Complex Demonstratives - UCL Working Papers in Linguistics.
Are complex demonstratives (expressions of the form ‘that F’) referring expressions or quantifiers? Their semantic behaviour seems to pattern sometimes with the former, sometimes with the latter. In this paper I examine both referential and quantificational accounts of complex demonstratives in order to show that neither side satisfactorily accounts for the all the data. I then outline an alternative analysis on which, although lexically univocal, complex demonstratives can give rise to genuinely singular or genuinely quantificational truth conditions according to context.
Over
the last few years, the philosophical debate on the referring-denoting
distinction has shifted its focus. Donnellan (1966) placed definite descriptions
at the heart of the debate, and for three decades or so, they stayed there.
Recently, however, definite descriptions have been usurped by what have
variously been called demonstrative descriptions, complex ‘that’-phrases
and complex demonstratives (I shall stick to this last throughout).
Complex demonstratives, which at a first pass we can think of as any expression
of the form ‘that F’, pose a particular problem for the truth-conditional
semanticist toiling on the border between referring and denoting: on the one
hand it would seem that they are semantically closely linked to simple
demonstratives, standardly taken to be the prototypical referring expressions,
but on the other they display the kind of syntactic complexity associated with
definite descriptions and, more broadly, with the class of quantifiers. For the
standard truth-conditionalist, who either overtly or tacitly accepts the
hypothesis that all noun phrases must either be referential or quantificational[1],
this is an alarming state of affairs, and much ink has been spilled in the
effort to show that these apparently anomalous expressions do, after all, behave
in familiar ways. None of the stories currently on offer, however, embodies the
truth, the whole truth and nothing but the truth: some, while offering nuggets
of truth, are empirically inadequate; others offer much of the truth as far as
they go, but seek to delimit their field of inquiry in ways that, under
scrutiny, reveal themselves as unprincipled; and yet others offer what seem to
me to be profound insights into the workings of complex demonstratives, but fall
at least at the hurdle of theoretical parsimony.
In
the first part of this paper, I shall examine a cross section of accounts from
along the theoretical spectrum. My aim will be to unearth the nuggets hidden
within these analyses at the same time as demonstrating the shortcomings of
each. In the second part of the paper, I want to take the insights offered by
these accounts and, with the help of a type of evidence used by none of them,
approach complex demonstratives from a different perspective, the perspective
offered by a general theory of communication. Approaching complex demonstratives
from a communication-theoretic perspective offers insights that are at least
obscured by a directly truth-theoretic approach. Towards the end of the paper,
I shall, however, address the questions at the heart of the
truth-conditional debate on complex demonstratives.
2
Referentialism
It
is more or less a given of philosophical semantics that if any expressions are
referential then simple demonstratives are. They seem to display more clearly
than any other expression-type what have been taken to be the two key features
of referentiality: they track individuals across possible worlds and their empty
uses lead to propositional failure. If I utter (1):
(1)
That is filthy
then
the truth of my utterance will depend, in any possible world, on the object I
have referred to in the actual world; if, say, I have pointed at a whiteboard
while uttering (1), then my utterance will be true at a possible world iff that
very whiteboard is filthy at that world[2].
If, on the other hand, I utter (1) without intending to refer to anything by my
use of ‘that’, then intuitively it seems that my utterance fails to express
any proposition; what would have to be the case, after all, for my utterance to
be true?
If we accept, then, that simple
demonstratives are referential, there would seem to be at least a strong pull
towards extending this story to complex demonstratives. Simple demonstratives,
after all, form a syntactic part of their complex brethren and furthermore their
uses seem semantically very closely allied. If I utter (1) while pointing at a
particular whiteboard, then it would seem that what I have said is much the same
as what I would have said by pointing at the same whiteboard while uttering (2):
(2)
That whiteboard is filthy
It
also seems that our use of a complex demonstrative in (2) passes the tests for
referentiality mentioned above: the truth of the proposition expressed by an
utterance of (2) will depend on the very same object across possible worlds and
if, in uttering (2), I fail to pick out a particular object by my use of ‘that
whiteboard’, then again it seems that my utterance fails to express any
proposition. Given this evidence, an entirely straightforward account of complex
demonstratives seems to present itself: they are referential, just as are simple
demonstratives; their contribution to propositional content goes no further than
their referent.
But
what role will be played on such a story by the nominal element of the complex
demonstrative (the ‘F’ in ‘that F’)? Maybe, on the model of simple
demonstratives, they play no more role than the contextual cues that guide us
toward the object a speaker intends to refer to as ‘that’; they may, in
other words, merely play a pragmatic role in helping the hearer assign
reference. This proposal, advocated by, for instance, Larson and Segal (1995),
has at least one immediate implication that has seemed unpalatable to many: the
nominal element of a complex demonstrative comprises a meaningful NP, so why
should the meaning of this NP simply disappear from utterance content? For the
compositional semanticist, and for some who are not so wedded to
compositionality, disappearing meanings have been seen as very unwelcome
theoretical artefacts. But there are other problems beyond this for the simple
referential thesis. Utterances with complex demonstratives in subject position
(henceforth complex demonstrative utterances) seem to give rise to
entailments which are incompatible with simple referentiality. Consider (3) and
(4):
(3)
That green car is very old
(4)
Some green car is very old
There
does on the face of it seem to be an entailment from (3) to (4), yet the simple
referentialist has no story to tell about this: for her, sentence (3) expresses
the same proposition as would (5), uttered while pointing at the appropriate
green car:
(5)
That is very old
and
there is surely no temptation to think that (5) entails (4) (there may, of
course, be an entailment once we add the extra premise that that is a green car,
but this is not what’s at issue here). How might the referentialist try to
rescue her story, which after all seems to be firmly rooted in intuition, from
this challenge? One strategy, adopted in differing forms by Kaplan (1978, 1989a,
1989b), Braun (1994) and Borg (2000), is to locate the contribution of the
nominal not in utterance content but in Kaplanian character. On this kind of
story, the nominal ‘green car’ in (4) would constrain the reference of ‘that
green car’ to something that is, in fact, a green car, and then drop out of
the picture. How does this help with the difficulties faced by our initial
story? Firstly, on this type of account the nominal does make a semantic
contribution, albeit not a contribution to propositional content. Secondly,
although it still sees no actual entailment between, for instance, the
propositions expressed by utterances of (3) and (4), it does at least have a
story to tell about why there should appear to be an entailment: since
the complex demonstrative ‘that green car’ can only be used to refer to
something that is, in fact, a green car, an utterance of (3) cannot express a
true proposition unless some green car is very old. In other words, although the
proposition expressed by (3) does not entail the proposition expressed by (4),
(6) is valid in the sense of Kaplan (e.g. 1989b):
(6)
If that green car is very old, then some green car is very old.
which
is, for the character-theorist, close enough to explain the illusion of
entailment.
For Dever (2001), however,
these solutions fail to do the work we want of them: firstly, what we’re
trying to explain about (3) and (4) is not the illusion of an entailment but an
actual entailment, and secondly, a story on which the nominal element of a
complex demonstrative contributes merely to character and not to content
violates semantic innocence, the principle that the same expression should
display the same semantic behaviour regardless of the linguistic context in
which it appears. Dever himself proposes a story which, although closely allied
to Braun’s character-theoretic analysis, can cope with these supposed
problems. His story falls within what seems to be a growing trend in
philosophical semantics for multiple-proposition analyses, i.e. analyses on
which single utterances standardly express more than one proposition.
Abstracting away from the syntactic minutiae, Dever’s story amounts to the
proposal that ‘that F is G’ expresses a sequence of two propositions: the
primary proposition that that is G, i.e. the referential proposition expressed
on all the accounts we’ve looked at so far, and the secondary proposition that
that is F.
I
am unsure as to how impressed we should be by the purported problems with the
character-theoretic account which motivate Dever’s story. Firstly, as Dever
himself points out, there’s little evidence to support the view that there is
an actual, rather than merely apparent, entailment between, for instance, (3)
and (4); and secondly, as I have argued in Powell (2001), there are good reasons
why we should not expect semantic innocence to hold at the level of
propositional content, but rather should relegate it to the lower level of
linguistic meaning. Whether or not you are inclined to agree with me on this,
there is a rather more fundamental problem facing Dever’s account, and indeed
facing all of the referentialist accounts we have so far looked at: as King
(1999, 2001) shows, complex demonstratives seem to have non-referential uses. In
particular, King points to three different types of uses of complex
demonstratives which, he argues, present the referentialist with difficulties.
Firstly there are what he terms no demonstration no speaker reference (NDNS)
uses. He asks us to consider Scott who, in the course of a lecture on hominid
history, utters (7):
(7)
That hominid who discovered how to start fires was a genius.
Here
Scott has no particular individual in mind to whom he intends to refer by
his use of ‘that hominid’ and equally, it seems, his hearer need not think
of any particular individual as ‘that hominid’ in order to understand
Scott’s utterance.
Secondly King points to quantifying
in (QI) uses, uses on which a quantifier outside the complex
demonstrative binds an anaphoric element inside:
(8)
Every father dreads that moment when his eldest child leaves home.
And
finally there are narrow scope (NS) uses, on which complex demonstratives seem
to occur within the scope of other quantifiers:
(9)
That professor who brought in the biggest grant in each division will be
honored
There
seem to be two possible interpretations of (9), one on which one and the same
professor brought in the biggest grant in each division and ‘that professor’
refers to her, and one on which ‘that professor’ occurs within the scope of
‘each division’ and thus picks out, for each division, the unique professor
who brought in the biggest grant in that division. On the latter interpretation,
the complex demonstrative has an NS occurrence.
On none of these types of use
does the complex demonstrative appear to be functioning referentially: NDNS uses
pattern with Donnellan (1966)’s attributive uses of definite descriptions in
allowing ‘whoever he is/was’ to be inserted after the NP:
(10)
That hominid who discovered how to start fires, whoever he was, was a
genius.
which
at least strongly suggests that they should receive a non-referential analysis.
As for QI and NS uses, there seems little temptation to analyse them
referentially: what could possibly count as the referent of ‘that moment when
his eldest child leaves home’ in (8) or ‘that professor who brought in the
biggest grant in each division’ in (9)?
On the face of it, the
existence of such non-referential uses of complex demonstratives seems to rule
out any straightforwardly referential account. Many referentialists, however,
have recognised such uses (in particular QI uses) without apparently feeling the
need to abandon their accounts and start again. The approach of Lepore and
Ludwig (2000) is typical:
Sometimes,
of course, “that” is pressed into service as a variant of “the”, and one
could imagine someone uttering (21) [‘Every man loves that woman who is his
mother’] with that in mind. We are not concerned with such uses of “that”,
but rather with demonstrative uses.
Lepore and Ludwig (200, p.219 fn.28)
We
can take this exclusion in two ways: either Lepore and Ludwig (and others who
support this approach to non-referential uses) are not in the business of
studying language itself, but are, rather, concerned with pinning down the
truth-conditions of particular uses of language, or there is an implicit
assumption in the passage above that there are two different ‘that’s in
play, one referential and one non-referential. If we are to take the first
reading, then Lepore and Ludwig are simply not engaged in the same enquiry as,
for instance, King, Dever and myself. If, however, we are to take the second
reading, explicitly advocated by Dever, then the referentialist may still have a
claim to provide a complete account of the expression-type he has in his sights,
i.e. referential complex demonstratives.
The
burden of proof is, here, on the referentialist. All things being equal, and
with our lesson on parsimony learnt from Grice and modified Occam’s razor, we
should be ready to reject a story which posits an ambiguity in ‘that’ for
one that posits none. Presuming we can tell a story about ‘that’ which doesn’t
require us to posit an ambiguity (and I aim to do just this in the latter part
of this paper), the referentialist must therefore convince us that there really
is an ambiguity if he is to overcome the methodological bias against his story.
Dever attempts to give at least some evidence to show that ‘that’ is
ambiguous (apart from his observation that the Oxford English Dictionary
supports this view, which I don’t intend to worry about too much). His key
evidence is that, in examples such as (8), substituting ‘this’ for ‘that’
leads to infelicity:
(11)
*Every father dreads this moment when his eldest child leaves home.
Given
that, for referential uses of ‘that’, this substitution is unproblematic,
Dever concludes that the ‘that’ in, for instance, (8) is simply a different
lexeme from the ‘that’ in, for instance, (2). There is, however, an
undefended assumption in Dever’s argument that seems to be at least
questionable: that because ‘this’ and ‘that’ are substitutable in
referential contexts, their semantics must be such that they are substitutable
in all contexts. There is, however, a clear and familiar difference between the
semantics of ‘this’ and the semantics of ‘that’: ‘this’ is used to
pick out objects which are by some standard near to the speaker, while ‘that’
is used to pick out objects which are not near the speaker. There is also some
evidence to suggest that ‘that’ is, in some sense, a default, whereas ‘this’
is marked: consider, for instance, that, however close something is to the
speaker, so long as there is not another further object which she might be taken
to be referring to, she may refer to it as ‘that’; whereas a speaker who
refers to something distant as ‘this’, whether or not there is a candidate
nearer object, is at least speaking infelicitously. It seems to me that it may
be just this dimension of the semantics of ‘this’ and ‘that’ that is
responsible for the infelicity of (11): on the one hand the speaker’s use of
the complex demonstrative is descriptive, thus picking out no particular
individual, but on the other she is indicating, by her use of ‘this’ rather
than ‘that’, that what she is talking about is relatively near to her. It
may well be the incompatibility between these two that leads to infelicity.
Whether or not you accept this
argument, the point I want to stress is that the infelicity of utterances like
(11) may well be explicable in terms other than the ambiguity of ‘that’.
Until we have stronger empirical evidence to support the view that ‘that’ is
ambiguous, therefore, we should, on methodological grounds, prefer a univocal
semantics for ‘that’. In the next section I shall look at some accounts of
this type.
3
Quantificationalism
The
point we seem to have reached is this: given the lack of evidence for an
ambiguity, we are on the hunt for a univocal semantics for complex
demonstratives, but one that allows for both referential and non-referential
uses. Where might this hunt lead us? On the assumption that all NPs are either
referential or quantificational, there seems only one direction open to us:
complex demonstratives are quantificational. But what might a quantificational
semantics for complex demonstratives look like? Neale (1993) outlines a story on
which ‘that F is G’ is equivalent to ‘the actual F I am demonstrating is G’
and on which complex demonstratives take mandatory wide scope. The advantage of
this account, for Neale, is that it gives the content of the nominal a role to
play in utterance content, thus fitting with the intuition mentioned earlier
that meanings shouldn’t simply vanish. It is, however, as Lepore and Ludwig
(2000) point out, not only ad hoc in that there is no clear reason why the
complex demonstrative should always take wide scope, but also empirically
inadequate: it has no story to tell about why the failure of a complex
demonstrative to designate should result in the failure of the utterance in
which it appears to express a proposition (this is not, after all, a property of
definite descriptions); and it predicts, wrongly, that an utterance by s of ‘that
F is G’ entails that s exists. We thus need something rather more subtle if we
are to place complex demonstratives within the class of quantifiers.
Two
recent accounts, those of King (2001) and Lepore and Ludwig (2000), attempt to
provide just that something. Lepore and Ludwig’s account is based on the
premise that, contrary to superficial appearance, complex demonstratives are not
syntactically formed by the concatenation of a quantifier ‘that’ and a
nominal. Instead they are restricted quantifiers with the quantifier element
suppressed and both ‘that’ and the nominal contributing to the restrictor.
In essence, as Lepore and Ludwig make clear, their story treats ‘that F is G’
as equivalent to [The x: x=that and x is F](x is G)[3].
I suspect that, as far as it goes, this story may have much to recommend it.
But, as mentioned earlier, Lepore and Ludwig are only in the business of
accounting for referential uses of complex demonstratives. If, therefore, we
accept, as I have argued we should, that there are truly non-referential uses of
complex demonstratives, then our hunt for a satisfactory account continues.
King (2001) offers a story
which is specifically tailored to account for both referential and
non-referential uses of complex demonstratives. On his analysis, complex
demonstratives are profoundly context-sensitive quantifiers, depending on
context not just for which properties are to be taken as determining the
intended referent, but also for whether that referent is to be tracked across
possible worlds or not. For King, the lexical meaning of ‘that’ (as it
appears in complex demonstrative constructions; he only tentatively suggests
that the outline of his account might be extended to simple demonstratives) is a
four-place relation:
_____
and _____ are uniquely _____ in an object x and x is _____
King (2000, p.43)
The
first argument place, on this story, is to be filled by the property expressed
by the nominal, the second by a property determined by speaker intention, the
third either by jointly instantiated or jointly instantiated in <w,
t> and the fourth by whatever property is expressed by the predicate.
There are for King two fundamentally different kinds of intention that can
determine the property to saturate the second argument place: either the speaker
may have a perceptual intention, which may include both intentions towards
objects of immediate perception and towards remembered objects of past
perception, or she may have a descriptive intention, i.e. an intention towards
whatever satisfies some descriptive condition. Which of these two kinds of
intention she has will make a difference not only to the first-order property
that saturates the second argument position but also to the second-order
property that saturates the third argument position. If the speaker has a
perceptual intention, towards the object a say, then the second argument
position will be saturated by the property of being identical to a and
the third by the property of being jointly instantiated in <w, t>, where w
and t are to be taken rigidly to designate the world and time of utterance. If,
however, the speaker’s intention is descriptive, then the second argument
position will be saturated by whatever property her intention determines, and
the third by the property of being jointly instantiated.
Let me work through a couple of
examples to illustrate how this story is supposed to go. Consider first a
classic referential use of a complex demonstrative: Janet and John are at a
party and, seeing a man across the room, Janet says:
(12)
That man wearing a kipper tie has no dress sense
What
proposition will Janet’s utterance have expressed on King’s story? The first
parameter will be saturated with the property of being a man wearing a kipper
tie, the second, since Janet has a perceptual intention concerning the man in
the kipper tie, with the property of being (identical to) a, where a
is the very individual in question, and the third, again because Janet’s
intention is perceptual, with are uniquely jointly instantiated in <w,
t>. Finally, the fourth parameter will be saturated with the property of
having no dress sense. What, then, will this amount to? It amounts to the claim
that Janet’s utterance will be true in any circumstance of evaluation iff
there is, in the context of Janet’s utterance, one and only one thing which is
both a man wearing a kipper tie and also a, and that things has no dress
sense in the circumstance.
Next let’s take a
non-referential use, say an utterance of the sentence in (7). What truth
conditions will King’s account predict for this? Here the first parameter will
be filled with the property of being a hominid who discovered fire, the second
with the same property, since the speaker’s intention is simply to talk about
whoever satisfies the nominal, and the third, since the speaker’s intention is
descriptive, with are uniquely jointly instantiated. Finally again the
fourth parameter will be filled by the property of being a genius. King’s
prediction here, then, is that an utterance of (7) will be true in a
circumstance iff there is one and only one hominid who invented fire in that
circumstance and he is a genius in that circumstance[4].
What
advantages does this complex semantics have over the simpler story told by Neale
(1993)? One clear advantage is that it falls at neither of the two hurdles
mentioned above: firstly, the proposition expressed by an utterance of ‘that F
is G’ will not include any mode of presentation of the speaker, and thus will
not entail that the speaker exists. Secondly, where there is nothing which the
speaker refers to with a perceptual use of ‘that F’ (the use that
referentialists are worried about), there is no property of being identical to
the object of the speaker’s perceptual intention to saturate the second
argument position, and thus no complete proposition expressed. It also satisfies
the referentialist’s other intuition about referential uses: that in such uses
the referent of the complex demonstrative is tracked across possible worlds. On
King’s story, the truth value of a complex demonstrative utterance backed by a
perceptual intention will, in any possible world, depend on the object of the
speaker’s intention in the context of utterance.
That it can handle
non-referential uses at the same time as accommodating the fundamental
intuitions upon which the referentialist bases his story seems to be strong
evidence in favour of King’s account. Unfortunately,
the account is empirically inadequate. Any story on which complex
demonstratives are quantificational is going to have to show that they enter
into scopal interactions with other quantifiers, modal operators etc. Views in
the literature are radically divided on this, with the dividing line not
surprisingly running between referentialists and quantificationalists. For Dever
(2001), for instance, complex demonstratives cannot enter into scopal relations,
whereas for King (2001) they can. Dever of course excludes King’s QI and NS
uses from the class of complex demonstratives, and with NS uses in particular,
we’ve already seen what seems to be a scope ambiguity involving a complex
demonstrative: (9) has two readings, one on which the complex demonstrative
falls within the scope of the quantifier ‘each division’, and one on which
the scopal relation is reversed. But what of those uses of complex
demonstratives which are supported by perceptual intentions, i.e. those uses at
the heart of the referentialist’s story? As Dever points out, cases in which
complex demonstratives occur in sentences with uncontroversial quantifiers are
not going to be of much help here. To see why, consider (13):
(13)
Every woman in the room likes that man over there
On
King’s story, this sentence should equate to two sets of truth conditions
which will look roughly like (14) and (15):
(14)
For all x such that x is a woman in the room the properties of being a
man over there and being identical to a are uniquely jointly instantiated
in <w, t> by y and x likes y
(15)
The properties of being a man over there and being identical to a
are uniquely jointly instantiated in <w, t> by y and for all x such that x
is a woman in the room x likes y
But,
although these sets of truth conditions may differ in their scope relations,
their truth values will covary across contexts. There is thus no way to
establish whether (12) really does equate to (14) and (15). Sentences in which
complex demonstratives co-occur with verbs of propositional attitude, however,
are a different matter. Dever asks us to consider the example in (16):
(16)
Albert believes that upright citizen is a spy
For
King there should be a reading of (16) on which Albert holds a contradictory
belief, i.e. the belief that the property of being an upright citizen and being
a are uniquely jointly instantiated in <w, t> by x and x is a spy. But
there just doesn’t seem to be any such reading.
For
both King and Lepore and Ludwig it is important that such readings should be
available, and both offer examples in which such narrow scope readings of
referentially-used complex demonstratives relative to verbs of propositional
attitude are supposedly available. Both, however, seem to me to be examples of
something else. King asks us to imagine a party at which Alan has just been
named CEO of Chanticleer. Sherry, a Chanticleer employee who believes Alan hates
her, arrives at the party to hear the bad news. Two other party-goers are in
conversation, when one looks over to Sherry and, seeing that she is looking very
glum, asks why. The other, pointing at Alan, replies:
(17)
Sherry believes that guy who was just named CEO of Chanticleer hates her.
King
then mounts an argument that runs like this: the utterance of (17) is an
explanation of Sherry’s behaviour; if the complex demonstrative in (17) were
interpreted referentially (or as taking wide scope) (17) would not be an
explanation of Sherry’s behaviour, since she has long believed that Alan hates
her without wasting any time moping about it; it must therefore be the case that
the complex demonstrative is interpreted as taking narrow scope with respect to
the belief operator, thus placing the property of being just named CEO of
Chanticleer within Sherry’s belief.
But
this cannot be the reason that (17) is taken as an explanation of Sherry’s
behaviour. Consider a slight variant on the context above: imagine that both
speaker and hearer know that Alan has not been appointed CEO of
Chanticleer, but that Sherry misguidedly believes he has. In this situation the
reading that King takes to be our natural interpretation of (17) will be plainly
and straightforwardly true, since Sherry does hold the belief that the
properties of being just named CEO of Chanticleer and being Alan are uniquely
instantiated in x and x hates her. Yet an utterance of (17) in such a context
seems at best highly infelicitous, if not plain false. Given this, we would do
well to look elsewhere for a story on why an utterance of (17) acts as an
explanation of Sherry’s behaviour, and we don’t have far to look. Let’s
assume that, on the evidence above, (17) expresses a singular proposition about
Alan. Since it appears at a point in the conversation at which the hearer is
expecting an answer to his question, he will interpret the utterance of (17) in
that light; in other words, he will ask himself why Sherry’s belief that Alan
hates her is making her glum. Now there is one property of Alan that the speaker
has made contextually highly salient by his choice of referring expression: the
property of being the newly-appointed CEO of Chanticleer. Putting together the
premise that Sherry believes that Alan hates her and the premise that Alan is
the newly-appointed CEO of Chanticleer, along with various other premises such
as that Sherry works for Chanticleer, it shouldn’t take the hearer long to
reach a conclusion about the causes of Sherry’s glumness. (17) can thus act as
an explanation of Sherry’s behaviour without any need for a narrow scope
interpretation.
Lepore and Ludwig try another
tack. They ask us to consider Tom, Mary and Mary’s companion in a restaurant.
Tom leans over to Mary and, pointing at a waiter in white sneakers, says ‘that
man wearing white sneakers is a good waiter’. Mary, who has failed to hear a
part of Tom’s utterance but who sees that the waiter Tom is talking about is
wearing Nike sneakers, turns to her companion and says ‘Tom believes that man
wearing Nike sneakers is a good waiter’, to which her companion, who has
better hearing, replies:
(18)
No, he thinks that man wearing white sneakers is a good waiter.
Lepore
and Ludwig’s argument then runs like this: if the complex demonstrative in
(18) is analysed referentially (or as taking wide scope), Mary and her companion
are both attributing the same belief to Tom, i.e. the belief of the man in
question that he is a good waiter; since they are in disagreement, this cannot
be the intended interpretation; the complex demonstrative in (18) must therefore
be interpreted as taking narrow scope in relation to the belief operator, thus
placing the property of being a man wearing white sneakers within Tom’s
belief. But this argument is based on a false assumption: that, if there is a
disagreement between Mary and her companion, it must be a disagreement on the
content of Tom’s belief. Consider the following example, adapted from Carston
and Noh (1995):
(19)
A: We went to the zoo and saw the hippopotamuses
B:
No, we went to the zoo and saw the hippopotami
In
(19) B is disagreeing with A, but her disagreement is not with the content of A’s
utterance, it is, rather, with the linguistic form A has used to express that
content: the negation in (19) is metalinguistic. This type of analysis would
seem to extend very naturally to (18): what Mary is taking issue with is not the
content of her companion’s utterance, which is simply the attribution of a
singular belief to Tom, but the linguistic form her companion has used to
express that content, which inaccurately mirrors the linguistic form used by Tom
to express his own belief. Given the availability of such an analysis, this
example seems to offer no firm evidence of scope interaction between a complex
demonstrative and a verb of propositional attitude.
Even clearer than propositional
attitude contexts are sentences in which complex demonstratives, used
perceptually, co-occur with negation. Consider (20):
(20)
That policeman is not John’s brother
uttered
while pointing, say, at a particular policeman. On any quantificational account,
(20) should admit of two interpretations. Sticking with King’s story, those
two interpretations will look something like:
(21)
The properties of being a and being a policeman are uniquely jointly
instantiated in <w, t> by x, and it is not the case that x is John’s
brother
(22)
It is not the case that the properties of being a and being a policeman
are uniquely jointly instantiated in <w, t> by x, and x is John’s
brother
In
other words, there should be a reading of (20) on which it will be true just in
case there is nothing which is a unique policeman who is a here and now and
which is John’s brother. But there doesn’t seem to be any such reading: this
is, after all, a reading which would be true in the circumstance that the
individual who is the object of the speaker’s intention is John’s
brother, just so long as he is not a policeman. King mounts a defence against
this kind of argument by invoking pragmatic processes: in perception-based
cases, speakers are interested in getting their hearers to pick out the objects
of their perceptual intentions; to this end, they will pick, as the nominal, a
predicate which they believe their intended referent to satisfy, and which will
thus help their hearers towards identifying the intended referent; given this,
they are not going to want their utterances to be interpreted in such a way that
they are true merely by virtue of their referents not satisfying the nominal
predicate. The key point of this argument is that, with perceptual uses, narrow
scope readings of complex demonstratives relative to negation are pragmatically
blocked. It seems to me, however, that such readings are not so much blocked as
absent. Consider (23):
(23)
There is a flag hanging out of every window.
The
sentence in (23) clearly has two scopal readings, one of which, the reading on
which ‘a flag’ takes scope over ‘every window’, is pragmatically blocked
in most contexts. But there are two things to note about (23): firstly, although
this reading is blocked, we can nevertheless make it out, and secondly, it is
possible to manipulate the context so that this reading becomes the natural one:
suppose we are discussing a factory across the road which prints huge flags and
dries them after printing by hanging them out of the window; the flags are often
so large that, when drying, they hang out of four of five windows, but this time
they’ve excelled themselves; wanting you to come and have a look, I utter
(23). In such a context, the wide-scope existential reading will be at least
much less strongly suppressed than in a neutral context. These two properties,
the properties of being make-outable and de-suppressible by contextual
manipulation, are typical of pragmatically suppressed scopal readings. Yet the
supposed suppressed reading of (20) has neither of these properties. Given this,
the only reason to suppose that there is such a reading is if one’s theory
requires it.
So where does our discussion of
the referentialist and quantificationalist strategies leave us? The
referentialist seems to have a lot that is right to say about referential uses
of complex demonstratives; in particular, as the discussion of scope above
demonstrates, she is right to claim that, on referential uses, complex
demonstratives contribute nothing but their referents to propositional content.
However, she has nothing to say on non-referential uses which, as I hope I have
convinced you above, are genuine uses of complex demonstratives. The
quantificationalist, on the other hand, has a lot that is right to say about
non-referential uses (I shall expand on just how much of the quantificationalist’s
story is right later in the paper), but, again, given the scopal data above, has
serious difficulties when trying to account for referential uses.
4
Hybrid accounts
If
referential uses are truth-conditionally singular, non-referential uses are
truth-conditionally general, and complex demonstratives are not ambiguous, what
kind of story is there left to tell? Maybe the way forward is an account on
which complex demonstratives are hybrids of referring terms and quantifiers.
Versions of this sort of account have been advocated by Richard (1993) and Neale
(1999)[5]. For Richard, complex
demonstratives are what he calls articulated terms: they introduce not
only their referent to propositional content, but also some quasi-Fregean way of
thinking of that referent, as expressed by the nominal. Quite what the truth
conditions of a complex demonstrative utterance will be on this story is
slightly obscure: Richard suggests that an utterance of ‘that F is G’ will
be truth-conditionally equivalent to an appropriately related utterance of ‘that
is F and that is G’, in which case his story is not really hybrid in the sense
we are interested in, and will fall at the same hurdles as, for instance, Dever’s
account (plus some others). The syntactic structures he assigns to articulated
terms, however, suggest that maybe his intention is to claim that complex
demonstrative utterances give rise to two parallel sets of truth conditions, one
singular and one general.
Whether this is Richard’s
view or not, it is clearly the view of Neale (1999). For Neale, the lexical
semantics of a complex demonstrative consist in a series of instructions:
initially, the hearer of ‘that F is G’ is instructed to build the
descriptive proposition in (24):
(24)
[the x: s is indicating x & Fx] Gx[6]
He
is then instructed to find whichever object is the x such that s is indicating x
and x is F in the context, and build a second, singular proposition containing
that object:
(25)
Ga[7]
The
thrust of Neale’s story is thus that complex demonstratives are not
referential or quantificational, they are referential and
quantificational. In tandem with this two-proposition account, Neale tells a
story about contextual weighting: in most contexts it will be the proposition in
(25) which will carry what he calls the contextual weight, but in some ‘exceptional
circumstances’[8]
it will be the proposition in (24).
This story will not, however,
do the work we need. And, to be fair to Neale, it is not intended to do so,
since, even in those contexts in which it is the descriptive proposition that
carries the contextual weight, Neale is still envisaging that there should be a
perceived object towards which the speaker has some referential intention; he
is, in essence, giving an account on which the kinds of uses of demonstratives
which lie at the heart of the referentialist story can be interpreted
(predominantly) descriptively, rather than one that can account for the purely
descriptive uses raised by King. On King’s descriptive uses, there is no
object towards which the speaker has any referential intention; when Scott uses
the expression ‘that hominid who discovered how to start fires’ he has no
individual in mind about whom he wishes to be understood to be talking, rather
he wants to talk about whoever is the hominid who discovered how to start fires.
In this kind of situation, Neale’s first proposition will be false, and his
second proposition will simply not get off the ground: since there is no x such
that s is indicating x, it is, a fortiori, not the case that ((there is a unique
x such that s is indicating x and x is F) and x is G); and, since nothing
satisfies the descriptive material in the general proposition, there is no
object for the singular proposition to be about.
We must therefore look for
another kind of hybrid story, and it is the business of the rest of this paper
to outline and defend an account of this sort.
5
Complex demonstratives and communication
As
I suggested at the beginning of this paper, the question I want to address
directly is what role complex demonstratives play in communication. Once we have
a clear idea on this, we can then move on to look at the key concern of the
philosophical semanticist, i.e. the truth conditions of the propositions
expressed by complex demonstrative utterances. Hopefully an answer to the first
question will point us in the direction of a principled answer to the second.
All the accounts we have looked
at so far agree, either tacitly or overtly, on one underlying principle: that,
whatever the meaning of complex demonstratives may be, it must be analysed along
more than one dimension. For character-theorists like Braun (1994) and Borg
(2000), a satisfactory account must recognise the different contributions
complex demonstratives make to character and content; for quantificationalists
like King (2001), complex demonstratives are semantically mandated to introduce
into propositional content not just the property expressed by their nominal, but
also a further property determined by context; and for hybridists like Richard
(1993) and Neale (1999), complex demonstrative contribute both the property
expressed by their nominal and their referent to propositional content, albeit,
on Neale’s account at least, to the content of different propositions. All of
these different versions of what we might call the two-tier story on
complex demonstratives are expressions of the same underlying intuition: that
understanding a complex demonstrative involves both the linguistically-given
nominal and some extra non-linguistically-given (i.e. contextually determined)
element; when we understand ‘that F’, we understand firstly that what is
being talked about is an F, and secondly that it is a particular F being talked
about, and, in order to establish which F, we need to make use of
something other than the nominal. I have intentionally attempted to leave this
intuition very vague: it can, as witnessed by the variety of accounts discussed
above, be developed in widely differing ways. But it nonetheless expresses what
is possibly the only common ground between all the accounts currently available.
And, given this, I suggest we should take it pretty seriously.
I propose, then, that we start
from the following working hypothesis: ‘that F’ is a tool that speakers use
to talk about particular Fs. The tool works by indicating two things: (i) that
the hearer is intending to talk about a particular F; and (ii) that being an F,
in the context of utterance, is not the only way in which the speaker is
thinking of her intended F. As things stand, this hypothesis is, obviously
enough, extremely vague: in particular we are going to need to make much more
explicit the notions of thinking about and talking about an F. To
indicate roughly the way things are going to go on this, I intend to cash these
notions out in terms of the disjunction of Russellian acquaintance and
description; to think of an F in the way indicated by the use of a complex
demonstrative, one must either be able to think of that thing by acquaintance or
by description. It is not, therefore, on the account I am going to outline,
necessary to be acquainted with the designata of one’s uses of complex
demonstratives and thus not necessary to be acquainted with those designata in
order to understand complex demonstratives utterances.
On
the face of it, this kind of view seems to have much in common with the
quantificational account of King (2001). For King, uses of complex
demonstratives can be backed by either acquaintance or description, in King’s
terms by perceptual intentions or descriptive intentions. And on
King’s account, too, understanding complex demonstratives involves grasping
both that what is being talked about, in the appropriate sense of talked
about, is an F and that there is some other way in which the F being talked
about is to be thought of, the property that saturates King’s second argument
place. Although the account I shall advocate diverges from King’s in at least
one fundamental respect, there is a fair amount of common ground. Given this, I
would like, before progressing to a more articulated formulation of my own
account, to highlight a point, in addition to those already discussed above, at
which King’s story seems to get things wrong. An examination of this
shortcoming will, I hope, point us in the direction of a more complete analysis.
For
King, it is typical of uses of complex demonstratives backed by descriptive
intentions that the property which saturates the second argument position, the
argument position reserved for first-order properties determined by speaker
intention, is redundant. He asks us to imagine Danielle who knows, on purely
general grounds, that there is currently one and only one person swimming across
Lake Tahoe. Intending to talk about whoever it is that is currently swimming
across Lake Tahoe, Danielle utters:
(26)
That person swimming across Lake Tahoe now must be cold
On
King’s analysis, the first argument position is, as ever, saturated by the
property expressed by the nominal, i.e. the property of being a person swimming
across Lake Tahoe now. But, for King, that is also the content of the
descriptive intention which backs up Danielle’s use of the complex
demonstrative: her intention is to talk about whoever has the property expressed
by the nominal. It is therefore the case that the second argument position is
saturated with just the same property as saturates the first argument position.
The speaker’s intention is thus redundant[9].
On the rough sketch of the
communicative role of complex demonstratives given above, this is at least an
undesirable result: what I have suggested is that complex demonstratives are
communicative tools designed for a particular purpose, that purpose being to
guide the hearer’s interpretive process in (at least) two ways. If the speaker
only has one way of thinking of whatever she wants to talk about, then it would
seem that, on the picture above, a complex demonstrative is the wrong tool for
the job. I accept, however, that if this were the only reason to resist the
redundancy in King’s account, it would certainly not be reason enough. But
there is some empirical evidence to suggest we might do well to think again.
Consider an example that seems to be parallel to King’s example in (26):
(27)
That oldest man in the world must be worried about mortality
Here,
as in King’s example, the speaker can, self-evidently, believe on purely
general grounds that there is one and only one oldest man in the world, given
any story on the semantics of superlatives[10].
And here, again, as in King’s example, the speaker’s intention seems
redundant, since, ex hypothesi, she wants to talk about whoever is the
oldest man in the world. And yet (27) seems infelicitous in a way that (26) isn’t.
Why might that be? What distinction can we draw between (26) and (27) that might
explain why one is felicitous and the other not?
Another type of evidence seems
to point in the direction of a plausible answer. Take the sentence in (28):
(28)
That dog with three legs is called ‘Lucky’
I
want to outline two possible contexts in which (28) might be uttered: in the
first context, Janet and John are standing in a room full of dogs, all but one
of which have the standard canine allocation of legs. Janet turns to John and
utters (28). In the second context Janet and John are again in a room full of
dogs, but now all of the dogs are three-legged. Pointing at a particular dog
across the room Janet utters (28). There seems to be an intuitive difference
between the work done by the nominal ‘dog with three legs’ in these two
contexts: in the first context, Janet is using the nominal not only to pick out
which kind of thing she’s talking about (she’s talking about a dog), but
also which individual within that kind she’s talking about (the one with three
legs). In the second context, by contrast, she is simply using the nominal to
indicate which kind of thing she’s talking about (she’s talking about a dog
with three legs), and John must turn to other, non-linguistic factors to
establish which particular individual within this kind Janet is talking about
(Janet’s demonstration is going to give him a big clue).
What might the distinction
between Janet and John’s two contexts tell us about complex demonstratives?
Let me suggest an answer: saturating King’s second parameter doesn’t always
have to be a purely non-linguistic matter; sometimes the speaker can give the
hearer information via her utterance that allows him either fully or partially
to saturate this parameter[11].
So, returning to (28), in the first context, the property expressed by ‘dog’,
not ‘dog with three legs’, will saturate the first parameter (the parameter
we’ve so far been thinking of as the nominal parameter), and the property of
having three legs will go some way towards saturating the second parameter. In a
context such as this, however, where speaker and hearer are in an immediate
perceptual relation with the object the speaker is intending to talk about,
perceptual/causal properties of the referent will no doubt also enter into the
saturation of the second parameter. Compare an utterance of (28) in the first
context to an utterance of (9), repeated as (29):
(29)
That professor who brought in the biggest grant in each division will be
honored
for
which linguistically given material (‘who brought in the biggest grant in each
division’) will entirely saturate the second argument position.
In
the second context above, the nominal of (28), as we’ve already seen, only
goes to identify the type of thing that Janet is intending to talk about; it is
left to contextual indicators to determine which individual of that type Janet
wants to talk about. Translating this into King’s terms, the property
expressed by ‘dog with three legs’ saturates just the first parameter, with
the second parameter left entirely to context, albeit a context manipulated by
Janet’s demonstration.
How does this story help us
with the data we started out with, the infelicity of (27) in comparison to (26)?
If we reanalyse (26) along the lines just sketched, it would seem natural to say
that the kind of thing Danielle is intending to talk about is a person. And
which person? Well, the one swimming across Lake Tahoe now. In other words, the
first parameter is saturated by the property of being a person, while the second
is saturated by the property of being swimming across Lake Tahoe now. Both
parameters are thus saturated. But what of (27)? In (27) there are no two
properties that are separable in this way: the property of being the oldest man
in the world is not the property of being both the oldest man and in the world.
It is this, I suggest, that makes (27) infelicitous. Why? Because, contrary to
King’s account, the parameter which corresponds, on his story, to speaker
intention, cannot be redundant: complex demonstratives are tools for picking
things out via two or more distinct routes, and uses of complex demonstratives
for which, as in (27), there is only one way of thinking of the speaker’s
intended referent, are thereby rendered infelicitous.
There
are two extra pieces of evidence to suggest that this sort of story is along the
right lines. Firstly, consider (30):
(30)
That person who is the oldest man in the world must be worried about
mortality
To
my ear, (30) is less infelicitous than (27). On the accounts we have looked at
so far in this paper, all of which treat the nominal as a whole unit, there is
no obvious explanation of this difference. On the kind of story I am sketching,
however, this contrast is to be expected: in (30), unlike in (27), there are two
separable properties (the property of being a person and the property of being
the oldest man in the world) which can be prised apart to saturate both of King’s
first two parameters.
Secondly,
imagine that on Tuesday Janet is reading the newspaper and sees an article about
the oldest man in the world which she discusses with John. On Wednesday, having
mulled things over in the last twenty-four hours, John utters (27) to Janet.
Again this seems less infelicitous than does (27) uttered in a context without
such a background. Why? Because in this context there is another ‘route in’
to the person John is intending to talk about: he’s the person they were
talking about yesterday. The second parameter thus doesn’t go unfilled. There
still does seem to me to be a certain oddness to an utterance of (27) even in
this context, albeit less than before, and this too can be explained in the
terms laid out above: complex demonstratives are designed in such a way that the
hearer interpreting a complex demonstrative first goes to a class to which the
thing talked about belongs, and then goes beyond that to discover which member
of the class is being talked about. But of course for any utterance of (27), the
nominal property will determine a unique individual, thus leaving the second
parameter, although now filled, not doing any semantic work over and above the
work already done by the nominal. Again this seems to go against the grain of
complex demonstratives, although not as seriously as where there is no property
to saturate the second parameter at all. Given this kind of data, it would seem
helpful to introduce terminology to distinguish that part of the nominal which
goes to saturate King’s first parameter, the part of the nominal which makes a
complex demonstrative a complex demonstrative, and that optional part of the
nominal which can go towards saturating King’s second parameter. I propose to
call the former a nominal sortal and the latter a nominal individuator.
Let
me then briefly summarise the kind of picture that the evidence above seems to
point towards: complex demonstratives are communicative tools of a particular
kind, they are tools for talking about individuals; given their meaning, they
indicate that the individual in question is being thought of in at least two
ways, firstly as a member of a particular class and then via some other route or
routes that distinguish(es) them from all other members of that class. This is,
of course, no more than the vaguest formulation of the outline of an analysis of
complex demonstratives: we have, for instance, said nothing yet about how to
cash out the two-routes notion, or about what the truth conditions of complex
demonstrative utterances might be. It is the purpose of the next section of the
paper to firm up the proposed analysis.
6
A semantics for complex demonstratives
Before
tackling the semantics of complex demonstratives head on, I want to put in place
one piece of preliminary groundwork. I take it as uncontroversial from a
cognitive perspective that having a concept of an individual is a necessary
prerequisite for entertaining thoughts about that individual[12].
However, it has been a philosophical commonplace since Russell that there are
two fundamentally different ways of thinking of individuals: in Russell’s
terms, one may think of an individual either as an object of acquaintance or via
description. This distinction, and the discussion it has given rise to, are
familiar and I don’t intend to go into them in detail here. The lesson I wish
to draw, however, is simple: there are ways of thinking of individuals which are
truth conditionally singular, in the sense that all they contribute to the
truth-conditions of propositional representations is their referents, while
there are other ways of thinking of individuals that are general, in that they
give rise to general, i.e. quantificational, truth conditions. I shall call the
former de re individual concepts[13]
and the latter descriptive individual concepts[14].
With that piece of groundwork
in place, I can now spell out in greater detail the story on complex
demonstratives which I have so far only sketched. I shall initially express the
details of my analysis as a more explicit version of the communication-based
account above, which will focus on the mapping from linguistic form to mental
representation. Once I have done that, however, I shall spell out the
truth-conditional implications. It is only by separating out the path from
language to mind and that from mind to world that, so I shall claim, we can
properly appreciate how complex demonstratives work. As far as the first of
these paths is concerned, my proposal looks like this: as for all linguistic
expressions, a complex demonstrative provides the hearer of an utterance in
which it occurs with information about the thought which the speaker is
entertaining and intending to express. The information carried by a complex
demonstrative ‘that F’, where F is a nominal sortal, comprises three
aspects:
i)
The component of the speaker’s mental representation
which corresponds to the complex demonstrative in her linguistic representation
is an individual concept, be it de re or descriptive;
ii)
That individual concept is a concept of an F;
iii)
That individual concept is a concept of an O, where O is a
further property or conjunction of properties over and above the property of
being an F.
We
know what sort of property is expressed by the nominal sortal, but what sorts of
properties can do service as O-properties, what can they be properties of, and
where can they come from? The possibilities are pretty well open ended. Take the
following sentences, some of which we’ve seen before:
(31)
That man has no dress sense
(32)
That man wearing a kipper tie has no dress sense
(33)
That hominid who discovered how to start fires was a genius
(34)
That oldest man in the world must be worried about mortality
(35)
That painter was great at painting hands
Janet
points to a man across the room and utters (31) to John. What is the O-property
in this case? In what way or ways other than simply as an F is Janet thinking of
the F she intends to talk about? She is presumably thinking of him in a way that
is linguistically unrepresented[15]:
she is thinking of him as the object with which she stands in a range of
perceptual/causal relations. The same sorts of causal relations will also appear
as an element of the O-property for (32), but in this case they will not be
alone. Here an element of the O-property is linguistically indicated by the
nominal individuator: Janet’s individual concept is a concept of something
wearing a kipper tie. Linguistic and non-linguistic elements thus combine in
cases such as (32) to give the O-property. In (33), depending of course on the
context of utterance, all the work may be done linguistically: the O-property
may simply be the property of having discovered how to start fires. For (34), I
would like you to think back to the context in which Janet and John have
previously been discussing the oldest man in the world. What will the O-property
be here? As I suggested above, the O-property (and thus what allows John
felicitously to use the complex demonstrative) is the property of being the
subject of a particular previous conversation. In (35) something rather
different is going on: imagine that Janet and John are in a gallery, and, seeing
a portrait in which the hands are particularly well painted, Janet utters (35).
What is the O-property here? What I would like to suggest is that it is the
property of standing in a particular relation, the relation in which x stands to
y iff x has painted y, to the painting which Janet has made
salient. The moral of this discussion is that there is, as far as I can see, no
antecedent restriction on the kind of property that can do work as an
O-property: it can be descriptive or perceptual, it can derive from a relation
that holds directly between the speaker and what she wants to talk about or from
a relation between the speaker and something that stands in a relation to what
she wants to talk about, and so on.
The picture painted so far,
then, looks like this: complex demonstratives are tools used by speakers to
indicate that the thought they intend to express contains an individual concept;
beyond this they indicate that the individual concept in question is a concept
of something that satisfies both the nominal sortal and another property which
may or may not be partly or wholly determined by a nominal individuator. But
what is the hearer supposed to do with this information? What, in other words,
constitutes understanding a complex demonstrative utterance? Firstly the hearer
must grasp the properties F and O, but what does it mean to grasp these
properties? Firstly the hearer must assign a structure to the nominal, i.e. he
must pragmatically infer whether the nominal is to be taken as a sortal or as
both a sortal and an individuator. Once he has done this, the process of
retrieving the F-property is straightforward: it is just the property expressed
by the sortal. Next the hearer must retrieve an O-property. This may, as we have
seen, be wholly or partly determined by a nominal individuator; if pragmatic
inference leads the hearer to conclude that the O-property is wholly determined
by the nominal individuator, then again all he need do is work out which
property is expressed by the individuator in order to establish which is the
O-property. If, however, the nominal individuator doesn’t wholly determine the
O-property, or if there is no nominal individuator, the hearer must look beyond
the linguistic content of the utterance itself to discover such a property in
the context, i.e. to discover, via contextual cues and pragmatic inference, the
other way, apart from as an F, that the speaker is thinking of what she intends
to talk about.
The first step of the
interpretation process is, then, to establish which are the F and O properties.
But, once the hearer has done this, what is he to do with the properties he is
left with? Given the information carried by complex demonstratives as an
expression type, he knows that the intended interpretation is an individual
concept. The next stage of the process must, therefore, be to find an individual
concept of an F and an O. But which individual concept? Or rather, which kind of
individual concept? In answering this question, the hearer’s inference will be
guided by the kind of property which is doing work as the O-property: if it is
the kind of property, i.e. perceptual/causal, typical of de re individual
concepts, then the hearer should infer that the individual concept entertained
by the speaker is de re and thus that the intended interpretation is a de
re individual concept; if, on the other hand, the O-property is the kind of
property, i.e. satisfactional, typical of descriptive individual concepts, then
he should infer that the hearer’s concept is descriptive. Although complex
demonstratives are, therefore, neutral between de re and descriptive
individual concepts as far as their lexical semantics are concerned,
establishing which kind of concept constitutes the intended interpretation is a
necessary part of understanding a complex demonstrative utterance. To summarise,
the hearer’s task, in interpreting a complex demonstrative, is to find an
individual concept of the appropriate kind of an F and an O.
So far, however, I have not
confronted the accounts outlined in the first half of this paper head on: I have
not laid out the truth-conditional implications of the story I am advocating. I
suspect that, from the truth-conditional perspective, my account comes close to
being a hybrid of the stories of King (2001) and Braun (1994). With King I agree
that the semantics of complex demonstratives involve two separate properties,
although we part ways over the details of this; I also agree with King that
which property is retrieved as that not expressed by the nominal (or the nominal
sortal in my terms) will have a knock-on effect on the kind of
truth-conditional content assigned to the utterance as a whole: for King, a
perceptual intention leads to truth conditions anchored in the context of
utterance while a descriptive intention does not, for me perceptual properties
lead (at least standardly) to de re individual concepts, while
satisfactional properties lead to descriptive individual concepts. However,
where my disagreement with King is the clearest is in the treatment of
referential uses of complex demonstratives. And it is here that the predictions
of my story have much in common with Braun’s account: for Braun, the nominal
acts as a dimension of character, i.e. acts to constrain reference, for me the
individual concept which constitutes the interpretation of ‘that F’ must be
a concept of an F (and, of course, an O). For both Braun and myself, therefore,
‘that F’ is a tool for talking about Fs, and cannot properly be used to talk
about non-Fs.
Having
mentioned the common ground between my account and those of King and Braun, let
me then spell out how the truth conditions go on my story. The lexical semantics
of complex demonstratives must be explicated in terms of the mapping from
linguistic to mental representations: complex demonstratives may be construed as
functions from contexts to individual concepts. Thus, ‘that F is G’ will be
true just in case G*IC(F*, O) is true, where G*IC(F*, O)
is a mental representation formed by combining the property expressed by G
with an individual concept of the appropriate type of an F and an O (F* being
the property expressed by F). This gives us one step towards a truth
condition for complex demonstrative utterances, but what of the step from mental
representation to world? When, in other words, will G*IC(F*, O) be
true? This will of course depend on which kind of individual concept IC(F*, O)
is: if it is a de re individual concept, G*IC(F*, O) will be
true just in case G* is a property of its referent; if it is a descriptive
individual concept, then it will, no doubt, have something like a standard
descriptive truth condition, i.e. it will be true iff [the x: F*x & Ox]
(G*x).
The
key truth-conditional prediction of this account is, therefore, that, although
they are lexically univocal, complex demonstratives can give rise to genuinely
referential or genuinely quantificational truth conditions, according to speaker
intention. And this is, given the evidence canvassed in the earlier part of the
paper, just what we should want from an analysis of complex demonstratives.
7
Defending the analysis
The
analysis I am advocating is consistent with the data, particularly the scopal
data, laid out above. However, I would like, in the last part of this paper, to
raise some potential objections to the account and to explain why I do not think
these objections pose any genuine threat.
The first potential objection
is that, having argued that we need a univocal story on complex demonstratives,
what I have actually provided is an ambiguity account. On my analysis, one and
the same complex demonstrative utterance can, in different contexts, have either
genuinely singular of genuinely quantificational truth conditions; what, so the
argument might go, could be more ambiguous than that?
There is certainly something
right in this objection: I am proposing an account on which complex
demonstratives can give rise to very different kinds of truth conditions, so, if
all that is being claimed is that, on my account, complex demonstratives are truth-conditionally
ambiguous, then I’m going to have to put up my hands to that. My claim,
however, is that, on my story, complex demonstratives are not ambiguous in any
way that we should worry about. Let me briefly rehearse the familiar
methodological claims about ambiguity to demonstrate why. According to Grice’s
Modified Occam’s Razor, ‘senses are not to be multiplied beyond necessity’
(Grice (1967, p.47)). Given two competing theories about the interpretation of a
linguistic expression, one of which posits an ambiguity and the other of which
posits none, we should be ready to accept the latter theory. How might a theory
without an ambiguity do the same work as one with an ambiguity? By handing
everything beyond the univocal semantics over to pragmatics. The essence of
Modified Occam’s Razor, then, is that an expression should be viewed as
ambiguous iff one cannot account for it via a univocal semantics plus
pragmatics. But this is just what my account of complex demonstratives does: it
takes a single semantics, a lexically encoded meaning which constrains the
mapping from linguistic to conceptual representation for all complex
demonstratives, and leaves the rest, i.e. which is the O-property, whether the
intended individual concept is de re or descriptive etc., to pragmatics.
In any theoretically significant sense, therefore, the account I have proposed
is not an ambiguity account.
A second potential objection
might be levelled at the predictions my account makes on referential uses of
complex demonstratives[16]. There are, first of all,
the kind of entailment data exemplified by (3) and (4), repeated here as (36)
and (37):
(36)
That green car is very old
(37)
Some green car is very old
I
shall say no more about these than to accept the sort of story told by Braun:
although there is not an entailment, there’s the next best thing, since the
complex demonstrative ‘that green car’ can only be interpreted via an
individual concept of a green car.
Secondly,
Lepore and Ludwig ask us to consider various situations in which the nominals of
referentially-used complex demonstratives appear to be playing some role in
truth-conditional content (beyond constraining reference, that is). Consider,
for instance, (38) uttered while indicating a particular individual:
(38)
Each woman in this room admires that man whom she sees at the podium[17]
Their claim concerning this example is that, since there is an element within the complex demonstrative that is anaphoric on an element outside, the content of the nominal cannot drop out of propositional content. This would certainly weigh heavily against accounts, such as mine, on which referentially-used complex demonstratives contribute nothing but their referents to propositional content. I do not believe, however, that there is any problem with examples such as these. Firstly, it seems to me that the referential reading that Lepore and Ludwig are after here, where a particular individual is being picked out as ‘that man who she sees at the podium’, is markedly less natural than a descriptive reading, and of course there is no problem, on my account, with anaphoric relations being set up between a quantifier outside a non-referentially used complex