortal status
2
in the The grammatical rule suggested below p. 212. Thomason, is an example of a linguist's "subcategorization" rule. See N. Chomsky, Mass.: M.I.T. of the Theory of Syntax 1965), Aspects Press, (Cambridge, pp. 148-160. text
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
incorrectness
II
classical The by now classic defense of bivalent logic?indeed, bivalent logic?as being fully equipped to handle sortally incorrect statements is given by Quine:
. . . there has been a concern among to declare mean philosophers as 'This stone is such rather than trivially false, predications ingless, drinks procrastina (Carnap) and 'Quadruplicity thinking about Vienna' tion' (Russell). Here we witness sometimes revul just a spontaneous a remote project of cutting sion against silly sentences and sometimes down to something like empirical size. But meaningful language since the philosophers who would build such categorial fences are not generally to banish from language all falsehoods resolved of mathe I fail to see much benefit matics and like absurdities, in the partial exclusions for the forms concerned would that they do undertake; remain still quite under control if admitted like self-contradic rather, if one likes). Tolerance of the tions, as false (and false by meaning, ... source of simplicity is a major in theory; and in the don't cares it counts double, instance present sparing us as it does both the and the of them.3 of settling categories respecting The suggestion seems simple enough: the once the deviant statements
to these
for the moment that it is atomic sortally incorrect statements which should be assigned the value false. Once we have made this assign
ment to the sortally The incorrect statements, they are no more obnoxious as
than contradictions
guage. any over, then may
that have
will or
atomic
V. O. Quine, Word and Object (Cambridge, Mass.: M.I.T. Press, to which Quine aludes here is taken 1960), p. 229. The "meaninglessness" to be a result of syntactic non-well-formedness. 4 This is an extension of the Fregean of allowing the signifi method or sortal range, of a predicate to include everything that can cance-range, count as an object. See especially in P. T. Geach "Function and Concept," and M. Black and eds.), the Philosophical Translations (trans, from (London: Basil Blackwell, of Gottlob Frege Writings 1970), pp. 33ff.
3 W.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
64
MERRIE BERGMANN One objection against any bivalent theory dealing with
statements If these
sortal
incorrectness are
is that
be admitted
assigned
into a semantically
wo truth-value
interpreted
system
they should be
systems) or a
(e.g.,
in supervaluational
incorrect
approach.
which we
not preclude us from developing a semantic theory in which they are simply treated as being false. Brushing this objection aside, then, it seems that the conse quences of any bivalent approach should be the decisive factors in accepting or rejecting it. R. Routley offers the following problem:
... a theory which simply reclassified non-significant as inconsistent: since false would be quickly discredited a sentence of definition, is, nonsignificant by negation be for some sentence p both p and ~p would nificant, theory.5 as sentences the classical also non-sig false in the
That is, if "Quadruplicity drinks procrastination" is sortally incorrect On the then so is "Quadruplicity does not drink procrastination." of the latter is the that classical the former, negation assumption
and the are to be designated statements incorrect sortally inconsistent. is rendered classical system that Note that in this quote, Routley makes an important as false, assump
tion: that the classical negation of a sortally incorrect statement is itself sortally incorrect. But if the Quinean theory treats sortally
incorrect sary atomic statements of sortal usual, as false, condition as incorrectness, and will produce a neces falsehood thus, making classical may simply negation a true statement from a false
function
one. No
reference
is made
of any
in the definition
the statement distinction
of classical
negation
other than between
aspect Thus,
5 Routley,
p. 368.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
SORTAL INCORRECTNESS
atomic sortally negation incorrect statements and other non-true statements
65
The distinction
between
internal and
can be seen most clearly negation by an examination such as "un-" or "dis-." with For prefixes negating
ample, from the nontruth of (3) The theory of relativity it can be inferred that (4) It is not the case that the theory of relativity
in classical music,
is interested
in classical music,
is interested
but not that (5) The theory of relativity is uninterested in classical music.
(4) and (5) may be said to express, respectively, the external nega tion and the internal negation of (3). In statement (4), the prefix "it is not the case that" is interpreted as denying the truth of the subsequent statement, while in the case (5) something more seems to be at stake. The assertion appears to entail that the theory of
can have dispositions?a clear relatively a of negation, that takes then, sortally another incorrect statement. sortally absurdity. incorrect It is this statement sort into
of negation
is tied to
a predicate
the predicate
things
is
the predicate could otherwise be significantly applied. The problem with negation, then, is that on one reading (inter nal negation) the negation of a predication can be taken as indicat falls outside of the ing that the denotation of the subject-expression
6
internal I, 46.
and external
negation
may
be traced
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
MERRIEBERGMANN
sortal the predicate's a of incorrect sortally On this range. can statement
not be true. But if bivalent logics only admit of classical negation, then internal negation is not expressible in those languages. Classical
negation discredits statements yields true statements from sortally evaluate incorrect statements
when the latter are treated as being false, and for this reason Routley
semantic as false. approaches which sortally incorrect
But it is not clear that Quine's position cannot be saved. Note that the examples of sortally incorrect statements given by Quine are atomic; these statements are to be admitted into the theory
as falsehoods. The question of the truth-values of non-atomic sortally in Word
Consider
the following
terms
of non-denoting
and Object:
and coped with, such truth-value per gaps can be admitted a But they values. truth of three like best logic by something haps are that promise no as complications remain an irksome complication, in understanding. gain issue merely that these various perplexities Let it not be supposed is neither between what is false and what from a pedantic distinction two cate those would be true nor false. gained by pooling Nothing under gories under the head of the false; for they are distinguished, of the negations contains in that the one category whatever names, of none of its all its members the other contains the negations while Even
members.7
non-truth-functionality
of
internal
serious
three
criticism of bivalent
are needed
values
to make
tinction between
Routley
. . . forces in . . . the addition of the limited [internal ] negation models. Con 3-valued to at least the theories of up terpretations a sort of not is discerned the falsity; just non-significance sequently be split off as a for short?can falsity?nonsense non-significance third value.8
Once
truth-functionally
is admitted, internal negation can operate (taking the third value into the third value). The
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
SORTAL INCORRECTNESS
67
sortally incorrect statements problem is not in deciding whether should be called false, but rather in deciding how many values the (It seems that underlying logic, if truth-functional, must have. Quine and Routley would not be in disagreement on this point.)
Ill
Within
for internal
a three-valued
and external
(truth-functional)
negation are:
semantics,
the matrices
~A T F N
internal
A T F N
external
F T N
negation
F T T
negation
The two matrices may be related by the use of the Bochvar-Frege horizontal h:9 hA T F N T F F
-A is defined as ~hA. With respect to the matrices for negation, the third value may indifferently be interpreted as signifying non
truth signifying The incorrectness). "it is true that. or as semantic internal deviance negation Then case, (in the present incorrect of a sortally the external sortal state
ment
read:
statement A may
sortally ?A, -A incorrect, involves
Where
A
unlike sortal
is
concerning
applicability. But adopting a three-valued semantics has an additional con sequence: the binary logical connectives must be redefined over
9
see H. G. Herzberger, "Truth For a discussion of this connective, in R. L. Martin and Modality in Semantically Closed Languages," (ed.), The Paradox Press, 1970), pp. 25-46. of the Liar (New Haven: Yale University
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
68 the three values.10 matrices is normal by that alternatives normal are varieties
MERRIE BERGMANN
Many for two I shall of these possible; A con of disjunction. statement which gov have
just connective
a compound components
been assigned classical values (T or F), the compound statement itself is assigned the value itwould be assigned by the corresponding
classical matrix. Normality is a minimal concession to classical logic;
it means
been given affect
in the present
to sortally the
the admission
the non-deviant
of these
statements
of the
will
lan
statements normal
three-valued
logic with
all tautologies
far are
itself be assigned
the value N:
AvB T F N
the assignment Letting is semantically deviant, the
I T T T N
F T F N
N N N N
mean of the that the statement
of A7 to a statement
of purity principle acceptance is de statement to saying amounts that if any part of a compound a as whole is the viant compound meaningless), (nonsignificant, This is the most natural rendered deviant. interpretation thereby case. of A7 in the present
Against
where the 10 F.
the principle
components
of purity,
it may
are
be objected
that only
that even
one is as
of a statement
such
an internal negation avoids this problem Sommers by making "Pre his a See on bivalent semantics. and predicates, retaining operation in America in M. Black (Ithaca: Cornell (ed.), Philosophy dicability," I develop semantics The alternative Press, 1967), pp. 262-281. University is defined over all in Section IV is likewise bivalent, but internal negation statements of the language. 11 and its Application D. A. Bochvar, "On a 3-Valued Logical Calculus Matematicheskii Sbornik to the Analysis of Contradictions," n.s., 4 (1938), pp. 287-308.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
SORTAL INCORRECTNESS
signed warrant a classical the value, this one
69
to be sufficient may assignment Take value to the compound. is of whose immediate components logic, the compound statement
of a classical one
In classical
the value T regardless of the value assigned to The principle of uniformity states that connec
matrices: whenever a given Kleene's row (or column) dis
uniform
of a matrix
F, the value junction
has the same entry for both the value T and the value
N will also have that entry. "strong" to this principle:12
conforms
AvB T F
IT T T
F T F
N T N
value the N latter in this case is non appears
N
The most natural truth, rather as interpretation than deviance. shown by the
IT N N
of the In following fact reading
anomalous,
example.13
(6) Either
When the disjunction is interpreted as Kleene's strong connective, (6) will be assigned the value T on Friday and A7 on every other day of the IfN signifies deviance, this will mean that the statement is week. deviant on every day of the week but Friday. Each of the two interpretations of the third value is plausible
as applied statements. to sortally incorrect atomic seems it that for disjunction is chosen, however, When a decision the matrix between
is forced.
H. The
A recent alternative
to the three
Thomason's approach
van to Metamathematics Introduction (Toronto: Kleene, 332-340. Nostrand, 1952), pp. 13 is from but related point, The example, together with a different p. 238. Thomason, 14 here that Martin I should mention contains an alterna Thomason. on H. "two-dimensional" G. based tive semantic Herzberger's analysis of Truth," Journal "Dimensions methods. (See H. G. Herzberger, of 2 [1973], The methodology is 535-556.) pp. Logic, Philosophical own a semantics I of similar have (Sect. my invaluable; analysis given in a forth and which will be presented dissertation IV) in my Ph.D. C.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
70 in the rejection systems sible valuations. The propositional language tive connectives. Define of truth-functionality scheme is roughly with
MERRIEBERGMANN
with as to admis respect Let L be a follows. as primi function
with the atomic statements of L in its domain, and the set of classical The partial function is defined for all and values in its range.
only members language. Let of the class a complete of sortally valuation correct atomic statements of the classical v generated by p be any =
valuation
where A
of L,
atomic
subject
statement,
to the constraint
p(A) v(A).
that
Let
V(p) be the class of all complete the super valuation s(p) generated = T iff = T for v(A) s(p)(A) = F iff = F for v(A) s(p)(A) s(p)(A)
Hence
value.
Then
is undefined
sortally
otherwise.
atomic statement is assigned a truth
no
incorrect
Let us say that a logical connective is truth-functional just in case the truth-values (or lack thereof) of the immediate components of a compound governed by that connective wholly determine the value (or lack thereof) of the compound. A logical system is truth-functional
are truth-functional with all of its connectives when respect exactly seen can be of the system. It to the admissible valuations by examina are the supervaluations where tion that a supervaluational system the admissible valuations is non-truth-functional.15 The divorce from of generating for the method is important, the set that exactly above guarantees
super of classi
cal tautologies will be tautologies over the supervaluations (with T and entailment coincides with the classical relation. designated), is (The proof straightforward). The virtue of the supervaluational approach lies in this fact.
Neither tems. of these For results will, example: coupling in general, hold the three-valued for three-valued for matrices sys internal
I and Two-Dimensional But article, Logic." "Presupposition coming is based on certain will not consider Martin's work here, for his analysis in this section. I am arguing which intuitions 15 against will both be and s(p)(~C) if C is sortally incorrect, s(p)(C) E.g., = T. is undefined, undefined. But while s(p)(CvC) C) s(p)(Cv~
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
SORTAL INCORRECTNESS
negation "internal" together is a proper and Bochvar's "internal" it Mw); disjunction and Kleene's comprise of classical we obtain Bochvar's
71
(call itMs).
Taking T as designated,
inMw
entail
ment relation is a proper subrelation of the classical. Taking both T andN as designated, the sets of tautologies coincide, but the entail
ment The relation same Yet ofMw results the is still a proper hold for Ms.16 subrelation of classical entailment.
supervaluational
concession
to classical
logic
seems
to be
too strong. Note that the results for statement (6) inM s hold also for the supervaluational system. (The statement is true on Friday and
receives no truth-value on other days of the week.) So here absence
of truth-value should be interpreted as signifying nontruth, than deviance. But why should the statement
rather
(7) Either the theory of relativity is interested in classical music or the theory of relativity is uninterested in classical music be true, if neither disjunct is true? Because the statement has the form of a classical tautology, it will be true on the supervaluational
approach even when the negation is interpreted as internal negation.
I fail to see the virtue in retaining the set of classical tautologies, if this is the result. Thomason has shown how to define external
negation statement over the supervaluations; and it seems correct that the
(8) Either the theory of relativity is interested in classical music or it is not the case that the theory of relativity is interested
in classical music
should be considered true. But this because the second disjunct (which involves external negation) is itself true. What we have here is a conflict of intuitions. Three different
criteria may be used for choosing the non-bivalent dyadic connectives,
and each has some intuitive plausibility. First, sortal deviance may be said to be contagious: statements such as (6)would be judged odd on sortal grounds by many native English speakers, and odd on any day
of the week. Second, uniformity is a plausible characteristic; al
though (6) is odd it also appears to be true by virtue of the first disjunct
16 For proofs, consult Chapter Two of my Ph.D. dissertation.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
72
MERRIE BERGMANN
(on Fridays). Third, classical tautologies expressible by statements such as (8) seem true?despite the apparent nontruth of their counter
parts containing internal negation?in this case, statement (7). Note
once again that the interpretation of the third value as indicating deviance in the case ofMw is the most plausible interpretation, while in the case ofMs and the supervaluational approach, the third value
and truth-value gaps are most plausibly interpreted as indicating nontruth.
But here is the key to the conflict: in all three systems, the sortal incorrectness of a statement has been held semantically identi fiable through the statement's truth-value. When this identification
is made, seems problems to be true squeeze is rendered out somewhere: either a statement which nontrue deviance (Mw), or a statement's a passion In the next sec for truth. in the
is obscured
semantically
through
IV
structure, sense.
all connectives
truth-functional
vocabulary:
predicates
each
i p ,r 2
integer
n p y . . .
n, a finite
,r m n
number
of n-adic
(Iwould prefer to restrict the number n to something smallish, but leave it open for generality. At any rate, for any number
n, the number m of n-adic ax,a<?, and Individual Logical constants connectives predicates . . . may be zero.)
punctuation:
-, D,
y,
(, ).
The statements
of Lc are
is an
predicate,
and P
"Meaning
Relations
Among
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
SORTAL INCORRECTNESS
If A and B are statements, then -A, yA, and A D B are
73
state
The basis for the semantics will be a logical space S. S is a non empty set, which we may assume to be the collection of all entities which occur in at least one possible world. (I assume that the domains
of all possible worlds are disjoint. However, non-disjoint domains
may be assumed, provided that each possible entity is included in the extension/sortal range of a given predicate in one world if and only if it is included in that extension/sortal range in every possible world of which it is a member.) The meaning, or intension, of a predicate determines a class of possible entities to which the predicate could I truly apply; or the meaning is that class of possible entities.18 because the if idea is here that those entities say "could," possible
were actual existent entities, the predicate would be true of them.19
Likewise, the sortal range of a predicate includes those possible indi viduals of which the predicate is significant, or sortally correct. Any predicate's intension will be included in its sortal range; the possible entities of which P could be true must be part of P's sortal range. Letting S be the logical space for Lc, define a sortal specification R on S: R is a function defined over the predicates of Lc. For each n-ary predicate P, R (P) is an ordered pair of subsets of Sn, such that the first member of the pair is included in the second. The first member of R(P), P's intension, will be denoted by [RKP)]; the second, P's sortal range, by [R2(P) J. Let refbe a function (reference function) defined over the individual constants of Lc: for arbitrary constant a, ref(a) E S. We may think of ref(a) as the individual
denoted by a. Let M = (R, ref) be a model relative to the sortal
specification R on S.
18 Van Fraassen's com have this ontological theory does not ostensibly But I believe that ontological commitment does squeeze out some so rather as nebulous than leaving intents entities where; only being in the theory, I prefer simply to think of logical space in such a represented can be identified with the intents of the predicates. way that its regions Van Fraassen's account com alternative of logical space is, of course, below. patible 19with the formal apparatus developed I am here taking the sense of predicates, as we ordinarily use them, to be existence-entailing. It is because of this existence-entailment that I say that possible in the logical space would entities be such-and-such of the actual world). See also H. S. Leonard, (as members if they existed and Predicates," Philo "Essences, Attributes, of the American Proceedings 37 (1964), pp. 25-51. sophical Association,
mitment.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
MERRIEBERGMANN
specifications logical space of distinguishing a bivalent language. a reasonable?be provide between internal and ex The internal negation of a
predication involves a derivative affirmation that the denotation of of the predication lies within the predicate's the subject-expression sortal range. in classical "The theory of relativity is uninterested music" is true if: it is not true that the theory of relativity is interested in classical music, and it is true that the theory of relativity is the sort External nega of thing that could be interested in classical music. So where Pa is a tion does not involve this derivative affirmation.
statement duce "~" of Lc, as and "~" stands for internal ?Pa a defined connective), (I will later intro negation will be true exactly when
That is, the referent of a is within P's R\P). refia) E ?2(P) sortal range but excluded from P's intension. Thus, if ref(a)
E R2(P), exactly one of Pa, ?Pa will be true and the other false. On
the other hand, if ref(a) ? R2(P), then both statements will be false. The external negation of a predication is defined only with respect
to the predicate's intension?viz., Pa is true if and only if Pa is
of Pa or ?Pa is that ref(a) E R2(P). This means that ?Pa is true iff Pa is false and Pa is sortally correct. Carrying this idea a bit further, ' ifwe had in the language for each statement B a statement B which
is sortally that B says to range over definition then we correct, ?B all statements: could is true extend iff B the is false above andB'
is true.
B tions
will mean:
of the valua
A bivalent valuation for Lc with fixed S is defined as follows. For each model M on S, v is the bivalent valuation generated by M iff for every statement A of Lc, v(A) E {T,F}; and the following con
ditions are met:
= T iff . . . , (9) v(P?% . . . td (refitj, ref(td) E R*(Pfl = T = T iff = F or v(B) v(A) (10) v(A DB) = F = (11) v(-A) Tiffv(A) . . . , refit,)) E R2(P2(P, . . . ii) = T iff (ref(tt), (12) viyPfa
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
75
= T = T.
connective, in the following sense (noted
= If we let A by van Fraassen). | |R {ref: A is true in (R,ref)}, we find that |-A \Rand \A D B |Rare functions, respectively, of |A |R, no corresponding function for yA and |J5 |Rand A | |R. There is | |R.
The operation bivalent theory. is non-truth-functional, and allows us to develop a
"y" is an unusual
value and sortal status of its immediate substatement(s), and whose sortal status depends on the sortal status of its
immediate substatement(s).
(b)
its immediate
depends ment(s). on the
substatement(s)
sortal status
and whose
immediate
sortal status
substate
(c)
A composite unlimited
forms a statement whose
connective
truth-value
is a connective which
depends on the sortal
truth-value
of its immediate
sortally Let me "-" now correct. some
substatement(s),
of the connectives It forms
and which
is always
is a simple
classify unlimited
connective.
negation is which
always sortally correct (14), and which is true iff the statement negated is false (11). "It is not the case that the theory of relativity
20 comes from A. Pap, The terminology "limited" and "unlimited" Mind 59 and n.s., (1960), p. 54. "Types Meaninglessness," Pap uses these terms to distinguish modes between of negation, where "limited" is internal negation, and "unlimited" is external nega negation negation
tion.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
MERRIEBERGMANN
sortally correct, despite sortal status of the
imbedded statement.
ternal
as a "sortally negation opaque" statement formed does not receive the The of a connective statement "D"
governed
is determined
classically (10). But the statement is sortally incorrect if either of its immediate components is sortally incorrect (13). Thus, "If today is less than ten, then today is Friday" is both true and sortally incorrect
on Friday.
Internal negation
nectives.
is defined with -A
have
con
(17) 1A =?tAD
As consequence, we
v(yA),
negation.
limited
Internal
negation
connective,
for
(22) v(-A)
Finally,
of a statement governed by this connective depends only upon the sortal status of the immediate substatement ((12)-(14)), and the statement formed is always sortally correct (15). "y" is also "sortally
opaque," or a semantic operator.
The way inwhich models have been defined allows for the intro duction of two modes of semantic entailment and of validity.21 The
second correspond to logical, or universal, relations; the first are more
21 This
distinction
is drawn
in Thomason,
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
SORTAL INCORRECTNESS
restricted. A set of statements T sortally entails a statement
77
A
relative to R, T |hR A, if: for every model M relative to R, if the bivalent valuation v generated by M is such that v(B) = T for all then v(A) = T also. A statement A is sortally valid relative PET, to R if $ h | RA. A set of statements T logically entails a statement A, r | \-A, if: for every logical space S and sortal specification R thereon, r |hR A. A. A statement A is logically valid if 4> \\behind the of intuitions and conflict Mw Ms is resolved by Thus,
status. and sortal The present truth-value system may separating in to various Note first that if classical also be related, ways, logic. statement correct for some model M, then the is sortally every atomic
simple limited and composite limited versions of any connectives will have identical truth-conditions on that model. (This is readily veri fiable.) Moreover, the truth-conditions for the simple limited con
nectives are identical with the classical truth-conditions. So on the
hypothesis, the resultant system would differ from classical logic only in the addition of a primitive "verum" operation y :v(yA) = T for allA.
Of correct. true") as (24) TA =??--A. course, So not define all models the simple will make unlimited statement every 'T" connective (read: sortally "is
We may replace every atomic statement A by TA, and define the connectives as in Lc (note that the clause (12) is then inoperative). The resultant system is then classical with the addition of the verum operation. That is, the simple limited and composite limited connec
tives again changeable Note collapse. salva veritate; that for TA and A are that not for it is possible in general inter some v and A,
v(yTA)
= T and = F. v(yA)
V A full quantificational system Lq may also be defined along the same lines. The vocabulary of Lq includes, in addition to the vocabu lary of Lc, the following:
Individual Possible Dyadic Logical variables individual predicate symbol xu x2, . . . wu w2, . . . variables "="
"V."
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
78
MERRIEBERGMANN of
The terms of Lq include the individual constants and both varieties variables. The formulas are defined recursively:
If for some n, tu . . . ,tn are terms and P is an n-adic predicate, . . . tn is a formula. then Pti If ii, t2 are terms then tx = t2 is a formula. are formulas, If A andP then -A, y A, and A D B are formulas.
If A
then
The occurs
statements free.
formulas
no variable
Let S (a non-empty
sortal specification on S, the
Let R be a
is some non
tion defined over the individual constants of Lq: for arbitrary constant = a, ref (a) E D. Let M (R,D, ref) be a model relative to the sortal specification R on S. The definition of the bivalent valuations generated by the model is twofold, since we have introduced quantification. First, let an over on M d the terms of be any function defined assignment function = ref (a), Lq with range S, such that for any individual constant a, d(a) and for any individual variable x, d(x) E D. Relative to each assign ment function d, a satisfaction function sd is defined for the model M:
for every variable formula of either A, sd(A) E {T,F}, and (where t j is a term, and y is a variety)
. . . , . . . ?i) = T iff (26) sd(P^ d(t?) E ?HPj1) (diU), = T = T iff = F or (27) sd(A D B) sd(B) sd(A) = F = Tiff*d(A) (28)sd(-A) = = = (29) sait, d(t2) t2) T iff d(ix) = T iff = T for all sd, (A) (30) sd((Vi/)A) assignment functions on are as same d' M which d except possibly in the value of the
d(y)
(31) s?(y(Pfli (32) sd(y(A Di)) = T (33)sd(y-A) . . . , . td) = T iff (?(ij, dtfi)) E ?W) = = T = T iff sd(yA) sd(yB)
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions
79
= T iff = T for all d' defined as in s* (yA) (30) (35) sd(y(Vy)A) = T. (36) sd(yyA) Finally, define the bivalent valuation v generated by M on S as: for = T for all satisfaction functions = T iff any statement A, v(A) sd(A) = F. The two modes of entailment and sd on M; otherwise, v(A)
validity are defined as earlier.
The reason for the introduction of two varieties of variables is that sortal specifications are then reflected in the language. Since individual variables range only over the domain in the logical space, the relations on the space determined by sortal specifications may not
be adequately reflected In general, bles. the in a language containing of D may be cardinality varia only individual smaller than that of S.
Suppose, then, that for some monadic predicates P and Q, R2(P) C R2(Q). Suppose also that for some D c M, [R2(Q) R2(P)]nD = (?>. Then for the individual variable x, the statement (\/x)(-yPx D -yQx), as well as (\fx)(yPx D yQx), will be true in that model. That
is, as far as the members of the domain are concerned, the sortal
ranges of P and Q seem to be identical. But they are not. The pos sible individual variables come into play here: ifw is a possible indi vidual variable, then (Vw)(-yPw D -yQw) is not true in the model.
Other relations in the sortal specification are also thus expressible. to R, and y
individual variable,
relative
by M. Then = T. n R2(Q) = <f> iftv((4w) (yPw&yQw)) = = = T. R2(Q) i??v((4w)(yPw yQw)) v yQw)) = T, U R2(Q) = S ffiv(Qtw)(yPw Dartmouth
College.
22 as part of my Ph.D. in this essay were developed The ideas contained "A Presuppositional dissertation of Uni Semantic Theory Categories," I should like to acknowledge of Toronto, 1976. my debt to my versity dissertation Dr. B. C. van Fraassen, and advisor, Dr. H. G. supervisor, Herzberger.
This content downloaded from 128.214.2.137 on Sat, 1 Mar 2014 09:19:49 AM All use subject to JSTOR Terms and Conditions