**Definition 2.23: Stalnaker’s uniqueness assumption: for every world ? and antecedent ?, that is entertainable at ?, there is a smallest ?–admitting sphere around ? containing exactly one**

**3.4 No trouble: the ‘marvellous mountain’ argument is unsound.**

Lewis assumes that all restricting modifiers such as ‘In Australia’, ‘On a mountain far away’ or ‘at world w’ (abbreviated ‘at w’) commute with respect to truth-functional connectives in regimes other than classical ones.

The following objection to Lewis’ ‘marvellous mountain’ argument, which he gave against the existence of inconsistent worlds, is not widely proposed among rival theories to GR.128

There is a tendency among authors working in impossible world semantics, especially those
who endorse non GR approaches to possible and impossible worlds to see the force of Lewis’
*argument stemming from the metaphysical aspect of its concretism – of the assumed nature *
of possible worlds in CT – as being a sufficient condition (and the key culprit), and only few
take issue with the commutativity of the 'at w’ modifier with the truth-functional connectives.
That is, it seems to me that some authors are a little too quick to accept Lewis’ derivation of
the literal contradiction, and blame its success on the GR ontology ― an alleged shortcoming
of GR they like to point out as the fulcrum of that derivation.129 For those authors, granting
Lewis this apparently absurd consequence seems just a little too convenient, and indeed often
serves as their cue to endorse less committal ontologies.130 For example, Nolan (1997, p.541)
thinks that Lewis’ emphatic rejection of impossible worlds largely flows from what he takes
them and their parts to be – namely as existing simpliciter. This, Nolan identifies as reason
enough to derive a contradiction from positing objects with inconsistent properties.

Extending this approach to impossible objects produces literal impossibilities, it seems: if the impossibilium corresponding to the blue swan-and-not-a-swan is literally a swan and is literally not a swan, then a contradiction is literally true. (Nolan 1997, p.541)

128_{ A similar version of the argument can be found in Kiourti (2010, Ch. IV, §4.41). Mares (2004, pp.84-87) sees }
*the problem with this property in the case of negation, which as a matter of fact is the key culprit. Lycan (1994, *
pp.39-41) believes that Lewis’ argument fails, observing that a truly inclusive quantifier would require the
invalidation of the entailment: at w: ~P entails ~(at w: P).

129_{ E.g. (Nolan 1997, p.541), (Jago 2012, p.64), (Vander Laan 1997, p.606). }
130_{ Kiourti (2010, p.102) also makes this observation. }

*But this analysis takes it for granted that Lewis’ reductio argument is sound. That is, Nolan *
*takes it for granted that it’s fine to go from there being some individual that is both a swan *
*and not a swan, to it being the case that the said individual is a swan and it not being the *
*case that it is a swan. However, the applicability of that inference doesn’t rest on the *
*metaphysical nature of the objects whose existence is being posited, but on the logical *
principles thought as correspondingly fitting the metaphysical view, and consequently
employed in the analysis. Vander Laan (1997, p.606) thinks that Lewis’ reasons for rejecting
impossibilia stem from his concretism and his insistence to take ‘at w’ as a restricting

modifier, i.e. as merely restricting quantification over concrete worlds that are said to exist much in the same way as our world.

Lewis goes on to say that ‘at so-and-so world’ is indeed a restricting modifier,
unlike ‘in such-and-such story’, since worlds are like the actual world, not like
stories. It is this last point that is of interest here. Lewis’s reasons for rejecting
impossible worlds stem from his concretism, that is, his view that worlds are
concrete objects much like us and our surroundings. (Vander Laan 1997, p.606)
But this analysis assumes as correct Lewis’ analysis of concrete impossibilia. Lewis’
*rejection of classically impossible worlds does not rely on the metaphysical nature of the *
*objects in the quantifier’s scope, but rather on the logical assumptions he makes about them, *
which are embodied by the posited properties of the ‘at w’ modifier. Vander Laan then
suggests an abstractionist approach that treats worlds more like stories rather than concrete
objects, i.e. where ‘at so-and-so world’ modifier is intended along the lines of ‘according to
such-and-such story’. Such an approach, he observes, would avoid the problem of ending up
in contradiction, since stories, unlike worlds, need not be maximal nor consistent.

*How should we read ‘on the mountain’? Let’s recall that such modifiers act by restricting the *
quantifiers in their scope to the domain of a single world, much in the same way as the ‘in
Australia’ restricting modifier restricts all talk to that which exists in Australia. So, all swans
are indeed black if we restrict our discourse only to what exists in Australia. Lewis assumes
the existence of a domain―the marvellous mountain―where contradictions are true, but for
simplicity we can speak of maximal domains, i.e. worlds, where contradictions are true. The
objection to Lewis doesn’t hinge in any way on such domain generalisation, but rather will
simplify the discussion – it will be a lot simpler to speak of worlds than their subdomains.

The premise that assumes the existence of a mountain such that ‘On the mountain P and not
P’, amounts to assuming the existence of an inconsistent world, since according to Lewis’s
version of GR anything that exists must do so at some world.131 So were anything to exist ‘on
the mountain’, it would exist ‘on the mountain at some world’. So, since we’re considering
*the marvellous goings-on ‘on the mountain’ then a fortiori we’re considering those goings-on *
at the world whose part is the marvellous mountain.

The key point to appreciate here is that since Lewis insists that the modifier and truth
functional connectives commute in general, then it follows that this commutativity holds for
*the particular case when the modifier restricts quantification to the entire domain of some *
world. This shift of domain does somewhat reduce the generality of the original argument,
but the refutation works equally well. That is, given that individuals are world bound in CT,
and assuming in line with Divers that the marvellous mountain is intended as a world-bound
impossibilium, extending the scope of the ‘on the mountain’ restricting quantifier to the
world 𝑤 of which the mountain is a mereological part of, and employing ‘at 𝑤’ instead, will
*not result in an omission of what happens on the mountain relative to the actual world. The *
point is that if a special case (here, about certain spatiotemporally related mereological sums
of individuals) of a general claim (here, about any individual) is refuted, then so is the general
claim. And the choice to lay out the refutation focusing on the special case is motivated only
by clarity and simplicity of presentation.

One further thing to note is that given some world 𝑤 (say the actual world) and some
*mereological subdomain of it (of the actual world X, say ‘Australia in the year 2015’) the *
*truth of ‘at X: 𝜑’ implies the truth of ‘at 𝑤: 𝜑’ if and only if 𝜑 expresses an existential *
proposition, and the converse is true if 𝜑 expresses a universal proposition, for any X that is
part of 𝑤. To see this, observe that ‘in Australia in 2015, there are wombats’ implies ‘at the
*actual world (i.e. in some spatiotemporal location), there are wombats’, but ‘in Australia in *
*2015 all swans are black’ doesn’t imply that ‘actually, all swans are black’. Conversely *
*actually ‘every human is a mammal’ implies ‘all humans in Australia are mammals’, but *
*‘there exist giant black holes’ doesn’t imply that ‘in Australia there are black holes’. In my *
shift of domain (to worlds, from mountains) in the present refutation, I have taken care to
avoid any possible issues that could arise due to the negligence of those relationships.

Lewis wants us to accept a certain property of the ‘at w’ modifier – namely that it commutes
with the truth-functional connectives. Let’s denote this alleged property of the modifier with
*MC, for modifier commutativity with truth-functional connectives: *

*(MC) For any domain 𝑋, and corresponding restricting modifier 𝜂*𝑋, truth-functional

connective/operator 𝑓, and sentence(s) 𝜑: 132

(𝜂_{𝑋}∘ 𝑓)(𝜑) iff (𝑓 ∘ 𝜂_{𝑋})(𝜑)

For clarification (examples), see the special cases of MC defined on the next page – namely MCC and MCN. Naturally, in this general definition worlds are just special kinds of domains. So, it is clear that if 𝑋 is a subdomain of some world and is equivalent to a proposition with the I or O form, then the truth of 𝜂𝑋(𝜑) entails the truth of 𝜂𝑤(𝜑), and the converse is true

when 𝜑 is either of the A or E form.133

What reasons does Lewis give in support of that property of the modifier? Well, given that
the modifier restricts the domain of all quantifiers within its scope to one possible world, and
given that worlds are (by definition of Lewis’ CT) just mereological sums of their parts, what
occurs at any given world, or region of some world, is amenable to purely mereological
*(extensional analysis). That is, saying ‘in the box there’s a green marble and a red marble’ is *
*the same as saying that ‘in the box there’s a green marble’ and ‘in the box there’s a red *
*marble’. That is ‘at w: A and B’ should be equivalent to ‘at w: A’ and ‘at w: B’, which indeed *
most people will accept as intuitively reasonable (by MC).

So far so (almost) good, it would seem. But Lewis then swiftly proceeds to harness our
intuitions further (as if that was a completely seamless, immediate step), whilst we’re still
under the spell of the apparently innocuous134 nature of the instance of MC just presented,
and asks us to accept as equally innocuous to have MC extended to apply to yet another truth-
*functional connective – namely negation, i.e. ‘at w: not A’ and ‘not at w: A’ being equivalent *
*– in other words, negation and the restricting modifier commute. That is, we’re supposed to *
*accept this broader applicability as equally unproblematic, and not only when applied to *
possible worlds, but apparently to impossible ones as well, since it is explicitly applied to the

132_{ Where ∘ denotes the function-composition relation. When f is a unary operator (e.g. negation) then it acts }*on a single sentence, else if f is a dyadic connective (e.g. conjunction) acts on a pair of sentences. For *

clarification (examples), see the special cases MCC and MCN, further below.
133_{ A: Every S is P. E: No S is P. I: Some S is P. O: Some S is not P. }

‘on the mountain’ restricting modifier, and the mountain has been stipulated to be an inconsistent regime. Let’s denote these instances of MC as follows:

(MCC) *‘at w: A and B’ iff ‘at w: A and at w: B’. *
(MCN) *‘at w: not A’ iff ‘not at w: A’. *

In fact, the argument hinges on MCN holding. But it cannot hold.135 To be sure, MC holds for
*classical worlds. That is, the commutativity of the ‘at w’ restricting modifier holds for *

*classical truth-functional connectives, i.e. whose properties do not violate the principles of *
*classical logic, but such principles are not guaranteed to obtain at non-classical worlds and *
the corresponding commutativity principle fails in such situations. Hoping that the reader’s
intuitions about classical mereology can work in his favour, in a way that they may be guided
*by the common-sense analogy: ‘in the box there is no green marble’ therefore it’s not the *
*case that ‘in the box there is a green marble’. But affirming this purported property of the *
restricting modifier, as I will show, makes the unwarranted assumption that it ought to hold
for inconsistent worlds, which ultimately boils down to wrongly assuming that inconsistent
objects, and worlds of which they are a part, are to be analysed classically; in particular, that
LNC ought to hold there. That is, MCN presupposes LNC, i.e. if LNC fails so does MCN.136

It will be of benefit to have a clear outline of the essential elements of Lewis’s argument and its immediate consequences.

*(Hyp) There is a world w such that at w: A and not A. *

(P1) MCC and MCN are true of the domain restricting modifier for all worlds.
*(C1) There is a world w such that at w: A and not at w: A. * (Hyp.+P1)
(P2) There is no subject matter whereby one can tell the truth by contradicting

themselves (LNC).

(C2) *Therefore, there is no such w. *

135_{ Kiourti (2010, pp.116-119) raises similar objections to accepting MCN and consequently follows a very similar }
line of reasoning to mine in demonstrating what this assumption amounts to. My conclusion differs by virtue of
*how I have formulated the objection – Kiourti concludes that Lewis begs the question against the hypothesis of *
concrete impossibilia, where I conclude that the argument is unsound on grounds of the falsity of hypothesis
*concerning the properties of the ‘at w’ modifier, where w needn’t be a classical world. *

136_{ Given that Lewis is committed to a classical truth conditional theory of meaning, his adherence to MCN is }
understandable. After all, denying MCN may require a denial of the classical truth conditional theory of meaning
*or denying that the connective ‘¬’ means not (classical negation) at impossible worlds. However, these *

alternatives cannot be ignored without the risk of begging the question. For example, one may reject MCN by adopting an information theoretic theory of meaning, e.g. see (Mares 1997, 2004) – a theory which requires dialethism and a paraconsistent view of negation, both of which Lewis rejects emphatically.

(C3) *Therefore there are no inconsistent worlds.*137_{ } _{(RAA) }

(C4) There are no impossible worlds.

*Why assume that MCN ought to always hold for all worlds, both possible and impossible?*138
*Let’s think of circumstances that would violate MCN. Consider a world w* that contains *
*some genuinely inconsistent object a. That is, let us assume that ‘at w*: Pa and not Pa’ for *
*some property P. Now, given that Pa is a truth value glut (both true and false, but in *
*particular true) at w*, we should analyse it accordingly, which means we have no justified *
*recourse to MCN, which presupposes LNC, precisely because the existence of a at w* is a *
*counterexample to LNC! That is, we cannot infer ‘not at w*: Pa’ from ‘at w*: not Pa’, *
*precisely because both Pa and not Pa hold at w.*139

In other words, the assumption of the universality of MCN, rests on the erroneous supressed
assumption that LNC ought to hold for inconsistent worlds in much the same way as it does
*for consistent ones, like boxes and fridges, which we are familiar with. But we are not *
*considering consistent worlds, so appeal to MCN is unjustified, precisely because w* is an *
*inconsistent world. We will run into literal contradictions, i.e. ‘at w Pa’ and ‘not at w Pa’ *
*being true simpliciter, only if we assume the existence of truth value gluts, and then proceed *
to analyse them classically by unwarranted appeal to MCN. This is what Lewis in fact does.
By refraining from any such questionable steps, we don’t end up in explicit contradiction. So,
*although w* is a genuinely inconsistent world, the theory remains consistent. That is, *

*assuming the existence of w*, doesn’t lead to a contradiction simpliciter. But if the only *
reason for Lewis’ rejection of the existence of genuine inconsistent worlds is the success of
*this reductio ad absurdum, then his argument fails, because the attempted reductio ad *
*absurdum from assuming the existence of w* is unsound (since P1 is false). Consequently, *
using only Lewis’s reasons for rejecting the existence of genuine inconsistent worlds, it
doesn’t follow that they don’t exist.

137_{ The argument has the form of reductio ad absurdum. }

138_{ An analogue of this argument has been made in the context of extended theories of objects that include }
non-existent objects, e.g. Parsons’s theory – a succinct summary of Parsons’s analogue to the foregoing
argument is given in (Zalta 1988, p.132).

139_{ An analogous issue arises, concerning a questionable principle in theories of non-existent objects, e.g. }
*Parsons (1980, p.19, p.105), Zalta (1988, pp.131-4). The analogy is along the lines possible vs. impossible and *

*existent vs. non-existent. Parsons points out that the analogous principle to MCN holds for all existing objects, *

*but not all objects (where impossible ones are among the non-existent). In that theory inconsistent objects are *
classified as non-existent objects, much in the same way as Lewis would classify them as classically impossible.

*For similar reasons, a “Lewis-type” reductio ad absurdum against the existence of genuine *
*incomplete worlds isn’t sound either.*140_{ That is, given the breakdown of Lewis’s argument }

above, the argument with the following conclusion is unsound, also due to P1’ being false.
*(Hyp’) There is a world w such that it’s neither the case that at w: A nor is it the case *

*that at w: not A. *

(P1’) MCC and MCN are true of the domain restricting modifier for all worlds.
*(C1’) There is a world w such that at w: A and not at w: A. (Hyp+P1)*
(P2’) There is no subject matter whereby one can tell the truth by contradicting

themselves (LNC).

*(C2’) Therefore, there is no such w. *

*(C3’) Therefore there are no incomplete worlds. * (RAA)
(C4’) There are no impossible worlds.

*Let us assume the existence of an incomplete world w° that has, as one of its parts, an object *
*a such that neither Pa nor not Pa is true at w°, and then let us investigate what is really doing *
*the work in a "successful” derivation of a contradiction. That is, Pa is a truth value gap at w°, *
*so we have both ‘not at w° Pa’ and ‘not at w° not Pa’. How is MCN justified here, and can it *
*be used to derive a contradiction? In particular, can we use MCN to infer ‘at w°: not Pa’ from *
*not ‘at w°: Pa’? Call that particular MCN instance MCNI for importing negation, i.e. going *
*from not ‘at w A’ to ‘at w not A’, as the dual move to MCNE from the previous, inconsistent *
*world w* example, which hinged on exporting negation, i.e. from ‘at w not A’ to not ‘at w A’. *

(MCNI) *Importing negation: If ‘not at w: A’, then ‘at w: not A’. *
(MCNE) *Exporting negation: If ‘at w: not A’, then ‘not at w: A’. *

*That would suffice for a derivation of a contradiction, since we already have ‘not at w° not *
*Pa’. This is also necessary since I have already argued that the other direction, i.e. MCNE is *
unwarranted, thereby disabling the only other way of deriving a contradiction. (In passing we
can quickly set out that other means of deriving a contradiction, were MCNE to be allowed in
*the case of incomplete worlds. Given MCNE we can go from not ‘at w° not Pa’ to not [not *
*‘at w° Pa’], and from there to ‘at w° Pa’, and given that we also have not ‘at w° Pa’, a *
contradiction ensues. But MCNE fails (as argued earlier), so let’s focus on the inapplicability
*of MCNI in the case at hand.) The use of MCNI would seem wrong, since w° is an *

140_{ The next two paragraphs can be skipped. }

*incomplete world, so we cannot assume that the failure of Pa being true at w° means that not *
*Pa should be true there. To do so would be to assume that there is a logical connection *
*between the failure of Pa being true at w° and the truth of not Pa at w° after all. As a matter *
*of fact, it would amount to assuming that LEM ought to hold at w°, but it needn’t because w° *
*isn’t a classically possible world, and as a matter of fact it doesn’t, precisely because Pa is a *
*truth value gap at w°. *

So, the above shows that Lewis-type arguments against the existence of impossible worlds fail, because they fail to rule out the existence of inconsistent and incomplete worlds, which are classically impossible.