Documente Academic
Documente Profesional
Documente Cultură
BOËR
0. I NTRODUCTION
1. ADEQUACY CONDITIONS
(2) Bel(A, f, R, a1 , . . . , an ).
(In what follows, belief will be our paradigm of a judgment relation.) Now
suppose – as will transpire in the next section – that MRTJ is set out as
a formalized theory in which the general notions of belief, logical form,
and determining a truth condition are taken as primitive, along with some
notation for describing the “logical complexity” of the logical forms of
particular beliefs. A reduction of MRTJ to another theory whose primitives
are taken as antecedently understood will not be adequate for the purpose
of vindicating the former’s notions of logical forms, multiple relations and
the like unless the reduced version of MRTJ provides plausible analyses
of beliefs of arbitrary complexity. One test of the plausibility of such ana-
lyses is their ability to accommodate and explain the intuitive validity or
invalidity of certain inferences about people’s beliefs. It is, e.g., notorious
that the Substitutivity of Identity seems to fail in inferences about psycho-
logical attitudes like belief. The reduced theory should offer some account
of this which is consistent with MRTJ’s commitment to the Substitutivity
ON THE MULTIPLE RELATION THEORY OF JUDGMENT 183
(n ≥ 0);
(P4) for each pure formula φ whose free variables (if any) are
α1t1 , . . . , αktk and each variable δ i foreign to φ, the quatern-
ary predicate [λα1 . . . αk δ : φ]
, of type t1 , . . . tk , i, i, i, i,
which is counted as syntactically simple (hence as not contain-
ing occurrences of α1t1 , . . . , αktk ).
[I] Beli,t1 ,...,tn ,i,i,i,i,t1 ,...,tn (α i , ρ t1 ,...,tn ,i,i,i,i , β1t1 , . . . , βntn ) →
LogicalForm(ρ).
[II] ∼Beli,t1 ,...,tn ,i,i,i,i,t1 ,...,tn (α0i , [λα1t1 . . . αntn δ i : φ & ∼φ],
β1t1 , . . . , βntn ).
αj1 , . . ., αjn are those of the variables α1 , . . ., αn free in ψ, and βj1 , . . ., βjn
are the corresponding terms from among β1 , . . ., βn . Again, under the as-
sumptions (i) and (ii), we might lay down [IV], which (roughly speaking)
requires that when objects are believed to satisfy neither of two conditions,
they should be separately believed not to satisfy each of those conditions:
Similarly, we might stipulate [V], which says (again roughly) that believ-
ing objects to satisfy some condition rules out believing that condition to
be unsatisfied:
No doubt numerous other principles of this ilk could be justified for beliefs
of other basic logical forms, but having a complete list of them is not
important for present purposes.
As for the nature of logical forms themselves, we may lay down at least
this much at the outset. A logical form f is a relational entity that “de-
termines” a particular ontological structure Rf , the latter being a (possibly
unexemplified) formal relation between ordinary properties, relations, and
individuals. The chief constraint on the identity of the structure Rf derives
from the role that a logical form is supposed to play vis-à-vis the condi-
tions for someone’s believing truly/falsely under it: viz., a case of believing
under f is a case of believing truly/falsely under f iff the entities believed-
about are/are not related by Rf . Accordingly, the logical form f must be
distinct from the determined structure Rf . For in a case of believing falsely
under f , the entities believed-about are of course not related by Rf . Yet
even if Rf is not exemplified by anything at all, f must still be the logical
form under which the subject believes, where this naturally suggests that
f (itself a relational entity) must somehow be “exemplified in the belief”.
By this we do not mean that there is an extra entity, “the belief”, which is a
relatum of f , but merely that any case of believing “under” f must involve
f ’s actually relating some items. (As it stands, MRTJ offers no account of
f ’s relata; in our proposed reduction, we shall identify them with certain
188 STEVEN E. BOËR
Now (4), which is univocal with respect to ‘Bel’ and the logical form
assignment in premisses and conclusion, not only is a plausible analysis of
the transparent reading of (3) but also is formally valid, being an instance
of the Substitutivity of Identity. This leaves (5)–(7) as candidates for rep-
resenting the wholly or partly opaque readings of (3) on which it is invalid.
Each of (5)–(7) either equivocates on ‘Bel’ and the logical form assignment
or “changes the subject” by making the statements about properties instead
of (or in addition to) individuals. None of (5)–(7) is valid solely in virtue of
our logical axioms, although this leaves it open that a deeper understanding
of what it is to have a belief “under” a logical form might lead us to accept
one of them. (5), which involves no equivocation, looks promising as a
rendering of the fully opaque construal of (3) and is intuitively invalid.
Here the intuition is that if George IV thinks of Scott as uniquely authoring
Waverly, then the mere fact that Scott uniquely authored both Waverly
and Marmion offers no assurance that George IV also thinks of Scott as
uniquely authoring Marmion. (7) likewise seems invalid. The fact that the
two novels had the same author and that George IV has an identity-belief
that happens to be about Scott and the author of Waverly is insufficient
grounds for concluding that George IV thinks of Scott as the author of
Marmion (or “as” of any other kind, for that matter). The intuitive status
of (6) is less clear and will be left to our reduction to settle.
(D3) FlsBeli,t1 ,...,tn ,i,i,i,i,t1 ,...tn =def [λα i ρ t1 ,...,tn ,i,i,i,i β1t1 . . . βntn :
Beli,t1 ,...,tn ,i,i,i,i,t1 ,...,tn (α, ρ, β1 , . . . , βn ) &
∼|ρ|(β1 , . . . , βn , α)].
Clearly, however, the definition schemes (D2) and (D3) will not sup-
ply what Russell wanted of MRTJ, viz. (stratified) notions of truth and
falsehood for beliefs that will serve in place of the traditional notions of
propositional truth and falsehood that he jettisoned along with the propos-
itions themselves. The obvious problem is that infinitely many truths and
falsehoods may go unbelieved, so that there will not be enough surrogate
“belief facts” to go around (hence not enough real cases of believing-truly
or believing-falsely for (D2) and (D3) to take up the slack). However,
even where there is no fact that someone believes, say, that Rab, there
is still the corresponding relation between a thinker x and a logical form f
that consists in (i) x’s bearing the appropriate multiple relation to R, a, b
under f and (ii) f ’s being such-and-such a logical form h. Taking h to be
[λF i,i y1i y2i d i : F (y1 , y2 )], this relation between thinker and logical form
would be the complex relation Bh:R,a,b defined by (D4):
In sentences of the sort ‘The belief that a bears R to b is true’, we can now
think of (11) as para-phraseable by (12), to which the relation Bh:R,a,b is
assigned as referent:
In general, then, for any logical form g t1 ,...,tn ,i,i,i,i and entities e1 , . . . ,
en of respective types t1 , . . . , tn , we will have as a belief-surrogate the
complex relation Bg:e1 ,...,en defined by (D5):
(D6) For any term ρ of type t1 , . . . , tn , i, i, i, i and terms
α1t1 , . . ., αntn :
Truei,t1 ,...,tn ,i,i,i,i (Bρ:α1 ,...,αn )
=def (∃β i )|ρ|(α1 , . . . , αn , β),
with ‘Falsei,t1 ,...,tn ,i ’ being defined as ‘∼Truei,t1 ,...,tn ,i ’. In other
words, a belief-surrogate Bg:e1 ,...,en is “true" just in case the value of g
actually relates e1 , . . . , en to some object w. Trivially, we will have as
theorems all instances of the corresponding truth scheme (T2), and from
the instance of (T2) for the particular belief-surrogate Bh:R,a,b discussed
above, theorem (T3) will follow:
(13) The belief that Desdemona loves Cassio is true. [I.e.: Beliefs
under the logical form of a binary predication involving loving,
Desdemona, and Cassio respectively are true.]
These results, together with the fact that MRTJ treats the logical com-
plexity of a belief’s content as a function of its logical form, ensure that
the foregoing account of truth for belief-surrogates applies not just to the
“atomic” ones but to the “molecular” ones as well.
3. THE THEORY I
Thus, since ‘f(x)’ is the shadow of ‘F (x)’, under [‘f’, ‘x’; ], it follows
from (A2) that if f, x, d are distinct variables of respective types i, i and
i in y’s Mentalese, (d foreign to «f(x)»), then y’s Mentalese name «[λfxd:
f(x)]» designates the type i, i, i relation [λF i x i d i : F (x)].
Given the adoption of LOT, it is natural to model the having of beliefs as
the “acceptance” of certain sentences of one’s Mentalese. Since, however,
most of a person’s beliefs at any moment are merely tacit, the correspond-
ing notion of acceptance must not require occurrent tokening of Mentalese
sentences. Accordingly, let us gloss Accepts(x, S)
as x is disposed as
if x inwardly and assertively tokens S
,8 where to token S – a property –
is to produce something exemplifying it. I will of course contain axioms
characterizing this relation. The following, admittedly incomplete list of
axioms (derived, with modifications, from Loar (1981: 72)) may serve to
convey the flavor of what would be involved:
(16) [λF i x i d i : F (x)] =def [λν i,i,i y i zd i : ν = [λF i x i d i : F (x)]
& (∃w){NameType(w, i, i, i, My ) & DesIn(w, ν, My ) &
[λrs i u: (∃f)(∃x)(∃d)(Distinct(f, x, d) & VarType(f, i, Ms ) &
VarType(x, i, Ms ) & VarType(d, i, Ms ) & ∼OccursIn(d, «f(x)»)
& r = «[λfxd : f(x)]») & u = «r»)]wyz}].
From (16) we learn that what MRTJ calls ‘[λF i x i d i : F (x)]’ – alias ‘the
logical form of a (first-order) unary predication’ – may be identified with
a certain complex relation R between a structure ν, person y, expression
z of My , and arbitrary individual d. In effect, Rνyzd obtains just in case:
(a) ν is the structure ‘[λF i x i d i : F (x)]; (b) for some distinct variables f,
x, d of respective My -types i, i, i, the term «[λfxd : f(x)]» designates the
structure ν in My , and (c) z is the term «[λfxd : f(x)]» of My .
According to MRTJ, [λF i x i d i : F (x)] determines the ontological
structure [λF i x i d i : F (x)]. But by (16), [λF i x i d i : F (x)] “contains”
[λF i x i d i : F (x)] in the sense of being a relation that requires its first term
to be that very structure. This suggests a general definition of determining
via “containing”, which is spelled out in (D8) with the aid of the predicate
‘SynRel’ expressing the higher-order property of being a syntactic relation
among expressions of a person’s Mentalese:
In light of (A1), (D8), and (D9), MRTJ’s axiom schemes [VIII] and [VI]
are theorem schemes of I+ .
Having embraced LOT, we shall naturally wish to speak not only about
a logical form determining a structure but also about its being a logical
form of something – viz., someone’s Mentalese sentence. While it would
be possible to define a relation FormOf that would apply to any Men-
talese sentence, our project here requires only a more restricted version
that applies to “pure” Mentalese sentences, i.e., those all of whose simple
terms are variables. Accordingly, we schematically define FormOf as the
following complex relation:
The idea about ‘Bel’ we want to articulate in I+ is that the truth of (18)
requires something like the following: (i) that Othello assertively tokens in
his language of thought some sentence S comprising names of loving, Des-
demona, and Cassio; and (ii) that these names are syntactically arranged in
S in exactly the way dictated by [λF i,i y1i y2i d i : F (y1 , y2 )], (so that S will
be a relational sentence of Othello’s Mentalese whose binary predicate,
first argument, and second argument are respectively the aforementioned
names of loving, Desdemona, and Cassio). This would be straightforward
were it not for a difficulty that arises in trying to generalize the account to
cover “nested” beliefs.
A purported vindication of MRTJ must surely take seriously its guiding
idea – that belief is a multiple relation involving thinker, logical form,
and the various entities thought about. If so, however, then that guiding
idea should be applied across the board, to any proffered analysis of belief
ascriptions in Mentalese as well as in English! (This is part of the point of
(C5), which bars appeal to any unanalysed Mentalese analogue of a ‘that’-
clause construction.) Now MRTJ analyses an iterated belief ascription like
(19) by something of the form (20):
In (20), however, there are two logical forms to contend with – f1 , which
determines how Bel1 relates Iago to all the other constituents, and f2 ,
which determines how Iago thinks of Bel2 as relating Othello, loving,
Desdemona and Cassio! In other words, logical forms must be capable
of figuring among the objects of a belief as well.
The difficulty facing the attempt to generalize the account to cover (20)
is this: it is not enough merely that Iago should accept a Mentalese sentence
whose structure is dictated by f1 and whose terms respectively designate
believing, Othello, loving, Desdemona, Cassio, and a certain logical form
f2 (for Othello’s belief). Rather, Iago must represent the logical form of
Othello’s alleged belief in some “canonical” way that reveals to Iago the
conditions under which Othello’s belief would be true, i.e., puts Iago in
a position to appreciate those truth conditions (if he considers the matter,
is smart enough, etc.). The need for such canonical representation of the
logical forms of others’ beliefs is intrinsic to any view of thinking as a kind
of “inner speech”. This is most easily seen when we pretend that the inner
202 STEVEN E. BOËR
For types t 5 not of the sort t1 , . . .tk , i, i, i, i, canonical naming may
be equated with ordinary designation, as in (D12), since no type t 5 entity
could be a logical form:
(D12) CNOi,t
5 ,i 5 5
=def [λwx t y i : DesIn(w, x t , My )].
ON THE MULTIPLE RELATION THEORY OF JUDGMENT 203
(D13) Beli,t1 ,...,tn ,i,i,i,i,t1 ,...,tn =def [λx i f t1 ,...,tn ,i,i,i,i y1t1 . . . yntn :
(∃S)(∃v1 ) . . . (∃vn ){Sentence(S, Mx ) & VarType(v1 , t1 , Mx )
& . . . & VarType(vn , tn , Mx ) & Distinct(v1 , . . . , vn ) &
(∀w)(FreeIn(w, S) ≡ (w = v1 ∨ · · · ∨ w = vn ) &
FormOf(f, S, x) & (∃b1 ) . . . (∃bn ){NameType(b1 , t1 , Mx )
& . . . {NameType(bn , tn , Mx ) & CNO(b1, y1 , x) & . . . &
CNO(bn, yn , x) & Accepts(x, S(b1 . . . bn /v1 . . . vn ))}}].
(n ≥ 0)
(D14) BeliefPredn =def [λbtx x: (∃tx1 ) . . . (∃txn )(t = i, t1 , . . . , tn , i,
i, i, i, t1 , . . . , tn & (∃e)(Expression(e, Mx ) & b = «et »
& ∼(∃d)(∃tx0 )(Expression(d, Mx ) & e = «dt0 »)) & R(x, e,
t1 , . . . , tn )))]. (n > 0)
(25) For any y1 , . . ., yn : being a u such that τ (u, [λν1 . . .νk δ : ψ],
y1 , . . ., yn ) = being a u such that β(u, [λν1 . . . νk δ : ψ],
y1 , . . ., yn ) where ψ(y1 , . . ., yn /ν1 . . .νk ).
Since we have neither appealed to propositions nor made use of any un-
defined predicates of a language that take nominalized sentences of that
language as arguments, condition (C5) has clearly been met. Moreover,
the definition schemes (D7)–(D13) provide a precise ontological account
of the “logical forms” posited by MRTJ and reveal (via the systematic
connection with the single relation ‘Acceptsi,i ’) what is structurally
common to the members of the family of differently typed belief relations
in virtue of which they are “belief” relations; so (C2) and (C3) are satisfied.
What remains to be shown is that (C1) and (C4) are met as well.
Let us begin with (C1). There are two issues here: viz., (a) whether
MRTJ, understood via the reduction to I+ , can provide analyses of ‘A
believes that p’ and ‘The belief that p is true/false’ for all grammat-
ically admissible replacements of ‘p’; and (b) whether the proffered
analyses are plausible. The answer to question (a) is clearly affirmat-
ive. Since logical forms themselves can be of arbitrary truth-functional
and/or quantificational complexity, and since according to MRTJ all of
a judgment’s structure derives from its logical form, it should be obvi-
ous that MRTJ is not restricted merely to “atomic” judgments but can
provide analyses of belief ascriptions with content-clauses of any degree
of complexity.14 Question (b) is not so easily settled. Even assuming, as we
have, the acceptability of the general framework of LOT, we can only argue
from representative examples in which the analyses of belief ascriptions
provided by the reduced version of MRTJ can be seen to plausible.
Let us begin with (18), an ascription to Othello of the “atomic” belief
that Desdemona loves Cassio. MRTJ’s analysis of (18) is (26):
(26) Beli,i,i,i,i,i,i,i,i,i, (Othello, [λF i,i x i y i d i : F (x, y)], Loves,
Desdemona, Cassio).
(29) Bel(A;
[λF x1 . . . xn Gy1 . . . yk d i : F (x1 , . . ., xn ) ∨ G(y1 , . . ., yk )];
R, a1 , . . ., an , S, b1 , . . ., bk ).
Now by (D7) the logical form [λd i : (∀ν i )(∃F i )F (ν)] is the relation
(38) True([λui f : f =
[λH t2 zi Gi,i,i,i,i J i,i x i y i d i : H (z, G, J, x, y)] & Belt1 (u, f ,
Belt2 , Othello, [λF i,i x i y i d i : F (x, y)], Loves, Desdemona,
Cassio)]) ≡ Belt2 (Othello, [λF i,i x i y i d i : F (x, y)], Loves,
Desdemona, Cassio).
And (38) follows from the corresponding instance of theorem scheme (T2)
in which g is the logical form
and e1 − e6 are respectively the relation Belt2 , Othello, the logical form
[λF i,i x i y i d i : F (x, y)], the relation Loves, and the individuals Desde-
mona and Cassio.
By forcing Mentalese to incorporate the kind of multiple rela-
tion treatment of belief ascriptions found in MRTJ, our account of
(33)/(34) requires Iago’s Mentalese representation of the logical form
[λF i,i x i y i d i : F (x, y)] of Othello’s belief to be of the canonical sort
«[λv1 v2 v3 d : v1 (v2 , v3 )]» This satisfies our original demand that Iago must
represent the logical form of Othello’s alleged belief in some canonical
way that reveals (to Iago) the conditions under which Othello’s belief
would be true. For suppose that in Iago’s Mentalese O, L, D, and C are
respectively names for, Othello, loving, Desdemona, and Cassio. Suppose
also that Iago’s Mentalese contains predicates B and T (both of Mentalese
type i, i,i,i,i,i,i,i,i, i,i,i,i) for believing and truly-believing re-
spectively. Then it follows from (T7) that Iago accepts (39):
accounting for the fact that certain possible orderings “make sense” but
others do not? Following Griffen (1985), we may call (a) the “Direction”
Problem and (b) the “Nonsense” Problem.
It should be clear that MRTJ, as reduced to I+ , does not suffer from the
Direction Problem. So long as ‘believes’ is translated by an appropriately
typed predicate, the ingredient type theory prevents any permutation of
arguments of unlike type. Moreover, the residual difference within a type
between, say,
and
(44) Beli,i,i,i,i,i,i,i,i,i (Othello, [λF i,i x i y i d i : F (x, y)], Loves,
Cassio, Desdemona)
In I+ , (45) and (46) are both equivalent to (44). And of course both (47)
and (48) are equivalent in I+ to (43):
receives via the reduction an interpretation that not only is intelligible but
is – so we have urged – plausible as well.
In light of our results, it may be hoped that multiple relation theories of
judgment will at least no longer be glibly dismissed as unworkable oddities
but will instead be accorded the respect they deserve. For whatever the
ultimate fate of multiple relation theories may be, we have seen how to
elaborate at least one of them in a way that renders it just as viable as many
of the more “standard” theories that have paraded through the literature
since Russell’s day.
ACKNOWLEDGEMENT
NOTES
13 As is customary, we pretend that the ingredient proper names designate real individuals.
14 For ease of exposition, we have restricted attention to replacements of ‘p’ whose logical
forms could be formalized in the base language L. But it should be clear that the addition,
e.g., of tense and modal operators would be straightforward, so that L could be brought
ever closer to full-blown (albeit regimented) English.
15 Of course, some sentences are called nonsensical for reasons that have nothing to do
with their type-theoretical well-formedness. Thus, e.g., if numbers are entities of type i
and having a beard is a property of type i, type restrictions cannot rule out such Rylean
“category mistakes” as ‘The number 6 has a beard’. Such residual oddities, however, are
perhaps best viewed not as literal nonsense but merely as gross absurdities – particularly
blatant necessary falsehoods.
REFERENCES
Crimmins, M.: 1992, Talk About Beliefs, Bradford/MIT Press, Cambridge, MA.
Fodor, J.: 1975, The Language of Thought, Thomas Y. Crowell Company, New York.
Fodor, J.: 1987, Psychosemantics, MIT Press, Cambridge, MA.
Griffin, N.: 1985, ‘Russell’s Multiple Relation Theory of Judgment’, Philosophical Studies
47, 213–247.
Jubien, M.: to appear, ‘Propositions and the Objects of Thought’, Philosophical Studies.
Maloney, J. C.: 1989, The Mundane Matter of the Mental Language, Cambridge University
Press, Cambridge.
Richard, M.: 1990, Propositional Attitudes: An Essay on Thoughts and How We Ascribe
Them, Cambridge University Press, Cambridge.
Russell, B.: 1906–1908, ‘The Nature of Truth’, Proceedings of the Aristotelian Society 7,
28–49.
Russell, B.: 1910, ‘On the Nature of Truth and Falsehood’, in his Philosophical Essays,
Simon and Schuster, New York.
Russell, B.: 1910/1911, ‘Knowledge by Acquaintance and Knowledge by Description’,
Proceedings of the Aristotelian Society 11, 108–128.
Russell, B.: 1912, The Problems of Philosophy, Oxford University Press, Oxford.
Russell, B.: 1918/1919, ‘The Philosophy of Logical Atomism’, Monist 28, 495–527.
Russell, B.: 1919, ‘On Propositions: What They Are and How They Mean’, Proceedings
of the Aristotelian Society, Supplementary Volume 2, pp. 1–43.
Russell, B.: 1992, E. R. Eames and K. Blackwell (eds), Theory of Knowledge: The 1913
Manuscript, Routledge, New York.
Russell, B. and A. N. Whitehead: 1910, Principia Mathematica, Vol. I, 2nd edn, Cambridge
University Press, Cambridge.
Salmon, N.: 1986, Frege’s Puzzle, MIT Press, Cambridge, MA.
Soames, S.: 1987a, ‘Direct Reference, Propositional Attitudes and Semantic Content’,
Philosophical Topics 15, 44–87.
Soames, S.: 1987b, ‘Substitutivity’, in J. J. Thomson (ed.), Essays in Honor of Richard
Cartwright, MIT Press, Cambridge, MA.
Department of Philosophy
The Ohio State University
Columbus, Ohio 43210, USA