Professional Documents
Culture Documents
THINKING
ABOUT
PROPOSITIONS
Jeffrey C . King, Scott Soames, & Jeff Speaks
King020513OUK.indd ii 11/23/2013 12:58:33 PM
New Thinking about Propositions
Jeffrey C. King
Scott Soames
Jeff Speaks
Introduction 1
Jeff Speaks
Bibliography 245
Index 251
Beginning in the 1970s and 1980s, many philosophers of language found themselves in
an increasingly difficult situation.
On the one hand, many came to believe that in order to do semantics properly as
well as to give an adequate treatment of the attitudes, one needed to posit certain enti-
ties—propositions—which could be the meanings of sentences (relative to contexts),
the contents of mental states, and the primary bearers of truth and falsity. But largely
due to the arguments of Scott Soames1 many also came to distrust the standard theo-
retical account of the nature of propositions, which treated them as sets of worlds, and
came to think of them instead as structured entities of some sort.
There was however no consensus about what these structured entities could be.
A standard way of talking around the problem was to point out that propositions could
be represented as ordered pairs. But it is pretty plain that this is just a way of talking
around the problem—to say that propositions can be represented as ordered pairs is
not to say what they are. The unsatisfactory situation persisted.
Jeff King’s 2007 book, The Nature and Structure of Content, had a decisive impact on
this situation in two ways. First, it made clear just how unsatisfactory the situation was,
by stating the desiderata on a theory of propositions—desiderata which thinking of
propositions as ordered pairs, for instance, plainly did not meet. Second, and perhaps
more important, it made clear that the situation did not have to remain unsatisfactory;
the book presented a novel view of propositions which both provided a clear meta-
physical account of their nature, and made the case that propositions, so understood,
could play the roles for which philosophers of language and mind wanted them in the
first place. It made clear that progress on the question of the nature of the proposition,
the question which so puzzled Frege, Russell, and the early Wittgenstein, is possible.
This book is a contribution to the ongoing discussion to which King’s book, as well as
the work of others, gave rise.2
The three co-authors of this book agree on the views which led to the unsatisfac-
tory situation described above. The aim of the first Part is to make the case for this
1
See especially Soames (1988).
2
We can’t, of course, hope to discuss all of this recent work. See, among other places, Collins (2011),
Gaskin (2008), Hanks (2009), Hanks (2011), McGlone (2012), Moltmann (forthcoming).
constellation of views in two ways: first, by arguing that we need propositions in our
accounts of language and the mind, and in our semantic theories; and, second, by
arguing that traditional accounts of propositions—the classical theories of Frege and
Russell, and the view of propositions as sets of possible worlds—are not up to the task.
A subsidiary aim of this part of the book is to make plain some of the constraints
which a successful theory of propositions should have to meet. By seeing how certain
views of propositions, or attempts to do without them altogether, fail—for example by
failing to provide suitable objects of the propositional attitudes, or to give an adequate
semantics for a certain class of sentences—we get clearer on the roles propositions
must play. Each of us thinks that we should be able to give some account of the entities
which play these roles. None of us is committed to the claim that this account should
respect commonsense or folk intuitions about what these entities are (to the extent that
such intuitions exist). But, on the other hand, the role these entities play—with respect
to our mental states and our language use—means that they must be the sorts of enti-
ties to which ordinary users can refer and so, in that sense, they must not be completely
beyond the ken of non-philosophers.
At the end of Part I the agreement ends. In Part II, each of us lays out and defends
his favored theory of propositions. King elaborates and refines his view of propositions
as facts; Soames extends his view (first defended in his 2010 book, What is Meaning?)
of propositions as cognitive event types; and I defend the view that propositions are
a kind of property. The essays in Part III criticize the views defended in Part II; and
in Part IV each of us presents some further thoughts on the task of giving a theory of
propositions.
Propositions are also thought to be the things that possess modal attributes such as
being possible, necessary and impossible. It certainly does seem as though the primary
bearers of truth and falsity are also the things that possess these modal attributes. We
do, after all, talk of something being necessarily true, thereby attributing both being
true and being necessary to one and the same thing.
Then there are the things we believe, doubt, assume and suspect. It is thought that
propositions are the things we bear these various attitudes to. Hence, e.g. believing that
snow is white is a matter of bearing a certain sort of cognitive relation to the propo-
sition that snow is white. Indeed, this view is so ingrained among philosophers that
attitudes such as belief and doubt are generally called propositional attitudes. Again,
this fits nicely with the other roles played by propositions, since it does seem that the
things we believe and doubt are true or false and possess modal attributes like being
necessary, possible and so on.
Our perceptual experiences seem to represent the world as being certain ways. My
current perceptual experience in some sense seems to represent the shirt I am wear-
ing as black. Further, it seems that my perceptual experience can represent the world as
being a certain way accurately or inaccurately. But this suggests that perceptual experi-
ences have accuracy conditions. This leads many to think that perceptual experiences
have contents that are accessible for accuracy. And this in turn leads many to believe
that perceptual experiences have as their contents some sorts of propositions. So here,
propositions play the role of being the contents of perceptual experiences and thus what
explains how perceptual experiences can represent the world accurately or inaccurately.
The relation between the propositional contents of perceptual experiences and of natu-
ral language sentences is a matter of some controversy and will come up in the sequel.
For philosophers who countenance possible worlds, propositions offer an appealing
way to understand possible worlds.1 Possible worlds can be identified with complete,
consistent sets of propositions.2 There are various ways to define completeness and
consistency, but the idea behind a set of propositions being complete is that any propo-
sition is to be true or false relative to a complete set of propositions. As to consistency,
the idea is that the members of the set could have been jointly true.
We turn now to some of the roles played by propositions in philosophy of language.
Having mentioned earler that propositions are the things we believe, doubt and so on,
they are also the things we assert when we utter sentences. Asserting, like believing, is
a relation between an individual and a proposition. It is widely agreed that to explain
a variety of phenomena we need some notion of context that gives us something like
the record of the conversation to that point. What exactly this record must contain
is a matter of some dispute, but at least since the pioneering work of Stalnaker [1978]
it is widely agreed that the record must include the material that the conversational
participants are taking for granted in the conversation and recognize each other to be
1
See Adams [1974]
2
Or, as in Soames [2007], possible worlds could be identified with the property of making all the proposi-
tions in a suitably defined set of propositions true.
so doing. This material will include things that have been asserted and accepted in the
conversation. But such things are propositions. Hence propositions will play a signifi-
cant role in characterizing context/conversational record.
We have mentioned that propositions serve as the semantic contents of sentences
relative to contexts. But propositions will figure in semantic theory in other ways as
well. In particular, we explain the semantics of “that” clauses by saying that they desig-
nate propositions.3 Thus, the standard view is that a belief ascription such as
1. Amy doubts that snowboarding is hard.
asserts that Amy stands in the doubting relation to the proposition designated by
“that snowboarding is hard,” viz., the proposition that snowboarding is hard. On this
account, “doubt” expresses a two-place relation between individuals and propositions,
with “Amy” and “that snowboarding is hard” designating an individual and a proposi-
tion, respectively. As many have noted, this account looks able to explain why infer-
ences such as the following are valid
Amy believes everything Carl said.
Carl said that skiing is hard.
So Amy believes that skiing is hard.
For on the account in question, this argument has the following structure:
For every x such that Carl said x, Amy believes x.
Carl said P.
So, Amy believes P.
Advocates of propositions hold that “that” clauses designating propositions are used
to attribute other properties to propositions as well. Consider the following:
2. That Harry is ignorant is true.
3. That Harry is ignorant is likely.
4. That Harry is ignorant is certain.
5. That Harry is ignorant is possible.
2 attributes the property of being true to the proposition that Harry is ignorant,
whereas 3, 4 and 5 attribute the properties of being likely, being certain and being pos-
sible to that proposition.
Those who believe in propositions will also hold that they sometimes are designated
by demonstratives. For example, suppose I say “Skiing is hard.” You may reply with
either of the following:
6. That’s true.
7. Glenn said that too.
3
I cautiously use the word “designate” to avoid committing myself to the claim that “that” clauses refer to
propositions. See King [2002a] and King [2007] Chapter 5. Further, it may be that “that” clauses sometimes
designate things other than propositions—facts—e.g. in so-called factive contexts such as “Jeff resents that
the ski season is over.” King and Soames debate the issue in subsequent chapters.
4
Again, some expressions of this form might designate facts. For example, “what Glenn regrets.” See
previous note.
In this paper I’ll focus on two arguments against semantic theories which wish to avoid
commitment to propositions. Each has been discussed in various forms in the litera-
ture. The first holds that on the most plausible semantics of a class of natural language
sentences, the truth of sentences in that class requires the existence of propositions;
and some sentences in that class are true. The second holds that, on the best under-
standing of the form of a semantic theory, the truth of a semantic theory itself entails
the existence of propositions.
It is the job of a semantic theory for a language to give the meanings of expressions of
the language, part of which task is revealing, in a sense which I will leave informal, the
logical structure of sentences of the language. It seems plausible that, at a first pass, the
logical forms of the sentences of the apple sentences are, respectively,
ූx John ate x.
[the x: John ate x] delicious apple(x)
[the x: John ate x] delicious apple(x) & (John waits another day before eating x ☐→ x is rotten)
[the x: John ate x] Mary gave x to John.
Some evidence for these interpretations of the apple sentences is that the apple sen-
tences seem to jointly entail:
(1) There is something which John ate, which was a delicious apple, which could have been rot-
ten, and which Mary gave him.
(1F) ූx (John ate x & x is a delicious apple & x could have been rotten & Mary gave x to John)
These sentences seem, at a first pass, to have something like the following as their
logical forms:
1
I’m ignoring for simplicity worries about incomplete descriptions, and the assumption that the anteced-
ent of the above counterfactual is possible.
As with the apple sentences, we can defend these interpretations of the proposition
sentences by noting that the proposition sentences seem to jointly entail:
(2) There is something which John said, which was true, which could have been false, and which
Mary believed.
And the logical forms assigned to the proposition sentences explain why they
entail (2F).
If (2F) is true, then there are things which are said and believed, which are the bear-
ers of truth values and have modal properties like being possibly true. So if (2F) is
true, there are propositions. Hence the claim that (2F) is true is a claim which any-
one who wants to develop a semantic theory without commitment to propositions
must deny.
The problem is that a plausible argument can be made that our semantic theory
should be committed to the truth of (2F). We can think of this argument as having
two independent premises. First, that sentences like the proposition sentences are
sometimes true, and second, that the correct semantic analysis of these sentences is
(roughly) the one given above. So one who wants to deny that (2F) is true must either
deny that the analyses of the proposition sentences given above is correct—and so
deny that the proposition sentences really do jointly entail (2F), even if they do entail
(2)—or must deny that the proposition sentences are true.
Before considering these two strategies for blocking the argument for (2F), let’s
pause for a moment to consider the claim that (2F) entails the existence of proposi-
tions. Though this claim might seem obvious, there is one complication here worth
noting. What (2F) and related claims say is that there is a type of entity which is what
speakers say, what subjects believe, and is the bearer of truth and falsity as well as pos-
sible truth and falsity. Couldn’t one say that there is such a type of entity—but that it is
not a proposition?
One could, because standardly propositions are taken not just to have the character-
istics attributed by (2F), but also to be language-independent abstract objects. We can
bring out the complication by considering a “sententialist” view according to which
sentences, rather than propositions, are the objects of attitudes like saying and believ-
ing, and the bearers of truth and falsity. According to this sort of sententialist view,
(2F) is true—but still, one might think, we’re not thereby committed to the existence of
propositions.
Fortunately for our purposes, we can avoid this complication by pointing out that
this version of sententialism is false.2 Were this theory true, then the truth of “Violet
believes that the sky is blue” would entail that Violet stands in the belief relation to
the sentence, “The sky is blue.” But this ascription could be true even if Violet were a
monolingual French speaker who stood in no special relation at all to this sentence—
and indeed could be true even if Violet spoke no language at all.
And given that sententialism is false, there don’t seem to be any other plausible can-
didates for objects which are not language-independent abstract objects which could
satisfy (2F). (This, of course, leaves completely open the question of the nature of the
abstract objects which satisfy sentences like (2F)—but at this stage, that’s what we
want.) Hence I’ll proceed by ignoring the sententialist alternative, and assuming that if
(2F) is true, then there are such things as propositions.
Let’s turn to our first strategy for blocking the argument from the proposition sen-
tences to (2F): denying the argument’s validity. One question for the the proponent
of a semantic theory which accepts the truth of the proposition sentences and (2) but
denies (2F) is whether they want to say similar things about the apple sentences, (1),
and (1F). Either way we go here, we seem to get in trouble. On the one hand, it is pretty
hard to deny that (1F) gives an accurate rendering of (1).3 On the other hand, it is pretty
hard to deny that the relationship between the apple sentences and (1F) appears to be
the same as the relationship between the proposition sentences and (2F). Semantic
theories, like other theories, are to be judged in part by their capacity to give expla-
nations which unify apparently similar phenomena. For this reason alone, we have
some reason to doubt a semantic theory which offers radically different explanations
of the fact that the apple sentences entail (1) and the fact that the proposition sentences
entail (2).
But, setting aside the point that the apple sentences and the proposition sentences
seem, on their face, to demand parallel treatment, there remains the problem of
how—if not via the analyses above or some notational variant thereof—to understand
the logical form of sentences like the proposition sentences and (2), while capturing
entailments like that between the proposition sentences and (2). Many attempts have
been made to provide analyses of sentences like the proposition sentences which,
unlike (2F), don’t entail the existence of propositions. Here I’ll focus on the most
2
The locus classicus for the following objection is Church (1950).
Not every view which is standardly classified as “sententialist” fits this mold. For example, the paratactic
analysis of Davidson (1968), which I discuss below, is often thought of as a sophisticated sententialist view,
but according to the paratactic analysis (2F) does not give the form of (2), and is not true. Hence I discuss
it under the heading of views which deny that the proposition sentences entail (2F), rather than as a view
which aims to reconcile (2F) with the denial of the existence of propositions.
3
Though, to be fair, some who doubt the existence of apples because of their views about mereology are
inclined to offer different interpretations of the apple sentences than the ones given above. (See, for example,
van Inwagen (1990), ch. 10.) Someone with metaphysical scruples about propositions—whether or not they
have doubts about apples—might on parallel grounds offer a reconstrual of the proposition sentences. I will
consider some candidate reconstruals below.
4
Note that that it’s not enough for the proponent of the paratactic analysis to point out that competent
speakers of a language have some grip on when a pair of utterances samesay each other. It is quite plausi-
ble that they do. The relevant question is not whether there is such a thing as the samesaying relation, but
whether this relation can be understood in a way which does not appeal to propositions.
learned the words to express her belief about the sky, and hence is not even disposed to
accept any sentence which samesays “The sky is blue”?5
For the friend of the paratactic analysis, the best move here is to appeal to token
belief states—thought of as states of or events involving the brain—rather than natural
language sentences.6 One can then say that the above ascription is true iff Violet stands
in the relation expressed by “believes” to some brain state which samesays “The sky
is blue.”
One might object that this view of belief ascriptions commits the paratactic anal-
ysis to a controversial theory of the mind—one according to which, necessarily, for
every belief of every subject, there is a corresponding belief state which (intuitively)
has the same content as the belief. Now this commitment by itself might not seem so
bad—after all, this sort of theory of belief is widely, even if not universally, held to be
plausible.
What seems to me a bit worse is the fact that the paratactic account is not only
committed to this theory of the mind, but also builds it into the meaning of belief
ascriptions. On this sort of theory, when ordinary speakers talk about belief, they’re
quantifying over the brain states of subjects of ascriptions, and making claims about
brain states samesaying particular utterances. I’m sympathetic to Stalnaker’s worry
that this sort of theory
makes a belief attribution carry more weight than it is plausible to assume that it carries. If it
were correct, belief attributions would be far more speculative, and believers far less authorita-
tive about their beliefs, than they seem to be. While theoretical and experimental developments
in cognitive psychology may someday convince me that I store my beliefs in a form that is struc-
turally similar to the form in which they are expressed and described in English, I don’t think
that my ordinary belief attributions commit me to thinking that they will.7
One way to bring out the oddness here is to imagine a philosopher expressing
Stalnaker’s skepticism by saying:
It is possible for a subject to have the belief that the sky is blue without having any particular
belief state with the content that the sky is blue.
Maybe our imaginary philosopher is making a false claim about what is possible for
believers; but could he really be uttering an outright contradiction?
But let’s set to one side these doubts about samesaying, and about the relation
between belief ascriptions and belief states. The main source of objections to the
5
For discussion see, among other places, Schiffer (1987).
6
See, for example, LePore and Loewer (1989), 112.
7
Stalnaker (1990), 230.
paratactic analysis is that the analysis seems unable to capture all of the entailments
which we should want our semantics for belief ascriptions to capture. Consider, for
example, the following inference:
1. Violet believes that the sky is blue.
2. The sky is blue.
————————————————
C. Violet believes something true.
This argument certainly appears to be valid—in the (usual) sense that it is impossi-
ble for its premises to be true and its conclusion false. But according to the paratactic
analysis, it is invalid.
For consider a world w in which “The sky is blue” means that the sky is red, and sup-
pose that in w Violet has a belief state which samesays (in w) “The sky is blue”—and
suppose that in w, as in @, the sky is blue. Then the second premise of our argument is
true; and on the paratactic analysis the first must be as well, since in w Violet does have
a belief state which samesays “The sky is blue.” But the conclusion of our argument will
be false—since, in w, Violet’s belief is not true. And this contradicts our claim that the
above argument is valid.
One might reply that this argument overlooks the fact that the paratactic analysis of
the first premise attributes to Violet a belief state which samesays a particular utterance
of “The sky is blue”—not the sentence type of which that utterance is a token. But this
doesn’t seem to help; just as sentences could have had different meanings than they
actually have, particular utterances or inscriptions could have had different meanings
than they actually have. To suppose the opposite would be to make the quite surprising
claim that particular sounds, or marks on a page, could have their meanings essen-
tially. Given that they don’t have their meanings essentially, we can let w be a world in
which the relevant utterance of “The sky is blue” has a different meaning than it actu-
ally does, and the argument proceeds as above.
A second reply would be to borrow a page from the post-Kripke descriptivist
playbook, and introduce rigidification into the analysis. One might say, roughly,
that “Violet believes that the sky is blue” is true iff Violet has a belief state which
samesays the utterance of “The sky is blue” in @ (at a particular time and location).
But this is open to what seems to me to be a decisive objection raised by Soames
(2002). Consider the ascription “Amelia knows that Violet believes that the sky is
blue.” This could be true at a world w in which Amelia has no thoughts about @, the
actual world. (Indeed, she may be in no position to refer to @, as opposed to worlds
similar to @.) But on the proposed modification of the paratactic analysis, this will
be impossible, since the analysis of “Violet believes that the sky is blue” will make
reference to @.
A quite different line of reply, defended by LePore and Loewer (1989) (110), is to
concede the conclusion that, according to the paratactic analysis, the above argument
is invalid, and to try to explain away its seeming validity. Their strategy is to say that
the argument seems valid to speakers because they are assuming the truth of an extra
premise like:
“The sky is blue” is true iff the sky is blue.
As competent English speakers, we know this premise to be true. And if we add this
premise to the argument this is enough to make the argument, given the truth of the
paratactic analysis, valid.
This seems to me unsatisfactory. In general, it is simply not true that arguments
which would be valid were we to add a premise which we all agree to be true seem valid
to us. Consider, for example,
Blue is my favorite color.
— — — — — — — — — — —
The color of the sky is my favorite color.
This argument does not even seem valid; students just introduced to the concept of
validity can see that it is not. Does that mean that such students harbor some doubts
about whether the sky is blue? Surely not. Rather, despite the fact that they know—and
know that each other knows—that the sky is blue, they are able to evaluate the validity
of this argument without importing this assumption as an extra premise. It is therefore
mysterious why we should not be able to do this with the argument above.8
Summing up: it seems to me plausible that even if we set aside doubts about the
samesaying relation, and about the use of belief states in the semantics of belief ascrip-
tions, the paratactic analysis fails to do the job which we wanted it to do: namely, to
account for the validity of inferences involving sentences like the proposition sen-
tences without appealing to propositions.9
Let’s set aside attempts to explain the truth of proposition sentences, and the fact
that they entail (2), without commitment to (2F) and the existence of propositions.
A quite different line of response to the inference from the truth of the proposition
sentences to the existence of propositions is not to deny its validity but simply to deny
its premise—namely, that sentences like the proposition sentences are true. This is,
strictly speaking, an issue outside the scope of semantic theory, since semantic theo-
ries are committed to claims about what it takes for a sentence of the object language to
be true rather than claims about which sentences of that language are true.10
But it is tempting to say that if any view deserves an incredulous stare, it is this one.
Those who deny that any proposition sentence can be true are committed to denying
8
One might reply that the difference is that while the students know that the sky is only contingently blue,
we mistake the contingent truth that “The sky is blue” is true iff the sky is blue for a necessary truth. But of
course philosophers are very familiar with the fact that biconditionals of this sort are only contingent, and
the argument about Violet’s beliefs still seems obviously valid to us.
9
For an excellent summary of other objections, along with replies to those objections, see LePore and
Loewer (1989).
10
With the exception of logical truths and, in the case of a possible worlds semantics, truths which are true
in every index.
that, strictly speaking, anyone ever says anything, believes anything, or wants any-
thing. Indeed, it is one of those views that is, unless you’re being careful, hard to state
without self-refutation—as the preceding sentence illustrates.
How might one make plausible the denial that the proposition sentences are true?
One currently popular strategy would be to endorse a fictionalist theory of our
proposition-talk. On this view, we should take our discourse about propositions—
i.e., our use of proposition sentences—as involving something other than outright
assertions of the propositions corresponding to the surface form of the sentences
uttered. Instead, in uttering those sentences we’re engaging in a kind of pretense or
pseudo-assertion. What exactly this comes to depends on the fictionalist view in
question. On some views, it will be a matter of asserting some proposition other than
the one corresponding to the surface form of the sentence uttered—like the proposi-
tion that that sentence is true according to a certain fiction. On other views, it will
be a matter of speakers standing in some attitude other than assertion to the relevant
proposition.11
Like the paratactic analysis, fictionalist approaches to various discourses have gen-
erated a large literature to which I can’t aim to do justice here. It should be admitted
that, in some ways, our proposition talk seems ripe for a fictionalist treatment—one
might, to borrow an example from Yablo (2000), feel impatient with someone worry-
ing about whether we have beliefs on the basis of worries about whether propositions
exist in much the same way as one might feel impatient with someone worrying about
whether “creatures of metaphorical make-believe” like “the green-eyed monster”
really exist. Surely, one wants to say, such skepticism misses the point of our talk about
belief! On the other hand, discourse about what we and others believe and desire, and
what is true, can seem like the very paradigm of genuine assertion, and hence a quite
implausible target for fictionalist treatment.12
The question of whether fictionalism about our use of the proposition sentences and
the like can give an adequate treatment of their (seeming) truth conditions and logical
properties is one which can only be answered by consideration of detailed fictional-
ist proposals.13 But, independent of such consideration, one might wonder whether
whether fictionalism about proposition talk faces a special sort of self-referential
11
On either construal, this would be a “hermeneutic” rather than a “revolutionary” fictionalism, since it
purports to describe proposition talk rather than to reform it. The distinction is due to Burgess (1983) and is
applied to the case of fictionalism in Stanley (2001). Fictionalism about propositions is defended in Balaguer
(1998). Crimmins (1998) and Kroon (2004) defend related but weaker theses.
Of course, one might worry in the present context that this description of fictionalism immediately entails
the existence of propositions. More on this below.
12
This sort of objection to various fictionalist views is voiced in Stanley (2001), who objects that at least
some fictionalist treatments of areas of discourse which we take to be perfectly literal involve an implausi-
ble attribution of “a novel and quite drastic form of failure of first-person authority over one’s own mental
states” (47).
13
For two excellent examples, see Richard (2000) and Stanley (2001).
problem which is not faced by fictionalism about other types of discourse. To see this,
consider Mark Richard’s well-known explanation of the notion of “piggy-backing,”
which is part of the explanation of how one sort of fictional use of a sentence might
work. Piggy-backing is, he says,
making an utterance u within a pretense in which u has a real world truth condition c, thereby
actually asserting a proposition which is (in fact) true iff c obtains14
The salient point, for our purposes, is that this account of what piggy-backing is—and
hence, in part, of what the relevant sort of fictionalist theory says—involves claims
about the assertion of propositions and hence can’t be endorsed by anyone who denies
that there are such things. And this is not just true of Richard’s way of setting the issue
up. Fictionalism is often introduced via a distinction between two different attitudes
toward a class of propositions—one which involves commitment to the propositions’
truth, and one which does not. But no such way of explicating the doctrine can work
if the goal is fictionalism about propositions—we can’t use talk of propositions in our
(presumably non-fictional) explication of the theory while explaining them away with
the theory so explicated.15
Similar remarks apply to Kendall Walton’s fictionalist account of discourse about
fictional characters, as Walton is well aware. As he says, “I have shamelessly helped
myself to properties and propositions in the preceding chapters, and will use them
now in explaining away fictional entities.”16 And the same goes for any fictionalist
account which either explicitly makes reference to contents or uses sentences like the
proposition sentences, since there’s at least a serious suspicion of circularity about the
method of stating a fictionalist theory of the proposition sentences by (presumably,
non-fictionally) using those very sentences.
It’s important not to overstate the importance of this point—not every fictionalist
account makes use of piggy-backing, defined as Richard defines it, or is modeled on
Walton’s fictionalism about fictional characters. But it’s at least not obvious that fic-
tionalism about proposition talk can be explicated in a way which does not smuggle
propositions in via the back door.
I will conclude this section with two general thoughts about the view that the propo-
sition sentences are false, whether or not this is accompanied by a fictionalist story
about our ordinary use of those sentences.
First, the best motivation for this view seems to be, in effect, to run the present argu-
ment in reverse and say that, precisely because the proposition sentences do entail
(2F), and because admitting propositions to our ontology is an unacceptably high cost
to pay, we are forced to deny the truth of the proposition sentences. Answering this
14
Richard (2000), 214. Richard introduces this notion in the process of explicating and criticizing the
fictionalism about modes of presentation defended in Crimmins (1998).
15
Related points are made in Richard (2000), §5.
16
Walton (1990), 390.
sort of argument involves showing that admitting propositions to our ontology is not
an unacceptably high cost to pay, and this is best done by developing a positive theory
of propositions—hence, to that extent, evaluation of this case for denying the proposi-
tion sentences must wait until the sketches of our positive theories of propositions in
Part II.
Second, there is a sense in which this strategy might be unstable. If the theorems of
our best semantic theory are themselves relevantly like proposition sentences, then
the semantic theorist, who is in the business of constructing a semantic theory, can
hardly avoid commitment to propositions by denying all of the proposition sentences.
(Unless, of course, she regards her own semantic theorizing as itself a kind of pre-
tense.) As we’ll see in the next section, a good case can be made that the most plausible
semantic theories which do not make explicit reference to propositions do, in fact,
have proposition sentences among their theorems.
The claim that sentences of this sort could be the theorems of semantic theories is,
on the face of it, puzzling. Why should a theory which issues T-sentences, but makes
17
Here, as above, I’m ignoring for simplicity the need to relativize the propositions expressed by sentences
to contexts. Put in these terms, what a semantic theory should provide is a pairing between sentences and
characters—where the latter are functions from contexts to propositions.
18
See Davidson (1967). One might think that another main alternative is the view that a semantic theory
should provide not a pairing between sentences and propositions, but rather a pairing between sentences
and a set of indices in which those sentences are true. I’m setting this aside for two reasons. First, I take this
to be not an alternative to a propositional semantics, but rather a version of it according to which proposi-
tions are sets of worlds, or situations, or whatever. Second, this view of semantic theory is (in part) the topic
of Ch. 3.
is true, but tells us hardly anything about the meaning of “Snow is white.” Rather, we
want a semantic theory to entail, for each sentence of the object language, exactly one
interpretive T-sentence: a T-sentence such that the sentence used on its right-hand side
gives the meaning of the sentence mentioned on its left-hand side. Our theory must
entail at least one such T-sentence for each sentence in the object language because
the aim is to give the meaning of each sentence in the language; and it must entail no
more than one because, if the theory had as theorems more than one T-sentence for
a single sentence S of the object language, an agent who knew all the theorems of the
theory would not yet understand S, since such an agent would not know which of the
T-sentences which mention S was interpretive. The extension problem is the problem
of designing a theory which can meet both of these requirements.
One reason why the extension problem is difficult is that it seems that any theory
which implies at least one T-sentence for every sentence of the language will also imply
more than one T-sentence for every sentence in the language. For any sentences p,q, if
the theory entails a T-sentence
S is T in L iff p,
then, since p is logically equivalent to p & ฏ(q & ฏq), the theory will also entail the
T-sentence
which, if the first is interpretive, won’t be. But then the theory will entail at least one
non-interpretive T-sentence, and someone who knows the theory will not know which
of the relevant sentences is interpretive and which not; such a person therefore would
not understand the language.
The information problem is that, even if our semantic theory entails all and only
interpretive T-sentences, it is not the case that knowledge of what is said by these theo-
rems would suffice for understanding the object language. For, it seems, I can know
what is said by a series of interpretive T-sentences without knowing that they are inter-
pretive. I may, for example, know what is said by the interpretive T-sentence
“Londres est jolie” is T in French iff London is pretty
but still not know the meaning of the sentence mentioned on the left-hand side of the
T-sentence. The truth of what is said by this sentence, after all, is compatible with the
sentence used on the right-hand side being materially equivalent to, but different in
meaning from, the sentence mentioned on the left.
In response to these objections, Davidsonians typically modify the bare sketch of
Davidsonian semantic theory given above in two related ways. First (in response to
the extension problem) they specify a set of canonical rules of inference which don’t
permit the derivation of non-interpretive T-sentences. Second (in response to the
information problem) they add to the theory a rule of inference which takes us from
canonically derived T-sentences to a different sort of sentence, like:
A theory meeting such-and-such formal and empirical constraints entails that S is T iff p
or a meaning theorem, like:
S means that p.19
This extra rule of inference helps with the information problem, because it seems that
someone who knows what is said by, for example, a meaning theorem will know the
meaning of S.
Let’s take a step back here. The initial idea was that Davidson’s approach to seman-
tics might provide a viable avenue for the semantic theorist who wants to avoid
commitment to the existence of propositions; the reason for thinking this was that
the theorems of a Davidsonian semantic theory—T-sentences, rather than pairings
of sentences with propositions—don’t involve any commitment to the existence of
propositions. But now we’ve given up the idea that T-sentences are the theorems of a
Davidsonian semantic theory and replaced these with sentences which are relevantly
similar to the “proposition sentences” discussed in the previous section.
A plausible case can be made, as noted earlier, that proposition sentences and others
of the same form entail the existence of propositions. And I noted earlier, also that, for
19
These responses are defended in, respectively, Davidson (1976) and Kolbel (2001).
this reason, one might avoid commitment to propositions simply by saying that no sen-
tence of this form is true. But this is plainly not something which our neo-Davidsonian
semantic theorist can say, since she would then be denying the theorems of her own
theory. The proposed modification of Davidsonian semantics thus ends up closing off
one route for the semantic theorist who wants to avoid commitment to propositions.20
Setting this aside, there are independent problems with the idea that a semantic
theory could include a rule of inference like one which takes us from T-sentences
to meaning theorems—at least when this is conjoined with the idea that a theory is
a satisfactory semantics for a language iff knowledge of the theory would suffice to
understand sentences of the language. The problem is that the sort of explanation of
semantic competence that we get from a neo-Davidsonian semantic theory can also be
given by a certain sort of translation manual; and giving a translation manual, which
maps sentences from one language onto their translations in the other language, is
not an adequate semantic theory for either. Hence the sort of explanation of semantic
competence provided by a neo-Davidsonian theory of the sort sketched above is not
sufficient to justify those theories.21
The point that providing a translation manual between two languages does not suf-
fice for providing a semantic theory of either has been defended elsewhere, and I won’t
go through the arguments again here.22 Given this, it is worth noting that a transla-
tion manual of this sort might provide an explanation of a kind of a speaker’s com-
petence with a language. After all, given my knowledge of English, if I am given, and
sufficiently internalize, a translation manual which maps sentences of Urdu onto their
English translations (and vice versa), this will give me an understanding of the rel-
evant Urdu sentences. So a translation manual can explain semantic competence with
a language—so long as we take for granted the speaker’s understanding of another lan-
guage. Let’s call this a derivative explanation of semantic competence, since it explains
competence with one language in terms of competence with another. Given that trans-
lation manuals are not semantic theories, if Davidsonian semantic theories are to be
defended on the grounds that knowledge of them would suffice for competence, the
sort of explanation they offer had better not be a derivative one.
But this is just the sort of explanation of semantic competence which can be pro-
vided by a Davidsonian semantics supplemented by the sort of extra rule of inference
mentioned above. To see this, consider how, precisely, the crucial extra rule of infer-
ence from canonically derived T-sentences to the theorems of the theory might be for-
mulated. Where L is the target language, we could try to formulate it as:
If S is T in L iff pis a theorem of the theory, then S means that pis a theorem of the theory.
20
This still leaves the Davidsonian the option, discussed in the previous section, of endorsing the truth of
proposition sentences while denying that these sentences entail claims like (2F).
21
This argument is inspired by Harman (1975). I develop this line of argument in more detail in Speaks
(2006).
22
See, among other places, Lewis (1970), LePore and Loewer (1981), and Speaks (2006).
But this will plainly not provide knowledge of the meaning of S. It will, perhaps, pro-
vide knowledge that a certain sentence about the meaning of S is a theorem of a true
theory, and hence true; but knowledge that a sentence about the meaning of S is true
doesn’t tell one what the meaning of S is unless one understands the sentence about
S. In this case, that sentence—a sentence of the form S means that p—is stated in
the language of the theory. Hence explaining competence via a theory which depends
crucially on this rule of inference relies upon the speaker’s prior competence with the
language of the theory—and this is the same sort of derivative explanation of semantic
competence as is provided by a translation manual.
We could try to avoid the problem by formulating the rule of inference in terms of
what is said by the theorems of the theory, without mentioning expressions in the lan-
guage of the theory, as follows:
If it follows from the theory that S is T iff p, then S means that p.
But there is a problem with understanding this formula. “S” can be understood as a
universally quantified variable over sentences, but how is “p” to be understood?
One idea is that “p” is a sentence letter, and that the above is a schema; on this formu-
lation, to master this rule of inference is to know that every instance of the schema is
true. But an instance of a schema is a sentence, and if what this rule of inference gets us is
that a certain sentence is true, then we are back in the problem we were trying to avoid.
Knowledge that a certain meaning theorem is true yields knowledge of the meaning of
the sentence mentioned in the theorem only if one understands the theorem; but if we
are making use of competence with the language in which the meaning theorems are
stated, we can give no non-derivative explanation of semantic competence.
One might object that this argument attacks a straw man. Any theory has to be stated
in some language or other; hence any explanation of semantic competence based on
knowledge of a semantic theory will presuppose understanding of the language of
the theory. To argue that knowledge of a neo-Davidsonian theory for L cannot give
a non-derivative explanation of semantic competence with L is thus only to point out
that this theory can’t do the impossible.
If this were true, this would be a problem for Davidsonian semantics rather than for
the present argument; for if no semantic theory can give a non-derivative explanation
of semantic competence, then Davidson’s principal argument in favor of his approach
to semantics—namely, that it can give such an explanation—fails. But, more impor-
tantly, the objection is based on a mistake. Consider a semantic theory whose theo-
rems pair sentences with the propositions which are their semantic contents. There’s
no reason why one can’t know the theorems of such a theory without knowing the
language of the theory, since in general there’s no reason why one can’t know a proposi-
tion without knowing the truth of some sentence which expresses that proposition.23
The troubles into which neo-Davidsonian explanations of semantic competence fall
23
Though such explanations of competence might face other problems. See Soames (1992).
are not the result of the fact that they are stated in a language. They are the result of
the particular rules of inference on which those theories rely to solve the information
problem.
And in fact we can show that there is something fishy about the inclusion of these
extra rules of inference without even bringing in issues about the explanation of
semantic competence. Suppose for reductio that a Davidsonian semantic theory, sup-
plemented with one of the sorts of rules of inference above, can serve as a satisfactory
semantic theory for L. Then any theory which provides as much information about the
meanings of expressions of L (and no extra, false information) must also be satisfac-
tory semantics for L. But consider a translation manual of the sort mentioned above,
which pairs synonymous sentences of two languages. All should agree that this sort of
translation manual is a semantic theory of neither of the two languages; it says which
expressions mean the same as which other expressions, but doesn’t say what any of
those expressions do mean. But, as Gilbert Harman suggested, we can imagine adding
to this translation manual a rule of inference parallel to the rules of inference which are
supposed to take us from T-sentences to meaning theorems, e.g.
If Sin L means the same as S*in L*is a theorem of the theory, then Sin L means that
S*is a theorem of the theory.24
But if we add this sort of rule of inference to our translation manual, we are able to use
it to derive as theorems all of the theorems of our neo-Davidsonian theory. Hence our
modified translation manual must be a satisfactory semantic theory for L. But it isn’t.
So our original supposition, that a Davidsonian semantic theory supplemented with
these sorts of extra rules of inference, might be a satisfactory semantic theory for a
natural language, must be rejected.25
In “Truth and Meaning,” Davidson remarked that, “paradoxically, one thing that
meanings do not seem to do is oil the wheels of a theory of meaning—at least as long
as we require of such a theory that it non-trivially give the meaning of every sentence
in the language. My objection to meanings in the theory of meaning is not that they
are abstract or that their identity conditions are obscure, but that they have no dem-
onstrated use.”26 The foregoing provides some reason for thinking that Davidson was
wrong about this—indeed radically wrong, if the preceding arguments show that
meanings as entities are not just useful, but necessary, for the construction of a seman-
tic theory. But, as Davidson’s quote suggests, there is another powerful motivation for
non-propositional semantics: this is the thought that, however useful propositions
may be in semantic theory, there can be no adequate metaphysical account of what
sorts of things these entities are. An adequate answer to this sort of skepticism requires
an answer to the question: What are propositions? This is the question to which much
of the following is addressed.
24
See Harman (1974). Here I am supposing that L* is the language of the theory.
25
For a less compressed exposition of this line of argument, see §II of Speaks (2006).
26
Davidson (1967), 20–21.
The two leading traditional conceptions of propositions are those found in classical
theories of structured propositions descending from Frege and Russell, and those
growing out of more recent theories of propositions as sets of possible worlds (or func-
tions from such to truth values). I will begin with the former.
statement, which lasted for over ten minutes, is entirely true, it is so because his ora-
tion resulted in the assertion of one or more propositions, each of which is true.
Propositions themselves then are taken to be timeless, unchanging, platonic entities
with which we are acquainted by a kind of passive intellectual awareness. As Russell
put it in (1904),
Suppose, for the sake of definiteness, that our judgment is ‘A exists,’ where A is something that
does as a matter of fact exist. Then A’s existence [by which he means the proposition that A exists],
it seems plain, subsists independently of its being judged to subsist . . . In this case the Objective
[i.e. the proposition] of the judgment—at least in the view of common sense—is as truly inde-
pendent of the judgment as is A itself. But the peculiarity of the cognitive relation, which is what
we wish to consider, lies in this: that one term of the relation is nothing but an awareness of the
other term.1
According to this conception, which Frege shared, the fact that propositions represent
things as being one way or another, and so are true (false) iff those things are (are not)
the way they are represented to be, is not derivative from, or attributable to, concep-
tually prior cognitive activities of agents who entertain them. On the contrary, since
propositions are the primary bearers of intentionality, the intentionality and truth con-
ditions of cognitive acts or states must be explained in terms of quasi-perceptual rela-
tions we bear to propositions. For Frege and Russell, all intentionality originates and is
grounded in an abstract “third realm.”
It is this difficult doctrine, more than any other, that generates the fatal difficulty
they encountered —known as “the problem of the unity of the proposition”. Since the
two philosophers struggled with the problem in different ways, I will deal with them
separately.
Russell on propositional unity
The problem of the unity of the proposition is to explain what propositions are in a way
that makes clear how they can have the intentional properties they do. Russell starts
with simple sentences like (1), one part of which—the predicate—is used to say, or
assert, something about the referent designated by its other part—the subject.
1. Socrates is human.
Corresponding to this, Russell thought, one part of the proposition expressed by (1)—
the concept/property humanity—is applied to, or asserted of, the other part of the
proposition—the man Socrates. This was his basic model for propositions.
In every proposition . . . we may make an analysis into something asserted and something about
which the assertion is made.2
1
Russell (1904), “Meinong’s Theory of Complexes and Assumptions,” Mind, 13: three installments: 204–
219, 336–354, 509–524; reprinted in Russell, Essays in Analysis, New York: George Braziller), 1973, cited
at p. 60.
2
Russell (1903), p. 43.
Although there are problems with this model, there is also something revealing about it.
When we assert that Socrates is human we may be said to assert the property being human
(a.k.a. humanity) of him. In so doing we represent him as human. Hence, our assertion has
truth conditions; it is true iff Socrates is the way he is represented to be. This is clear enough.
However, Russell’s task of explaining the intentionality of our speech act in terms of the
conceptually prior intentionality of the proposition expressed was more daunting. Roughly
put, he needed to translate commonsense talk about what we do—assert of things that they
are so-and-so, and thus represent them as being a certain way—into talk about what propo-
sitions “do,” in a way that reveals them to be the fundamental bearers of intentionality.
However, the idea needs fine tuning. As the examples in (2) illustrate, the notion Russell
needs is not assertion, but predication.
2a. Socrates is human.
b. If Socrates is human, then Socrates is mortal.
c. That Socrates is human is widely believed.
d. I wonder whether Socrates is human.
e. Is it true that Socrates is human?
f. Is Socrates human?
Whereas one who assertively utters (2a) asserts that Socrates is human, one who asser-
tively utters (2b), (2c), or (2d) does not assert this, while one who uses (2e) or (2f) to ask a
question doesn’t assert anything. Nevertheless, there is something common to the cases.
In each case, the speaker uses the name “Socrates” to refer to Socrates, and the predicate
“is human” to represent him as being a certain way. In (2a) Socrates is asserted to be that
way; in (2b) it is asserted that if he is that way, then he is mortal; in (2c) it is asserted that
he is widely believed to be that way; in (2d) I indicate that I wonder whether he is that
way, and in (2e) and (2f), I ask whether he is. In each case, part of what a speaker does in
performing his or her overall speech act (of asserting, questioning, etc.) is to represent
Socrates as being human by predicating the property humanity of him. Since, for Russell,
the intentional properties of the resulting speech acts are derived from the propositions
they express—which include the proposition that Socrates is human—the relation he
needs to “unify” that proposition is predication. In the proposition that Socrates is human,
humanity is predicated (not asserted) of Socrates.
The problem is that while predication, as I have used it here, is something that agents
do, what Russell needs is “a logical sense of predication.” It is, admittedly, difficult to see
what this might amount to. Perhaps because of this he was drawn to assertion, rather
than predication, as the crucial “unifying” relation. Since for him propositions are
entities the intentional properties of which are conceptually independent of us, it was
natural for him to look to “assertion in the logical sense”—which he took to be linked
in a mysterious way to truth—to do the job.3 The obvious problem, of course, is that no
3
At this stage, Russell was strongly tempted by the idea that just as the copula, “is,” in “Socrates is human”
signals that “human” is functioning as a predicate of the referent of “Socrates,” so the proposition expressed
relation that fails to apply to false propositions can possibly “unify” them. In addition,
since, for Russell at this stage, there was no intentionality without propositions, and
hence no conception of truth prior to an account of its bearers, the unifying notion he
needs is explanatorily prior to truth. For these reasons, we are better off shifting the
burden of “unifying” propositions to predication, while waiting until later to decide
whether propositions so conceived can have whatever ontological independence from
agents turns out to be needed.
This brings us to Russell’s most famous remark on propositional unity.
Consider, for example, the proposition “A differs from B.” The constituents of this proposition,
if we analyze it, appear to be only A, difference, B. Yet these constituents, thus placed side by side,
do not reconstitute the proposition. The difference which occurs in the proposition actually relates
A and B, whereas the difference after analysis is a notion which has no connection with A and B. It
may be said that we ought, in the analysis, to mention the relations which difference has to A and
B, relations which are expressed by is and from when we say A is different from B. These relations
consist in the fact that A is referent and B relatum with respect to difference. But A, referent,
difference, relatum, B, is still merely a list of terms, not a proposition. A proposition, in fact, is
essentially a unity, and when analysis has destroyed the unity, no enumeration of constituents will
restore the proposition. The verb, when used as a verb, embodies the unity of the proposition, and is
thus distinguishable from the verb considered as a term, though I do not know how to give a clear
account of the precise nature of the distinction.4 (my emphasis)
A central point here is that there is more to the proposition that A differs from B than
the fact that its constituents are A, B, and difference. In addition, there is both the man-
ner in which these constituents occur, and how their occurring as they do represents
A and B as being different.Modifying Russell, we may put this by saying that in the
proposition, difference is predicated of A and B, with the result that they are represented
as being different. In a mere list, nothing is predicated of anything, so the list doesn’t
represent the items as being one way rather than another. Consequently propositions
are true or false, while lists are neither.
Although we are presently taking predication to be primitive, one can still reason-
ably ask for an account of what, in a proposition, indicates which constituent is predi-
cated of which things. Since adding predication as an extra propositional constituent
would do nothing to confer intentionality on what is otherwise a mere list, there seems
by the sentence contains something functioning as “a verb” that somehow unites Socrates and humanity
into a coherent propositional whole. This “verb” is thought to be a special sort of assertion/predication rela-
tion by which the concept humanity is brought to bear on the man Socrates, rendering the whole repre-
sentational. See section 53 of Russell, The Principles of Mathematics, 1903. Of course, it wouldn’t do for this
“verb” simply to occur as another constituent of the proposition, since that would generate a further unity
problem. Rather it must “really apply to,” and hence really relate, its arguments. Russell’s struggle in The
Principles of Mathematics with “the logical notion of assertion” as the element needed to explain the unity
of the proposition is discussed in section 3.5 of chapter 7 of Volume 1 of Soames, The Analytic Tradition,
Princeton: Princeton University Press.
4
Russell (1903), pp. 49–50.
to be only one answer to this question that is roughly Russellian in spirit.5 Just as the
structural relations holding among syntactic constituents of a sentence show how they
are to be understood, so the structural relations holding among the constituents of the
proposition must show what it predicates of what.6
So far, so good. What structural features of a proposition do show what is predicated
of what? Consider the proposition expressed by (3), the constituents of which are the
relation identity together with the relation difference occurring twice over.
Any of these could be used as a formal model of the Russellian proposition expressed
by (3), as could any number of tree structures, two of which are pictured in (5).
5
It is worth contrasting this example—A differs from B—with Russell’s earlier example—Socrates is
human. In the earlier case, Russell was inclined to import something he called “a verb” corresponding to the
copula into the proposition. The function of this new element was to really relate, via assertion or predica-
tion, the man Socrates to the concept humanity. However, in the present case we have no copula, but rather
what is grammatically already a verb—“differs.” This compounds Russell’s problem, since here the job of
relating the constituents of the proposition falls to the relation difference itself. And how can difference do
this job unless A really does differ from B? Russell has no answer. My suggested revision on Russell’s behalf
assigns the job of relating propositional constituents to one another (in a way that forms a representational
whole) not to something that is itself a propositional constituent, but rather to the structural relationships
the constituents bear to one another in the proposition.
6
Frege’s answer—that some constituents of propositions are essentially predicative (and so cannot them-
selves be subjects of predication)—was regarded by Russell with deep suspicion, believing it to lead to “inex-
tricable difficulties.” (Section 49 of The Principles of Mathematics.) His idea, developed in that section and
following, is essentially that such a doctrine is self undermining. Since to state it we have to be able to refer to
purely predicative propositional constituents and predicate properties of them, stating the doctrine requires
recognizing propositions in which they occur not predicatively, but as predication targets. That said, it must
be admitted that Russell himself found it difficult to avoid such undermining himself. See, for example, sec-
tion 52 of The Principles of Mathematics.
5a. Prop
difference
identity difference
5b. Prop
Which of the structures of the sort illustrated by (4) and (5) is the proposition
expressed by sentence (3)? Expressed in this direct and uncompromising way, the
question is absurd. The problem is not that any of these could serve, and hence that
there is no determinate answer. The problem is that it is hard to see how any formal
structure of this, or any similar, sort could possibly be the proposition we are looking
for. Proposition (3) is something that represents the relation of identity as being different
from the relation of difference by virtue of the fact that difference is predicated of the two
relations. But there is nothing in the sets or sequences of (4), in the tree structures of
(5), or in any other formal structure we might construct to organize the constituents
of the Russellian proposition which, by its very nature, indicates that anything is predi-
cated of anything. Hence, there is nothing inherent in such structures that makes them
representational, and so capable of being true or false. Structures of this sort can’t pos-
sibly be the primary bearers of intentionality.
We could, if we wished, adopt rules that would allow us to read off the needed
information about predication from such structures, and so interpret them. To do
this would be to endow the structures with representational meaning or content,
thereby making them bearers of truth and falsity. This would not, however, make them
Russellian propositions. For Russell, propositions are not things that have meanings, or
get interpretations from us; they are the meanings that sentences come to express when
we initially endow them with meaning, or that we discover when we come to under-
stand sentences previously so endowed. The real problem with Russell’s conception of
propositions is that it makes it difficult or impossible to answer a question that can’t
be avoided: “What makes propositions representational, and, hence, bearers of truth,
objects of the attitudes, and meanings of sentences?”
Frege on propositional unity
Frege had his own problem with “the unity of the proposition.” He too had to “unify”
the constituents of a proposition without either adding a further constituent relating
the others, or relying on a mysterious, unspecified structural configuration to do the
7
For detailed discussion, see chapter 2 of Soames (2010), also chapter 2 of Soames, Volume 1 of The
Analytic Tradition.
of propositions must play predicative roles. Of course, entities of certain kinds can
never be predicated of anything; but entities of other kinds can. Since those that can
may themselves be subjects of further predications, Frege’s idea of inherently predica-
tional constituents can’t solve the problem of propositional unity. This brings us back
to the Russellian idea of a structure of constituents, at least one of which is predicated
of the others, even though that which is predicated can itself be the target of other
predications.
b c
One might, of course, wonder what these relations amount to, and try to determine
whether they have informative analyses. But since the same questions arise for all com-
plex entities, there is no special “uniting” problem of this sort for propositions. What is
special is that propositions must be—inherently and without further interpretation by
us—capable of being true or false. Since it would seem absurd to characterize any set,
sequence, or abstract tree as inherently representing things as being certain ways—and
so as being true or false—the idea that propositions are any of these structures is a
non-starter. Of course, propositions can’t be sentences either, since it is only by virtue
of expressing propositions that sentences are supposed to be bearers of truth condi-
tions themselves. So the problem remains.
In my opinion, the key to solving it is to recognize the obvious fact that predica-
tion is something that agents do. Properties don’t predicate themselves of anything;
nor, unless we have it explained to us, do we understand what it is for a complex of
which various properties are constituents to predicate one of them of the others. This
is what Frege and Russell were up against. They needed predication to make sense
of propositions, but their conception of propositions made it impossible for them to
find appropriate agents for the needed predications. The solution to their problem
is to retain their idea of propositions as structurally complex entities that are inher-
ently intentional, and hence the bearers of truth conditions, while giving up their idea
that propositions are the primary bearers of intentionality. Instead of explaining the
intentionality of the cognitive activity of agents in terms of an imagined conceptu-
ally prior intentionality of the propositions they entertain, we must explain the inten-
tionality of propositions in terms of the genuine conceptually prior intentionality of
the cognitive activity of agents who entertain them. This can’t be done on the tradi-
tional Frege-Russell model of robustly platonic propositions passively apprehended
by agents. It also can’t be done by tying propositions too closely to agents. In my view,
we need a conception of propositions that (i) recognizes unentertained propositions,
including the truth or falsity of propositions at world states at which no propositions
are entertained, while (ii) explaining the intentionality of propositions in terms of the
intentionality of the cognitive acts of possible agents who entertain them. This will be
my approach in chapter 6.
members. What does the set containing worlds 1, 2, and 3 represent? Is it true or false?
These questions are bizarre. If we wanted, we could use the set to represent the actual
world as being in the set—and so to make the claim that no world outside the set is
actual. But we could equally well use it to represent the actual world as not being in the
set—and so to make the claim that no world inside the set is actual. Independent of
interpretation by us, the set doesn’t represent anything, doesn’t make any claim, and so
doesn’t have truth conditions.
What about the function f that assigns worlds 1-3 truth and all others falsity? On the
face of it, f would seem not to be a proposition either, for surely if sets aren’t inherently
representational then neither are their characteristic functions. Suppose we replaced
the values truth and falsity with the North and South Poles. What does the function
that assigns worlds 1-3 the North Pole and all other worlds the South Pole represent?
Independent of interpretation by us, it doesn’t represent anything. Why, then, should
the original function assigning truth and falsity be representational? What, after all,
are truth and falsity but properties we grasp primarily through their application to
propositions? But surely, if propositions are needed to illuminate truth and falsity, they
can’t be among the building blocks for constructing propositions.
The illusion that propositions can reasonably be construed as functions from worlds
to truth values is fed by a natural way of thinking about worlds—as maximal states
(properties) the universe might conceivably be in (or have).8 Under this conception,
each assignment of a truth value to a world-state w can be correlated with the proposi-
tion that predicates w of the universe, and is thus true or false depending on whether
or not w is the maximal state the universe really is in. A function from world-states
to truth values can then be associated with the (possibly infinite) disjunction of the
propositions correlated with its assignments of truth to world-states.
This picture is fine, if one already has independent accounts of propositions and truth.
But it doesn’t provide a foundational account of what propositions are. Instead, it pre-
supposes propositions by appealing to them as things that predicate properties, in
this case a world-state, of other things, in this case the universe—while further pre-
supposing that propositions are true when their predication targets have the proper-
ties predicated of them. When (wrongly) taken as a foundational story, this account
also (i) inverts the conceptual order relating propositions and truth by taking the latter
to be an unexplained primitive involved in the construction of the former, while (ii)
merely correlating functions from world-states to truth values with propositions, with-
out providing a basis for identifying the two. In addition, the story is inconsistent with
8
See Stalnaker, Inquiry, Cambridge: MIT Press, 1984, and Ways a World Might be, Oxford: Oxford
University Press, 2003. See also, pp. 200–209 of Soames Reference and Description, Princeton and Oxford,
2005; Soames, “Actually,” in Mark Kalderon, ed., Proceedings of the Aristotelian Society, supplementary vol-
ume, 81, 2007, 251–277, reprinted in Soames, Philosophical Essays, Vol. 1: Natural Language: What it Means
and How we Use it, Princeton and Oxford: Princeton University Press, 2009; also chapters 5 and 6 of Soames,
Philosophy of Language, Oxford and Princeton: Princeton University Press, 2010.
the parallel attempt to identify properties with functions from world-states to sets of
individuals. Since it is part of the story that world-states are properties attributed to the
universe, these properties can’t be functions from world-states to sets. But if properties
aren’t reducible to functions from world-states to extensions, there is no reason that
propositions should be reducible to functions from world-states to their extensions.
So far I have made two negative points: (i) the possible-worlds conception of propo-
sitions fails to explain how propositions can be representational, and so have truth
conditions, and (ii) it wrongly takes what it calls “worlds” and “truth values” as unex-
plained primitives from which it tries to construct properties and propositions, when
in fact properties and propositions are needed to explain and illuminate both truth
and worlds as world-states. I now turn to related points about the conceptual founda-
tions of possible-worlds semantics.
9
In this paragraph, as in the previous paragraph, “S” is a metalinguistic variable. “A” and “B” are used as
schematic letters.
that so-and-so is true iff so-and-so. This allows us to derive information about meaning
from statements about truth conditions. For example, we derive ⌜“S” doesn’t mean that
R⌝ from ⌜“S” is true iff Q⌝, when Q and R are known to differ in truth value (e.g. when
they are inconsistent).10
Modalized truth conditions provide further information about meaning. 11
Assuming the usual connection between modal operators and world-states, we derive
(6a,b).
6a. ∀w [at w (the proposition that S is true iff S)]
b. ∀w [at w, the proposition that S is true iff at w, S)]
The truth predicate is monadic, “at w” is a sentential operator with the force if w were
instantiated it would be the case that, and (6c) and (6d) come to the same thing.
6c. ∀w [at w, the proposition that Plato philosophized is true iff at w, Plato
philosophized]
d. ∀w [the proposition that Plato philosophized is true at w iff Plato philosophized
at w]
This is the starting point for understanding statements like (7) about the truth condi-
tions of sentences made by possible worlds semantics.
Claim (7) carries information about the meaning of the sentence it mentions in virtue
of our antecedent understanding of what it is to be true, and what it is to philosophize,
at a world-state. To say that x philosophizes at w is to say that if w were instantiated,
then x would philosophize. What is it for S to be true at w? The possible-worlds seman-
ticist can’t quite say that for the English sentence S to be true at w is for it to be such that
if w were instantiated, then S would be a true sentence of English (i.e. one which expresses
a true proposition). In possible-worlds semantics, S can be true at w even if S means
nothing at w, or means something different from what it actually means. This shows
that the dyadic truth predicate of possible worlds semantics is a technical substitute
for our ordinary notion. Using our ordinary notion, we say that S is true at w iff at w,
S expresses a proposition that is true. Since what S could have meant (or expressed)
is no help in illuminating what S actually means (expresses), the possible-worlds
semanticist doesn’t follow us in this. Instead, he dispenses with the ordinary notion
10
Here, “Q” and “R” are metalinguistic variables along with “S.”
11
“S” is used as a schematic letter in NT and (6), while being used as a metalinguistic variable later in the
paragraph.
of sentential truth, and introduces the technical predicate of sentences “is true at w” to
mean the proposition p that S actually expresses is true at w—otherwise put: S expresses
a proposition at @ and that proposition is true at w.
In short, the dyadic truth predicate of possible-worlds semantics is parasitic on the
prior notions: the proposition actually expressed by a sentence and the monadic property
truth of propositions. It is by taking these for granted that we extract useful informa-
tion about meaning from the truth conditions provided by such a semantic theory.
We derive ⌜”S” doesn’t mean that R⌝ from ⌜∀w “S” is true at w iff at w, Q⌝ when Q and
R aren’t necessarily equivalent—tacitly assuming ⌜if “S” means that P, then necessar-
ily the proposition “S” actually expresses is true iff P⌝.12 Although this doesn’t identify
what S does mean, it does so up to necessary equivalence, thereby providing informa-
tion about meaning that restricts the range of acceptable alternatives. Without the prior
notions of truth and propositions here employed, even this limited information about
meaning extracted from the semantic theory would be lost.
This is the heart of the problem with the analysis of propositions as functions from
world-states to truth values. If one takes world-states and truth values to be unex-
plained primitives, with the goal of using them to provide reductive analyses of proper-
ties, propositions, and meaning, then one can’t use the method just given for extracting
claims about meaning from possible-worlds semantic theories. Without prior accounts
of propositions, truth, and the connection between meaning and truth, the theorems of
such a semantic theory won’t carry any information about meaning. Having denied
themselves such accounts, proponents of this approach, like Bob Stalnaker in his book
Inquiry, are reduced to telling us that any two objects—0 and 1, or the North and South
Poles—can serve as values assigned by the intensions of sentences to unexplained indi-
ces called “worlds.”13 But, surely, we don’t learn anything about the meaning or genuine
truth conditions of a sentence by associating it with a function that assigns some unex-
plained indices the North Pole and others the South Pole.
As I explain more fully in chapter 6, the proper explanation starts with the idea that
agents predicate properties of objects in cognition and perception, thereby entertain-
ing propositions. Agents do this before they have the concept proposition. Focusing
on similarities and differences in our experience, agents like us are able to acquire the
concept, after which we are able to make propositions objects of thought and subjects
of predication. This, in turn, allows us to acquire the concept of truth. Given truth,
properties can be conceptualized as things true of other things. With the concepts
truth, property, proposition and modality (what could be but isn’t) in place, we can
characterize world-states as ways for things to be—maximally informative properties
that the world could have had. Such a world-state w can be defined as the property of
12
Again, “P,” “Q,” “R,” and “S” are metalinguistic variables.
13
On p.2 we are told, “There are just two truth values—true and false. What are they: mysterious Fregean
objects, properties, relations of correspondence and non-correspondence? The answer is that it does not
matter what they are; there is nothing essential to them except that there are exactly two of them.”
making true a set w* of basic propositions that tell a complete world-story. Roughly
put, a proposition p is true at w iff p is an a priori consequence of w*. So, we can come
to know that p is true at w by deriving p from w*. As for the actual world-state @, we
can come to know p to be true at @, given knowledge of p, by noting that since p is true,
it must be true at this very world-state—the one that is instantiated.
This is a satisfying foundational picture of how the notions proposition, property,
truth, and world-state are related to one another, and to cognition, including how we
are able to know the things about them that we do. It is also the conceptual background
needed to extract useful information about meaning from possible-worlds semantic
theories. The problem for the philosophically ambitious proponent of such theories—
who wants to tell us what propositions and properties are, but not what his primitive
dyadic notion of truth amounts to—is that he can’t adopt this foundational picture,
since to do so he would have to invoke a rival conception of propositions as bearers
of monadic truth, antecedent to what his theory provides. His project of using what
he calls “truth” and “worlds” to define propositions prevents him from, at the same
time, taking real propositions and ordinary truth to be prior to those supposed primi-
tives. Since he typically characterizes properties as functions from worlds to sets of
individuals, he also can’t presuppose an antecedent conception of worlds as proper-
ties. With no way of explaining the crucial notions taken to be primitive, the reduc-
tive possible-worlds semanticist lacks the conceptual resources needed to explain how
his own elaborate technical machinery yields any information whatsoever about the
genuine semantic properties of sentences and other expressions.
14
Situations and Attitudes, Cambridge: MIT Press, 1983.
15
For criticism of influential metalinguistic accounts of belief ascriptions see chapter 7 of Soames, Beyond
Rigidity, New York: Oxford University Press, 2002.
16
For details, see Soames, “Direct Reference, Propositional Attitudes, and Semantic Content,” Philosophical
Topics 15, 1987, 44–87; and “Why Propositions Can’t be Sets of Truth-Supporting Circumstances,” Journal of
Philosophical Logic 37, 2008, 267–276, both reprinted in Philosophical Essays, Vol. 2, Princeton: Princeton
University Press, 2009.
17
Robert Stalnaker, “Assertion,” in Peter Cole, ed., Syntax and Semantics, vol. 9 Pragmatics,
New York: Academic Press, 1978; reprinted in Stalnaker, Context and Content: Essays on Intentionality in
Speech and Thought, New York: Oxford University Press, 1999, 78–95.
input (i) the meaning of a sentence S that is uttered, and (ii) the possible world-states
compatible with everything known or presupposed in the conversation at the time of
utterance. These, in turn, are supposed to generate a contingent proposition that is true
(false) at any world-state w in (ii) (and so compatible with all presuppositions in the
context of utterance) iff the proposition S would express, if w turned out to be the actual
world-state, is true (false) at w. This, it is alleged, is the contingent content asserted and
informatively communicated by the utterance of S.
The problem with this strategy is that there are many crucial cases in which it can-
not generate the required content. It will generate such a content only if there are
world-states w compatible with everything known or presupposed by the conversational
participants at which the proposition that would be expressed by S, if w were actual,
is false at w. With this in mind, consider a case in which I say to you ⌜He/she/it/that is
F⌝, rigidly designating an individual x and predicating a property of x that is, in fact,
one of x’s essential properties (without which x could not exist). Suppose further that
the situation in which I make this informative remark is one in which both of us are
intimately acquainted with x, and have many singular thoughts about x—including
the thought that x exists, is the subject of our conversation, and is the individual about
whom we are speaking. In this scenario, every world-state in (ii) is one in which S
would express a proposition attributing F-hood to x (and nothing else). Since x exists
at all these world-states and F-hood is essential to x’s existence, the propositions that
would be expressed, if any of them were actual, are true at those states. Thus, we are
left with the possible-worlds proposition that is true at all relevant world-states, which
means that diagonalization has failed.18
As indicated above, the source of the problem is the inability of the model to
accommodate attitudes to singular propositions that predicate essential properties
of objects—even though its chief proponent, Robert Stalnaker, recognizes both that
some properties are essential in the relevant sense, and that the model requires singu-
lar thought about world-states.19 Since these states are specified in terms of objects and
properties, this should mean that singular thought about world-states bottoms out in
singular thoughts about objects and properties. Being at the center of the model, such
singular thoughts can’t legitimately be excluded in cases, like those just discussed, in
which they lead to indigestible results. The problem can be shown to persist even when
epistemically possible world-states replace metaphysically possible world-states as the
truth-supporting circumstances.20 Thus, combining this failed strategy with the first
failed strategy for dealing with hyperintensionality doesn’t save the ability to model
propositions within the possible-worlds framework.
18
For discussion, see Soames, “Understanding Assertion,” in J. Thomson and A. Byrne, eds., Content and
Modality, Oxford: Clarendon Press, 2006; reprinted in Philosophical Essays, Vol. 2.
19
Stalnaker’s reply to my argument in “Understanding Assertion” is also in Content and Modality.
20
This result is also established in “Understanding Assertion.”
The high-water mark for the final strategy for dealing with hyperintensionality while
retaining propositions as sets of possible world-states is a semantic view I have previ-
ously called “strong two-dimensionalism”—suggested by certain influential writings
of Frank Jackson and David Chalmers.21 The view was prompted in part by the prob-
lem posed by the Kripkean necessary a posteriori for the view that propositions are
sets of metaphysically possible world-states. Since, on this view, there is only one nec-
essary proposition, which everyone knows a priori, no proposition can be both neces-
sary and knowable only a posteriori. The alleged “illusion” to the contrary is in failing
to recognize that context-sensitive sentences are associated not with one proposition
but with two. One of these propositions, called “the primary intension” of S, is the set
of possible world-states at which S expresses a truth. This proposition is true at w iff S’s
Kaplanian character maps w (considered as a context) onto a proposition that is true
at w. In this sense, the primary intension of S is more or less equivalent to the claim
that S’s Kaplanian character expresses a truth. The other proposition associated with S,
called “the secondary intension of S” at a given context C, is the proposition expressed
by S at C, which is the set of possible world-states w such that S is true relative to C,w.
Given that primary and secondary intensions come apart when S is context sensitive,
strong two-dimensionalists analyze names and natural kind terms as context-sensitive,
rigidified descriptions ⌜the actual D⌝. Since (8a) and (8b) express truths in precisely
the same contexts, their primary intensions are identified, even though their second-
ary intensions will, typically, be different.
8a. The D is F (if the D exists)
b. The actual D is F (if the actual D exists)
Since (8b) is the analysis of (9), when N is a name or natural kind term, the primary
intension of (9) is the primary intension of (8a) while its secondary intension is the
secondary intension of (8b).
9. N is F (if N exists)
Thus, when F expresses an essential property of the individual or kind designated by
N, and ⌜the D⌝ is nonrigid, the secondary intension of (9) is necessary, and knowable
a priori, while the primary intension of (9) is typically contingent, and knowable only
a posteriori. In this way, the thesis that propositions are sets of metaphysically possible
world-states while also being objects of the attitudes was defended against the chal-
lenge posed by the Kripkean necessary a posteriori.
This defense depends on taking the secondary intensions of sentences to provide
the arguments of modal operators and the primary intensions of those sentences to
21
Chalmers (1996), The Conscious Mind, Oxford University Press; Jackson, From Metaphysics to Ethics,
Oxford University Press, 1998. In Reference and Description, I reconstruct several precise and explicit ver-
sions of two-dimensionalism based on their work. The simplest, and, I believe, truest to the original motivat-
ing idea, is the one I there call “strong two-dimensionalism.”
provide the arguments of epistemic operators (“assert,” “believe,” “know,” etc.) The
incorrectness of such semantics can easily be shown by embedding a single sentence
under both modal and epistemic operators. When we do this, we find that in natural
language such operators require a single proposition to be supplied as argument to
operators of both types. For example, consider the (10a) and (10b).
10a. It is a necessary truth that if the actual husband of Stephanie Lewis was the actual
author of COUNTERFACTUALS and Mary believes that the actual husband of
Stephanie Lewis was the actual author of COUNTERFACTUALS, then Mary
believes something true.
b. It is a necessary truth that if the actual husband of Stephanie Lewis was the
actual author of COUNTERFACTUALS and Mary believes that the husband of
Stephanie Lewis was the author of COUNTERFACTUALS, then Mary believes
something true.
Although (a) is true, strong 2D semantics wrongly take it to express the same proposi-
tion as the false (10b). The same is true for the pair in (11).
11a. It is a necessary truth that if Mary believes that the actual husband of Stephanie
Lewis was the actual author of COUNTERFACTUALS, and if that belief is
true, then the actual husband of Stephanie Lewis was the actual author of
COUNTERFACTUALS.
b. It is a necessary truth that if Mary believes that the husband of Stephanie Lewis
was the author of COUNTERFACTUALS, and if that belief is true, then the actual
husband of Stephanie Lewis was the actual author of COUNTERFACTUALS.
The sentences in (12) bring names (or natural kind terms) into the picture.
12a. Although John truly believes that n is D, had the world been in state w, n would
not have been D and John would not have believed that n was D.
b. Although John truly believes that the actual D is D, had the world been in state w,
the actual D would not have been D and John would not have believed that the
actual D was D.
c. Although John truly believes that the D is D, had the world been in state w, the
actual D would not have been D and John would not have believed that the
D was D.
Here we let o be uniquely denoted by the nonrigid description ⌜the D⌝, we let n be a
name of o, and we let the strong 2D analysis of n be ⌜the actual D⌝. We stipulate that
⌜John truly believes that n is D⌝ is true (which assures the truth of what he believes).
We further take w to be a world-state at which some object other than o is uniquely
denoted by ⌜the D⌝, and in which John does not believe of o that it has the property
expressed by D, though he does believe the proposition expressed by ⌜the D is D⌝.
Given all this, (12a) should be true, even though strong 2D semantics wrongly assimi-
lates (12a) to (12b), which is in turn assimilated to the false (12c). This is the same problem
that was illustrated in (11), but put in a different form.
Finally, I extend the point to include the relationship between proper names and variables
of quantification. Here we let n be a name, and F be a predicate such that the truth of ⌜n is F⌝
guarantees the truth of ⌜There is such a thing as n⌝. I further assume that for any context C
and world-state w, if the following (a) sentences are true at C,w, then the (b) sentences are too.
13a. John truly believes that n is F, but had the world been in state w, n would not have
been F.
b. There is an x such that John truly believes that x is F, but had the world been in state w,
x would not have been F.
14a. John truly believes that n is F, but had the world been in state w, John would not have
believed that n was F.
b. There is an x such that John truly believes that x is F, but had the world been in state w,
John would not have believed that x was F.
Putting these facts together we see that if (15a) is true at C,w, then (15b) must also be
true there.
15a. John truly believes that n is F, but had the world been in state w, n would not have
been F and John would not have believed that n was F.
b. There is an x such that John truly believes that x is F, but had the world been in state w,
x would not have been F and John would not have believed that x was F.
Strong 2D semantics misses the entailment of (15b) by (15a), This results from a pair of
facts: (i) that whereas the primary and secondary intensions of variables (relative to
assignments) are identical, the primary and secondary intensions of names (and natural
kind terms) are different in 2D semantics, and (ii) that primary intensions are taken to be
the arguments of propositional attitude verbs, but not modal operators, in the system.
The lesson of this argument is easily summed up. Let n rigidly designate o. Then for
any world-states w, ⌜n is F⌝ is true at w iff at w, o has the property expressed by F. So,
it ought to be the case that for any world-state w, ⌜John’s belief that n is F⌝ stands for a
belief about o—one that comes out true when evaluated at w only if at w, o has the prop-
erty expressed by F. Surprisingly, this is not so on strong two-dimensionalist semantics,
which wrongly allows something other than what we might call ⌜the fact that n is F⌝ to
verify (at w) the truth of what strong two-dimensionalism identifies as ⌜the belief that n
is F⌝. In and of itself, this refutation of strong two dimensionalism doesn’t refute other
varieties of two-dimensionalism. But those either don’t identify propositions with sets
of world-states, or haven’t been fully enough specified in order to determine what view
of propositions they presuppose. Hence, they need not be considered here.22
22
See pp. 267–282 of Reference and Description for a battery of arguments, including those given here,
against strong semantic two-dimensionalism. Other versions of two-dimensionalism are defined and
rejected on pages 290–325.
That propositions represent—that is, that they have truth conditions—is something
that needs to be explained. According to “classical” conceptions of propositions of
the sort championed by Frege and Russell, propositions have truth conditions by
their very natures and independently of minds and languages. But no one has ever
been able to explain how anything could have truth conditions by its very nature and
independently of minds and languages. Thus, the classical conception of propositions
leaves unexplained something very much in need of explanation. As a result, in King
[2007, 2009] I rejected classical conceptions of propositions.
To be clear, I am of course not saying that any time a thing has a property, there
must be some explanation for how or why it has that property. I seriously doubt, for
example, that there is any substantial explanation of why/how I have the property of
being identical to Jeff King. But certain sorts of properties are such that we feel com-
pelled to give an account of how/why something manages to possess them and perhaps
even what the possession of them consists in. Some things in the world—names, maps,
mental states, perceptual experiences and so on—represent other things. Properties of
this sort—representing something else or representing it as being a certain way—are
precisely the sort of properties the possession of which we rightly feel compelled to
explain. Perhaps it is not entirely clear what it is that makes a property such that pos-
session of it is something that needs to be explained. But it seems utterly clear that the
property of having truth conditions is the sort of property whose possession is in need
of explanation. Hence, again, since the classical conception of propositions as things
that have truth conditions by their very natures and independently of minds and lan-
guages is incapable of explaining how or why propositions have truth conditions, it is
unacceptable.
These same considerations show that propositions are neither sets (of possible
worlds or anything else), nor n-tuples nor any other such things. “Formal” entities of
this sort are just not the kinds of things that have truth conditions. For example, there
is nothing about the 3-tuple consisting of me, the loving relation and Annie that makes
it true iff I love Annie (as opposed to true iff Annie loves me).1 In general, there is noth-
ing about a set or n-tuple that determines that it is true under this or that circumstance.
Having for these sorts of reasons rejected classical conceptions of propositions, as
well as views according to which propositions are sets or n-tuples, in King [2007, 2009]
I formulated a new account of propositions according to which they are endowed with
representational powers—truth conditions—by thinking agents. I call the account an
account of naturalized propositions. In formulating this theory, I assumed a broadly
Russellian view of propositions on which they are complex, structured entities with
individuals, properties and relations as constituents. The proposition that Michael
swims, for example, has Michael and the property of swimming as constituents; the
proposition that Barry loves Michelle has Barry, the loving relation and Michelle as
constituents; and so on.
As I made clear in King [2007]2 and King [2009], the challenge in formulating this
new account of propositions is to say exactly what relation holds together the con-
stituents of a proposition; and, most importantly, why/how this complex—the propo-
sition—consisting of the constituents standing in the relation in question manages to
have truth conditions. At the time those works appeared, the account of propositions
articulated in them was the only account in which there was an explanation of how/
why propositions have truth conditions. I consider this to be a very significant strength
of the view sketched in those works.
Nonetheless, since the publication of these works, I have come to think that a slightly
different version of my view is superior to the one I had been defending. One of the
goals of the present chapter is to articulate this new view. A second goal is to consider
and respond to some objections to the view. In so doing, I hope to facilitate a better
grasp of my account.
Because the new view I’ll be sketching here is in many respects similar to the view in
King [2007, 2009] and is the result of amending the latter in a number of ways, the best
way to explain the new view is by sketching the account of propositions given in King
[2007, 2009] and then explaining how the new view differs from this one. Idealizing a
lot, let’s begin by considering the sentence “Michael swims” with the syntactic struc-
ture as follows:3
1
I made this point in King [2007] p. 8. More generally, the points just made regarding why propositions
cannot have truth conditions by their very natures and independently of minds and languages, as they were
classically thought to, and why no merely “formal” entity such as an n-tuple can by itself have truth con-
ditions are covered in greater detail in King [2007, 2009 and 2012]. Many of these points are repeated by
Soames in Chapter 3 of the present work.
2
Pp. 3–4, 25–26 and 59–64.
3
John Collins [2007] has recently argued that my pretending syntax is much simpler than it is for exposi-
tory purposes is far from innocent, since the real complexity of syntax ends up being a problem for me.
I respond to Collins below.
1.
Michael swims
Call the syntactic relation that obtains between “Michael” and “swims” in the sentence
here R. I call relations like R that lexical items stand in to form sentences sentential rela-
tions. It is worth pointing out, and we’ll return to this later, that the syntactic relation R
itself has a certain semantic significance in English. That is, English speakers interpret
R in a certain way: they take R to ascribe the semantic value of “swims” to the seman-
tic value of “Michael.” This is part of the reason that the English sentence is true iff
Michael possesses the property of swimming. Further, it is a contingent matter that R is
interpreted by English speakers in the way it is in the sense that there might have been a
language that included the sentence 1, but whose speakers took the sentence to be true
iff Michael doesn’t swim. In so doing, they would have been interpreting R differently
from the way English speakers do.
It is worth saying a bit more about the idea that English speakers interpret R and
syntactic concatenation generally. That English speakers interpret R as ascribing the
semantic value of “swims” to the semantic value of “Michael” results in the fact that
they take 1 to be true iff Michael possesses the property of swimming. Similarly, when
English speakers confront other cases of syntactically concatenated expressions, they
spontaneously and unreflectively compose the semantic values of the concatenated
expressions in characteristic ways. For example, when English speakers confront
“brown cow” they do something like conjoin the properties that are the semantic val-
ues of the two expressions; when they confront “some man,” they do something like
saturate an argument of the relation expressed by “some” with the property expressed
by “man,” resulting in the (relational) property of properties that is possessed by a
property A iff some man has A. That speakers interpret syntactic concatenation in the
ways they do consists in the fact that they spontaneously and unreflectively compose
the semantic values of the concatenated expressions in the ways described. Hence, this
is how my talk of R above being interpreted by English speakers as ascribing the prop-
erty of swimming to Michael should be understood. I’ll put the fact that speakers of
English so interpret R by saying that R encodes ascription in English.
But why do English speakers interpret syntactic concatenation in the small handful
of ways they do? I suspect it will turn out that speakers of different natural languages
interpret syntactic concatenation in the same small handful of ways. This would make
it a reasonable hypothesis that doing so is part of our biologically endowed language
faculty. That this is so would make language acquisition significantly easier. When
encountering concatenated expressions, speakers would be hardwired to compose the
semantic values of the concatenated expressions in a small handful of ways. Hence,
speakers would only need to learn which way to do it in specific cases.
Returning to the main theme, in virtue of the existence of the English sentence 1,
there is a two-place relation that Michael stands in to the property of swimming. The
relation is this: ___is the semantic value of a lexical item e of some language L and ___
is the semantic value of a lexical item e’ of L such that e occurs at the left terminal node
of the sentential relation R that in L encodes ascription and e’ occurs at R’s right terminal
node. Because we also wish to talk about the two-place relation that Michael stands in
to the property of swimming in virtue of the existence of the English sentence “I swim”
taken in a context with Michael as speaker, we should really suppose that in virtue of
the existence of sentence 1, Michael stands in the following relation to the property of
swimming (boldface indicates new additions): there is a context c such that ___is the
semantic value in c of a lexical item e of some language L and ___ is the semantic value in
c of a lexical item e’ of L such that e occurs at the left terminal node of the sentential rela-
tion R that in L encodes ascription and e’ occurs at R’s right terminal node.4 This relation,
I claimed, is the relation that holds Michael and the property of swimming together in
the proposition that Michael swims. As such, I’ll call it the propositional relation of the
proposition that Michael swims.
As I did in King [2007, 2009], I’ll call an object possessing a property, or n objects
standing in an n-place relation, or n properties standing in an n-place relation or etc. a
fact. Then the proposition that Michael swims is the fact consisting of Michael and the
property of swimming standing in the two-place relation mentioned above: there is a
context c such that Michael is the semantic value in c of a lexical item e of some language
L and the property of swimming is the semantic value in c of a lexical item e’ of L such that
e occurs at the left terminal node of the sentential relation R that in L encodes ascription
and e’ occurs at R’s right terminal node.5 Note that this fact is distinct from the fact that
is Michael possessing the property of swimming. The latter fact makes the former fact
qua proposition true.
One might complain that the account I have just given of how constituents are held
together in a proposition simply trades in one problem for another. I have answered
the question of what holds the constituents of propositions together by specifying the
relations that I claim do that job. But, one might complain, this leaves unanswered
the general question of what holds together a relation and its relata when they are
so related.6 It is true that I haven’t answered this question, and in this sense I have
traded in the question of what holds together the constituents of a proposition for the
question of what holds together the components of a fact. My excuse is that I think that
anyone who believes that things stand in relations and possess properties must face the
question, if only to dismiss it, of what holds an object and a property together when the
4
The quantification over contexts here is over possible contexts of utterance. See King [2007] pp. 42–45.
More on this below.
5
I’ll qualify this slightly below.
6
Jim Higginbotham raised this sort of worry at an Author Meets Critics Session on King [2007] at the
Pacific Division Meetings of the American Philosophical Association in Vancouver on April 11, 2009.
object possesses the property or what holds an n-place relation and n objects together
when the objects are so related, etc. So I claim to have reduced the mystery of what
holds propositions together to a mystery that all of us who think that objects possess
properties and stand in relations need to face in any case. Reducing two mysteries to
one seems like progress to me.
I have claimed that the proposition that Michael swims is the fact described above
consisting of Michael standing in the two-place relation mentioned to the property
of swimming. We must now face the question of how/why this fact has truth condi-
tions. In general, such facts aren’t the sorts of things with truth conditions. Consider
the fact consisting of me standing in the two-place sitting in front of relation to my
computer. This fact, of course, obtains but it doesn’t have truth conditions. So how
is it that the fact that I claim is the proposition that Michael swims does have truth
conditions and so is the sort of thing that is true iff Michael possesses the property of
swimming?
One of the most radical and provocative features of the account of propositions in
King [2007, 2009] is the idea that it is something speakers do that endows the fact
that is the proposition that Michael swims, and propositions generally, with truth con-
ditions. This will explain why this fact has truth conditions, while many other facts
do not.
Though the two-place propositional relation binding together Michael and the
property of swimming is highly complex (e.g. it has the sentential relation R of 1 as a
component or “part”), let’s suppress that complexity for a moment and simply focus on
the idea that on the present view the proposition that Michael swims is a fact consisting
of Michael standing in the (complex) two-place propositional relation to the property
of swimming. We can represent this fact/proposition thus:
1P.
2007
Melbourne 200 7
nee 2000
Melbourne 200
2007
220
2007
20
2007
Melbourne 20
0007
2
07
7
Mellbouurne
Melbbourrn
Meelbo
M
(where the picture on the left is Michael; that on the right is the property of swim-
ming and the branching tree structure is the propositional relation). Now one way this
fact could have truth conditions is if speakers interpreted the propositional relation
here as ascribing the property of swimming at its right terminal node to Michael at its
left terminal node. Then the fact would be true iff Michael possessed the property of
swimming. Recall that the sentential relation of the sentence 1 is interpreted by English
speakers as ascribing the property that is the semantic value of “swims” to the seman-
tic value of “Michael,” which we expressed by saying that the sentential relation R of
1 encodes ascription in English. What we are now saying is that if the propositional
relation of 1P were interpreted as ascribing the property at its right terminal node
to the individual at its left terminal node, and so itself encoded ascription, the fact/
proposition would have truth conditions. Encoding ascription understood in this way,
note, is a relational property of the propositional relation itself: the property of being
interpreted as ascribing what is at its right terminal node to what is at its left terminal
node. So henceforth, let’s understand the proposition that Michael swims to be the fact
described above, taken together with the propositional relation having the relational
property of encoding ascription (this means that the fact that is the proposition that
Michael swims is a slightly “larger” fact than we have taken it to be to this point, since it
now includes the propositional relation possessing a certain relational property). In so
doing, we can explain why the proposition/fact has truth conditions.
But the explanation is still preliminary and unsatisfying until we explain what con-
stitutes our interpreting the propositional relation of 1P as ascribing the property of
swimming to Michael. What exactly makes it the case that we so interpret the proposi-
tional relation? Let me sketch my explanation, which comes in two steps.
Call the fact that I claim is the proposition that Michael swims FAST. What we first need
to explain is why it is FAST, rather than some other fact, whose propositional relation we
interpret as ascribing the property of swimming to Michael so that it is true iff Michael
swims. I believe that there are a number of conditions a fact must satisfy in order to be
the one whose propositional relation we so interpret, including being a fact consisting
of Michael standing in a two-place relation to the property of swimming.7 But a crucial
condition is that we must be able to make sense of the idea that speakers have some sort of
cognitive connection to the fact in question. Surely it would be bizarre to hold that speak-
ers are interpreting the propositional relation of a fact in a certain way, where we claim
that they have no cognitive connection or access to it. Further, since we want speakers of
different languages to in some cases grasp the same proposition, we must be able to make
sense of speakers of different languages interpreting the propositional relation of the same
proposition/fact. And this requires them to be cognitively connected to the same fact in
order that we can make sense of their interpreting its propositional relation.
In addition, it seems reasonable to hold that the required cognitive connection to
the fact that is the proposition that Michael swims comes about in virtue of speakers
7
See King [2007] pp. 62–64 and King [2009] p. 268 for discussion.
deploying sentences of their languages. For by the time speakers deploy sentences of
their languages, they presumably must have propositional attitudes whose contents
are the semantic contents of the sentences they are using. But this means that proposi-
tions must exist by that time. That in turn means that speakers must be interpreting the
propositional relations of the facts that are propositions in certain ways by that time.
And in turn, this means that speakers at that time must be cognitively connected to
the relevant facts. The most straightforward explanation of why speakers have cogni-
tive connections to the facts that are propositions as soon as they deploy sentences of
languages is that by deploying sentences of their languages they thereby have cognitive
access to the relevant facts.8
To summarize, then, for a fact to be the proposition that Michael swims, we must
be able to make sense of the idea that speakers of different languages all have cognitive
access to it and do so in virtue of deploying the relevant sentences of their languages.
I’ll now argue that FAST is preeminently a fact of this sort.
To see this, note first that sentences (types) themselves are likely facts in my sense.
For it seems plausible that word types are properties and hence that sentences are prop-
erties standing in sentential relations. Obviously, speakers of e.g. English and German
have cognitive access to the facts that are sentences in their languages, like “Michael
swims” and “Michael schwimmt.” More importantly, as a result, they also have access
to the following “interpreted sentences”:
These are just the sentences, together with the semantic relations the lexical items
bear to their semantic values (including the semantic values themselves—these
relations are represented by vertical lines connecting “Michael” to Michael and
“swims”/“schwimmt” to the property of swimming). Hence these interpreted sen-
tences are just “bigger” facts than the sentences themselves in virtue of including
the semantic relations between lexical items and their semantic values, as well as the
semantic values themselves. We can describe the fact 1IE as follows: there is a context
c such that Michael is the semantic value of “Michael” in c, which occurs at the left
terminal node of the syntactic relation R that in English encodes ascription and the
English word “swims” occurs at the right terminal node of R and has as its semantic
value in c the property of swimming.9 It seems to me that by having cognitive access to
the sentences “Michael swims.” and “Michael schwimmt.” and being competent with
them, English and German speakers thereby have cognitive access to the facts that are
the interpreted sentences IE and IG respectively.
8
Of course the explanation cannot be that they have cognitive access to the facts that are propositions
because they are expressed by the sentences they are deploying. For we are now trying to explain how certain
facts came to be propositions (by having their propositional relations interpreted in certain ways, etc.) and
so we cannot appeal to the fact that they already are propositions expressed by sentences of the relevant
languages.
9
The figures IE and IG fail to capture that “Michael swims.”/“Michael schwimmt.” is English/German
and that Michael is the semantic value of “Michael” relative to a context of utterance (this qualification is
unnecessary here, but would be crucial if we considered the sentence “I swim” in a context with Michael as
the speaker).
1IE.
Michael swims
2007
Melbourne 200 7
nee 2000
Melbourne 200
2007
220
2007
20
2007
Melbourne 20
0007
2
07
7
Mellbouurne
Melbbourrn
Me lbo
Me
1IG.
Michael schwimmt
2007
Melbourne 200 7
nee 2000
Melbourne 200
2007
220
2007
20
2007
Melbourne 20
0007
2
07
7
Mellbouurne
Melbbourrn
Me lbo
Me
Let’s say that the fact of object o possessing property P is a witness for the fact of
there being P’s (i.e. the fact of the property P having the property of being instantiated);
similarly for the fact of o bearing R to o’ and the fact of there being an x and y such that
xRy, and so on. The facts IE and IG are both witnesses for the fact that I claim is the
proposition that Michael swims, namely, FAST. FAST is the result of “existentially gen-
eralizing” on the words in IE/IG and the languages involved.
The crucial point for current concerns is that that having cognitive access to a wit-
ness for a fact is a way of having cognitive access to the fact witnessed: having cogni-
tive access to the fact of o possessing P is a way of having cognitive access to the fact of
there being P’s. But then having cognitive access to IE or IG above suffices for having
cognitive access to FAST, the fact I claim is the proposition that Michael swims. Thus
we can see how English and German speakers can all have access to FAST in virtue
of deploying the relevant sentences of their languages. Hence, we are in a position to
make sense of their interpreting its propositional relation as encoding ascription, and
so make sense of the claim that it is FAST whose propositional relation we so interpret.
But even if we are now convinced that it is FAST’s propositional relation that we
interpret as ascribing the property of swimming to Michael, we need to say what con-
stitutes our so interpreting it. That is, what is it we do that amounts to our so interpret-
ing it? It is simply that we compose the semantic values at the terminal nodes of the
propositional relation in the way we do. In the end, this is just a reflex of the sentential
relation R having the semantic significance it does. When we entertain a proposition,
we work our way up the propositional relation, combining semantic values to yield
new semantic values for further combining. Obviously we must combine or com-
pose those semantic values in some way. In the case of FAST, were we to do anything
other than ascribe the property of swimming to Michael, we would not be combining
semantic values in a manner that is consistent with the way we interpret the syntax of
the sentence 1. It just isn’t coherent to interpret the sentential relation R as ascribing the
semantic value of “swimming” to the semantic value of “Michael,” while composing
the semantic values Michael and the property of swimming in some other way as one
moves up the propositional relation of FAST. Semantic values only get composed once
in understanding the sentence 1, and hence entertaining the proposition FAST. We
either do so in the way dictated by the way we interpret the sentential relation R or not.
To do so in the way dictated by our interpretation of the sentential relation R just is to
interpret the propositional relation as encoding ascription.
To summarize, FAST has truth conditions because speakers interpret its proposi-
tional relation as ascribing the property of swimming to Michael. The account of what
constitutes speakers doing this is in two steps. First, reason was given for thinking that
it is FAST’s propositional relation that gets interpreted as ascribing the property of
swimming to Michael. Second, an account was given of what so interpreting FAST’s
propositional relation consists in. There are facts closely related to FAST that probably
satisfy these conditions as well, so here we would have to claim that FAST is the most
eligible to be the proposition that Michael swims of the facts satisfying all relevant
conditions.
Having sketched the account of propositions I endorsed in previous work, let me
now note a couple ways in which I would alter that account. The fact that I claimed
is the proposition that Michael swims, FAST, is very roughly the result of compos-
ing the syntactic relation R that obtains between “Michael” and “swims” in the sen-
tence “Michael swims” with the semantic relations obtaining between “Michael” and
Michael and “swims” and the property of swimming (relative to a context c), while
existentially generalizing away the lexical items “Michael” and “swims.” In this way,
semantic relations between words and their semantic values play a role in binding
together the constituents of a proposition. In order that there be propositions con-
taining constituents that have never actually been referred to (even using demonstra-
tives or etc.), I had to make use of semantic relations between lexical items and their
semantic values in possible contexts of utterance. That is why propositions are facts of
the following sort: There is a (possible) context c and lexical items a,b of some language
L such that.. .
I now think that I was hasty to take this route. In so far as the relation ___being the
semantic value of ___ relative to context ___ is a perfectly acceptable semantic relation,
so too is ___being the semantic value of___ relative to assignment___ and context ___,
where assignments are functions assigning objects to variables. Hence, if the latter is
an acceptable semantic relation, it can play a role in binding constituents together in
propositions, just as I claimed the semantic relation ___being the semantic value of
___ relative to context ___ did. For a language containing as singular terms names,
contextually sensitive singular terms and variables, we can define the relevant relation
as follows.
For any context c, assignment g, and singular term e, o is the semantic value of e relative to g and c
iff (i) e is a contextually sensitive singular term and o is the referent of e in c; or (ii) e is a name and
o is the bearer of e; or (iii) e is a variable and o is g(e).
We can now say that the proposition that Michael swims is the following fact: there is
a context c, assignment g and language L such that for some lexical items a and b of L,
Michael is the semantic value of a relative to g and c and the property of swimming is the
semantic value of b relative to g and c and a occurs at the left terminal node of syntactic
relation R that in L encodes ascription and b occurs at R’s right terminal node.
Several points should be emphasized about this characterization of propositions.
First, as before, the existence of the English sentence “Michael swims” suffices for the
existence of the above fact/proposition. For given the existence of this sentence, that
the lexical items “Michael” and “swims” have as their semantic values relative to any
c and g Michael and the property of swimming, respectively, and the semantic sig-
nificance of the syntactic relation R the lexical items stand in in the English sentence
“Michael swims,” there is a context c, assignment g (namely, any context and assign-
ment) and language L (English) such that for some lexical items a and b of L (“Michael”
and “swims”), Michael is the semantic value of a relative to g and c and the property of
swimming is the semantic value of b relative to g and c and a occurs at the left terminal
node of the syntactic relation R that in L encodes ascription and b occurs at R’s right
terminal node. But of course, this is just to say that the fact that I claim is the proposi-
tion that Michael swims exists.
Second, it suffices for the existence of this fact/proposition that there be a syntactic
structure of the following sort in English:
1a.
x swims
where the left terminal node is occupied by a variable, the right terminal node is occu-
pied by “swims” and the syntactic relation obtaining between them as before has the
semantic significance of encoding ascription. For given that this is so, again, there is
a context c, assignment g (namely, any context and an assignment that maps “x” to
Michael) and language L (English) such that for some lexical items a and b of L (“x”
and “swims”), Michael is the semantic value of a relative to g and c and the property of
swimming is the semantic value of b relative to g and c and a occurs at the left termi-
nal node of syntactic relation R that in L encodes ascription and b occurs at R’s right
terminal node.
Third, we no longer need to make use of semantic relations between lexical items
and entities in merely possible contexts of utterance. Recall that we needed to do that
before in order to secure the existence of propositions that contained constituents that
had never actually been the semantic value of any expression (relative to a context).
But now syntactic structures like 1a insure e.g. that the proposition that o swims exists
even where o is an object that has never been the semantic value of any expression
relative to any actual context of utterance.10 For even if that is so, there is a context c,
assignment g (namely, any context and an assignment that maps “x” to o) and language
L (English) such that for some lexical items a and b of L (“x” and “swims,” respectively),
o is the semantic value of a relative to g and c and the property of swimming is the
semantic value of b relative to g and c and a occurs at the left terminal node of the syn-
tactic relation R that in L encodes ascription and b occurs at R’s right terminal node.
And this is just to say that the proposition that o swims exists.11 This means that our
account of propositions only commits us to the existence of actual contexts of utter-
ance and so doesn’t presuppose or require the existence of possible worlds. That in turn
means that on the new account, we leave open the option of analyzing possible worlds
in terms of propositions, (see e.g. Adams, 1974). On the old account, we could not do
that since in characterizing propositions we appealed to possible contexts of utterance,
which were effectively centered possible worlds. Hence, the old account was precluded
10
I don’t mention assignments here precisely because, as mentioned above, on the old view the only
semantic relation that played a role in binding constituents together in proposition was ___being the seman-
tic value of ___ relative to context___.
11
What about propositions with properties as constituents that have never been expressed by any predi-
cate nor been actually designated by contextually sensitive expressions? Well, given sentences like “Several
properties are had by John” will have syntactic structures like “x was had by John,” where the “x” takes prop-
erties as values. Given what was said in the main text, this is enough for the proposition that P was had by
John to exist, where P is a property that no predicate expresses and that has never been actually referred to by
contextually sensitive expressions.
from analyzing possible worlds in terms of propositions.12 That the new account leaves
open this possibility the old account precluded strikes me as a point in its favor.
Finally, the explanation for why our facts/propositions have truth conditions goes
through as before.13
Let’s now turn to a second emendation of the view defended in King [2007, 2009],
henceforth assuming the above emendation is accepted. As was said before, on the
current view of propositions, the facts that are propositions very roughly result from
composing syntactic relations of sentences with semantic relations between the lexical
items in the sentences and their semantic values (relative to a context and assignment),
while “existentially generalizing away” the lexical items (context, language and assign-
ment). But the syntactic relation between the lexical items is not existentially general-
ized away and so remains a component of the fact that is the proposition. This has the
result that any syntactic difference between sentences results in the sentences express-
ing distinct propositions. Though some may think that this results in propositions
being individuated too finely, elsewhere I have argued that this isn’t so.14 Nonetheless,
there is a way to alter the account with the result that propositions are individuated
somewhat less finely.15 The idea is simply to “existentially generalize away” the syn-
tactic relations like R that to this point have been components of the facts that are
propositions. Thus we now claim that the proposition Michael swims is the follow-
ing fact: there is a context c, language L, syntactic relation R, assignment g and lexical
items a and b of L such that the semantic values of a and b relative to c and g are Michael
and the property of swimming, respectively, and a and b occur at the terminal nodes of
R, where R ascribes the semantic value of b relative to g and c to the semantic value of a
relative to g and c in L.
This account, when extended to other sentences and propositions, has the result that
sentences containing lexical items with the same semantic values (relative to param-
eters) but e.g. whose lexical items occur in different orders, may nonetheless express
the same proposition. For example, suppose there were a language, Renglish, in which
the following is a sentence that is true iff Michael swims:
1b.
swims Michael
12
Soames [2011] criticizes the old account for ruling out analyzing possible worlds as sets of propositions.
13
The explanation needs to be slightly complicated in so far as we have to say something about how cogni-
tive access to a syntactic structure like 1a (which may or may not be sentential) gives us cognitive access to
the fact of there being a context c, assignment g and language L such that for some lexical items a and b of L,
o is the semantic value of a relative to g and c and the property of swimming is the semantic value of b rela-
tive to g and c and a occurs at the left terminal nod of syntactic relation R that in L encodes ascription and b
occurs at R’s right terminal node.
14
King [2011]
15
I might owe this idea to Jeff Speaks.
Then on the new account, 1b and 1a above express the same proposition, despite the
syntactic differences between the two sentences. Henceforth, this is the account of
propositions we’ll be working with.
I’ll now turn to objections to this view of propositions. I called the fact with which
I identified the proposition that Michael swims on my old view FAST; let’s call the fact
I now identify with the proposition that Michael swims FAST.
I’ll begin by considering two closely related objections to the explanation as to how/
why propositions have truth conditions. The first objection runs as follows. According
to the explanation offered, e.g. English comes into existence bringing into existence
facts like FAST above. But in order for the fact FAST to have truth conditions and so
be a proposition, the relation binding together its components—its propositional rela-
tion—must be interpreted by speakers of a language, thereby endowing it with seman-
tic significance. This, however, will only happen after languages exist and speakers are
using and understanding sentences. Thus, the account of propositions on offer postu-
lates a time when English sentences existed, had truth conditions and were being used
and understood by speakers, but at which time no propositions existed and so speakers
had no propositional attitudes. This means speakers were using and understanding
language without having any beliefs or intentions, and indeed without asserting any-
thing (since making assertions and having beliefs and intentions requires bearing rela-
tions to propositions). Surely this is implausible.16
This objection rests on a misunderstanding of my view. I do not claim that there was
a time when language existed and speakers were using and understanding sentences,
but propositions did not exist. Quite the contrary. I claim that languages (at least
some—see below), propositions and propositional attitudes came into existence at the
same time. Once speakers are deploying sentences of their languages, they thereby are
cognitively connected to the facts that I claim are propositions and are interpreting
their propositional relations in such a way that the structured contents of those sen-
tences have truth conditions. In so doing, language users endow the propositions with
truth conditions.
It is true that my explanation of why speakers interpret the propositional relation of
the proposition that Michael swims in the way they do appeals covertly to the way in
which they interpret the sentential relation in the sentence “Michael swims.” Speakers
work their way up the propositional relation composing semantic values in the way
they do, thus interpreting the propositional relation, because it is the way the semantic
significance of the syntax dictates that they compose these semantic values. In this way,
the explanation of why speakers interpret the propositional relation the way they do
appeals to how they interpret sentential relations. (Recall that the explanation of how
and why speakers interpret sentential relations in the way they do appeals to our bio-
logically endowed language faculty.) So in this sense, speakers interpreting syntactic
16
Soames [2011] raises a version of this objection.
17
I had already made this explicit in King [2009] pp. 266–268.
18
Greenberg [2011] argues that certain pictures have semantic contents.
19
All this suggests that a better title for King [2007] would have been The Nature and Structure of Linguistic
Content.
In the first part, I shall consider the claim that syntactic structure is the actual structure of con-
tent. In a series of articles and a recent book, King has argued that it is. My riposte will be that
there is just far too much structure in the syntax for this to be plausible.23
My substantive complaint against this proposal [King 2007] is that syntax provides far too much
structure, much more than can be accommodated as the values that determine a sentence’s truth
conditions, the proposition it expresses. This, further, is not a mere quirk of some construc-
tions: the surplus structure is generally exhibited and reflects the fact that current syntactic
theory is not in the business of providing structures that answer to philosophical conceptions
of propositions.24
These quotes suggest that Collins misunderstands my view in a way that matters to
the debate. He seems to think that I have some philosophical conception of propo-
sitions and their structures, and that I think that somehow syntax just happened to
deliver structure exactly answering to this conception.25 Such a view appears to require
a miraculous convergence and so does seem implausible; but it isn’t my view. To antici-
pate, I think that there are these syntactic structures that developed the way they are
20
p. 806
21
p. 806
22
p. 807–808
23
p. 808
24
p. 810
25
Another quote suggesting the same thing occurs on p. 810: “The proposal under discussion presumes
that syntactic theory arrives at the same point, from a different direction, as the philosopher is aiming at.”
for who knows what reason and we press them into service as providing the structures
of propositions.
Let’s now consider the details of Collins’s objection. The first worry Collins raises
for my view concerns movement.26 The problem is that even in quite simple sentences,
there is movement; and movement is understood as copying (and deletion at PF but
not LF). Collins assumes that movement is driven by the requirement that uninter-
pretable features be valued (and uninterpretable features must be valued if a syntactic
structure is to be interpretable). Collins provides the following illustrative derivation
for “Bill sleeps” (where +/- mark interpretable and uninterpretable features, respec-
tively; and underlining marks the valuing of uninterpretable features):27
(3)a. Bill sleeps
b. [VP Bill{+1st, +sing, -Case} sleep] (I think Collins intends +3rd and –Nom here)
c. [T’{+Pres, -3rd Per, -Sing Num, -EPP} [VP Bill{+3rd, +sing, -Nom} sleep]]
d. [TP Bill{+3rd, +sing, -Nom} [T’{+Pres, -3rd Per, -Sing Num, -EPP}[VP <Bill{+3rd, +sing, -Nom}> sleep]]]
In line c, the uninterpretable features of T’ are valued by matching the interpretable
features of “Bill.” The movement of “Bill” to TP in d is driven by the requirement that
the uninterpretable EPP feature of T be valued. This all means that for a sentence like
“Bill sleeps,” we get the following syntax:
3e.
+Pres, etc.
Bill sleeps
Collins writes:
There are two crucial things to note here. Firstly, the derivation is driven to value uninterpretable
features, not to provide a propositional structure. This motive is semantic in nature, for an unin-
terpretable feature is precisely a feature that has no semantic significance, but the structure pro-
duced departs from what we have been imagining is required for propositional structure. This is
the second point. The copying of Bill higher up the structure creates an item that looks to be sur-
plus relative to property instantiation; that is, the lower copy of Bill exhausts what propositional
structure appears to be interested in, namely, Bill is fixed as the agent of sleep, the instantiater of
26
Actually, prior to this Collins raises another objection. He begins by making the point that syntax has
jettisoned the notion of a sentence (S) in favor of thinking of it as the projection of tense (T) or inflection (I)
(pp. 811–812). Collins seems to think this creates problems for philosophers, apparently including me (on
p. 812 he writes “As we’ll see, it is precisely this well motivated move in linguistics that causes trouble for the
philosophers.”). But I don’t know why he thinks this. I agree that what we call sentences are TPs or IPs; and
I don’t see at all why that is any sort of problem for me. Since this objection seems to me not to have any force,
I don’t discuss it in the main text.
27
P. 814
the property of sleeping. Further, after the introduction of tense, we have a temporal dimension.
Still, Bill gets copied for a reason that is not recorded in the proposition, as it were (perhaps Case
valuation or EPP elimination). Of course, per the first point, this is no problem or mystery at all
from the perspective of the syntax. The syntax operates to value uninterpretable features, and
copying is part of that mechanism, a displacement of items to meet interface demands, not to
create a propositional structure.28
Again here, Collins suggests that I claim that syntactic derivations are driven to cre-
ate propositional structure. Again let me say that my view is that we exploit syntactic
structure, however derived for whatever reason, and make use of it to produce propo-
sitional structure. So I can happily agree with Collins that “the derivation is driven
to value uninterpretable features” and that “the syntax operates to value uninterpret-
able features, and copying is part of that mechanism, a displacement of items to meet
interface demands.” The crucial question is: can I build the propositional relation of
the proposition expressed by “Bill sleeps” out of the “sentential relation” in 3d above?
I think it is clear that I can. Here is the proposition: There is a context c, assignment
g, lexical items a,b and sentential relation R of language L such that copies of a and b
occur at the terminal nodes of R, where the semantic value of a relative to c and g is Bill
and the semantic value of b relative to c and g is the property of sleeping and R ascribes
the semantic value of b to a.
So I think that taking account of what Collins takes to be real syntactic complexity
(including copied material) will complicate my account, (indeed, that is why I ignored
real syntactic complexity in outlining my view), but I do not see that it refutes or even
damages it.29
That said, it should be remembered that syntactic theory is likely far from finished.
My account requires that propositional relations can be built out of the syntactic struc-
tures the true syntactic theory assigns to sentences of natural languages. However, my
point is that even if, as unlikely as it may be, syntax as described by Collins is the true
syntactic theory, this requirement can be met.
Collins considers another example (4a-j p. 815) and writes:
The details of these constructions and movement/copying in general are quite complex, but a
generalization can be readily extracted: copying occurs after everything relevant to a proposi-
tional (truth conditional) structure has been determined. But copying creates new structure,
and so there is more structure than appears to be demanded by the requirements of the encoding
of structure relevant to propositions. Again, the crucial point is that, according to recent theory,
copying is motivated by semantic requirements, that is, the valuing of uninterpretable features
on functional heads, such as Tense. So, all of the structure is playing a semantic role in one sense,
just not the sense apparently required for the encoding of propositions.30
28
P. 816
29
I have ignored tense here, but there is no reason to think this will cause problems (I just don’t know
exactly what the tense head looks like nor exactly how the semantics of tense works).
30
Pp. 815–816
I just don’t see an argument here against my view. Collins seems to take me to claim
that syntax must produce only as much structure as is required to get enough and the
right sort of propositional structure. But I claim no such thing! Again, you might think
this if you thought there was some pretheoretical notion of propositional structure and
you thought it was the job of syntax to deliver it. But again that isn’t my view. To repeat,
my view is that syntax produces structures, driven by whatever mechanisms, and we
exploit these structures by building propositions out of them. There is no claim that we
get exactly the degree/amount of structure required for structured propositions. The
claim is just that syntactic structure gives us enough structure and so can be made to
do the job. From my perspective, then, it is as though I had claimed that people press
pianos into service as tables and Collins argued that I can’t be right because pianos
have more structure than tables require.
At one point, Collins does seem to touch on something like my view.32 The problem
he raises for this view is that “...if propositions are structured by syntax, then proposi-
tions appear to contain elements that play no role in the determination of truth condi-
tions. Any bullet can be bitten, but if this one is, I cease to understand the position.”33
But here I don’t understand what Collins doesn’t understand! If Collins is right about
the real complexity of syntax, on my view propositions contain elements/structure
that go beyond what would be strictly required to determine their truth conditions.
I don’t see anything hard to understand about that position; I don’t see that it involves
biting any bullet; and I don’t see that Collins has shown any problem with it. I conclude
that Collins’s objection to my account of propositions based on the real complexity of
syntax can be successfully countered.34,35
Finally, let me turn to an objection to my view raised by Soames [2011]. Because
Soames states the objection so clearly and concisely, let me quote it in full:
Since English contains both ⌜ the fact that S⌝ and ⌜ the proposition that S⌝ —which for King des-
ignate different things—his view seems to require “that”-clauses to be ambiguous between the
readings they bear in ⌜ Pam regrets (the fact) that S⌝ and ⌜ Pam believes(the proposition) that S⌝.
31
P. 816
32
First paragraph of section 2.2 p. 818—he calls this “biting the bullet,” but I have no idea why.
33
P. 818
34
On pp. 819–825 Collins considers several responses to his objection. I want no part of his suggested
responses, as I think his objection can be countered in the way I have just done. Collins also complains that
“biting the bullet” (which seems to be something like my view, though again I can’t at all see what bullet I am
biting) has bad empirical consequences. Specifically, he argues that my view individuates propositions too
finely (see p. 820). Since I respond to this objection in King [2011], I won’t discuss it here.
35
Sometimes Collins’s point seems to be that according to current syntactic, the syntactic structures of
sentences (TPs, IPs) aren’t appropriate for any sort of truth conditional semantics (see pp. 805–806 and 807).
Since this isn’t an objection to my account of propositions specifically, I haven’t tried to address it here.
But he neither gives any linguistic argument that this ambiguity exists, nor rebuts seeming evi-
dence to the contrary—e.g., “Pam regrets that she is pregnant. Although her parents don’t realize
it yet, in time they will come to believe it.” Here, the fact regretted is described as something that
will eventually be believed—a proposition. How, given the supposed difference between the two,
can that be?36
Two things need to be said in response to this objection. First, Soames assumes here
without argument that expressions like “the fact that Annie is smart” designate the
things I call facts. If that were so, then e.g. “the fact that Rebecca swims” and “the prop-
osition that Rebecca swims” would designate different things on my view (the former
would designate the fact (in my sense) that makes the latter true) and Soames’s objec-
tion would be up and running. However, it is important to see that it is not at all clear
that expressions like “the fact that Annie is smart” do designate the things I call facts.
I picked out a class of things of the following sort: n objects standing in an n-place
relation, n properties standing in an n-place relation, and so on. I then stipulated that
I would call these things facts. But then clearly it is a very substantial claim that the
ordinary language English locution “the fact that Annie is smart” designates the sort of
thing that I stipulatively called a fact.37 Perhaps it is easier to see this if we imagine that
I had called the things I call facts states of affairs instead. Then clearly the question of
whether the expression of ordinary English “the fact that Annie is smart” designates a
state of affairs in my sense is a substantial question. I really don’t know what arguments
might be given that such expressions do designate what I call facts. But the impor-
tant point for our purposes is that Soames’s objection here depends on this claim; but
Soames has given no reason to accept it, nor is it clear what reason might be given for
accepting it.
So Soames is correct that in King [2007] I didn’t offer evidence that “the fact that
Rebecca swims” and “the proposition that Rebecca swims” designate different things,
nor did I rebut seeming evidence to the contrary. But that is because, as I explicitly said
in King [2007], I was not assuming that expressions like “the fact that Rebecca swims”
designate the sort of thing that I call a fact.38 As just indicated, that this is so is a sub-
stantial claim that I do not know how to argue for (or against). Since I was not assum-
ing that these expressions designate different things, I was not committed to offering
an argument that they do (or to rebutting seeming evidence that they don’t).
But let’s waive this worry and simply assume that expressions like “the fact Annie
is smart” really do designate the things that I call facts. Then Soames is correct in
claiming that on my view e.g. “the fact that Rebecca swims” and “the proposition that
Rebecca swims” would designate different things. Soames then objects that I give no
reason for thinking that these expressions do designate different things nor do I rebut
what Soames offers as seeming evidence to the contrary. Unfortunately for Soames’s
36
Soames [2011] p. 9
37
I actually made this point already in King [2007] p. 149 note 30 and p. 150 note 31.
38
See note 37.
objection here, there is lots of evidence that such expressions do designate different
things.39 Further, Soames’s seeming evidence to the contrary can be explained on the
hypothesis that the expressions in question designate different things.
Let’s begin with the evidence that expressions of the form “the fact that...” and “the
proposition that...” designate different things; and that “bare” that-clauses can desig-
nate either sort of thing.
The first bit of evidence concerns one of the diagnostics for so-called factive con-
texts. Generally speaking, factive contexts allow that-clauses that begin “the fact that,”
whereas non-factive contexts do not:40
4a. Isabel regretted the fact that she didn’t buy insurance.
b. The fact that Amy won the race surprised us.
c. The fact that the moon creates the tides is well known.
d. Isabel believed [*the fact] that she didn’t buy insurance.
e. [*The fact] that Amy won the race is likely.
f. [*The fact] that the moon creates the tides is true.
On the view that “the fact that...” and “the proposition that...” designate different sorts
of things, this data is easy to explain. Factive contexts predicate properties of the sort of
things designated by expressions of the form “the fact that...,” which aren’t propositions
but rather are facts. Non-factive contexts predicate properties of propositions. The
properties appropriately predicated of facts are different from the properties appropri-
ately predicated of propositions.
If it isn’t already clear, the same sort of data suggests that it can’t just be that while
expressions of the form “the fact that...” designate true propositions, expressions of the
form “the proposition that...” designate propositions:
5a. ?Isabel regretted the true proposition that she didn’t buy insurance.
b. ?The true proposition that Amy won the race surprised us.
c. ?The true proposition that the moon creates the tides is well known.
d. Isabel believed the true proposition that she didn’t buy insurance.
e. ?The true proposition that Amy won the race is likely.
f. The true proposition that the moon creates the tides is true.
To the extent that 5a-c are coherent, they clearly don’t mean the same thing as 4a-c.
But if expressions of the form “the fact that...” designated true propositions, it is
utterly unclear why there would be this difference between 4a-c and 5a-c. Similarly,
39
There is lots of evidence that the expressions in question designate different things, contrary to what
Soames suggests. What there isn’t, to my mind, is evidence that expressions like “the fact that Rebecca
swims” designate what I call facts.
40
This is discussed in Parsons [1993]. As with many diagnostics, this one is not perfect as Parsons
himself notes.
5d-f are fine.41 But if expressions of the form “the fact that.” designate true proposi-
tions, then we would expect 4d-f to be fine as well.
Finally, since sentences like 4a-f are all fine with “bare” that clauses (“that...”), in so far
as we are convinced that expressions of the form “the fact that...” and “the proposition
that...” designate different things and that they must designate facts in sentences like
4a-c in order that they are felicitous whereas they must designate propositions in sen-
tences like 4d-f in order that they are felicitous, we should hold that “bare” that-clauses
may designate facts or propositions.
A second bit of evidence that expressions of the form “the fact that...” and “the prop-
osition that...” designate different things, presented in Parsons [1993], is that when we
attempt to quantify across both a factive and a nonfactive context, the results are often
quite anomalous:
6a. *Everything John says Glenn discovers.
b. *Everything John believes is tragic.
c. *Everything John knows is likely.
By contrast, as Parsons notes, quantifying over two factive or non-factive contexts is
generally fine:
7a. Everything John says Glenn believes.
b. Everything Sue regrets John discovers.
c. Everything John says is likely.
d. Everything John discovers is tragic.
If the things that factive contexts appropriately predicate properties of, and the things
that nonfactive contexts appropriately predicate properties of, are different, the
anomolousness of 6a-c would be explained.42
A final bit of evidence that expressions of the form “the fact that...” and “the proposi-
tion that...” designate different things is given by sentences such as:43
8a. That the rock hit the window caused the window to break.
b. The fact that the rock hit the window caused the window to break.
c. *The true proposition that the rock hit the window caused the window to break.
d. *The proposition that the rock hit the window caused the window to break.
41
5e’s oddity appears to have a straightforward pragmatic explanation. Use of “The true proposition that
Amy won the race” commits the speaker to the claim that Amy won the race. To then say that her winning
the race is likely is odd.
42
Below we’ll consider the fact that sometimes sentences that quantify into factive and non-factive con-
texts sound fairly good. Examples include “There is something that Glenn knew and Cris doubted” and
“John believes everything he knows.” Presumably many philosophers would even agree with the latter.
43
Such evidence is discussed in Harman [2003] and Asher [2000], both of whom conclude that the things
designated by expressions of the form “the fact that...”—facts—are distinct from the things designated by
expression of the form “the [true] proposition that...”—[true] propositions.
8b suggests that the things designated by expressions of the form “the fact that...” are things
that can cause other things. 8c and d suggest that neither propositions nor true propositions
can do so. This strongly suggests that the things designated by expressions of the form “the
fact that...” are distinct from the things designated by expressions of the form “the [true]
proposition that....” Finally, the felicity of 8a strongly suggests that the “bare” that-clause in
it designates the same thing that “The fact that the rock hit the window” designates in 8b.
Hence, contrary to what Soames suggests, there is ample evidence that expressions of
the form “the fact that...” and expressions of the form “the proposition that...” designate
different things; and that that-clauses are capable of designating both kinds of things.
However, there are two pieces of counter-evidence to these claims. First, there is
Soames’s example:
9. Pam regrets that she is pregnant. Although her parents don’t realize it yet, in time they
will come to believe it.
Here the occurrences of “it” appear to have “that she is pregnant” as their antecedent. If the
latter designates a fact (“Pam regrets...” is a factive context), then so should the former. But
then why is it felicitous to say that this fact is believed in the final clause? Second, there are
examples in which quantifying across a factive and a non-factive context seems felicitous:
But this would seem to require that there be some one thing that both the factive and
non-factive contexts predicate a property of, contrary to what has been claimed.
Before saying something in response to these considerations, there are a couple
points that should be stressed. First, as indicated above, we have seen quite substan-
tial evidence that expressions of the form “the fact that...” and “the [true] proposition
that...” designate different things and that “bare” that-clauses are capable of desig-
nating either kind of thing. Hence in encountering two bits of counter-evidence to
these claims, the weight of the overall evidence is still very much in favor of these
claims. This strongly suggests that we should explain the apparent counter-evidence
while maintaining the claims in question. Second, because I think the considerations
here are subtle and complex, I won’t be trying to provide a definitive rebuttal to the
counter-evidence. Rather, I’ll indicate some general strategies of rebuttal.
The first strategy, due to Parsons [1993], takes as its starting point the idea that e.g. the
fact that gasoline is lighter than water and the true proposition that gasoline is lighter
than water are intimately related. Following Parsons, let’s say that the relevant fact in
this case corresponds to the true proposition that gasoline is lighter than water.44 Now
44
Here we come up against one of the subtleties alluded to above. For any that-clause designating a true
proposition, it seems we can form a phrase of the form “the fact that...,” which apparently designates the
corresponding fact. Does this means that for every true proposition, there is a corresponding fact? Those
skeptical about the claim that every true proposition is made true by a corresponding fact (e.g. because they
don’t think true negative existential propositions are made true by corresponding facts; or they don’t think
that true propositions about long past things (e.g. that Socrates was a philosopher) are made true by corre-
sponding facts) presumably would also be skeptical about answering this question affirmatively.
let’s suppose that we have an object language in which sentences refer to truth-values.45
We then add the expression “^” that fronts sentences (and perhaps other expressions
too) with the stipulation that “^[S]” refers to the proposition that S.46 Finally, letting f
be a function that maps true propositions to their corresponding facts, we let “c” name
f. Then “c^[S]” refers to the fact that S. The idea is that Soames’s example can be repre-
sented as follows:
9. Pam regrets c^ [she is pregnant]. Although her parents don’t realize it yet, in
time they will come to believe it.
where the final occurrence of “it” is anaphoric on a term of the from “c^[S].”47 Now
instead of having the entire term as its antecedent, and hence referring to a fact, the
pronoun could have as its antecedent the embedded term “^[S]” and so refer to the
proposition this expression refers to. This would explain the felicity of the example.48
A second strategy for explaining Soames’s example while maintaining that expres-
sions of the form “the fact that...” and “the proposition that...” designate different things
(and that bare that-clauses can designate either) is to appeal to other cases in which
something similar seems to happen. Again, we begin by assuming that e.g. the fact
that snow is white and the proposition that snow is white are intimately related. Now
consider the following example, inspired by examples discussed by Chomsky [2000]:
11. The book I just stole from the library is on my desk. It was written in 1801 and
has been translated into many languages.
It is hard to resist the idea that the occurrence of “The book I just stole from the library”
in the first sentence designates the particular book (token) on my desk. It, after all, is
the thing I stole from the library and that is on my desk. However, equally, on reflec-
tion it is hard to resist the idea that the second sentence predicates a property of a
quite different thing: a thing that was written in 1801 and has been translated into many
languages. Surely, this thing isn’t the particular book on my desk that I just stole. And
this is so despite the fact that the pronoun in the second sentence is anaphoric on “The
book I just stole from the library” in the first sentence. In such cases, it looks like the
antecedent of a pronoun designates one thing and the pronoun anaphoric on it desig-
nates something else, albeit something intimately connected to the former. But if this
can happen in a mundane case of this sort, there is no reason to think it can’t happen in
Soames’s example 9.
Concerning our example 10
45
Parsons [1993] also supposes that names refer to individuals and predicates refer to functions from
n-tuples of objects to truth values.
46
So that-clauses referring to propositions are terms of the form “^[S].”
47
Here we suppose that bare that-clauses that refer to facts are terms of the form “c^[S].”
48
On such a view, both expressions of the form “the fact that...” and “bare” that-clause that designate facts
would be thought of as having the form “c^[S].” See previous note.
1
Though this is complicated a bit by the fact that proponents of the non-reductive views might analyze
entities of some other type as a kind of proposition, as with views which take facts to be true propositions,
and further complicated by the fact that proponents of reductive views might as well—so long, of course, as
the category to be reduced to propositions isn’t the same as the category to which the reductive theorist aims
to reduce propositions.
2
See Russell’s discussion of “that Charles I died in his bed,” in Russell (1910), p. 151.
the fact of the thing’s having the property; but there is no room for some other thing,
the proposition that the object has that property. To be sure, Russell’s worry here rests
on a bare metaphysical intuition: that once we cross the relevant objects, properties,
and facts off the list, there’s no room left for entities of some other kind, which are
“about” the relevant situation, to squeeze their way in. But, unargued as it is, the intui-
tion seems to me to be a powerful one—and the best way to answer it seems to be to
identify that Gore won in 2000 with some object, property, fact, or event in which we
have independent reason to believe.
A second reason to be skeptical about sui generis propositions, separate from but
related to Russell’s, comes from the oddness of supplementing an ontology of objects,
properties, facts, events, etc. with another category of entity whose sole roles are to
be the contents of mental states and sentences, and the bearers of truth and falsity. It
seems oddly anthropocentric to add a category to our ontology just to play these roles;
surely it would be better, all things being equal, to find a class of entity to play these
roles among the types in which we already have independent reason to believe.3
My aim will be to advance a reductive view of propositions. In my view, proposi-
tions are a kind of property. As is well known, defenders of (close relatives of) this view
have included Roderick Chisholm and David Lewis—and, more recently, Peter van
Inwagen has defended the view that propositions are 0-place properties.4 My view is in
many ways similar to these. But as will become clear, my view also borrows a lot from
an apparently quite different view of propositions—Jeff King’s view that propositions
are a kind of fact.5
3
Though here I am setting aside issues about the nature of the constituents of propositions, this is one
place in which these issues come to the fore. This is because a defender of a Fregean view of content—on
which the constituents of propositions are modes of presentation of objects and properties rather than those
objects and properties themselves—might reasonably claim to have an answer to Russell’s skeptical argu-
ment; after all, she might say that true propositions are, in effect, modes of presentations of facts, whereas
false propositions are modes of presentation of nothing—but which could be modes of presentations of
facts, were the facts different. But this won’t help with the second worry, which focuses on the oddness of
postulating a class of entities to serve as the contents of mental states and sentences.
4
Lewis (1979), Chisholm (1981), van Inwagen (2004).
5
See King (2007) and Chapter 4 above.
property of talking.6 In this sort of case, let’s call Amelia and the property of talking the
constituents of PROP; and, in general, we can say that if e is a semantically simple expres-
sion which is a part of some sentence S, then, in our sense, the content of e will be a
constituent of the proposition expressed by S. This gloss on “constituent” is intended to
make the claim that propositions have constituents pretty metaphysically unexciting,
and therefore common ground between various views of propositions; it does not, for
example, build in any assumptions to the effect that the relationship between proposi-
tions and their constituents is analogous to the relationship between composite mate-
rial objects and their parts.7
Whatever view we take of the nature of PROP, it seems that it should bear some close
relation to its constituents, namely Amelia and the property of talking. After all, there
are some interesting necessary connections between the two, like the following:
Necessarily, anyone who entertains a thought with content PROP entertains a thought about
Amelia.
Necessarily, anyone who entertains a thought with content PROP entertains the thought of an
object that has the property of talking.
The existence of necessary connections of this sort indicates that, in some sense or
other, the constituents of a proposition are part of its identity.
But, on the other hand, it would be a mistake to respond to the necessary connections
between propositions and their constituents by simply identifying propositions with
the collection of their constituents, or saying that propositions are “nothing over and
above” their constituents. That this would be a mistake is shown by the fact that there
are distinct propositions with the same constituents. The propositions expressed by:
Jane loves John.
and
John loves Jane.
each have John, Jane, and loving as their constituents, but are plainly not identical.
The same lesson can be drawn from Russell’s observation in the Principles of
Mathematics (§52) that substitution of one expression for another with the same
content can transform a sentence—which expresses a proposition relative to a
6
Here, for simplicity, I am setting aside some qualifications—first, the qualification that sentences typi-
cally express propositions only relative to contexts and, second, that perhaps sometimes sentences express
only fragments of propositions (even relative to contexts).
This discussion also obviously presupposes a particular (Russellian) view about what sorts of thing
the constituents of propositions are. But the foregoing could be restated to accommodate other views; for
example, a Fregean will want to replace talk about objects as constituents of propositions with talk about
individual concepts, and will want to replace talk about properties and relations with talk about modes of
presentation of those properties and relations. Nothing in what follows will depend upon one choice or
another here, though I will stick with the Russellian view for simplicity. (And because I think that it is true.)
7
For some discussion of whether we should take “constituents” talk more seriously than this, see Ch.
11 below.
context—into a string of words which does not express a proposition. To adapt his
point to the present example, “loves” and “loving” are apparently both terms for the
relation of loving; but, while
John loves Jane
expresses a proposition,
John loving Jane
expresses nothing. If propositions really were nothing over and above their constit-
uents, it is puzzling why this should be so. Why should switching out one term for
another with the same content change a string of words from one which is proposi-
tion-expressing to one which is not?
This all seems to indicate that propositions are—speaking somewhat metaphori-
cally—their constituents plus some extra ingredient. The problem of giving a theory
of the nature of propositions then becomes the problem of saying what this extra
ingredient is.
Jeff King’s theory of propositions is based in part on the idea that this extra ingre-
dient has something to do with the syntax of proposition-expressing sentences. This
seems plausible especially when we think about examples like Russell’s; it seems quite
plausible that the difference between “John loves Jane” and “John loving Jane” is to be
explained in terms of the fact that strings with the syntactic form of the former—those
consisting of a name, concatenated with a two-place predicate and another name—
express propositions in English whereas strings with the syntactic form of the latter—
those consisting of a name, followed by an abstract singular term, followed by another
name—do not.
But the “extra ingredient” we’re looking for can’t just be a syntactic relation—a point
which King brings out nicely via the example of a possible language, Nenglish, which
is like English but for the fact that concatenation of a name and a predicate expresses a
proposition which is true iff the referent of the name does not instantiate the property
expressed by the predicate. Hence “Amelia talks” would express a different proposition
in Nenglish than it does in English—despite the fact that in both languages Amelia and
the property of talking are the constituents of the proposition expressed by the sen-
tence, and that in both languages the syntactic form of the sentence is that it is a name
concatenated with a monadic predicate.
What this example brings out is that syntactic relations, just like linguistic expres-
sions, can make different semantic contributions in different languages. Perhaps, then,
our extra ingredient has something to do with the semantic contribution of the syntac-
tic form of the sentence which expresses the relevant proposition.
So far, so good. But this view raises some further questions. Exactly how should we
think about the semantic contribution of syntax? And what does this tell us about what
propositions are?
Simplifying a bit, King answers these questions as follows. He suggests that we think
of the semantic significance (in English) of the relation between “Amelia” and “talks”
in “Amelia talks” as the following instantiation function from objects, properties, and
worlds to truth values: the function which, given as argument an object o and property
F, determines the truth value true at w iff o instantiates F at w. He then embeds this
view of the semantic significance of syntax in a view that propositions are a certain
kind of fact. We can, in King’s view, describe the proposition expressed by “Amelia
talks” as follows: it is the fact of there being words x and y of some language such that
x has Amelia as its content, y has the property of talking as its content, R(x,y), and R
encodes the instantiation function.8
Though I think that this view has many attractive features, it’s here that King and
I part company. So long as we stick with the view that propositions are a kind of fact,
then it is hard to avoid the view that the relevant facts are in part about the linguis-
tic items that express propositions. But I think that a simpler view, which avoids the
detour through existential quantification over linguistic expressions, is available if we
think of propositions as properties, rather than as facts.
Consider again the example of “Amelia talks.” If we think of the semantic content of
this sentence as a property, one natural view is that the property is the property of being
such that Amelia talks. On this kind of view, what is contributed by the syntax of a sim-
ple predication—the semantic significance (in English) of this bit of syntax, in King’s
terms—is something like the three-place relation corresponding to the open sentence
“__ is such that __ instantiates __.” In the case of the sentence “Amelia talks,” the con-
tents of the name and predicate fill in the second two slots to deliver the monadic prop-
erty expressed by “__ is such that Amelia instantiates the property of talking.”
This view accommodates the two pieces of data about the relationship between
propositions and their constituents mentioned above. As on King’s view, the differ-
ence between “John loves Jane” and “John loving Jane” is explained by the fact that the
syntactic form of the former has, in English, a semantic content, whereas the syntactic
form of the latter does not. And we get an explanation of the distinctness of the propo-
sitions that John loves Jane and that Jane loves John in terms of the distinctness of the
properties of being such that John loves Jane and being such that Jane loves John.9
Though the view that propositions are a sort of property may sound odd at first, it
actually fits rather naturally much of our talk about propositions. We might say that
believing a proposition, for example, is taking the world to be a certain way. But if, as
8
See, e.g., King (2007), 37.
9
One might object that this is not much of an explanation; isn’t this just a relabeling of the fact which was
supposed to need explanation? I don’t think so. Anyone who believes in the existence of complex properties,
like the property of being such that John loves Jane, must already accept the fact that the property of being
such that John loves Jane is distinct from the property of being such that Jane loves John. It’s an explanatory
gain if we can show that the distinctness of the propositions that John loves Jane and that Jane loves John is
nothing over and above this fact about properties. Maybe this fact about the identity conditions of properties
is itself deeply mysterious, and in need of explanation; but even if this is so, it is so whether or not we identify
propositions with properties—and it is better to have one mystery than two.
it seems, “ways things are” are properties, this indicates that having a belief is taking a
certain attitude toward a property. Parallel points might be made about mental states
other than belief, and speech acts like assertion; we can hope, intend, and desire that
the world be a certain way; and we can say, assert, and suggest that the world is a certain
way. Again, if ways things can be are properties, this suggests that the objects of mental
states and speech acts—i.e., propositions—are properties.
But why hold this view about the correct order of explanation? We can all agree—mod-
uloworries about the existence of the property of being such that Amelia talks—that
I instantiate this property iff I exist and the proposition that Amelia talks is true. Taking
the left hand side of this biconditional as basic has the advantage that we need postu-
late no new category of entities, in addition to properties. And it is hard to see what
advantage taking the right hand side as basic could have; after all, any analysis of what
it is for the proposition that Amelia talks to be true will equally serve as an analysis of
what it is for any existent to instantiate the property of being such that Amelia talks.
There are several different ways in which this view of propositional truth can be gen-
eralized to an account of truth with respect to a world (or arbitrary circumstance of
evaluation), but the simplest is as follows. Propositions are properties which are true iff
they are instantiated. Propositions are true with respect to a world w iff, were w actual,
that property would be instantiated—or, equivalently, iff, were w actual, the proposi-
tion would be true.10 Given this view of truth at a world, thinking of propositions as
10
There are, however, reasons not to go for this simple generalization, especially if one thinks that (a) prop-
ositions can’t be true without existing, (b) singular propositions can’t exist unless their constituents do, and
(c) some singular propositions have contingently existing objects as constituents. I discuss these issues in
Speaks (2012).
properties does not seem to require any serious revision in the way that we think about
entailment relations between propositions, or semantics more generally. Propositions
are necessary iff they are true with respect to every possible world; just so, on the
present account, the propositions are necessary iff the properties which they are, are
instantiated in every possible world. One proposition F would entail another proposi-
tion G iff any world in which F is instantiated is also a world in which G is instantiated.11
I mentioned at the outset that views of propositions which assimilate propositions
to members of an ontological category—in this case, properties—in which we have
independent reason to believe, enjoy the advantage of (relative) ontological parsi-
mony. While this is, I think, a genuine advantage, it is worth noting that—given the
account of truth, and truth at a world just sketched—it comes with a significant string
attached. If propositions are properties which are true iff they are instantiated, then it
seems clear that we must think that there are uninstantiated properties, and indeed
properties which could not be instantiated. Otherwise, there would be no account of
the propositions expressed by false, and necessarily false, sentences.12
Even if this view of propositions does make available clear explanations of truth and
truth at a world, there is one aspect of the traditional theory of propositions which it
does not capture. This is the view that, as Scott Soames puts it, propositional attitudes
are representational “because of their relations to inherently representational proposi-
tions.”13 Properties like the property of being such that Amelia talks are not inherently
11
An anonymous reader suggested the following alternative view:
A proposition is necessary iff every actually existing thing necessarily instantiates it.
The idea is that this better captures the intuitive connection between truth and necessary truth.
I doubt that any pre-theoretic intuition favors this view and, in any case, it (plus the assumption that some
actually existing things could have failed to exist) entails the falsity of the plausible principle (often called
“Serious Actualism”) that, necessarily, if an object instantiates a property, that object exists. One might try
instead
A proposition is necessary iff every actually existing thing instantiates it in every world
where it exists.
but this is equivalent to the view defended above, given the premise that some actually existing things exist
necessarily.
A different view in the neighborhood of the view that I defend (and which I used to hold) is the view that
properties are properties of worlds, like the property corresponding to the open sentence
Were w actual, it would be the case that Amelia talks.
On this sort of view, truth of course can’t be identified with instantiation, on pain of making possible truth
entail truth. On this sort of view, a proposition is necessarily true iff it is instantiated by every world. I don’t
have the space here to discuss my reasons for favoring the present version of the property view over the
“properties of worlds” version.
12
Though it is worth noting that the view is not committed to the existence of simple necessarily uninstan-
tiated properties, which may well seem more troubling than complex necessarily uninstantiated properties.
A view of properties which I think would suit my purposes is outlined in van Inwagen (2004). But I think
that the view of propositions I am developing would be consistent with various views of what properties are,
so long as those views are not committed to any sort of principle of instantiation (or possible instantiation).
13
Soames (forthcoming).
representational things; hence if propositions are properties of this sort, this aspect of
the view of propositions common to Frege and (the early) Russell must be rejected.
This may seem like a cost; but there is also a benefit here. The idea that an entity can
be intrinsically representational has seemed to be a puzzling one to many. If we can
give an account of truth and propositional attitudes (about which more below) with-
out making use of entities of this sort, this is a good thing.
Propositional attitudes
My sketch of the view that propositions are properties has been, in one respect, quite
different than the way this topic is discussed in, for example, Chisholm and Lewis.
Both Chisholm and Lewis approach the topic not by asking what sorts of things are
expressed by sentences, but rather by asking what sorts of things are the objects of
propositional attitudes.
This is more than a difference in order of exposition. Both Chisholm and Lewis—
as well as contemporary defenders of their view14—advance the view that the objects
of the attitudes are properties without identifying propositions with properties. This
view can sound paradoxical, since it sounds like the denial that propositional attitudes
are relations to propositions. But this is a superficial objection; there’s nothing incoher-
ent in the idea that one sort of a thing, a proposition, is expressed by sentences, while
another sort of thing, a property, is the content of mental states. In this respect, the
Chisholm/Lewis view about properties is less ambitious than the view that I’m defend-
ing; they think that properties can play some of the roles standardly assigned to propo-
sitions, whereas I think that properties can play all of the roles—including, crucially,
the role of being the things expressed by sentences.
There are some general reasons to prefer the “pure” view that properties can play all
of the roles assigned to propositions over a “mixed” view of the sort defended by Lewis
and Chisholm, according to which properties are the objects of the attitudes, and some
other sort of thing is what is expressed by sentences.15
One sort of reason for preferring the pure property theory over a mixed theory is
based on the sort of ordinary language considerations often used to introduce talk
about propositions in the first place. The mixed view is forced, for example, to deny
face value readings of apparently true bits of ordinary language discourse like:
That sentence expresses my belief perfectly.
14
See especially Feit (2008, 2010) and Turner (2010).
15
Of course, Chisholm and Lewis had different views about the nature of the propositions expressed by
sentences—it’s just that they agreed that these entities were distinct from the properties that were the objects
of the attitudes.
which seems to be inconsistent with the mixed view that propositions are expressed by
sentences, properties are the objects of belief, and propositions ≠ properties.16
A second reason for preferring the pure property theory is simply that the mixed
theory leaves us without a theory of propositions. This is, obviously, not a pressing
objection if one feels able to give an independent account of propositions, or if one is
happy to take them as a primitive category in one’s ontology. But if one is (perhaps for
the reasons sketched at the outset) unhappy with non-reductive views, and (perhaps
for reasons like those given in chapter 3 above) dissatisfied with the reductive alterna-
tives to the view that propositions are properties, these considerations will push one
towards a pure version of the property theory of propositions.
Even if a pure theory is to be preferred, and even if Lewis and Chisholm each went
for a mixed theory, that doesn’t show that we can’t learn from their treatment of the
attitudes, since there is no reason (for all we’ve said) why one couldn’t integrate their
theory of the attitudes into a pure theory. So let’s turn to the Lewis/Chisholm theory of
propositional attitudes.
If the objects of the attitudes are properties, then it is natural to think that belief must
be believing of something that it is a certain way; supposing must be supposing that
something is a certain way; guessing must be guessing that something is a certain way;
and so on. A basic question for defenders of this view of the attitudes is: what is this
“something”?
Lewis and Chisholm gave basically the same answer to this question: in their view,
propositional attitudes are ascriptions of properties to oneself. Chisholm expressed the
theory like this:
“Believing must be construed as a relation between a believer and some other thing . . . What
kind of thing, then? . . . The simplest conception, I suggest, is one which construes believing as
a relation between a believer and a property—a property which he may be said to attribute to
himself.”17
It is understandable why the view was introduced in this way, since its principal initial
motivation was the explanation of the distinction between first-personal beliefs and
third-personal beliefs about oneself—or, as Chisholm put it, between the emphatic
and non-emphatic reflexive.
There seems to be a distinction, to use one of many standard examples, between my
believing of myself that I am on fire, and my believing that Jeff Speaks is on fire. It seems
that I might have either belief without the other, if I am sufficiently confused about
my identity. And this point is not limited to belief; parallel remarks could be made
16
See also the discussion of the “proposition sentences” in Ch. 2 above.
17
Chisholm (1981), 27. For a similar statement of the view, see Lewis (1979).
On the property theory, what A believes—the content of A’s belief—is a property. This
in itself need not be problematic if we adopt the account of what it is for a property
to be true sketched in the preceding section. But consider what sort of properties are
assigned as the contents of beliefs in the key case of first personal belief. In the case of
the first-personal belief that one is on fire, the Lewis/Chisholm theory assigns as the
content of the belief the property of being on fire. What would it take for this property
to be true?
We can’t say that this property is true iff it is instantiated, since (given the above plat-
itude) this would entail that my belief that I am on fire can be made true by someone
other than me being on fire. And we can’t require for its truth that everyone instantiate
the property, since the truth of my belief doesn’t require that everyone be in flames.
And we can’t require that I instantiate the property, since that would make the truth
of others’ first-personal beliefs that they are on fire hostage to my temperature rather
than their own. We could, of course, get round this sort of problem by letting the prop-
erty self-ascribed in the case of my first-personal belief that I am on fire be the property
of being such that Jeff Speaks is on fire, since then we could say that what the subject
believes is true iff the property is instantiated—but this would remove the wanted con-
trast between first- and third-personal beliefs, and hence sacrifice the motivation for
18
The idea is that we can analyze beliefs about what is happening here as beliefs about what is in the same
place as me, and beliefs about what is happening now as beliefs about what is happening at the same time
as the time at which I exist—though the latter analysis requires us to think of the subjects of the attitudes as
time-slices of persons. See the discussion of the example of the insomniac in Lewis (1979), §VII.
the view that propositional attitudes are self-ascriptions of properties.19 The only plausible
move here seems to be to abandon the platitude, and say that talk about the truth of a
subject’s belief comes apart from talk about the truth of what the subject believes; but this,
I think, is hard to accept.
A second sort of problem for the Lewis/Chisholm theory—which has been developed
by Markie (1988) and Nolan (2006)—stems from the observation that, while in many cases
nothing is lost by thinking of a propositional attitude with the content p as a belief that I am
such that p is the case, in other cases, understanding the content of the belief requires that
we consider worlds where p is the case, but in which I am not such that p is the case, because
I do not exist. The most striking case is perhaps the example of the desire that I not exist. On
the Chisholm/Lewis theory, this is interpreted as the desire that I instantiate the property of
nonexistence. But this seems wrong, since it is impossible that I have the property of nonex-
istence, and the desire that I not exist is not a desire for something impossible.20
How might the defender of the Chisholm/Lewis theory of the attitudes reply to this crit-
icism? One interesting suggestion, defended in different ways in Feit (2010) and Turner
(2010), is that we can defuse this objection by appealing to something like the distinc-
tion, employed in a different context by Kit Fine, between truth in a world and truth at a
world—where the latter does not require existence at that world.21 If we accept this distinc-
tion, then we might say that my desire that I not exist is true at w iff I can instantiate the
property of not existing at w iff the proposition that I don’t exist is true at (not in) w.
I’m sympathetic to Fine’s distinction and to the idea that a defender of the idea that
properties are the objects of the attitudes might use this distinction to avoid the result
that the desire that I not exist be a desire for something impossible. But this move
comes at a cost. Instantiating a property at a world is defined in terms of the truth of
propositions at a world; hence it seems that anyone who appeals to this notion of hav-
ing a property at a world must adopt what above I called a “mixed” theory, according
to which, while properties are the objects of the attitudes, propositions still exist as
members of some other ontological category. For the reasons given above, I think that
it would be better to avoid this result.22
19
Of course, the proponent of this view might try to distinguish the properties of being such that Jeff
Speaks is on fire and the property of being such that I am on fire. But then all the work is being done by this
distinction between first-personal and third-personal properties, and none by the idea that the attitudes are
self-ascriptions of properties.
20
Here I’m setting aside the view of Williamson (2001) that everything which exists does so necessarily.
21
See, among other places, Fine (1985).
22
This needn’t be the end of the story. One might adapt the essentials of Turner’s response to Nolan and
then try to give a theory of what it is for an object to have a property at a world which did not go via some
proposition’s being true at that world. I think that it is plausible that this could be done for the case of the
desire that I not exist. But, as Turner recognizes, a full solution to Nolan-type problems requires an account
not just of overtly first personal mental states, but also for desires like the desire that there be world peace—
since we don’t want this to entail a preference, ceteris paribus, for worlds in which I exist and there is world
peace over worlds where there is world peace without me. And though it is somewhat plausible that I can
have the property of nonexistence at a world without existing in that world, it seems less plausible that this is
so for every property, and, in particular, it seems implausible that I can instantiate the property of being such
that there is world peace at a world without existing in that world.
Finally, the foregoing objections aside, I think that the idea that all of my mental
states involve attributions of properties to myself should strike us as odd. Imagine
someone engaged in purely abstract reasoning—say, a mathematician. It’s very coun-
terintuitive to think of the mathematician as trying to figure out what she is like—it
seems like what she cares about is not which properties she has, but what the world is
like, or what numbers are like. It should be possible for thinking to be less self-involving
than the Lewis/Chisholm view makes it out to be.23
Fortunately, there is no reason why the view that propositions are properties must be
tied to the view that all thought is self-ascription. But if we reject this view, this leaves
us with the question raised earlier this section: given that belief is believing something
to be a certain way, if this “something” is not oneself, then what is it?
There is a sense, I think, in which this question has no answer. To believe that Amelia
talks, there is no special object which one must believe to be such that Amelia talks; it
is enough if one takes there to be something which is such that Amelia talks. Of course,
often one takes a property to be instantiated by something because one takes it to be
instantiated by something in particular, and this is as true of the property of being
such that Amelia talks as it is of other properties. One might, for example, believe the
proposition that Amelia talks by taking the world to be such that Amelia talks; this
fits nicely with the intuitive idea that in, for example, the case of belief, one “takes the
world to be a certain way.”
One might, however, wonder whether Nolan’s objection to the Chisholm/Lewis
theory of the attitudes can be resurrected as an objection to the present view. Just as
one can desire that one never have existed, can’t one desire that the world never have
existed, or that there should have been nothing rather than something? This seems to
be a coherent desire; but on the present interpretation, this would be the desire that
the property of there being nothing be instantiated—which, given that a property is
instantiated only if it is instantiated by something, is impossible.
However, I think that this objection can be defused by distinguishing two interpre-
tations of the desire that there be nothing rather than something. On the first inter-
pretation, the desire that there be nothing rather than something is something like
the desire that there be no concrete things, or no material things. But (presuming that
it is possible that there be no concrete things, or no material things) the properties
assigned as contents to this desire are possibly instantiated, and so the desire comes
out—as it should, on this interpretation—as a desire for something possible. After
all, in a world in which there are no concrete things, there will be things—necessarily
existing abstract objects, for example—which instantiate the property of being such
that there are no concrete things.
23
Though see Nolan (2006), 678–679, for some interesting discussion of how the view that all thought is
self-involving in the relevant sense seems to follow from Lewis’ views about modality.
On the second interpretation, the desire that there be nothing rather than some-
thing really is the desire that there be nothing—no concrete things, no abstract things,
no things of any sort. It is true that, on the present view, the content of this desire
turns out to be a necessary falsehood (if we assume, as I do, that a property cannot
be instantiated unless something instantiates it). But this consequence does not seem
to be objectionable, since it seems independently plausible that some things—like at
least some abstract objects—exist necessarily.24 So on neither interpretation of Nolan’s
problematic desire does it pose a problem for the view that propositions are properties
which, if instantiated, are instantiated by everything.
Now it should be noted that the fact that my view avoids this sort of puzzle for the
Chisholm/Lewis view—as well as the other objections raised against that view—comes
at a cost. That theory was explicitly developed to explain the distinction between
first-personal and third-personal beliefs and, more generally, the distinction between
indexical and non-indexical beliefs. And, as it stands, the view of the attitudes just
sketched offers no explanation of this distinction. In the next section I’ll return to the
question of whether the sort of “property theory” I’ve been developing can capture
some of the explanatory advantages of the Chisholm/Lewis theory.
should not express binary relations between subjects and propositions, but rather ter-
nary relations, and hence to be of the form
A believes of o that it is F.
where “F” stands for the property expressed by “S” in the context. However, there is no
plausible candidate for the value of “o,” for two reasons. First, (as noted above) there is
nothing to stop two different subjects from each believing that Amelia talks, but to do
so by believing the property of being such that Amelia talks to hold of distinct things.
Second, we would get into trouble with the modal profiles of attitude ascriptions if we
supplied as value for “o” anything whose nonexistence was consistent with the truth of
the ascription.
24
Though, even if it is not objectionable that the desire that there be nothing come out as impossible on
this second interpretation, there still may be some oddness in treating it as the desire that a certain property
be instantiated. But this oddness seems to be shared with the traditional view of desire as a propositional
attitude, since on such a view the desire that there be nothing will be equivalent to the desire that a certain
proposition be true.
We can get around this problem by analyzing attitude ascriptions as existential gen-
eralizations of the form
ූx A believes of x that it is F.
which are true with respect to a world w iff the referent of “A” in w believes of some object
in w that it is F (where F is the proposition expressed by the complement of the ascrip-
tion). This sort of view also stays a bit closer to the standard semantics for belief ascrip-
tions, in that it does not take them to predicate a ternary relation of a subject, a world, and
a proposition, but rather a binary relation between a subject and a proposition—albeit a
binary relation defined by existential generalization on a ternary relation.25
But this view also faces a serious problem. It appears that, given the truth of an
ascription
A believes that S
The problem is that (setting aside the, at present, irrelevant point that languages don’t
contain singular terms for every object) such examples of quantifying into attitude
ascriptions seem to entail, for some singular term n, the truth of
A believes that n is such that S.
But then, if we re-apply our analysis of ordinary attitude ascriptions, we find that this
entails the truth of
ූx A believes of x that it is such that n is such that S.
which in turn, for some singular term n*, entails the truth of
and we’re off on a regress, the upshot of which is the implausible conclusion that hav-
ing a single belief entails having infinitely many beliefs of ever-increasing complexity.26
25
In this respect, the view of belief is similar to that defended in Salmon (1986).
26
Alexis Burgess suggested to me the possibility that this regress could be blocked by taking note of the fact that
the initial existential generalization is a sentence of the meta-language, and that when we “re-apply our analysis”
we’ve ascended to the meta-meta-language—and from the truth of some sentence of the meta-meta-language
we are not automatically licensed to infer the truth of the corresponding sentence of the object language. There
may well be something to this idea—and it would be very nice, for my purposes, to have a response to this regress
argument. But there’s an intuitive worry about this strategy. The sentences of the meta- (and meta-meta-) lan-
guage will, like the sentences of the object language, presumably express propositions. So, for example, some
proposition P will be expressed by the sentence “ූx A believes of x that it is such that n is such that S” of the
meta-meta-language. And some proposition Q will be expressed by just this string of words in the object lan-
guage. The problem, it seems to me, is that it looks quite plausible that P will entail Q. But, if it does, then the
regress argument goes through—even if we pay attention to the distinction between object and meta-language.
27
We could solve this problem by adopting a sufficiently coarse-grained view of properties to block the
claim that the beliefs ascribed at various stages in the regress are genuinely distinct. But this would plug one
hole in the theory only to introduce problems elsewhere, since we’d then face all the problems with belief
ascriptions faced by coarse-grained theories of propositions. See, for discussion, Soames (1988) and Ch.
3 above.
28
Though this is just an informal gloss; the idea is not that belief is to be analyzed as a three-place relation
between subjects, properties, and the property of being instantiated, since this would lead to the same sort of
regress just discussed.
29
One might think that this is overkill: once we have the difference in attitudes, why also distinguish
between their contents? Why not, in other words, also let the content of the first-personal state be the
This does not, by itself, solve parallel problems about indexicality which arise at the
level of propositional attitude ascriptions—in particular, it does not explain why it
seems to many competent speakers that, out of my mouth, “I believe that Jeff Speaks
is F” can differ in truth value from “I believe that I am F.” But it might still help. One
popular style of explanation of these speaker intuitions explains them away in terms of
a confusion between the truth of the semantic content of a sentence and the truth of
some proposition which would, in the relevant context, be pragmatically conveyed by
an utterance of that sentence. Proponents of this strategy owe some account of the rel-
evant pragmatically conveyed propositions, as well as an account of the mechanism by
which those propositions are pragmatically conveyed by the relevant utterances. If we
recognize the existence of the attitude of self-predication, then this might give us the
beginnings of such an account. Perhaps when I utter an ascription I believe that I am
F, I pragmatically convey the proposition that JS self-attributes the property denoted
by “F.”30 In the relevant cases, the truth-value of this proposition will come apart from
the truth-value of the proposition that JS believes-to-be-instantiated the property of
being such that JS is F.
One might object that this supposed explanatory advantage is one shared with any
reasonable theory, since any theory—whatever it says about the nature of proposi-
tions—can recognize the existence of the relation between subjects and properties
which I’m calling “self-attribution.” But, to do this, proponents of some other view of
propositions—whether they defend a non-reductive view, or aim to reduce proposi-
tions to an ontological category other than properties—must introduce a distinction
between different sorts of contentful mental states, treating some as relations to propo-
sitions, and others as relations to properties. Echoing Lewis, I “protest that the advan-
tages of uniform objects are not to be lightly forsaken.”31 The fact that the property
theory can recognize the existence of the attitude of self-attribution, while preserving
a consistent treatment of contentful mental states as relations to properties, is a point
in its favor.
property of being such that JS is on fire? The reason is simply that I might self-attribute the property of being
such that JS is on fire without, intuitively, taking myself to be on fire, if I’m sufficiently confused about my
identity. Thanks to an anonymous reader for pressing this concern.
30
One might also take this to be part of the semantics of attributions of first-personal beliefs; I’m skepti-
cal about this sort of view, since it would treat “believes” as ambiguous, which would lead to problems with
capturing certain valid inferences from pairs of first-personal and third-personal ascriptions. The problems
here are analogous to the problems for Quine’s view that notional and relational senses of “believes” should
be distinguished which are pointed out in Salmon (1995).
31
Lewis (1979), 532.
But the conclusion seems barely coherent, let alone entailed by the premise.
One might think that this is a place in which the present theory of propositions can
come to its own rescue. After all, given that loving is the relation denoted by “loves,”
one might have thought that the following should be valid:
A loves B.
A loving B.
But this is not valid, since the conclusion fails to express a proposition—and this fact
(as noted above) is a fact which the view that propositions are properties promises to
explain. Might we offer a similar explanation of the invalidity of the first argument in
terms of the fact that “Bob believes the property of being such that Amelia talks” fails
to express a proposition?
Perhaps, but it is not quite this easy. We can’t simply say that
Bob believes the property of being such that Amelia talks.
are ungrammatical since, in general, they aren’t—we can meaningfully and truly say
Bob believes the proposition expressed by “Amelia talks.”
This suggests that the conclusion of the problematic argument above is not ungram-
matical, but simply false.
This places a constraint on the sort of semantics for attitude ascriptions which the
defender of the present view—who will have a hard time denying that “that Amelia
talks” and “the property of being such that Amelia talks” refer to the same thing—can
accept. In particular, it looks like the proponent of the view of propositions that I have
been sketching will have to deny the following principle:
If an attitude ascription A V’s xis true—where A is the name of the subject of the ascription, V
is the attitude verb, and x is some term which refers to a proposition—then any other well-formed
ascription which differs from this only by the replacement of x with another term for the same prop-
osition must also be true.
It can’t be denied that this principle has a great deal of initial plausibility. However, there is
some independent reason to deny it.32 Consider, for example, the pair of sentences
Joe fears that the Mets will win the World Series this year.
Joe fears the proposition that the Mets will win the World Series this year.
These are both grammatical, but the first is true, and the second is false—Joe may be afraid
of many things, but propositions are not among them. And this is so despite the fact that,
on any view, “that the Mets will win the World Series this year” and “the proposition that
the Mets will win the World Series this year” refer to the same proposition.
The proponent of the view of propositions as properties might seize on examples like
this, and say that whatever explains the fact that these sentences about Joe’s fears differ in
truth-value can also explain the fact that our initial pair of sentences,
Bob believes that Amelia talks.
Bob believes the property of being such that Amelia talks.
can differ in truth-value, despite the fact that “that Amelia talks” and “the property of
being such that Amelia talks” both designate the same thing. To be sure, it is not obvious
that this line of response is satisfactory, since it is not obvious that the explanation of the
“fears” substitution failures will carry over to the example we are interested in; but the two
sorts of examples do seem similar, so it is perhaps not unreasonable to hold that the right
treatment of the “fears” examples will solve our own problem.33
32
For a much more in-depth discussion of these issues, see King (2007), chapter 5.
33
One plausible explanation of the examples involving “fears,” which is defended in King (2007) (153–163)
is that some attitude verbs, including “fears,” are ambiguous. This will only help with our “believes” example
if “believes” is also ambiguous; but it may not be implausible to say that it is. Consider, for example:
Joe believes that the Reds will win the World Series this year.
Joe believes Dusty.
If we take these sentences at face value, then it looks like “believes” expresses different (though obviously
related) relations in these two sentences. Perhaps the problem with our troublesome pair of “belief ” sen-
tences is that
Bob believes the property of being such that Amelia talks.
forces the interpretation of “believes” exemplified by “Joe believes Dusty.”
This is, obviously, a view on which “believes” is ambiguous (though the two meanings are of course related).
But it is nowhere near a systematic account of this ambiguity. A systematic account would have to explain
why the first interpretation of “believes” is triggered not just by belief ascriptions involving that-clauses, but
also by sentences like the following:
Bob believes what Amelia said.
Bob believes the proposition expressed by what Amelia said.
Bob believes the first thing he hears every day.
and so on. For an interesting discussion of some alternatives to this view that “believes” is ambiguous in this
way, see King (2007), p. 157, note 39.
A demarcation problem
On the present view, propositions are properties. But are all properties propositions?
Presumably not. Redness is a property, but, unlike the property of being such that
Amelia talks, not a proposition. This view about the relationship between propositions
and properties has, of course, many analogues in philosophy. Proponents of the view
that mental states are brain states don’t for that reason think that every brain state is a
mental state; and proponents of the view that propositions are functions—e.g. from
worlds to truth-values—don’t think that addition, just because it is a function, must
also be a proposition.
But this leads to a question for the proponent of the view that propositions are prop-
erties: if not all properties are propositions, what distinguishes the ones that are from
the ones that are not?
A first answer which suggests itself is that the properties which are propositions are
the ones which are such that, if they are instantiated at all, are instantiated by every-
thing.34 But a moment’s thought shows that this won’t work, since the property of exist-
ence (if such there be) and the property of being self-identical are, if instantiated at all,
instantiated by everything, and yet these properties don’t seem to be propositions.
One might reply by pointing out a difference between the property of being such
that Amelia talks and the property of self-identity: the latter, but not the former, is such
that it is necessarily instantiated. So maybe we could say that propositions are such that
(i) if they are instantiated at all, are instantiated by everything, and (ii) they are possibly
not instantiated. The obvious objection, though, is that though (i) and (ii) may work
for the property of being such that Amelia talks, they won’t work for the property of
being such that 2 + 2 = 4, since this, like the property of being self-identical, satisfies
(i) but not (ii).
A quite different approach would be to demarcate the propositions not using an
intrinsic criterion, but an extrinsic one, e.g. by saying that a certain property is a prop-
osition iff it can be expressed by a sentence. But this looks similarly hopeless, partly
because it is not obvious that every proposition is possibly expressed by a sentence, and
partly because it seems to get the order of explanation backwards. Why, one wonders,
is “My pants are on fire” a sentence, whereas “Pants” is not? Presumably because the
former expresses a proposition, and the latter does not.35
34
This is parallel to the way that Chisholm demarcates propositions from states of affairs; see Chisholm
(1976), 122–124.
35
For an attempt to explain the distinction between sentences and non-sentences which does not appeal to
propositions, see Davidson (2005).
Other “extrinsic criteria” might be more plausible. One might (as Alexis Burgess suggested to me) try to
explain which properties are propositions in terms of certain, perhaps modal, relations holding between the
instantiations of properties. For the use of a strategy like this in demarcating the laws of nature, see Lange
(2009).
So, while it would be very nice to have a demarcation criterion, it’s not at all easy to
see how to state one.36 But, while this is a vice, it is not a vice unique to the view that
propositions are a kind of property.
Proponents of the view that propositions are a sui generis category of entities also owe
a statement of a demarcation criterion—and, typically, fail to provide one. Suppose we
ask a proponent of that view why that grass is green is a proposition while being green is
not. What should they say? They might say: that grass is green belongs to the sui generis
category of propositions, whereas being green does not. But is this any better than the
proponent of the property theory of propositions taking the distinction between those
properties which are propositions, and those which aren’t, as primitive?37
And, in fact, lots of views which identify the F’s with some subset of the G’s do so
without identifying a criterion for distinguishing the G’s which are F’s from the G’s
which aren’t—a case in point is the view that mental properties are a subset of the phys-
ical properties. Proponents of that view typically don’t also provide an explicit criterion
to divide the physical properties which are also mental properties from those which
aren’t. If this is no decisive objection to that view, it’s not easy to see why it should be a
decisive objection to the view that propositions are properties—even if this is some-
thing we should want a fully satisfactory theory of propositions to provide.38
36
This would be one reason to prefer van Inwagen’s version of the property theory over mine. van Inwagen
takes propositions to be 0-place properties; this provides a neat way of distinguishing the propositions that
are properties from those that are not. Despite this, I prefer my version because I don’t understand what a
0-place property could be.
37
Friends of the view that propositions are sui generis entities might try to explain this distinction in terms
of the fact that that grass is green is true or false, whereas being green is not—an explanation of which I can-
not avail myself, since I take truth for propositions to be instantiation, and these properties are, equally,
instantiated. But this move does not seem plausible since it involves taking truth as explanatorily prior to
propositions—see for discussion Ch. 3 above.
38
Thanks to Alexis Burgess, Ben Caplan, Lorraine Juliano Keller, Matthew Lee, and Juhani Yli-Vakkuri for
comments on previous versions of this essay, and to the participants in my graduate seminars at Notre Dame
in the spring of 2008 and fall of 2009 for discussion of issues surrounding the metaphysics of propositions.
1
The conception in this chapter elaborates, extends, and modifies the one introduced in Soames, What is
Meaning, Princeton and Oxford: Princeton University Press, 2010.
yield the propositions the sentence is used to assert or express in different contexts.
Sometimes little or no supplementary information is required to reach these proposi-
tions. Sometimes, no proposition is expressed without it. Sorting out the contributions
made by conventionally-encoded versus contextually-varying information to the
propositions asserted or communicated by a use of a sentence is the business of seman-
tics and pragmatics. A rough and ready criterion distinguishing the two is this: the
meaning of an expression is information about it that an ideally rational agent would
have to know in advance—independent of the varying information available in dif-
ferent contexts—in order to reliably grasp what uses of sentences containing it assert,
express, or convey.
Since propositions are pieces of information that are asserted or expressed, they are
the primary bearers of truth conditions, which are inherited by sentences or utterances
that express them. As explained in chapter 3, the fundamental connection between
truth, meaning, and propositions is expressed by the simple schema (1) (where “S” is
a metalinguistic variable over sentences, and “P” is a schematic letter replaceable by
sentences).
1. If a sentence S of L means that P, then S is true iff P.
Roughly put, for S to mean that P is for S to express the proposition that P, and for S to
be true is for the proposition S expresses to be true. Hence, the a priori obviousness of
instances of (1) is reducible to that of the corresponding instances of (2).
2. The proposition that P is true iff P.
Different sentences can, of course, express the same propositions, which may be
assumed, asserted, known, or believed. What are propositions, and what is the rela-
tion of expressing that sentences or utterances bear to them? A naturalistic account
must avoid characterizing propositions as inhabitants of a Platonic third realm beyond
mind and matter, with no explanation of how we come to bear attitudes to them, as well
as how we are acquainted with, and come to know things about, them. It would be nice
to be able to regard talk of the propositions expressed by sentences as simply talk about
sentences at a level of abstraction that groups them together in terms of their repre-
sentational features. However, since propositions have a life beyond language, this
can’t be the whole story. One challenge facing any plausible theory of propositions is to
make good on their independence from language without turning them into eternal,
unchanging, representational contents that somehow become attached to sentences by
an otherworldly expressing relation. This is one of the challenges I will address.
These general remarks about the role of propositions in theories of language hold
whether or not the language under investigation contains singular terms referring to
propositions, quantifiers ranging over them, operators operating upon them, or predi-
cates taking them as arguments. Whether or not a language contains such expressions,
propositions are needed to explain the uses to which speakers of the language put it.
In other words, propositions are needed to state the goals of semantic and pragmatic
theories for any language. Of course, when natural languages like English are involved,
they are also needed within semantics proper as referents of that-clauses, arguments
of attitude verbs, referents of some names and uses of indexicals, members of the
domains of some quantifiers, and so on.
Propositions are also crucial to cognitive theories. To think about something is
to think about it as being a certain way. So propositions, which represent things as
being one way or another, are the contents of many cognitive states—such as one’s
belief, doubt, or uncertainty that the economy will recover soon. Each of these atti-
tudes is a relation in which an agent stands to a proposition. Although the picture
generalizes, it is complicated by syntactic variation among verbs designating rela-
tions to propositional objects expressed by their complement clauses. One axis of
variation divides verbs according to whether these clauses are finite (tensed) or
non-finite (infinitival). Further variation among verbs taking finite clauses separates
those that also take complex noun-phrase objects such as the proposition/claim/
statement that the Earth is round, and the proposition/claim/statement that Martin
asserted. Attitude verbs that take both finite clauses and complex nominal objects
include assert, believe, know, deny, accept, reject, doubt, assume, refute, prove, estab-
lish; verbs that take finite clauses but not complex nominal objects include say, think,
judge, see, perceive, desire, hope, expect, anticipate, suppose, hypothesize, imagine;
verbs that take finite clauses plus a few restricted complex nominal objects—but
not the proposition/claim that S—include predict (that/the result that), regret, real-
ize (that/the fact that); verbs that take both finite and non-finite clauses include
believe, expect, assume, suppose, imagine, desire, prefer, wish. There are also verbs
like want that take only non-finite clauses. Despite this variation, it is plausible to
suppose that these verbs have readings in which they express cognitive relations that
hold between agents and propositions. If so, each should correspond to a cognitive
state-type with propositional content.
Propositions are also central to our understanding of perception. Seeing and hear-
ing are relations between an agent and something else, often an object or event. These
perceptual states also represent the agent’s immediate environment, or things in the
environment, as being certain ways. Imagine seeing a poster on the wall as red—in
one case because it is red and the lighting is normal, and in the other case because it is
illuminated by light that makes it appear red, even though it is white. If one’s phenom-
enal experience in the two cases is the same, then one’s visual experience represents
the poster as red in both cases, even though it is actually red in only one. The fact that
one’s perception represents it in this way is independent of whether or not one forms
the perceptual belief that it is red—which one might do in either case, both cases, or
neither. So we can specify the content of the perception, and evaluate its veridicality,
whether or not the agent forms a belief with that content. The agent’s perception is
veridical—i.e. accurate or truthful—iff the poster is the way that perception represents
it to be—i.e. iff it is red. This suggests that although what one perceives is typically an
object or event, the content of one’s perception is a proposition or set of propositions
representing things as being certain ways. Since, perception, like cognition, is repre-
sentational, it is a bearer of propositional content.2
In this way, we come to see propositions as the common thread tying together lan-
guage, thought, and perception. Their essential feature is that they represent things as
being certain ways, and so have truth conditions, which allow them to serve as con-
tents not just of sentences, but also of thoughts and perceptions. Perceptual and cogni-
tive contents are also among the things we use language to think and talk about. The
fact that the same propositions can simultaneously function as linguistic, perceptual,
and cognitive contents provides us with a systematic way of doing this. In the simplest
case, we choose a sentence that expresses a proposition p that is part of the content of
the perceptual or cognitive state we wish to characterize. Using a complement clause
(e.g. a that clause in English) to designate p, we characterize the cognitive or perceptual
content as having certain properties and standing in certain relations, e.g. to agents. In
this way, we use sentences to express complex cognitive contents that represent other
cognitive or perceptual contents as satisfying various conditions.
As important as it is to recognize the commonality in linguistic, cognitive, and per-
ceptual content, it is also important not to overlook their differences. Propositions
are (or at least can be) bite-sized bits of information; they are the minimal units of
their representational type. Individual sentences are thus their natural vehicles. Visual
perception, the content of which is inherently holistic, stands at the other end of the
spectrum. Although individual propositions can be abstracted as constituents of
that content, the content of a visual state is most closely approximated by a network
of propositions the elements of which are counterfactually connected to one another.
Removing or changing one may be impossible without drastically modifying the rest,
and hence the overall picture. The contents of belief and other cognitive states stand
somewhere between atomistic language and holistic visual perception. Although
belief reports are typically atomistic, the propositions correctly said to be believed are
often parts of larger cognitive structures that exhibit some of the holism and counter-
factual interconnection exhibited by perceptual states. Finally, the atomistic character
of language—with propositions assigned to sentences (or utterances) one by one—is
balanced by the vast syntactic and semantic resources that language puts at our dis-
posal. As individual sentences become more abstract in content, and more syntacti-
cally and semantically complex, they come to express many propositions that cannot
be constituents of the contents of any perceptual state, and—due to our cognitive limi-
tations—that cannot be cognized by us except when presented linguistically. A good
account of propositions should make room for all of this.
2
A temporal element is needed for the content of a perceptual state to be a complete proposition -- typi-
cally the time at which the perceptual experience occurs. Since that moment isn’t seen it may not be strictly
part of the content of one’s (purely) visual experience. Thus, we may need a slightly broader conception in
which perceptual and cognitive experiences are temporally supplemented. Thanks to Francois Recanati for
raising this issue.
Before leaving perception, a further point should be noted. Perception, like cog-
nition, is, I think, a cognitive activity in which we do something that results in the
world being represented in one way or another. Think of Wittgenstein’s duck/rabbit
example, or of an Escher drawing of a complex geometric structure. In the former
case, a curved line can look either duck-shaped or rabbit-shaped. First we see it one
way, then another. Once we realize that it can be seen both ways, we may try, with
varying degrees of success, to move at will from a perception with one representa-
tional content to a perception with the other. The same is true when an Escher draw-
ing of a building with a set of stairs that appears at one moment to be descending, and
at another to be ascending. Similar experiences can be had in specially constructed
rooms designed to create perceptual anomalies. To see what is before us first one way,
and then the other, is to first predicate one property of what we see, and then to predi-
cate a different property of it. Somehow perception makes properties and relations
available to us to put together in different predicative patterns. How we see things—
the predication we make—is usually automatic, unconscious, and so better described
as a kind of cognitive operation than as a species of intentional action. But sometimes
our experience makes multiple properties or relations available for predicating of the
same things, either unconsciously or with a degree of conscious control—in which
case our predications occasionally qualify as intentional. Either way, the important
point is that putting together representational structures in perception and cognition
is always a cognitive operation of some kind.3 The simplest cases are those in which we
predicate properties or relations of things that are given to us in perception or cogni-
tion, and thereby entertain a simple proposition, like the proposition that o is red, or
that o1 is bigger than o2.
3
Thus, in what follows it should be understood that when I speak of “acts of predicating” properties of
things I am not assuming that such “acts” are intentional.
proposition represents and that the agent does too—are related. Presumably, one
is the basis for the other. Which is fundamental? Is the fact that the agent repre-
sents o as red explained by (a) the fact that the agent has a certain attitude to the
proposition that o is red, plus (b) the fact that the proposition—in and of itself
and without interpretation by us—intrinsically represents o as red? Or is it the
other way around? Is the fact that the proposition represents o as red explained by
the fact that for an agent to entertain it is for the agent to represent o as red? The
traditional answer is that propositional representation is primary and the agent’s
representation is to be explained in terms of it. In What is Meaning? I argued that
this approach is unsustainable. Here I will elaborate a theory based on the oppo-
site approach. The guiding ideas are (i) that the perceptual and cognitive activity
of agents is the conceptual basis of all representation and (ii) that propositions
are representational in virtue of the relations they bear to this representational
activity.
The key move is to define what propositions are in a way that makes the deri-
vation of their representational properties from the representational activities of
agents plausible. Think again about the proposition that o is red. It is the content of
an occurrent perceptual or cognitive state whenever the agent predicates being red
of o. Whenever an agent does this, a concrete event occurs, at a specific time and
place, in which the agent predicates this property of that object. This suggests that
the proposition that o is red is simply the minimal event type in which an arbitrary
agent predicates being red of o. This event-type is representational because every
conceivable instance of it is one in which an agent represents something as being
a certain way. What it represents is what is representationally common to all such
instances. Since every such instance is one in which o is represented as being red,
we speak, derivatively, of the proposition itself representing o as red. Since nothing
else is representationally common to all conceivable instances of the proposition,
representing o as being red exhausts its representational content. Otherwise put, to
entertain the proposition that o is red is to predicate redness of o, which is to do
something that results in an instance of the event type that the proposition is. The
representationality, and hence truth conditions, of the proposition are due to the
representational features of these possible instances. The proposition represents o as
red, and nothing further, because what it is to entertain it is simply to predicate red-
ness of o, and so to represent o as red.
From this we derive its truth conditions: the proposition is true iff whatever (namely
o) it represents to be a certain way is that way (red); it is false iff o isn’t red. These condi-
tions can be modalized. For every metaphysically or epistemically possible world-state
w the proposition that o is red is true at w iff at w, o is red—which is just to say that for
each such world-state, if it were instantiated, then the proposition that o is red would
be true iff o were red. Thus, we explain how it is that the proposition that o is red has its
truth conditions essentially.
In this way, we solve the real problem of the unity of the proposition that defeated
Frege and Russell—which, as I argued in chapter 3, also defeats (along with other
problems) the possible-worlds conception of propositions. As indicated there, the
problem is to explain how propositions manage to represent the world, and so
have truth conditions from which the truth conditions of sentences, utterances,
and cognitive states can be inherited. While traditional accounts correctly rec-
ognize that propositions must have their intentional properties inherently—
independent of any need for further interpretation by us—such accounts err in
taking this to mean that their intentionality can’t be explained by the relation they
bear to the actual and possible cognitive activity of agents. The conception of
propositions as cognitive event types saves us from this error by identifying the
intentionality-explaining relation that propositions bear to agents, not with being
interpreted by us (to mean so and so), but with having instances in which we repre-
sent things (as being so and so). Since the proposition that o is red is the event type
in which an agent predicates redness of o, it represents o as being red because all
conceivable instances of it are events in which an agent does so. The intentionality
of the event type is inherent to it in the sense that the event-type couldn’t be what
it is without bearing its intentional properties, even though it does so by virtue of
a relation it bears to agents. Being inherently intentional, it can be the interpreta-
tion of sentences and utterances, without itself being the sort of thing for which an
interpretation is needed.
entertain p that guarantees that one can think thoughts about p. The minimal form of
acquaintance with propositions is the ability to cognitively or perceptually represent
the world by predicating properties of objects, thereby generating tokens of event-types
corresponding to those predications. To gain a more robust form of acquaintance one
must be able to make propositions targets of predication. This requires the ability to
focus on the concrete events in one’s cognitive life, recognize their similarities, and
group together those bearing relevant similarity relations into units or types. Since the
proposition that o is red is an event type in which one predicates redness of o, one, who
can focus on particular events in one’s cognitive life, and reliably group together those
in which one predicates this property of this object, is in a position to make the propo-
sition that is the event type of which they are instances, an object of thought. Given the
means both of thinking of o as red, and of becoming aware in this way of so doing, one
can then make further predications about the proposition that o is red, which was the
content of one’s initial thought. For example, one may think, “That’s not true,” thereby
predicating untruth of the proposition that is the type of cognitive event one has just
experienced.
So far, I have mentioned two operations involved in proposition formation, negating
properties and predicating them of objects. In addition to negating properties, agents
also conjoin them. We entertain the proposition that o is red and round by conjoining
being red and being round, and predicating the result of o. In these cases, functions
are applied to properties. In others they are applied to objects, or to other functions.
Thus, cognitively primitive agents don’t need to predicate properties of propositions to
believe that o isn’t green or that o is red and round. Think of the function Neg as 2-place
relation in which the identity of its first argument determines the second. When P
is a property, NegP is a property uniquely true of the property which is P’s negation.
An agent acquainted with NegP can predicate the property it determines of an object.
Property conjunction is similar.
What about conjunctive and disjunctive propositions? One way of approaching the
problem would be to start with relations R& and RV. Predicating these of a, redness,
and b, roundness represents a as red and b as round (and only this), and a as red or b as
round, respectively. To perform this predication is to do something that approximates
entertaining the conjunctive and disjunctive propositions that a is red and b is round,
and that a is red or b is round. To believe the propositions one entertains is to be dis-
posed to endorse the predications. To believe their negations is to believe propositions
in which one negates R& and RV to get the relations ~R& and ~RV, which are then
predicated of the relevant arguments, just as R& and RV were. None of these beliefs
requires making propositions predication targets. By taking R& and RV to be 2-place
relations each argument of which is an n-place property followed by an ordered
n-tuple, one can get the effect of embedding propositions formed using them under
R& and RV themselves, thereby making the equivalent of full truth-functional cogni-
tion possible for agents that can’t, for whatever reason, reflect on their own cognitive
acts or experiences.
However, we don’t have to rest content with this approximation of more famil-
iar thoughts about truth-functional cognition. R& and RV are complex relations
predicated of pairs of n-tuples of the constituents of arbitrary pairs of propositions,
where each such proposition is itself the predication of a property or relation of its
other constituents. Using this model, one can, for each truth functional compound
of propositions, generate an equivalent proposition that is itself a predication of such
a pair of n-tuples. However, we can also generate propositions that are genuinely
truth-functional compounds of other propositions. Let p be a proposition that rep-
resents things as being so-and-so (and nothing more) and q be a proposition that rep-
resents things as being such-and-such (and nothing more). Next consider a certain
disjunctive operation the application of which to p and q represents things as being
so-and-so or things as being such-and-such (and nothing more). To entertain this prop-
osition is to entertain p, to entertain q, and to operate on them in this way—where
operating on them isn’t predicating anything of them. Let the result be a disjunctive
proposition. Conjunctive propositions can be treated in the same way. So can nega-
tions of propositions; when p represents things as being so-and-so, to negate p is to
represent things as not being so-and-so. On this model, all propositions involve predi-
cations at some level, but some propositions are properly characterized as operations
on constituents that themselves are, or depend on, predications. In what follows, I will
sometimes ignore this complication, e.g. when speaking of entertaining an arbitrary
proposition as predicating a property of certain things. In all such cases, a more com-
plicated statement involving predicating a property of those things or operating on
them can be supplied, without affecting the larger point at issue.
We are now ready for a more general sketch of propositions. The simplest are those
in which properties are predicated of objects. Complex propositions may involve
other operations such as conjoining, disjoining, and negating properties or proposi-
tions, as well as operating on, for example, a two-place relation R to form the reflex-
ive, one-place property self-R-ing. They may also involve applications of functions to
objects, or to properties (or propositional functions). In addition, some complex prop-
ositions involve the ascription of higher-order properties to lower-order properties
(or propositional functions) as in quantification. Propositions of any sort may also be
arguments of further predications, which we find in modal propositions and attitude
ascriptions. For example, the proposition that necessarily it is not the case that Kripke is
Kaplan is the event type of (i) predicating identity of the pair of Kripke and Kaplan (ii)
predicating untruth of, or applying the negation operation to, the event type of which
the previous predication is an instance, and (iii) predicating being necessarily true of
the complex event type of which the second predication or operation is an instance.
The proposition that John believes that Kripke is Kaplan is the event type of (i) predi-
cating identity of the Kripke and Kaplan, and (ii) predicating the belief relation of the
pair consisting of John and the event type of which the first predication is an instance.
Further detail is provided by the following illustrations. The proposition that Cicero
is wise is the event type of predicating being wise of Cicero; the proposition that he is
eloquent and wise is the event type of first conjoining being eloquent and being wise,
and then predicating the result of Cicero; the proposition that Tully shaved Cicero is
the event type of predicating the shaving relation of Cicero and Cicero; and the propo-
sition that he shaves himself is the event type of operating on the shaving relation to get
the property being one who shaves oneself, and predicating it of Cicero. The proposition
that 6 cubed is greater than 14 squared is the event type of applying the cubing function
to the number 6 and the squaring function to 14, and predicating being greater than
of what results from these applications. Functional application is also at work with
Fregean definite descriptions, which are singular terms formed from attaching “the”
to a formula. “The” denotes a function fthe that maps a propositional function g onto
the unique object to which g assigns a true proposition, if there is one; otherwise fthe
is undefined. The proposition that the G is H is the event type of applying fthe to g, and
predicating being H of whatever results from that application. The proposition that
all Gs are H is the event type of (i) applying the function fall to g, yielding the property
being true of all objects to which g assigns a truth, and (ii) predicating this property of
the propositional function h.4 More generally, the proposition that some G is so and so—
expressed by the sentence ⌜(Some x: Gx) (. ..x. . .x. . .)⌝—is the event type in which one
predicates the property being true of some object to which g assigns a truth (which results
from applying fsome to g) of a certain semantic value associated with (. ..x. . .x. . .). Which
value? Consider the proposition po expressed by the formula relative to an assignment
of object o to “x” (which, for simplicity, I stipulate to be the only variable with free
occurrences in the formula). Po will be an event type consisting of a sequence of event
types involving predications and other cognitive operations, where event types i and j
in the sequence involve operations on o corresponding to the free occurrences of “x”
in the formula.5 Let fso-and-so be the function that assigns to any object o’ the proposition
po’ that differs from po only (if at all) in that the ith and jth event types in the sequence
of event types that comprise po’ involve cognitive operations on o’ (rather than o). The
proposition that some G is so and so—expressed by ⌜(Some x: Gx) (. ..x. . .x. . .)⌝—is the
event type in which being true of some object to which g assigns a truth is predicated of
the propositional function fso-and-so. The same basic mechanism accounts for proposi-
tions expressed by sentences involving lambda abstraction, as well as those containing
anaphora of arbitrarily long distance.6
4
These functions are properties, not ordered sets. An n-place function is an n+1-place property R such
that o’ = o*, if R o1. . .ono’ and Ro1. . .ono*. The cubing function is a 2-place property that combines with n to
determine a property being the product of n times n times n. This determines the subject of the predication in
the proposition that n cubed is odd. With Fregean definite descriptions, the function fthe is a 2-place property
that combines with an argument g to determine a property being an object that is unique in determining a
true proposition when taken as argument of g. This determines the subject of predication in the proposition
the F is G. With quantification, applying the function fall to g gives us the property being true of all objects to
which g assigns a truth, which is the property predicated (rather than merely determining that property).
These points are connected with distinctions made in the final section of this chapter.
5
When both occurrences of “x” are in the same simple clause, event type i = event type j.
6
My use of propositional functions, rather than complex properties, is merely a convenience. I take no
stand on which way of filling out the theory is to be preferred.
At this point, a word must be said about how I am using the verb “predicate.” I begin
here with the account previously given in What is Meaning? According to that account,
the verb “predicate”, needed by the conception of propositions as cognitive-event
types, is analogous to the intensional transitive “look for.” If Bill is looking for Maria,
and Maria is Mary, who, in turn, is the chief of police, then Bill is looking for Mary,
but it doesn’t follow (on one reading) that he is looking for the chief of police. It also
doesn’t follow from the fact that he is looking for the fountain of youth that there is
such a thing. Analogously, if Bill predicates P of x, and x is identical to y, which, in turn,
is the unique F, then Bill predicates P of y, but it doesn’t follow that he predicates P of
the F. It also doesn’t follow from the fact that he predicates P of the F that there is an F.7
Like an intensional transitive (which expresses a cognitive relation between an agent
and a content) the verb “predicate” expresses a cognitive relation between an agent,
a property, and a content. So, if we treat definite descriptions as singular terms, the
proposition that the king of France is wise will be the event type predicating being wise
of the king of France, even though there is no king. The truth of the proposition depends
on there being something of which being wise is predicated, but its existence doesn’t. (This
account of the content of “predicate” will suffice until the final section of this chapter
when it will be modified to accommodate special examples considered there.)
More can be said about the existence conditions of propositions and other event
types. First, if an event type E has instances that exist, then E exists. For example, since
I can (directly) refer to Socrates even though he no longer exists, if I do so refer, then
a concrete event e exists that is an instance of the minimal event type in which one
refers to Socrates—which must also exist. Ditto for the (minimal) event type in which
one predicates no longer existing (directly) of Socrates—which is the proposition that
Socrates no longer exists. Since, in certain cases, one can also (directly) refer to merely
possible individuals, there exist propositions—event types of predicating properties
of those individuals—the “constituents” of which have never existed and never will.8
To understand this, one must not confuse failing to refer with referring to a
non-existent. “The present king of France” fails to refer, and so has no referent; “Socrates”
has a referent, just one that doesn’t exist. The view endorsed is not Meinongian—
there is no such thing as the golden mountain, whether existent or non-existent. This
non-Meinongian view eliminates an alleged problem for Millians: namely, that if one
of the so-called “constituents” of a singular proposition fails to exist then the proposi-
tion also fails to exist. This false claim relating the existence of a proposition to the
existence of its constituents comes from thinking of propositions in the wrong way.
7
“P” is here used as a variable over properties, while “F” is used as a schematic predicate letter.
8
For more on referring to, or quantifying over, the non-existent, see Nathan Salmon, “Existence,”
Philosophical Perspectives, 1, 1987, 49–108; Scott Soames, “Actually,” in Mark Kalderon, ed., Proceedings of
the Aristotelian Society, supplementary volume, 81, 2007, 251–277, reprinted in Soames, Philosophical Essays,
Vol. 1: Natural Language: What it Means and How we Use it, Princeton and Oxford: Princeton University
Press, 2009; and pp. 128–129 of Soames, Philosophy of Language, Oxford and Princeton: Princeton University
Press, 2010.
9
Although Fregean definite descriptions are singular terms in some possible languages, I do not assume
that definite descriptions in English are Fregean. Rather, I take them to be generalized quantifiers.
10
In a complete account, this principle—which covers only atomic propositions—would be extended to
include complex propositions as well. The idea behind the extension is this: Let the event-type p be a propo-
sition; let the basic acts the performance of which define the sequence of event types that make up p be
targeting a given object o, predicating a particular simple property p, applying a certain function f, targeting
a specific argument g, negating something, conjoining a pair of things, etc.; then p will exist if each of those
acts has been performed.
true. But they don’t. Although many properties require things that have them to exist,
some don’t. An individual can have the properties being dead, being referred to by me,
and being admired by someone despite not existing. Similarly, a pair of individuals—
Plato and Aristotle—can instantiate the relation of non-identity without existing. By
the same token, a proposition can represent something as being a certain way, and so
be true because the thing is that way, whether or not the proposition exists. Thus, there
is nothing to prevent the nonexistent proposition that m is a molecule from being true.
Since we can quantify over the merely possible, we can quantify over possible proposi-
tions, and say that if p predicates being so-and-so of o, then p is true (at world-state w)
iff (at w) o is so-and-so, whether or not p exists (at w). Propositions that couldn’t be
entertained by any agent might require a further story, as do metaphysically impossible
objects generally, but even here it is not obvious that there are irresolvable difficulties.
Perhaps we will have to draw a line excluding propositions that couldn’t conceivably be
entertained by any agent, but if so, would that really be a loss?
Although a proposition p can be true at a world-state w without existing at w, p can’t
be entertained at w, accepted at w, asserted at w, denied at w, or judged at w to be true
without existing at w. To bear any of these attitudes to p at w an agent must entertain p
at w. Since in each case, this involves producing an instance of the event type that p is,
bearing any of these attitudes guarantees the existence of p. So, apart from a few minor
complications—e.g. to believe p doesn’t strictly require one to have entertained p, but
only to be disposed to bear the judging relation to p—when we really need the existence
of propositions as objects of attitudes they are (nearly enough) guaranteed to exist.
11
See Soames, “Understanding Deflationism,” Philosophical Perspectives, 17, 2003, 369–383, also in Soames,
Philosophical Essays: Vol. 2: The Philosophical Significance of Language, Princeton: Princeton University
Press, 2009.
12
See Soames, “Actually,” and chapters 5 and 6 of Philosophy of Language for fuller explanations.
sentences that express them. Just as propositions are event types in which agents per-
form certain representational cognitive operations, so sentences may plausibly be
taken to be event types instances of which are utterances and inscribings—thought of
as concrete events occurring at particular times and places in which agents produce
auditory, visual, or tactile tokens endowed with semantic and syntactic properties.
On this picture, since both sentences and propositions are event types, they can share
common instances: e.g., cases in which the event that is one’s referring to (targeting) o
and predicating P of o is the event of one’s using expression E to refer to (target) o and
expression F to predicate P of o. The cognitive acts or operations—referring to (target-
ing) o and using E to do so—are different, as are the acts or operations of predicating P
of o and using F to do so. Thus, the event type p that consists of one’s performance of the
propositional acts or operations (of referring and predicating) differs from the event
type S that consists of one’s performance of the sentential acts or operations (of using
E to refer and F to predicate)—even though some instances of one are also instances
of the other. When an event is an instance of both a sentential and a propositional type,
there is no extra inner event of “grasping the proposition” over and above using the sen-
tence meaningfully. So, when S expresses p, one who understands S can entertain p by
tokening S. For some propositions, this may be our only feasible way of entertaining
them. In such cases what distinguishes p from S is the possibility that an event could be
an instance of one but not the other. More generally, the heretofore mysterious express-
ing relation holding between a sentence and a proposition may be grounded in some-
thing like the by relation that holds between two things that are done when an agent
can do one of those things (entertaining the proposition) by doing the other (uttering
or inscribing the sentence).
The fourth advantage I will mention here is the prospect for illuminating otherwise
puzzling semantic phenomena. Here is an example.13
3a. Russell defended the proposition that arithmetic is reducible to logic.
b. Russell defended logicism.
4a. Mary believes that Russell defended the proposition that arithmetic is reducible
to logic.
b. Mary believes that Russell defended logicism.
“Logicism” is a Millian proper name for the proposition that arithmetic is reducible
to logic, which is also designated by the directly referential that-clause. Nevertheless
sentences (3a) and (3b) express different propositions, and the truth of (4a) guarantees
the truth of (4b), but not vice versa. “Logicism” and the that-clause contribute the same
proposition L to those expressed by the sentences in (3) and (4). But the clause somehow
13
This example is discussed in Mark Richard, “Articulated Terms,” Philosophical Perspectives, 7, 1993,
207–230; and Soames, “What are Natural Kinds?,” Philosophical Topics, 1 and 2, 2007, 329–342.
also contributes something else to the propositions expressed. The view of propositions as
cognitive event types explains what and why.
According to it, understanding sentence (3b) and entertaining the proposition it
expresses requires one to think of L, and to predicate having defended of the pair consist-
ing of Russell and L. Since one can think of L simply by possessing the name “logicism,”
without knowing much about its referent, one who is competent with the name, and
accepts sentence (3b), can entertain, and even believe, the proposition it expresses with-
out being able to state, or informatively identify, L. By contrast, in order to understand
sentence (3a) and entertain the proposition it expresses, one must first predicate being
reducible of the pair consisting of arithmetic and logic—thereby entertaining the propo-
sition L expressed by the that-clause. Next, one predicates having defended of the pair of
Russell and L. This difference carries over to (4a) and (4b), with the result that the truth
of the former requires the truth of the latter, but not vice versa. Because propositions are
event types that involve thinking of things and predicating properties of them, two proposi-
tions can place different constraints on how an agent thinks about their common predica-
tion targets, even if the truth conditions of the two propositions result from predicating the
very same properties of the very same targets. Although (3a) and (3b) predicate the same
thing of the same targets, the former is an event type in which the propositional coor-
dinate must be cognized by entertaining it, while the latter is an event type that doesn’t
require this. The difference in truth value between (4a) and (4b) is sensitive to this.
In this way, taking propositions to be cognitive event types brings together the two
related but distinct aspects of linguistic and cognitive content. On the one hand, such
content faces the world—imposing conditions that must be satisfied, if the world is to
conform to the way it is represented to be. On the other hand, this content also faces
the mind, imposing conditions on what it takes for an agent to entertain it. Whereas
the worldly aspect of content has long been accommodated in semantics, it has been
difficult to do justice to the mental aspect of content when integrating the two. The
conception of propositions as representational cognitive event types provides us with
a natural way of achieving this. Being representational, the truth conditions of propo-
sitional event types– in virtue of which they “face the world”—are essential to them.
Being event types in which one performs cognitive acts or operations, propositions
can impose different conditions on the cognitive operations it takes to entertain them,
even when they are representationally identical in the sense that their truth conditions
are derived from predicating the very same properties of the very same things. The fact
that we need to recognize propositions that do differ in this way strongly supports a
metaphysics of propositions that explains how this is possible.
14
For classic discussion, see David Lewis, “Attitudes De Dicto and De Se,” The Philosophical Review, 88,
1979, 513–543.
15
John Perry, “The Essential Indexical,” Nous, 13, 1979, 3–21.
16
See also, John Perry, “Frege on Demonstratives,” The Philosophical Review, 86, 1977, 474–497.
(A) All there is to a proposition is its representational content; hence propositions the truth
conditions of which arise from predicating precisely the same properties of precisely the same
things, are identical.
One reasonable (though not theoretically neutral) description of all de se cases is that
they turn on systematically different ways of believing/asserting representationally
identical things—in particular on believing or asserting in the special first person, or
immediate present tense, way versus believing/asserting in a person-time neutral way.
Although believing/asserting in the first way is generally thought to guarantee believ-
ing/asserting in the second way, the converse doesn’t hold. De re cases are those in
which we have the latter without the former. In the presence of (A), this means that
the agent’s de se epiphany can’t be a matter of coming to assert, believe, or know any
proposition not already asserted, believed, or known. Thus the conventional wisdom
about the supermarket example has been that there is no proposition that Perry came
to assert, believe, and know that he hadn’t already asserted, believed, and known.
This “wisdom” comes in two opposing forms. According to Perry, the epiphany is
not one of coming to believe or know a new proposition; it is one of coming to believe
or know an old proposition in a new (first-person/present-tense) way. According to
David Lewis, it involves coming to bear the primitive attitude self-ascribing to a certain
property P—where this primitive attitude must not be confused with the ordinary atti-
tude ascribing to an individual who happens to be oneself (which Perry already bore to
P when he saw his image in the mirror). The property P is, of course, being one who is
making a mess, which, according to Lewis, is what Perry came to know and believe in his
moment of epiphany. Both the views of Perry and those of Lewis are revisionary, in the
sense of explaining away, rather than preserving, some of our pre-theoretic thoughts
on the matter. Whereas Perry’s view is theoretically conservative, it flies in the face of
the irresistible urge to describe the messy shopper as coming to learn (know, believe)
something he didn’t know (believe) before. Whereas Lewis’s view respects our judg-
ment about this, it does so at the cost of reconstruing all cognitive attitudes previously
taken to be relations between agents and propositions—the representational nature of
which is readily explainable—as relations between agents and properties (some ger-
rymandered). Unfortunately, we are given no explanation of how a property like being
one who is making a mess can truly or falsely represent anything as being one way or
another.
Rather than dwelling on the challenges facing these views, I will try to enlarge the
space of alternatives. Since we know already from the discussion of (3) and (4) that
assumption (A) is false (because representationally identical propositions can differ in
the cognitive requirements for entertaining them), the tacit premise from which Perry
and Lewis derive their common conclusion—that in de se cases no new proposition is
asserted, believed or known—is no longer available. Without it, their conclusion no
longer follows. If propositions are cognitive event types, the possibility remains that at
the moment of epiphany the agent does come to assert, believe, and know a special de
The idea is that each person p has a (first-person) way of thinking of p that no one else
can use to think about p. This idea is plausible—as is the idea that for each time t there
is a special way of thinking about t, at that very time, that is not available at any other
time. The difficult point for the neo-Fregean is to show that these ways of thinking of
oneself or the present time are Fregean senses: one that uniquely picks out p and is a
constituent of propositions entertained by p when p thinks about p in this first-person
way, and one that uniquely picks out t, and is a constituent of propositions about t that
are entertainable only at t. John Perry and David Kaplan have explained why ordinary
Fregean senses of definite descriptions can’t be identified with these special ways of
thinking.18 Although Saul Kripke has recently argued that special “acquaintance-based
senses” can be so identified, this appears to be false, as I have argued elsewhere.19 This
suggests that there simply are no propositions, as traditionally conceived, that can be
objects of newly acquired beliefs in de se cases.
But propositions are not what they have traditionally been conceived to be; they
are cognitive event types. Consider the de re proposition entertainable by anyone who
predicates being one who is making a mess of John Perry. Its constituents are the man
Perry and the property being one who is making a mess. There are no special constraints
on how one must think of Perry in order to entertain this proposition, beyond the abil-
ity to refer to him directly. In this respect, the de re proposition about Perry is analo-
gous to the proposition (3b) that Russell defended logicism. By contrast, if there is a de
17
Frege, “Der Gedanke,” Beitrage zur Philosophie des deutschen Idealismus, 1, 1918, 58–77; trans. “The
Thought” by A. and M. Quinton, in Mind 65, 1956, 289–311; trans. “Thought,” in Beaney, The Frege Reader,
Oxford: Blackwell, 1997, 325–345, quoted at p. 333.
18
From Perry we get the example of Rip Van Winkle, who awakens on October 20, 1823 after sleeping
20 years, and thinks (wrongly) “Today is October 20, 1803.” Here, the belief is about the day on which it
occurs, no matter what day, if any, satisfies the qualitative temporal description Mr. Van Winkle has in
mind. See Perry, “Frege on Demonstratives,” Philosophical Review, 86, 1977, 474–497, at p. 487. From Kaplan
we get the example of Castor and Pollux, raised in qualitatively identical environments to be molecule
for molecule identical, and to associate the same purely qualitative descriptions with their correspond-
ing uses of the same terms. Despite this, each refers to himself and not to the other when he uses “I.” See
Kaplan, “Demonstratives,” in Joseph Almog, John Perry, and Howard Wettstein, eds., Themes from Kaplan,
New York: Oxford University Press, 1989, 481–563;, at p. 531.
19
Kripke, “Frege’s Theory of Sense and Reference,” Theoria 74, 2008, 181–218; Soames, chapter 2 section 7
of The Analytic Tradition in Philosophy, Vol. 1.
20
For example, an agent who self-ascribes F-hood and thereby believes the de se proposition expressed by
a use of “I am F,” also counts as believing the de re proposition expressed by “x is F” relative to an assignment
of the agent to the variable “x.” This is evidenced by the fact that whenever an agent A believes that A is F (de
se) there is some individual x such that A believes that x is F (de re). Although it may often be the case that
when A has the de se belief, A also believes (de re) the de re proposition he would use ⌜t is F⌝ to express, for
some name, indexical, or demonstrative used to refer to A, this need not always be so. Whether or not it is so
in a particular case depends on the semantics and pragmatics of ⌜t is F⌝, as used by A.
21
For simplicity in specifying the proposition expressed I will adopt Russell’s technique of letting propo-
sitional functions stand in for complex properties, and so will take the quantified proposition to predicate a
property of such a function, rather than taking it to predicate a higher-order property of a lower-order one.
Nothing hinges on this. “F” and “G” are used as schematic letters.
The proposition (5 de re) predicates the property assigning a true proposition to every
F of the propositional function Bde re that assigns to each individual o the proposition
(5Matrix/o de re). That proposition, in turn, predicates the believing relation of the pair of
o and the proposition (5Complement/o de re)—which itself predicates being G of o (without
placing further restrictions on how o must be cognized in order to be entertained).
Obviously, entertaining proposition (5 de re) doesn’t require one to entertain all propo-
sitions in which being G is predicated of an individual, or even all propositions in which
believing that one is G is predicated of an individual of which F is true. It is enough that
the reporter be able to think about, and refer to, the propositional function Bde re (or the
complex property it represents). Although I haven’t discussed what this amounts to, no
special problems are presented by examples like (5).
The same is true when (5) is understood as reporting de se attitudes. Proposition (5
de se) predicates the property assigning a true proposition to every F of the propositional
function Bde se that assigns to each individual o the proposition (5Matrix/o de se). This
proposition, in turn, predicates the believing relation of the pair consisting of o and
the proposition (5Complement/o de se)—which itself predicates being G of o, while requiring
any agent who entertains it to think of o in the first-person way. Thus, in making simple
propositions like (5Complement/o de se) available, the conception of propositions as cogni-
tive event types also makes propositional functions like Bde se and complex propositions
like (5 de se) available. Since agents may use (5 de se) to report or think about the de se
attitudes of others—thereby entertaining, asserting or believing the proposition they use
the belief ascription to express, without entertaining the de se propositions it represents
others as believing—there is no general problem reporting or thinking about the de se
attitudes of other agents.
Next consider (6a), in which “he” is anaphoric on its antecedent, the name “John
Perry.”
6a. John Perry came to believe that he was the messy shopper.
The semantic effect of the anaphoric relation between antecedent and pronoun is to
introduce a variable-binding operation making (6a) equivalent to (6b).22
6b. λx (x came to believe that x was the messy shopper) John Perry
Thus, to use (6a) on its de se understanding is to use it to assert the proposition (6 de se)
in which being an individual to which the propositional function Bλ de se assigns a truth is
predicated of John Perry—where Bλ de se assigns an individual o the proposition (λMatrix/o
de se) that predicates the believing relation of the pair consisting of o and the proposi-
tion (λComplement/o de se). Finally, this proposition predicates being the messy shopper of
22
This account of anaphora is defended on independent grounds in Soames, “Attitudes and Anaphora,”
Philosophical Perspectives, 8, 1994, 251–272; reprinted in Philosophical Essays: Vol. 2: The Philosophical
Significance of Language.
o, while requiring any agent who entertains it to think of o in the first-person way.23 The
end result is that the proposition expressed by (6a) predicates the property believing of
oneself (in the special first-person way) that one is the messy shopper of John Perry. As
before, the de se proposition reported to be believed doesn’t have to be entertained by
the reporter.24
Perry can, of course, report his own de se attitude using (6c), taking the complement
clause to express the de se proposition he entertains.
6c. I believe that I am the messy shopper.
If he does this, his utterance will assert both that he believes the de se proposition only
he can entertain, and that he believes the de re proposition, entertainable by all, that
predicates being the messy shopper of him. His hearers will understand the de re asser-
tion in the normal way, while also realizing that he endorses the being-a-messy-shopper
predication thinking of its target, Perry himself, in the special first-person way. This, it
may be argued, is our pretheoretic way of describing the de se proposition he believes,
without having to entertain it ourselves.
A further twist is provided by a version of the case in which Perry’s epiphany comes
in two stages. In this version he is accompanied on his trip around the supermarket by
his daughter, who is also intent on finding the messy shopper. At the crucial moment,
her face lights up with the realization that her father is the culprit. Noticing the shock of
recognition on her face, Perry mutters (7a) under his breath, wondering if she is right.
7a. She thinks that I’m the messy shopper.
This is Perry’s first epiphany, in which he seems to believe a new proposition (about
his daughter’s beliefs) that he had not previously believed. Since it is a new belief, all
the old arguments can be recycled to show that the proposition believed can’t be an
ordinary (non-de se) proposition. What proposition is it? Although “I” occurs in the
complement clause of (7a), Perry’s use of (7a) surely doesn’t attribute to his daughter
23
If we conduct the analysis in terms of complex properties rather than propositional functions, the de
se proposition (6b) predicates, of Perry, the property being one who believes the proposition that (i) predi-
cates being the messy shopper of one, while (ii) requiring any agent who entertains it to think of oneself in the
first-person way.
24
One way of developing these ideas about (5a,b) and (6a,b) in a semantic theory would be to allow two
kinds of occurrences of variables—normal ones, and ones that result from adding “*” to occurrences in
complement clauses that represent pronouns anaphoric on matrix subjects of attitude ascriptions. Although
adding “*” would not affect the referent of the variable relative to an assignment, a complement clause
. . .x*. . . containing a *ed occurrence of “x” would express a singular proposition, relative to an assignment of
o to “x”, that can be entertained only by the referent of “x” relative to the assignment. The difference between
de se and de re readings of attitude ascriptions in cases like (5a,b) and (6a,b) could then be captured by a
difference between *ed and non-*ed occurrences of variables representing pronouns in the complement
clause. In effect, pronouns anaphoric on matrix-subject antecedents will be ambiguous between de se and de
re inducing occurrences of variables. By contrast, the phonologically empty syntactic constituent PRO -- as
in “He expects PRO to win” (meaning “He expects to win”) could be treated as unambiguous, and thus as
always corresponding to a *ed occurrence of a variable bound by a higher subject.
a belief in a de se proposition that she couldn’t possibly entertain. Nor, if Perry’s use of
(7a) is to express a new belief, can one maintain that his use of “I” is purely de re.
There is, however, another option. Perry’s new belief can be identified with the de se
proposition expressed (from Perry’s point of view) by (7b).
7b. λx (she believes x is the messy shopper) me
Understood de se (from Perry’s perspective), (7b) expresses the proposition that is the
cognitive event type of predicating being someone believed by her [his daughter] to be
the messy shopper, of Perry, thinking of the predication target, Perry, in the special first-
person way. Here, I have used lambda abstraction to give the first-person pronoun
what is, in effect, wide-scope over “believe.” The mechanism used to achieve this is not
crucial—there are multiple ways of encoding scope distinctions for terms of various
types (not just quantifiers). What is needed is simply that one’s semantic or pragmatic
theory provides a principled way of applying one of them here.
This completes my preliminary sketch of the possibilities for the analysis of de se atti-
tudes that are opened up by the conception of propositions as cognitive event types. By
providing a natural explanation of how a proposition p can constrain the way one of its
constituents (e.g. a predication target) must be cognized by an agent who entertains p,
the conception makes available de se (first-person, present-tense) propositions distinct
from those that have previously been recognized. Since these special ways of cognizing
a predication target do not, for the reasons indicated above, involve any new predica-
tions of that target, the new propositions are representationally identical to ordinary
de-re propositions. Of course, my sketch falls far short of a complete analysis. However,
if it seems promising, the general lesson to take from it is worth emphasizing; coming
to understand what propositions are can be important not only in providing philo-
sophical foundations for linguistic and cognitive theories, but also in enhancing their
explanatory potential by providing new tools for empirical analyses.
contextually determined inquiry. Let “n” be a proper name of w, and let “PW” abbre-
viate an articulated term “the property making it true . . .” that articulates each of the
basic propositions used in defining w. Finally, let “p” designate any proposition true
at w—where for a proposition to be true at w is for the proposition to be an a priori
consequence of the set of basic propositions used to define w.25
Then consider (8).
8a. p is true at PW.
b. p is true at n.
c. p is true at this very world-state—said at w referring to w
The relationship between the propositions expressed by (8a) and (8b) is like the rela-
tionship between propositions (3a) and (3b). In both cases, the (a) and (b) propositions
predicate the same properties/relations of the same things; in both cases the (a) propo-
sition differs from the (b) proposition only in adding the constraint that in order to
entertain it the agent must also entertain its propositional constituent (in the case of
3a), or its propositional sub constituents (in the case of 8a). Moreover, in both cases,
anyone who knows or believes the (a) proposition knows or believes the (b) proposition,
but not conversely.
Now consider the relationship between the propositions expressed by (8b) and (8c).
I have already made room for a special first-person way of thinking about, and refer-
ring to, an individual x, plus singular propositions about x that can be entertained only
by x when thinking about himself or herself in the first-person way. I have similarly
made room for a special present-tense way of referring to a time t—as now, or this
very time—plus singular propositions about t that can be entertained only by one who
thinks of t in that way at t. Having done so, I introduce a parallel hypothesis about
world-states which, although I neither accept nor reject it, is, I think, worth further
examination. The hypothesis is that there is a special world-bound way of referring
to a world-state w—as this very world-state—plus singular propositions about w that
can be entertained only by those at w who think about it in this special world-bound
way. On this hypothesis, the relationship between proposition (8b) and proposition
(8c) parallels the relationship between (i) the ordinary singular proposition that
Scott Soames wrote What is Meaning? and the de se proposition that I wrote What is
Meaning? (entertained by me in the first-person way), and (ii) the ordinary singular
proposition about the present time t that Martha is working at t and the temporally de
se proposition I express to myself by saying Martha is working now (at this very time).
Remember, these special ways of thinking of things are not special descriptive ways
of thinking about them. Castor and Pollux can be in qualitatively identical cognitive
states when thinking about themselves in the first-person way, even though each refers
to himself and not the other. Rip Van Winkle can be in qualitatively identical cognitive
25
See Soames, “Actually,” summarized in chapter 6 of Philosophy of Language. There is, of course, no
requirement that the propositions in terms of which a world-state is defined exist at w.
states at different times when thinking of, and referring to, those different times in the
same special present-tense way. By the same token, the world-state to which I actu-
ally refer in the hypothesized special world-bound way—by saying to myself “this
very world-state”—is different from the world-state to which I refer at a world-state
that is merely possible, even though my cognitive state at that world-state is identical
with the one I am actually in. On the hypothesis under investigation, all of these cases
involve special ways of thinking about and referring to things the identities of which
are not determined by any description imposed by the agent. On the contrary, the
items picked out are determined by who the agent is, when the thoughts are occurring,
and the world state at which the agent entertains the proposition. Hence, the hypothesis
maintains, including these special ways of thinking about or referring to the predica-
tion targets of propositions do not introduce any new descriptive content, or any new
truth-condition-determining predications, into the ordinary singular propositions to
which they are added.
Given all this, we can distinguish propositions (8a), (8b), and (8c). Although all are
singular propositions in which being something at which p is true is predicated of the
world-state w, propositions (8a) and (8c) impose constraints on how agents who enter-
tain them must think about w, while proposition (8b) imposes no such constraint.
Proposition (8a) requires the agent to think of w by entertaining the propositions
that are themselves constituents of the property that w is. Proposition (8c) requires
thinking about w in the special actual-world-state way that parallels the present-tense
way of thinking about a time and the first-person way of thinking about an agent. We
have seen that knowing or believing proposition (8a) guarantees knowing or believing
proposition (8b), but not vice versa. The same relation holds between proposition (8c)
and proposition (8b). As in the ordinary de se case, where knowing or believing de se
guarantees knowing or believing de re, but not the other way around, so knowing or
believing the worldly-de-se proposition (8c) guarantees knowing or believing (8b), but
not conversely.
We can now connect this with an interesting result about the a priori. On the account
of world-states indicated above, the propositions true at a world-state are those that are
a priori consequences of the basic propositions that define it. With this in mind, let w be
the actual world-state, and let the proposition that Plato was a philosopher be an a priori
consequence of the basic propositions defining w. We then get the following results:
(i) The proposition that it is true at PW that Plato was a philosopher (correspond-
ing to (8a)) is knowable a priori but the proposition that it is true at this very
world-state that Plato was a philosopher (corresponding to (8c)) is not knowable
a priori;
(ii) The proposition that Plato was a philosopher iff it is true at this very world-state
that Plato was a philosopher is knowable a priori but the proposition that Plato
was a philosopher iff it is true at PW that Plato was a philosopher is not knowable
a priori;
(iii) It follows from (i) that the proposition that that Plato was a philosopher is true
at w is knowable a priori. It follows from (ii) that the proposition that that Plato
was a philosopher iff it is true at w that Plato was a philosopher is knowable a
priori.26 But neither their conjunction nor the proposition that Plato was a phi-
losopher is so knowable.
Since we have two propositions that are knowable a priori, even though their conjunc-
tion isn’t, the set of a priori truths is not closed under a priori consequence. Although
this result isn’t new, the simple explanation of it provided by the hypothesis that
this-very-world-state cognition parallels special first-person and present-tense cogni-
tion is, as is the recognition that this result is of a piece with those discussed in the
previous two sections.27 The fact that the conception of propositions as cognitive event
types allows us to tie these phenomena together as three aspects of the same thing pro-
vides some reason for taking the hypothesis seriously.
Semantic relationism
Another case in which there may turn out to be a productive interplay between a satis-
fying metaphysical conception of what propositions are and an empirically-informed
account of the kinds of propositions needed in semantics and cognitive science comes
from Kit Fine’s fascinating work in Semantic Relationism.28 Although he doesn’t give
a metaphysics of propositions there, he does argue that we must recognize a class of
coordinated propositions that place special constraints on how an agent is required to
think of their predication targets in order to entertain them. It is central to his concep-
tion that these propositions share the structure, constituents, and truth conditions of
their uncoordinated counterparts, which lack special constraints on how their predica-
tion targets are cognized. Though his theory doesn’t apply to examples like (3a) and
(3b), what he says about the difference between coordinated and uncoordinated prop-
ositions is importantly similar to what I have said about (3a) and (3b). This raises the
question of whether the conception of propositions as cognitive event types is capable
of accommodating Fine’s coordinated propositions.
I noted earlier that the proposition that Tully shaved Cicero is the event type of
predicating the shaving relation of Cicero and Cicero, and that the proposition that
26
From (ii) we get the a priority of the proposition that λy [Plato was a philosopher iff it is true at y that Plato
was a philosopher] this very world-state, which guarantees the a priority of the proposition that λy [Plato was
a philosopher iff it is true at y that Plato was a philosopher] w, which guarantess the a priority of the proposi-
tion that Plato was a philosopher iff it is true at w that Plato was a philosopher.
27
A version of this result is explained and established in chapter 6 of Soames, Philosophy of Language.
The failure of closure of a priority under a priori consequence does not upset the claim that for a proposi-
tion to be true at w is for it to be an a priori consequence of the basic propositions defining w. Failures of
closure always involve propositions about world-states, which aren’t among the propositions that define
world-states.
28
Kit Fine, Semantic Relationism, Malden, MA: Wiley-Blackwell, 2007.
he shaves himself is the event type of reflexivizing the shaving relation to get the prop-
erty being one who shaves oneself and predicating it of Cicero. According to Fine, the
(coordinated) proposition that Cicero shaved Cicero is different from both of these.
Expressed in my terms, it is the event type of predicating the shaving relation of Cicero
and Cicero, thinking of the two as the same. What is it to do this? It is not to predicate
the shaving relation while assuming that the individual one’s predication represents as
shaver is the same as the individual one’s predication represents as being shaved. To pred-
icate a relation of a pair, one must think of the relation and the pair; one doesn’t also
have to make a higher-order judgment about what one’s predication represents. Surely
there are agents who predicate properties of things, and thereby have propositional
attitudes, without bearing attitudes to propositions about their own cognitive activi-
ties. Given the importance for thought and action that Fine takes coordination to have,
he would, I am confident, not wish to exclude such agents from the benefits of bearing
attitudes to coordinated propositions.
Nor can the coordinated proposition be the event type in which one predicates shav-
ing of Cicero and Cicero, while judging Cicero to be identical with Cicero. The content
of that judgment can’t be the uncoordinated proposition that Cicero is Cicero (which
is just the proposition that Cicero is Tully). Nor can it be the coordinated proposition,
since that would involve using coordination to explain coordination. So Fine must
take thinking of the members of a pair as the same to be primitive, understanding that
to predicate shaving of Cicero and Cicero while bearing this attitude to them is differ-
ent from predicating self-shaving of Cicero, and also that predicating being F of o, and
being G of o while taking them to be the same is different from predicating being F and
G of o.29
Suppose, for the sake of argument, that there is such a primitive attitude of taking
things to be the same. Since it is a kind of cognition, it may seem that there must be
propositions the entertaining of which requires one to cognize things in this way. Are
there really such propositions? This question can be taken in two ways: (i) Is there
a need within semantics and cognitive science to recognize such propositions? (ii)
If there is a need, does the conception of propositions as cognitive event types make
room for them? Elsewhere I have argued that the case for coordination in natural
language semantics (and cognition) remains inconclusive (at best).30 Thus question
(i) remains open. Here, I will confine myself to saying a word about (ii).
29
This is Fine’s view. On page 59 he says: “But the coordinative aspect of the coordinated content of a
sentence, such as ‘Cicero wrote about Cicero’ is entirely lacking in any special descriptive or truth conditional
character and relates entirely to how its truth conditions. . .are to be grasped [entertained]. It is a significant
feature of the traditional Fregean view that there can be no difference in what it is to grasp [entertain] the
sense of an expression without there being a difference in how the sense has application to [or represents]
the world.. . . But under the relational view, these two aspects of sense come completely apart. There is no dif-
ference in what it takes for the sentences “Cicero wrote about Cicero” and “Cicero wrote about Tully” to be true,
even though there is a difference in their coordinated content.”
30
Soames, “Two Versions of Millianism,” in Michael, O’Rourke, ed., Topics in Philosophy, Vol. 10: Reference
and Referring, Cambridge: MIT Press, 2012, 83–118.
Surely, there are limits on what cognitive acts or operations propositions can
encode. Thinking of a certain tune while predicating redness of an object is a (com-
plex) cognitive activity of some sort, as is predicating a relation of a pair, while feel-
ing affection toward its members. Presumably, the event types of doing these things
are not propositions because one of their cognitive components is orthogonal to how
the agent represents things to be. By contrast, propositions (3a) and (3b) differ—
despite the fact that their truth conditions arise from predicating the same relation
of the same things—because the condition placed on a predication target in (3a)
is that it be entertained, which is itself a cognitive activity in which certain things
are represented to be a certain way. Since the cognitive activities that make up (3a)
are all representational, it is a genuine proposition that differs from (3b). This sug-
gests that whether there are (or could be) Finean coordinated propositions depends
on whether the putative attitude of taking objects to be the same is appropriately
representational.
How should we think about it? It may help to ask about cases in which one takes
non-identical objects to be the same. Certainly non-identical objects can appear the
same way identical objects do, and so, one would think, provoke the same cogni-
tive responses. Are we then to suppose that there are propositions we can entertain
only by predicating something of a pair of non-identical things, mistakenly taking
them to be the same? Suppose an agent mistakenly takes Cicero and his brother
(each of whom shaves the other but neither of whom shaves himself) to be the same,
while predicating the shaving relation of them. If this sequence of cognitive acts is
encoded by a genuine “coordinated” proposition, what is its truth value? To say it is
true ignores the fact that entertaining it requires one to be disposed to mistakenly
judge non-identical things to be identical, and so to think of them as related in a
way they are not—which seems very much like representing them falsely or incor-
rectly. To say that the putative proposition is not true is a non-starter if, as I have
indicated, the truth conditions of a proposition are derived from what it predicates
of what (which in Fine’s system is reflected by the requirement that coordinated
propositions share the truth conditions of their uncoordinated counterparts).
For this reason, it is unclear that there are (or could be) propositions in which
non-identical objects are coordinated—which, in turn, casts some doubt on the
existence of propositions in which identical objects are coordinated.I am not sure
whether this doubt can be overcome. If it can, then the conception of propositions
as cognitive event types doesn’t, as far as I can see, create any further difficulties
for Fine’s view. Thus, I tentatively conclude that the metaphysics of propositions
offered here provides important philosophical grounding for Fine’s general concep-
tion of representationally identical but cognitively distinct propositions without
raising any new problems for his particular conception of coordinated propositions
that it doesn’t already have.
31
Russell, “On Denoting,” Mind 14, 1905, 479–493. The argument comes from “On Fundamentals,” writ-
ten in June 1905, first published in Russell, Foundations of Logic, London: Routledge, 1994. Exposition and
criticism of the argument, along with an account of its role in leading Russell to his theory of descriptions
is found in chapters 7 and 8 of The Analytic Tradition in Philosophy, Vol. 1. See also Richard Cartwright,
“On the Origins of Russell’s Theory of Descriptions,” in Philosophical Essays, Cambridge: MIT Press, 1987,
95–133, and Nathan Salmon, “On Designating,” Mind, 114, 2005, 1069–1133.
32
I here assume that (9b) unambiguously expresses the true proposition indicated above, and that (9c)
unambiguously expresses the false proposition there indicated. I will return to question of ambiguity below.
and (ii) these meanings can occur in propositions only in the role of present-
ing their denotations as the ultimate subjects of predication in the propositions.
Therefore, these meanings can never be subjects of predication in any proposi-
tion in which they occur.
Accepting R leads Russell to deny that (9b) is true and to conclude that there is no true
singular proposition in which being what “the first line of Gray’s Elegy” means is predi-
cated of anything. From here it is a short step to the conclusion that no one can know of
anything that it is what the description means. But surely, Russell thinks, in order for
an expression E to mean something it must be possible to know of what E means that E
means it. Thus, he concludes, it is impossible for meaningful definite descriptions (of
any language) to be singular terms. (The argument generalizes to all complex singu-
lar terms formed by combining a function symbol with one or more terms serving as
arguments—e.g. “32.”)
Since Russell’s conclusion is clearly false, and R is the likely culprit, we need a con-
ception of propositions that provides an explanation of why it is false. Although purely
technical moves could be made, no conception of propositions I know of provides a
plausible explanation of why this should be so. I believe that the conception of proposi-
tions as cognitive event types can do better. Let us start by taking proposition (9a) to
be the event type of (i) thinking of the function fthe and the function g (that assigns an
object a truth iff it is a line in Gray’s Elegy preceding all others), (ii) combining fthe and
g into a single constituent in which g is understood to play the role argument-of-fthe,
and (iii) predicating the identity relation of the pair consisting of “the curfew tolls the
knell of parting day” plus the result of applying fthe to g. Let it be part of the theory that
the constituent fthe-plus-g is the meaning of the definite description. Let it also be part
of the theory that this constituent occurs in the propositions expressed by each of the
sentences (9a,b,c). Since “the curfew tolls the knell of parting day” is uniquely deter-
mined by fthe-plus-g, proposition (9a) comes out true. Similar reasoning gives us the
falsity of (9c). What about proposition (9b)? It seems that it should be the event type of
(i) predicating means of the pair “the first line of Gray’s Elegy” and fthe-plus-g. However,
since these are also the (major) constituents of proposition (9c), the assumption that
in both cases the same relation is predicated of the same arguments requires the two
propositions to have the same truth value.33 This is what leads to Russell’s false conclu-
sion. So far, the problem is still with us.
The conception of propositions as cognitive event types does, correctly, block the
conclusion that propositions (9b) and (9c) are identical. The case is analogous to (3a)
33
The major constituents of a proposition p are those in terms of which its truth conditions are directly
defined. Sometimes a major constituent of p itself has constituents that are sub constituents of p in a weaker
sense; they are elements that have to be cognitively accessed in the process of identifying, and putting to
use, the major constituent of which they are constituents. The constituents of the argument of defending in
proposition (3a) are sub constituents of that proposition, but not of proposition (3b); the functions that are
constituents of M are sub constituents of proposition (9c), but not of proposition (9b). See chapters 8 and 9
of The Analytic Tradition, Vol. 1 for further discussion.
and (3b)—with “M” in (8b) playing the role of “logicism” in (3b). Applying that lesson
here, we see that one who entertains proposition (9c), but not one who entertains prop-
osition (9b), must think of fthe, and of g, and combine them into a function-argument
structure to be used to determine the predication target. However recognizing this
difference is not enough, since nothing I have said up to now explains how proposi-
tion (9b) can be true (which it must be), while proposition (9c) is false. Since the two
propositions have the same major constituents, and since a difference in the way that
one of those constituents is cognized in the two propositions can’t, by itself, affect their
truth values, they must also differ in structure.
The structure of a proposition is the manner in which its constituents are related to
one another. Since propositions are cognitive event types, the structural relationships
in which their constituents stand to each other are not relationships in which things
occupying certain positions in an n-tuple stand to those occupying other positions;
nor are they relationships that certain nodes in an abstract tree structure bear to other
nodes. Although nothing prevents using formal constructions of these or other sorts
to model propositional structures, the structures being modeled are something else.
The structural relationships between the constituents of a proposition are given by the
roles the constituents play in the sequence of cognitive operations performed by an
agent who entertains it—roles like being predicated (of certain things), being targets (of
certain predications), being applied (to certain arguments), being arguments (to which
certain things are applied), as well as being sub constituents of larger constituents which
may themselves play these roles, or of propositions that are constituents of larger prop-
ositions. This is the sense in which propositions (9b) and (9c) must differ in structure.
The required difference is a difference in the sense in which the relational property
being what “the first line of Gray’s Elegy” means is predicated of its argument in the
two propositions. In (9b) it is directly predicated of the complex that is the meaning
of the description; in (9c) it is indirectly predicated of whatever is determined by that
complex (which is the referent of the Fregean description). In order for the relation
direct predication to hold of an agent A (who entertains a proposition p), a property
F (to be predicated of something) and an item x (of which F is predicated), A must
have x in mind as the thing to be represented as having F. By contrast, the indirect
predication relation holds between A, F, and an item x which is the kind of thing (e.g. a
function-argument complex) that determines something else (e.g. a value). In order for
this relation to hold, A must have x in mind, and use it to represent whatever, if any-
thing, is determined by x as having F. The direct predication is veridical iff x has F; the
indirect predication is veridical iff there is something uniquely determined by x and
that thing has F.
Let “PredD” and “PredI” express the relations direct predication and indirect predica-
tion, respectively, and let “T” be a schematic letter to be replaced by a singular term.
When I say something of the form (10), the proposition I express is one that directly
predicates PredD of the triple consisting of A (or p), the property so-and-so, and the
referent, if any, of the term replacing “T”.
34
Remember, to predicate a property of something one doesn’t have to think that it really has the property.
9d. The denoting complex identified by Soames in the final section of chapter 6 of
New Thinking about Propositions is what “the first line of Gray’s Elegy” means.
15a. Proposition (9d) directly predicates being what “the first line of Gray’s Elegy”
means of the denoting complex identified by Soames in the final section of
chapter 6 of New Thinking about Propositions.
15b. Proposition (9d) indirectly predicates being what “the first line of Gray’s Elegy”
means of the denoting complex identified by Soames in the final section of
chapter 6 of New Thinking about Propositions.
The distinction between direct and indirect predication replaces my earlier univocal
characterization of predication as being a 3-place relation between an agent, a property,
and (as I put it) “a content.” Though indirect predication has some of the features I men-
tioned in connection with that characterization, neither direct nor indirect predication
has all of them. Thus, my earlier use of “predicate” must be recast in terms of these two
new notions. To that end, one should understand claims made in previous sections by
sentences of the form (16a) in which “T” is replaced by a Fregean definite description,
or other function-argument singular term, as making the claim expressed by (16b).
16a. Agent A (or proposition p) predicates the property so-and-so of T
b. Agent A (or proposition p) indirectly predicates property so-and-so of T
In all other cases, claims made by sentences of the form (16a) that I have used earlier
should be understood as claims involving direct predication.
This reconstrual requires making explicit something I have so far taken for granted
without comment. When an n-place predicate is paired with n arguments—some of
which may be Millian and some non-Millian—we must think of the predication as
proceeding in stages. This technique, familiar from Montague, treats the proposition
expressed by a sentence of the form
(17) A loves B
as arising first by combining the two-place relation loves with the content/referent of
the term replacing “B,” and then predicating the resulting one-place property of the
content/referent of the term “A.”35 When “B” is replaced by a Millian singular term the
content and referent of which is x, the resulting one-place property is loving x, which
may then be predicated directly, or indirectly, of the referent or content of the term
that replaces “A,” depending on whether that term is Millian or non-Millian. When “B”
is replaced by a non-Millian singular term—e.g. something the content of which is a
complex consisting fthe combined with an argument g—the resulting one-place prop-
erty is loving whomever is the value of fthe at g—which may, of course, also be predicated
35
Richard Montague, “The Proper Treatment of Quantification in Ordinary English,” in J. Hintikka
and J. Moravcsik, eds., Approaches to Natural Language, Dordrecht: Reidel, 1973; reprinted in Richmond
Thomason, ed., Formal Philosophy: Selected Papers of Richard Montague, New Haven: Yale University Press.
directly, or indirectly, of the referent or content of the term that replaces “A.” Thus the
operation, call it “reduction,” that maps an n-place relation plus an argument to the rel-
evant n-1 place relation subdivides into direct and indirect reduction, on analogy with
direct and indirect predication.
Although this sketch of propositions as cognitive event types is far from complete,
I hope to have given some idea of its promise and flexibility. What makes the conception
natural for resolving Russell’s problem involving (9b,c) is that the distinction required
to solve the problem—between what an agent who entertains proposition (9b) aims
to represent as having a certain property vs. what an agent who entertains proposition
(9c) aims to represent as having that property—is a cognitive difference encoded in the
cognitive acts that provide the structure of the event types with which propositions are
identified. By contrast, what made the problem seem insoluble to Russell in 1905 was
his conception of propositions as platonic objects the intentional properties of which
are prior to, and independent of, the agents who entertain them. Thinking of proposi-
tions in this way, and asking himself the question “What does M do in the proposition
expressed by (9c) that it doesn’t do in the proposition expressed by (9c)?” he naturally
answered “Nothing!”—which led him astray. Although the purely platonic conception
of propositions has lasted for a very long time, we are now, I hope, beyond that. Armed
with a more accurate conception of propositions, we pose the crucial question differ-
ently—“What do agents use M to do when entertaining the proposition expressed by
(9c) that they don’t use it to do when entertaining the proposition expressed by (9b)?”
Since the answer to this question is obvious, what had been a problem ceases to be.36, 37
36
I return to the question of whether sentences (9b) and (9c) are ambiguous. Sentence (9b) is not; it can’t
express the false proposition that (9c) does because entertaining the latter requires one to cognize the con-
stituents of M, whereas no proposition expressed by sentence (9b) does. Nor does it make sense to suppose
that (9b) expresses a proposition the truth conditions of which are the same as those of the false proposition
expressed by (9c) even though it doesn’t require cognizing the constituents of M. It is less clear whether (9c)
has a reading in which it can express the true proposition that (9b) unambiguously expresses. People do say
things like “red” means red, and “on the counter” means on the counter. Whether these are legitimate uses or
mistakes is not entirely clear to me. But if they are correct, then there may be a similar use of (9c) in which it
expresses a truth—indeed, the proposition that results from adding to the proposition expressed by (9b), the
requirement that the constituents of M be cognized.
37
I am indebted to Brian Bowman for helpful comments on this chapter.
1
The demand that an explanation be given of how/why propositions have truth conditions actually goes
all the way back to King [1995], which defended an earlier version of the view defended in King [2007].
his explanation of exactly how they manage to have truth conditions differs signifi-
cantly from mine. Despite my sympathy to his overall approach, my purpose here is to
criticize Soames’s account of propositions and his explanation of how they manage to
have truth conditions.
Soames begins with the notion of the mental act of predication, which he takes to
be primitive.2 However, by way of illustration, if an agent perceives an object o as red,
and so has a perceptual experience that represents o as being red, the agent predicates
redness of o.3 Similarly, if an agent “thinks of ” o as red4, or “form[s] the nonlinguis-
tic perceptual belief that o is red.”5 For Soames, predicating redness of o does not
amount to believing that o is red. To believe that o is red, one must predicate redness
of o and do something like endorse the predication.6 Unfortunately, it is hard to say
precisely what predicating amounts to since the notion of predicating is primitive for
Soames.7
An agent predicating redness of o is an event token. Of course there may well be
many event tokens of agents predicating redness of o by an agent perceiving it as
red, an agent thinking of it as red, a different agent thinking of it as red and so on.
Soames claims that the proposition that o is red is the event type of an agent predicat-
ing redness of o. Other more complex propositions are identified with event types of
agents performing sequences of primitive mental acts on the various constituents.8
One entertains a proposition by performing these sequences of acts on the relevant
constituents.
With only this much of Soames’s view on the table, I can raise the first problem with
his account. But before turning to that let me note as an aside that it is odd that Soames
appears to assume with little comment or argument that the contents of perceptual
experiences are expressible by sentences of natural languages.9 This claim is contro-
versial among those who work on perceptual experience. Some deny that perceptual
experience has content at all. Some allow that it has content, but hold that the content is
not propositional. And some allow that perceptual experience has propositional con-
tent, but deny that the content is expressible by means of sentences of natural language.
Hence, it is a bit surprising that with so little comment or argument Soames adopts
the view that the perceptual experience has contents that are expressible by natural
language sentences.
2
Soames [2010] p. 81
3
Chapter 6 of present work p. 96
4
Soames [2010] p. 103.
5
Soames [2010] p. 81
6
Chapter 6 of present work p. 97
7
Soames ends up with two notions of predication: direct predication and indirect predication. See
Chapter 6 pp. 120–123.
8
Hence Soames needs a number of primitive mental acts beyond predication. See [Soames 2010a] p. 115,
122; and Chapter 6 of present work pp. 97–99
9
Chapter 6 pp. 93–94
In any case, Soames claims that one way of entertaining the proposition that o is
red is to see o as red.10 Having the perceptual experience of seeing o as red consti-
tutes predicating redness of o. When I think the thought that o is red, or understand
the sentence “o is red,” I likewise entertain the proposition that o is red and thereby
predicate redness of o according to Soames. This means that in all three cases—per-
ceiving, thinking, and understanding a sentence—the same primitive cognitive act
of predication is occurring. Indeed, since the propositions that are the contents of
perceptual experiences in general involve primitive mental acts all of which also
occur in entertaining propositions in thought and understanding sentences of natu-
ral language, Soames is committed to the claim that all the primitive cognitive acts
that occur in perception when one entertains a proposition by seeing things a cer-
tain way, also occur in entertaining the proposition in thought, and understanding a
natural language sentence that expresses that proposition. I have three related con-
cerns with this. First, what sort of claim is the claim that e.g. the same cognitive act of
predication occurs in perceptual experience, thinking and understanding sentences
of natural language? It sounds like a (high level) empirical claim; and I am not sure
how else to take it. But then the claim seems highly speculative and I wonder what
evidence there is for it. Second, the claim that the same cognitive act of predication
occurs in thinking and in perception seems quite dubious. In perception, except in
some very limited range of cases, I have no choice as to what things appear to have
what properties. In thinking, I have complete freedom in what properties and rela-
tions I think about things as possessing. This would seem to me evidence right off
that the cognitive “acts” by means of which properties are attributed to objects in
perception and the cognitive acts by means of which properties are attributed to
objects in thought are very different kinds of acts, contrary to what Soames claims.
Perhaps Soames would admit that at this lower level the cognitive acts in percep-
tion and thinking differ, but that at a higher level of abstraction, the same cognitive
act is being performed. But then I am back with the first worry: this claim seems
highly speculative and I wonder what evidence there is for it. Finally, the claim that
all the primitive cognitive acts that occur in perception as a result of which percep-
tual experiences have propositional content also occur in entertaining propositions
in thought and understanding sentences of natural language entails a degree of cog-
nitive unity and convergence in perception, thought and language that seems to me
implausible.
A second problem with Soames’s view is that it requires commitment to idiosyn-
cratic metaphysical views that many, and probably most, would reject. Since Soames
identifies propositions with event types, in asking whether all the propositions exist on
his view that intuitively should exist, we are asking whether all the relevant event types
10
Soames [2010] p. 81
exist. Soames provides two principles regarding the conditions under which event
types qua propositions exist:
1. If an event type E has instances that exist, then E exists.11
2. if (i) R is an n-place property for which there have been events in which an agent
predicates R of things, and (ii) o1. . .on are objects for each of which there have been
events in which an agent refers to it, then (iii) there exists a proposition p which is
the (minimal) event type of referring to o1. . .on and predicating R of them—even if
no one has ever performed that predication, and hence there exist no instances of p.12
These principles seem plausible enough. However, there are lots of objects that have
never been referred to and properties that have never been predicated of anything.
Principles 1 and 2 don’t secure the existence of event types qua propositions involving
such objects and properties. Hence, on Soames’s account such propositions don’t exist.
But surely, many such propositions exist and are true (e.g. the proposition that some
never-referred-to molecule o is a molecule, which Soames himself considers). Soames
responds to this worry by saying that this would only be a problem if propositions
had to exist to be true. But they don’t!13 These propositions that have as constituents
never-referred-to objects and never-predicated relations are merely possible and so
don’t exist on Soames’s view. But he holds that many non-existent things possess prop-
erties. Soames gives as examples an individual having the property of being dead, being
referred to by me and being admired by someone. He also gives the example of Plato
and Aristotle instantiating the relation of non-identity. So, according to Soames, there
is nothing to prevent non-existent, merely possible propositions—like the proposi-
tion that the never-referred-to molecule o is a molecule—from having the property of
being true.
I think it is fair to say that most philosophers would reject Soames’s alleged exam-
ples of non-existent things possessing properties (e.g. four dimensionalists would
deny that Nixon possessing the property of being dead is an example of a nonexist-
ing thing possessing a property; nihilists would deny that a sentence like “Nixon is
dead” expresses (or seems to express) a truth because a nonexistent thing possesses a
property). Relatedly, I think it is also fair to say that most philosophers reject the view
that non-existents may possess properties. It is precisely because they wish to reject
such a view that most actualists are so-called serious actualists and most presentists
are serious presentists. (The former are actualists who hold (roughly) that if a thing
o possesses a property at a world w, it exists at w; the latter are presentists who hold
(roughly) that if a thing o possesses a property at a time t, it exists at t.) So adopting
Soames’s view of propositions requires one to hold a metaphysical view that most
philosophers reject.
11 12 13
Chapter 6 p. 101 Chapter 6 p. 102 Chapter 6 p. 103
A third problem with Soames’s account is that it appears to allow there to be too
many propositions. More specifically, it distinguishes propositions that intuitively
shouldn’t be distinguished; and it is forced to posit the existence of propositions that
intuitively don’t exist. Taking these points in turn, Soames claims that propositions
are event types in which an agent performs a sequence of cognitive acts such as predi-
cating a property of an object, applying a function to an argument, and performing the
operation of reduction on a relation and an argument, whereby an n-place relation and
an argument are mapped to the relevant n-1 place relation.14 Obviously, on this view
different sequences of cognitive acts yield different event types and so different propo-
sitions. But this forces advocates of the view to claim that there are more propositions
than there intuitively are. Consider this sequence of cognitive acts: an agent first per-
forms the cognitive action of reduction by combining the argument Juliet with the
2-place relation loves (on the second “argument place”), yielding the 1-place property
___loves Juliet. The agent then predicates this (directly, in Soames’s sense) of Romeo.
That’s the (or a) proposition that Romeo loves Juliet. Call this proposition Romeo 1. But
now consider this different sequence of cognitive acts: an agent combines Romeo with
the relation loves by reduction (on the first “argument place”), yielding the property
Romeo loves__. The agent then predicates this of Juliet. That also looks like the propo-
sition that Romeo loves Juliet. But it is a different sequence of cognitive acts than the
first, and so is a different proposition. Call this proposition Romeo 2. Finally, consider
this sequence of cognitive acts: an agent predicates loves of the pair Romeo, Juliet in
that order. This again looks to be the proposition that Romeo loves Juliet. But it too is
a different sequence of cognitive acts than the first two. Call this proposition Romeo 3.
So how does Romeo love Juliet? Let’s count the ways. It looks like there are three dis-
tinct propositions, because of three distinct event types of agents performing sequences
of cognitive acts, that are all propositions to the effect that Romeo loves Juliet. This is
surely implausible. Soames might try to block there being the proposition Romeo 2
by claiming that the cognitive act of reduction when applied to an n-place relation
and an argument has to give us back the n-1 place relation that results from the argu-
ment “saturating” the nth argument place of the relation. So if I perform reduction on
Romeo and loves, it can only yield the property loves Romeo. To this I can only respond
that I find that I can in thought combine Romeo with the relation loves yielding the
property Romeo loves__ and that I can then predicate that property of Juliet.15 Soames
also might try to block there being a proposition Romeo 3 by denying that I can predi-
cate a 2-place relation of Romeo and Juliet, in that order. Again, I can only report that
I can do this in thought. So here Soames’s account delivers three propositions where it
14
Soames [2010] p. 115, 120; and Chapter 6 pp. 99–100 and especially 123–124.
15
Frege seems to agree. Concerning thoughts expressed by sentences containing multiple names, Frege
[1979] writes “A thought that is put together in this way is just what traditional logic calls a singular judge-
ment. We must notice, however, that one and the same thought can be split up in different ways and so can be
seen as put together out of parts in different ways.” (pp. 201–202).
should deliver one. Similar remarks apply to relations of higher adicity and arguments,
where we will get an even greater proliferation of “the same” proposition.
A second case of Soames’s theory delivering propositions where it shouldn’t
concerns his account of the de se. On Soames’s view, a proposition can require to
be entertained that some of its constituents be cognized in a particular way or it can
fail to do so. So for example, what Soames calls the de re proposition that John Perry
is the messy shopper is entertainable simply by predicating being the messy shopper of
Perry, where there are no constraints on how Perry is thought of (though one must
refer to him directly). What Soames calls the de se counterpart of this proposition is
entertainable by predicating being the messy shopper of Perry while thinking of Perry
in “the first person way.”16 Only Perry can entertain this proposition. So entertaining
this proposition requires one to cognize Perry in a special way. Apparently worried
that the claim that propositions can require in order to be entertained that things be
cognized in a certain way could lead to there being all manner of strange proposi-
tions, Soames writes:
Surely, there are limits on what cognitive acts or operations propositions can encode. Thinking
of a certain tune while predicating redness of an object is a (complex) cognitive activity of some
sort, as is predicating a relation of a pair, while feeling affection toward its members. Presumably,
the event types of doing these things are not propositions because one of their cognitive compo-
nents is orthogonal to how the agent represents things to be.17
Later Soames says that whether a proposition can require that something be
thought of in a given way in order to be entertained depends on whether the way of
thinking of the thing is representational: whether it is a matter of the thing being
represented in a certain way. In entertaining the de se proposition that Perry is a
messy shopper, the required way of thinking about Perry is the first person way.
Since this is a matter of representing Perry in a certain way, this proposition is
kosher. However, in predicating being smart of Annie while feeling affection for
her, my feeling affection for Annie is apparently for Soames not representing her
in a certain way. Hence, there is no proposition in which being smart is predicated
of Annie, and which to be entertained requires me to feel affection for Annie in so
doing. That all sounds good.
The problem is that there are all kinds of ways of thinking of things that constitute
representing them as being a certain way, so the way of thinking/representing meets
Soames’s criterion for being allowably “encoded” in propositions, where it is just very
implausible to suppose that there are propositions such that to be entertained requires
thinking of things in those ways. Hence Soames’s account commits him to the exist-
ence of propositions that it is implausible to think exist. For example, Soames seems
committed to a proposition the entertaining of which requires me to predicate redness
16 17
Chapter 6 pp. 110 Chapter 6 p. 118
18
Don’t say that there is such a proposition: that o is red and inhabits a world in which water is H2O. On
Soames’s view, just as we want to distinguish the proposition that I am the messy shopper and am thinking
of myself in the first person way from Soames’s (alleged) de se proposition that I am the messy shopper, so
we want to distinguish the proposition the entertaining of which requires me to think of o as inhabiting a
world in which water is H2O while I predicate redness of o from the proposition that o is red and inhabits a
world in which water is H2O. On Soames’s view, the former does not predicate a conjunctive property of o;
the latter does.
19
Soames [2010] pp. 6, 32, 55–56, 64, 106–107; Chapter 6 pp. 96–97
20
Soames [2010] p. 107. Similar remarks occur on pp. 96–97 in Chapter 6
21
See p. 104 for Soames’s [2010] explanation of why the proposition that snow is white is true at a world w
where it has no tokens, where Soames explicitly appeals to possible tokens of the proposition qua event type
that snow is white. See also Chapter 6 p. 96 where the same appeal is made.
story—where by “complete” I mean one that answers all questions relevant to a contextually
determined inquiry.22
Soames’s analysis of possible worlds obviously presupposes that there are proposi-
tions with truth conditions. Indeed, Soames even claims that the basic propositions
mentioned in the quotation above are constituents of the property that he identifies
with a possible world in that same quotation.23 But Soames’s explanation of how/why
propositions have truth conditions presupposes that there are possible worlds. So for
Soames, that there are propositions with truth conditions presupposes that there are
possible worlds (since the explanation for why at least some propositions have truth
conditions appeals to merely possible tokens of them). But that there are Soamesian
possible worlds presupposes that there are propositions with truth conditions (since
these are constituents of the property that is the possible world—it is the property of
making these propositions true, and surely that presupposes that there are these prop-
ositions with truth conditions). That is circular.
More generally, that Soames’s explanation of how/why propositions have truth con-
ditions appeals to possible worlds precludes anyone who adopts Soames’s view from
constructing possible worlds by using propositions in any way (e.g. as sets of propo-
sitions as in Adams [1981]). Thus, endorsing Soames’s view here precludes adopting
types of views in another area that many find attractive.
As I’ve indicated several times, Soames agrees with me that some explanation must
be given of the fact that propositions have truth conditions and so represent things as
being a certain way. The final difficulty I’ll raise for Soames’s account of propositions is
that his explanation of how/why propositions have truth conditions fails.
Soames states his explanation of why propositions represent/have truth conditions
as follows:
This suggests that the proposition that o is red is simply the minimal event type in which an arbi-
trary agent predicates being red of o. This event-type is representational because every conceiv-
able instance of it is one in which an agent represents something as being a certain way. What it
represents is what is representationally common to all such instances. Since every such instance
is one in which o is represented as being red, we speak, derivatively, of the proposition itself rep-
resenting o as red.24
22
Chapter 6 pp. 113–114
23
Chapter 6 p. 115
24
Chapter 6 p. 96
25
Chapter 6 p. 96
Soames doesn’t say so here, but he claims that instances/tokens of the event type of
predicating redness of o are things that “inherently” have truth conditions and so rep-
resent.26 This is the first step in his explanation of how/why propositions qua event
types have truth conditions. For the second step, Soames claims that since every event
token of the event type predicating redness of o has truth conditions, we say—deriva-
tively (see the last sentence of the first quotation above)—that the event type itself has
truth conditions. I’ll be challenging both the steps of Soames’s explanation here.
The first step of Soames’s explanation is the claim that event tokens of agents predi-
cating redness of o inherently have truth conditions. However, Soames gives no
argument for this claim nor does he explain how/why such event tokens have truth
conditions inherently. The most he says is that in predicating redness of o, an agent
represents o as red.27 But how do we argue from that claim to the claim that the event
token of the agent predicating redness of o itself represents o as red and so is true iff o is
red?28 Soames doesn’t say; and it just doesn’t seem to follow at all. In general, when an
agent bears a relation R to something o at a time, the event token of the agent bearing
R to o does not itself bear R to o. If I hug Annie, the event token of my hugging Annie
doesn’t hug Annie. So why, from that fact that an agent represents o as red (by predicat-
ing redness of it), would it follow that the event token of the agent representing o as red
itself represents o as red? In any case, Soames offers no argument that event tokens of
predicating redness of o have truth conditions.
Further, the claim that the event tokens in question inherently have truth condi-
tions is mysterious. I can see how event tokens could have truth conditions in virtue of
agents interpreting them in certain ways. But how could an event token inherently have
truth conditions? How could an event token have truth conditions by its very nature?
That seems as mysterious as the claim that propositions are sui generis abstract entities
that have truth conditions by their natures and independently of minds and languages,
which I and Soames himself both reject as unintelligible. Given that it is a mystery how
event tokens could have truth conditions inherently, in the absence of any argument
for the claim that the event tokens of agents predicating redness of o inherently have
truth conditions or any explanation of how this could be so, we should reject the claim.
26
See Soames [2010] p. 107. In the final sentence of the above quotation Soames appeals to the “represen-
tational features” of possible event tokens of the event type of an agent predicating redness of o. These repre-
sentational features presumably include the having of truth conditions.
27
“. . .what it is to entertain it [the proposition that o is red] is simply to predicate redness of o and so repre-
sent o as red.” Chapter 6, p 96. Here it is an agent doing the entertaining, predicating and representing.
28
Looking at the quotations from Soames, it might be thought that all he wants to claim is that in an event
token of an agent predicating redness of o the agent represents o as red and not that the event token of her predi-
cating redness of o does so as well. But given that o is represented as being red, something must be such that it
is true iff o is red (Soames often moves from the claim that something represents o as red to the claim that it is
true iff o is red). When Soames talks about representational features of event tokens, he suggests that the thing
that is true iff o is red is the event token of predicating redness of o. And in Soames [2010] he is quite explicit in
claiming that event tokens in which agents predicate things of other things are representational (p. 107).
Finally, what evidence there is suggests that the event tokens in question do not
have truth conditions inherently. Suppose Vicky is perceiving o as red and that this
event token is quite salient to us. If I say, nodding at Vicky “What is occurring is true
(false)” or “The event of Vicky seeing o as red is true (false),” this just sounds like a
category mistake. (Indeed, the second sounds like a misguided attempt to say that the
event occurred.) And the same is true of any attempt to predicate truth or falsity of an
event token of an agent predicating redness of o.29 But if Soames is right that such event
tokens are inherently things with truth conditions, why would predicating truth or fal-
sity of them sound so anomalous as to seem like a category mistake? I conclude that
what evidence there is suggests that event tokens in which agents predicate redness of o
do not have truth conditions inherently. Hence the first step in Soames’s explanation of
how/why propositions have truth conditions fails.30
Recall that the second step of Soames’s explanation as to how/why propositions have
truth conditions was to say that the event type of an agent predicating redness of o,
which Soames identifies with the proposition that o is red, itself has truth conditions
derivatively, because every possible event token of an agent predicating redness of o
has truth conditions. Of course, we have seen reason to reject the claim that such event
tokens do inherently have truth conditions. But suppose we were to grant that they did.
Would it follow that the relevant event types derivatively represent and so have truth
conditions as Soames claims?
I think there is good reason to think not, at least in any substantive sense, even by
Soames’s lights. Further, there is reason to think the claim that event types have truth
conditions in any substantive sense (derivatively or otherwise) is false, again, even by
Soames’s lights.
Taking the former point first, suppose that event tokens of predicating redness of
o represent o as being red. Does it follow that event types of predicating redness of o
derivatively represent o as being red? Only in a very stipulative and non-substantial
way. Soames tells us that the sense in which the event type of predicating redness of o
derivatively represents o as red is that in every instance of the event type o is represented
as being red. That means that we could in an exactly similar sense say that the event
type of an agent hitting Alan derivatively touches Alan because in every instance of it
29
E.g. Suppose I ask Vicky to think of o as red. As she is doing so, if I say “The event of Vicky thinking of o
as red is true (false).” or “The event Vicky is now bringing about is true (false).” again this sounds like a cat-
egory mistake.
30
Someone might attempt to use this sort of argument against my account of propositions. Doesn’t predi-
cating truth or falsity of the facts that I claim are propositions sound anomalous too? Doesn’t that show that
the facts I claim are propositions don’t have truth conditions? But I have an explanation of this not available
to Soames. I can be acquainted with the fact that is the proposition that Rebecca swims qua fact or qua prop-
osition. When I am acquainted with it as just another fact in the world, I have no reason to think it has truth
conditions since facts generally do not have truth conditions. However, when I am acquainted with this fact
qua proposition that Rebecca swims, I cannot fail to see that it has truth conditions, (this is discussed in King
[2007] pp. 50–52). Note that this explanation works precisely because facts do not have truth conditions
inherently, but rather are endowed with truth conditions by us. Since Soames holds that his event tokens
inherently have truth conditions, this sort of explanation is not available to him.
Alan is touched. Clearly there is no substantive sense in which the event type of hitting
Alan touches Alan. But then there is also no substantive sense in which the event type
of predicating redness of o represents o as being red.
This is closely related to another reason why, even by Soames’s lights, event types
don’t represent, and so have truth conditions in anything but a nonsubstantial, stipu-
lated sense even if we grant that event tokens do. Recall why Soames thinks that event
tokens of agents predicating redness of o have truth conditions and so represent o as
red. The reason is that by predicating redness of o, the agent represents o as red and
thereby we have something that is true iff o is red.31 So that something is predicated of
something “in” an event token is the key to its having truth conditions. The problem is
that nothing is predicated of anything “in” the event type of an agent predicating red-
ness of o. This is easy to see by supposing this event type exists at a world without agents
and so has no instances. It will still be the case that nothing is predicated of anything at
that world. But then these event types, unlike their tokens, lack exactly what is required
for the tokens to represent and have truth conditions according to Soames: predicating
something of something. Perhaps Soames would say that the event type of predicating
redness of o derivatively predicates redness of o, and so has truth conditions, in the
sense that in all its instances, redness is predicated of o.32 But we have already seen how
non-substantial and stipulative this sense of derivative predication is bound to be. The
crucial point is that the sense in which the event type of predicating redness of o deriva-
tively predicates redness of o is simply that in each of its instances redness is predicated
of o. The fact remains that the event type itself predicates nothing of anything. But then
it remains true that the event types lack what Soames thinks is the key to having truth
conditions: predicating something of something. That in instances of the event type
predication occurs, and in this sense something is derivatively predicated of something
in the event type, is quite beside the point. Hence, by Soames’s own lights, the event
types he takes to be propositions lack what is essential for having truth conditions.
Putting aside Soames’s argument for the claim that the event type of predicating red-
ness of o has truth conditions, let’s turn to the second point mentioned above: there
is reason to think that the claim that the relevant event types have truth conditions is
false, even by Soames’s lights. The problem is that, as was the case with event tokens of
predicating redness of o, what independent evidence there is suggests that the event
31
Soames makes clear that predicating is the key to these event tokens having truth conditions, and so
representing, when he writes “. . .what it is to entertain it [the proposition that o is red] is simply to predicate
redness of o, and so represent o as red.” Chapter 6, p. 96 my emphasis. Soames [2010] writes “When we see an
object o as red, we predicate redness of o. It is in virtue of this that our perceptual experience represents o as
being red.” (p. 81, italics my emphasis). Soames is talking about the so-called deflationary theory of proposi-
tions here and not his own “cognitive realist” theory. But the alleged insight that the presence of predicating
is the key to an event token having truth conditions is preserved in his theory, as the previous quotation
makes clear.
32
See Soames [2010] pp. 100–101 where he is discussing whether the act type of predicating redness of o
predicates anything of anything.
type of predicating redness of o does not have truth conditions. To see this, note first
that we can speak of event types as occurring or happening. Thus, if I see a pedestrian
get hit by a car at the corner of Amsterdam and 87th, I can say “That happens every
week.” and mean that the event type of a pedestrian getting hit at Amsterdam and 87th
has instances every week. So the demonstrative “that” here picks out an event type.33
Similarly, for locutions like “what just occurred” (e.g. my electricity goes off and I say
“what just occurred happens every day at this time.”). So by indicating a token of an
event type and using expressions like those just mentioned, I can talk about the event
type. And by talking about an event type happening, I am talking about its having
instances.
Now according to Soames, entertaining a proposition is simply tokening an event
of the type that is the proposition (by performing the acts of predication involved in
tokening the type).34 So suppose I ask Vicky to entertain the proposition that arithme-
tic is incomplete, if consistent, and she complies. Hence she tokens an event that is of
the type that is the proposition that arithmetic is incomplete, if consistent (roughly, by
predicating being incomplete if consistent of arithmetic). I now say “What just occurred
is true.” This is bizarre to the point of being incoherent. But if Soames is right, this
should sound fine and be true. For the expression “what just occurred” should be capa-
ble of being used to talk about the event type that Soames claims is the proposition that
arithmetic is incomplete if consistent. And of course since Godel proved this proposi-
tion, it is true.35
Finally, evidence similar to the above strongly suggests that the event types that
Soames identifies with propositions are not propositions.36 As before, imagine that we
ask Vicky to entertain the proposition that arithmetic is incomplete if consistent and
she complies by tokening the event type that Soames claims is the proposition. Again,
we should be able to talk about the event type using the expression “what just occurred.”
I say “What just occurred was proved by Godel.” Again, this is incoherent and surely
is a category mistake. But again, if Soames were right this should be true: again, Godel
did prove the proposition that arithmetic is incomplete if consistent. Similar remarks
apply to predicating of the relevant event types many things that can sensibly be predi-
cated of the propositions Soames identifies with these event types.37 Further, if we
predicate of propositions properties that are had by the event types that Soames claims
are propositions, again the predications are bizarre to the point of incoherence: “What
33
If it picked out the event token, I would have asserted the absurdity that the relevant event token happens
every week.
34
Soames [2010] p. 106
35
Note that even if “what just occurred” picked out the event token of Vicky entertaining the proposition
that arithmetic is incomplete, if consistent (as it surely could), the sentence “What just occurred is true.”
should still be true in this situation on Soames’s view since according to him, the event token in question
is true.
36
Of course, the above evidence that the relevant event types don’t have truth conditions is also evidence
that they aren’t propositions. My point here is that there is additional evidence that they are not propositions.
37
For example, “What just occurred entails that Hilbert’s program is impossible.”
Godel proved occurred twice today.”; “What Godel proved just happened.”; “What
Godel proved occurred all over the world today.” All of this strongly suggests that the
event types identified by Soames are not propositions.
The arguments just given that Soames’s event types don’t have truth conditions and
are not propositions amount to noting the following. When the events types are picked
out with language appropriate to picking out event types, it sounds incoherent to pred-
icate truth or falsity of them. It is also incoherent to predicate of them other things that
are coherently and truly predicated of propositions. The explanation of the incoher-
ence is that the event types are neither true nor false, nor are they propositions. And
when we pick out propositions using language appropriate to picking out propositions
(“the proposition that. . .”), it is incoherent to predicate of them things that are coher-
ently and truly predicated of event types. The explanation of this incoherence is that
propositions are not event types. Now Soames uses arguments of exactly this sort to
refute the view that propositions are act types. Soames [2010] writes
Whereas it is perfectly coherent to say that predicating brilliance of John is what I just did or that
the proposition that John is brilliant is false, it would be incoherent to say “*The proposition that
John is brilliant is what I just did.” or “*What I plan to do (when I plan to predicate brilliance of
John) is false.” Moreover, although it is correct to say that I believe, and Godel proved, the propo-
sition that arithmetic is, if consistent, incomplete, it would be absurd to say “What I believe,
and Godel proved, is something I just did.” The source of the absurdity is not hard to locate. Act
types—like kissing Martha or predicating incompleteness of arithmetic—are either themselves
a certain kind of property, or something closely akin to properties. As such, they are not the
kinds of things that have truth conditions.38
Hence, if Soames thinks that his arguments refute the view that propositions are act
types, he should think that my arguments refute the view that propositions are event
types.39
Having now explained and criticized Soames’s account of propositions, I want to
reiterate the important points of convergence between his view and mine mentioned at
the outset. Soames and I both claim that classical conceptions of propositions, accord-
ing to which they represent things being a certain way, and so have truth conditions
independently of minds and languages and by their very natures, must be rejected. The
problem is we have no idea how anything could do this, and so such theories make the
fact that propositions have truth conditions a complete mystery. It is important to see
that if Soames and I are right about this, it takes virtually every theory of propositions
except his and mine off the table (though see the last paragraph of my comments on
Speaks’s view below). Soames and I also think that propositions have truth conditions
in virtue of things agents do. This makes certain facts about agents explanatory prior
38
Pp. 101–102
39
Couldn’t an argument like this be given against my view of propositions as well? See my discussion in
note 30 of why I think I can explain the anomalousness of predicating truth and falsity of the facts I claim are
propositions in a way that is consistent with my claim that they are propositions.
40
Speaks indifferently goes back and forth between saying the proposition that Amelia talks is the prop-
erty being such that Amelia instantiates talking and the property being such that Amelia talks. I will as well.
41
To paraphrase Wang Chung.
42
This simple view really isn’t Speaks’s because of propositions like the proposition that Rick Santorum
doesn’t exist. Pick a world w where Santorum doesn’t exist. If this world were actual, arguably the property
being such that Rick Santorum doesn’t exist itself wouldn’t exist. But then arguably this property wouldn’t be
instantiated and so the proposition that Rick Santorum doesn’t exist wouldn’t be true. However presumably
we want the proposition that Rick Santorum doesn’t exist to be true at w. Hence a different account is called
for here.
that one believes instantiates the property. So according to Speaks, two people may
believe Amelia talks, where one believes o instantiates being such that Amelia talks and
the other believes that o’ instantiates being such that Amelia talks, where o≠o’ (and the
first has no beliefs involving o’ and the second has no beliefs involving o). Speaks seems
to think that the normal case might involve believing-instantiated the property being
such that Amelia talks because of believing that the actual world instantiates it. Below
I’ll cast doubt on this claim.
Speaks also holds that there are other cognitive relations we bear to properties
that are not propositions. (Here he wants to give an account of “first person” men-
tal states that others have called belief, desire, etc. de se.) Agents can bear the relation
of self-attribution to properties like being on fire. This gives us “first personal” men-
tal states of the sort discussed in the literature on de se belief. But note that Speaks’s
account here is not one of de se belief because the properties self-attributed are not
propositions and self-attribution is not belief. On Speaks’s account, I could believe Jeff
King is on fire, which for Speaks amounts to my believing-instantiated the property
being such that JK is on fire. Or I could take myself to be on fire, which for Speaks
amounts to self-attributing the property being on fire. The former is my standing in a
relation to a proposition; the latter is not. It is worth noting here that Speaks does not
really capture all the “first personal” mental states that one who goes in for this sort of
thing might want. First person desires for Speaks amount to “self-desiring” proper-
ties. So my first person desire to be a professional skier is my self desiring the property
being a professional skier (and this latter property is not a proposition).43 However, such
an account stumbles on the Nolan [2006] example of the “first person desire” not to
exist. For Speaks, this would be understood as self-desiring the property not existing.
Since this has to be understood as something like (first personally) wanting to possess
a property, and since I could not possess this property (for Speaks as for me, I must
exist to possess a property), in self-desiring not to exist I am doing something incoher-
ent on Speaks’s account. But intuitively, I am not.44
As to the semantics of verbs of propositional attitude, Speaks takes “believes” to
express the following two-place relation between a person and a property: ___takes to
be instantiated the property___ (similar remarks apply to other verbs of attitude).45 It is
important to see that for Speaks e.g. belief ascriptions do not by means of their seman-
tics ever assert that one self-attributes a property (similar remarks hold for desire
ascriptions etc.). Thus for Speaks, both “I believe JK is on fire” and “I believe I am on
fire” (uttered by me) express the proposition that JK stands in the believes-instantiated
relation to the property (proposition) being such that JK is on fire. Speaks suggests
that perhaps the proposition that JK self attributes being on fire can be pragmatically
43
Chapter 5 p. 85
44
There may be many things bad about desiring not to exist, but it being incoherent to do so is not one
of them.
45
Chapter 5 p. 85
conveyed by my uttering “I believe that I am on fire.” But Speaks is upfront about the
fact that he has no explanation of the mechanism by means of which this proposition
is pragmatically conveyed by uttering the sentence in question, while acknowledging
that an account of the mechanism is required in order to sustain the suggested prag-
matic approach.
Having sketched Speaks’s account, and before turning to critical remarks, let me say
that I think Speaks’s view has a lot to recommend it. Indeed, it is the best version of the
propositions-as-properties view known to me.
Let me turn to critical remarks.
My first and second objections concern features of Speaks’s account of proposi-
tional attitudes. Recall that Speaks claims that to believe that Amelia talks is to believe
that something is such that Amelia talks. I suggested above that Speaks should say
instead that to believe Amelia talks is to believe-instantiated the property being such
that Amelia talks. Here is why. If we say that believing that Amelia talks is believing
something is such that Amelia talks, what do we say about this latter belief? Applying
the account again, we get that believing something is such that Amelia talks is believ-
ing that something is such that something is such that Amelia talks. But now we must
again apply the account to this latter belief, which will be believing something is such
that something is such that something is such that Amelia talks. Once again, the
account applies to the latter belief and so on. To avoid this regress, we must understand
Speaks to be saying that to believe that Amelia talks is to believe-instantiated the prop-
erty being such that Amelia talks. (As mentioned above, this is a matter of bearing the
two-place believe-instantiated relation between individuals and (propositions) prop-
erties to the property being such that Amelia talks.) That is why in explicating his view
above, I took him this way.
Call a belief like believing-instantiated the property being such that Amelia talks a
general belief (since you are merely “taking the property to be instantiated” without
necessarily taking any particular thing to instantiate it). Now recall that Speaks claims
that we often have such a general belief because we believe of some particular o that o
instantiates being such that Amelia talks. Call this a specific belief (since you take a par-
ticular thing to instantiate the property being such that Amelia talks). So Speaks says that
sometimes we have the general belief that Amelia talks (by believing-instantiated the
property being such that Amelia talks) because we have the specific belief that o instanti-
ates being such that Amelia talks. But believing that o instantiates being such that Amelia
talks again on Speaks’s view amounts to believing-instantiated the property being such
that o instantiates being such that Amelia talks. This means that the so-called specific
belief that o instantiates being such that Amelia talks has really simply turned out to be
the general belief of believing-instantiated the property being such that o instantiates
the property of being such that Amelia talks. In short, contrary to what Speaks suggests,
all belief is general on his view: it is always a matter of believing-instantiated some
property (that is a proposition). It is important to appreciate how peculiar this view
is. It would be the analogue of holding that for non-proposition properties, we only
ever believe that something instantiates them. We never believe of particular things,
that they instantiate them. That is a quite strange view. At the very least, Speaks should
explain how we come to e.g. believe instantiated the property being such that Amelia
talks without believing that a particular thing instantiates the property.
Of course, as we saw, Speaks did hope to explain how one comes to believe-instantiated
the property being such that Amelia talks precisely by claiming that one often does so
because one believes that some particular thing o instantiates the property. He hoped
to explain having general beliefs as the result of having specific beliefs. But given what
I have just argued, that specific belief collapses into general belief on Speaks’s view,
the proposed explanation now looks unnatural and implausible. For given what I have
just argued, Speaks’s claim that we sometimes believe that something instantiates being
such that Amelia talks because we believe that some particular thing o does so ends up
in more official terminology as being the claim that we sometimes believe-instantiated
being such that Amelia talks because we believe-instantiated being such that o instanti-
ates being such that Amelia talks. But as Speaks himself admitted (p.c.), this order of
explanation doesn’t seem very natural when stated in the more official terminology
as we just did. This leaves Speaks without an explanation of how we come to simply
believe-instantiated the property being such that Amelia talks.
A second problem again concerns the propositional attitudes. As we just saw,
for Speaks believing must be understood as bearing the believes-instantiated rela-
tion to properties (that are propositions). Similarly, desiring will be understood as
desires-instantiated; assuming will be assumes-instantiated, and so on. So proposi-
tional attitudes for Speaks will be, roughly speaking, a matter of bearing various atti-
tudes towards the instantiation of certain properties.46 For lots of attitudes, that seems
quite plausible. But for some, it does not. I can consider, explain and understand the
claim that arithmetic reduces to logic and in none of these cases does it seem that my
attitude has anything to do with whether the alleged property being such that arith-
metic reduces to logic is instantiated. Take considering: when I consider the claim that
arithmetic reduces to logic I can do so without any attitude at all about the instantia-
tion of the property being such that arithmetic reduces to logic. I may simply have an
interest in really thinking about what the claim comes to. So here Speaks’s account of
the attitudes seems strained at the very least.
My third criticism is that it just does not seem as though the property being such that
Amelia talks is something that is true or false. To say that it is seems like some sort of
category mistake. Perhaps someone would respond that despite this, the property in
fact is (say) true. They might add that when we consider the claim that Amelia talks,
we take it to be true; and the “that” clause here designates the property in question.
So we do take properties like being such that Amelia talks to be true and false. The
problem with this is that properties are most transparently expressed by predicates. If
46
Chapter 5 p. 85
propositions really are properties as Speaks claims, why when we consider the predi-
cate “being such that Amelia talks” that allegedly expresses the proposition in question
are we not inclined to say that it expresses something true or false? It looks like Speaks
will have to claim that when we encounter a proposition qua property as the thing
designated by a “that” clause, we treat it as something that is true or false. But when we
encounter it as the thing expressed by a predicate, we don’t do so. This is made all the
more peculiar by the fact that, as indicated above, it is predicates that canonically and
transparently express properties. Hence, when we encounter a proposition qua prop-
erty as the thing expressed by a predicate, we will be more aware that it is a property.
If propositions are properties, why when we encounter them via the linguistic devices
that most make clear that they are properties do we precisely not want to treat them
as things that are true and false? Why must we encounter propositions via linguistic
devices that disguise the fact that they are properties (e.g. “that” clauses) in order to
treat them as things that are true and false? I don’t see what answer Speaks can give to
these questions.
Fourth, and related to the third worry, Speaks himself holds that not all properties are
propositions—things that are true and false and possess modal properties. For exam-
ple, Speaks holds that being red is a property but not a proposition.47 This seems right,
as being red does not seem to be something that is true or false. However, this raises the
question of which properties are propositions on Speaks’s view. Speaks admits that he
has no account of what distinguishes the properties that are propositions from other
properties. I think this is problematic for two reasons. First, absent such an account
we really haven’t been told what propositions are. We can’t identify propositions with
properties, because some properties aren’t propositions. And we can’t identify propo-
sitions with properties of type X, since Speaks admits he can’t give any account of what
type X is. So there is an important sense in which Speaks’s view leaves us without an
account of what propositions are. Second, not only do properties like being red not
seem to be propositions—things that are true and false and possess modal proper-
ties—but above I claimed that even the properties Speaks takes to be propositions, like
the property being such that Amelia talks, don’t seem to be things that are true and false
and possess modal properties. If in general it doesn’t seem like properties are the kinds
of things that can be true or false and bear modal properties, then we need to be told
why/how it is that certain properties can be true or false and bear modal properties. We
need to be told what is “special” about them such that they can do these things other
properties can’t do. But, again, this is exactly what Speaks is unable to do.
It seems to me this is deeply unsatisfying. We have entities—properties—that don’t
seem to be the kinds of things that are true or false and bear modal properties. We are
then told that some of them in fact are true and false and do possess modal properties.
47
Chapter 5 p. 90.
But nothing can be said about which properties are like this; and how/why they have
these features that other properties lack.
Now Speaks has a clever reply to this type of objection. He correctly points out that
those who think that mental properties are a subset of physical properties do not in
general provide a criterion for distinguishing the mental physical properties from the
non-mental physical properties. We don’t take this as a decisive objection to the view
that the mental properties are a subset of the set of physical properties. Hence that
Speaks provides no criterion for distinguishing the proposition properties from the
non-proposition properties is not a decisive objection to his view that propositions
comprise a subset of the set of properties. In response, let me say that I do think that
no criterion has been given for distinguishing the two is an objection to the view that
mental properties are a subset of the physical properties. And I am sure that those
who hold the view agree that this must eventually be done if the theory is to be suc-
cessful. It is however reasonable for those who hold this view to think that continued
study of mental and (non-mental) physical properties of the brain will yield further
understanding; and that, as a result, an account of what distinguishes mental physical
properties from non-mental physical properties will be forthcoming. But I, at least,
am much less sanguine about the idea that further study of properties and proposi-
tions will provide a deeper understanding of them that will yield an account of how to
distinguish proposition properties from non-proposition properties. So I don’t think
that failing to provide a criterion for distinguishing mental physical properties from
non-mental physical properties is on a par with Speaks failing to provide a criterion for
distinguishing proposition properties from non-proposition properties.
Fifth, there are conjunctive propositions, negated propositions, and disjunctive
propositions.48 If propositions are properties, as Speaks claims, there must be conjunc-
tive, negated and disjunctive properties.49 However, as is well known, many who believe
in properties do not think there are conjunctive, and especially negated and disjunc-
tive, properties.50 One reason for this is that it is widely thought that when two things
both possess the same “real” property, they should resemble each other or have a com-
mon nature; but common possession of negated and disjunctive properties does not
in general make for similarity or possession of a common nature. So because Speaks’s
account commits one to negated and disjunctive properties, it can only be adopted
by the most promiscuous in their views about what properties there are. I would have
hoped that a theory of propositions could remain neutral on this question.
A sixth and final objection concerns Speaks’s identification of truth and instantiation
for propositions.51 It is generally thought that in having truth conditions, a proposition
48
Less contentiously, there are conjunctive, disjunctive and negated sentences that express propositions.
49
Less contentiously, there are properties expressed by conjunctive, disjunctive and negated predicates.
50
Or properties expressed by conjunctive, disjunctive and negated predicates.
51
The qualification here is due to the fact that on Speaks’s view, for properties like being red instantiation
isn’t truth.
in some sense specifies conditions that have to be met by a world for the proposition to
be true there. Consider a proposition P specifying such conditions and a world w that
meets them. Surely, we want to say in such a case that P is true at w because w is a cer-
tain way. Indeed, this seems like a truism. However, one would also think that a thing’s
possessing an intrinsic property generally explains why the thing is a certain way. That
I possess the property of being 6 feet tall explains why I am a certain way. Possession of
the property constitutes my being a certain way. Now surely the same should be true of
worlds possessing intrinsic properties; that the world possesses an intrinsic property
constitutes its being a certain way. Suppose a world w possesses the property being
such that snow is white. This is an intrinsic property of w.52 Then, just as in other cases,
that should explain why w is a certain way. However, Speaks claims that w possessing
or instantiating a property like being such that snow is white is just this property qua
proposition being true at w. But then on the Speaks account, we should say w is a cer-
tain way, because the property/proposition being such that snow is white is true at w (i.e.
is instantiated at w). Unfortunately, this precisely reverses what we said was the proper
order of explanation mentioned above; the proposition that snow is white is true at
w because w is a certain way. Surely, this is the right order of explanation. Hence, that
Speaks’s account has it that w is a certain way because a proposition is true there, is a
strong reason for rejecting the account.
Let me close by saying why I think a view of the sort Speaks sketches, according to
which propositions are some kind of world properties, is the best alternative to views of
the sorts that Soames and I favor on which propositions have truth conditions because
of some facts about agents. The advantage that Soames and I see with such views is
that they promise to offer an explanation of how/why propositions have truth condi-
tions; and we both think the latter is something that very much needs to be explained.
There is a sense in which a view like Speaks’s offers an explanation for why propositions
have truth conditions. Propositions are properties of worlds (and everything or noth-
ing). As such, they are by their nature the sorts of things that are instantiated or not by
worlds. They are also the kinds of things that would be instantiated or not if this or that
world were actual. But instantiation just is truth for propositions/properties; and that a
proposition/property would be instantiated if w were actual just is truth at w. So prop-
ositions by their nature are true or false (i.e. instantiated or not) at worlds, and so have
truth conditions, simply because they are the kinds of things that would or would not
be instantiated if this or that world were actual. The explanation here takes as primitive
that properties by their natures are and would be instantiated or not by things, but that
does not seem to be such a bad thing to take as primitive.
52
E.g. a duplicate of w would have to possess it.
Here’s one thing that the three of us have in common: we all dislike the idea that prop-
ositions could be entities which are intrinsically representational—in the sense that
they both are representational and would exist and be representational, even if there
were no subjects around to do any representing.
My response to this negative claim is pretty straightforward: I deny that proposi-
tions are representational entities and, instead, identify the proposition expressed by
“Amelia talks” with the property of being such that Amelia talks. This property is not—
any more than any other property—a representational entity; properties are not about
anything. I then argue that we can use properties of this kind to give an account of
the sorts of phenomena for which we might have thought we needed representational
entities: the representational mental states of subjects, and the fact that sentences, like
the contents of these states, have truth conditions.
King and Soames both give a different kind of response to the denial that there are
intrinsically representational entities. They think that even if there are no intrinsi-
cally representational propositions (in the above sense), we still need representational
propositions. It’s just that we need to explain how these entities come to have repre-
sentational properties; and each tries to explain this by saying how the relevant enti-
ties come to have the representational properties they do in terms of the mental acts
(broadly construed) of thinking subjects.
King’s and Soames’s theories differ in many important ways; in particular they differ
on the question of which mental acts provide the relevant representational proper-
ties. (In King’s case, these are interpretations of sentential and propositional relations,
and in Soames’s they are events of predication.) But they both satisfy the very abstract
sketch of the preceding paragraph, and so both are theories which fill in the following
diagram in different ways:
Representational
properties of
non-fundamental
mental states
Representational
properties of
sentences
I’ll turn to the details of King’s and Soames’s theories below. But the fact that they share
the above structure is enough to raise two general questions about their views.
The first is a question of motivation. Given that both King and Soames want to sup-
ply propositions with representational properties, there’s a pretty clear sense in which
both theories say more than mine. But once we have intrinsically representational
propositions off the table, why should we try to come up with derivatively representa-
tional surrogates?
This question can be brought into sharper focus by pointing out the ways in which
King and Soames accept the basic ontology of the theory I (adapting the views of
Chisholm, Lewis, van Inwagen, and others) provide. Neither King nor Soames
expresses any skepticism about properties; on the contrary, each makes free use of
properties in his account of propositions. (Each, like me, thinks of properties as among
the constituents of propositions.1) Neither King nor Soames expresses any skepticism
about our ability to have cognitive access to properties; again, each presupposes this
sort of access in giving his account of propositions. So it seems that each should agree
that the properties with which I identify propositions exist, and that they are the sorts
of things to which we have cognitive access.
This last point bears emphasis. Both King and Soames emphasize, in different ways,
the naturalistic credentials of their theories. King thinks of his account as one of natu-
ralized propositions because it avoids saying that entities are representational inde-
pendently of thinking subjects—still more naturalistic, then, is a theory which avoids
saying that any entities are representational, full stop. Both King and Soames wonder
how our cognitive access to propositions as traditionally conceived might be under-
stood, and take pains to explain how we might have access to the entities with which
they identify propositions. But in giving this account, both presuppose rather than
explain our cognitive access to properties—hence it’s hard to see how it could be easier
to have cognitive access to the entities with which they identify propositions than to
1
Though see Ch. 11 for some further discussion of the meaning of “constituent” in this context.
the properties which I take propositions to be. It may well be that our cognitive access
to properties is something which needs explanation; but, if so, this is a challenge for my
theory no more than for those of Soames and King.
I suggest, then, that the ontology of the theory I prefer is a proper subset of the ontol-
ogies of the theories of King and Soames, and our cognitive access to the entities with
which I identify propositions can be no harder to understand than our cognitive access
to propositions as conceived by King and Soames. It then seems to me that if we are to
prefer their theories to the property theory, we must think that their theories explain
something which my admittedly sparser theory cannot. What I don’t quite see is what
this crucial explanandum could be.
A second sort of question for theories of the kind that Soames and King provide can
be presented by way of a dilemma. If propositions have representational properties,
then either they have them essentially, or they don’t. But there is trouble either way.
If propositions have their representational properties essentially, then it is hard to
see how any theory could explain why they have those properties. This is just because
there seem to be limits on the sort of explanation we can give of why something has a
property, if it has that property essentially.
Suppose that my desk—call it “Fred”—is essentially wooden. Given this, can we
explain why Fred is wooden? One might well think that we can’t. Or, at least, we can’t
explain this in the same way that we can explain why a book is sitting on a particu-
lar table—after all, Fred simply couldn’t exist without being wooden, and so we can’t
recount events in the history of Fred’s existence which led to its woodenness.
That said, there is, of course, another sense in which we can explain how Fred came
to be wooden—and that is just that we can explain how Fred came to be. But, given that
Fred could not exist without being wooden, this is the only sort of explanation of Fred’s
woodenness which we can give.
Hence, on the view that propositions have their representational properties essen-
tially, it seems that the only explanation we can give of their possession of those prop-
erties is an explanation of how the relevant proposition came to be. But one might
think that giving an explanation of why a proposition exists is not the same thing as
giving an explanation of its possession of certain representational properties. This is
best brought out by thinking about an example.
Consider a proponent of propositions as traditionally conceived: someone who
thinks that propositions have their representational properties independently of any-
thing that anyone thinks or does. Such a person could, it seems, think of propositions
as structured and he could also think that some propositions—singular propositions—
have contingently existing objects as constituents. Reflecting on these commitments,
our traditional proposition theorist might also come to believe that singular proposi-
tions can’t exist unless their constituents do. So far, it seems, none of these commit-
ments force our imaginary theorist—who of course has a lot in common with many
actual philosophers—to give up her traditional view of propositions, according to
which “propositions have truth conditions by their very natures and independently
of minds and languages” (63) and “propositional representation is primary and the
agent’s representation is to be explained in terms of it” (136)
But now consider a singular proposition about some artefact—say, the paper air-
plane I just made. How should our traditional proposition theorist think about this
proposition? Given the commitments just sketched, she should say that by making
the paper airplane I brought into existence a bunch of singular propositions with that
airplane as a constituent—including, for example, the proposition that that paper air-
plane is white.
Now — should our traditional proposition theorist say that her construction of the
airplane explains the representational properties of the proposition that that airplane
is white? It seems to me that she should not. She should say that her construction of
the airplane explains the existence of that singular proposition, but that once it came
to exist, it brought its representational properties with it; it had those properties, as it
were, automatically. Nothing in the making of the airplane explained the proposition’s
having those representational properties, and hence nothing in this story should force
our traditional proposition theorist to stop being a traditional proposition theorist.
This example can be used to formulate a challenge to those who want to explain
the representational properties of propositions and who take the first horn of our
dilemma, saying that propositions have their representational properties essentially.
Let E be something proposed as the explanation for the representational properties
of the proposition that that airplane is white. We should agree that my making the
paper airplane does not explain the representational properties of this proposition. So
we should also agree that if E is to provide a genuine explanation of this proposition’s
representational properties, it must bear some relation to that proposition’s having its
representational properties which my act of making the paper airplane does not. (So it
can’t just be the relation of bringing that proposition into existence.) The challenge is to
say what this relation could be.
If one thinks that it’s hard to see what this relation could be, this might push us
toward the second horn of the dilemma, according to which propositions don’t have
their representational properties essentially. This view fits much more easily with the
idea that we can explain the representational properties of propositions: if they don’t
have those properties essentially, then it is no harder, in principle, to understand how
we could imbue them with these properties than it is to explain how we could paint a
fence white. But this view faces other problems.
If propositions don’t have their representational properties essentially, then it is
possible for some proposition to exist, and either lack representational properties
altogether, or to have different representational properties than it actually has. But (if
propositions have representational properties) there is a presumably necessary con-
nection between a proposition’s representational properties and its truth conditions.
Hence if the proposition that grass is green does not have its representational proper-
ties essentially, it could presumably exist without being true iff grass is green. But this
leads to very odd results; in particular, it seems to make claims like the following true:
Possibly, John believes that grass is green, and grass is green, but John’s belief is not true.
Possibly, grass is green, and the proposition that grass is green is false (or exists and lacks a
truth-value).
[PR] there is a context c such that x is the semantic value (relative to c and assignment g) of a
lexical item e of some language L and y is the semantic value (relative to c and assignment g) of a
lexical item e’ of L such that e occurs at the left terminal node of the sentential relation R that in L
encodes ascription and e’ occurs at R’s right terminal node.3
A first pass at King’s view then identifies the proposition that Michael swims with the
fact obtained by assigning Michael as value to the free variable “x,” and the property of
swimming as value to the free variable “y.” As noted above, King denies that proposi-
tions like this have representational properties on their own; rather, we do something
to give them these representational properties. So what is it that we do?
The central thing that we do is that we interpret the propositional relation [PR] as
encoding ascription; as King puts it, “FAST has truth conditions because speakers inter-
pret its propositional relation as ascribing the property of swimming to Michael.” But,
one might wonder, how do we manage to interpret propositional relations?
2
One point worth noting here: above I’m using “essential property” as equivalent to “property a thing has
necessarily.” But it is plausible, following Fine (1994) and others, that we can understand the essential prop-
erties of a thing as a proper subset of the properties that it has necessarily; and perhaps we can explain why a
thing necessarily possesses certain properties in terms of its essential properties. Given all of this, one might
try to escape the dilemma above by saying that propositions have their representational properties neces-
sarily but not essentially, and that the essential properties of propositions can explain their representational
properties. This seems to me like a promising general strategy; but I’m not quite sure how to make it work in
the context of Soames’s or King’s theory. Thanks to Ben Caplan for helpful discussion of this point.
3
As King notes, we might think of propositional relations as either involving specific sentential relations,
or as involving existential generalization over sentential relations. I prefer the second version of the theory,
which is what I use in what follows—but nothing hangs on this choice here.
“is simply that we compose the semantic values at the terminal nodes of the proposition in the
way we do” (13)
Here we’ve reached the “fundamental mental states” mentioned above: they are a mat-
ter of speakers composing the semantic values of simple expressions in a way which
results in their assigning truth conditions to the relevant sentences.
Note that for King’s view to be plausible, we must think of the interpretation of sen-
tential relations—i.e., the composing of certain semantic values of expressions stand-
ing in certain syntactic relations—as not just explaining how the relevant propositional
relation gets interpreted, but also as metaphysically sufficient for the interpretation of
that propositional relation.4 For suppose that it were not. Then a speaker could inter-
pret the syntactic relation in “Michael swims” as encoding ascription—thereby ensur-
ing that there is some language in which a term for Michael is concatenated with a term
for swimming, and in which concatenation encodes ascription—without the proposi-
tion that Michael swims existing. (Since this proposition, on King’s theory, could not
exist unless [PR] encodes ascription.) But this would be absurd.5
In what follows, I want to discuss three sorts of worries about this theory: (1) that
propositional relations necessarily have certain representational properties, so their
possession of these properties can’t be explained by what speakers do; (2) that the best
4
Hence I think that we can’t say, as King (2007), 60 does, that we can explain the fact that “the propo-
sitional relation inherits its significance from the sentential relation” in terms of “something that we and
our linguistic ancestors did.” (Unless it was not just something that we’ve done, but something which it was
metaphysically impossible for us not to do.)
5
This point is also relevant to a distinction between two different ways in which King sometimes states his
theory. King sometimes suggests that the facts with which he identifies propositions include this fact about
how the relevant propositional relations are interpreted. (See King Ch. 4 p, 9, and King (2009), 265.) This
suggests that FAST is not the fact
(a) [PR](Michael, swimming)
but rather the more complex fact
(b) [PR](Michael, swimming) & [PR] encodes ascription
In what follows, I’ll ignore this complication in King’s view, and discuss the simpler version (a) of the theory.
I don’t think that, in the end, the difference between these versions of the view matters much. This is because,
given the way that King defines the relevant notions, it is impossible for (a) to obtain without (b), or for
(b) to obtain without (a). The second direction is trivial; the first holds because, by the argument just given,
interpreting the relevant sentential relation as encoding ascription must be metaphysically sufficient for also
encoding the propositional relation involving that sentential relation. But then since [PR] could not hold
between Michael and swimming unless R was interpreted as encoding ascription, it follows that, necessarily,
if (a) is a fact, then (b) is as well.
So propositional relations are necessarily existing entities which necessarily have rep-
resentational properties—like the property of being such that for any x and y, the fact
that they relate x and y is true iff the first instantiates the second, and is about x and is
about y. Because these are necessary truths, they are not under the control of speak-
ers, and speakers do not bring them about. But at this stage, one might worry that
propositional relations are the sorts of things whose existence King is concerned to
deny—entities which have their representational properties “by [their] very nature
and independently of minds and languages.”
King might reasonably, and correctly, reply that on his account there are no proposi-
tions, and there is no truth and falsity, without speakers doing something to bring it
about that there is. I’m a bit skeptical that this reply is enough to allay the worry that
propositional relations are a bit too much like the primitively representational propo-
sitions King wants to do without. One way to bring out the reasons for this is to focus
on just what speakers do to bring truth and falsity into the world.
The role played by speakers is simply to make it that case that [PR] relates some
things; once they have brought into existence a fact of the form [PR](x,y) there’s no fur-
ther work for them to do in giving this fact its representational properties. But, given
this, it seems like the role played by speakers on this theory is uncomfortably close to
the role played by our paper airplane maker in the example discussed above.
We can focus this worry in the way mentioned above, by asking: what relation does the
interpretation of [PR] bear to the representational properties of FAST which the making
of the paper airplane does not bear to the proposition that that paper airplane is white?
It seems to me that this is a difficult question for the proponent of King’s view to answer;
and this, in turn, suggests that we are getting no more of an explanation of the represen-
tational properties of propositions than is provided by the making of the paper airplane.
King might reply that the crucial difference here is a difference between the relevant
action types—making a paper airplane, on the one hand, and interpreting a propositional
relation, on the other. Making a paper airplane, he might object, simply isn’t the sort of
action which can explain the representational properties of the many singular propositions
involving the airplane—interpreting a sentential and a propositional relation, however, is.
But this focus on exactly which actions of language-using subjects do the relevant
explanatory work leads to another, related sort of worry about King’s view. In order to
avoid appeal to merely possible contexts, King allows that it is sufficient for an arbitrary
singular proposition that o is swims to exist that there be some syntactic structure
x swims
in English, where the “x” is a variable which has a semantic value only relative to an
assignment. For any object o, there will be, after all, some assignment which assigns o
as the value of “x.” But that will be sufficient for [PR] to hold between o and the prop-
erty of swimming, which will be sufficient for the relevant singular proposition to exist.
But this means that the first time someone used a variable expression, it was—meta-
physically speaking—a much more momentous event than one might have thought.
For this event brought a host of representational entities into the world: the infinitely
many singular propositions which predicate swimming of something.
In one sense, this is nice for King’s view: it secures the existence of a great many
propositions, and at very little cost. But this virtue has a corresponding vice, and rein-
forces the worries raised in connection with the example of the paper airplane above.
The worry is that we are getting too much for free from the nature of [PR]—to which
no thinking subjects contribute anything—to get a satisfying explanation, in terms of
thinking subjects, of the representational properties of propositions.
We can push the point one step further. As King notes (note 11), natural languages
plausibly contain expressions which are best treated as variables over properties. But
then we could presumably have a sentence in a natural language, like “Something is
some way” which involves the syntactic structure
x y
where “x” is a variable over objects and “y” a variable over properties. But could utter-
ing “Something is some way” really suffice to explain the representational properties of
every singular monadic predication?
2.These have all been worries about the extent to which King’s theory explains rather
than assumes the representational properties of propositions. Let’s turn our focus
now to the nature of the fundamental mental state which is supposed to be doing the
explaining: speakers’ interpretation of sentential relations.
Our understanding of the interpretation of sentential relations is sharply constrained
by King’s explanatory ambitions. Because the interpretation of sentential relations is
supposed to explain the existence and representational properties of propositions,
interpreting a sentential relation cannot be, or presuppose, any propositional attitudes.
This constraint seems to be inconsistent with some things King says about the inter-
pretation of sentential relations. Recall that the interpretation of sentential relations is
simply a matter of speakers composing semantic values in the way that they do. This is
supposed to explain how certain syntactic relations acquire semantic significance, which
in turn explains how propositional relations and propositions acquire representational
properties. But sometimes King talks as though the explanatory order runs in the other
direction, and that speakers compose semantic values in the way that they do because the
relevant syntactic relations have a certain semantic significance. He says, for instance, that
“Speakers work their way up the propositional relation composing semantic values in the way
they do, thus interpreting the propositional relation, because it is the way the semantic signifi-
cance of the syntax dictates that they compose these semantic values.”6
But this can’t be right. If interpretations of sentential relations are supposed to be what
brings the properties of sentential relations into the world, we can’t in general under-
stand interpretations of sentential relations to be responses to those very properties.
King might, without serious cost, simply take back those remarks about the inter-
pretation of sentential relations. But there’s another circularity worry in the vicinity
which, I think, can’t be as easily dismissed.
The interpretation of sentential relations is supposed to be a matter of speakers
composing the semantic values of the terminal nodes of a propositional relation. But
that means that the interpretation of sentential relations can only get off the ground if
those terminal nodes have a semantic value. But this looks, from the point of view of
King’s theory, very odd. If the interpretation of sentential relations is supposed to be
what brings representational properties into the world, how can we understand the
interpretation of sentential relations in terms of an antecedent assignment of seman-
tic values to expressions? If we’re reluctant to take as primitive the representational
6
King, p. 86. See also pp. 77–78: “That is, what is it that we do that amounts to our so interpreting it? It is
simply that we compose the semantic values at the terminal nodes of the propositional relation in the way we
do. In the end, this is just a reflex of the sentential relation R having the semantic significance it does.” Better,
I think, for King to say that R’s having the semantic significance that it does is a reflex of the way we compose
the relevant semantic values.
7
I’m thinking of accounts in the spirit of Lewis (1975) and Schiffer (1972). One way around these problems
would be for King to take a different view of the interpretation of sentential relations, and understand it in
terms of language users taking certain sentences to have certain truth conditions. This has the advantage that
it does not obviously make use of explanatorily prior facts about the assignments of semantic values to the
terminal nodes of certain logical forms. The disadvantage, though, is that “taking sentences to have certain
truth conditions” seems itself to be a propositional attitude.
3.This worry leads into the question of how, on King’s view, we should understand
the mental states of subjects who existed prior to the existence of natural languages,
and hence (one might think, given the above story) prior to the existence of proposi-
tions. Do we really want to deny that such subjects had beliefs, desires, and perceptual
experiences, each of which plausibly involves a relation to a proposition?
King is well aware of this problem, and tries to make room for the existence of such
subjects. In particular, he finds plausible the idea that, prior to the existence of lan-
guages like English, organisms enjoyed perceptual experiences which had proposi-
tions as their contents. King further suggests that “there will be an account of those
contents in the spirit of the present account of the contents of natural language sen-
tences.” I agree with King’s thought that an account of propositions should be consist-
ent with pre-linguistic but genuinely propositional perceptual experiences, but am less
sure that I can see how his account meets this constraint.
Let’s consider how we might adapt the preceding story to the case of perceptual expe-
riences; to fix ideas, let’s think about a frog’s visual experience of a fly sitting on a green
leaf, and let’s simplify by supposing that the content of this visual experience is a singular
proposition which predicates the property of being in a certain location (“L”) of the fly.
Presumably this, like any such singular proposition, could be expressed by a sentence.
Hence, on King’s view, the content of the frog’s experience will then be the proposition
expressed by such a sentence, i.e. the fact that the fly and L stand in the propositional rela-
tion [PR]. Our question is: what did the frog do to bring this proposition into existence?
I take it that the frog must have done something to interpret the propositional rela-
tion [PR]. But this seems to show that we need to change our view of what the relevant
propositional relation is—after all, the frog isn’t interpreting any sentential relations.
And it won’t do to say that [PR] works fine for the propositions expressed by sentences
but that we need to invoke some other propositional relation for mental states, since
it is quite plausible that some propositions can both be expressed by a sentence and be
the content of a mental state, like a perceptual experience.
I think that the best thing for King to say here is that even if the frog does not inter-
pret any sentential relations, he does do something similar: he interprets his experi-
ence. This suggests that we should then revise our view of propositional relations like
[PR] so that they will be as applicable to the case of interpretations of experiences as to
the case of interpretations of sentences.
This is obviously non-trivial. But let’s suppose that we’ve done it. A problem still
remains, which has to do with a fundamental difference between sentences and per-
ceptual experiences. Presumably, the phenomenal character of the frog’s experience is
fixed independently of the frog’s interpretation of his experience; were it not, it’s hard
to see what the frog could be interpreting. But according to a plausible intentional-
ist thesis, phenomenal properties are identical to certain representational properties,
which involve relations to propositions. Hence it seems that if the phenomenal char-
acter of the frog’s experience is fixed independently of his interpretation of his experi-
ence, the content of his experience must be as well.
But if this is right, and the analogue of interpretation of sentential relations plays no
role here, we lose our explanation of how the fly and L come to stand in whatever the
relevant propositional relation turns out to be. And then we lose our explanation of the
existence of the relevant proposition, as well as of its representational properties.
To be sure, King does not pretend to a fully worked out theory of the contents of per-
ceptual experience, and it may be that his story can be told in some way other than the
way I’ve just argued to be objectionable.8 But I do think that the disanalogies between
perceptual experiences and sentences cast some doubt on the idea that King’s theory
can be smoothly generalized—as it surely must be, if it is to be acceptable—to the case
of perceptual experience.
Consider the token event of an agent judging that o is red. The idea is that token events
of this sort are always underwritten by distinct token events of predicating redness of
o. And what’s true of judgement also holds for a wide variety of mental events—including
entertaining the proposition that o is red, accepting that o is red, asserting that o is red,
denying that o is red, judging that o is red, and visually representing o as red. Each token
of these event types is accompanied by a distinct token event of predicating redness of o.
The primitive representational facts are facts about the representational properties of
these token mental events of predication. Propositions are types of such events, and inherit
their representational properties from the representational properties of their tokens.
I’d like to raise four sorts of questions about this account: (1) questions about whether
we have reason to believe that the relevant token events exist; (2) questions about their
8
One idea would be to say that the frog interprets his own brain state rather than the phenomenal char-
acter of the experience, and that the connection between that brain state and its content is contingent rather
than necessary. Maybe something like this would work, but there are at least two worries. One is the sheer
implausibility of the idea that frogs interpret their own brain states in the way that we interpret sentences.
The second and more fundamental is that, even if the connection between the underlying brain state and its
content is contingent, this does not imply that it is mediated by the interpretation of the frog. More plausibly,
it is fixed by facts about (counterfactual and actual) causal relations in which that brain state stands, which
are fixed independently of the frog’s interpretation.
frequency; (3) questions about how, if they have these properties necessarily, these rep-
resentational properties of propositions could be explained by anything we do; and
(4) questions about why, given that types don’t invariably inherit the properties of their
tokens, propositions should inherit the representational properties of theirs.
1.On reading Soames’s description of the fundamental events of predication, one
might be tempted to reply Hume-style, that “when I enter most intimately into what
I call my mental states, I always stumble on some particular propositional attitude
or other, a judgement or an assertion or an experience. I never can catch an event of
predication, and never can observe anything but the attitudes.” When I think about
various propositional attitudes—events of visually representing that o is red, judging
that o is red, and asserting that o is red—I really don’t notice an event of predication
which accompanies each one. But if they are always accompanied by such an event,
what explains this?
One might reply that I do notice the event of predication—it’s just that I don’t
notice it as such. After all, I do notice the similarity in content between these differ-
ent mental states—and isn’t this, on his view, just to notice the events of predication?
Not quite: noticing sameness of content, on Soames’s view, is noticing sameness in
associated event types (contents being identical to event types). But to see that various
mental states are all relations to the same event type is not to notice the occurrence of
tokens of that type.
A different and more promising reply is is suggested by some remarks Soames makes
when explaining predication: he might say that every occurrent propositional attitude
is accompanied by a certain basic propositional attitude: roughly, the attitude of “call-
ing the proposition to mind.” Soames calls this attitude entertaining. Perhaps we could
simply identify the event of coming to have this propositional attitude—entertaining
the proposition that o is red—with the token event of predicating redness of o. Soames
says things which suggest this; for instance, he says that “to entertain the proposition
that o is red is to predicate redness of o.”9 If such an identity claim were true, this would
make the existence of the token events of predication much less dubious—for, in each
of the cases discussed above, we obviously do call the relevant proposition to mind.
But it is hard for me to see how this identity claim could be true. For suppose that
token events of predication were identical to token events of a subject coming to enter-
tain a proposition. Then, on Soames’s view, a certain token event e would be identical
to a token event e* of a subject coming to stand in a relation to an event type E of which
e is a token. But if token events x, y are identical, then any type of which x is a token
must also be a type of which y is a token. But one type of which e is a token is the propo-
sition that o is red. It then follows that e* must be a token of the proposition that o is
red. But then propositions are event types with the surprising property that each of its
token events is some subject coming to stand in a relation to that type.
9
136–137. See also Soames (2010), 81–82.
This is worrying for a few reasons. First, it leads to troubles understanding what
the view says. Soames claims that the proposition that o is red is a certain event type
E. If we want to know what E is, we naturally look to its tokens. But when we look to
its tokens we find that they are events of subjects coming to bear a certain relation to
E,which leaves our question about the nature of E unanswered.
This regress is not obviously vicious. But it does, I think, make trouble for some of
the explanatory claims that Soames seems inclined to make. Soames is inclined to say
that the representational properties of propositions are explained by the representa-
tional properties of token events of predication, and is at least open to the idea that the
existence of the relevant propositions is also explained by the existence of the relevant
token events of predication. But it is hard to see how this is going to work if the token
events of predication are identical to events of subjects coming to stand in relations
to the propositions whose existence and representational properties the token acts of
predication were supposed to explain.
A further, related worry comes from reflection on the propositional attitude of enter-
taining a proposition, on the supposition that events of coming to stand in this attitude
are identical to events of predicating properties of objects. These events of predicating
properties of objects are supposed to have their representational properties intrinsi-
cally—in particular, on pain of circularity, they are not supposed to have their rep-
resentational properties explained in terms of the representational properties of any
proposition. But since these events of predication are identical to events of coming to
be in a certain propositional attitude state—entertaining the relevant proposition—
it follows that at least one propositional attitude state also must have its representa-
tional properties intrinsically, and to be such that its representational properties are
not explained by the representational properties of any proposition. But this raises a
question. If we are willing to say this about one propositional attitude state, why not say
this for all of them? Why not—so far as the explanation of the representational prop-
erties of mental states are concerned—do away with the primitive acts of predication
entirely, and let each of the familiar propositional attitudes—belief, assertion, etc.—be
intrinsically representational states, whose status as representational is not explained
by the representational properties of the propositions to which they are relations?10
Now nothing in his theory forces Soames to identify predicating redness of o with
entertaining the proposition that o is red; his statements to the effect that coming to
be in the second state just is a matter of coming to be in the first state are also consist-
ent with holding that the two states (and events of coming to be in those states) are
10
One might say this while still giving the proposition that grass is green a role in determining exactly
which representational property one has when one believes that grass is green. This is of course analogous to
what Soames says about predication: the objects of the predication — in our example, o and redness — are
not themselves representational entities, but the objects of a given event of predication obviously play a role
in determining the representational properties of that event. More on this sort of account of the attitudes in
Ch. 11 below.
11
Soames (2010), 29.
about the existence conditions of propositions, given Soames’s interest in explaining the
representational properties of propositions. After all, if event types did exist necessarily,
then propositions would exist, and have all of their representational properties, whether
or not there were any thinking subjects around to do any representing. And if this were
true, it would be hard to see how Soames’s theory could live up to the guiding ideas
“(i) that the perceptual and cognitive activity of agents is the conceptual basis of all representa-
tion and (ii) that propositions are representational in virtue of the relations they bear to this
representational activity.” (8)
What role could agents play, if all of the propositions existed and had all of their repre-
sentational properties, no matter what they did?12
The problem is that this pair of commitments—that belief does not require predica-
tion and that existence does require certain facts about predication—invalidates some
intuitively valid inferences.
Consider some property which could be such that no one has ever predicated of
anything—say, the property of being 726-sided. Can someone in the relevant world
still have beliefs involving this property? It seems that they can; in particular it seems
that they can believe that no circles are 726-sided without having ever predicated the
property of being 726-sided of anything. Now consider the inference:
This certainly seems to be valid. Indeed, as noted in Chapters 1 and 2, it’s the sort of
trivially valid inference often used to defend the existence of propositions. But on the
present theory it comes out invalid—even though Jeff and Scott both bear the belief
relation to the proposition that no circles are 726-sided, this proposition need not exist,
and hence there is no guarantee that there is something that Jeff and Scott both believe.13
To my mind, this consequence of the theory is made more worrying by the fact that
superficially quite similar inferences, like
12
Though, as mentioned in note 2 above, one might try to answer this question via a theory according to
which some sorts of necessary truths are explained in terms of others.
13
One might think that there is an interpretation on which this argument is, even on Soames’s view,
valid: namely one in which we interpret the existential quantification in the conclusion as possibilist. But if
we force an actualist interpretation of the quantifier, by changing the conclusion to “there is actually. . .,” and
changing the premises from “believes” to “actually believes,” the argument seems no less valid.
come out valid—after all, assertion, being an occurrent mental state, does require an
event of predication, and hence entails the existence of the relevant propositions. The
fact that we are forced to treat the “believes” inferences differently than the “asserts”
inferences seems to me to be a cost of the view.
3. Setting aside these worries about the validity of certain inferences, there’s a more
purely metaphysical worry arising from this discussion of belief and predication,
which is just that Soames’s theory is committed to the claim that subjects can believe
propositions which do not exist, and never did. This is a commitment of which Soames
is well aware. But it does show that the view that propositions as event types is one
which no philosopher committed to serious actualism—the thesis that nothing can
have a property or stand in a relation in w without existing in w—should accept.
But rather than focusing on the plausibility of denying serious actualism, I’d like to
focus on its consequences for Soames’s explanatory claims. While Soames does not of
course think that propositions can have just any property without existing, he does
think that they can have quite a few: not just being believed, but also being true or
false, and representing the world as being a certain way. But one might think that this
reinstates the worry above—which echoes one of the objections to King’s view—that
the role actually played by thinking subjects on Soames’s view is thinner than it might
at first seem.
Consider a barren world w, with no subjects doing any thinking or talking. As far as
I can tell, on Soames’s view, every proposition has all of its representational properties
in w: the proposition that grass is green is, in w, about grass and about greenness; in w,
it represents grass as green; it w, it is true iff grass is green. No subjects in w are to thank
for the proposition having these representational properties, there being no such sub-
jects to thank. How can this be squared with Soames’s claim that propositions are rep-
resentational in virtue of the relations that they bear to representational activity?
There are really two problems here. The first is the problem of explaining the fact
that the proposition that grass is green has representational properties in the barren
world w. The second is about whether the fact that propositions have representational
properties in w impugns our explanation of the representational properties of proposi-
tions in the actual world.
Let’s focus first on what explains the representational properties of the proposi-
tions in w. I think that the best thing for Soames to say here is that propositions can
have representational properties in a barren world like w because certain counterfac-
tual claims about those propositions are true in w: in particular, the proposition that
grass is green is about grass in w because it is true in w that were some subject to, for
example, judge that grass is green, that judgment would involve predicating green-
ness of grass.
But here it seems to me that we risk reversing Soames’s preferred order of explana-
tion. Setting aside worries about the existence of acts of predication, I don’t doubt that
this counterfactual is true in w. What I wonder is what makes it true. It seems to me
that something about w must explain the truth of this counterfactual—but it’s hard
(in the absence of any actual predications of the right sort) to see what could explain
its truth other than facts about the representational properties of the proposition that
grass is green. But we can’t both explain the representational properties of propositions
in terms of certain counterfactuals, and explain the truth of those counterfactuals in
terms of the representational properties of propositions.
Now return to the second problem, about propositions in the actual world. What
role did the activity of we, the thinking and talking subjects of the actual world, have
to play in explaining the representational properties of propositions? The preceding
remarks about w put some constraints on the way in which we answer this question: we
can’t say that, were it not for our thought and talk, these propositions wouldn’t have
those representational properties; after all, they have them all in our barren world
w. Our contribution is only this: that we brought them into existence. But this seems
to me a less significant achievement when we know that they already had all of their
representational properties before they came to be.
This cluster of worries is obviously analogous to the the first objection to King
raised above. The situation for Soames’s view seems to me in one way better and in
one way worse than King’s theory. Soames’s view avoids the problem that saying
“Something is some way” brings into existence every singular monadic proposi-
tion; and, in this way, he manages to secure a closer, and potentially more explana-
tory, relation between the fundamental mental states and the propositions whose
representational properties they are supposed to explain. But this virtue is con-
nected to a corresponding vice. By making the connection between the fundamen-
tal mental states and propositions more demanding, Soames makes it harder for
propositions to exist; which in turn makes it more tempting to say that propositions
have all of their representational properties whether or not they exist, and no mat-
ter what we (actually) do—which casts doubt, from a different angle, on the central
explanatory claim.
4. Let’s set these questions aside, and grant for purposes of argument that there are
token events of predication of the sort Soames describes, that they occur when the
theory says they do, and that their having representational properties is explained by
the nature of predication rather than by any built-in representational properties of
its objects. There are still residual worries about how these acts of predication could
explain the representational properties of propositions. Soames explains the relation
between the two succinctly:
“This suggests that the proposition that o is red is simply the minimal event type in which an
arbitrary agent predicates being red of o. This event-type is representational because every con-
ceivable instance of it is one in which an agent represents something as being a certain way.” (136)
This seems straightforward: there are event tokens which are inherently representa-
tional, and the event types of which these are tokens—the propositions—inherit their
representational properties from those token events.
But, as Ben Caplan has pointed out (in private communication), this is not quite as
straightforward as it might at first sound. After all, it is not in general true that event
types inherit the properties of tokens of that type. A token event of predication has
the property of being a token event, whereas the corresponding type does not; a token
event of eating dinner must have a certain duration, whereas the type does not. So if
types do not in general inherit the properties of their tokens, why should we think, as
Soames encourages us to, that the representational properties of token events of predi-
cation explain the properties of propositions?
Soames might reply that representational properties are special: perhaps often event
types don’t inherit the properties of their tokens, but they do when the relevant prop-
erties are representational ones. But this is doubly problematic. On the one hand, it
seems ad hoc; why should representational properties be different in this way? And, on
the other hand, it seems false—even if we restrict ourselves to representational proper-
ties, event types don’t always inherit the representational properties of their tokens.
To see this, consider again a token event of predicating redness of some particular
object o. Suppose that here Bob performs the predication, and does so while eating an
Oreo cookie. This token event is, it seems, an event of the following types:
(i) the event type of Bob predicating redness of o while eating an Oreo.
(ii) the event type of someone predicating redness of o while eating an Oreo.
(iii) the event type of someone predicating redness of o.
(iv) the event type of someone predicating something of o.
(v) the event type of someone predicating something of something.
(vi) the event type of someone doing some predicating.
(vii) the event type of someone doing something.
On Soames’s theory, I think, only (iii)-(v) are propositions, and hence inherit repre-
sentational properties from the token event described above. But we want to know
why just these get that privilege: (i), (ii), (vi) and (vii) are, after all, equally event types
of which our token event is a token. No appeal to the specialness of representational
properties, as opposed to properties like having a certain duration, will answer this
question.
As it stands, this is less an objection to the theory than a question to which the theory
ultimately owes an answer which, at present, I don’t think it provides. The question
is: given that types often do not invariably inherit the properties of their tokens, what
explains the fact (if it is a fact) that propositions inherit the representational properties
of their tokens? The foregoing examples are enough to show that this is not a question
which has a trivial answer, and hence that the token/type relations to which Soames
appeals in his theory are not as obviously explanatory as one might have thought.14
14
Thanks to Ben Caplan for very helpful discussion of these issues, and for comments on a previous draft
of this paper.
Propositions as properties
I begin with a friendly amendment. According to Speaks (p.5) the proposition expressed
by “Amelia talks” is the property: being such that Amelia instantiates talking, which is also
the semantic content of the predicate “is such that Amelia talks.” Since the predicate is
complex, its content should be a structured complex of the contents of its grammatical
constituents. This suggests that the content of “Amelia instantiates talking”—namely,
being such that <Amelia, talking> instantiates Instantiation—should be included in the
content of “Amelia talks.” Since we don’t want this, Speaks should drop instantiation, and
identify the content of “Amelia talks” with being such that Amelia talks, while treating “is
such that” as a syncategorematical element that turns sentences into predicates without
adding an extra constituent to the content.
Speaks motivates his view, in part, by noting:
“believing a proposition is taking the world to be a certain way. But if, as it seems, ‘ways things are’
are properties, this indicates that having a belief is taking a certain attitude toward a property.” (p. 6)
There is something right about this, but it doesn’t favor the view that propositions are
properties (of the sort he has in mind). The claim that believing a proposition is taking
the world to be a certain way approximates the more discriminating claim that believing
a proposition is taking something (or things) to be a certain way (or ways). Sometimes
the thing so taken is the entire universe, but often it’s not. To believe that o is red is to
take o to be a certain way, which involves taking a certain stance toward a property.
To believe the proposition is to (be disposed to) predicate redness of o, thereby repre-
senting o as red (while endorsing that predication). In this way, we can accommodate
Speaks’s truism without taking the properties he identifies to be the things believed.1
1
My view of propositions as event types also accommodates Speaks’s truism. I here leave it open whether
event types might themselves be properties of events, or of agents (if they are identified with acts). Whatever
For Speaks, propositions are a certain kind of property, and truth is instantiation.
This makes it difficult to capture the fact that truth is a kind of accuracy in representa-
tion. A map or portrait is accurate, or veridical, when it represents its subject matter as
being how it really is; a proposition is true when it represents things as they really are.
This parallel seems to be lost when propositions are identified with properties that,
as Speaks admits, aren’t intrinsically representational.(p. 7) Unless he can identify
some sense in which they are representational, he will lose the pretheoretic connection
between truth and accuracy.
Just as believing that so-and-so involves taking (representing) things to be a certain
way, so does doubting/denying/imagining that so-and-so. Since all are attitudes to the
same proposition, there is a sense in which all involve taking things to be the same
way. However, since agents who believe, deny, doubt, or imagine that so-and-so have
different takes on things, there is also a sense in which the ways they take things to be
are different. Although both senses are legitimate, only the former is at issue when one
asks “Is what is believed/denied/doubted/ imagined true?” To ask this is to ask whether
the way things are taken to be that is common to these cases is the way things really
are. Capturing this common way things are represented to be requires either (i) tak-
ing propositions to be intrinsically representational independent of the attitudes of
agents, or (ii) postulating an ur-attitude—like entertaining a proposition—that is both
inherently representational and part of the characterization of the other attitudes. The
former strategy, which Speaks correctly rejects, was that of the early Russell. The latter
strategy is central to my theory of propositions as cognitive event types. On my view,
for an agent A to entertain the proposition that Amelia talks is for A to cognitively
represent her as talking. A’s act, and the event type of performing that act, represent
Amelia as talking because agents who perform it do. Just as torturing someone is said
to be a violent act because agents who do that are violent and events in which it is done
are violent episodes, so predicating talking of Amelia, and the event type of so doing,
may be said to represent Amelia as talking. From this we derive the truth conditions of
the proposition. It is true at w iff at w Amelia is as she is represented to be by one who
entertains it (at any world-state). If propositions are to be identified with properties of
the world, Speaks must provide a similar story connecting truth, representation, and
the attitudes.
De se attitudes are a special case. At first blush, Speaks’s view that they are 2-place
relations between an agent and a property might seem to hold promise in uniting the
de se with the non de se. In the end, however, this seemingly promising idea founders
on the evidently correct principle TB (true belief).
TB. A’s belief is true iff what A believes is true.
one decides about that, propositions in my sense are not the properties with which Speaks identifies
propositions.
Although those properties that are objects of ordinary beliefs are, on Speaks’s
account, true iff they are instantiated, this does not extend to putative objects of de
se beliefs—like being Rudolf Lingens or being in danger—which aren’t truth bearers
at all. Nor, as he recognizes, does the property view of propositions provide a good
way of retaining the de se and the non de se as beliefs in the same sense. Instead, he
is forced to posit another attitude distinct from belief in order to capture what Lewis
would call belief de se. (The same proliferation is required for other attitudes.) As
I argue in chapter 6, there is no such problem when propositions are taken to be cog-
nitive event types. Not only is TB preserved, but the relation between the de re and
the de se falls out automatically. On the cognitive account, concrete cognitive events
in which an agent entertains a de se proposition are always simultaneously events
in which the agent entertains the corresponding de re proposition (though the con-
verse does not hold). From this it follows that both propositions are truth evaluable,
and that believing the de se proposition guarantees believing its de re counterpart
(but not vice versa). Far from facilitating a satisfying account of the de se, taking
propositions to be the kind of properties that Speaks identifi es them with is an obsta-
cle to giving such an account.
At the end of chapter 5, he raises two interesting problems for his own view of propo-
sitions that extend to other theories as well. The demarcation problem for his theory
is to specify which properties are propositions and which are not. Having identified
truth with instantiation of properties to which we bear certain cognitive attitudes, he
needs to explain why other properties (to which we bear different but related cognitive
attitudes) aren’t true when they are instantiated. After considering some inadequate
ways of drawing the distinction, he takes the problem to remain unsolved. The situa-
tion is different for propositions as cognitive event types.
Although some open questions concerning demarcation remain, there is no funda-
mental problem explaining why certain event types have truth conditions and others
do not. Those that have truth conditions are those that represent things as being one
way or another, in virtue of the fact that they are event types in which agents perform
acts that represent things, in part by predicating properties of predication targets. This
is not the whole story because predication (which involves cognizing certain targets in
one or another way) hasn’t been fully specified, and because the range of other cogni-
tive operations (including applying functions and performing certain function-like
operations) hasn’t been exhaustively explored. I think that enough has been done to
justify optimism about future progress—though, of course, the proof will be in the
pudding.
Speaks’s other unsolved problem—the substitution problem—also generalizes
to my theory. If Bob believes that Amelia talks, then for some x, x = the proposition
that Amelia talks and Bob believes x. Since, on my view, the proposition that Amelia
talks = the cognitive event type in which an agent predicates talking of Amelia, it
follows that for some x, x = that event type and Bob believes x. Since this is what is
expressed by
Propositions as facts
I agree with two tenets of King’s view: that propositions represent things as being cer-
tain ways, and so have truth conditions, and that their intentionality is due to the inten-
tionality of agents. But I don’t fully understand his explanation of the second. It begins
with the claim that sentence (7a) is the fact given in (7b), which consists in “Michael”
standing in a certain “sentential relation” to “swims.” (p. 11)
7a. Michael swims.
b. the fact consisting in “Michael” occurring as the left terminal node that is the
daughter of a node that also dominates the right terminal node at which “swims”
occurs.
Let us call the relation in which “Michael” stands to “swims” in (7a/b) “R.” Next King
identifies another fact, given by (7c), called “an interpreted sentence” (corresponding
to (7a)).
7c. the fact consisting in there being a possible context of utterance c such that
(i) Michael is the semantic value of “Michael” in c, (ii) the property swimming is
the semantic value of “swims” in English, and (iii) “Michael” stands in R, which
in English encodes ascription (predication), to “swims”
2
Speaks raises this possibility, which he attributes to King.
(7c) includes quantification over possible contexts, while also including the English
language, the man Michael, the property swimming, the notion the semantic content of
an expression relative to a context, R, and the notion encoding ascription.3 This was the
view of King (2007); here, he modifies it, explaining how quantification over possible
contexts can be eliminated in favor of quantification over assignments of objects to
variables. Thus (7c’) replaces (7c).
7c. the fact consisting in there being a context of utterance c and an assignment
f of values to (individual) variables such that (i) Michael is the semantic value
of “Michael” relative to c,f, (ii) the property swimming is the semantic value of
“swims” in English, and (iii) “Michael” stands in R, which in English encodes
ascription (predication), to “swims”
The first crucial claim in King’s explanation of the intentionality of the proposition
that Michael swims is the claim that speakers of English have “cognitive access” to the
facts he calls interpreted sentences, including (7c’). (pp. 10–13). Next, he articulates a
principle that allows him to conclude that speakers have cognitive access to proposi-
tions from the fact that they have cognitive access to interpreted sentences. Call a fact
that consists simply of objects o1. ..on instantiating an n-place property P a witness of the
related fact consisting in there being some x1. . .xn Px1.. .xn. King’s principle is that “having
cognitive access to a witness for a fact is a way of having cognitive access to the fact
witnessed.” (p13). His third claim is that (7c’) is a witness to the proposition (7d) that
Michael swims.
7d. the fact consisting in there being some language L, some expressions e and e’ of
L, some syntactic relation R of L, and some context c and assignment f of objects
to variables such that (i) Michael is the semantic value of e relative to c,f, (ii) the
property swimming is the semantic value of e’ in L, and (iii) e stands in R, which
in L encodes ascription (predication), to e’ in some sentence of L
Although there is one further step needed to explain the intentionality of proposi-
tions, there are already some matters to attend to. First, (7c’) is not a witness of (7d),
since the move by which we reach the latter from the former is not bare existential
generalization (as the definition requires); it is more complex, explicitly introducing
the notions of an expression, a language, and a syntactic relation of the language. Of
course, “Michael” and “swims” are expressions, English is a language, and R is a syntac-
tic relation (in which some expressions stand to others in sentences). But (7c’) doesn’t
say that they are, so the move to (7d) imports content not found in (7c’). Since I suspect
King would be happy to add this content to (7c’), I will here take that to have been done.
Second, although (7d) speaks of some language, more is needed to ensure the exist-
ence of propositions in which properties not expressed by a predicate of any existing
3
For R to encode ascription is, I take it, for R to be used by speakers to predicate the property expressed by
the predicate expression that stands in R to the term or terms in question, of the referents of those terms.
language are predicated of objects. This may be done either by adding quantification
over all possible languages to (7c’) and (7d) (in a manner analogous to the treatment of
contexts in King (2007)), or by invoking assignments of properties to predicate vari-
ables in actual languages (in a manner analogous to the use of assignments of objects to
individual variables). Let this also be assumed.
These are matters of detail. The serious questions are What kind of cognitive access
is King talking about? and Is the witness principle true for that kind of access? Suppose
(i) that “cognitive access” to the fact consisting in such-and-such being so-and-so
involves knowing or believing that such-and such is so-and-so. Then the witness prin-
ciple will follow from the claim that when one knows or believes that Po1. ..on, one also
knows or believes that that for some x1. . .xn Px1. . .xn, plus the principle (ii) that one who
knows or believes the latter thereby has “cognitive access” to the fact that for some x1.
x Px1.. .xn. Although this story might sound plausible, I doubt it is King’s view. For one
.. n
thing, he is highly skeptical of the idea that that the proposition that S (which can be
known or believed) is the same as the fact that S (which he seems to suggest cannot).
But without this identity the story requires further explanation, which he doesn’t give,
in order to be plausible. Worse, it is implausible to suppose that fledgling language
users mastering simple sentences like “Michael swims” are acquainted with the “inter-
preted sentence” (7c’) in a way that depends on knowing, believing, or even imagining:
that there is a context of utterance c and an assignment f of values to (individual) variables such
that (i) Michael is the semantic value of “Michael” relative to c,f, (ii) the property swimming
is the semantic value of “swims” in English, and (iii) “Michael” stands in R, which in English
encodes ascription (predication), to “swims”.
This is too complicated for neophyte language users; it also contains concepts—con-
text, assignment, semantic value, R, ascription (i.e. predication), English—that such
language users cannot be assumed to possess. More importantly, the fact that belief
and knowledge are relations to propositions disqualifies them from playing a role in
the explanation of how agents endow otherwise non-intentional facts with intentional
properties by “interpreting” them. To take King’s “cognitive access” to presuppose
propositional attitudes as prerequisites for such “interpretation” would be to destroy
the explanation by assuming what is to be explained. Being aware of this, he doesn’t
appeal to such attitudes, contenting himself with the unexplained phrase “cognitive
access.” To my mind, this trades one problem for another. Although cognitive access is
central to his explanation, the notion has been left too underspecified to bear the load
placed on it.
Setting this aside, we have reached the stage of the explanation at which speak-
ers have cognitive access to the general metalinguistic fact (7d). Somehow “cognitive
access” to this extraordinarily complex fact is supposed to lead fledgling speakers to
“interpret” the relation R* expressed by the following formula.
λxy [there is some language L, some expressions e and e’ of L, some syntactic relation R of L, and
some context c and assignment f of objects to variables such that (i) x is the semantic value of
e relative to c,f, (ii) y is the semantic value of e’ in L, and (iii) e stands in R, which in L encodes
ascription (predication), to e’ in some sentence of L]
R* is a complex two-place relation that holds between Michael and swimming, if (7d)
is a fact. Since King identifies (7d) with the proposition that Michael swims, he calls
R* “the propositional relation.” The idea is, in effect, to treat the fact (7d) as a kind
of pseudo sentence made up of two pseudo words, the man Michael and the property
swimming, standing in the pseudo grammatical relation R*. As always, “grammatical
relations” carry semantic significance. Just as the real grammatical relation R is used by
English speakers to predicate the property expressed by a predicate expression P of the
items designated by Ps argument expressions, so the pseudo grammatical relation R*
is used by anyone who entertains the proposition to predicate swimming of Michael.
For King, this is what it is to entertain the proposition that Michael swims. In short,
in King’s view as in mine, the proposition represents Michael as being one who swims
because agents who entertain it do so.
Although I find this high-level agreement between King and me to be satisfying,
I worry about his account of entertaining a proposition. The account begins with
the problematic claim that by virtue of understanding (7a) we have cognitive access,
in some robust but unexplained sense, to what King calls “the interpreted sentence,”
which is the fact (7c’). Since (7c’) specifies which expression plays the referring role,
and which the predicating role, while further indicating the property predicated and
the object that is its predication target, this “cognitive access” presumably involves
understanding the sentence (7a) as predicating swimming of Michael. Step 2 gets us
from “the witness” (7c’) to the complex general fact (7d) to which it is claimed we also
have the required “cognitive access.” Step 3 portrays us as picking out the highly com-
plex relation R* and conceptualizing (7d) as consisting of Michael standing in R* to
swimming. Step 3 sees agents as undertaking, for some unknown reason, to endow
this fact with intentionality by using it (the pseudo-sentence/proposition) to represent
something else—Michael—as being a certain way—a swimmer—in the way that par-
allels their use of the ordinary sentence (7a) to do the same thing.
Not only is there no explanation of why agents do this, it is anomalous that anything
of this sort should be required. Merely understanding (7a)—which, on King’s account,
is analytically prior to any of his further steps—involves agents using the sentence to
predicate being one who swims of Michael. One who does this represents him as a
swimmer and hence thinks of him as one who swims—which surely is itself a propo-
sitional attitude. Thus, one who understands the sentence (in King’s sense) should be
seen as already bearing a propositional attitude of the most basic kind to the proposition
that Michael swims. To understand the situation in this way is to recognize that mak-
ing sense of the very first step in King’s putative explanation involves presupposing that
agents bear propositional attitudes to the proposition whose intentional properties he
sets out to explain. Since this presupposes some other, conceptually prior, way of enter-
taining the proposition that Michael swims, it seems to me that his explanation fails.
King must not see it this way. Perhaps he takes the first step in the process—namely
as understanding the sentence, in the sense of having “cognitive access” to (7c’)—as
not involving one’s predicating anything of Michael, not representing him as one who
swims, and hence as not entertaining, or bearing any attitude, to the proposition that
he swims. How this can be so is a mystery to me. Since the agents in question use and
understand the sentence which, in effect, tells them to predicate swimming of Michael,
I would have thought that their understanding the sentence and “cognitive access”
to (7c’) would already have them thinking of Michael as a swimmer—and hence as
entertaining the proposition that he is, independent of any explanatorily downstream
“interpretation” of the “propositional fact” (7d). Since, by contrast, King thinks such
interpretation is required, the final step in his journey is to argue that once agents have
embarked on interpreting the propositional fact and relation, there is only one reason-
able way for them to do so. He says:
“But even if we are now convinced that it is [7d]’s propositional relation that we interpret as
ascribing the property of swimming to Michael, we need to say what constitutes interpreting
it. . .It is simply that we compose the semantic values at the terminal nodes of the propositional
relation in the way that we do. [Note the treatment of the fact (7d) as a kind of pseudo sentence
and the relation R* as a pseudo grammatical construction.] In the end this is just a reflex of the
sentential relation R having the semantic significance that it does [i.e. ascription/predication].
When we entertain a proposition, we work our way up the propositional relation, combining
semantic values to yield new semantic values for further combining. [Again propositions as
pseudo sentences.] . . . In the case of [7d], were we to do anything other than ascribe the property of
swimming to Michael, we would not be combining semantic values in a manner that is consistent
with the way we interpret the syntax of the sentence [7a]. It just isn’t coherent to interpret the sen-
tential relation R as ascribing the semantic value of ‘swimming’ to the semantic value of ‘Michael,’
while composing the semantic values Michael and the property of swimming in some other way
as one moves up the propositional relation of [7d]. Semantic values only get composed once in
understanding sentence [7a], and hence entertaining the proposition [7d]. We either do so in a
way dictated by the sentential relation R or not. To do so in the way dictated by the way we interpret
the sentential relation R just is to interpret the propositional relation as encoding ascription. To
summarize, [the proposition 7d] has truth conditions because speakers interpret its proposi-
tional relation as ascribing the property of swimming to Michael.” (13–14, my emphasis)
The sentences emphasized in the passage make it sound as if there are two (simul-
taneous?) interpretations going on here (running in parallel?)—one of the sentence,
which involves the sentential relation R, and one of the proposition, which involves the
propositional relation R*. The two are brought into harmony by the need for consist-
ency, which dictates that the latter agrees with the former. My problem is that I don’t
understand the need for the apparent duality in the first place, in which a pair of inter-
pretations must be brought into harmony. By King’s own account, his elaborate con-
struction requires some conceptually antecedent understanding of the sentence to
provide the facts needed to construct the proposition the interpretation of which must
be made consistent with the conceptually prior understanding of the sentence. If, as
I believe, this understanding of the sentence already requires one to bear an elemen-
tary attitude (entertaining) to a, or the, proposition that represents Michael as a swim-
mer (and nothing more), then, what is to be explained is presupposed at the first step.
Perhaps King will explain where precisely, and why, he disagrees.
In addition to this worry, there is a further, elementary point to be emphasized.
However “interpreting,” and hence entertaining, propositions is ultimately explained,
King is committed to the idea that it always involves understanding sentences. For me,
this is too logo-centric. There are many actual and possible agents, including human
beings, who bear propositional attitudes to propositions presented to them in percep-
tion and non-verbal thought that seem not to be presented to them by any spoken or
written sentence they understand. Although some philosophers may be tempted to
speculation about the “languages of thought and perception” of all possible agents with
perceptions, beliefs, desires, and expectations, one’s theory of propositions shouldn’t
force one to this extremity.4 I will return to a related point below.
The critique offered here is, of course, directed at King’s claims that the use and
understanding of language by agents is explanatorily prior to their attitudes to prop-
ositions, and to their endowment of propositions with intentionality. In an earlier
exchange between us I took the presence of passages like the following, from King
(2007), to pose a temporal problem as well.
“Consider the time at which sentences like ‘Rebecca swims’ first came into existence . . . As should
now be clear, the existence of sentences such as [this] brings into existence facts such as 4b’’
[that bear the same relation to ‘Rebecca swims’ as (7d) bears to (7a)] where let us suppose the
propositional relation doesn’t yet encode the instantiation function [now called ‘ascription’] but
the sentential relation of ‘Rebecca swims’ does. Since we now claim that the propositional rela-
tion encoding the instantiation function [ascription] is part of the fact that is the proposition
that Rebecca swims, 4b’’ is not yet that proposition. Indeed neither the proposition that Rebecca
swims, nor, we may suppose, any other proposition exists yet. Thus it must be that the language does
not yet contain verbs of attitude, modal operators, or that-clauses. However, sentences have truth
conditions . . . As verbs of attitude enter the language speakers begin to talk about structured
contents . . . In short, when English came into existence and prior to it having the resources to talk
about propositions, it brought into existence facts like 4b” . . . As speakers began to attempt to talk
about structured contents by means of that-clauses, they implicitly took these contents to have
the same truth conditions as the sentences with those contents . . . Perhaps it was indeterminate
at first which eligible facts are the structured contents of sentences.But the facts that in the end
4
The point is not to deny that predications that occur in perception or “non-verbal” thought are con-
stituted by the agent’s use of an internal language-like representational system some elements of which
designate the predication targets while other elements designate the properties predicated. The point is to
remain neutral on such speculation. I certainly believe that some instances in which natural language users
predicate properties of objects are constituted by the agents’ use and understanding of the natural language
expressions they employ. However, I neither assert nor deny that all predications of properties, and instances
of entertaining propositions, are similarly constituted by the agent’s employment of an internal representa-
tional system that mediates between the propositions entertained and their worldly subject matter.
are most eligible to be structured contents, propositions, must share the truth conditions of the
sentences whose contents they are eligible to be . . .”5
On this basis, I interpreted King as implausibly claiming that there was a time before
propositions existed when speakers used and understood sentences of primitive lan-
guages (without modal operators, attitude verbs, or that-clauses). Since this seems
clearly in line with the words just quoted, I was surprised to read that he now claims my
earlier interpretation was a misunderstanding. Nevertheless, I am pleased with what
seems to me to be his change in view.
However, as he notes in chapter 4 above, there is still a temporal problem to be faced.
Languages are cognitively complex social institutions. To speak and understand them
agents must have beliefs and intentions about expressions and what they are used to
talk about. They must also have beliefs and expectations about what other agents know
and what they don’t, as well as what is of interest to them and what isn’t. Speakers need
further beliefs and expectations about how their hearers will interpret their words, and
how well they will read the speakers’ beliefs and intentions. Because of this, there is,
I think, no speaking a language by agents who don’t first possess a rich store of propo-
sitional attitudes. This was a problem for King (2007). Since the existential quantifiers
used in specifying the facts with which he identified propositions did not range over
the nonexistent, the existence of propositions at t depended on the existence of one or
more languages at t. From his account of what it is to entertain a proposition it further
followed that no one can entertain a proposition at a time when it doesn’t exist. Thus,
he was forced to the implausible suggestion that there were no propositions or agents
with propositional attitudes before there were languages.
Recognizing the language-of-thought hypothesis as a possible way out, he rightly
did not rest his case on it, and so admitted that we must take seriously the idea “that
strictly speaking our proto-linguistic ancestors did not have propositional attitudes
because propositions didn’t exist then.”6 Nevertheless, he suggested that “they had
some sort of “proto-intentional states”: proto-beliefs and proto-intentions.”7 Summing
up, he put his tentative conclusion as follows:
“Propositions and real intentional states with propositional content come into existence
together. Hence we need not suppose that our proto-linguistic ancestors literally had propo-
sitional attitudes prior to the existence of language and propositions. It is enough to suppose
that they had proto-intentional states not too different in kind from those had by many animals
today.”8
As I said in our earlier exchange, I don’t find this convincing. Whatever these
“proto-intentional states” are, they can’t be relations to representational bearers of
5
King (2007) pp. 60–61, my emphasis. See also the paragraph spanning pp. 66–67.
6
Ibid., p. 66
7
Ibid., p. 66, my emphasis.
8
Ibid., p. 67.
truth conditions, lest they raise the same problems that genuine propositional atti-
tudes do. If the postulated primitive states are not relational in this way, we need to be
told how, if at all, they are representational, and how, if they are not, they provide the
rich conceptual resources necessary to give birth to language. Whereas I don’t think
this can be done, King argued that it must be possible since everyone faces a version of
the same problem.
“[E}ven if propositions existed eternally, there was a time at which no creatures had mental
states with propositional content. Hence, some account must be given of how creatures came
to have propositional attitudes. If we consider creatures immediately prior to the time when
creatures had propositional attitudes and creatures who first had them, some explanation will
have to be given of how the latter managed to get in cognitive contact with propositions. But
in sketching such an account one is faced with the challenge of describing the minds of our
ancestors without using verbs of propositional attitudes. Here again it seems one would have
to invoke proto-intentional states and proto-intentional action as part of the explanation,just
as I did above. Hence, on this score my account is not in any worse shape than an account of which
propositions are eternal.”9
I don’t think this is quite right. From my perspective, King’s focus on language as
the loci of propositions led him to misconceive the problem. The cognitive require-
ments required to speak and understand even a very simple language are complex.
Because of this, antecedent propositional attitudes are required to explain both the
birth of language and the way children acquire it. It was because King denied this,
while tying propositional attitudes to understanding sentences, that he was confronted
with a puzzle. How can agents have the complex cognitive abilities needed to master a
language, and thereby come to have propositional attitudes, without having the attitudes
to begin with? His 2007 answer embraced the deus ex machina of “proto-beliefs” and
“proto-intentions,” which somehow have just the power needed for agents to master
language, and so to acquire propositional attitudes, without having whatever features
of propositional attitudes that make understanding a language necessary for having
them. The suggestion that we all face this sort of puzzle is simply not true.
On my account, all thought and perception involves propositional attitudes.
Consider vision. To see something is to see it as being a certain way (e.g. as red, round,
etc.)—which on my story is to represent it as being so-and-so by virtue of predicating
the property being so-and-so of it. Since this is one of the basic ways of entertaining
a proposition, any creature that can see has propositional attitudes. The same can be
said of nonverbal thinking. To think of something is to think of it as being a certain
way, which is to represent it as being that way by virtue of predicating a property of it.
This, too, is to bear an attitude to a proposition. Hence, nonlinguistic agents capable of
perception and cognition may, depending on their capacities, bear cognitive attitudes
to more or less conceptually rich sets of propositions. This, I maintain, is what makes it
possible to explain the birth and acquisition of language.
9
P. 67, my emphasis.
For this reason, I am pleased to see the big step King takes in this direction with the
following words from chapter 4.
“I believe that many things have content other than sentences of natural languages. Maps, dia-
grams, perhaps pictures and, most importantly for present purposes, perceptual experiences
have contents. In the case of each sort of thing that has content, there will be an account of those
contents in the spirit of the present account of the contents of natural language sentences . . . it is
plausible to suppose that the contents of perceptual experiences have truth conditions. Finally
it seems reasonable to suppose that the contents of perceptual experiences can be objects of
attitudes like belief, desire, etc. But then our prelinguistic ancestors could have had beliefs and
desires whose objects are the contents of perceptual experiences. These attitudes could then fig-
ure in the account of how language, and the contents of natural language sentences came into
existence.” (22–23)
Because his new views concerning types of contents/propositions not tied to language
haven’t yet been presented, no serious assessment of them can now be made. There
is, however, a worry to be registered. Since his account of propositions expressed in
language ties them, and attitudes we bear to them, so completely to sentences and our
cognitive relations to them, it seems likely that the class of propositions to which we
are, on his approach, related by visual experience will be entirely disjoint from the
class of linguistically expressed propositions. The same, I suspect, can be said about
the class of linguistically expressed propositions and the class of propositions to which
we are related by nonverbal thought (if he also recognizes these). This has the poten-
tial for creating problems for the account of how sentential clauses are used to report
the contents of perceptual and cognitive experiences. It may also create problems for
explaining the ways in which language users effortlessly integrate the propositional
information brought to them through language, perception, and nonverbal cognition.
For these reasons, I suspect it is a mistake to start with a thoroughly linguistic account
of propositions expressed in language, with the hope of grafting it on later accounts of
those with which we are nonlinguistically acquainted. Instead, I prefer to start with
a notion of propositions not tied to any single mode of presentation, and to work for
further specification from there. Time will tell which of these research programs is the
more successful.
King closes chapter 4 discussing an objection in my earlier article “Propositions”
that took it for granted that facts in his sense are things that can be referents of ⌜the fact
that S⌝. He says that this was a mistake, indicating that what he then meant (and now
means) by “fact” is “n objects standing in an n-place relation, n properties standing in
an n-place relation, and so on.” (King p.30) On this understanding—which indeed is
what I took him to mean—Annie’s being smart qualifies as a fact, as does Jeff ’s being dif-
ferent from Scott. These, I assumed, were regarded as complex entities the existence of
which were taken to make the propositions that Annie is smart and that Jeff is different
from Scott true. Taking it to be obvious that ⌜the fact that S⌝ designates a fact, if it des-
ignates anything at all, I assimilated Annie’s being smart and Jeff ’s being different from
Scott to the fact that Annie is smart and the fact that Jeff is different from Scott—thereby
reaching the familiar conclusion that the fact that Annie is smart and the fact that Jeff is
different from Scott are what philosophical defenders of facts take to make the proposi-
tions that Annie is smart and that Jeff is different from Scott true. (King pp. 29–30) If
I now understand him correctly, it is this last step that he disputes, when “fact” is taken
as he understands it. Though he doesn’t deny it—either now or in King (2007)—he
doesn’t affirm it either. This being so, the argument to which he objects was based on a
premise to which he was not committed.10
Nevertheless, in chapter 4 above he is, for the sake of argument, willing to assume
that ⌜the fact that S⌝ does designate a fact, while providing evidence that even so, what
it designates is different from what ⌜the proposition that S⌝ designates. It is good that
he does this, because the constructions he considers in attempting to establish this con-
clusion support, by and large, the substitutability of ⌜so-and-so’s being such-and-such⌝
and ⌜the fact that so-and-so is such-and-such⌝ for one another—which, on his meth-
odology, supports the substantive view that he complains my misunderstanding
forced on him. Viewed in this light, his response is an attempt to provide precisely the
empirical evidence my objection requested.11
His evidence consists in the problematic results of substituting one of ⌜the fact that
S⌝ and ⌜the proposition that S⌝ for the other, or for substituting either for ⌜that S⌝,
under various verbs. In many cases such substitution changes meaning, truth value, or
grammaticality. Although King is cautious about interpreting these results, he is right
to suggest that, taken at face value, they do make a prima facie case for distinguishing
the referents of ⌜the proposition that S⌝ from those of ⌜the fact that S⌝, while taking
⌜that S⌝ to be capable of designating those of either. To that extent, his examples pro-
vide a reasonable response to my earlier objection.
Nevertheless, I remain uncertain what facts are supposed to be, and what relation
they bear to propositions. I am also troubled by the observation that his substitution
tests can cut in directions different from those he indicates; as they do with a venge-
ance when one of the designators, K1—K4 (of what he takes to be the proposition that
Michael swims) is substituted for “the proposition that Michael swims” or for “that
Michael swims” in examples (8a,b,c)
K1. there being some language L, some expressions e and e’ of L, some syntactic rela-
tion R of L, and some context c and assignment f of objects to variables such
that (i) the property swimming is the semantic value of e in L, (ii) Michael is
10
In footnote 30, p. 149 of King (2007), to which he directs us in chapter 4, he says “I am not assuming that
expressions of the form ‘the fact that p’ designate what I have called facts throughout the book. It is a sub-
stantive claim that they do so. I will remain neutral on that question here. But I shall call the things that they
designate ‘facts’ in this chapter and assume that they are not propositions.” I am afraid I picked up his usage
in the text while overlooking this footnote.
11
The reader is invited to make such substitutions in (4a-f) and (8b) of chapter 4.
the semantic value of e’ relative to c,f and (iii) e stands in R, which in L encodes
ascription (predication), to e’ in some sentence of L
K2. Michael’s standing in the relation there being some language L, some expressions
e and e’ of L, some syntactic relation R of L, and some context c and assignment f
of objects to variables such that x is the semantic value of e relative to c,f, y is the
semantic value of e in L, and e stands in R, which in L encodes ascription (predica-
tion) to e’ in some sentence of L to the property swimming
K3. the fact that there is a context c and assignment f of objects to variables such that
Michael is the semantic value, relative to c and f, of some expression e of some
language L, the property of swimming is the semantic value of some expression
e’ of L, and e stands in R, which in L encodes ascription (predication), to e’ in
some sentence of L
K4. the fact that Michael stands in the relation there being some language L, some
expressions e and e’ of L, some syntactic relation R of L, and some context c and
assignment f of objects to variables such that x is the semantic value of e relative to
c,f, y is the semantic value of e in L, and e stands in R, which in L encodes ascription
(predication) to e’ in some sentence of L to the property swimming
These substitutions produce apparent absurdity. Does this show that the fact desig-
nated by K1—K4 really isn’t the proposition King takes it to be? If not, why do the
changes rung by the substitutions he mentions in his examples (4–8) show that ⌜the
proposition that S⌝ and ⌜the fact that S⌝ designate different things? As I said in connec-
tion with Speaks’s discussion of the substitution problem, a theorist like King will have
to accept some of the seemingly absurd results of substitution as true (if the unprob-
lematic sentences into which they are substituted are true). This contributes to my
uncertainty about what the results of substitution in his examples (4–8) really show.
I am similarly unconvinced by his rebuttal of my example (9), which appears to
indicate that what is regretted—namely the fact that Pam is pregnant—is some-
times believed, and hence is nothing more than the proposition that Pam is pregnant
(contra King).
9. Pam regrets that she is pregnant. Although her parents don’t realize it yet, in time
they will come to believe it.
King mentions two ways of accommodating this data. The first, derived from Parsons
(1993), takes the antecedent of the occurrence of “it” in the final clause to be an occur-
rence of “^[Pam is pregnant]” in the complex singular term “c^[Pam is pregnant]”
that is the object of “regret.” The first of these designates the proposition that Pam is
pregnant, while the second designates the fact that Pam is pregnant—which the func-
tion designated by “c” assigns as value to the proposition as argument. Thus, King con-
cludes, the truth of (9) can be made compatible with the distinctness of the fact from
the proposition.
This strikes me as too quick. Although the explanation requires there to be a com-
plex term for the fact in a position in which its argument expression can serve as the
antecedent of a later pronoun, this is an accidental feature of the example chosen. For
example, consider (10)
10a. Pam regrets something that her parents don’t yet realize, but will soon come to
believe.
b. Something Pam regrets is now merely suspected by her parents, but will soon be
believed by them.
Not only do these seem fine, they seem to entail that some one thing can be both regret-
ted and believed, or both regretted and suspected. It also seems obvious that the same
things that can be known can, and are, believed—despite the fact that King takes belief
to require propositions as objects, while seeming to take examples like (11) (his (4c)) to
show that the objects of knowledge are the non-propositional referents of clauses ⌜the
fact that S⌝.
11. The fact that the moon causes the tides is well known.
11. The book I just stole from the library is on my desk. It was written in 1801 and
has been translated into many languages.
According to King, the antecedent of “it” is the occurrence of “the book I just stole
from the library” in the first sentence, which he takes to designate a concrete object—
a copy of said book—while “it” designates the abstract object itself. This is taken to
show that antecedents and anaphors can refer to different but related things, which, he
thinks, is how (9) may be understood. Again, I am skeptical. I may truly remark “The
book I stole from the library that is now on my desk is the same as the book Mary stole
from her library that is now on her desk.” Here I am talking about a book type—e.g. War
and Peace—which is both sitting on my desk, and sitting on Mary’s, having been stolen
by each of us from our respective libraries. What this illustrates is also at work when
one says that one wrote the same word on the board twice on separate days. Abstract
objects can have properties—like being written on the board, being on a desk, being
stolen from the library, and the like, by virtue of properties had by their tokens. Thus,
the definite description in (12) can be understood as denoting the book itself, rather
than a copy of the book, even though the book is truly said to be sitting on the desk (by
virtue of the fact that the copy is).
This completes my critical remarks. Despite the inconclusive sparring over the ref-
erents of ⌜the fact that S⌝, ⌜the proposition that S⌝ and ⌜that S⌝, the crucial points
in my discussion of King’s chapter 4 are the challenges I pose for his explanation of
the intentionality of propositions, and my related concern that he ties the proposi-
tions expressed by sentences, and the attitudes we bear to them, too closely to those
sentences and our understanding of them. Despite these points of disagreement, I am
pleased that King now recognizes propositions expressed by pictures and diagrams,
and that he takes perceptual experience to have propositional content. These additions
expand the broad areas of agreement between our two views, both of which remain
works in progress.12
12
Thanks to Brian Bowman for helpful comments on this chapter.
Response to Speaks
Speaks begins by raising two objections to Soames’s and my views that he thinks apply
to both of them in virtue of a common structure they share. I’ll only consider his objec-
tions as they apply to my view.
The first objection runs as follows. I make free use of properties. They are constit-
uents of my propositions. Hence, Speaks claims, I accept the ontology of his theory
according to which the proposition that Amelia talks is the property being such that
Amelia talks. Speaks, by contrast, does not accept the existence of the “representational
surrogates” that I claim are propositions. So:
(S1) The ontology of Speaks’s theory is included in my theory and not vice versa.
Further, for me, properties are constituents of propositions and in trying to explain
the cognitive access agents have to propositions, I simply assume that agents have
cognitive access to the properties that are their constituents. So since Speaks thinks
that propositions are just properties, it seems that I must admit that it is no easier to
have cognitive access to the entities I claim are propositions than it is to have cognitive
access to the entities Speaks claims are propositions. Hence:
(S2) Our cognitive access to the entities Speaks claims are propositions can-
not be harder to explain than our cognitive access to the entities I claim are
propositions.
From (S1) and (S2), Speaks reasonably concludes that the only reason for preferring
my theory to his would be that my theory explains something not explained by his
((S1) and (S2) together seem to entail that Speaks’s sparser-than-my ontology yields
a theory of propositions and explains how we have cognitive access to them at least as
well as my theory). But Speaks does not see what that would be.
In response, let me say that I am happy enough to accept (S2), but I reject (S1).
Perhaps more importantly, I reject a presupposition of Speaks’s argument.
Taking the former point first, as my criticisms of Speaks in Chapter 7 suggested,
when I get serious about thinking about properties, I ultimately think that sharing of
properties should make for real, objective similarity. But this means that, first, I find it
hard to swallow the idea that there are negated or disjunctive properties.1 For sharing
properties like not being a skier or being a skier or a planet obviously does not make for
real similarity. That in turn means that I will want no part of properties like being such
that George W. Bush is not a skier or being such that George W. Bush is a skier or there
are eight planets in Earth’s solar system, which Speaks alleges are expressed by the sen-
tences “George W. Bush is not a skier.” and “George W. Bush is a skier or there are eight
planets in Earth’s solar system.” Second, even sharing of the properties Speaks alleges
are expressed by simple sentences without negation or disjunction does not make for
objective similarity. Everything possesses the property being such that George W. Bush
was a poor student, which Speaks alleges to be expressed by the sentence “George
W. Bush was a poor student.” But this does not make everything objectively similar to
everything else. So I would want no part of the property being such that George W. Bush
was a poor student either. Hence, I do not accept all of Speaks’s ontology, and so (S1) is
false.2
Perhaps more importantly, Speaks’s objection here presupposes that the properties
he alleges are propositions are able to play the role propositions are intended to play
as well as my candidates for propositions do. After all, the appeal to Speaks’s sparser
ontology only has bite if his ontology can do everything that mine can. However in
Chapter 7 I gave reasons for thinking that Speaks’s properties are not suited to play
the role of propositions. But then I have given reason for rejecting the presupposition
Speaks’s argument requires here.
Speaks gives a second objection to my theory and that of Soames that he thinks
applies to both theories in virtue of their shared structure before turning to specific
criticisms of each theory. However, in this case he gives a more specific version of the
objection when he turns to specific objections to my view. Hence, I consider this more
specific version below rather than the more general version stated at the outset.
The first objection Speaks raises specifically to my account of propositions must be
stated with some care, since it raises some complex issues. Speaks correctly claims that
I hold that propositions have their representational properties essentially. He then
1
More cautiously, properties expressed by negated or disjunctive predicates, like “is not a skier” and “is
either a skier or a planet”.
2
One response to my concerns here would be to invoke something like a Lewisian theory of naturalness.
One could then say that sharing of the perfectly natural properties makes for objective similarity, but not
so when one gets down to non-natural properties like those Speaks alleges are expressed by the sentences
mentioned in the text. I have tended to resist this picture because I tend to be attracted to a sparse account of
properties whereby the only properties that exist are highly natural. The remarks in the text here reflect this
attraction.
worries that as a result of this, I cannot explain why/how propositions have represen-
tational properties/truth conditions. For, he says, you cannot explain why a thing has
any essential property. Speaks considers his essentially wooden desk—Fred—and asks
whether we can explain why it is wooden. Speaks thinks that we cannot, since Fred
couldn’t exist without being wooden and so we cannot explain how it came to be so. At
most we can explain how Fred came into existence. Similarly, Speaks claims, given that
I hold that propositions have their truth conditions essentially, at most I can explain
how propositions came into existence. Further, Speaks adds, explaining how proposi-
tions came into existence is not the same thing as explaining how they have truth con-
ditions. To show this, he provides the following example. Suppose there is a proponent
of the view that propositions are structured, having objects, properties and relations as
constituents, and who thinks of propositions as traditionally conceived: namely, they
have their truth conditions by their very natures and independently of minds and lan-
guages. Now suppose I make a white paper airplane. Consider a singular proposition
about the airplane, say the proposition that o is white, where o is the airplane in ques-
tion. In explaining how this proposition came to exist, the theorist should say that by
making the plane, the proposition was brought into existence. Let’s call this the white
plane explanation. But, Speaks says, in explaining in this way how the proposition that
o is white came into existence, have I thereby explained how/why the proposition has
truth conditions? He correctly answers that I clearly have not. Nothing about the mak-
ing of the plane explains how/why the proposition that o is white has truth conditions.
Hence, in this case explaining how/why the proposition that o is white came into exist-
ence does not explain how/why that proposition has truth conditions.
Now Speaks claims that my explanation of how/why the proposition that Michael
swims came into existence is relevantly similar to the white plane explanation above
and so, like the latter, does not explain how/why the proposition that Michael swims
has truth conditions. If Speaks were right here, this would be a very damning conse-
quence since I have insisted that a theory of propositions must explain how/why prop-
ositions have truth conditions and have rejected various theories for not being able to
do this.
Speaks’s argument for the claim that my explanation of how the proposition that
Michael swims (came into existence and) has truth conditions is relevantly like the
white plane explanation, and so doesn’t really explain why that proposition has truth
conditions, begins with some claims about what I call the propositional relation of the
proposition that Michael swims (PR). I claim that the proposition that Michael swims
consists of Michael standing in a certain relation—PR—to the property of swimming.3
I claim that the following two-place relation is PR—the propositional relation of the
3
I sometimes add that the proposition consists of Michael standing in the propositional relation to the
property of swimming and the propositional relation possessing the property of encoding ascription. However,
I agree with Speaks that which of these things I identify with the proposition makes little difference. See
Chapter 8 note 5 of the present work.
4
In Chapter 4 I suggested quantifying over syntactic relations instead of having a particular syntactic
relation be a component of PR. I am going to ignore that here and I don’t think it affects anything I say (see
note 9 below). If it does, that may be a reason for not going the route suggested in Chapter 4 and quantifying
over syntactic relations.
5
Chapter 8 p. 153 of the present work. I follow Speaks here in talking about constituents of complex
properties and relations. I prefer to reserve the word “constituents” for talking about (certain) elements of
propositions. In what follows, I eventually lapse into my favored terminology and talk of the components of
complex properties and relations.
6
I am not sure why Speaks thinks K3 follows from K2, since K2 would seem to allow PR to exist and
encode nothing. But if that were the case and if PR related two things x and y, I don’t see why the resulting
fact would be true iff x instantiates y. Perhaps Speaks thinks that for PR to relate anything, it must encode
ascription.
plane provides no explanation of why the proposition that o is white has truth condi-
tions. Speaks makes these points in the following passage:
The role played by speakers is simply to make it that case that [PR] relates some things; once they
have brought into existence a fact of the form [PR](x,y) there’s no further work for them to do
in giving this fact its representational properties. But, given this, it seems like the role played by
speakers on this theory is uncomfortably close to the role played by our paper airplane maker in
the example discussed above.
We can focus this worry in the way mentioned above, by asking: what relation does the inter-
pretation of [PR] bear to the representational properties of FAST which the making of the paper
airplane does not bear to the proposition that that paper airplane is white?7
Now I agree with Speaks that if I held K1-K3, I would not have explained how/why
propositions have truth conditions. Details aside, that PR encodes ascription is a big
part of the explanation of how/why propositions have truth conditions. But if PR exists
necessarily and encodes ascription (at least whenever it relates things), then speakers
could play no role in its so doing. We would in effect be positing a necessarily existing
relation that gives propositions their representational oomph. That doesn’t look like
much of an explanation of that oomph.
But I don’t hold all of K1-K3. In particular, I reject K1. As to K2, I hold that PR
encodes ascription essentially—in every world in which it exists, it encodes ascription.
Finally, I accept K3. I am sure Speaks will find my rejection of K1 puzzling. Recall that
his reason for thinking me committed to K1 is that PR does not have any contingently
existing concrete objects as constituents. This suggests that Speaks thinks that the only
reason a property or relation could fail to exist is if it is something like the property of
loving Annie, which contains the concrete, contingently existing Annie as a constitu-
ent. At deprived worlds where Annie sadly fails to exist, so too will properties that have
her as a constituent. Now I agree that properties and relations can fail to exist for this
reason. But I think that some properties and relations that don’t (or don’t obviously)
have contingently existing concrete objects as constituents nonetheless exist only con-
tingently. Before moving on, let me note that I am not alone in this. Robert Stalnaker
[2010] writes:
Consider, for example, colour properties—paradigms of purely qualitative
properties. If there are metaphysically possible worlds with radically
different physical laws, perhaps worlds without light or other kinds of
electromagnetic radiation, then there will be worlds in which nothing is
or could be colored, and I think it would also be reasonable to conclude
that the color properties would not exist in such a world.8
7
Chapter 8 p. 154 of the present work. FAST is the proposition that Michael swims.
8
Stalnaker [2010] p. 23
I won’t argue that matter here, but I also think that if certain views about what it is to be
a species are correct (e.g. if it involves reproductive isolation) then it is plausible that
properties like being a member of species X exist only contingently.
Fine and good, but why think relations like PR exist contingently? Because it is plau-
sible to think that some of the properties and relations that are components of PR exist
contingently. Here I’ll focus on the syntactic relation R that the words in the English
sentence “Michael swims” stand in. But I think that a similar case could be made for
other properties and relations that are components of PR.9
It seems pretty clear that at lifeless possible worlds, English doesn’t exist. At such
worlds, there are no language users and no facts about language users; and it seems
implausible that English could exist in the absence of such facts. Consider a lifeless
world w. Does the fact that English doesn’t exist at w consist in there being no tokens of
English expressions at w? It doesn’t seem so. There are many English sentences (types)
that have no tokens in the actual world. But it seems very plausible that these sentences
exist in the actual world nonetheless. After all, we do say that there are English sen-
tences that have no tokens. But this suggests that for English to fail to exist at w, it can’t
merely be a matter of tokens of English sentences failing to exist at w. The types must
fail to exist there too. But now if the English sentence type “Michael swims” fails to
exist at w, it seems plausible to suppose that this is because the word types “Michael,”
“swims” and the syntactic relation R all fail to exist at w. But since R is a component of
PR, this means that PR won’t exist at w either.10
So the picture is that agents brought language into existence and in so doing brought
into existence expression types and the syntactic relation R interpreted as encoding
ascription. In so doing, they brought into existence PR, also interpreted as encoding
ascription. Having done this, they provided the proposition FAST with its representa-
tional properties. So here, unlike the case of the white plane explanation, the explana-
tion of how speakers brought the proposition into existence does explain how/why the
proposition has truth conditions. This is because in the present case, unlike the case of
the white plane explanation, the explanation of how the proposition came into exist-
ence includes an explanation of how speakers brought PR into existence, where the
latter gives the proposition FAST its representational oomph.11
Speaks’s second objection concerns how expressions like “Michael” and “swims” get
their semantic values. Speaks correctly points out that on my view, part of the expla-
nation as to why/how propositions have truth conditions is that speakers interpret
9
E.g. I suspect the property of being a syntactic relation that encodes ascription exists contingently. See
note 4.
10
Obviously, I presuppose here that if a component of a relation fails to exist at a world, so does the rela-
tion. Speaks seems to accept this too.
11
It should be clear, then, that I disagree with Speaks when he says “The worry is that we are getting
too much for free from the nature of [PR]—to which no thinking subject contributes anything—to get a
satisfying explanation, in terms of thinking subjects, of the representational properties of propositions.”
(Chapter 8 p. 154 of the present work; my emphasis). Thinking subjects brought [PR] into existence. So we do
get an explanation of the representational properties of propositions in terms of thinking subjects.
syntactic relations, and so propositional relations, in the way that they do. But in
order to explain how speakers interpret the syntactic relation in the sentence “Michael
swims,” we must assume that “Michael” has Michael as its semantic value and “swims”
has the property of swimming as its semantic value. In general, in order to explain
speakers interpreting any syntactic relations we must assume that lexical items have
semantic values. But then this means that (some) lexical items having semantic values
is explanatorily prior to speakers interpreting syntactic and propositional relations,
and so explanatorily prior to the existence of propositions. This in turn means that
the account of how (some) lexical items come to have semantic values cannot appeal
to agents having propositional attitudes whose objects are propositions expressed by
natural language sentences. Speaks worries that plausible accounts of how expressions
get their semantic values will have to appeal to such propositional attitudes, and so will
be inconsistent with my account of propositions.
Speaks is correct that on my view some lexical items having semantic values is
explanatorily prior to the existence of propositions expressed by natural language
sentences. However, as I said in Chapter 4, I think that things other than natural lan-
guage sentences have truth evaluable content. As I indicated, maps, diagrams, perhaps
pictures and, most importantly for present purposes, perceptual experiences all have
contents.12 For each sort of thing that has content, there will be an account of those con-
tents in the spirit of the present account of the contents of natural language sentences.
In the case of perceptual experiences, there is still considerable controversy regarding
what their contents are like. I believe it is plausible to suppose that the contents of per-
ceptual experiences have truth conditions; and that these contents can be the objects of
propositional attitudes. As I indicated in Chapter 4, this means that our prelinguistic
ancestors could have had beliefs and desires whose objects are the contents of percep-
tual experiences. These attitudes could then figure in the explanation of how lexical
items acquired semantic values.
So I am committed to explaining how some lexical items acquired semantic values
without appealing to propositions expressed by sentences of natural language. Agents
having propositional attitudes towards the contents of perceptual experiences can fig-
ure in the explanation of how agents secured semantic values for these lexical items.
Note that once we have explained how agents secure semantic values for a few lexical
items, form sentences, interpret syntactic and hence propositional relations, and so
bring propositions expressed by natural language sentences into existence, attitudes
towards these existing propositions expressed by natural language sentences can then
play a role in the account of how new lexical items get semantic values. So the restric-
tion that only attitudes towards the contents of perceptual experiences can be appealed
to in explaining how lexical items get semantic values only applies to the first lexical
items that get semantic values. As I’ve said, there is still considerable controversy about
12
Of course some deny that perceptual experiences have contents.
the nature of the contents of perceptual experiences but I am here assuming they are
truth evaluable. However, many other questions remain including what sorts of prop-
erties figure in these contents.13 Still, the claim that one can give an account of how the
first lexical items got their semantic values by appealing only to propositional attitudes
that have as their objects the contents of perceptual experiences does not strike me as
implausible. Hence, I can reasonably hope that my theory of propositions is consistent
with an account of how agents got language up and running and hence brought propo-
sitions expressed by natural language sentences into existence, contrary to the worry
Speaks is raising here.
As just indicated, I think that perceptual experiences have truth evaluable contents;
and as I indicated both above and in Chapter 4, I would expect there to be a theory of
their contents that is in the spirit of my theory of the contents of natural language sen-
tences. Speaks’s final objection purports to cast doubt on whether such an account of
the contents of perceptual experience will be forthcoming.
In considering how a view of the contents of perceptual experiences that is in the
spirit of my view of the semantic contents of natural language sentences might look,
Speaks begins with a frog’s visual experience of a fly sitting on a green leaf. Call this
experience FLY. Speaks proposes to simplify the example by supposing that the con-
tent of FLY is a singular proposition that predicates the location L of the fly. Call this
proposition fly-at-L. Speaks just assumes that fly-at-L can be expressed by some sen-
tence S of natural language. (As we’ll see, I do not make this assumption.) On a view
like mine, fly-at-L consists of the fly standing in some propositional relation [PR] to
location L.
Now, Speaks asks, what did the frog do to bring this proposition into existence?
Since on my view, in the case of natural language sentences and the propositions they
express, speakers endow propositional relations with truth conditions, and so bring
propositions into existence, by interpreting syntactic relations and thereby interpret-
ing propositional relations, Speaks assumes that in the present case I must say that the
frog does something that constitutes interpreting the propositional relation of fly-at-L,
thereby endowing the proposition with truth conditions. Obviously it cannot be the
frog interpreting sentential relations. And so we cannot view [PR] as being in part built
out of sentential relations as we have been doing to this point. Given Speaks’s assump-
tion that fly-at-L can be expressed by a sentence S, [PR] is the propositional relation
of the proposition expressed by sentence S. This means, Speaks thinks, that we must
somehow alter our account of propositional relations so that [PR] can be the proposi-
tional relation of the proposition expressed by S and the proposition that is the content
of FLY, since these are the same proposition (fly-at-L). Suppose, Speaks says, that we
have somehow done that.
13
See Siegel [2006] for a good discussion of the relevant issues and an argument to the effect that the class
of properties that figure in the contents of perceptual experience is considerably wider than one might have
thought. If Siegel is right, that is good news for me!
But we still haven’t said how the frog brings fly-at-L into existence by endowing
[PR] with semantic significance thereby endowing fly-at-L with truth conditions. As
I noted above, Speaks assumes that this will be a matter of the frog interpreting [PR].
But what constitutes his doing that? Speaks thinks the best answer I can give is that the
frog interprets his perceptual experience, thereby endowing it with truth conditions/
content. And Speaks thinks that this claim is implausible. For, as Speaks notes, the
phenomenal character of FLY is fixed independently of the frog allegedly interpreting
his experience (if not, Speaks rightly asks, what is the frog interpreting?). On one ver-
sion of intentionalism Speaks finds plausible, for FLY to have a certain phenomenal
character just is for it to have a certain content.14 But then if phenomenal character
is fixed independently of the frog interpreting his experience, its content is too. This
means that the explanation of how FLY gets its content by the frog interpreting his
experience and thereby [PR] that Speaks attributes to me cannot be right. Thus, Speaks
claims, I no longer have an explanation of the existence of representational properties
of fly-at-L.
Before providing a detailed response to Speaks’s objection here, it will be use-
ful to think about what it would be to provide an account of the content of some
non-linguistic entity, say a map, that is in the spirit of my account of the semantic
contents of natural language sentences. Suppose we are given a very simple map with
points on it representing three small towns. These dots are the analogues of words in
a sentence and like them have semantic values; the towns they represent. The spatial
relations between the dots in the map are the analogues of sentential relations and
like the latter have semantic significance; they represent spatial relations between the
towns in question.15 The “propositional relation” of the proposition that we assign to
the map will be built out of the spatial relations between the dots on the map. Just as
in the case of the sentential relation and the propositional relation of a proposition
expressed by a sentence, the propositional relation of the map is interpreted the way
it is because the spatial relations among the dots on the map are interpreted the way
they are. Crucial point: is there any guarantee that there is some notion of propositional
relation such that the propositional relation of the proposition expressed by the map is
the same as the propositional relation of a proposition expressed by a natural language
sentence? No, and in the present case it seems to me doubtful that there is such a notion
of propositional relation. If that is right, then no natural language sentence will express
the same proposition as the map. Personally, I find this claim quite plausible.
14
Speaks only appears to need the weaker claim here that some phenomenal properties of a perceptual
experience are identical to representational properties of the perceptual experience. For if that is so, and
the former are fixed independently of the frog interpreting [PR], the latter must have been as well. And that
means that at least some of the representational properties of a perceptual experience are fixed indepen-
dently of the frog interpreting his experience.
15
I don’t think this is actually quite right and things are a bit more complicated (I think the spatial relations
between dots in the map are actually more like words too than like sentential relations; the analogue of a
sentential relation for the map seems to me to be that the dots and spatial relations are configured in a certain
way in the map). But this is good enough for present purposes.
about Speaks’s worry that the account I will be forced to give of how the frog brings
the proposition fly-at-L into existence will be inconsistent with a plausible inten-
tionalist thesis? Recall that Speaks assumed that I would have to claim that the
propositional relation of the proposition that is the content of FLY—the frog’s
perceptual experience—is endowed with semantic significance, thus bringing the
proposition fly-at-L into existence, by the frog interpreting his perceptual experi-
ence. But, Speaks claims, the phenomenal character of FLY is fixed independently
of his interpreting his experience. And it is plausible to hold the intentionalist
claim that fixing phenomenal character of a perceptual experience just is fixing its
content (at least in part). But then the claim cannot be right that the frog brings the
proposition fly-at-L, that is the content of his perceptual experience, into existence
by interpreting his experience. Hence, according to Speaks, I am left without an
account of how the frog brings fly-at-L into existence and endows it with represen-
tational properties.
Where Speaks’s objection goes wrong is in assuming that I will claim that the
propositional relation of fly-at-L gets its semantic significance—thus bringing fly-at-
L into existence—by the frog interpreting his perceptual experience. As indicated
above, I think Speaks assumed this because my account of how speakers endowed
the propositional relation of the proposition that Michael swims with semantic
significance appeals to speakers interpreting syntactic relations. From this, I think
Speaks inferred that my account of the contents of other things will always appeal
to the same mechanism as to how propositional relations are endowed with seman-
tic significance, bringing propositions into existence: an agent interpreting some-
thing as a result of which the propositional relation is itself interpreted by the agent.
Hence, the only plausible thing he saw for me to say in this case was that the frog
interpreted his perceptual experience as a result of which the propositional rela-
tion of fly-at-L is endowed with semantic significance. However, as we saw from the
map example discussed above, accounts of contents of non-linguistic things in the
spirit of my account of the contents of natural language sentences need not, and in
general could not, explain how the propositional relations of propositions expressed
by non-linguistic entities acquire semantic significance by appealing to the same
mechanism whereby propositional relations of propositions expressed by natural
language sentences acquire semantic significance. In the present case, assuming the
intentionalist claim that Speaks finds plausible to the effect that fixing phenomenal
character of a perceptual experience is fixing its content, my account of the contents
of perceptual experience can hold that whatever fixes the phenomenal character of a
perceptual experience endows its propositional relation with semantic significance
and brings the proposition that is its content into existence. We thereby explain how
the proposition that is the content of the perceptual experience comes into existence
with representational properties in a way that is consistent with the intentionalist
thesis Speaks finds plausible.
Response to Soames
Soames’s first objection concerns my notion of speakers having cognitive access to sen-
tences of their languages, construed as facts in my sense; to what I called interpreted
sentences in Chapter 4 of the present work and elsewhere; and to the facts that I claim
are propositions. In the end my explanation for how/why propositions have truth
conditions is that speakers interpret the propositional relation of the proposition that
Michael swims as having a certain semantic significance. As I’ve said, speakers inter-
pret the complex two-place relation between Michael and the property of swimming
as ascribing the property to Michael. Now I claimed that in order for it to be plausi-
ble that speakers interpret the propositional relation of the proposition that Michael
swims, they must have some sort of cognitive connection or cognitive access to the fact
that I claim is the proposition. The idea here seems to me simple and obviously correct.
Consider some fact that I have no cognitive connection to: say some fact on the other
side of the galaxy. Surely it would be very implausible to suppose that I interpret some
relation that is a component of this fact in a certain way. I take it that it is implausible
precisely because I have no cognitive connection to the fact in question. I also claimed
that it appears that speakers have cognitive access to the fact that is the proposition that
Michael swims in virtue of deploying sentences of their language, since by the time
they are deploying such sentences they have propositional attitudes towards the con-
tents of the sentences they are using. Hence, propositions must exist at that time, and
that means that speakers must be interpreting their propositional relations.
My thought was that speakers clearly have a cognitive connection to the sentences
of the languages they employ, where I thought of sentences themselves as facts in my
sense (since they are arguably properties (word types) standing in a (syntactic) rela-
tion). I then introduced the notion of an interpreted sentence, which is just the sen-
tence taken together with the semantic relations the lexical items in the sentence bear
to their semantic values, as well as the semantic values themselves. Thus, the inter-
preted sentence is just a slightly “larger” fact than the fact that is the sentence. I claimed
that having cognitive access to a given sentence also gives one cognitive access to the
interpreted sentence corresponding to it. Finally, I considered a fact consisting of
objects o1,. . .on standing in the n-place relation R and called it a witness for the fact of
there being x1,. . .,xn such that Rx1,. . .,xn. In such a case, let’s say that first fact is a witness
with respect to o1,. . .,on for the second. I claimed that having a cognitive connection to a
witness fact is sufficient for having cognitive access to the fact it witnesses.
Now since interpreted sentences are witnesses to the facts that are the propositions
expressed by the corresponding sentences (with respect to language, lexical items, con-
text etc.), this means that having cognitive access to an interpreted sentence suffices for
having cognitive access to the relevant proposition. We thereby have an explanation
of how speakers have cognitive access to the facts I claim are propositions in virtue of
deploying the relevant sentences—speakers’ cognitive access to sentences gives them
cognitive access to interpreted sentences which in turn gives them cognitive access
to the facts I claim are propositions. Hence we can see how speakers could interpret
the propositional relations of these propositions, thereby endowing them with truth
conditions.
Soames objects that the notion of cognitive access/connection that is central to my
explanation here of how propositions acquire truth conditions is unexplained and
hence “too underspecified to bear the load placed on it.”16 Further, given that I haven’t
explained what sort of cognitive access I am talking about, Soames sees no reason to
think that having cognitive access in the sense I require to a witness fact suffices for
having cognitive access to the fact it witnesses. There is a sense in which I agree with
his complaint. In King [2009], where I first introduced the notion of cognitive access/
connection in giving the explanation of how/why propositions have truth conditions,
I wrote:
I’m going to continue to speak loosely about speakers having cognitive access or a cognitive con-
nection to this or that fact. At some point I need to get more serious about the sort of cognitive
connection required. But this is not that point.17
So I acknowledge that in using the notion of cognitive access/connection that I have not
fully specified in my explanation of how/why propositions have truth conditions, I am
issuing a promissory note that at some point needs to be cashed in. Though I can’t fully
cash it in presently, perhaps I can make a small down payment by way of saying a few
things that may be of help.
As we have seen, the cognitive access to the sentence, interpreted sentence and prop-
osition that I appeal to is used to explain how we go about interpreting the sentential
and propositional relations. Hence, the operative notion of cognitive access must be
such that having cognitive access to a fact puts one in a position to interpret features
of the fact in certain ways. You must in some sense have the fact in mind in such a way
as to be in a position to interpret its features. But, as Soames noted, since the cogni-
tive access in question is being used to explain how propositions expressed by natural
language sentences are representational, and so have truth conditions, the cognitive
access in question cannot be a matter of having any propositional attitudes towards
propositions expressed by natural language sentences (though it could involve having
attitudes towards the contents of perceptual experiences).
Now I take it that it is clear that speakers have sentences, and even interpreted sen-
tences, in mind in the relevant sense. They are cognitively in a position to interpret the
sentential relations of these facts. But now why think they are thereby in a position to
interpret the propositional relations of the facts that I claim are propositions, which
the relevant interpreted sentences witness?18 That is, why think that if one has a witness
16
Chapter 9 p. 171
17
p. 269
18
Soames pointed out that interpreted sentences are not quite witnesses for the relevant propositions. But
he and I agree, I think, that that they are close enough to being so that the ways in which they fall short won’t
affect my argument.
fact in mind in the relevant sense, one is thereby in a position to have the fact witnessed
in mind? Here is a tentative, rough suggestion.When one has a given fact in mind, one
is able to abstract from certain features f1,. . .,fn of a fact one has in mind, thereby having
in mind the fact witnessed with respect to f1,. . .,fn by the original fact. Not only does it
seem plausible that one is able to abstract from features of a witness fact in this way,
thereby having the fact it witnesses in mind, but it also seems plausible that we do this
in moving from having a sentence in mind to interpreting the propositional relation
of the proposition expressed by the sentence and thus understanding the sentence.
Consider the sentence “Michael swims”:
1.
Michael swims
and suppose a speaker has the sentence in mind. The speaker then accesses the seman-
tic values of the sentence—relative to the context of utterance—with the result that the
speaker now has the interpreted sentence in mind:
1IE.
Michael swims
2007
Melbourne 200 7
nee 2000
Melbourne 200
2007
220
2007
20
2007
Melbourne 20
0007
2
07
Mellbouurne
7
Melbbourrn
Me lbo
Me
At this point, the speaker abstracts from all the now irrelevant features of the inter-
preted sentence: the words in it, what language it is in, the context in which the utter-
ance was made, and so on. The speaker thereby has the proposition that Michael swims
in mind, which the interpreted sentence witnessed with respect to a language, context,
lexical items and so on. Hence, the speaker is in a position to interpret its propositional
relation, and thereby entertain the proposition.
As I said, this account, of why having cognitive access to the witness for a fact puts
one in a position to have cognitive access to the fact it witnesses, is tentative and it may
be that there is a better account to be found. However, it does seem to me plausible,
along the right lines and hence promising so far.
Let me close my response to Soames’s present objection by noting that he doesn’t
seem well placed to make this sort of objection to my view. He objects that I use an
unexplained notion, cognitive access/connection, as a central part of my explanation
as to how/why propositions have truth conditions. As indicated, he complains that
the notion “has been left too underspecified to bear the load placed on it.”19 But cen-
tral to Soames’s own explanation of how/why propositions have truth conditions is his
notion of predication, which he explicitly takes to be primitive. Hence Soames’s com-
plaint about a primitive/unexplained notion playing a central role in the explanation
of how/why propositions have truth conditions applies to his own view.
Soames’s second objection to my view concerns what it is to entertain e.g. the propo-
sition that Michael swims. As we have just seen, one way to entertain the proposition
starts with having cognitive access to the sentence “Michael swims,” which I have just
cashed out in terms of having the sentence in mind in such a way as to be in a position
to interpret its sentential relation. One then is in a position to have the interpreted sen-
tence mentioned above in mind. This in turn places one in a position to have the prop-
osition itself in mind and interpret its propositional relation as ascribing the property
of swimming to Michael, thereby entertaining the proposition. Soames objects that
the first step here, having the sentence “Michael swims” in mind, already presupposes
that one understands the sentence and so entertains the proposition. Hence he claims
my account of entertaining the proposition that Michael swims already assumes at the
first step what it sets out to explain.
As the remarks just made in response to Soames’s first objection hopefully made
clear, the first step of my explanation of one way to entertain the proposition that
Michael swims—acquiring cognitive access to the sentence “Michael swims”—does
not presuppose that one understands the sentence. As I indicated, it just requires one
to have the sentence in mind and so be in a position to interpret the sentential relation.
It does not e.g. require one to actually do so. The idea is that understanding a sen-
tence begins with having it in mind in such a way as to be in a position to interpret its
sentential relation. So having the sentence in mind in this sense just does not require
understanding it. Looking through Chapter 4, I couldn’t find any passages in which
I said or implied that having cognitive access to a sentence required understanding
it, but perhaps it is somehow suggested by something I said. In any case, it isn’t my
view. So Soames is wrong that the first step of my explanation of this way of entertain-
ing the proposition that Michael swims presupposes that one already entertains the
proposition.
19
Chapter 9 p. 171
In making the above objection Soames raises some other issues that are worth clear-
ing up. He cites a passage from Chapter 4 in which I am trying to explain why the prop-
ositional relation of the proposition that Michael swims is interpreted in the way it is,
endowing the proposition with truth conditions.20 Soames seems puzzled by what I say
there, so let me try again. I said there that interpreting the propositional relation of
the proposition that Michael swims as ascribing the property of swimming to Michael
is a reflex of the fact that the sentential relation of the sentence “Michael swims” has
the semantic significance it does. Let me try to explain this idea. As we have seen, one
way of entertaining the proposition that Michael swims is to first have the sentence
“Michael swims” in mind. (Recall that the sentential relation of this sentence has a
certain semantic significance: it ascribes the semantic value of “swims” to the semantic
value of “Michael.”) One then accesses the semantic values of “Michael” and “swims.”
One thereby has the interpreted sentence 1IE above in mind. Abstracting from the fact
that “Michael swims” is an English sentence, containing certain lexical items, that was
uttered in a certain context and so on, one thereby has the fact that is the proposi-
tion that Michael swims in mind. Only now are semantic values composed, and so the
semantic significance of the sentential relation of “Michael swims” is now cashed in.
Thereby, the property of swimming is ascribed to Michael. However, since it is the fact
that is the proposition that Michael swims that we have in mind when semantic val-
ues are composed, we count as interpreting its propositional relation as ascribing the
property of swimming to Michael. So interpreting the sentential relation in the way we
do, by composing semantic values in a certain way at a certain point in the process just
described, just is interpreting the propositional relation in the relevant way.
Soames’s third objection concerns the contents of natural language sentences and
the contents of perceptual experiences. Soames thinks that, on my view, the class of
propositions that are contents of perceptual experiences and the class of propositions
that are the contents of natural language sentences will likely be disjoint. As I indicated
in responding to Speaks, this certainly could be a consequence of my view. Soames
thinks there are two potential problems with this consequence. First, a view with this
consequence may not be able to explain the fact that we appear to use sentential clauses
in characterizing the contents of perceptual experiences.21 Second, a view with this
consequence may have problems explaining how we integrate propositional informa-
tion we acquire through language, perception and thought.
As for the first worry, Soames seems to be thinking that on a view like his where
the contents of perceptual experiences are also the contents of natural language sen-
tences, it is easy to explain why we use sentential clauses to characterize the contents
20
Chapter 9 p. 173
21
Soames puts the point by saying that views with the consequence in question might have problems giv-
ing an account of “how sentential clauses are used to report the contents of perceptual . . . experiences.” But
putting it this way seems to beg the question against views on which sentential clauses are not used to liter-
ally report the contents of perceptual experiences (e.g. because contents of sentential clauses are never the
contents of perceptual experiences).
22
See Chapter 4 pp. 66–68
occur in so-called factive contexts and cannot so occur in non-factive contexts, the oppo-
site is the case for expressions of the form “the proposition that. . . .”23 The second is that
expressions of the form “the fact that. . .” can happily occur in subject position in “causal
statements” (“The fact that the brick hit the window caused it to break”), whereas expres-
sions of the form “the proposition that. . .” cannot. The third is that quantifying across a
factive and non-factive context is usually terrible, (e.g. *’Everything Glenn regrets John
believes.’); whereas quantifying across two factive or two non-factive contexts is virtually
always fine (“Everything John says Glenn believes.”/“Everything John regrets Glenn dis-
covers.”).24 Call the first two sorts of evidence the substitution data; and call the third the
quantificational data. The most straightforward explanation of the substitution and quan-
tificational data is that e.g. “the fact that the brick hit the window” and “the proposition
that the brick hit the window” designate different sorts of entities; that factive contexts
predicate properties appropriate for what the former designate and not the latter, whereas
the opposite is true of non-factive contexts; and that causal statements express relations
between what the former designate and something else and not what the latter designate
and something else.
Now curiously, Soames didn’t comment on the quantificational data. But he agrees
that the substitution data provides prima facie reason for thinking the relevant expres-
sions designate different things. However, he worries that what he calls “substitu-
tion tests” can “cut in different directions,” as when you take the true sentence “Fred
believes that Michael swims” and substitute for the “that” clause a definite descrip-
tion (or some other expression) designating the fact that I claim is the proposition that
Michael swims. The result appears absurd or false. But then, if I think the substitution
data shows that expressions of the form “the fact that. . .” and “the proposition that. . .”
designate different things, shouldn’t I take this to show that the “that” clause and the
definite description designate different things, so that the “that” clause does not des-
ignate the fact I claim is the proposition that Michael swims? Call this the substitu-
tion problem. Of course, as Soames notes, he, Speaks, and I all have the same problem
here. When you substitute a certain definite description designating the thing any of us
claims is the proposition that Michael swims for “that Michael swims” in the sentence
above, the result is seemingly absurd or false. The substitution problem leaves Soames
uncertain about whether the substitution data supports the conclusion I claim it does.
If you think, as I do, that the substitution data is strong support for the conclusion
that expressions of the form “the fact that. . .” and “the proposition that. . .” designate
different kinds of things, what is wanted is to find some difference between that data
23
Since bare “that” clauses can occur in both, this provides evidence that they can designate either the
things designated by expressions of the form “the fact that. . .” or the things designated by expressions of the
form “the proposition that. . . .”
24
I say “virtually always fine,” because you can always create some sort of pragmatic anomaly, as with
“Everything John believes Glenn says.” I take it this sounds a bit odd out of context because of the strange-
ness of one person saying everything another person believes. But even here you can generally rig the con-
text to make it okay. Imagine the sentence preceded by “I know why John likes Glenn so much.”
and the data cited in the substitution problem that suggests different things are going
on in the two cases. I think there is such a difference. In the case of the substitution
data, we have expressions of ordinary English that someone claims designate the same
thing such that when we substitute them for each other in the contexts I described
we get the various anomalies discussed in Chapter 4 and above. The explanation
I have given of this data, which again involves expressions of ordinary English, is surely
the most natural and straightforward. To repeat, the reason expressions of the form
“the fact that. . .” can occur in factive contexts and not in non-factive contexts and
the reverse is true of expressions of the form “the proposition that. . .” is that expres-
sions of these two forms designate entities of different sorts; and factive contexts
predicate properties of the one sort of thing, whereas non-factive contexts predi-
cate properties of the other sort of thing. Finally, the properties felicitously predi-
cated of the one sort of thing cannot be felicitously predicated of the other sort of
thing. This will be discussed further below, but this explanation is supported by the
quantificational data as well; if factive contexts predicate properties of one kind of
entity and non-factive contexts predicate properties of another kind of entity, and
the properties appropriately predicated of the one kind of entity cannot be appro-
priately predicated of the other kind of entity, then we wouldn’t expect to be able to
quantify into both a factive and a non-factive context with a single quantifier. The
examples given above and others show that this is indeed what we find in ever so
many cases. However, again, Soames did not comment on the quantificational data
in responding to me. Finally, the causal statement data suggests the same thing. The
reason expressions of the form “the fact. . .” can occur in subject position in causal
statements and expressions of the form “the proposition that. . .” cannot is that these
expressions designate different sorts of things and only things designated by the for-
mer can cause things.
The data in the substitution problem, on the other hand, does not involve two
expressions of ordinary English that someone claims designate the same thing. It
involves one expression of ordinary English (a “that” clause) and another expression
in “philosophiclese” that someone claims designates the same thing as the ordinary
language expression and that purports to provide something like an analysis or philo-
sophical characterization of the thing designated by the ordinary language expres-
sion. When we take a true sentence containing the expression of ordinary English and
substitute the expression of philosophiclese, we get something seemingly absurd. The
reason that this seems like a very different phenomenon from the substitution data
is that the present phenomenon is ubiquitous in attempts to provide philosophical
analyses or characterizations of things. But then you would expect the explanation
of the data here to involve the nature of philosophical analysis or characterization.
It would seem crazy to think the explanation of the substitution data would involve
such notions.
To see that the phenomenon is ubiquitous in attempts to give philosophical analyses,
consider the following two examples:
1. Some have been attracted to the view that people are material-filled regions of
space-time. But consider the following two sentences:
a. I had lunch with Annie
b. I had lunch with material-filled region x of space-time.
Since a is true and b is absurd, “Annie” and “material-filled region x of space-time”
designate different things. So Annie is not material-filled region x of space-time.
2. Some have thought that events are properties of time intervals. But consider the
following two sentences:
a. The event of Caesar’s dying took place in Rome in 44 BC.
b. The property of being a time interval during which Caesar died took place in
Rome in 44 BC.
Since a is true and b is absurd, “the event of Caesar’s death” and “the property
of being a time interval during which Caesar died” do not designate the same
thing. So the event of Caesar’s death is not the property of being a time interval
during which Caesar died.
It is often the case that when one can give arguments like this, one can give the argu-
ment against virtually any proposed analysis. We noted above that the substitution
problem afflicts the theories put forward by Speaks, Soames and me. It would also
afflict any other theory of propositions. But then if such arguments worked across the
board, they would show the impossibility of any philosophical analysis of the relevant
notion. That seems absurd. So a resolution of the present issues, including the substitu-
tion problem, would require reconciling the possibility of a correct philosophical anal-
ysis with data like 1 and 2 above (I reject the analysis in 2 at any rate, but not because
of the above arguments). To repeat, to think that the substitution data has anything to
do with philosophical analysis seems to me on its face to be ludicrous. This strongly
suggests that the substitution problem and the substitution data involve quite different
mechanisms. In turn, this means that the substitution problem should not lead us to
question the natural, straightforward explanations offered above of the substitution
data, contra Soames.25
The other critical remarks Soames makes concerning expressions of the form “the
proposition that. . .” and “the fact that. . .” concerns the evidence he gave in Soames
[2012] for the claim that they designate the same sort of entity, and sometimes even
25
On the one hand, I am expressing skepticism here that the substitution problem really is a problem for
any of my, Speaks’s and Soames’s theories of propositions, and more generally that arguments like 1 and 2
above should be taken seriously. On the other hand, in Chapter 7 I gave arguments against Soames’s view
that event tokens and types have truth conditions, where these arguments might seem to resemble 1 and
2 above. But they really don’t. The important feature of the substitution problem and arguments 1 and 2
above is that they involve substituting an expression of “philosophiclese” for an ordinary language expres-
sion that is claimed to designate the same thing. By contrast, my arguments against Soames in Chapter 7
involved only expressions of ordinary language (“the proposition that o is red”; “what Godel proved”;
“what just occurred”). Hence they are importantly unlike the substitution problem and arguments 1 and
2 above.
the same entity, which I discussed in Chapter 4. Soames [2012] offered the following
sentence
9. Pam regrets that she is pregnant. Although her parents don’t realize it yet, in time
they will come to believe it.
Since the final “it” seems to have “that she is pregnant” as its antecedent, it presumably
designates the same thing as the latter. But this seems to show that the thing regretted
and the thing believed are the same, contrary to what I have claimed. In Chapter 4,
I described two strategies for explaining the felicity of 9 in a way that is consistent with
the view that things regretted are different kinds of things from things believed. In
Soames’s critical remarks in Chapter 9, he criticizes those strategies. Though I continue
to think both strategies are plausible despite Soames’s criticisms, I won’t re-litigate that
case except for one point. Soames criticizes my first strategy by saying that it isn’t gen-
eral enough to handle the following example in which there is quantification into a
factive and non-factive context:
10.b. Something Pam regrets is now merely suspected by her parents but will soon
be believed by them.
This appears to show that one and the same thing is regretted and believed. But Soames
apparently overlooked the fact that in Chapter 4 I explicitly discussed how the first
strategy could be extended to handle examples like 10b.26 Since this is the only objec-
tion Soames makes to my first strategy, this leaves him with no objection to it.
In any case, rather than trying to provide more detail as to how to explain 9 and 10b
in a way that is consistent with the view that factive and non-factive contexts predicate
properties of different kinds of things, let me focus on why we should explain 9 and
10b in this way. That is, I want to explain why we should not take 9 and 10 to show that
factive and non-factive contexts sometimes predicate properties of the same thing, but
instead consider it to be anomalous data that should be explained by maintaining the
view that factive and non-factive contexts predicate properties of different kinds of
things. Though I discussed this in Chapter 4, I wish to be a bit more explicit here.
The first point to make, which I made in Chapter 4, is that the substitution data and
quantificational data provide lots of quite good evidence that expressions of the form
“the fact that. . .” and “the proposition that. . .” designate different kinds of things and
that factive contexts predicate properties appropriate to only what the former desig-
nate, while non-factive contexts predicate properties appropriate to only what the lat-
ter designate. Hence, this gives us good reason to think that examples like 9 and 10b
should be explained in a way that is consistent with these claims.
But the second point to make is that even if we confine our attention to data of the
sort Soames’s 9 and 10b represent, careful consideration of that data actually supports
26
See my discussion in Chapter 4 of example 10 on p. 70.
the claims that expressions of the form “the fact that. . .” and “the proposition that. . .”
designate different kinds of things and that factive contexts predicate properties
appropriate to only what the former designate, whereas non-factive contexts predi-
cate properties appropriate to only what the latter designate. To see this, let me first
note that we really should be considering data simpler than 9 and 10b. In the case of 9,
we want to see whether a pronoun in a non-factive context can felicitously have as its
antecedent a “that” clause in a factive context. Inserting a clause containing negation
and an anaphoric pronoun in a factive context (“her parents don’t realize it yet”) whose
antecedent is the “that” clause in the first factive context in 9 (and a similar clause in
10b) between the two clauses we are interested in can only produce noise. After all, the
pronoun in “her parents don’t yet realize it” should be felicitous (the pronoun and its
antecedent are both in factive contexts), and this may improve the felicity of the fi nal
pronoun (I’ll add that a fair number of informants I checked with, already found 9 “not
so good”). So to avoid noise produced by unnecessary and irrelevant complexity, let’s
try the most minimal example relevant: namely an example with just one factive con-
text, one non-factive context and no negation or anything else. Consider the following,
focusing on the reading where “it” is anaphoric on “that she is pregnant” (and not on
“Pam regrets that she is pregnant”—see note 27):
9’. *Pam regrets that she is pregnant. Her parents believe it.
I find this quite awkward. When we try other such examples, with a very few excep-
tions (discussed below), they all sound bad to varying degrees:
9’a. *Pam resents that she is pregnant. Her parents doubt it.
9’b. *Pam hates that she is pregnant. Her parents think it.
9’c. *Pam comprehends that she is pregnant. Her parents assume it.
9’d. *Pam forgot that she is pregnant. Her parents denied it.27
9’e. *Pam made clear that she is pregnant. Her parents said it.
On the other hand, it is easy to produce examples that are immaculate with two fac-
tives or two non-factives:
9’f. Pam likes that she is pregnant. Her parents hate it.28
9’g. Pam remembered that she is pregnant. Her parents forgot it.
9’h. Pam resents that she is pregnant. Her parents like it.
9’i. Pam is aware that she is pregnant. Her parents resent it.
9’j. Pam hopes that she is pregnant. Her parents doubt it.
27
This, like some (or maybe all?) of the other asterisked sentences, is okay when “it” takes as its antecedent
the entire first conjunct (“Pam forgot that she is pregnant.”). But this reading is predicted to be fine and so
irrelevant. Recall that we are focused on the reading where the antecedent of the pronoun is the “that” clause
(“that she is pregnant”). Here I have gone to past tense because the present tense (“Pam forgets she is preg-
nant.”) has a habitual-like reading and we are trying to avoid noise. I do the same thing in at least one other
case for similar reasons.
28
Thanks to Annie King for the excellent example.
29
I’ve already said that many informants find Soames’s original example at least a bit awkward; and the
simpler, less noisy 9’ is quite awkward. This suggests that to the extent Soames’s 9 is better, this needs to be
explained. I suggested above that perhaps having an intervening clause with a factive context containing a
pronoun anaphoric on the “that” clause, which should be felicitous, helps explain why the final pronoun in 9
is not too bad.
30
Even here, I find a bit of awkwardness as compared to “Pam knows that she is pregnant. Her parents
believe that she is.”
31
Again, the examples that seem most felicitous involve “knows” and a non-factive. The best seem to
involve “believes” as in “Everything John knows Frank believes.”
32
J&J [2011] p. 220
Let “Harry” name the proposition that Hannah is a philosopher. Harry contains Hannah as a
constituent. Suppose that John tells Bill that he is happy that Harry is true. John knows that
“Harry” refers to a proposition, but is not himself acquainted with Hannah. If Bill trusts John, it
seems that he can come to know by testimony that John is happy that Harry is true, and a fortiori
grasps that proposition. But Harry contains Hannah as a constituent, and by transitivity, so does
the proposition that John is happy that Harry is true. So Bill’s thought that John is happy that
Harry is true is not a singular thought about Hannah, though it has as its content a Russellian
singular proposition containing Hannah as a constituent. Bill can therefore grasp a proposition
containing an object as a constituent, without having acquaintance with that constituent.33
This argument fails given my account of propositions and the resulting account of
propositional constituency. Seeing how and why will illuminate these accounts of
mine. For simplicity, let’s make one change in J&J’s example and let “Harry” name the
proposition that Hannah swims (instead of the proposition that Hannah is a philoso-
pher—things are simpler if we consider a proposition with only two constituents).
Consider the sentence “Hannah swims” represented in tree form:
11.
Hannah swims
Call the syntactic relation between subject and predicate here R. As we have seen,
on my view R will be a component of the propositional relation of the proposition
expressed by 11, which can be represented as follows:
11P.
33
J&J [2011] p. 220
where Hannah is at the left terminal node of the propositional relation and the prop-
erty of swimming is at the right terminal node. As should be familiar, the syntactic
relation R provides all of the significant structure to the proposition 11P. The vertical
lines here represent the semantic relation being the semantic value of ___ relative to
f and c.
Now it is pretty clear that given our theory of propositions and propositional rela-
tions, we don’t have to take constituency as primitive as J&J claim Russellians must
do. We can characterize rigorously what it is to be a constituent of a proposition. Let’s
begin with what I call the simple constituents of a proposition: all entities at terminal
nodes of the propositional relation of the proposition P are simple constituents of
P. Thus, 11P’s simple constituents are Hannah and the property of swimming, just
as one would want and expect. For more complex propositions, it is probably use-
ful to have a notion of complex constituents of propositions. In order to characterize
these, we need a bit of terminology. Propositional relations, like the syntactic rela-
tions that give them their structure, have nodes. A node (immediately) dominates
its daughter nodes. A node dominates the daughters of every node it dominates.
We can now characterize complex constituents of a proposition as follows: for every
non-terminal node n in the tree of the propositional relation of the proposition P, the
subtree rooted in n that includes all simple constituents at terminal nodes of P domi-
nated by n is a complex constituent of P. Given a proposition P, whose propositional
relation looks like this:
1 2
the sub-trees rooted in nodes 1 and 2 are complex constituents of P. There would be four
simple constituents of P occupying the terminal nodes of the propositional relation.
Let’s return to J&J’s example. “Harry” is the name of the proposition that Hannah
swims. On the present account, Harry has Hannah as a simple constituent. Where PH,s
is the proposition that Hannah swims, the proposition that John is happy that Harry is
true, expressed by the sentence “John is happy that Harry is true.”, looks roughly as fol-
lows (some detail suppressed, nodes numbered and italicized expressions represent-
ing the semantic values of those expression (relative to parameters)):
12P. 1
2
3
4
John happy that
PH,s
True
As you would expect, given that the name “Harry” has the proposition that Hannah
swims as its semantic value, the sentence “John is happy that Harry is true” expresses a
proposition that has the proposition that Hannah swims as a simple constituent, along
with John, happy, that, and true. But Hannah is not a simple constituent of 12P, since she
does not occupy a terminal node of the propositional relation of 12P. Nor of course is
she a complex constituent: those are the sub-trees of 12P rooted in nodes 2, 3 and 4 (and
1, if we don’t rule out a proposition being a complex constituent of itself). So if, in J&J’s
example, Bill knows that John is happy that Harry is true without being acquainted
with Hannah, it just doesn’t follow that Bill grasps a proposition (12P) that has an
object as a constituent (Hannah) that Bill is not acquainted with. Thus, J&J’s argument
that one can grasp singular propositions with respect to o without being acquainted
with o fails on the present view of propositions and constituency.
On the present view, in cases like 12P, a proposition, PH,s, is a (simple) constituent of
another proposition, 12P, where the former’s (simple) constituents aren’t (simple) con-
stituents of the latter. (So constituency is not transitive on the present view. I’ll return
to this below.) As a result, the following sentences express different propositions, since
the proposition expressed by 13b has a (simple) constituent (Hannah) not had by the
proposition expressed by 13a:
34
Of course J&J attempted to argue against this view, but as we saw, their argument fails on my view of
propositions and constituency. See also Hawthorne and Manley [2012] for a dissenting view.
On the present view, 14a and 14b express different propositions in just the way that 13a
and 13b do (the proposition expressed by 14b has (simple) constituents not had by the
proposition expressed by 14a: e.g. the reducing relation). Suppose Glenn knows that
logicism is a doctrine to the effect that there is some intimate relationship, he isn’t sure
what, between arithmetic and logic. Suppose he would have no idea what it means to
say that arithmetic reduces to logic. Finally suppose he has heard that Russell was a
defender of logicism. Then it seems to me quite plausible that Glenn knows the propo-
sition expressed by 14a but not that expressed by 14b. If we were to say this (and hold
on to the view that to grasp a proposition, one must be acquainted with its constitu-
ents), we would have to say that in knowing the proposition expressed by 14a, Glenn is
acquainted with the proposition that arithmetic reduces to logic but not with at least
some of its constituents (e.g. the reduction relation). Again, this strikes me as a quite
plausible thing to say. Finally, note that for these reasons, we should say that 13a and
13b can diverge in truth value. If John can be acquainted with Harry without being
acquainted with some of its constituents (e.g. Hannah), then John could grasp the
proposition that Harry is true without grasping the proposition that the proposition
that Hannah swims is true. But then in such a situation 13a would be true and 13b would
be false (assuming that to be happy that P requires grasping P, as seems undeniable).
On the present view, that one can grasp the proposition expressed by 14a and not
grasp the proposition expressed by 14b is an instance of the more general phenom-
enon whereby in order to grasp a proposition one is required to have a robust cogni-
tive connection to its constituents, but not to things that are in some sense parts of
those constituents.35 To take the simplest case, to grasp 13a I must be acquainted with
John but I need have no acquaintance whatsoever with the metal screws in his ankle.
A more interesting case, given certain assumptions, concerns properties and relations.
Suppose that some properties and relations are complex and have other properties as
components. Suppose, for example, that the property of being an instance of knowl-
edge is the property of being a belief that is true and justified. On this way of thinking
the property of being justified is a component of the property of being an instance of
knowledge. Of course, the property of being justified may itself be complex and have
components. Now suppose that e.g. Michael Smith [1997] is right in claiming that the
relation x has a reason to Ψ is the complex relation of x being such that in nearby possible
worlds in which she has a maximally informed, consistent and unified set of desires, she
desires to Ψ. This complex property has as components the (complex) property of being
a maximally informed, consistent and unified set of desire, that of being a possible world
and so on. Now the proposition expressed by:
15. Glenn has a reason to ski.
35
See King [2002b] and King [2007] Chapter 7 for discussion.
has as a constituent the relation x has a reason to Ψ, but it does not have the properties
of being a maximally informed, consistent and unified set of desires or of being a possible
world as constituents, even though the latter are components/parts of the former. Thus
it is that John can grasp the proposition expressed 15 while having no notion of a maxi-
mally informed, consistent and unified set of desires nor of a possible world. He simply
has a much more robust epistemic connection to the complex property x has a reason
to Ψ than he does to some of its components.
Further, the account of constituency being discussed resolves at least some of the
issues involved in the paradox of analysis, as is argued in King [2007]. One issue is how,
if knowledge is justified true belief, the following can express different propositions:
16a. Knowledge is knowledge.
16b. Knowledge is justified true belief.
On the present account it is easy to see how this is so. The proposition expressed by 16a
does not contain the property of being justified as a constituent (it is rather a compo-
nent of a constituent of that proposition, but it is itself neither a simple nor complex
constituent of the proposition); that expressed by 16b does. They thereby are different
propositions and have different requirements on being grasped. Of course, this does
not resolve all the issues surrounding the paradox of analysis, but it is a good start. See
King [2007] Chapter 7 for a fuller discussion of those additional issues.
One final point. I mentioned that the on the present account, propositional constitu-
ency isn’t transitive (see the discussion of 12P above). But what about the motivation
J&J offered for it being transitive? Consider the proposition that Rebecca swims. On
the present view, Rebecca is a simple constituent of it. But it is easy to see that on the
present view, Rebecca is also a simple constituent of both the proposition that it is not
the case that Rebecca swims and the proposition that Rebecca swims and Shane skis.
In both cases she will be at a terminal node of the relevant propositional relation. In
general, if P is any proposition and Q is a proposition built up of P, other propositions
and truth functions, all simple constituents of P will be simple constituents of Q on the
present view of constituency. Further, P will be a complex constituent of Q. Now we
can see the difference between this case and 12P above. Hannah is a simple constituent
of a simple constituent of 12P: the proposition PH,s. Thereby Hannah can’t be a simple
constituent of 12P: she can’t occur at a terminal node of 12P’s propositional relation,
since she is a proper part of something that occurs at one. Nor can she be a complex
constituent. However, in the present case not only is Rebecca a simple constituent of
the proposition that Rebecca swims, P, but in virtue of P being a complex constituent of
the proposition that Rebecca swims and Shane skis, Q, Rebecca is a simple constituent
of Q as well. So we both eat our cake with P and Q and have it with 12P too! Thereby, we
deny the transitivity of propositional constituency, while capturing the motivation J&J
gave for it.
In summary, on the present account of propositions we can rigorously characterize
the notion of (simple and complex) constituents of a proposition and so do not have to
take it as primitive as J&J claim Russellians must. The resulting account distinguishes
between propositions that intuitively should be distinguished (those expressed by
13a and 13b; 14a and 14b; and 16a and 16b). Further, when coupled with the plausible
view that grasping a proposition requires being acquainted with its constituents, our
account makes plausible claims about the different conditions under which the propo-
sitions distinguished can be grasped (with the result that one can grasp 13a and 14a
without grasping 13b and 14b). In the case of the difference between the propositions
expressed by 16a and 16b (assuming knowledge is justified true belief) and the differ-
ent requirements on grasping them (assuming that to grasp a proposition one must be
acquainted with its constituents), the account of constituency contributes to a resolu-
tion of the paradox of analysis. That this account of propositional constituency, which
drops out of the present account of propositions, has these various desirable features
can only provide further support for the theory of propositions that yields it.
My aim in this essay will be to critically examine two aspects of current orthodoxy
about propositions: that they are representational and that they are structured. Along
the way I’ll also pause to discuss a few of the objections raised by Soames and King to
the view of propositions I defend.
it’s not obvious why we shouldn’t just say this about all mental states—that is, it’s hard
to see why we wouldn’t treat belief, desire, etc. as intrinsically representational atti-
tudes to propositions which are not, in themselves, representational.
But let’s set this polemical point to the side, and ask: would the representational
properties of beliefs be objectionably mysterious if not explained in terms of the repre-
sentational properties of propositions?
First, let’s get clear on what we’re saying has representational properties when we
say that, for example, my belief that South Bend is lovely is about South Bend. There
are, I think, two sorts of things that we can be saying when we make claims like this.
On the one hand, we might be saying that I stand in a certain “aboutness” relation to
South Bend—which we might call the “believes-about” relation. On the other hand,
we might be saying that there’s a certain entity—perhaps a mental sentence or some-
thing of the sort—which has the dual properties of being a belief of mine and of being
about South Bend. I’ll consider these interpretations in turn.
Consider first the fact that I stand in the “believes-about” relation to South Bend.
To stand in the believes-about relation to South Bend is, on my view, to stand in the
belief relation to a proposition one of whose constituents is South Bend.1 Hence the
fact that I stand in the believes-about relation to South Bend is not to be explained
solely in terms of the properties of the proposition that South Bend is lovely—part of
the explanation is my standing in the belief relation to this proposition. On my view
these relations, rather than the entities to which the relations are borne, are the source
of the representational properties of subjects associated with the attitudes. Hence, to
explain these representational properties of subjects, we don’t need to appeal to the
representational properties of propositions, but only to representational ways of being
related to those propositions.
This does not mean, of course, that the proposition has no role to play in explain-
ing the representational properties of a subject: the representational properties that
I instantiate when I believe that grass is green are different than the representational
properties that I instantiate when I believe that snow is white in virtue of differences
between the propositions that grass is green and that snow is white. But this does not
mean that these propositions are themselves representational, or about anything.
An analogy might help. Consider the view that all theft is morally wrong. On this
view, the wrongness of an act can be explained simply by its being an act of theft. But
this is consistent with the moral properties of the act also being partly explained by
the object of the theft—one might think that stealing a car is morally worse than steal-
ing a piece of bubble gum. But the fact that the moral properties of an act of theft is
partly a function of the moral properties of the object of the theft does not imply that
1
Though see below for some second thoughts about “constituent” talk. Others, of course, will specify the
relevant property of propositions differently—in terms of, for example, singular Fregean modes of presenta-
tion of South Bend. I’m ignoring such differences for simplicity here—they are independent of the question
at issue.
“I can consider, explain and understand the claim that arithmetic reduces to logic and in none
of these cases does it seem that my attitude has anything to do with whether the alleged property
being such that arithmetic reduces to logic is instantiated. Take considering: when I consider the
claim that arithmetic reduces to logic I can do so without any attitude at all about the instantia-
tion of the property being such that arithmetic reduces to logic. I may simply have an interest in
really thinking about what the claim comes to. So here Speaks’s account of the attitudes seems
strained at the very least.”
instantiated—but surely, as King says, one can consider a proposition without taking
any attitude at all toward the question of whether the relevant property is instantiated.
But here I am inclined to think that King exaggerates the distance between belief
and consideration. Of course, considering a proposition does not involve believing that
the relevant property is instantiated; but that is just because belief and considering
are different propositional attitudes, and hence involve different attitudes toward the
instantiation of the relevant property. It seems to me plausible that when we consider
a proposition, we are considering what would be the case were that proposition true;
and it is natural, on my sort of view, to take this to be an attitude toward the instantia-
tion of the relevant property.2
One might think, though, that there is a different sort of attitude toward a proposi-
tion, which we might also express using “consider,” which is different than consider-
ing whether a certain proposition is true. This is the attitude we might have toward a
proposition when we’re simply examining its properties—for example, we might con-
sider its structure, or consider whether this or that is a constituent of the proposition.
We might, as King puts it, simply be interested in “really thinking about what the claim
comes to.” But I think that examining, in this sense, the proposition that arithmetic
reduces to logic is best understood as a propositional attitude to some proposition
other than the proposition that arithmetic reduces to logic. It might, for example, be a
matter of judging that (or hypothesizing that, or wondering whether) the proposition
that arithmetic reduces to logic has a certain structure. And these propositional atti-
tudes can, without strain, be understood as involving the instantiation of the relevant
property (which will, in this case, be the property of being such that the proposition
that arithmetic reduces to logic has the relevant structure).
The case is analogous to one in which I consider the properties of some object, like
my desk. This need not involve me being in some mental state whose content is my
desk; rather, the contents of such mental states are propositions which attribute one or
another property to my desk.
Let’s consider the other interpretation of argument (i), according to which “my
belief ” refers to something like a mental sentence which qualifies as one of my beliefs
and has the representational property of being about South Bend.3 Do we need to
explain what it is for such a mental sentence to be about South Bend in terms of the
representational properties of the proposition which that mental sentence has as its
content?
2
It’s worth adding that the “believes-to-be-instantiated” bit of Ch. 5 is a dispensable add-on to the the-
ory of propositions sketched there; it is an attempt to explain how, if propositions are properties, we might
understand propositional attitude relations to propositions. But one might well give a theory of propositions
without also giving a theory of the attitudes. That, I think, is what King does; he gives an account of proposi-
tions, and of how we can have cognitive access to them, but never explains exactly what it is to believe a fact
of the relevant sort.
3
Nothing much is built into “mental sentence” here—the following remarks would apply just as well to
theories which want to avoid commitment to anything like a language of thought, and so make use of syntac-
tically unstructured “belief states” instead.
On my view, for one of my mental sentences to be about South Bend is for that sentence
to have as its content a proposition one of whose constituents is South Bend. Hence—
as with the representational properties of subjects—the representational properties of
mental sentences are not explained solely by a proposition, but rather by those sentences
standing in a certain relation—which we express by phrases like “has the content that”—
to that proposition. But then, as in the case of subject-level representational properties, it
is open to us to trace the source of the representational properties of mental sentences not
to the proposition to which it is so related, but to the relation itself.
The above remarks about mental sentences generalize to give us a response to argu-
ment (ii) above: the argument that we need to appeal to the representational properties
of propositions to explain the representational properties of sentences of public lan-
guages like English. Such sentences also have their representational properties in vir-
tue of standing in certain relations—like that expressed by “_ is the semantic content
of _ in C”—to propositions. Hence, as above, the representational properties of sen-
tences can be explained not in terms of the representational properties of the proposi-
tions they semantically express, but by the representational properties of the relation
of semantically expressing.
This style of explanation can, it seems to me, be further generalized to bearers of
contents other than sentences. In chapter 9, Soames objects to my view of propositions
that it
“makes it difficult to capture the fact that truth is a kind of accuracy in representation. A map or
portrait is accurate, or veridical, when it represents its subject matter as being how it really is; a
proposition is true when it represents things as they really are. This parallel seems to be lost when
propositions are identified with properties that, as Speaks admits, aren’t intrinsically represen-
tational. Unless he can identify some sense in which they are representational, he will lose the
pre-theoretic connection between truth and accuracy.”
On my view, maps and portraits, like sentences, have contents, and I’m inclined to
think that, despite the fundamental differences in the way these contents are encoded,
maps and portraits have the same kinds of contents as sentences—namely, the sorts
of properties discussed in chapter 5. But, given this, the parallel between maps and
sentences seems pretty direct; each are representational in virtue of standing in certain
relations to propositions which are not themselves representational. For a map to be a
map of South Bend, for example, just is for that map to have as its content some propo-
sition one of whose constituents is South Bend. And just as a sentence is true iff the
property which is its content is instantiated, a map is accurate iff the property which is
its content is instantiated.4
4
There remains the awkward question of why, if maps are this closely related to sentences, it sounds a bit
odd to say that a map is “true.” Parallel worries arise, as is well-known, about the construal of perceptual
experiences as a propositional attitude. I’m inclined to think that this oddness does not reflect anything
deep about the subject matter—and in any case Soames’s worry here is not that my view makes maps and
sentences too similar, but that it does not make them similar enough.
5
See, respectively, Grice (1968), Grice (1969), Lewis (1975), and the essays in Fodor (1990).
One might worry, though, that saying that propositions can have truth condi-
tions without being about anything will force us to deny some platitudes about
truth. For example, it seems plausible that something is true just in case it rep-
resents the world as being some way, and the world is that way—but that can’t be
right if, as I think, propositions are true or false but don’t represent the world as
being any way at all.
I think that this is the strongest argument in favor of ascribing representational
properties to propositions. But I don’t think that it is decisive. The right thing to say, it
seems to me, is that these seeming platitudes are really platitudes—but only when we’re
talking about the notion of truth applicable to sentences or beliefs. On my view, it is a
platitude that a sentence is true iff it represents the world as being some way, and the
world is that way—what it is for a sentence to represent the world as being some way is
for that sentence to have a certain property—a way things could be—as its content, and
what it is for the world to be that way is for the property to be instantiated. But no such
claim holds about the truth of propositions.
Like many, I think that propositions are the fundamental bearers of truth and falsity,
and that the truth and falsity of sentences, mental states, maps, etc. is to be analyzed
in terms of the truth and falsity of propositions to which these entities stand in the
relevant relations.6 On a view like this, it will be no surprise if the right account of
truth and falsity for propositions is different from the right account of truth and falsity
for the non-fundamental bearers of truth and falsity. Hence, I think, it should not be
surprising if platitudes which are correct about the truth and falsity of sentences don’t
hold when we’re talking about the truth and falsity of propositions.7
6
For defense, see Soames (1999).
7
It might also be worth noting that it seems not to be a platitude that propositions are about anything. The
following hardly counts as commonsense: “That South Bend is lovely is about South Bend.” By contrast, it is
hard to deny either of “ ‘South Bend is lovely’ is about South Bend” or “The belief that South Bend is lovely is
about South Bend.”
sentence S as the contents of the subsentential expressions in S. Now, this sort of use of
the word “constituent” does not make wholly trivial the claim that propositions have
constituents—given that propositions are identical only if they have the same constitu-
ents, the claim that propositions have constituents in this sense entails that if two sen-
tences S, S* express the same proposition, then S contains an expression with a certain
content iff S* does. And this will be denied by proponents of “coarse-grained” concep-
tions of propositions, like the view that propositions are sets of worlds. For on that
view necessarily equivalent sentences will always express the same proposition—but
two sentences can be necessarily equivalent even if one contains terms with a content
which no expression in the other sentences has. The obvious examples here are math-
ematical truths—but there are plenty of others.8
So there is a substantive debate to be had about whether propositions have constitu-
ents in the lightweight sense. But one might reasonably object that this is not really
a debate about whether propositions really have constituents at all. For surely to say
that propositions have constituents must be to say about propositions something at
least analogous to what we say about tables and chairs when we say that they have
constituents. In this metaphysically heavyweight sense, to say that propositions have
constituents is to say that there are entities—paradigmatically the contents of subsen-
tential expressions figuring in the sentence which expresses the proposition—to which
propositions stand in some relation which either is, or is closely analogous to, the rela-
tion between parts and wholes.
There’s no obvious contradiction in accepting the claim that propositions have con-
stituents in the metaphysically lightweight sense, but denying that they have constitu-
ents in the metaphysically heavyweight sense. One might, after all, adopt the following
view about propositions:
Primitivism. Propositions are a sui generis category of entity, and hence are not identical to sets
of possible worlds. Further, a pair of sentences can be necessarily equivalent despite expressing
distinct propositions. Indeed, propositions are much more fine-grained than sets of possible
worlds; two sentences can express the same proposition only if those sentences are synonymous,
and two sentences can be synonymous only if they are composed of synonymous subsentential
expressions. But propositions, like other abstract objects, are simple, and have no parts; hence
propositions have (in the metaphysically heavyweight sense) no constituents.9
Intuitively, this sort of primitivist agrees with the traditional doctrine of structured
propositions about the individuation of propositions—about when, for example, the
proposition expressed by one sentence is identical to the proposition expressed by
8
One can adopt the well-known recipe for constructing such cases from Soames (1987)—just pick any
sentence expressing a necessary truth, and conjoin it with an arbitrary sentence S. The conjunction and S
will be equivalent, and hence according to the view that propositions are sets of worlds, will express the same
proposition. But in the standard case our necessary truth will contain some expression which is not synony-
mous with any expression in S.
9
This is the view of, among others, Plantinga (1974), and is defended at length in Keller (2012).
another. So the primitivist can agree that propositions have constituents in the light-
weight sense. But they disagree about whether these claims about the individuation of
propositions are best explained by attributing genuine complexity—i.e., the having of
parts—to propositions.
Do the theories of propositions defended in Part II by Soames, King and me entail
that propositions have constituents in the metaphysically heavyweight sense? I don’t
see that they do. We defend, respectively, the views that propositions are event-types,
facts, and properties. But I think that each of our theories leaves open the question of
whether these event-types, facts, and properties have constituents in the metaphysi-
cally heavyweight sense—that is, we leave it open whether event-types, facts, and prop-
erties stand in part/whole relations (or some closely analogous relation) to anything.10
To see this, consider the following thought-experiment. Imagine being presented
with a compelling argument for the conclusion that no abstract objects, and in particu-
lar no properties, facts, or cognitive event types, have parts. Now ask yourself: would
this affect any of the claims made in the chapters in Part II of this book, or help us to
decide which of those theories was most likely to be true? I am not sure that it would.
Of course, we’d now have to understand the use of “constituent” in those chapters in a
metaphysically less-than-serious way, as a useful term for specifying just which prop-
erties, facts, or cognitive event-types propositions are supposed to be, but not as imply-
ing that there’s any genuine complexity in the entities so specified. But, other than that,
it doesn’t seem as though anything would be different.
I think that this openness of our views on this score is in one sense a virtue, and
in another, a vice. It is a strategic virtue because it makes our views consistent with
(respectively) various views of the nature of event-types, facts, and properties. And,
relatedly, it means that we don’t have to confront difficult questions which confront
Millians who are also believers in the claim that propositions have constituents in the
heavyweight sense. For example: on standard views, part/whole relations are transi-
tive, so that if x is a constituent of y and y is a constituent of z, then x is a constituent of
z. But it looks as though Millians will have to deny this principle, since I can be a con-
stituent of a singular proposition about me without every part of me also being a con-
stituent of that proposition. One can, fortunately, entertain singular thoughts about
me without thereby entertaining singular thoughts about every part of me.11
The vice corresponding to this virtue, though, is pretty obvious, and that is just
that the above virtue is obtained only by failing to explain the relationship between
10
Though I think that the theories leave this question open, I’m not saying that Soames and King them-
selves have no views on the matter. For example, King expresses sympathy with the idea that the constituents
of propositions are parts of those propositions in King (2007), p. 120 note 42.
11
See for discussion Gilmore (2011). For a more wide-ranging discussion of the problems with assimilat-
ing talk about the constituents of propositions to standard theories of part/whole relations, see Keller (2012).
It’s also worth noting that some views of propositions as in some sense structured—like the view on which
they are ordered pairs—don’t run into this problem. For criticism of “ordered pair” views, see King (2007).
propositions and the entities which the three of us all refer to as that proposition’s constitu-
ents. If it isn’t parthood, what is it?
Let’s return for a moment to the primitivist view of propositions sketched above. The
primitivist agrees with Soames, King, and I that propositions have constituents in the
metaphysically lightweight sense, but denies that propositions have constituents in the
heavyweight sense. But I’ve just said that the theories which Soames, King, and I give are
consistent with the denial of the claim that propositions have constituents in the heavy-
weight sense. Does that mean that our theories are consistent with this primitivist view?
Not quite. The primitivist says something which each of us denies, namely that prop-
ositions are a sui generis category of entities, and hence rules out the sorts of theories
that each of us try to provide by, respectively, assimilating propositions to cognitive
event-types, facts, and properties. This brings out the fact that we should distinguish
three different questions on which a theory of propositions should take a stand.
The first is a question about whether propositions have constituents in a lightweight
sense, which boils down to questions about the individuation of propositions—e.g.
questions about the conditions under which a pair of sentences express the same prop-
osition, or a pair of subjects believe the same proposition.
The second is a question about whether propositions have constituents in the heavy-
weight sense, which is just the question of whether propositions are simple (lack parts)
or complex (have parts).
The third is a question about what sorts of things propositions are. Are propositions
a basic category of our ontology, or are they a subclass of some other more fundamen-
tal category of thing?
The independence of these three questions deserves some emphasis. Occasionally
the first two are implicitly conflated, as when one infers from the falsity of the possible
worlds view of propositions (and hence the falsity of the claim that propositions lack
constituents in the lightweight sense) that propositions must be structured (and hence
have constituents in the heavyweight sense).12 And occasionally the second two are
conflated, as when the view that propositions are simple (and hence lack constituents
in the heavyweight sense) is taken to imply the claim that propositions are irreducible
to any other category of entity.
The focus of this book has been overwhelmingly on the third of the above questions.
(Though, in each case, our answers to this third question plausibly entail answers to the
first.) We’ve managed, by contrast, mainly to ignore the second. In a way, this emphasis
makes sense. Just as we can ask whether propositions are complex or simple, we can
ask whether properties, facts, and other categories of abstracta are complex or simple;
and, plausibly, whatever considerations lead us to say that propositions are or are not
complex will lead us to the same conclusion about properties and facts.13 Hence one
12
See, for discussion, Keller (2012).
13
Of course, I think that propositions are properties, and King thinks that they are facts. But both of us
think that there are properties and facts which are not propositions, and questions about complexity will
arise for these as well.
might well think that our second question is not really a question about propositions in
particular, but rather a question about abstract objects more generally.
This doesn’t change the fact that the second question is one to which any complete
theory of propositions, especially ones which are expressed using talk about struc-
ture and constituents, owes an answer. But I wonder whether the pervasive use of
structure-talk has made this second question seem more fundamental than it really is.
What hangs on the question of whether propositions are simple or complex, unstruc-
tured or structured?
If anything hangs on this question, it must be that the claim that propositions have
constituents in the metaphysically heavyweight sense explains something. So our ques-
tion reduces to: What might the claim that propositions have constituents in this sense
explain?
A natural, and popular, suggestion is that structure explains representational prop-
erties; we can explain what it is for a proposition to be about Socrates by saying that the
proposition contains Socrates as a constituent.14
But, as Keller (2012) convincingly argues, this sort of explanatory claim is problem-
atic. One might put her point like this: either constituency is parthood, or it isn’t. If
it is, then we can’t explain aboutness in terms of constituency, for many things have
parts without being in any sense “about” those parts. And if it isn’t, then, absent a the-
ory of the constituency relation, we’ve just replaced one primitive—aboutness—with
another. Where’s the explanatory gain there?
This dilemma does not, of course, show that there is no good answer to the ques-
tion of what the claim that propositions have constituents in the metaphysically heavy-
weight sense explains. But I think that it does strongly suggest that the explanatory
value of the attribution of structure to propositions has been a bit under-explained.
While I still find the view that propositions are structured attractive, I’m less and less
sure exactly what is explained by adding to one’s theory of propositions the claim that
they have parts.15
14
For versions of this thought, see, among many other places, King (2007), 6, and Braun (1993), 461. We
can set aside in the present context the fact that I don’t think that propositions have representational proper-
ties to be explained; we might still think that we can explain the representational properties of beliefs and
sentences in part in terms of the constituents of the propositions that they have as their content
15
Thanks to Lorraine Juliano Keller for helping to engender the skeptical worries just voiced. Discussion
of these topics with Lorraine, along with a reading of her excellent dissertation, Whence Structured
Propositions? has prevented me from making at least some of the metaphysically irresponsible claims which
I would otherwise have made in this book.
In this chapter I reply to the probing and provocative critiques of Speaks and King,
which I use (i) to clarify aspects of my position that had previously been underde-
veloped and insufficiently clear, (ii) to introduce improvements needed to properly
understand the sense in which propositions are genuinely representational, and (iii)
to discuss the explanatory burden in terms of which any theory of propositions must
be judged.
In reply to Speaks
The real explanatory burden
Speaks notes that I agree with him that being such that Amelia talks and other proper-
ties of this sort exist, and that we have cognitive access to them. Since both of us must
explain this access, while I must also explain our access to the cognitive event types
I have identified as propositions, he concludes that his explanatory debt is less than
mine. I disagree; the explanation I provide of our access to (degenerate) properties like
being such that Amelia talks is a trivial extension of the explanation already provided
of our access to the propositions from which they are derived. Whereas I explain our
access to these properties, he doesn’t. Although we both presuppose agents’ access to
simple properties, he is silent, whereas I am not, about how complex properties are gen-
erated, how they are individuated, and how we access them.
In my system negating a property being so-and-so, predication of which represents
its target as being so-and-so, generates the property not being so-and-so, predication
of which represents its target as not being so-and-so. Similar stories can be told about
conjoining and disjoining properties, generating n-1 place properties from n place
properties, and forming complex properties in other ways. In each case, agents’ access
to complex properties is explained by operations they perform on more fundamental
properties, while the complex properties themselves are individuated by the ways in
which predication of them represents their targets, from which their contributions to
the truth conditions of propositions can be read off. This idea is extended by operations
generating properties whose representational contents, individuation conditions,
and cognitive accessibility are parasitic on those of already generated propositions.
Operating on the proposition that John loves Mary or Bill hates Mary, agents can gen-
erate the property—λx [John loves x or Bill hates x]—predication of which represents
its target as being one whom John loves or Bill hates. Taking the degenerate case of
this operation in which no constituent is abstracted from the original proposition, we
generate the property being such that John loves Mary or Bill hates Mary predication
of which of any target represents precisely what the proposition represents. When S
expresses p, ⌜is such that S⌝ stands for the property predication of which of represents
what p represents (and nothing further). Since the individuation of, and our cogni-
tive access to, the property are parasitic on the already explained individuation of, and
access to, the proposition, these properties don’t add any further explanatory burden.
It is Speaks who faces the problem of individuating and explaining our cognitive
access to the properties expressed by ⌜is such that S⌝ and ⌜is such that R⌝ for arbitrary
S and R. How can he do so without invoking propositional intermediaries? Well, how
are Amelia and being a talker related to being such that Amelia talks? This Speaksian
property/proposition is not the property that x instantiates iff Amelia instantiates
being a talker. There is no unique property satisfying that condition, since being such
that Amelia talks and 1st-order arithmetic is incomplete does too. Nor is it enough to say
that the proposition is a property satisfying the condition, since Speaks needs different
properties satisfying the condition to be objects of different attitudes. One could build
formal structures—trees, tuples, sets of sets—out of simple properties and objects, and
then stipulate their instantiation conditions. But unless the structure and the conditions
assigned can be shown to be nonarbitrary, this will, at most, model propositions, not
identify them.1
One could, of course, go representational by maintaining (i) that being such that
Amelia talks is the property that represents Amelia as a talker (without representing
anything further) because one who predicates it of anything represents Amelia that
way, (ii) that being such that the earth is round is the property that represents the earth
as round (and nothing further) for a similar reason, (iii) that being such that Amelia
talks or the earth is round is the property that represents Amelia as a talker or the earth
as round (and nothing further) because one who predicates it of anything does, and
so on. But this would take Speaks down my road, which he doesn’t wish to travel—in
part, I suspect, because it leads to the question, “Since predicating being a talker of
Amelia represents her as a talker, shouldn’t we already have the proposition that she
1
The problem of identifying propositions (be they properties or not) with essentially arbitrary structures
intended to individuate them, which are then assigned truth conditions, is discussed on pp. 52–55 of What is
Meaning? (Soames 2010).
talks, before reaching the property being such that she does?” Eschewing this represen-
tational route, he needs his own compositional theory of the structures and instan-
tiation conditions of such-that properties that nonarbitrarily distinguishes them from
one another and explains how agents with limited cognitive resources can access
indefinitely many. Until he provides one, his explanatory burden is more, not less,
daunting than mine.
the sense of adopting that way of thinking—of o as red—as a potential basis for further
thought or action. In the case of assertion, the something else is an act of publicly com-
mitting oneself to o’s being as one represents it to be. In each case, we (i) cognitively
represent o to be red (which we also do when we merely imagine o as being red), and
(ii) take a further stance toward that representation (which we do not take when we
merely imagine o to be red). So, all events of judging or asserting that o is red involve
a distinctive kind of cognizing—“predicating redness of o”—accompanied by other
cognitive doings analytically distinguishable from it. This point generalizes to many
related attitudes including questioning, denying, and so on. Of course, it does not fol-
low that any of these cognitive events involve an initial event of predication succeeded
in time by another event of endorsing, questioning, or what have you. How these cog-
nitive acts are performed—simultaneously or in sequence—is not for philosophy to
decide. Thus the theory generates no expectation that even the most powerful intro-
spector should be able to notice events consisting of an agent’s performance of an act
of predication in splendid isolation (in the absence of the performance of any further
acts) in cases in which an agent judges, asserts, questions or denies something. So the
fact that Speaks doesn’t notice them tells us nothing about the theory.2
The point is strengthened when one considers simpler cognitions—seeing, visu-
alizing, or imagining o as red. Any event consisting of an agent’s doing one of these
things is an instance of the agent’s cognizing o as red—a.k.a. predicating redness of o.
Of course, not all instances of predicating are instances of seeing, not all instances of
predicating are instances of visualizing, etc. How then do these different instances of
predicating differ from one another? Does seeing o as red consist of predicating red-
ness of o plus doing something else (the doing of which is no part of the predicating),
while visualizing o as red consists of that same predicating plus doing a different some-
thing else, and similarly for imagining o as red? I think not. To see o as red is to predi-
cate redness of o (i.e. to cognize o as red) in a certain way, while to visualize o as red is
to predicate redness of o in a different way. But these different ways no more involve the
performance of different acts (the doing of which is no part of the predicating), than
the difference between punching a bag with one’s right hand and punching it with one’s
left hand involves the performance of different additional acts (the doing of which is
no part of either punching). When it comes to events the lesson is clear. Just as there is
no bare event of my punching the bag that is not identical with an event of my punch-
ing it with my right hand or identical with an event of my punching it with my left, so
there is no bare event of my predicating redness of o (i.e. of cognizing o as red) that is
not identical with my seeing o as red, my visualizing o as red, imagining o as red, or my
2
Events of predication occur when an agent does something representational—like judge, assert, and the
like. Since not all propositional attitudes—e.g. believing and assuming—require occurrent cognitive events,
but may sometimes be fully dispositional, agents may sometimes hold propositional attitudes without per-
forming any acts of predication, or experiencing any predicational events constitutive of the propositions to
which the attitudes are born.
cognizing o as red in some other way.3 In short, there are no events of bare predication
(i.e. of cognizing but not of cognizing in any particular way) of the sort Speaks is seek-
ing. To seek them is to misunderstand the theory.
3
To pursue the analogy, just as I can, on reflection, determine that I have punched with my right hand
(which is different from punching with my left), so, I claim, I can determine, on reflection, that I have visually
represented o as red (which is different from conceptually representing o as red). Hence I am aware of repre-
senting in different ways, even though I notice no bare acts of representing, because there are none. Above
and throughout, I use “predicate” to underline that representing is something we do.
of the event type of entertaining p (which, in chapter 6 is identified with p), I take him
to be characterizing its intentionality as intrinsic. I now find this way of putting things
unfortunate. The proper way to proceed is, I think, (i) to identify the act of predicating
being so-and-so of o as representational because for an agent to perform it is for the
agent to represent o as so-and-so, and (ii) to explain whatever intentionality is pos-
sessed by the event type of performing this act and the individual instances of that type
in terms of (i).
Next Speaks asks:
“why not say this [that they are intrinsically representational] for all... [such cognitive acts/
event-types]? Why not... let [not just the attitude of entertaining, but also] each of the famil-
iar attitudes—belief, assertion, etc.—be intrinsically representational states [acts/event types],
whose status as representational is not explained by the representational properties of the prop-
osition to which they are relations?” (13–14)
Consider the attitudes affirming, denying, and occurently doubting that o is red. Each
involves taking a cognitive stance toward the proposition that o is red. Although
the stance differs from case to case, what it is a stance toward is the same. Because
the attitude entertaining abstracts away from any stance, it is perfectly suited to cap-
turing what we focus on when we ask whether what is affirmed, denied, or doubted
is true. This is the fundamental sense in which affirming, denying, and occurently
doubting that o is red all represent o as being the same way. This representational
commonality is captured by taking those attitudes to involve entertaining the propo-
sition that o is red plus a further cognitive ingredient that varies from one attitude to
the next. The presence of this further ingredient is why we shouldn’t treat the richer
attitudes Speaks has in mind as we treat entertaining. There may be a further sense
in which the richer attitudes are representationally different from one another. After
all, an agent’s take on the world will differ markedly depending on whether the agent
affirms, denies, or doubts that o is red. But this further representationality results
from how the cognitive stances involved in these attitudes interact with their propo-
sitional object.
What these cognitive stances amount to is the least developed aspect of my view.
So far, I have said that to judge that o is red is to predicate redness of o while affirm-
ing or endorsing that predication, to believe that o is red is to judge, or be disposed to
judge, that it is, and so on. To this, I here add a further cautionary note. To endorse
or affirm a proposition p that one has entertained is not to predicate a property or
relation of p, or to perform any representation-modifying operation on p. It is to
entertain p in a certain way, which results in that cognitive event playing a certain
committing role in one’s cognitive life. The same is true of other stances, such as
wondering. This must be so, since even cognitively unsophisticated creatures that
are unable to identify and target the types of which the cognitive events of their own
experience are instances can bear the attitudes of judging, believing and wondering
to propositions.
Existence and belief
Speaks’s next objection focuses on the conjunction of my views (i) that an agent can
believe a proposition p without ever entertaining p and (ii) that this can happen even
in cases in which the p doesn’t exist. The objection is that if this were so, then cer-
tain arguments that are clearly valid wouldn’t be. For example, Speaks says, Argument
A would not be valid.
Argument A
Coming up with a better example isn’t easy. On the account I have offered, there
are real but nonexistent propositions, some of the simple constituents of which nei-
ther have been cognized already nor are cognitively accessible by any systematic
means mastered by agents. There are also propositions that agents believe without
having entertained them. But it isn’t easy to show that some propositions are mem-
bers of both classes. Displaying them in an argument is out of the question, since to
do so is to guarantee their existence. There is, of course, no such bar to displaying
an existing proposition that could have been believed without existing, or was in
fact believed before it existed. However, as the discussion of Argument A illustrates,
even this is daunting. Suppose, for the sake of argument, we find such a proposi-
tion, expressed by some sentence S. Switching the premises to the past tense gives
us Argument A*.
A1*. At t, Jeff believed that S.
A2*. At t, Scott believed that S
Next consider the following conclusions.
a* There is something that Jeff and Scott both believed at t.
b* There exists something that Jeff and Scott both believed at t.
c* At t, some proposition was believed by both Jeff and Scott.
d* At t, there was a proposition that both Jeff and Scott believed.
e* At t, there existed some proposition that both Jeff and Scott believed.
Suppose that A1* and A2* are true, and that the proposition expressed by S exists
now, but didn’t exist at t. Then a* and b* will be true, and we won’t have a counter-
example. Whether or not c* is a counterexample depends on whether we can at t
quantify over things not existing at t. Since I have argued we can, I can recognize the
truth of c*. I also believe that we can, and sometimes do, use “there is/are” to range
over domains that include nonexistent things. Thus, d* has a reading in which it is
true. What Speaks needs is e*. If the right sort of sentence S can be produced, A1*
and A2* will be true and e* will be false. Such a result, though mildly surprising,
wouldn’t be a weighty objection.4 What we are engaged in is theory construction,
not ordinary language analysis. Sometimes philosophical theory leads to correct,
but surprising and even mildly counterintuitive results about which pretheoretic
opinion isn’t determinative. So long as wholesale rejections of commonsense con-
victions are avoided, as they are here, philosophical explanation may sometimes
prevail.
4
As Brian Bowman has pointed out to me, one can probably find the right sort of sentence S if one con-
structs an argument in which one concludes that there (now) exists a proposition that Jeff and Scott didn’t
believe at t, from ⌜~ Scott believed S at t⌝ and ⌜~ Jeff believed S at t⌝. However, this doesn’t add much to the
weight of the objection.
5
This explanation supersedes my careless comments (from 2010 on) about propositions inheriting inten-
tionality from their possible instances—comments that, doubtless, contributed to Speaks’s impression of
circularity.
6
I will say more about this extended sense below.
7
The reason that Speaks’s event types (i), (ii), (vi), and (vii) on page 18 are not good candidates for propo-
sitions, and are not naturally assigned representational content that makes them bearers of truth conditions,
is that they are not connected closely enough to the cognitive lives of agents to serve the function that is the
raison d’être of this extended sense of representing.
In reply to King
Expressible perceptual content
King is surprised that I assume, without extended argument, that the content of per-
ceptual experience is expressible in natural language. I think the worry is exaggerated.
Though complications exist, and questions can be raised, it is, I think, overwhelm-
ingly plausible that much perceptual content is linguistically expressible, provided one
observes certain niceties. Suppose I truthfully report “This looks red” on the basis of
seeing an object o, which is red. When one sees something as red, one typically, per-
haps always, sees it as some finely individuated shade of red. Since not all these shades
are (nonindexically) named, there is no guarantee that the particular shade I visually
predicate of o is nonindexically expressed by any English term. But surely, it is express-
ible. If we can see it, attend to it, and discriminate it from other shades, we can name it,
if the need arises. That is the sense in which it is nonindexically expressible—as well as
being indexically expressible as “that shade (of red).” In addition to visually predicating
this shade of o, do I also visually predicate the redness of o? I think so. For any par-
ticular red-shade, predicating it of o also counts as predicating being red of o. Hence,
visually entertaining the proposition that o is that shade of red also counts as visually
entertaining the proposition that o is red.8
Other questions about visual content are more challenging. One of these, which
I raised in chapter 6, is whether the representational content of a visual perception of a
complex scene can be encompassed by a complex web of related propositions. Though
I am not certain that it can, I do think that much of that content is propositional. Other
important and far-reaching questions arise concerning how the vague color terms of
natural language come to encode the different properties they do (in different contexts
as used by different speakers), and how we should think of the not-fully-determinate
clouds of propositions asserted by utterances of sentences containing them in different
contexts. However that is work for another time.9
8
As for the vexed but much discussed question about the general relation between perceptual and
conceptual content, I largely align with the position outlined by Jeff Speaks in “Is There a Problem about
Nonconceptual Content?,” The Philosophical Review, 114, 2005, 359–398.
9
See my Rethinking Language, Mind, and Meaning, Princeton University Press, forthcoming.
philosophical truths. That said, the resulting account remains a package deal, some
aspects of which are detachable from others. For example, my student Justin Dallmann
has shown that one can recapitulate my account of propositions in a “serious actualist”
framework by trading my distinction between existent and non-existent truth bearers
for a distinction between propositions the existence of which (at the actual world) is
grounded in what is actually concrete and those the existence of which (at the actual
world) is grounded in what is merely possibly concrete.10 Though I don’t favor this
reconstrual, I commend it to serious actualists who find themselves unable to recant
their metaphysical error.
In addition to raising worries that I make room for too few propositions, King gives
two reasons for thinking that I countenance too many. He claims (i) that my account
of the de se and related cases opens the floodgates to too many proposition-building
operations which allow too many proposition-candidates to be constructed, and (ii)
that sometimes performing legitimate proposition-building acts in different orders
produces different cognitive event types where there is only one proposition. I will take
up these points separately.
First the de se. What makes the first-person way of thinking of oneself a
proposition-building operation is its direct, non-descriptive role in inference and
action. Because of this, admitting it (and related de se ways of cognizing) doesn’t com-
mit me to other garden-variety “ways of thinking” as analogous proposition builders.
De se ways of identifying predication targets differ from King’s examples—thinking of
o as inhabiting a world in which water is H2O and thinking of o as self-identical—in
not introducing any new predications or functional applications. One can, of course,
think of o as the x: x = o & x inhabits a world in which water is H2O, or as the x: x = o
& x is self-identical. To do so in the service of predicating redness of o is to entertain a
proposition that requires one to apply fthe to the propositional function corresponding
to the extra descriptive condition with the intention of predicating being red of the
result.11 Though these descriptive propositions exist, they are of a different kind than
the de se propositions to which King assimilates them. Thus, he is wrong to imagine
that I am committed to propositions the constituents of which are simply o and red-
ness, the entertainment of which requires o to be cognized in one of his ways.
Comparing (1) and (2) provides further perspective.
1a. Scott Soames is the messy shopper.
b. I am the messy shopper. (used de se by SS)
2a. Russell sought to prove logicism.
b. Russell sought to prove that arithmetic is reducible to logic.
10
Justin Dallmann, “Existence and the Cognitive Event Type Theory of Propositions,” (unpublished
manuscript, USC).
11
I here employ the Fregean definite description operator fthe. Other choices are possible.
The (immediate) constituents of the (a) and (b) propositions are the same in both
cases, as is the form of predication (direct). The propositions differ only in that the
(b) propositions impose an extra requirement on the way in which a predication target
must be cognized by one who entertains the proposition. Proposition (1b) requires SS
to be cognized in the first-person way; (2b) requires its propositional constituent to be
entertained. Although in (1b) this extra requirement doesn’t involve any further predi-
cations (functional applications, etc.), in (2b) it does. However, in neither case does the
extra requirement involve predicating anything further of (or operating in any further
way on) the relevant predication target. This, I suspect, is what King missed in wrongly
concluding that, for me, extra requirements on how predication targets are cognized
can introduce new predications of, or operations on, them. They can’t. Nor is it arbi-
trary that proposition-building acts can involve targeting propositions in a way that
involves entertaining them (as in (2b)). This is simply the combination of two cognitive
acts, both of which we know independently to be proposition building—entertaining
propositions and directly targeting them (as we can do with anything with which we
are acquainted or for which we have a name). Once all this is clear, King’s contention
that proposition-building acts employed in my analyses of (1b) and (2b) overgenerate
propositions, and so leave us with too many, can be seen to be groundless.
This is not true of his second worry, which raises a real issue, albeit a minor one.
Is the proposition that Romeo loves Juliet the cognitive act (or event type) in which
loving is predicated of the pair consisting of Romeo followed by Juliet? Is it the act (or
event type) in which one first operates on loving and Juliet to form the property lov-
ing Juliet, which is predicated of Romeo? Or is it the act (or event type) in which this
order is reversed? Since there are three slightly different acts (or event types), it might
seem that I am saddled with three different propositions where there should be only
one. Though puzzling, this issue is not, I think, very serious. One response would be
to allow three different but related Romeo-loves-Juliet propositions, while character-
izing attitudes like judging, believing, and asserting in a way that guarantees that an
agent who bears them to one of the three propositions bears them to all three. A differ-
ent response would be to identify the proposition that Romeo loves Juliet with the act
(or event type) in which one either predicates loving of <Romeo, Juliet>, or combines
loving with Juliet and predicates the resulting complex property of Romeo, or com-
bines loving with Romeo and predicates being one whom Romeo loves of Juliet. Short of
investigating how to extend these (and perhaps other) strategies generalize, I will not
here attempt to adjudicate between them.12 Still, I see no reason to think that the issue
(which arises for most accounts of structured propositions including King’s) can’t be
resolved.
12
As Brian Bowman has reminded me, particular languages, like English, which recognize verb phrases as
sentential constituents but not subject+transitive verb combinations, might constrain the proposition can-
didates expressible by their sentences. Even so, using such languages to report the attitudes of others would
itself raise the issue.
Circularity?
King’s next contention—that my account of propositions is circular because the expla-
nation of their representational properties presupposes possibility, while my notion
of a possible world-state presupposes propositions—is misguided on two counts.
First, and foremost, his critique tacitly assumes that ordinary modal notions like what
could be and what could not possibly be are conceptually dependent upon, and so to be
analyzed in terms of, the conceptually prior notion of a possible world-state. As I have
repeatedly argued, this Lewisian assumption couldn’t be further from the truth.13
Although there are both epistemically and metaphysically possible world-states—ways
the world could be (or have been)—they are defined in terms of our ordinary modal
notions, rather than the other way around. As I explain in chapter 3, on my analysis,
the notion of a proposition conceptually depends on objects, properties and cognitive
acts of agents, while the notion of a possible world-state—i.e. a maximal property of a
certain sort that the universe could have instantiated—conceptually depends on truth,
propositions, and our ordinary modal notions. Since propositions don’t conceptually
depend on possible world-states, an explanation of how they manage to be representa-
tional can make use of ordinary modal notions, including the possibility of a cognitive
act being performed and an event type having instances, without circularity.
Second, as I made clear in my reply to Speaks, a proposition represents things as
being a certain way because it is either the cognitive act of representing things as being
that way, or the event type of performing that act. Though it follows from this explana-
tion that any possible performance of the act is one in which an agent represents things
as being a certain way, it is not obvious that the explanation conceptually presupposes
any modal notions at all (though even if it did, there would be no circularity).
13
See Soames, “The Place of David Lewis in Analytic Philosophy,” forthcoming in “The Place of David Lewis
in Analytic Philosophy,” David Lewis, eds. Barry Loewer and Jonathan Schaffer, Oxford: Wiley-Blackwell;
chapter 5 of Soames, The Philosophy of Language, Princeton and Oxford: Princeton University Press,
2010, and Soames, “Actually,” Aristotelian Society Supplementary Volume 81, 2007, 251–277, reprinted in
Soames, Philosophical Essays, Volume 2, Princeton and Oxford: Princeton University Press, 2009.
does—namely represent o as red, which I call “predicating redness of o.” At this point,
we appeal to a derivative sense of “represent” in which this act itself represents o as red.
Though distinct from the primary sense in which an agent represents o as red, this
extended sense is related to that primary sense in a way analogous to the way in which
the sense in which some acts are intelligent, stupid, thoughtful, or kind is related to the
primary sense in which agents who perform those acts are intelligent, stupid, thought-
ful, or kind. Very roughly, (i) for an act to be intelligent or thoughtful is for it to be one
the performance of which marks one as behaving intelligently or thoughtfully, and
(ii) for a cognitive act to represent o as red is for it to be one the performance of which
marks one as representing o as red.
As indicated in my reply to Speaks, we, as agents, need this extended sense of rep-
resentation in part because we wish to isolate individual aspects of the thought and
perception of ourselves and others in order to assess them for accuracy. When o is such
that to perceive or think of o as red is to represent it accurately, it is both enormously
useful and very natural to seek an entity—a particular sort of perceiving or thinking—
plus a property that entity has when this sort of perceiving or thinking is accurate. The
entity is a proposition, which is either the cognitive act of representing o as red or the
cognitive event type of so doing. The property is truth, which the act (or event type)
has iff to perform it (or to bring about an instance of the event type) is to represent o as
o really is.
In What is Meaning? I ruled out acts as propositions on the basis of a short-sighted
ordinary-language argument about what is or isn’t an absurd “category mistake” of the
sort that fills the last few pages of King’s current critique of my view.14 As I said in my
response to Mark Richard at the session on What is Meaning? at the Eastern Division
Meetings of the APA in December of 2011, I now see the error of those ways. Because
our task is theory construction—which in philosophy as well as empirical science can,
when successful, usher in new, surprising, and sometimes counterintuitive truths—
ordinary-language style arguments that deny this have no more force against the act
view of propositions than they do against the event-type view.15 Since I no longer see
a compelling reason to analyze propositions as event types as opposed to acts, I no
14
What is Meaning?, pp. 101–102.
15
It is, for example, common in the philosophy of language for propositions to be said to be the meanings
of non-indexical sentences, despite the fact that this goes strongly against the grain of some of our ordinary
ways of speaking about meaning. This is noted in Richard Cartwright’s classic article, “Propositions,” in
R. J. Butler, ed., Analytical Philosophy, First Series, Oxford: Basil Blackwell, 1962; reprinted in his Philosophical
Papers, Cambridge, MIT Press, 1987. On pp. 49–50 of the latter he says, “If what someone asserts, on some
occasion [namely a proposition] is itself the meaning which the words he utters have, on that occasion
of their utterance, then anything predicable of what he asserts must also be predicable of the meaning
of his words. But it is obvious on very little reflection that ever so many things predicable of what is
asserted cannot (on pain of nonsense) be predicated of the meaning of a sentence. And the fundamental
point to be noticed in this connection is that although we may predicate of something asserted that it is
(or was) asserted, this cannot be predicated of the meaning of a sentence. It simply makes no sense to say
that someone asserted the meaning of a sentence [my emphasis]...Just as the meanings of sentences can-
not be asserted, neither can they be affirmed, denied, contradicted, questioned, challenged, discounted,
confirmed, supported, verified, withdrawn, or repudiated; and whereas what is asserted can be said to be
accurate, exaggerated, unfounded, overdrawn, probable, improbable, plausible, true, or false, none of these
can be said of the meaning of a [i.e. any] sentence.” Try it. Bill asserted/proved/contradicted/supported/
questioned/withdrew the proposition that mathematics is reducible to logic vs. *Bill asserted/proved/con-
tradicted/supported/questioned/withdrew the meaning of the sentence “Mathematics is reducible to logic.”
Whereas the former sound fine, the latter sound like category mistakes—incoherent or without sense (when
they are not taken as suggesting some entirely different content). Similarly for *The meaning of the sentence
“Mathematics is reducible to logic” is plausible, probable, or untrue. But these are not incoherent or without
sense, as Cartwright himself came to realize between 1967 and 1986. (See the addenda on pp. 52–53 for per-
suasive argument.) But then, what reason is there to deny that some meanings may be propositions even
though certain things truly attributable to propositions initially sound as if they couldn’t be true of meanings
(and conversely)? There is no good reason; the results of fruitful and systematic theorizing justify the revi-
sion of some of our ordinary, pretheoretic thought and talk. This is just as true in the case of successful theo-
ries that identify propositions with cognitive acts or event types as it is in the case of theories that identify the
meanings of some sentences with propositions.
16
Chapters 3 and 6 were submitted in September of 2011 to Speaks and King for their criticism, before my
change of mind on this point. Since those chapters could not be altered to reflect this change of mind after
my co-authors had begun working on their critiques, my restatement had to wait for this chapter on “fur-
ther thoughts.” In fact, I now am now more inclined to identify propositions with cognitive acts than with
event types.
17
King says, “In general, when an agent bears R to something o at a time [and so has the property bearing
R to o], the event token of the agent bearing R to o does not itself bear R to o. If I hug Annie, the event token
of my hugging Annie doesn’t hug Annie. So why, from the fact that an agent represents o as red (by predicat-
ing redness of it), would it follow [my emphasis] that the event token of the agent representing o as red itself
represents o as red?” Later he makes the same argumentative move concerning event types.
18
One might construe King in a slightly different way—not as holding that agents use F in order that they
may represent o as red, but rather that, already being able to represent o as red themselves, they use this ability
to stipulate that the otherwise brute fact F is henceforth to be understood as representing o as red. However,
What is to be explained?
As I see it, the issue between King and me is which (if either) of these imagined ways of
extending the primary sense in which agents represent is, or should be, in play when
we think of, and theorize about, propositions as representational. This is a matter not
of arbitrary stipulation, but of theoretical insight. More generally, the three theories
of propositions sketched in this book are attempts to outline sound and fruitful con-
ceptions capable of playing the roles for which propositions are needed in both phi-
losophy and empirical science. It is true that each of us holds views that are at least
mildly revisionary. Up to now, propositions—what is said, believed, etc.—have not
ordinarily been thought to be either the cognitive acts/events that are central to my
account, the complex linguistic facts that are central to King’s, or the complex proper-
ties central to Speaks’s conception. But, since the task is theory construction, this is of
no great consequence. To assess our theoretical accounts, one must determine which
best accommodates the most important features of our uncontentious pretheoretic
talk of propositions, while providing us with entities that can play the theoretical roles
for which we need propositions in philosophy, psychology, biology, linguistics, and
philosophical logic.
The following are a few of the facts that I think need to be explained by any successful
theory of propositions:
(i) that one who judges or affirms that that o is red, himself represents o as red,
and cognitively commits himself to o’s being so;
(ii) that such an agent may thereby stand in the judging, affirming, and believ-
ing relations to the proposition that o is red without having any conception
of propositions, and without having the ability to represent them as bearing
properties and standing in relations to anything;
(iii) that agents with sufficient cognitive sophistication can acquire the ability
to represent propositions as having properties and standing in relations by
focusing on their own cognitive acts and experiences of representing things
as being one way or another, by grouping these acts and experiences into sim-
ilarity types (on the basis of what things in the different cases have been taken
to be what ways), and by treating the different types as units, thereby implic-
itly identifying propositions as what similar types have in common without
forming any worked out positive conception of what these unities are;
(iv) that judging, affirming, or believing that that o is red does not require an agent
to have mastered any language; agents could stand in these attitude relations
to the proposition even if there were no sentences or languages at all;
this is not a plausible story for him to tell, in part because it presupposes that agents already bear proposi-
tional attitudes to the propositions his account is supposed to explain, and in part because agents do not have
his enormously complex linguistic facts in mind as things to be endowed with representational properties
by their stipulations.
(v) that all propositions represent things as being certain ways and so are true iff
the things in question are as they are represented to be;
(vi) that the proposition that o is red would represent o as being red, and could be
true, even if there were no agents;
(vii) that it is possible for one and the same proposition to be the content of a per-
ceptual experience, a nonlinguistic thought, and an assertive utterance of a
sentence;
(viii) that the proposition that some past philosophers, including Socrates and Plato,
don’t exist itself both exists and is true, even though Socrates and Plato no
longer exist;
(ix) that the proposition (a) that o is red is distinct from the propositions (b) that
o is red and o is self-identical and (c) that o is red and 1st-order arithmetic is
incomplete, and that one may stand in the affirmation, judgment, or belief
relations to (a) without standing in those relations to (b) and (c), but one can-
not stand in those relations to (b) or (c) without standing in them to (a).
(x) that the proposition that Cicero shaved himself, represents Cicero as having
the property being a self-shaver, and so is distinct from the proposition that
Cicero shaved Cicero, even though the same are true in the same metaphysi-
cally and epistemically possible world-states.
(xi) That the propositions that Russell sought to prove that arithmetic is reducible
to logic and that Russell sought to prove logicism are different—since the latter
can be asserted or believed by someone who doesn’t believe or assert the for-
mer—even though they represent precisely the same things as being precisely
the same ways, and hence have identical truth conditions.
(xii) that the points just made in (xi) also hold for the propositions that I wrote this
chapter and that Scott Soames wrote this chapter.
In taking (i—xii) to be facts, I am not claiming that they are the contents of privileged
intuitions that must, if at all possible, be preserved by theories of propositions. As far
as I can see, few, if any, of our strongly held pretheoretic convictions are so privileged.
Rather, what are often called intuitions are things we strongly believe, frequently but
not always with good reason, prior to conscious theorizing of the sort found in logic,
philosophy, linguistics, or psychology. For this reason, it makes considerable sense
that we should seek to preserve and explain most of them, while being ready to revise
some of them when necessary
This is the background against which one should judge pretheoretically surprising
identifications of propositions with one or another class of entities. Since the expla-
nations provided by the theory I have sketched seem superior to those provided by
other theories, I judge it to be more likely to be correct than they are. According to
it, propositions are, very roughly, ways of thinking, conceiving, or perceiving things to
be. Although this doesn’t sound terribly surprising or counterintuitive, it becomes so
when one gets more specific—identifying propositions with cognitive acts (in which
case they become a species of things done) or with the event types in which one per-
forms those acts (in which case they become a species of things that happen). It is true
that both of these claims sound jarring at first. However, the unreflective opinion that
propositions can be neither things we do nor things that happen is not sacrosanct and
may itself be due either to a failure to theorize, or to a tendency to do so incorrectly.
It may also be true that any theory of propositions that leads to plausible explanations
of facts like those illustrated by (i-xii) will lead to jarring surprises of its own. Since we
are not in a position to rule this out in advance, we must not hobble ourselves by pro-
hibiting surprises of the sort that I am willing to accept, or that my co-authors are. As
I see it, success in our common enterprise will be success in identifying what agents
have been referring to all along when speaking of propositions, and what properties
they have ascribed to these entities when characterizing them as having been asserted
or believed, or as having truth conditions—even if little of the theoretical detail about
what these entities are, or how precisely we or they manage to represent the world, is
something we are in a position to know without careful theory construction.19
19
This chapter has been greatly improved by Brian Bowman’s many helpful comments.
______ (2010). “Selfless Desires and the Property Theory of Content”. Australasian Journal of
Philosophy 88 (3):489–503.
Fine, K. (1985). Plantinga on the reduction of possibilist discourse. In J. Tomberlin and P. van
Inwagen (eds.), Alvin Plantinga (Dordrecht: D. Reidel): 145–186.
______ (1994). “Essence and modality”. Philosophical Perspectives 8:1–16.
______ (2007). Semantic Relationism. Oxford: Blackwell.
Fodor, J.A. (1990). A Theory of Content and Other Essays. Cambridge, MA: MIT Press.
Foster, J. (1976). Meaning and Truth Theory. In Truth and Meaning: Essays in Semantics.
Oxford: Oxford University Press.
Frege, G. (1918). Der Gedanke, Beitrage zur Philosophie des deutschen Idealismus, 1, 1918, 58–77;
trans. “The Thought” by A. and M. Quinton, in Mind 65, 1956, 289–311; trans. “Thought,” in
Beaney, The Frege Reader, Oxford: Blackwell, 1997, 325–345, quoted at p. 333.
______ (1979). A brief Survey of my logical Doctrines. In Gottlob Frege: Posthumous Writings.
Chicago:The University of Chicago Press..
Gaskin, R. (2008). The Unity of the Proposition. Oxford: Oxford University Press.
Gilmore, C. (forthcoming). Parts of Propositions. In Shieva Kleinschmidt (ed.), Mereology and
Location. Oxford: Oxford University Press.
Greenberg, G. (2011). Semantics for Pictures. Unpublished ms.
Grice, P. (1968). “Utterer’s Meaning, Sentence Meaning, and Word Meaning”. Foundations of
Language 4: 225–242.
______ (1969). “Utterer’s meaning and intention”. Philosophical Review 78 (2):147–177.
Hanks, P. (2009). “Recent work on propositions”. Philosophy Compass 4 (3):469–486.
______ (2011). “Structured Propositions as Types”. Mind 120 (477):11–52.
Harman, G. (1974). Meaning and semantics. In Milton K. Munitz & Peter K. Unger (eds.),
Semantics and Philosophy. NY: New York University Press. Reprinted in Harman (1999).
______ (1999). Reasoning, Meaning, and Mind. Oxford: Oxford University Press.
______ (2003). “Category mistakes in metaphysics and epistemology”. Philosophical Perspectives
17 (1):165–180.
Hawthorne, J. and Manley, D. (2012), The Reference Book. Oxford: Oxford University Press.
Jackson, F. (1998). From Metaphysics to Ethics. Oxford: Oxford University Press.
Johnston, M. (2004). “The obscure object of hallucination”. Philosophical Studies 120
(1–3):113–183.
Kaplan, D. (1977/1989). Demonstratives. In Joseph Almog, John Perry & Howard Wettstein
(eds.), Themes From Kaplan. Oxford: Oxford University Press, 1989
Keller, L.J.. (2012). Whence Structured Propositions? Dissertation, University of Notre Dame.
King, J.C. (1995). “Structured propositions and complex predicates”. Noûs 29 (4):516–535.
______(2002a). “Designating propositions”. Philosophical Review 111 (3):341–371.
______ (2002b), Two Sorts of Claims about Logical Form. In Logical Form and Language, G.
Preyer and G. Peter (ed.), New York: Oxford University Press, (2002) pp. 118–131.
______ (2007). The Nature and Structure of Content. Oxford: Oxford University Press.
______ (2009). “Questions of Unity”. Proceedings of the Aristotelian Society, Vol. CIX, Part 3
______ (2011). “On Fineness of Grain”. Philosophical Studies DOI: 10.1007/s11098-011-9844-9.
______ (2012). “Propositional unity: what’s the problem, who has it and who solves it?”
Philosophical Studies DOI: 10.1007/s11098-012-9920-9
______(ms.), “Acquaintance, Singular Thought and Singular Propositions,” unpublished ms.
______ (1998). “Is Belief De Re Reducible to Belief De Dicto?” Canadian Journal of Philosophy,
Supplementary volume 23: 85–110.
______ (2005). “On designating”. Mind 114 (456):1069–1133.
Salmon, N.U. and Soames, S. (eds.) (1988). Propositions and Attitudes. Oxford: Oxford
University Press.
Schiffer, S.R. (1972). Meaning. Oxford: Clarendon Press.
______ (1987). Remnants of Meaning. Cambridge, MA: MIT Press.
Siegel, S. (2006). Which properties are represented in perception? In Tamar S. Gendler & John
Hawthorne (eds.), Perceptual Experience. Oxford: Oxford University Press.
Soames, S. (1987). “Direct reference, propositional attitudes, and semantic content”. Philosophical
Topics 15 (1):47–87. Reprinted in Salmon & Soames (1988) and in Soames (2009), v. 2.
______ (1992). “Truth, meaning, and understanding”. Philosophical Studies 65 (1–2):17–35.
______ (1994). “Attitudes and anaphora”. Philosophical Perspectives 8:251–272.
______ (1999). Understanding Truth. Oxford: Oxford University Press.
______ (2002). Beyond Rigidity: The Unfinished Semantic Agenda of Naming and Necessity.
Oxford: Oxford University Press.
______ (2003). “Understanding deflationism”. Philosophical Perspectives 17 (1):369–383. Reprinted
in Soames (2009), v. 2.
______ (2005). Reference and Description: The Case Against Two-Dimensionalism. Princeton,
NJ: Princeton University Press.
______ (2006). Understanding assertion. In Judith Jarvis Thomson & Alex Byrne (eds.), Content
and Modality: Themes From the Philosophy of Robert Stalnaker. Oxford: Oxford University
Press. Reprinted in Soames (2009), v. 2.
______(2007a). “Actually”. Aristotelian Society Supplementary Volume 81 (1):251–277. Reprinted in
Soames (2009), v. 1.
______ (2007b). “What Are Natural Kinds?” Philosophical Topics 35 (1–2):329–342.
______ (2009). Philosophical Essays: Natural Language: What It Means and How We Use It, vol-
umes 1 and 2. Princeton, NJ: Princeton University Press.
______ (2010). What Is Meaning? Princeton, NJ: Princeton University Press
______(2011). Propositions. In Delia Graff Fara & Gillian Russell (eds.), The Routledge Companion
to Philosophy of Language. Routledge.
______ (forthcoming-a) Language, Mind, and Meaning: The Hempel Lectures. Princeton
University Press, forthcoming.
______ (forthcoming-b) The Analytic Tradition in Philosophy, volume 1. Princeton University
Press, forthcoming.
______(forthcoming-c). David Lewis’s Place in Analytic Philosophy. In Barry Loewer & Jonathan
Schaffer (eds.), David Lewis. Chichester: Wiley-Blackwell.
Speaks, J. (2005). “Is there a problem about nonconceptual content?” Philosophical Review 114
(3):359–398.
______ (2006). “Truth theories, translation manuals, and theories of meaning”. Linguistics and
Philosophy 29 (4):487–505.
______ (2012). “On Possibly Nonexistent Propositions”. Philosophy and Phenomenological
Research 85 (3):528–562.
Stalnaker, R. (1978). “Assertion”. Syntax and Semantics 9:315–332. Reprinted in Stalnaker (1999).
______ (1984). Inquiry. Cambridge: Cambridge University Press.
______ (1990). “Mental content and linguistic form”. Philosophical Studies 58 (1–2):129–146.
______ (1999). Context and Content: Essays on Intentionality in Speech and Thought.
Oxford: Oxford University Press.
______ (2003). Ways a World Might Be: Metaphysical and Anti-Metaphysical Essays.
Oxford: Oxford University Press.
______ (2010). Merely Possible Propositions. In Modality: metaphysics, logic and epistemology, ed.
by B. Hale and A. Hoffmann. Oxford: Oxford University Press
Stanley, J. (2001). “Hermeneutic factionalism”. Midwest Studies in Philosophy 25 (1):36–71.
Turner, J. (2005). “Strong and weak possibility”. Philosophical Studies 125 (2):191–217.
______ (2010). “Fitting attitudes de dicto and de se”. Noûs 44 (1):1–9.
van Inwagen, P. (1990). Material Beings. Ithaca, NY: Cornell University Press.
______ (2004). A Theory of Properties. In Dean W. Zimmerman (ed.), Oxford Studies in
Metaphysics, Volume 1. Oxford: Clarendon Press.
Walton, K.L. (1990). Mimesis as Make-Believe: On the Foundations of the Representational Arts.
Harvard University Press.
Williamson, T. (2002). Necessary existents. In A. O’Hear (ed.), Logic, Thought, and Language.
Cambridge: Cambridge University Press.
Yablo, S. (2000). A Paradox of Existence. In Anthony Everett and Thomas Hofweber (eds.),
Empty Names, Fiction and the Puzzles of Non-existence, 275–312. CSLI.
facts 2, 7–8, 20, 44, 51–9, 65–72, 75, 87, 95, 136, negation 98–9, 103, 161, 186, 206
139, 146, 151–2, 156–8, 162–81, 190, 196–7, Nolan, Daniel 81–3, 141
223–4, 241 non-existents 101, 130, 237
Feit, N. 78, 81
fictionalism 17–18 paratactic analysis 13–16
Fine, Kit 81, 116–8, 151 Parsons, T. 66–9, 179
perceptual experience 26 Russell, Bertrand 25–30, 32–3, 47, 48, 61, 71–4,
content of 6, 8, 47, 60, 93–8, 128–9, 137, 153, 78, 97, 119–24, 167, 208, 237–43
157–8, 177, 181, 191–7, 200–1, 219, 236
Perry, John 38, 107–113, 132 Salmon, N. 84, 86, 101, 119
Plantinga, A. 222 Schiffer, S. 14, 156
possible worlds 6, 8, 57–8, 77, 96–7, 133–4, Segal, Gabriel 20
189–90, 212–3, 222–4, 239 semantic relationism 116–18
conception of propositions as sets of 2, 16, sentential relations 50–5, 59–60, 63, 151–8, 169,
33–44, 47 173–4, 188, 192–3, 197, 199–200
predication 27–34, 37, 72, 75, 86, 95, 97–100, 102, Siegel, S. 192
104, 110–124, 128–30, 137–8, 147, 155, 158–74, Stalnaker, Robert 6, 14, 34, 37, 40–1, 189
179, 199, 226–31, 237, 238 Stanley, J. 17, 208–14
proposition syntax
-al functions 99–100, 111–2, 237 relations of 29, 48–9, 54–64, 74–5, 93–4,
-al relations 50–63, 147, 152–60, 172–7, 187, 152–5, 170–9, 188–96, 209–10
191–200, 209–13 semantic significance of 49, 55, 57, 59, 75, 111,
entertaining a 55, 97–9, 104–8, 110–1, 115–8, 155, 172–3, 193–6
122, 124, 129, 132–3, 135, 138, 158–61, 167,
172–6, 199–200, 223, 230–2, 234, 236, 238 “that” clauses 7–8, 9–19, 64–70, 87–8, 93, 105–6,
unity of 26–33, 97 174–5, 178–80, 201–8
representational properties of 29–31, 33–5, 48, truth
77–8, 92–6, 105–9, 113, 118, 127, 132–5, 147–9, conditions 17, 27, 31–7, 47–8, 51–5, 58–61, 64,
150–67, 174–6, 185–6, 189–90, 193, 195, 197, 80, 91–7, 106, 108, 110, 116–8, 120, 124, 127–8,
215–21, 225–31, 234–6, 239 133–40, 145–52, 156, 167–9, 173–7, 187–93,
singular 41, 76, 101, 107, 112, 114–5, 120, 196–200, 204, 220–1, 227, 235
149–50, 154, 157, 187, 192, 209, 211, 223 deflationism about 103–4
propositional attitudes 2, 6, 39–40, 44, 53, functions 98–9, 119, 208, 213
59–60, 77–86, 117, 140–3, 155–61, 171–6, monadic vs. dyadic 35–8
191–7, 217–9, 229–30 T-sentences 19–24
de se 106–15, 132–3, 141, 167–8, 237. See also Turner, Jason 78, 81
propositional attitudes, first personal Two-dimensional semantics, 38, 42–4
first personal 80–1, 85–6, 141
ascriptions of 7, 12–15, 39–40, 79–88, 99, van Inwagen, Peter 12, 72, 77, 90, 148
112, 141 variables 39–40, 44, 56, 112, 154, 170–1, 178–9
assignment of values to 44, 56–8, 63,
Quine, W.V. 86 100, 110, 112, 151, 154, 170–1, 178–9,
188, 217
Recanati, Francois 94
representational properties Walton, K. 18
as intrinsic 33, 78, 133, 146–8, 160, 167, 215–6 Williamson, T. 81
of maps 47, 60, 167, 191, 201, 219, 221
See also perceptual experience, proposition Yablo, S. 17
Richard, Mark 17–18, 105, 211, 240 Yli-Vakkuri, Juhani 90