Topic-Neutrality
Roderick Batchelor
University of São Paulo
[email protected]
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
The paper suggests a definition of the idea of topic-neutrality, and indicates some
of the consequences of identifying logicality with topic-neutrality so defined.
In discussions of the distinction between logical and non-logical
notions or entities, the idea of ‘topic-neutrality ’ is often alluded to
as possibly providing a characteristic mark, or perhaps even a defin-
ition, of logicality. Intuitively, a ‘topic-neutral’ notion is a notion
which may occur in propositions about any topic or subject-matter
whatever. But there does not seem to have been any attempt to pro-
vide a precise definition of this intuitive idea of topic-neutrality.
In what follows we will suggest that such a precise definition can be
obtained by building on some definitions of related notions given by
David Lewis (Sect. 1); and will then consider some of the consequences
of identifying logicality with topic-neutrality so defined (Sects 2
and 3).
1. A definition of topic-neutrality
The intuitive characterization of the idea of topic-neutrality may
be put, in a more regimented form, as follows: entity x is topic-
neutral = for every topic M, there is a proposition p such that: p is
entirely about M, and x is a constituent of p. From this a precise
definition would result, if we could give suitable definitions or char-
acterizations of four notions: (1) topic; (2) proposition; (3) entirely
about; and (4) constituent.
Lewis (1988a, 1988b) has suggested definitions of (1), (2), and (3).
A proposition he defines in the usual manner of intensional logic as a
set of possible worlds. Then he defines a topic (in his terminology
‘subject-matter’) in effect thus: set of propositions is a topic =df
the disjunction of all propositions in is necessary, and the conjunc-
tion of any two distinct propositions in is impossible. And where
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
doi:10.1093/mind/fzr013 Advance Access publication 21 April 2011
2 Roderick Batchelor
p is a proposition and M a topic, he defines: p is entirely about M =df p
is a disjunction of (zero or more) elements of M. Here it is actually not
too important whether propositions are really identified with sets of
worlds, so long as there is a one-to-one correspondence between the
propositions and the sets of worlds. But if we do make that identifi-
cation, the notions of topic and entirely about acquire a more vivid
appearance. Disjunction is now union, conjunction intersection, and
necessity and impossibility identity with respectively the set of all
worlds and the empty set. A topic then is simply any partition of
the set of all worlds: i.e. a family of sets of worlds any two of which
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
are disjoint and the union of all of which is the full set of all worlds.
(We depart slightly from customary usage here in that we allow the
partition to contain the empty set.) The members of a partition are
called its ‘blocks’, or ‘cells’. A proposition entirely about topic M is
then simply a union of M-blocks.
The intuitive motivation of these definitions can be seen as follows.
Given any topic or subject-matter in an intuitive sense, there will be a
corresponding equivalence-relation among worlds, namely the rela-
tion of perfect agreement with respect to the topic in question; and
this equivalence-relation then induces a partition of the worlds — that
is, a topic in the sense defined by Lewis. Conversely, given a topic in
Lewis’s sense, M, there will always be a corresponding topic in an
intuitive sense: it can be described, if not more naturally, as the ques-
tion as to which M-block is true. So Lewis’s notion of topic is neither
too narrow nor too wide. (It is of course somewhat too blunt: it does
not distinguish topics which one might in certain contexts wish to
distinguish. But that as we know is the way of intensional logic.) As for
entirely about, suppose first that, intuitively, p is entirely about M: so p
concerns nothing that goes beyond M. Then of course p cannot differ
in truth-value in two worlds which are exactly alike with respect to M,
i.e. which belong to the same M-block; which is in effect to say that p
is a union of M-blocks, i.e. that p is entirely about M in the sense
of Lewis’s definition. Conversely, suppose p is a union of M-blocks.
Then p must be, intuitively, entirely about M: if p concerned anything
beyond M, then two worlds could disagree on what p says about that
further issue without disagreeing on M, and so p would cut through
some M-block after all.
(See Lewis 1988a, 1988b. For some further developments of
the theory of topics, see Humberstone 1996, 2000, and Yablo
forthcoming.)
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
Topic-Neutrality 3
This gives us, then, quite solid definitions of (1), (2), and (3). But
when we come to the fourth notion, namely constituent, a difficulty
arises. For the notion of constituency applies most naturally to
Russellian propositions — structured complexes of a sui generis
sort — and not to intensional propositions, which are somewhat
amorphous; but as we saw it is intensional propositions that figure
in Lewis’s theory of topics. Now in fact an intensional proposition is
not wholly amorphous, especially because the worlds which it selects
are not mere atoms, and so one might try to define some recherché
notions of constituency applicable directly to intensional propositions
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
(cf. Fine 1977); and this then might provide a suitable completion of
the definition of topic-neutrality. This approach might suit an inten-
sional logician averse to Russellian propositions; but obviously it is
somewhat far-fetched, and here we will follow a more straightforward
course: we will assume the legitimacy both of the Russellian notion
of propositions with their constituents and of the intensional notion
of propositions as equivalent to sets of worlds; and from the notion of
constituency for Russellian propositions we will define a derivative
notion of constituency for intensional propositions. (Another possi-
bility, which we will not pursue here, would be to reformulate the
theory of topics in terms of Russellian rather than intensional
propositions.)
For each Russellian proposition r, there is a corresponding inten-
sional proposition [r], tantamount to the set of all worlds in which r is
true; and each intensional proposition corresponds to many different
Russellian propositions. Now, there are two ways in which one might
try to transfer the notion of constituency from Russellian to inten-
sional propositions: the constituents of an intensional proposition
might be defined as the entities which are constituents of every, or
alternatively of some, Russellian proposition corresponding to it. The
definition in terms of ‘some’ is in fact of no interest, because under it
every entity would be a constituent of every intensional proposition.
(Let p be any intensional proposition, x any entity, r any Russellian
proposition corresponding to p, and (x) any Russellian proposition
having x as constituent. Then e.g. r & ((x) _ ‰ (x)) is a Russellian
proposition corresponding to p and having x as constituent.) But the
definition in terms of ‘every ’ is well suited to our purposes. Thus a
constituent of an intensional proposition in this sense is an entity
which occurs in every Russellian proposition corresponding to the
intensional proposition: the intension cannot be ‘expressed’ without
the entity.
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
4 Roderick Batchelor
The similar notion applied to Russellian propositions themselves
is also of interest: we may say that x is an essential constituent of the
Russellian proposition r if, for every Russellian proposition s strictly
equivalent to r, x is a constituent of s. So x is a constituent of [r]
amounts to: x is an essential constituent of r.
Let us give some examples. Suppose that ja and cb are two atomic
Russellian propositions, and that Negation and Conjunction are cer-
tain simple entities which combine with Russellian propositions
to form other, more complex Russellian propositions. Then, for in-
stance, Negation is an inessential constituent of ‰ ‰ ja, since ‰ ‰ ja is
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
strictly equivalent to ja which does not contain Negation; likewise
Conjunction is an inessential constituent of ja & ‰ ‰ ja. But a and j
are essential constituents of ja, since the intension [ja], although it
can be expressed by many different Russellian propositions (ja itself,
‰ ‰ ja, ja & ‰ ‰ ja, ja & (cb _ ‰ cb), etc., etc.), yet cannot (pre-
sumably) be expressed by any Russellian proposition not containing a
or not containing j. Likewise, it is plausible to think that Negation
is an essential constituent of ‰ ja, Conjunction of ja & cb, etc.
(We take the modal independence of atomic propositions as a sort
of default assumption.)
Putting the present definition of constituency for intensional prop-
ositions together with the definitions of topic etc., we then complete
our definition of topic-neutrality :
x is topic-neutral =df for every topic M, there is a proposition
(i.e. intensional proposition) p such that: p is entirely about M,
and x is a constituent of p
Now, curiously, topic-neutrality in this sense turns out to be
equivalent to a much simpler notion, in which the notions of topic
and entirely about no longer appear at all. Let us use the term ‘rigid’ to
mean non-contingent (i.e. necessary or impossible). We can then state
the following criterion of topic-neutrality :
x is topic-neutral iff x is a constituent of some rigid intensional
proposition (or equivalently, iff x is an essential constituent of
some rigid Russellian proposition)
For suppose first that x is topic-neutral. Let us write > for the neces-
sary intensional proposition, and o for the impossible one. Now {>}
is a topic, and the propositions entirely about that topic are > and o
only. Hence x, being topic-neutral, is a constituent of either > or o
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
Topic-Neutrality 5
(or both), i.e. of some rigid proposition. For the converse, suppose x is
a constituent of some rigid proposition, i.e. of > or of o. But for any
topic M, both > and o are entirely about M. Therefore, for any topic
M, x is a constituent of some proposition entirely about M, i.e. x is
topic-neutral.
2. Logicality and its sparseness
So much then for topic-neutrality itself. We turn now to a brief con-
sideration of the prospects of equating logicality with topic-neutrality
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
as here defined.
Since the above criterion has given us a simpler equivalent of
topic-neutrality, let us use that instead and put:
x is logical =df x is a constituent of some rigid intensional prop-
osition (or equivalently, =df x is an essential constituent of some
rigid Russellian proposition)
Thus a logical entity would be an entity which is needed in order to
express >, or needed in order to express o, or both.
Now, the first thing to note about this definition is that it requires
the domain of logical entities to be very sparse.
Thus, in particular, if this definition is correct then there can hardly
be any complex logical entities. It is commonly thought that logical
entities can combine with one another quite freely to produce other,
more complex logical entities. So, for instance, if there is a logical
entity Negation (‰ ) and a logical entity Necessity (h), then there is
a logical entity Non-Necessity (‰ h), a logical entity Non-Necessity of
the Negation (‰ h‰ ), and so on. But this would doubtless lead to
logical entities which are not needed to express > or o. For instance:
surely Triple Negation (‰ ‰ ‰ ) is not needed to express > or o: single
Negation (‰ ) would always do instead. Since Triple Negation, if there
were such an entity, would be as paradigmatic a logical entity as any,
the conclusion here would be that there is no such entity as Triple
Negation — and similarly for other logical complexes. There can still
be, of course, Russellian propositions ‰ ‰ ‰ p (distinct from though
strictly equivalent to ‰ p), but there is no single complex entity Triple
Negation which occurs as a constituent in such propositions.
But if one finds this consequence of the above definition of logicality
undesirable, one can easily avert it. One can take that definition as a
definition of logicality for simple entities; and then define the logicality
of a complex as the logicality of all simples which constitute it.
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
6 Roderick Batchelor
Even so we still have very strict constraints on the extent of the basic
(i.e. simple) logical entities. If some basic logical entities are sufficient
to express > and o (possibly of course with the help of non-logical
entities), then there are no other basic logical entities. So, for instance,
if ‰ and & are basic logical entities then there are no others: for if
p is any Russellian proposition made up entirely of non-logical entities
(e.g. some atomic proposition ja), then p & ‰ p expresses o, and
‰ (p & ‰ p) expresses >. (The same holds of course for Negation
and Disjunction, or for Sheffer’s stroke, etc.) Or again, if Identity is
a basic logical entity then it is the only one: > can be expressed by
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
a = a, and o by a = b (where a and b are distinct non-logical simples).
An even droller example is Contingency (O): > could be expressed by
Oja, and o by OOja. And we may add as a final example — nothing :
if there is at least one necessary Russellian proposition not contain-
ing any logical entity and at least one impossible Russellian propos-
ition not containing any logical entity, then there are no logical entities
at all.
Now, most philosophers today tend to be (perhaps often uncrit-
ically) rather pluralistic about basic logical entities or notions: they
think that these include negation, conjunction, quantification, iden-
tity, necessity, set-membership, and maybe a few more or a few
less. Such philosophers will, I fear, be prone to take the conse-
quences just noted as a reductio ad absurdum of the adequacy of
the above definition of logicality (or at most they might be moti-
vated to find some modification of the definition which avoided the
consequences). However, more reductionistic philosophers may find
these consequences perfectly acceptable. Thus, for instance, in the
context of atomistic metaphysics, the sparse set of primitives ‰ and
& is by no means implausible (see Batchelor 2010 and Batchelor
forthcoming); and of course the radical alternative nothing was the
view (the ‘Grundgedanke’) of Wittgenstein’s Tractatus. Or to give a
less extreme example, we may consider the Priorian conception of
modality (as in Prior 1957). Here propositions like ‰ (ja & ‰ ja)
and ja & ‰ ja are not really rigid in the strong sense of either true
in all possible worlds or false in all possible worlds, because they
are supposed to lack truth-value in the worlds where a does not
exist. To obtain rigid Russellian propositions we might need gen-
erality, as in 8x ‰ (jx & ‰ jx) and ‰ 8x ‰ (jx & ‰ jx); so the lo-
gical simples might well be ‰ , &, and 8 — a plausible-sounding
enough set.
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
Topic-Neutrality 7
3. Logicality and modality
Now, besides these constraints on sparseness, the above definition of
logicality impinges also upon the interesting issue of the connection
between modality and logicality.
One feels that there should be some interesting connections be-
tween, on the one hand, the modal value of a Russellian propos-
ition — whether it is necessary, contingent, or impossible — and, on
the other hand, the nature of its constituents — whether they are lo-
gical or non-logical. More specifically, one feels that rigidity (necessity
or impossibility) should be intimately connected to logical constituents,
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
and correlatively contingency to non-logical constituents. For surely the
source of contingency is always the haps and mishaps of empiricalia,
whereas Logic alone is responsible for necessity (and impossibility).
(Here I suppose I ought to say, ‘pace Kripke & Co.’.)
The most naive attempt to give a more precise formulation of this
connection — naive perhaps almost to the point of childishness —
would be the following equivalence: for every Russellian proposition p:
p is rigid iff all constituents of p are logical
But of course this will not do. It is often said that there are proposi-
tions which involve only logical constituents but which are neverthe-
less contingent: the standard examples are propositions about the
number of things, such as that there are at least five things, understood
in the usual way in terms of quantification and identity. Now these
supposed counterexamples are very possibly misguided: I myself am
inclined to think that they are, and that the right-to-left part of the
above equivalence is in fact highly plausible. But in any case the
left-to-right part is surely untenable. Take any proposition containing
non-logical constituents, say ja; then ja & ‰ ja is rigid and yet not
all its constituents are logical.
Still, surely something can be salvaged even from that left-to-right
part. It must be admitted that a and j occur in the Russellian prop-
osition ja & ‰ ja; but clearly they do not occur there quite so
robustly as they do in, say, ja itself. One might try to explicate this
lack of ‘robustness’ in terms of invariance of truth-value, or invariance
of corresponding intensional proposition, under replacement of con-
stituents. Although I do not by any means wish to deny the general
interest of this sort of idea, still the attempt to apply it to the present
task raises some delicate problems. We need not go into this
matter here; for we have at hand a simpler and less problematic
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
8 Roderick Batchelor
explication: namely one using our notion of essential constituency.
Thus we can give the following refinement of the left-to-right part of
the above naive equivalence: for every Russellian proposition p:
If p is rigid then all essential constituents of p are logical
And indeed the correctness of this principle is an immediate conse-
quence of our definition of logicality.
It should not be thought, however, that the converse of this principle
is also true — i.e. that for every Russellian proposition p, if all essential
constituents of p are logical then p is rigid. This is refuted by the fol-
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
lowing example (a similar example is given in Fine 1977, pp. 179–80).
Suppose p1, p2, p3, … is an infinite sequence of wholly distinct
Russellian atomic propositions. Now let p be the disjunction _i 1
(pi & pi + 1 & pi + 2 & … ). This says that the sequence p1, p2, p3, … con-
tains only true propositions from some point onwards; and that is
surely a contingent proposition. But, for any n 1, p is strictly equiva-
lent to _i n (pi & pi + 1 & pi + 2 & … ); and therefore none of the pi or
their constituents is an essential constituent of p. Thus all essential
constituents of p are logical, and yet p is not rigid.
Obviously this example makes essential use of infinite propositions.
It may perhaps be a plausible conjecture that, for every finite
Russellian proposition p, if all essential constituents of p are logical
then p is rigid.
The above example serves also to refute the natural thesis that every
Russellian proposition has a strictly equivalent ‘canonical form’ in
which all constituents are essential constituents. A further counter-
example to this canonical form thesis seems to be provided by rigid
propositions. We saw that, if Negation and Conjunction are basic
logical entities, then o can be expressed by the Russellian proposition
ja & ‰ ja, and > by ‰ (ja & ‰ ja); and that if Identity is a basic
logical entity then > can be expressed by a = a, and o by a = b; and so
on. Here instead of a and b and j we may of course use other things;
but it seems plausible to think that we have to use something : that
from Negation and Conjunction by themselves, or from Identity by
itself, no Russellian proposition can be constructed at all, and so in
particular no necessary proposition and no impossible proposition.
Thus no rigid Russellian proposition would have a canonical form.
(An imaginative metaphysician might suggest that there is the
empty conjunction: a Russellian proposition consisting of Conjunction
applied to zero arguments. We would then have [&( )] = >,
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
Topic-Neutrality 9
[‰ &( )] = o: thus & would be the only constituent of >, and ‰ and &
the only constituents of o; so that the indicated Russellian propos-
itions would be canonical. And with Identity one could try
> = [Identity is identical with Identity], o = [Identity is identical
with (Identity is identical with Identity)]. But in neither case is the
legitimacy of the relevant Russellian propositions altogether
indubitable.)
Still, we should not be over-impressed by such failures. Rigid prop-
ositions are a rather special case, of familiarly explosive character; and
the infinite proposition described above is obviously an exceptional
Downloaded from http://mind.oxfordjournals.org/ at Duke University on May 6, 2012
anomaly. Perhaps every Russellian proposition which is both contin-
gent and finite has a canonical form. In any case propositions which
have canonical forms seem to be the norm, and the rest exceptions: so
we should expect some restricted version of the canonical form thesis
to be true.
References
Batchelor, Roderick 2010: ‘Grounds and Consequences’. Grazer
Philosophische Studien, 80, pp. 65–77.
—— forthcoming : ‘Some Questions about Attributes’.
Fine, Kit 1977: ‘Properties, Propositions and Sets’. Journal of
Philosophical Logic, 6, pp. 135–91.
Humberstone, Lloyd 1996: ‘A Study in Philosophical Taxonomy ’.
Philosophical Studies, 83, pp. 121–69.
—— 2000: ‘Parts and Partitions’. Theoria, 66, pp. 41–82.
Lewis, David 1988a: ‘Relevant Implication’, in his 1998, pp. 111–24.
Originally published in Theoria, 54.
—— 1988b: ‘Statements partly about Observation’, in his 1998,
pp. 125–55. Originally published in Philosophical Papers, 17.
—— 1998: Papers in Philosophical Logic. Cambridge: Cambridge
University Press.
Prior, A. N. 1957: Time and Modality. Oxford: Oxford University
Press.
Wittgenstein, Ludwig 1922 : Tractatus Logico-Philosophicus. London:
Routledge.
Yablo, Stephen forthcoming : Philosophy of Aboutness. Princeton, NJ:
Princeton University Press.
Mind, Vol. 120 . 477 . January 2011 ß Batchelor 2011
